SCHRIFTENREIHE Institut für Pflanzenernährung und Bodenkunde ...

SCHRIFTENREIHE 

Institut für Pflanzenernährung und Bodenkunde 

Universität Kiel 

Nr. 77 (2008) 

Ying Zhao 

Grazing effects on hydraulic, thermal and 

mechanical soil properties at multiple scales 

— a case study in Inner Mongolia grassland — 

Herausgeber: R. Horn und K. H. Mühling

(Nur für Dissertationen: ) 

Gedruckt mit Genehmigung der Agrar- und Ernährungswissenschaftlichen Fakultät 

der Christian-Albrechts-Universität zu Kiel 

Vertrieb: Institut für Pflanzenernährung und Bodenkunde 


Hermann Rodewald Str. 2 

D - 24118 Kiel 

(e-Mail: h.fleige@soils.uni-kiel.de) 

ISSN 0933-680 X 

Preis: Eur 15 (incl. Versandkosten)

Aus dem Institut für Pflanzenernährung und Bodenkunde 

Christian-Albrechts-Universität zu Kiel 

Grazing effects on hydraulic, thermal and 

mechanical soil properties at multiple scales 

— a case study in Inner Mongolia grassland — 

Dissertation 

zur Erlangung des Doktorgrades 

der Agrar- und Ernährungswissenschaftlichen Fakultät 


vorgelegt von 

M.Sc. Ying Zhao 

Aus Xihe, Gansu, V.R. China 

Kiel, 2008

Dekan: Prof. Dr. J. Krieter 

1.Berichterstatter: Prof. Dr. R. Horn 

2.Berichterstatter: Prof. Dr. R. Duttmann 

Tag der mündlichen Prüfung: 14.02.2008

I Table of contents 

I Table of contents i 

II Summary - Zusammenfasung iii 

Summary iii 

Zusammenfassung vii 

III List of figures xi 

IV List of tables xv 

1 General Introduction 1 

Chapter 2: Spatial variability of soil properties affected by grazing intensity in 

Inner Mongolia grassland 11 

Chapter 3: Spatio-temporal variability of soil moisture in grazed steppe areas 

investigated by multivariate geostatistics 39 

Chapter 4: Temporal stability of soil moisture in a semi-arid steppe and its 

application in model result validation 65 

Chapter 5: Modeling Grazing Effects on Coupled Water and Heat Fluxes in Inner 

Mongolia Grassland 85 

Chapter 6: Modeling of Coupled Water and Heat Transfer in Freezing and 

Thawing Soil 115 

7 General discussion and conclusions 139 

8 Acknowledgments 151 

9 Curriculum Vitae 153 

i

II Summary – Zusammenfassung 

Summary 

Over-grazing has been regarded as a main cause for grassland degradation in 

Inner Mongolia because of the increase in population and shift in the 

socio-economic system in recent years. Given the vital importance for the 

production of live stock and the environmental changes, it is crucial to have a 

thorough understanding of the mechanisms that maintain or change the 

ecosystem in response to the changes of land management. 

Intensive observations, analysis and modeling of environmental changes in 

response to grazing effects in Inner Mongolia grassland were carried out within 

the framework of the collaborative research project MAGIM: Matter fluxes in 

grasslands of Inner Mongolia as influenced by stocking rate (2004-2007). Based 

on this, we concentrated into investigations of the soil physical, hydrological and 

ecological processes in general and modeling the effects of grazing 

management (i.e. ungrazed since 1979=UG 79, ungrazed since 1999=UG 99, 

winter grazed=WG, continuously grazed=CG and heavily grazed=HG) on water 

and heat fluxes in particular. 

Geostatistical analysis showed that soil properties including soil water content 

(SWC), hydraulic conductivity (Ks), water drop penetration time (WDPT), shear 

strength (SS), soil organic carbon (SOC) concentration, bulk density (BD) and 

soil texture exhibited a moderate or strong spatial dependency. Soil compaction 

induced by sheep trampling, resulted in a more homogenous spatial distribution 

of soil properties, which will possibly enhance risks for soil erosion. 

Multiple regression analysis showed significant correlations among SWC, Ks, 

WDPT, SOC and BD; as well as between SS and silt content, indicating that soil 

physical and mechanical properties are interrelated with each other at the same 

location. Furthermore, multivariate geostatistical analysis revealed 

scale-dependent correlations among various parameters of which the ones 

iii

controlling soil moisture are of particular concern as they are related to 

grassland productivity. It was observed that the soil and plant variables (e.g. soil 

texture) are main contributor to the variations of soil moisture. Furthermore, 

those processes, affected by land managements, were interrelated at different 

spatial scales. For instance, random (

changes. The model results indicated that soil heat fluxes increased with 

increasing grazing intensity. In comparison with the two ungrazed sites, winter 

grazing did not show any effect on the water household components, while 

heavy grazing remarkably decreased soil water storage by 25-45%, interception 

by 50-55% and transpiration by 20-30%, and increased evaporation by 25-40%, 

runoff by 100-300% and drainage by 100-200%. We conclude that intense 

grazing deteriorates soil functions, consequently reduces plant available water 

and thus grassland productivity. 

The incorporation of an extended freezing code in the HYDRUS-1D model 

resulted in an improved understanding of the coupled hydraulic and thermal 

processes under frozen condition. The freezing model is capable of predicting 

the changes in soil water and heat fluxes accompanied by soil freezing and 

thawing behavior, as well as grazing effects. However, soil water content after 

spring snowmelt is overestimated and thus surface runoff might be 

underestimated. 

v

Zusammenfassung 

Überweidung wird als einer der Haupgründe für die Degradation von Grassland 

Ökosystemen in der Inneren Mongolei angesehen. Die zunehmende 

Bevölkerungszahl und ein sich im Wandel befindendes sozio-ökonomisches 

System haben in den vergangenen Jahren zu einem zunehmenden 

Beweidungsdruck geführt. Da Produktionssteigerungen nicht ohne 

Konsequenzen für die Umwelt bleiben ist ein grundlegendes Verständnis der 

beteiligten Prozesse, die bei der Beweidung auf das Ökosystem einwirken, von 

entscheidender Bedeutung für eine nachhaltige Entwicklung von Weideflächen 

in der Inneren Mongolei. 

Im Rahmen der Forschergruppe MAGIM (Matter fluxes in grasslands of Inner 

Mongolia as influenced by stocking rate) wurden seit 2005 intensive Messungen 

und Modellsimulationen auf Flächen der „Inner Mongolian Grassland Research 

Station (IMGERS)“ durchgeführt. Ziel ist es möglichst alle relevanten Parameter 

von Tier- über Pflanzenproduktion und Umwelteinflüssen auf Boden und 

Atmosphäre gemeinsam zu untersuchen. In diesem Teilprojekt wurden 

insbesondere bodenphysikalische und hydraulische Parameter erfasst, die eine 

systematische Modellierung der Einflüsse unterschiedlicher 

Beweidungsintensitäten (unbeweidet seit 1979, unbeweidet seit 1999, Winter 

Beweidung, kontinuierliche Beweidung und Überweidung) auf den Wasser- 

und Wärmehaushalt auch unter Berücksichtigung von Gefrier- und 

Auftauprozessen ermöglichen sollten. 

Geostatistische Analysen von in situ Feldmessungen in der obersten 

Bodenschicht zeigten für alle gemessenen Parameter, d.h. hydraulische 

Leitfähigkeit (Ks), Benetzungshemmung (WDPT), Scherwiderstand (SS), 

organischer Kohlenstoffgehalt (SOC), Lagerungsdichte (BD) und Textur eine 

moderate bis starke räumliche Abhängigkeit. Durch Schaftritt induzierte 

Bodenverdichtungen führten zu einer Abnahme der räumlichen Varianz der 

untersuchten Parameter, was unter anderem auf eine mechanische 

vii

Homogenisierung des Bodens zurückzuführen ist. Die damit verbundene 

Strukturzerstörung erhöht zusammen mit dem deutlich ausgedünnten 

Pflanzenbestand das Risiko für Bodenerosion durch Wind und Wasser 

möglicherweise erheblich. 

Multiple Regressionsanalysen zeigten signifikante Korrelationen zwischen SWC, 

Ks, WDPT, SOC und BD, sowie zwischen SS und Schluffgehalt. Des Weiteren 

ließen sich mit Hilfe multivariater geostatistischer Analysen skalenabhängige 

Korrelationen zwischen Bodenfeuchte und Textur, BD, SOC und 

Bedeckungsgrad feststellen. Die Skalenabhängigkeit der untersuchten Boden- 

und Pflanzenvariablen, die insbesondere die Bodenfeuchte steuern, sind 

offentsichtlich ebenso eine Funktion der Beweidungsintensität, mit räumlich 

zufälliger Variabilität (< 5 m) in UG79, kleinskaliger Variabilität (90 m) in UG99, 

und mesoskaliger Variabilität (165 m) in CG. Aus diesem Zusammenhang lässt 

sich schließen, dass die Bodenfeuchteverteilung im Raum durch die Beweidung 

beeinflusst wird, was bei der Regionalisierung hydrologischer Modelle eine sehr 

wichtige Rolle spielt. 

Ergänzend zu den geostatistischen Verfahren, in denen die räumliche 

Abhängigkeit von Boden- und Pflanzenparametern untersucht wurden, sollte ein 

in situ Monitoring der Bodenfeuchte und –temperatur an ausgewählten 

Standorten in Verbindung mit einer prozessorientierten Modellierung Aufschluss 

über die zeitliche Entwicklung der Wasser- und Wärmeflüsse bei 

unterschiedlicher Beweidungsintensität geben. Die Repräsentativität der 

gewählten Standorte wurde mit Hilfe einer Zeitstabilitätsanalyse untersucht. 

Dabei wurde bestätigt, dass die Monitoring Standorte mit sogenannten 

zeitstabilen Punkten (Time Stable Points = TSPs) übereinstimmen, die die 

mittlere Feldbodenfeuchte mit hoher Genauigkeit widerspieglen. Die Simulation 

der Wasserflüsse mit Hilfe eines hydraulischen Modells in den TSPs ist daher, 

sowohl die mittleren Feldwassergehalte sowie die zeitliche 

Wassergehaltsänderung betreffend, ebenfalls als repräsentativ anzusehen. 

viii

Die Beweidung beieinflusst besonders in den oberen 10-15 cm hydraulische, 

thermische und mechanische Bodeneigenschaften und -funktionen. Bei den 

Untersuchungen wurden eine Verringerung von SWC, SOC, WDPT und eine 

Zunahme von BD und SS festgestellt. Änderungen in der Bodenstruktur sind 

durch geringere gesättigte Wasserleitfähigkeiten (Ks) und Verschiebungen in der 

Wasserrentionsfunktion angezeigt. Im Vergleich zu den unbeweideten 

Standorten nehmen das Gesamtporenvolumen und der Anteil an weiten 

Grobporen auf den beweideten Standorten ab. 

Das Richards basierte hydraulische Model Hydrus-1D wurde anhand 

bodenphysikalischer Kenngrößen parametrisiert und mit Hilfe der Monitoring 

Daten verifiziert. Simulierte und gemessene Wassergehalte und 

Bodentemperaturen stimmen gut überein. Auch die beweidungsbedingten 

Bodenstrukturänderungen wurden zufriedenstellend widergespiegelt. Dabei 

zeigte sich, dass die Bodenwärmeflüsse mit zunehmender Beweidungsintensität 

anstiegen. Bei den Wasserhaushaltskomponenten konnte im Vergleich zu den 

unbeweideten Flächen bei starker Beweidung eine Abnahme des 

Wasserspeichers um 25-45 %, der Interzeption um 50-55 % und der 

Transpiration um 20-30 % sowie eine Zunahme der Evaporation um 25-40%, 

des Oberflächenabflusses um 100-300% und der Tiefensickerung um 100-200% 

festgestellt werden. Eine intensive Beweidung führt demnach durch die 

Verschlechterung von Bodenfunktionen zu einer Reduktion des 

pflanzenverfügbaren Wasserspeichers im Boden und damit zu einer 

Beeinträchtigung der Weidelandproduktivität bei gleichzeitig erhöhtem Risiko für 

Bodenerosion durch Oberflächenabfluss. 

Die Integration eines Frost/Tau-Moduls in HYDRUS-1D verbesserte die 

Modellierung gekoppelter hydraulischer und thermischer Prozesse in Phasen 

mit Bodengefrornis deutlich. Die numerischen Simulationen mit dem „freezing 

model“ konnten Änderungen in den Bodenwassergehalten und Wärmeflüssen 

ix

unter Gefrier- und Wiederauftaubedingungen sehr gut vorhersagen wobei sich 

erwartungsgemäß Unterschiede als Funktion der Beweidungsintensität ergaben. 

Allerdings wurden die Bodenwassergehalte nach der Frühjahrsschneeschmelze 

leicht überschätz was auf eine unzureichende Prozessabbildung bei der 

Entstehung von Oberflächenabflüssen auf noch gefrorenen Bodenlagen 

zurückgeführt werden kann. Das Modell ist an dieser Stelle noch 

verbesserungswürdig, was Ziel weiterer Untersuchungen sein sollte, um 

genauere Vorhersagen lateraler Wasserflüsse auf gefrorenen Bodenlagen bei 

der Schneeschmelze machen zu können. 

x

II List of figures 

Fig. 2.1. Location of geostatistical areas in the five plots. UG 79: ungrazed since 

1979, UG 99: ungrazed since 1999, WG: winter grazed, CG: continuous grazed, 

and HG: heavily grazed. Each plot has an area of 105 m×135 m. Sampling 

points are spaced every 15 m with a subgrid spacing of 5 m (the orthogonal 

boxes with dash line in WG and CG). Two larger geostatistical grids in WG and 

CG have an area of 300 m×550 m with a sampling points spacing of 50 m and a 

subgrid spacing of 10 m. 15 

Fig. 2.2. Semivariograms of (a) SWC, (b) Ln(K), (c) Ln(WDPT), (d) SS, (e) SOC, 

(f) BD, (g) sand content (%), and (h) silt content (%) in UG 79. K: hydraulic 

conductivity (cm d -1 ); WDPT: water drop penetration time (s); SS: shear strength 

(kPa); SWC: soil water content (cm 3 cm -3 ); SOC: soil organic carbon (g kg -1 ); BD: 

bulk density (g cm -3 ). 26 

Fig. 2.3. Ordinary kriged maps of (a) SWC, (b) Ln(K), (c) Ln(WDPT), (d) SS, (e) 

SOC, (f) BD, (g) sand content (%), and (h) silt content (%) in UG 79. K: hydraulic 

conductivity (cm d -1 ); WDPT: water drop penetration time (s); SS: shear strength 

(kPa); SWC: soil water content (cm 3 cm -3 ); SOC: soil organic carbon (g kg -1 ); BD: 

bulk density (g cm -3 ). 27 

Fig. 2.4. Scattergram plots and regression analysis for selected soil properties in 

UG 79 and HG between Ln(WDPT) and Ln(K) (a); Ln(K) and BD (b); Ln(WDPT) 

and SOC (c); and SS and BD (d). K: hydraulic conductivity (cm d -1 ); WDPT: 

water drop penetration time (s); SS: shear strength (kPa); SOC: soil organic 

carbon (g kg -1 ); BD: bulk density (g cm -3 ). 28 

Fig. 3.1. Location of the sampling plots delineated as:UG 79: ungrazed since 

1979, UG 99: ungrazed since 1999, WG: winter grazed, CG: continuous grazed, 

and HG: heavily grazed. Each plot has an area of 105 m×135 m. Sampling 

points were spaced at 15 m with a subgrid spacing of 5 m (the orthogonal boxes 

with dash line in WG and CG). Two larger geostatistical grids in WG and CG had 

an area of 300 m×550 m with a grid spacing of 50 m and a subgrid spacing of 10 

m. 43 

Fig. 3.2. Time series of (a) mean soil moisture content, (b) soil moisture variance 

and (c) soil moisture range for five plots during the sampling period 2004-2006. 

48 

Fig. 3.3. Correlations of variance to mean soil moisture content for five plots. 

51 

Fig. 3.4. Semiariograms and cross-semivariograms maps of soil variables and 

models for the linear Model of Coregionalization exemplified by UG 99. SOC: 

xi

soil organic carbon (g kg -1 ), BD: bulk density (g cm -3 ). 52 

Fig. 4.1. Demonstration of sample grids in the experimental site (example for UG 

79). Each site has an area of 105 m×135 m. Sampling points are spaced every 

15 m with a subgrid spacing of 5 m (Hollow circles donate the field long-term 

monitoring positions). 69 

Fig. 4.2. Ranked MRD of soil moisture in UG 79 in different water conditions 

during 3-yr measurement period (All: all measurements, D: dry, M: medium, W: 

wet). Vertical bars correspond to associated time standard deviations, labeled 

numbers are the time stability points that at least appear three times from four 

soil water conditions. 73 

Fig. 4.3. Ranked MRD of soil moisture in UG 99 in different water conditions 

during 3-yr measurement period (All: all measurements, D: dry, M: medium, W: 

wet). Vertical bars correspond to associated time standard deviations, labeled 



Fig. 4.4. Ranked MRD of soil moisture in WG in different water conditions during 

3-yr measurement period (All: all conditions, D: dry, M: medium, W: wet). Vertical 

bars correspond to associated time standard deviations, labeled numbers are 

the time stability points that at least appear three times from four soil water 

conditions. 74 

Fig. 4.5. Ranked MRD of soil moisture in HG in different water conditions during 

3-yr measurement period (All: all measurements, D: dry, M: medium, W: wet). 

Vertical bars correspond to associated time standard deviations, labeled 



Fig. 4.6. Distribution of drier points than overall average moisture content in WG 

under different water conditions during 3-yr measurement period (All: all 

measurements, D: dry, M: medium, W: wet; labeled numbers are the points that 

soil moisture drier than field average moisture content). 75 

Fig. 4.7. Field mean soil moisture versus soil moisture of time stable point in four 

sites during the period of 2004-2006 (a:UG 79; b: UG 99; c: WG; and d: HG). 

77 

Fig. 4.8. Cumulative infiltration and fitting curve for the p72 location in UG 79. 

80 

Fig. 4.9. Soil moisture comparison between measured and simulated results in 

UG 79 during the growing period in 2006 (Measured: measured value in field 

long-term monitoring site; Modelled: modelled result based HYDRUS-1D; TSP: 

xii

measured soil moisture in time stability point). 81 

Fig. 5.1. Pore size distribution for the four grazing intensities in four soil depths. 

Sparse: large pore size (>50 μm), pF

WG from August 2005 to July 2006. 124 

Fig. 6.3. Soil temperature and amplitude in topsoil (2 cm) as a function of grazing 

intensity (UG 99 and WG) from August 2005 to July 2006. 125 

Fig. 6.4. Measured and simulated soil moisture and temperature at 5 cm (a, b, c), 

20, 40, and 100 cm depth in UG 79 during the whole year of 2006 (M: Measured 

liquid water content; S: Simulated total water content running snow routine; and 

F: Simulated liquid water content running freezing model). 128 

Fig. 6.5 Measured and simulated soil moisture and temperature at 5 cm (a, b, c), 

20, 40, and 100 cm depth in WG during the whole year of 2006 (M: Measured 

liquid water content; S: Simulated total water content running snow routine; and 

F: Simulated liquid water content running freezing model). 129 

Fig. 6.6. Rainfall, measured air and soil temperature, simulated snow depth and 

runoff during the whole year of 2006. 131 

xiv

III List of tables 

Table 2.1. Descriptive statistics of soil water content measured on three 

sampling soil water statuses in 2004. 19 

Table 2.2. Descriptive statistics of shear strength measured on three sampling 

soil water statuses in 2004. 20 

Table 2.3. Descriptive statistics of Ln (WDPT) measured on three sampling soil 

water statuses in 2004. 21 

Table 2.4. Isotropic variogram parameters for soil water content on three 

sampling soil water statuses in 2004. 22 

Table 2.5. Isotropic variogram parameters for shear strength on three sampling 

soil water statuses in 2004. 23 

Table 2.6. Isotropic variogram parameters for Ln(WDPT) on three sampling soil 

water statuses in 2004. 24 

Table 2.7. Descriptive statistics of different parameters in UG 79 in 2005. 25 

Table 2.8. Isotropic variogram parameters for different properties in UG 79 in 

2005. 25 

Table 2.9. Correlation matrix for the analyzed variables in the five plots. 29 

Table 3.1 The main characteristics of the different grazing intensity plots 

sampled: UG 79: ungrazed since 1979, UG 99: ungrazed since 1999, WG: 

winter grazed, CG: continuous grazed, and HG: heavily grazed 47 

Table 3.2 Descriptive statistics of soil moisture for three moisture conditions 

measured during 2004-2006 49 

Table 3.3 Isotropic variogram parameters of soil moisture for three moisture 

conditions measured during 2004-2006 50 

Table 3.4 Correlation matrix between soil moisture and water-related factors for 

three soil moisture conditions (P

variation (%) at different spatial scales for different grazing intensities 57 

Table 4.1. Correlation matrix between soil moistures of the sampling grid for 

different water conditions (P

1.1 Background 

Chapter 1 Introduction 

1. Introduction 

Rangelands occupy about 30–50% of the earth’s land area, and they supply 

more than 80% of the feed of the livestock in Asia and Africa, 25% in North and 

Central America and 50% in the rest of the world (World Resources Institute, 

2000). Inner Mongolia grassland (North China, Fig. 1), one key part of the 

temperate Eurasian steppe belt from Eastern Europe to Eastern Asia, accounts 

for the largest Chinese grassland with nearly 20% of Chinese grassland areas 

(400 Mha). Grazing, therefore, is a major local land use form historically. This 

land is considered to be a fertile pasture and particularly flourishing in Yuan 

dynasty (A.D. 1271-1368). However, in the past few decades, due to the 

increase in population and shift in the socio-economic system, it is subject to an 

increased grazing pressure that leads to the broad degradation and risk of 

severe soil erosion, nutrient depletion and desertification (Li et al., 2000). 

Striking evidence for this process is given by numerous sandstorms attacking, 

e.g. Beijing with an increasing frequency. Yet, processes and factors responsible 

for the grassland degradation are not fully been understood, although there is no 

doubt that grazing management makes a major contribution. 

Fig. 1.1. Location of the investigated experimental site, Xilin River catchment, Inner 

Mongolia, North China. 

1

1.2 State of the art 

Given the vital importance for the production of live stock and the environmental 

changes, it is crucial to have a thorough understanding of the mechanisms that 

maintain or change the ecosystem in response to the changes of land 

management. In order to develop sustainable management strategies, it is 

essential to investigate the processes emerging from pasture practices. Since 

many of the rangelands are located in arid or semi-arid areas, water is one of the 

key variables determining the fate of ecosystem situated in such regions. 

Therefore, the understanding of the hydrologic processes is critically important 

(Sugita et al., 2007). Especially, the identification and prediction of soil moisture 

patterns (spatio-temporal organization) at different scales are necessary to 

understand underlying processes controlling the hydrological response of a 

region or even catchment (Grayson et al., 2002; Lin et al., 2006). 

It has been accepted that environmental variables that are used to evaluate an 

ecosystem, e.g. soil properties, are spatial dependent (Goovaerts, 1998). The 

spatial dependency is commonly characterized and quantified by geostatistical 

methods such as autocorrelation and variogram analysis. Furthermore, 

multivariate geostatistics can deal with multivariate spatial data to classify the 

correlation of regionalization variables and to define scale-dependent 

relationships (Goovaerts, 1992). This is needed to extrapolate or interpolate 

variables from different spatial scales when employing either upscaling or 

downscaling procedures (Western et al., 1999). For instance, when matter fluxes 

on various scales are investigated, it could be very useful to upscale water flux 

with the aid of physically-based hydrological models from plot to catena and 

finally to a regional scale (Fig. 2). 

Furthermore, it is possible to select monitoring site(s) representing the true field 

mean conditions in terms of temporal stability concept. If this is assumed, it is 

suitable to turn to the detailed scenario analysis on the plot scale to evaluate 

effects of grazing on hydraulic process after investigating a general natural 

process in the regional or field scale. In addition, a hydrological model normally 

selected the modeled location without a prior analysis so that the modeled result 

2


is also uncertain. Thus, the hydraulic model, linked with representative of 

monitoring locations, should be more valuable to get a comparable conclusion 

considering spatial variability. 

year 

month 

day 

hour 

aggregate 

soil moisture 

regime 

grazing effect 

aggregation 

1D- modeling 

soil moisture 

water flux 

scaling 

flow 

separation 

HYDRUS 

model 

2D- modeling 

water budget 

computing 

hydrograph 

rainfall etc. 

plot catena catchment 

Fig. 1.2. Methodology to understand and quantify the hydrological processes at different 

spatio-temporal scales (personal communications with Prof H. Zepp, Ruhr-University 

Bochum, Germany; 2004). 

Land management is reported to influence soil hydraulic and thermal properties 

by altering soil and plant functions (Flerchinger et al., 2003). In pasture areas, 

soil mechanical disturbances due to animal trampling, interlinked with 

hydrological changes, often have detrimental effects on soil properties. 

Particularly in the topsoil, soil deformation is characterized by a decrease of pore 

volume and an altered pore size distribution, which both affect water and air 

conductivity (Willat and Pullar, 1983; Krümmelbein et al., 2006), and soil water 

retention characteristics (Martinez and Zinck, 2004; Kutilek et al., 2006). These 

disturbances of soil structure further decrease water infiltrability and increase 

surface runoff. This will not only cause soil erosion by water and nutrient 

3

depletion, but also decrease the water availability and further increase soil 

erosion by wind. Thus, it is essential to quantify and predict management effects 

on soil properties to assess their consequences on plant production and the 

environment. Many studies have focused on the effects of land management on 

soil properties (Green et al., 2003). However, few studies have dealt with the 

consequences of these practices on water and heat fluxes (Chung and Horton, 

1987; Peth and Horn 2006; Ndiaye et al., 2007). Especially, considering the 

stress-dependent changes of the environment, an adequate account of water 

flow processes and its spatial variability based on hydrological models is still 

lacking due to the complexity of the soil system (Ndiaye et al., 2007), although 

there are widely applicable hydrological models like SHAW (Flerchinger and 

Saxton, 1989), HYDRUS (Šimůnek et al., 1998) and SWAT (Neitsch et al., 

2002). 

Especially, considering the amount of plant available water as one of the most 

limiting factors for sustainable grassland development, the effects of grazing on 

water-related mechanisms and water budgets is urgently needed to be 

understood in our studied area (Chen and Wang, 2000). Some studies have 

shown that grazing increases evaporation but decreases transpiration because 

of increasing bare ground area and reducing amount of biomass. Therefore, the 

heavily grazed sites are more susceptible to drought, and prone to being 

degraded due to the reduction of available soil water (Snyman, 2005; Zhang et 

al., 2005). However, to which extent grazing affects evapotranspiration, and how 

far it is partitioned into transpiration and evaporation remains unclear (Leenhardt 

et al., 1995). Furthermore, although water interception by the canopy and plant 

residues might occupy large volumes of water in steppe ecosystems, it normally 

is neglected or mixed with soil evaporation. 

Moreover, it is widely recognized that the movement of water and heat are 

closely coupled and strongly affected by each other, but their mutual interactions 

are rarely considered in practical applications. Frozen soil normally reduces 

infiltration capacity dramatically due to blocking of pores by the formation of ice 

lenses. As a result, lateral flow and snow melt may release huge quantities of 

4


water in spring and early summer, and cause surface runoff (Lewkowicz and 

Kokelj, 2002; Bayard et al., 2005). Especially, in cold and arid regions like in our 

studied area, quantification of the freezing and thawing effects on the water 

availability, runoff generation and groundwater recharges are essential. Currently, 

the model of snow hydrology and soil freezing and thawing processes is 

comparable lacking not only for model complexity but also for lacking data to 

fully parameterize or validate the model. 

1.3 Objectives 

In the light of the absence of interdisciplinary studies of rangelands in 

northeastern Asia, a project called MAGIM (Matter fluxes in grasslands of Inner 

Mongolia as influenced by stocking rate) has been funded by DFG (Deutsche 

Forschungsgemeinschaft) for 3 years (2004-2006; website: 

http://www.magim.net). The principal objective of MAGIM is to understand how 

grazing of steppe ecosystem feedbacks on water, nitrogen and carbon fluxes on 

site and regional scales. This comprehensive task requires research on various 

scales involving detailed laboratory and field experiments as well as concurrent 

model development and application. 

Under the framework of the project MAGIM, our objectives are to explore soil 

physical, hydrological and ecological processes by combining detailed field and 

laboratory measurements with process oriented modeling techniques to 

evaluate the influence of rangeland management (i.e. ungrazed since 1979=UG 

79, ungrazed since 1999=UG 99, winter grazed=WG, continuously grazed=CG 

and heavily grazed=HG; Fig. 1) on water and heat fluxes in particular. The study 

presented here is focusing on following objectives: 

Chapter 2: “Spatial variability of soil properties affected by grazing 

intensity in Inner Mongolia grassland” 

Although many investigations have been dealt with the productivity and stability 

of this steppe ecosystem, conclusions aiming at the development of sustainable 

management strategies are hard to be derived from those studies. It is reasoned 

5

that firstly the complex and dynamic interaction of soil properties influencing 

plant growth often remains unidentified, and secondly spatial auto- and 

cross-correlations of soil properties is usually ignored. This chapter serves to 

close this gap by performing a combined analysis of soil hydraulic and 

mechanical processes as well as their spatial variation. Therefore, our objectives 

are (i) to estimate the effect of grazing intensity on soil mechanical and hydraulic 

properties on the plot scale; (ii) to investigate the spatial variability of soil 

properties in the surface soil; and (iii) to explore correlations between soil 

properties. 

Chapter 3: “Spatio-temporal variability of soil moisture in grazed steppe 

areas investigated by multivariate geostatistics” 

Spatio-temporal soil moisture patterns are not only fundamentals for a spatially 

distributed hydrological modeling, but also keystones for the exploration of 

hydraulic processes and mechanisms in site-specific ecosystems. Identification 

of factors that control soil moisture spatio-temporal patterns is therefore 

important. The role of controlling factors of soil moisture is either unclear or even 

contradicting between different studies. It is reasoned for the lack of consistent 

soil moisture sampling in both space and time and of comprehensive data sets of 

possible controlling factors. For this approach, a combined grided and nested 

sampling of topsoil moisture and various environmental attributes was 

conducted at five sites during the growing seasons of 2004-2006. The objectives 

of this study are (i) to quantify the spatio-temporal variability of topsoil soil 

moisture as affected by grazing intensity, (ii) to ascertain the main factors 

controlling soil moisture patterns in semiarid environment, and (iii) to explore the 

multivariate spatial scale of controlling factors by multivariate and geostatistical 

approaches. 

Chapter 4: “Temporal stability of soil moisture in a semi-arid steppe and its 

application in model result validation” 

The high spatial variability of soil moisture and the small measurement support 

6


requires therefore appropriate sampling strategies and monitoring (or modeling) 

sites. The temporal stability concept might aid to estimate field mean water 

content whether hydrological model is applied in a time stability point. However, 

until now this has not been fully used and tested to estimate the water content 

and flux for a given probability level by a hydrological model. The purposes of 

this study are: i) to investigate the temporal stability of soil moisture under 

different water conditions, ii) to explore the grazing impact on temporal stability 

of soil moisture, and iii) to infer how far the time stability concept can be applied 

for the hydrological model to make reliable estimates of soil moisture. 

Chapter 5: “Modeling grazing effects on coupled water and heat fluxes in 

Inner Mongolia grassland” 

What are the grazing influences on soil bulk density, crusting, aggregation and 

hydrophobic features? How do parameters like soil texture, water retention 

properties and infiltration capacity change the modeling results of soil water 

movement? In this paper, in situ measurements of soil, plant and weather data 

were used to parameterize the model HYDRUS-1D, by which water and heat 

fluxes were simulated as a function of grazing intensity. The modeling results 

were verified by comparison with the measured data of soil water and 

temperature. The objectives of this study are (i) to quantify water and heat 

budgets affected by grazing intensity via calibrating HYDRUS-1D, and (ii) to 

define water use mechanisms in Inner Mongolia grassland. 

Chapter 6: “Modeling of Coupled Water and Heat Transfer in Freezing and 

Thawing Soil” 

In this study, we further address the field application of a new freezing code 

incorporated into HYDRUS-1D, which numerically solves coupled equations 

governing phase changes between water and ice, and heat transport with a 

mass- and energy-conservative method. The seasonally frozen soils in Inner 

Mongolia allow us to evaluate the effect of snowmelt and soil thawing water on 

surface runoff and seasonal water cycle. Specifically, we will focus on discussing 

7

the following questions: (i) How well does HYDRUS-1D with and without frozen 

soil scheme simulate soil water and temperature? (ii) What is the impact of 

frozen soil scheme on soil temperature, soil moisture and runoff simulations? (iii) 

What is the role of land use in the simulation of soil water and heat fluxes 

concerning the effects of grazing on heat and water transfer rate? 

1.3 References 

Bayard, D., M. Stahli, A. Parriaux, and H. Fluhler. The influence of seasonally 

8 

frozen soil on the snowmelt runoff at two Alpine sites in southern 

Switzerland. 2005. J. Hydrol. 309:66–84. 

Chen, Z.Z., and S.P. Wang. 2000. Chinese Typical Grassland Ecosystem. China 

Science Press, pp. 106–109. 

Chung, S., and R. Horton. 1987. Soil heat and water flow with a partial surface 

mulch. Water Resour. Res. 23 (12):2175–2186. 

Flerchinger, G.N., and K.E. Saxton. 1989. Simultaneous heat and water model of 

a freezing snow-residue-soil system I. Theory and development. Trans. 

ASAE 32:565–571. 

Flerchinger, G.N., T.J. Sauer, and R.A. Aiken. 2003. Effects of crop residue cover 

and architecture on heat and water transfer. Geoderma 116:217–233. 

Goovaerts P. 1992. Factorial kriging analysis–a useful tool for exploring the 

structure of multivariate spatial soil information. Journal of Soil Science 143, 

597–619. 

Goovaerts, P. 1998. Geostatistical tools for characterizing the spatial variability 

of microbiological and physico-chemical soil properties. Biol Fertil Soils 27, 

315–334. 

Grayson, R., G. Blöschl, A.W. Western, and T.A. McMahon. 2002. Advances in 

the use of observed spatial patterns of catchment hydrological response. 

Adv. Water Resour. 25:1313–1334. 

Green, T.R., L.R. Ahuja, and J.G. Benjamin. 2003. Advances and challenges in 

predicting agricultural management effects on soil hydraulic properties. 

Geoderma 116:3–27.


Krümmelbein, J., Z. Wang, Y. Zhao, S. Peth, and R. Horn. 2006. Influence of 

various grazing intensities on soil stability, soil structure and water balance 

of grassland soils in Inner Mongolia, P. R. China. In: R. Horn, H. Fleige, S. 

Peth, and X. Peng (Eds.). Advances in Geoecology, Vol. 38, pp. 93–101. 

Leenhardt, D., M. Voltz, and S. Rambal. 1995. A survey of several agroclimatic 

soil water balance models with reference to their spatial application. Eur. J. 

Agron. 4:1–14. 

Li, S.G., Y. Harazono, T. Oikawa, H.L. Zhao, Z.Y. He, and X.L. Chang. 2000. 

Grassland desertification by grazing and the resulting micrometeorological 

changes in Inner Mongolia. Agric. For. Meteorol. 102:125–137. 

Lin, H. 2006. Temporal stability of soil moisture spatial pattern and subsurface 

preferential flow pathways in the Shale Hills catchment, Vadose Zone 

Journal 5:317–340. 

Neitsch, S.L., J.G. Arnold, J.R. Kiniry, J.R. Williams, and K.W. King. 2002. Soil 

and water assessment tool theoretical documentation version 2000, 

Agricultural Research Service and Texas Agricultural Experiment Station, 

Texas. 

Ndiaye, B., J. Molénat, V. Hallaire, C. Gascuel, and Y. Hamon. 2007. Effects of 

agricultural practices on hydraulic properties and water movement in soils 

in Brittany (France). Soil Till. Res. 93:251–263. 

Peth, S., and R. Horn. 2006. Consequences of grazing on soil physical and 

mechanical properties in forest and tundra environments. In: B.C. Forbes, 

M. Bölter, L. Müller-Wille, J. Hukkinen, F. Müller, N. Gunslay, Y. Konstatinov 

(Eds.). Ecological Studies, Vol. 184, pp. 217–243. 

Snyman, H.A. 2005. Rangeland degradation in a semi-arid South Africa - I: 

influence on seasonal root distribution, root/shoot ratios and water-use 

efficiency. J. Arid Environ. 60:457–481. 

Šimůnek, J., M. Sejna, and M.Th. van Genuchten. 1998. The HYDRUS-1D 

software package for simulating the one dimensional movement of water, 

heat, and multiple solutes in variably-saturated media. Version 2.0. 

9

10 

IGWMC-TPS-70. Int. GroundWater Modeling Center, Colorado School of 

Mines, Golden. 

Sugita, M., J. Asanuma, M. Tsujimura, S. Mariko, M.-J. Lu, F. Kimura, D. Azzaya, 

T. Adyasuren. 2007. An overview of the rangelands 

atmosphere–hydrosphere–biosphere interaction study experiment in 

northeastern Asia (RAISE). J. Hydrol. 333:3–20. 

Western, A.W., and G. Blöschl. 1999. On the spatial scaling of soil moisture. J. 

Hydrol. 217:203–224. 

World Resources Institute, 2000. World Resources 2000–2001. People and 

Ecosystems: The Fraying Web of Life. World Resources Institute, 

Washington, DC, USA, 400 pp. 

Zhang, Y., E. Munkhtsetseg, T. Kadota, and T. Ohata. 2005. An observational 

study of ecohydrology of a sparse grassland at the edge of the Eurasian 

cryosphere in Mongolia. J. Geophys. Res. 110, D14103. 

doi:10.1029/2004JD005474.

Chapter 2 Spatial variability of soil properties affected by grazing intensity in Inner Mongolia grassland 

2. Spatial variability of soil properties affected by 

grazing intensity in Inner Mongolia grassland 

Ying Zhao, Stephan Peth, Julia Krümmelbein, Rainer Horn, Zhongyan Wang, 

Markus Steffens, Carsten Hoffmann, Xinhua Peng 

Ecological Modelling. 205:241–254. 

Abstract 

Analysis of the spatial variability of soil properties is important to interpret 

the site-specific ecosystems not only with respect to process investigations but 

also to model upscaling. This paper aims to study the effects of the grazing 

intensity on soil physical and mechanical properties and their interactions in a 

Leymus chinensis steppe of the Xilin river watershed, Inner Mongolia, China. 

The investigated sites were subjected to five grazing intensities (Ungrazed since 

1979, Ungrazed since 1999, Winter grazing, Continuous grazing and Heavy 

grazing). Soil water content (SWC), hydraulic conductivity (K), water drop 

penetration time (WDPT), shear strength (SS), soil organic carbon (SOC) 

concentration, bulk density (BD), and soil texture were measured at a grid with 

15 m sampling distance on the surface soil during the period of 2004-2005. The 

data were analyzed using descriptive statistics and geostatistics. The correlation 

and interaction between soil properties were analyzed by the methods of 

Pearson correlation, partial correlation and multiple regression analysis. The 

results showed that spatial distributions of soil properties could be well described 

by spherical or exponential models. The ranges of spatial dependence were the 

highest for WDPT and the lowest for SS. Grazing resulted in decreasing SWC, 

SOC and WDPT but increased BD and SS. Multiple regression analysis showed 

significant correlations among SWC, K, WDPT, SOC and BD; as well as 

between SS and silt content. Soil compaction induced by sheep trampling, 

especially in the heavily grazed site, inclined to a homogenous spatial 

distribution of soil properties, which will possibly enhance soil vulnerability to 

water and nutrient loss, and consequently reduce the plant available water and 

11

thus grassland productivity. 

Keywords: Grazing intensity; Spatial variability; Geostatistics; Soil properties; 

Correlations 


12 

Inner Mongolia grassland has been severely deteriorated in recent years 

because of inappropriate management, that resulted in an increasing 

degradation of the vegetation cover followed by pronounced soil erosion and 

nutrient depletion (Li et al., 2000). These processes may cause low water 

storage capacity and loss of soil fertility, consequently decrease the grassland 

productivity. Therefore, to develop sustainable management strategies, the 

processes emerging from pasture degradation by grazing are urgent to be 

understood as they play a key role in the ecosystem stability. 

Grazing associated with animal activity altered hydraulic and mechanical 

soil properties. Greenwood et al. (1997) found significant differences in 

unsaturated hydraulic conductivity, soil strength and bulk density for the surface 

soil between ungrazed and grazed pastures. Warren et al. (1986) reported that 

trampling animals caused soil deformation by exerting high ground contact 

pressures under their hooves. Besides soil compression, shear stresses further 

destroy soil structure by kneading and homogenization, which mainly related to 

a change of macroporosity and the connectivity of the pore system. Destruction 

of soil structure was observed for sheep grazing (Proffitt et al., 1995), cattle 

trampling (Pietola et al., 2005), and reindeer herding (Peth et al., 2003). Such 

changes depend not only on the magnitude of applied stresses, but also on the 

soil moisture controlled aggregate stability at the time of trampling (Horn and 

Fleige, 2003; Richard et al., 2001). Furthermore, animal trampling is mostly not 

static but must be considered as a short time dynamic process, where the soil 

undergoes repeated sequences of loading, unloading and reloading events 

(Peth and Horn, 2006). This dynamic soil compression is more intense due to a 

pronounced inter-particle shear deformation compared to the compaction under 

static loading. Finally, this will result in an intensive reduction of soil infiltrability


and in a water surplus in the topsoil layer after rainfall, which in turn decreases 

aggregate strength in the topsoil. 

Plant available water capacity may be reduced by soil compaction. Reduced 

plant available water combined with short vegetation periods in the semi-arid 

grassland limits plant growth and reduces the input of litter, consequently 

influence the storage of organic carbon in soil. Particularly upper soil layers 

show high concentration of soil organic carbon and are first affected by 

management (Kelly et al., 1996). High concentration of organic carbon, 

especially when it is rich in hydrophobic compounds, restrains water from 

entering into the soil by increasing the water repellency (Ritsema, 1996). 

Although this is an important “indirect” effect of grazing on the soil infiltrability it 

has not yet been well described and quantified. 

Soil properties have often been reported to show a strong spatial 

dependence (Shouse et al., 1995; Lophaven et al., 2006). The spatial 

dependency is commonly characterized and quantified by geostatistical methods 

such as autocorrelation and variogram analysis. Such spatial analysis is 

necessary to perform sound interpolation when producing contour maps and to 

simultaneously provide an estimate of the variance of the interpolated values 

(Goovaerts, 1998). Kriging, an interpolation procedure, which provides best 

linear and unbiased estimation, has been universally applied in the 

environmental sciences to analyze spatial variability and to resolve site-specific 

problems, e.g. with respect to SWC (Goovaerts and Sonnet, 1993; Famiglietti et 

al., 1998; Western et al., 1998, Buttafuocoa et al., 2005), shear strength (Cassel 

and Nelson, 1985; Júnior et al., 2006) and water repellency (Gerke et al., 2001). 

The correlations between soil properties are very complex, especially on 

plot or regional scale. These characteristics make it very hard to interpret the 

various studies and to predict the relevant processes and mechanisms (Ludovisi 

et al., 2005). Many attempts have been made to establish the relation among 

texture, structure, aggregate stability, bulk density and organic matter content of 

the soil and its resistance to water infiltration. Dekker and Ritsema (1994) did not 

find a significant relationship between the persistence of potential water 

13

epellency and the organic matter content. However, Greiffenhagen et al. (2006) 

showed a good correlation between soil organic matter, bulk density and water 

characteristics which could be easily used to predict plant available water. Lavee 

et al. (1996) found a correlation between aggregate stability and soil organic 

matter content and presumed that grazing and human activities interfere with the 

development of aggregate stability. Fullen et al. (1996) argued that soil organic 

matter tended to be associated with silt and clay contents rather than with sand 

content for Hilton soils. However, aggregate stability also might relate with soil 

texture. A good relation among soil strength, texture and stress dependent 

changes of the pore size distribution were reported by Horn and Fleige (2003). 

14 

The current research focuses on the seasonal changes and spatial 

dependence of soil properties. Although many investigations have been dealt 

with the productivity and stability of grassland ecosystems, conclusions aiming 

at the development of sustainable management strategies are hard to be derived 

from those studies. It is reasoned that firstly the complex and dynamic 

interaction of soil properties influencing plant growth often remains unidentified, 

and secondly spatial autocorrelation/variation of soil related parameters is 

usually ignored. The paper serves to close this gap by performing a combined 

analysis of soil hydraulic and mechanical processes as well as their spatial 

variation. Therefore, our objectives are (i) to estimate the effect of grazing 

intensity on soil mechanical and hydraulic properties on the plot scale; (ii) to 

investigate the spatial variability of soil properties in the surface soil; and (iii) to 

explore the correlation between soil properties. 

2. Materials and methods 

2.1. Experimental site and layout 

The experimental area is located within the Xilin River Basin (43º38´N; 

116º42´E), Eastern Inner Mongolia and administered by the Inner Mongolia 

Grassland Ecosystem Research Station (IMGERS), Institute of Botany, Chinese 

Academy of Sciences. Vegetation is characterized as a Leymus chinensis 

steppe, which is typical semi-arid grassland in the Eurasian mid-latitude zone. 

The growing season usually starts in May and ends in late September. Under


temperate semi-arid continental climate, the annual mean air temperature was 

0.7 °C and the annual total precipitation was 343 mm, of which 60-80% was 

received from June to August. The experimental site is on a smooth rolling 

landscape, approximately 1270 m in altitude with a mean relative height of 20-30 

m and slope of an inclination 100 cm, followed by an Ach-layer. 

Fig. 2.1. Location of geostatistical areas in the five plots. UG 79: ungrazed since 1979, UG 99: 

ungrazed since 1999, WG: winter grazed, CG: continuous grazed, and HG: heavily grazed. Each 

plot has an area of 105 m×135 m. Sampling points are spaced every 15 m with a subgrid spacing 

of 5 m (the orthogonal boxes with dash line in WG and CG). Two larger geostatistical grids in WG 

and CG have an area of 300 m×550 m with a sampling points spacing of 50 m and a subgrid 

spacing of 10 m. 

Before 1979, the whole area is lightly grazed by herds that are composed of 

70-90% sheep and 10-30% goats. Our measurements were conducted in five 

field plots of different grazing intensities: two ungrazed plots, fenced since 1979 

and 1999 respectively (hereinafter referred to as UG 79, 24 ha and UG 99, 35 

ha); and three grazed plots, one grazed only during winter with 0.5 sheep units 

15

(ewe and lamb) ha -1 yr -1 (WG, 40 ha), one continuously grazed site with 1.2 

sheep units ha -1 yr -1 (CG, 250 ha) and one heavily grazed site with 2.0 sheep 

units ha -1 yr -1 (HG, 100 ha). In each plot, a regular sampling grid (Fig. 1), 

positioned by differential GPS with UTM system, has been set up for the 

geostatistical analysis. The individual grids covered an area of 105 m×135 m 

with a sampling spacing of 15 m and subgrid spacing of 5 m. 100 points in each 

plot were used for geostatistical analysis. In addition, two larger geostatistical 

grids in WG and CG covering an area of 300 m×550 m with a sampling spacing 

of 50 m and subgrid spacing of 10 m had been installed to further explore 

possible scale effects (data not shown). 

2.2. Measurement of soil properties 

16 

On the grid sites, soil moisture was measured at the surface soil layer (0-6 

cm) by a HH2 Moisture Meter (Theta-probe Type ML2x, Delta-T devices Ltd, 

England) after soil specific calibration. Water repellency was quantified using the 

water drop penetration time test by recording the elapsed time of a distilled water 

droplet (0.5 mm 3 , using a standard glass pipette) infiltrating completely into a 

smooth soil surface. According to De Bano (1981), the water repellency was 

classified as wettable (600 s). Shear strength was measured by a Hand-held vane tester (Geonor 

H-60, Norway). Hydraulic conductivity was measured by a Mini-disk Infiltrometer 

(Decagon devices, USA) at a suction of 2 hPa. The soil surface was cleaned 

from litter prior to the measurement to ensure a good contact between the soil 

and the infiltrometer. Infiltration time was recorded at regular volume intervals 

from which the hydraulic conductivity has been calculated according to Zhang 

(1997). The calibrated van Genuchten parameters were obtained from RETC 

software based on laboratory analysis of the water retention curve. In addition, 

undisturbed samples were taken in triplicate at each grid point using a stainless 

steel cylinder (100 cm 3 ) for lab analysis. Bulk density of the soil was calculated 

with the mass of the oven-dry soil (105° C) divided by the core volume. Total C 

concentration was determined in duplicate by dry combustion on a Vario Max 

CNS elemental analyzer (Elementar Analysensysteme GmbH, Hanau). All


samples were free of carbonate so that the total C concentration equals the 

organic carbon concentration. Soil particle size distribution was measured with 

the pipette method. 

2.3. Statistical analysis 

Measured variables in the data set were firstly analyzed using descriptive 

statistical methods. The Shapiro-Wilk test revealed that all measured variables 

were approximately normally distributed, only K and WDPT showed a negative 

skewness. Natural logarithmic transformation Y*=Ln(y+1) (where y is the 

recorded value) for WDPT and Y*=Ln(y) for K were performed to obtain a nearly 

normal distribution before proceeding with the geostatistical analysis (Jongman 

et al., 1987). Data were not transformed back in order to simplify multiple linear 

regression analysis. Correlations among different analyzed parameters were 

tested using Pearson’s correlation coefficient. Semivariograms was used to 

determine the degree of spatial variability. Appropriate model functions were 

fitted to the semivariogram obtained by the maximum likelihood cross-validation 

method (Samper and Carrera, 1990). Before applying the geostatistical tests, 

each variable was also checked for drift, trend and anisotropy (Iqbal et al., 2005). 

The semivariogram (γh) was calculated as follows: 

N ( h) 

⎧ 

⎫ 

1 2 

γ ( h) 

= 2N 

( h) 

⎨∑ 

[ Z( 

xi 

+ h) 

− Z( 

xi 

)] ⎬ 

(1) 

⎩ i= 

1 

⎭ 

where γ(h) is the semivariance for interval class h, N(h) is the number of 

pairs separated by the lag distance h, Z(xi) and Z(xi+h) are values of the 

measured variable at spatial locations i and i+h, respectively. A semivariogram 

model consists of three basic parameters which describe the spatial structure: 

γ(h)=Co+Cs. Co represents the nugget effect; Cs is the structural component; 

Co+Cs is the sill; and the distance at which the sill is reached is the range. Value 

of the proportion of spatial structure Cs/(C0+Cs) is a measure of the proportion 

of sample variance (C0+Cs) that is explained by spatially structured variance 

(Cs). Following Cambardella et al. (1994), The classes of spatial dependence 

were distinguished: strongly spatial dependence (Cs/(C0+Cs) >75%), 

moderately spatial dependence (Cs/(C0+Cs) >25% and

dependence (Cs/(C0+Cs)


Table 2.1. Descriptive statistics of soil water content measured on three sampling soil water 

statuses in 2004. 

Treatment 

Sampling soil 

Water 

statuses 

Min. Max. Mean SD CV Skewness Kurtosis 

UG 79 Dry 0.03 0.09 0.06 0.01 20.00 0 -0.11 

Wet 0.24 0.37 0.30 0.03 9.00 0.16 -0.45 

Medium 0.10 0.24 0.17 0.03 17.75 0.23 -0.28 

UG 99 Dry 0.04 0.10 0.07 0.02 21.43 -0.04 -0.07 

Wet 0.31 0.40 0.36 0.02 6.39 -0.13 -0.79 

Medium 0.12 0.31 0.22 0.03 12.73 -0.04 0.43 

WG Dry 0.02 0.07 0.04 0.01 23.75 0.16 0.52 

Wet 0.29 0.38 0.34 0.02 5.29 -0.09 -0.50 

Medium 0.15 0.23 0.19 0.02 9.47 -0.02 -0.28 

CG Dry 0.02 0.08 0.05 0.01 22.22 0.55 1.27 

Wet 0.29 0.38 0.34 0.02 5.33 -0.10 -0.46 

Medium 0.12 0.24 0.19 0.03 13.30 -0.05 -0.32 

HG Dry - - - - - - - 

Wet 0.21 0.34 0.28 0.03 10.79 -0.02 -0.45 

Medium 0.07 0.21 0.13 0.03 20.54 0.59 0.74 

Min.: Minimum (cm 3 cm -3 ); Max.: Maximum (cm 3 cm -3 ); SD: Standard deviation (cm 3 cm -3 ); CV: 

Coefficient of variation (%); -: No data available. 

changes of WDPT for the three water statuses. Generally, with increasing water 

status the mean WDPT decreased. But WDPT was the highest during the 

medium moisture (Sept, 12), which might be due to more litter accumulation or 

decomposition corresponding with the late season of growth. WDPT was higher 

in the ungrazed plots showing a slight to strong water repellency than in the 

grazed plots where the soil was wettable to slight water repellency. The 

micro-relief of plant surface covered soil surface as a cuticle, which was 

composed of soluble, hydrophobic lipids embedded in a polyester matrix, often 

induces effective water repellency (Holloway, 1994; Dekker and Ritsema, 2000). 

Because of protection from grazing for 25 years in UG 79 and for 5 years in UG 

99, much accumulation of organic matter with hydrophobic components on the 

soil surface could have led to the observed relative strong water repellency. 

19

Table 2.2. Descriptive statistics of shear strength measured on three sampling soil water 


Treatment 

Sampling soil 

water 

statuses 


UG 79 Dry 12.16 45.05 29.43 7.51 25.51 0.46 -0.49 

Wet 16.67 41.74 28.53 5.83 20.42 0.21 -0.45 

Medium 20.00 37.24 28.32 3.81 13.47 -0.08 -0.27 

UG 99 Dry 20.42 53.45 37.24 6.82 5.14 -0.19 0 

Wet 16.01 39.64 27.72 5.32 3.15 0.03 -0.38 

Medium 20.66 43.85 29.85 5.44 3.28 0.54 0.06 

WG Dry 25.23 64.87 42.34 7.84 6.8 0.35 0.41 

Wet 20.42 44.45 30.63 5.53 3.4 0.43 -0.47 

Medium 26.43 51.05 33.63 5.98 3.97 0.05 0.11 

CG Dry 28.53 64.87 45.05 7.81 7.01 0.41 -0.30 

Wet 18.32 54.36 32.13 6.85 5.20 0.53 0.23 

Medium 21.62 45.35 31.08 5.20 3.00 0.36 -0.55 

HG Dry - - - - - - - 

Wet 25.23 54.00 39.04 5.95 3.94 0.24 0.02 

Medium 29.34 53.45 38.14 5.59 3.4 0.73 0.22 

Min.: Minimum (kPa); Max.: Maximum (kPa); SD: Standard deviation (kPa); CV: Coefficient 

of variation (%); -: No data available. 

The proportion of spatial structure derived from the semivariogram indicates 

the existence of moderate to strong spatial dependency for all selected soil 

properties (Tables 4-6). High coefficients of determination (R 2 ) indicated that 

exponential and spherical models fitted the experimental semivariogram data 

well. Furthermore, the small values of reduced sums of squares (RSS) indicated 

the differentiation between the model and the semivariogram was low for nearly 

all parameters and treatments. The proportion of the spatial structure for SWC 

decreased slightly with increasing grazing intensity, indicating that the SWC was 

more homogeneously distributed in grazed plots than in ungrazed plots (Table 4). 

The proportion of the spatial structure for SS slightly decreased with increasing 

grazing intensity, especially in CG (Table 5). The results suggested that the 

20


Table 2.3. Descriptive statistics of Ln (WDPT) measured on three sampling soil water 


Treatment 

Sampling soil 

water 

statuses 


UG 79 Dry 0.47 3.44 1.93 0.69 35.46 0.20 -0.04 

Wet 0.90 2.52 1.34 0.25 18.01 1.26 3.36 

Medium 1.11 5.45 3.16 1.01 31.84 0.05 -0.75 

UG 99 Dry 1.01 4.01 2.07 0.70 33.62 0.57 -0.11 

Wet 1.12 2.36 1.48 0.26 17.36 1.11 1.51 

Medium 1.74 5.99 3.56 0.82 22.92 0.40 1.24 

WG Dry 0.38 2.36 0.97 0.47 47.94 0.96 0.49 

Wet 1.15 2.73 1.93 0.33 17.25 0.30 -0.17 

Medium 1.13 4.98 3.02 0.84 27.72 0 -0.40 

CG Dry 0.36 3.43 1.09 0.63 58.07 1.47 2.36 

Wet 0.59 4.38 2.60 0.72 27.74 0.01 0.15 

Medium 1 2.63 1.78 0.33 18.54 0.09 -0.01 

HG Dry - - - - - - - 

Wet 0.56 2.34 1.24 0.44 35.13 0.62 -0.40 

Medium 0.80 3.24 1.47 0.50 34.31 1.47 2.16 

Min.: Minimum (s); Max.: Maximum (s); SD: Standard deviation (s); CV: Coefficient of 

variation (%); -: No data available. 

continuous grazing might lead to a more homogeneous spatial distribution of SS 

owing to the animal trampling. However, this trend was interrupted for very high 

stocking rate (HG) where spatial dependence increased again at the heavy 

grazing plot. A similar trend was also observed for the range values, which firstly 

increased and then decreased with increasing grazing intensity. This trend 

implies that heavy grazing initiated strong compaction effects with a partial 

heterogeneity likely caused by non-random repeated trampling. Similar to that of 

the SS, the order of spatial dependence for WDPT was higher in the ungrazed 

plots than in the grazed plots and the lowest in CG (Table 6). This result 

indicated an interaction of organic compounds and cohesion is to be assumed 

and hence an influence of the vegetation on soil surface strength (r = -0.68**, 

21

Table 9). In the topsoil layer, water repellency might be caused mainly by 

hydrophobic compounds from decomposed soil organic matter (Piccolo and 

Mbagwu, 1999). Together with the initially existing heterogeneous distribution of 

various repellent substances, subsequent relocation and the transport along 

preferential flow paths might contribute to the spatial variability of WDPT and 

SWC. 

Table 2.4. Isotropic variogram parameters for soil water content on three sampling soil 

water statuses in 2004. 

Treatment 

Sampling 

soil 

water 

statuses 

Nugget 

Co 

Sill 

Co+Cs 

Range 

A0 (m) 

Proportion 

Cs/(Co+Cs) 

R 2 RSS γ-Model 

UG 79 Dry 7.0E-06 1.5E-04 28 0.95 0.55 1.3E-09 Spherical 

Wet 1.0E-06 7.2E-04 28 1.00 0.73 1.8E-08 Spherical 

Medium 4.9E-05 9.0E-04 20 0.95 0.46 9.0E-10 Spherical 

UG 99 Dry 4.0E-06 2.3E-04 30 0.98 0.72 2.6E-09 Spherical 



WG Dry 1.5E-05 9.4E-05 25 0.84 0.19 4.7E-10 Exponential 

Wet 5.1E-05 3.8E-04 29 0.87 0.65 1.8E-09 Exponential 

Medium 3.4E-05 2.6E-04 29 0.87 0.54 6.0E-10 Exponential 

CG Dry 4.0E-05 1.5E-04 110 0.77 0.53 6.0E-09 Exponential 

Wet 3.0E-05 4.0E-04 100 0.93 0.93 5.8E-09 Exponential 

Medium 3.4E-04 5.7E-04 74 0.50 0.64 2.3E-08 Exponential 

HG Dry - - - - - - - 



Cs: structural variance; R 2 : determination coefficient; RSS: reduced sum of squares; -: No 

data available. 

Temporal variations in soil properties during the vegetation period showed a 

clear rainfall-dependent change for SWC and WDPT while this was not the case 

for the SS (Tables 1-6). This was in agreement with other studies where a 

pronounced seasonal variability was for SWC and WDPT (Buczko et al., 2005; 

Buttafuocoa et al. 2005). The results implied that hydraulic parameters are more 

sensitive than resilient mechanical parameters with respect to actual soil water 

22


content. 

Table 2.5. Isotropic variogram parameters for shear strength on three sampling soil 

water statuses in 2004. 

Treatment 

Sampling 

soil 

water 

statuses 

Nugget 

Co 

Sill 

Co+Cs 

Range 

A0 (m) 

Proportion 

Cs/(Co+Cs) 


UG 79 Dry 6.84 57.60 27 0.88 0.50 78.57 Exponential 

Wet 5.94 35.10 31 0.83 0.36 78.57 Exponential 

Medium 1.98 14.40 40 0.86 0.73 7.29 Exponential 

UG 99 Dry 5.58 47.70 38 0.88 0.48 315.90 Exponential 



WG Dry 8.73 63.00 87 0.86 0.26 178.20 Exponential 



CG Dry 36.00 72.90 400 0.50 0.93 81.00 Spherical 

Wet 26.10 52.20 270 0.50 0.80 97.20 Spherical 


HG Dry - - - - - - - 


Medium 1.53 31.50 26 0.95 0.68 25.11 Spherical 

Cs: structural variance; R 2 : determination coefficient; RSS: reduced sum of squares; -: 

No data available. 

3.2 Spatial distribution of soil properties in UG 79 

The spatial distribution of various soil parameters and their characteristics 

are exemplary shown for the ungrazed plot UG 79 (fenced for 25 years, 

considered to represent the typical ecosystem conditions without grazing). Table 

7 shows the descriptive statistics for all measured soil properties. The mean 

contents of sand, silt and clay were 48.5, 35.3 and 16.2%, respectively. A small 

standard deviation (SD) of the normally distributed data indicated that the 

sampling area could be considered as homogeneous in term of soil texture. Soil 

organic carbon (SOC) concentration was also normally distributed with a mean 

of 31.2 g kg -1 , but had a higher coefficient of variance (CV) compared to soil 

23

texture. The mean value of bulk density (BD) was relatively low with a mean of 

0.94 g cm -3 due to a high SOC concentration. After log-natural transformation of 

hydraulic conductivity (K) and water drop penetration time (WDPT), their mean 

values were 3.15 cm d -1 and 2.42 s, respectively, both showing a high SD and 

CV. 

Table 2.6. Isotropic variogram parameters for Ln(WDPT) on three sampling soil water 


Treatment 

Sampling 

soil 

water 

statuses 

Nugget 

Co 

Sill 

Co+Cs 

Range 

A0 (m) 

Proportion 

Cs/(Co+Cs) 


UG 79 Dry 0.21 0.54 250 0.62 0.66 2.2E-02 Exponential 

Wet 0.00 0.06 18 0.95 0.21 9.0E-04 Spherical 

Medium 0.15 1.00 17 0.85 0.51 1.8E-02 Exponential 

UG 99 Dry 0.06 0.50 35 0.87 0.64 6.7E-03 Exponential 



WG Dry 0.04 0.24 100 0.85 0.80 2.0E-04 Exponential 

Wet 0.01 0.11 92 0.99 0.56 1.2E-04 Exponential 

Medium 0.00 0.65 60 1.00 0.25 3.0E-03 Spherical 

CG Dry 0.18 0.52 250 0.66 0.94 2.1E-03 Spherical 



HG Dry - - - - - - - 

Wet 0.02 0.20 36 0.88 0.58 5.5E-04 Exponential 

Medium 0.14 0.53 310 0.74 0.83 9.0E-04 Spherical 

Cs: structural variance; R 2 : determination coefficient; RSS: reduced sum of squares; -: 

No data available. 

The isotropic semivariograms for the eight measured soil parameters in UG 

79 present a moderate to strong spatial dependence (Table 8). The ranges of 

spatial dependence were the highest for WDPT and the lowest for SS. The 

highest variability was obtained for SOC concentration and the lowest for soil 

texture. The spherical and exponential models fit the experimental data well 

according to R 2 values (0.67-0.96; Fig. 2), which was in agreement with many 

24


Table 2.7. Descriptive statistics of different parameters in UG 79 in 2005. 

Parameters Min. Max. Mean SD CV Skewness Kurtosis 

Ln(K) 2.27 4.36 3.15 0.49 15.56 0.30 -0.43 

Ln(WDPT) 0.69 5.89 2.42 1.25 51.65 0.83 0.16 

SS 36.34 64.87 52.55 6.73 12.81 -0.42 -0.33 

SWC 0.02 0.08 0.04 0.01 25.00 1.33 0.94 

Sand 41.00 56.70 48.50 3.42 7.05 -0.14 -0.7 

Silt 28.80 42.10 35.30 2.61 7.39 0.24 0.18 

SOC 18.40 45.10 31.20 5.70 18.27 0.27 -0.15 

BD 0.66 1.15 0.94 0.10 10.64 -0.40 -0.02 

K: hydraulic conductivity (cm d -1 ); WDPT: water drop penetration time (s); SS: shear 

strength (kPa); SWC: soil water content (cm 3 cm -3 ); Sand: sand content (%); Silt: silt content 

(%); SOC: soil organic carbon (g kg -1 ); BD: bulk density (g cm -3 ). 

Table 2.8. Isotropic variogram parameters for different properties in UG 79 in 2005. 

Parameters 

Nugget 

Co 

Sill 

Co+Cs 

Range 

A0 (m) 

Proportion 

Cs/(Co+Cs) 


Ln(K) 3.50E-02 2.70E-01 91 0.87 0.82 5.20E-03 Exponential 

Ln(WDPT) 1.07E+00 2.54E+00 311 0.58 0.87 9.47E-02 Spherical 

SS 6.03E+00 4.64E+01 43 0.87 0.78 6.67E+01 Exponential 

SWC 9.00E-05 2.50E-04 142 0.62 0.96 7.20E-10 Spherical 

Sand 1.00E-02 1.43E+01 94 1.00 0.96 1.90E+01 Spherical 

Silt 1.00E-02 7.99E+00 86 1.00 0.95 6.15E+00 Spherical 

SOC 1.89E+01 3.89E+01 191 0.52 0.94 1.01E+01 Exponential 

BD 1.40E-03 1.00E-02 57 0.87 0.67 3.90E-06 Exponential 

K: hydraulic conductivity (cm d -1 ); WDPT: water drop penetration time (s); SS: shear 

strength (kPa); SWC: soil water content (cm 3 cm -3 ); Sand: sand content (%); Silt: silt content 

(%); SOC: soil organic carbon (g kg -1 ); BD: bulk density (g cm -3 ). 

former studies (e. g. Qiu et al., 2001; Ferrero et al., 2005; Iqbal et al., 2005). 

Each variogram was the basis for the mapping by a kriged interpolation scheme 

of corresponding parameters. The ordinary kriged contour maps revealed 

moderate positional similarity of different soil properties, with complex positional 

effects in the plot’s interior (Fig. 3). The left part (close to fence, implied the 

formation of islands of fertility around fence) showed lower values of K and BD, 

but higher values of SWC, SOC and WDPT than that in the right part. Sand 

content was high in the central part extending from the upper-right to nether-left 

direction, but low at the two edges while a right opposite trend was observed for 

25

Semivariance 

26 

Semivariance 

Semivariance 

2.5x10 -4 

2.5x10 -4 

2.0x10 -4 

2.0x10 -4 

1.5x10 -4 

1.5x10 -4 

0 20 40 60 80 100 120 140 

2.5 

Separation Distance(m) 

1.0x10 -4 

1.0x10 -4 

Semivariance 

2.0 

1.5 

1.0 

0.5 

a 

c 

0.0 

0 20 40 60 80 100 120 140 

40 

35 

30 

25 

20 

e 


15 

0 20 40 60 80 100 120 140 

16 

12 

8 

4 

g 


0 

0 20 40 60 80 100 120 140 


Semivariance 

Semivariance 

0.30 

0.25 

0.20 

0.15 

0.10 

0.05 

b 

0.00 

0 20 40 60 80 100 120 140 

Semivariance 

21 

18 

15 

12 

9 

6 

3 

0.015 

0.012 

0.009 

0.006 

0.003 

Semivariance 

d 


0 

0 20 40 60 80 100 120 140 

10 

8 

6 

4 

2 

f 


0 20 40 60 80 100 120 140 

h 


0 

0 20 40 60 80 100 120 140 


Fig. 2.2. Semivariograms of (a) SWC, (b) Ln(K), (c) Ln(WDPT), (d) SS, (e) SOC, (f) BD, (g) 

sand content (%), and (h) silt content (%) in UG 79. K: hydraulic conductivity (cm d -1 ); WDPT: 

water drop penetration time (s); SS: shear strength (kPa); SWC: soil water content (cm 3 cm -3 ); 

SOC: soil organic carbon (g kg -1 ); BD: bulk density (g cm -3 ).


Fig. 2.3. Ordinary kriged maps of (a) SWC, (b) K, (c) WDPT, (d) SS, (e) SOC, (f) BD, (g) sand 

content (%), and (h) silt content (%) in UG 79. K: hydraulic conductivity (cm d -1 ); WDPT: water 

drop penetration time (s); SS: shear strength (kPa); SWC: soil water content (cm 3 cm -3 ); SOC: 

soil organic carbon (g kg -1 ); BD: bulk density (g cm -3 ). 

27

silt content. SS showed a relatively patchy distribution with intermittent local 

moderate positional similarity of different soil properties, with complex positional 

effects in the plot’s interior (Fig. 3). The left part (close to fence, implied the 

formation of islands of fertility around fence) showed lower values of K and BD, 

but higher values of SWC, SOC and WDPT than that in the right part. Sand 

content was high in the central part extending from the upper-right to nether-left 

direction, but low at the two edges while a right opposite trend was observed for 

silt content. SS showed a relatively patchy distribution with intermittent local 

highs and lows which did not clearly match the distribution of the other parameter. 

The correlations between different parameters will be discussed in more detail in 

the following section. 

3.3 Correlation between the soil properties 

28 

Spatial variability of soil properties is somewhat inherent in nature because 

of variations in soil parent materials and microclimate. However, some of the 

variability may be caused by grazing and management practices. Both 

geologic/pedologic and land use factors interacted with each other on spatial 

and temporal scales, are additionally affected by plant covering and topography, 

which, in turn, are further modified locally by erosion and deposition processes. 

Therefore, non-stationary condition and spatial heterogeneity may occur and 

correlation analysis is necessary. Table 9 presents the correlations between all 

analyzed parameters for the topsoil in the five studied plots (site-specific 

regressions). Fig. 4 also shows the inter-effects between the treatments 

(management-specific regressions). Finally, multiple stepwise regressions and 

partial correlation analysis are conducted to show the correlation among 

parameters and to evaluate the correlation of each variable excluded the 

inter-effects. 

SWC related with K, BD and SOC. 

SWC is crucial to model the atmospheric and hydrological processes. 

However, SWC is a highly temporal-spatial variable as it is affected by weather, 

soil texture, soil porosity, vegetation and topography. This was confirmed by a 

significant correlation between SWC and bulk density (r = -0.41**), sand content


(r = -0.54**) and SOC (r = 0.73**) (Table 9). Unexpectedly, a negative correlation 

between K and SWC (r = -0.74**) was observed. Generally, the reduced SWC at 

grazed areas was attributed to the reduction of the infiltration as livestock 

trampling compacted the soil surface (Naeth et al., 1991). This suggested that 

the litter might play an important role in impeding water infiltration and 

evaporation since there was a significant relationship between SWC and SOC (r 

= 0.73**). 

Table 2.9. Correlation matrix for the analyzed variables in the five plots. 

Parameters Ln(K) Ln(WDPT) SS SWC Sand Silt SOC BD 

Ln(K) 1 

Ln(WDPT) -0.86** 1 

SS 0.58** -0.68** 1 

SWC -0.74** 0.52** -0.77** 1 

Sand 0.41* -0.36** 0.78** -0.54** 1 

Silt -0.32* 0.44** -0.80** 0.42** -0.95** 1 

SOC -0.83** 0.79** -0.32* 0.73** -0.38** 0.46** 1 

BD 0.51* -0.43** 0.48** -0.41** 0.28** -0.40** -0.94** 1 

P

the ungrazed plot (Fig. 4b). Certainly, this result on the one hand was related to 

the actual SWC at the measuring time, On the other hand to the measurement 

method of Mini-disk Infiltrometer, where large pores (≥ 1.5 mm pores in diameter 

at a suction ≥ 2 cm) did not take part in the transporting water. Many studies 

reported increased animal trampling decreased infiltration rates because of 

higher bulk densities, higher penetration resistances and low volume of 

macropores (Naeth et al., 1990; Daniel et al., 2002; Binkley et al., 2003). Our 

result argued with those results because of reduced water repellency with 

increasing grazing intensity, which significantly decreased infiltration rate in the 

study area. 

Ln(K) (cm d -1 ) 

SOC (g kg -1 ) 

4.5 

4.0 

3.5 

3.0 

2.5 

2.0 

UG 79 

r =-0.78 

HG 

r =-0.84 r =-0.87 a 

0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 

40 

30 

20 

10 

Ln(WDPT) (s) 

UG 79 HG 

r =0.62 r =0.60 r =0.81 

0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 

Ln(WDPT) (s) 

c 

BD (g cm -3 ) 

BD (g cm -3 ) 

1.6 

1.4 

1.2 

1.0 

0.8 

0.6 

1.6 

1.4 

1.2 

1.0 

0.8 

0.6 

UG 79 HG 

r =0.52 r =0.23 r =0.86 

2.0 2.5 3.0 3.5 4.0 4.5 

Ln(K) (cm d -1 ) 

UG 79 HG 

r =0.31 r =0.13 r =0.61 

b 

13.5 15.0 16.5 18.0 19.5 21.0 

SS (kPa) 

Fig. 2.4. Scattergram plots and regression analysis for selected soil properties in UG 79 and 

HG between Ln(WDPT) and Ln(K) (a); Ln(K) and BD (b); Ln(WDPT) and SOC (c); and SS and 

BD (d). K: hydraulic conductivity (cm d -1 ); WDPT: water drop penetration time (s); SS: shear 

strength (kPa); SOC: soil organic carbon (g kg -1 ); BD: bulk density (g cm -3 ). 

30 

d


WDPT related with SOC and SWC. 

The organic matter deriving from roots and foliage might additionally 

contribute to the hydrophobicity by releasing aliphatic compounds (Rumpel et al., 

1998). The significant positive relationship between WDPT and SOC (r = 0.79**) 

(Table 9 and Fig. 4c) suggested that the organic substance reduced the rate of 

wetting (Peng et al., 2003). Furthermore, the results also confirmed a positive 

relationship between WDPT and SWC (r = 0.52**), indicating the occurrence of 

water repellency was closely related with the water content which in itself might 

be influenced by SOC concentration (“critical SWC”, Dekker et al., 2001; Täumer 

et al., 2005). These results underlined the interaction of bio-hydrological 

processes that a high SOC concentration reduced water infiltration through 

increasing water repellency, but also limited water evaporation through the 

organic litter mulch by which more water was retained in the soil. 

SS related with Soil texture, SWC and BD, SOC. 

SS was positively related with sand content (r = 0.78**) and BD (r = 0.48**, 

Table 9 and Fig. 4d), while it was negatively related with silt content (r = -0.80**), 

SOC (r = -0.32*), and SWC (r = -0.77**), respectively. However, no clear 

correlation with SWC was found for the shear vane tests in each plot, which 

might be explained by the high variance of SS within each plot at nearly identical 

values of SWC. This indicated that the mechanical properties were more related 

with soil texture than with the hydraulic properties. The compaction effects of 

animal trampling (land use) were mainly determined by the soil types. Willat and 

Pullar (1984) postulate structural damages by sheep hoof pressures leading to 

increased bulk density and a decrease in total pore volume in Australian silty 

loam soils. We suggest that an increase in bulk density, reduced macropores, 

and in combination with high hoof pressures leads to compacted topsoil. 

Multiple regression analysis among the soil properties. 

By multiple stepwise regression analysis for selected eight soil parameters 

following pedotransfer functions (PTFs) showing very significant 

interrelationships were derived: 

31

32 

Ln(WDPT)=-1.45Ln(K)+0.047Silt+5.3 (r =-0.89**) 

Ln(K)=-3.07Ln(WDPT)-0.26SOC+4.68 (r = -0.88**) 

SS=-8.16silt+81.3 (r = -0.80**) 

SWC=0.04SOC+0.92BD-0.006Ln(K)-0.142 (r = 0.82**) (2) 

Combined with the correlation matrix in Table 9, multiple regression analysis 

improved the correlation coefficients between WDPT and K, and between SWC 

and SOC, which underlined the strong interaction between these soil parameters. 

Partial correlation analysis also showed a pronounced correlation between K 

and WDPT (r = 0.84**), between SS and silt (r = -0.58**), and between SWC and 

SOC (r = 0.40*). The derived equations (2) presented here and former 

correlation analysis neglected the regional geochemistry process (e.g. cross 

correlation between adjacent observations with respect to the lag effect; 

Goovaerts, 1998) of the analyzed parameters, therefore it should be handled 

carefully when unmeasured parameters were predicted from other measured 

parameters. In this study, slight differences in topography have negligible effects 

on analyzed soil properties (data not shown). For a regional scale, 

cross-variogram analysis is better which, however, was beyond the scope of this 

paper but will be investigated more closely in further studies. 

Many variables in hydrology are to be considered as spatiotemporal 

parameters. SWC, SS, WDPT and K measurements and their interactions were 

identified in this study as complex spatiotemporal functions. We showed that 

geostatistical techniques, together with correlation analysis, were useful tools for 

(i) determining spatial variability characteristics of physical soil properties and (ii) 

recognizing significant interactions between soil properties. Although a large 

amount of data is required for this kind of analysis we deem it necessary 

because it is a prerequisite when matter fluxes on a regional scale is 

investigated. The spatial variation of soil physical properties, ideally expressed 

as maps, is valuable information not only for deriving a conceptual 

understanding of landscape fluxes (e. g. water) but also the basis for spatial 

discretization and parameter estimation in the modeled domain. Furthermore, 

the analysis of correlation length is very useful to determine appropriate model


element sizes, and to extrapolate and upscale with the aid of physically-based 

hydrological models from plot (e.g. Hydrus-1D) to catena (e.g. Hydrus-2D) and 

finally to a regional scale (e.g. SWAT) (Western et al. 1999). Further studies will 

therefore have to concentrate on questions like what size model elements have 

in different spatial scales (also scale effect) and what is the correlation structure 

between the variables using multivariate geostatistics. 

With respect of land use effects on site properties we summarize our results 

as follows. Heavy grazing resulted in a high resistance of soil to further 

deformation caused by compaction owing to intense trampling. Animal trampling 

led to a change in soil structure which is suggested to have a detrimental impact 

on soil functions (Krümmelbein et al., 2006). We showed that grazing increased 

SS and BD while it decreased SWC, SOC and WDPT. Grazing is therefore 

considered on the one hand reduce soil water storage through high evaporation 

because of the reduced soil organic carbon and intensive water redistribution 

(drainage or surface runoff) because of high infiltration rate by low water 

repellency, and low rainfall interception by low vegetation or litter coverage. On 

the other hand, deterioration of soil structure has negative long-term effects on 

soil stability and functions, which with increasing degree of compaction and soil 

homogenization by trampling, will possibly increase the risk of soil degradation 

and erosion. 

4. Conclusion 

In summary, it is important for the small-scale distribution of soil properties 

across the field to interpret the effects of various grazing intensities on soil 

functions. Descriptive statistics and geostatistical methods revealed spatially 

related variability for several soil properties on the plot scale with moderate to 

strong spatial structures. The highest variability was observed for SOC 

concentration while soil texture showed the lowest spatial variation. The ranges 

of spatial dependence were the highest for WDPT and the lowest for SS. 

Regression analysis showed significant correlations among SWC, K, WDPT, 

SOC and BD; as well as between SS and silt content. We concluded that heavy 

grazing resulted in a more homogenous spatial distribution of soil properties by 

33

soil compaction and soil homogenisation accompanied by a reduced input of 

organic matter and a reduced soil water storage capacity. This is assumed to 

increase soil vulnerability to further intensive grazing. Particularly, a further 

diminishing amount of plant available water is regarded as one of the most 

limiting factors for a sustainable development of steppe grassland ecosystems. 

In order to maintain a functioning of soil water and nutrient household, future 

land use in Inner Mongolia therefore needs to focus on protecting and restoring 

degraded grassland ecosystem from intensive grazing and animal trampling. 

Acknowledgements 

34 

The grants for this project were provided by the German Research Council 

(DFG) in the framework of the Interdisciplinary Research Project MAGIM (Matter 

fluxes in grasslands of Inner Mongolia as influenced by stocking rate), 

sub-project P8 (project code Ho 911/35). Dr. Angelika Kölbl is acknowledged for 

the manuscript revision. 

References 

Binkley, D., Singer, F., Kaye, M., Rochelle, R., 2003. Influence of elk grazing on 

soil properties in Rocky Mountain National Park. Forest Ecol. Manag. 185(3), 

239-247. 

Buczko, U., Bens, O., Hüttl, R.F., 2005. Variability of soil water repellency in 

sandy forest soils with different stand structure under Scots pine (Pinus 

sylvestris) and beech (Fagus sylvatica). Geoderma 126, 317-336. 

Buttafuocoa, G., Castrignano, A., Busonic, E., Dimased, A.C., 2005. Studying the 

spatial structure evolution of soil water content using multivariate 

geostatistics. J. Hydrol. 311, 202-218. 

Cambardella, C.A., Moorman, T.B., Parkin, T.B., Karlen, D.L., Turco, R.F., 

Konopka, A.E., 1994. Field scale variability of soil properties in Central Iowa 

soils. Soil Sci. Soc. Am. J. 58, 1501-1511. 

Cassel, D.K., Nelson,L.A., 1985. Spatial and temporal variability of soil physical 

properties of norfolk loamy sand as affected by tillage. Soil Till. Res. 5(1), 

5-17. 

Daniel, J.A., Potter, K., Altom, W., Aljoe, H., Stevens, R., 2002. Long-term


grazing density impacts on soil compaction. Transactions of the Asae 45(6), 

1911-1915. 

DeBano, L.F., 1981. Water repellent soils: a state-of-the-art. Gen. Tech. Rep. 

PSW-46. Pacific outhwest Forest and Range Experiment Station, Berkeley, 

California, pp. 21. 

Dekker, L.W., Doerr, S.H., Oostindie, K., Ziogas, A.K., Ritsema, C.J., 2001. 

Water repellency and critical soil water content in a dune sand. Soil Sci. Soc. 

Am. J. 65, 1667-1674. 

Dekker, L.W., Ritsema, C.J., 1994. How water moves in a water repellent sandy 

soil: 1. Potential and actual water repellency. Water Resour. Res. 30, 

2507-2517. 

Dekker, L.W., Ritsema, C.J., 2000. Wetting patterns and moisture variability in 

water repellent Dutch soils. J. Hydrol. 231/232, 148-164. 

Famiglietti, J.S., Rudnicki, J.W., Rodell, M., 1998. Variability in surface moisture 

content along a hillslope transect: Rattlesnake Hill, Texas. J. Hydrol. 210, 

259-281. 

Ferrero, A., Usowicz, B., Lipiec, J., 2005. Effects of tractor traffic on spatial 

variability of soil strength and water content in grass covered and cultivated 

sloping vineyard. Soil Till. Res. 84, 127-138. 

Fullen, M.A., Zheng, Y., Brandsma, R.T., 1996. Comparison of soil and sediment 

properties of a loamy sand soil. Soil Technology 10, 35-45. 

Gerke, H.H., Hangen, E., Schaaf, W., Hüttl, R.F., 2001. Spatial variability of 

potential water repellency in a lignitic mine soil afforested with Pinus nigra. 

Geoderma 102, 255-274. 

Goovaerts, P., Sonnet, P., 1993. Study of spatial and temporal variations of 

hydrogeochemical variables using factorial kriging analysis. In: Soares, A. 

(Ed.), Geostatistics Troia ’92, vol. 2. Kluwer Academic Publishers, Dordrecht, 

pp. 745-756. 

Goovaerts, P., 1998. Geostatistical tools for characterizing the spatial variability 

of microbiological and physico-chemical soil properties. Biol. Fertil. Soils 27, 

315-334. 

35

Greenwood, K.L., Macleod, D.A., Hutchinson, K.J., 1997. Long-term stocking 

36 

rate effects on soil physical properties. Aust. J. Exp. Agr. 37, 413-419. 

Greiffenhagen, A., Wessolek, G., Facklam, M., Renger, M., Stoffregen, H., 2006. 

Hydraulic functions and water repellency of forest floor horizons on sandy 

soils. Geoderma 132, 182-195. 

Holloway, P.J., 1994. Plant cuticles: physiochemical characteristics and 

biosynthesis. In: Percy, K.E. (Ed.), Air Pollution and the Leaf Cuticle. 

Springer, Berlin, pp. 1-13. 

Horn, R., Fleige, H., 2003. A method for assessing the impact of load on 

mechanical stability and on physical properties of soils. Soil Till. Res. 73, 

89-99. 

Iqbal, J., Thomasson, J.A., Jenkins, J.N., Owens, P.R., Whisler, F.D., 2005. 

Spatial variability analysis of soil physical properties of Alluvial soils. Soil Sci. 

Soc. Am. J. 69, 1338-1350. 

Jongman, R.H.G., ter Braak, C.J.G., van Tongeren, O.F.R., 1987. Data analysis 

in community and landscape ecology. Wageningen, Pudoc, 299 pp. 

Júnior, V.V., Carvalh, M.P., Dafonte, J., Freddi, O.S., Vidal Vázquez, E., 

Ingaramo, O.E., 2006. Spatial variability of soil water content and 

mechanical resistance of Brazilian ferralsol. Soil Till. Res. 85, 166-177. 

Kelly, R.H., Burke, I.C., Lauenroth, W.K., 1996. Soil organic matter and nutrient 

availability responses to reduced plant inputs in shortgrass steppe. Ecology 

77(8), 2516-2527. 

Krümmelbein, J., Wang, Z., Zhao, Y., Peth, S., Horn, R., 2006. Influence of 


of grassland soils in Inner Mongolia, P. R. China. Advances in Geoecology 

38, 93-101. 

Lavee, H., Sarah, P., Imeson, A.C., 1996. Aggregate stability dynamics as 

affected by soil temperature and moisture regimes. Geografiska Annaler 

Series a-Physical Geography 78A (1), 73-82. 

Li, S.G., Harazono, Y., Oikawa, T., Zhao, H.L., He, Z.Y., Chang, X.L., 2000. 

Grassland desertification by grazing and the resulting micrometeorological


changes in Inner Mongolia. Agric. For. Meteorol. 102, 125-137. 

Lophaven, S., Carstensen, J., Rootzéna, H., 2006. Stochastic modelling of 

dissolved inorganic nitrogen in space and time. Ecol. Modell. 193, 467-478. 

Ludovisi, A., Minozzo, M., Pandolfi, P., Taticchi, M.I., 2005. Modelling the 

horizontal spatial structure of planktonic community in Lake Trasimeno 

(Umbria, Italy) using multivariate geostatistical methods. Ecol. Modell. 181, 

247-262. 

Naeth, M.A., Chanasky, D.S., Rothwell, R.L., Bailey, A.W., 1991. Grazing 

impacts on soil water in mixed prairie and fescue grassland ecosystems of 

Alberta. Can. J. Soil Sci. 71, 313-325. 

Naeth, M.A., Rothwell, R.L., Chanasky, D.S., Bailey, A.W., 1990. Grazing 

impacts on infiltration in mixed prairie and fescue grassland ecosystems of 

Alberta. Can. J. Soil Sci. 170, 593-605. 

Peng, X., Zhang, B., Zhao, Q., Horn, R., Hallett, P.D., 2003. Influence of types of 

restorative vegetation on the wetting properties of aggregates in a severely 

degraded clayey Ultisol in subtropical China. Geoderma 115, 313-324. 

Peth, S., Horn, R., 2006. Consequences of grazing on soil physical and 

mechanical properties in forest and tundra environments. In: B.C. Forbes, M. 

Bölter, L. Müller-Wille, J. Hukkinen, F. Müller, N. Gunslay, Y. Konstatinov 

(Eds.). Ecological Studies, Vol. 184, pp. 217-243. 

Peth, S., Horn, R., Bölter, M., 2003. Impacts of grazing on soil structure water 

infiltration and soil heat flow as a result of reindeer herding in Northern 

Scandinavia. In: Proceedings of 16th Conference of ISTRO, 13-20 July 

2003, Brisbane, Australia, pp. 911-917. 

Piccolo, A., Mbagwu, J.S.C., 1999. Role of hydrophobic components of soil 

organic matter in soil aggregate stability. Soil Sci. Soc. Am. J. 63, 

1801-1810. 

Pietola, L., Horn, R., Yli-Halla, M., 2005. Effects of trampling by cattle on the 

hydraulic and mechanical properties of soil. Soil Till. Res. 82, 99-108. 

Proffitt, A.P.B., Bendotti, S., McGarry, D., 1995. A comparison between 

continuous and controlled grazing on a red duplex soil. I. Effects on soil 

37

38 

physical characteristics. Soil Till. Res. 35, 199-210. 

Qiu, Y., Fu, B.J., Wang, J., Chen, L.D., 2001. Spatial variability of soil moisture 

content and its relation to environmental indices in a semi-arid gully 

catchment of the Loess Plateau, China. J. Arid Environ. 49, 723-750. 

Richard, G., Cousin, I., Sillon, J.F., Bruand, A., Guérif, J., 2001. Effect of 

compaction on the porosity of a silty soil: influence on unsaturated hydraulic 

properties. Eur. J. Soil Sci. 52, 49-58. 

Ritsema, C.J., Dekker, L.W., 1996. Influence of sampling strategy on detecting 

preferential flow paths in water-repellent sand. J. Hydrol. 177, 33-45. 

Rumpel, C., Knicker, H., Kögel-Knabner, I., Skjemstad, J., Hüttl, R.F., 1998. 

Types and chemical composition of organic matter in reforested lignite-rich 

mine soils. Geoderma 86, 123-142. Samper, F.J., Carrera, J., 1990, 

Geoestadística-Aplicaciones a la hidrogeología subterránea: CIMNE. 

Barcelona, Spain, 484 pp. 

Shouse, P.J., Russell, W.B., Burden, D.S., Swlim, H.M., Sisson, J.B., van 

Genuchten, M.Th., 1995. Spatial variability of soil water retention functions 

in a silt loam soil. Soil Sci. 159(1), 1-12. 

Täumer, K., Stoffregen, H., Wessolek, G., 2005. Determination of repellency 

distribution using soil organic matter and water content. Geoderma 125, 

107-115. 

Warren, S.D., Nevill, M.B., Blackburn, W.H., Garza, N.E., 1986. Soil response to 

trampling under intensive rotation grazing. Soil Sci. Soc. Am. J. 50, 

1336-1340. 

Western, A.W., Blöschl, G., 1999. On the spatial scaling of soil moisture. J. 

Hydrol. 217, 203-224. 

Western, A.W., Blöschl, G., Grayson, R.B., 1998. Geostatistical characterisation 

of soil moisture patterns in the Tarrawarra Catchment. J. Hydrol. 205, 20-37. 

Willatt, S.T., Pullar, D.M., 1984. Changes in soil physical properties under grazed 

pastures. Aust. J. Soil Res. 22(3), 343-348. 

Zhang, R., 1997. Determination of soil sorptivity and hydraulic conductivity from 

the disk infiltrometer. Soil Sci. Soc. Am. J. 61, 1024-1030.

Chapter 3 Spatio-temporal variability of soil moisture in grazed steppe areas investigated by geostatistics 

3. Spatio-temporal variability of soil moisture in grazed 

steppe areas investigated by multivariate geostatistics 

Y. Zhao, S. Peth, R. Horn, P.D. Hallett, X.Y. Wang, M. Giese, Y.Z. Gao 

Revision for Journal of Hydrology. 

Abstract 

Land use has a significant impact on spatio-temporal soil moisture patterns, 

particularly in sensitive and poorly managed regions such as the grassland of 

Inner Mongolia. This study identified the factors that control soil moisture 

patterns under different grazing intensities in the semi-arid Xilin River Catchment, 

Inner Mongolia, China. Five different levels of sheep grazing intensity were 

sampled for various soil and vegetation properties at high spatial and temporal 

resolution during 2004-2006: (1) ungrazed since 1979 (UG 79); (2) ungrazed 

since 1999 (UG 99); (3) winter grazed only, with 0.5 sheep units (ewe and lamb) 

ha -1 yr -1 (WG); (4) continuously grazed, with 1.2 sheep units ha -1 yr -1 (CG); and 

(5) heavily grazed, with 2.0 sheep units ha -1 yr -1 (HG). The data were analyzed 

using correlation and geostatistical analysis. Results showed that spatial 

variance and, to a lesser extent, the correlation length were related to mean soil 

moisture content. Under different grazing intensities, (a) soil moisture patterns 

had weak to moderate spatial structures with typical correlation lengths ranging 

from 20 to 178 m; (b) soil and plant properties were the main factors controlling 

the distribution of soil moisture; and (c) factorial kriging analysis further revealed 

scale-dependent correlation structure of the controlling parameters. Specifically, 

the soil and plant properties strongly controlled the variation of soil moisture for 

UG 99 at short-scale (90 m), for CG at long-scale (165 m) and for HG at 

averaged random, short-scale and long-scale components, however, weakly 

controlled the variation of soil moisture for UG 79 at the micro-scale (random;

Keywords: Soil moisture; Grazing intensity; Inner Mongolia grassland; 

Controlling factor; Scale-dependency 


Over-grazing has depleted vegetation coverage over large areas of 

grassland in Inner Mongolia, leaving the exposed soil vulnerable to wind and 

water erosion (Li et al., 2000; Gao et al., 2002). Recent research by Zhao et al. 

(2007) found that heavy grazing in this region also degrades soil hydraulic and 

mechanical properties. Based on studies in other regions, the degradation of 

these soil properties could influence the spatio-temporal variation in soil 

moisture (Williams et al., 2003; Hébrard et al., 2006). As a result, hydrological 

processes governing plant water availability could greatly influence plant 

stresses and hence heterogeneity in establishment caused by grazing (Evenari 

et al., 1971). 

The impact of grazing intensity, or other differences in land management, on 

the spatio-temporal variability of soil moisture has been shown in previous 

research to depend on numerous plant and soil factors (Hawley et al., 1983; Qiu 

et al., 2001; Williams et al., 2003). However, these studies have reported 

conflicting results, so there is a need for further research to examine the causes 

of spatio-temporal variability in soil moisture under different land uses. As 

grazing intensity, particularly on the grasslands of Inner Mongolia, has such a 

large impact on hydrology, this area offers an ideal study site. Moreover, land 

degradation in Inner Mongolia is recognised globally as a major problem 

because of the local threat to agriculture and the wider impact on dust storms. 

Understanding the factors controlling the spatio-temporal variability in soil 

moisture will be essential in assessing the benefits of different land restoration 

practices. 

The patterns of soil moisture that develop are also fundamental to 

physically-based hydrological models (Blöschl and Sivapalan, 1995; Giacomelli 

et al., 1995; Western et al., 1999; Herbst and Diekkrüger, 2003). Several authors 

have shown the characteristics of soil moisture variability, using spatial 

40


measurement scales ranging from meters to kilometres (Hills and Reynolds, 

1969; Famiglietti et al., 1998; Qiu et al., 2001; Western et al., 2004), and 

temporal scales ranging from days to years (Entin et al., 2000; Cantόn et al., 

2004; Parent et al., 2006). However, the results appear to be influenced by the 

specific study area examined and the sampling time. 

The spatio-temporal variability of soil moisture has been shown to be 

associated with many controlling factors such as soil, vegetation, topography 

and weather (Francis et al., 1986; Crave and Gascuel-Odoux, 1997; Bádossy et 

al., 1998; Famiglietti et al., 1998). However, disentangling the influence of each 

controlling factor is marred by complexity, as some properties are 

interdependent and many have site specific spatio-temporal variations 

(Famiglietti et al., 1998; Western et al., 1999). In particular, the correlation length 

of soil moisture may be related to different processes that operate at particular 

scales or moisture conditions. Soil moisture patterns at small scale will be 

regulated by local factors (i.e. soil, vegetation, and surface topographical factors), 

whereas at larger scale processes such as subsurface flow and weather will 

have a greater impact. The weight of factors also depends on the antecedent 

moisture conditions at the time of sampling. In wet conditions, subsurface lateral 

flow may be important, whereas dry conditions will be influenced more by 

surface vertical flow at the onset of precipitation (Grayson et al., 1997; 

Gómez-Plaza et al., 2001). However, there is still great uncertainty about the 

scale-dependent relationship between soil moisture and its controlling factors 

(Zeleke and Si, 2006). 

Geostastitical techniques allow for complicated spatio-temporal soil 

moisture patterns to be characterized (Western et al., 1998; Entin et al., 2000; 

Webster and Oliver, 2001; Anctil et al., 2002). Using multivariate geostatistics, 

the interrelation of soil properties can be evaluated, e.g. soil moisture, in the 

presence of multivariate spatial data of regionalized variables (Castrignano et al., 

2000; Buttafuoco et al., 2005; Casa and Castrignanò, 2007). However, as 

Famiglietti et al. (1998) noted, research using geostatistics to link soil moisture 

variability to its controlling factors, may be limited by low sampling frequency in 

41

oth space and time. Land use may further complicate such analysis as 

recognized by Hébrard et al. (2006). 

This study addresses the current weakness in understanding by quantifying, 

at high resolution, the spatio-temporal variability of soil moisture and various 

controlling factors on plots receiving different grazing intensities in the Inner 

Mongolia grassland. A combined grid and nested sampling was conducted on 

five plots, ranging from heavy grazing to restored ungrazed grassland, from 

2004 to 2006. The objectives of this study were (i) to quantify the spatio-temporal 

variability of topsoil moisture as affected by grazing intensity, (ii) to ascertain the 

main factors controlling soil moisture patterns, and (iii) to explore the multiple 

spatial scale of controlling factors by multivariate and geostatistical approaches. 

Understanding the spatio-temporal variability of soil moisture and its controlling 

factors under different grazing intensities provides valuable information for 

hydrological modeling, plant establishment and the environmental impacts of 

different land management practices in the highly degraded and sensitive Inner 

Mongolia grassland. Moreover, the research has generic application to the 

overall understanding of how land use influences soil moisture patterns over 

space and time. 

2. Materials and methods 

2.1. Field descriptions and measurements 

Fieldwork was conducted at the long-term experimental sites in the Xilin 

River Catchment, which are managed by the Inner Mongolia Grassland 

Ecosystem Research Station (IMGERS; 43 o 37′50′′N, 116 o 42′18′′E). A detailed 

description of the study area can be found in Bai et al. (2004). The vegetation 

was dominated by perennial grasses, e.g. Leymus chinensis and Stipa grandis. 

The harsh local climate limited the growing season from May to September. For 

the last two decades, the mean annual air temperature was 0.7°C, and the mean 

annual precipitation was 343 mm, of which more than 85% fell during the 

growing season. The soils are Calcic Chernozems according to IUSS Working 

Group (2006). 

Five different levels of sheep grazing intensity were sampled (Fig. 1): (1) 24 

42


ha area ungrazed since 1979 (UG 79); (2) 35 ha area ungrazed since 1999 (UG 

99); (3) 40 ha area winter grazed only, with 0.5 sheep units (ewe and lamb) ha -1 

yr -1 (WG); (4) 250 ha area continuously grazed, with 1.2 sheep units ha -1 yr -1 

(CG); and (5) 100 ha area heavily grazed, with 2.0 sheep units ha -1 yr -1 (HG). 

The plots were adjacent to each other, apart from HG, which was 2.5 km away 

but on similar soils and topography. Prior to 1979, all plots were grazed at low 

grazing intensity. All plots except UG 79 were grazed at moderate grazing 

intensity from 1979 to 1999 and fenced in 1999. 

Fig. 3.1. Location of the sampling plots delineated as: UG 79: ungrazed since 1979, UG 99: 

ungrazed since 1999, WG: winter grazed, CG: continuous grazed, and HG: heavily grazed. Each 

plot has an area of 105 m×135 m. Sampling points were spaced at 15 m with a subgrid spacing of 

5 m (the orthogonal boxes with dash line in WG and CG). Two larger geostatistical grids in WG 

and CG had an area of 300 m×550 m with a grid spacing of 50 m and a subgrid spacing of 10 m. 

Measurements in each plot were arranged on a rectangular sampling grid, 

consisting of 100 points with a grid spacing of 15 m and an enclosed subgrid 

spacing of 5 m (Fig. 1). Volumetric soil water content (0-6 cm) was measured 

both weekly and after rainfall events exceeding 3 mm, for 3 years during the 

growing season, starting in 2004. These measurements used theta-probes 

43

(Type ML2x, Delta-T Devices, Cambridge, UK) that were calibrated for the soil 

and inserted into the surface at each point. In total, measurements were made at 

52 different times. 

Soil water content was compared to other properties that were anticipated to 

influence water transport and storage. At least 3 times during the growing 

season the following measurements were made in situ. Shear strength was 

measured with a hand-held shear vane tester (Geonor H-60, Norway). Water 

repellency was estimated with the water drop penetration time (WDPT) test, by 

measuring the time taken for a 0.5 mm 3 drop of water to enter the soil. A longer 

time indicates greater water repellency. Hydraulic conductivity was measured 

using a Mini-disk Infiltrometer (Decagon devices, USA) at a suction value of 0.5 

cm. 

Once each year, in July or August, the vegetation coverage was analyzed 

using a non-destructive method based on Braun-Blanquet (1964). Aboveground 

biomass was sampled on 0.25 m x 0.25 m plots by cutting plant at 10 mm height, 

including the standing dead material. At the first sampling time in 2004, 0-4 cm 

depth samples were taken for soil organic carbon determined in duplicate by dry 

combustion on a Vario Max CNS elemental analyzer (Elementar 

Analysensysteme GmbH, Hanau), and soil texture with the pipette method. 

Intact soil cores, 56 mm diameter x 40 mm heigh were taken for bulk density and 

water retention. The cores were dried to pF 1.8 (field capacity) on a sand table 

and 4.2 (wilting point) on a pressure plate to describe water retention. The main 

mean characteristics (100 points) for each plot are described in Table 1. 

A terrain elevation model was derived from Real Time Kinematic dGPS 

(RTK-GPS) measurements with a lateral resolution of 2 m and vertical resolution 

of 0.1 m. Using this information, topographical variables such as slope, aspect, 

curvature, and upslope contribution area were computed using ArcGIS 

(Johnston, et al., 2001). A wetness index, which is a surrogate for lateral 

subsurface flow process, was also derived. This index was defined as ln(α/tanβ), 

where α is the upslope contribution area and β is the local slope angle 

(Gómez-Plaza, et al., 2001). 

44


2.2. Data analysis 

All measured variables were analyzed with descriptive statistics and 

geostatistics. Correlations were tested using Pearson’s correlation coefficient. 

Prior to calculating spatial variation, data were tested for normal distribution with 

a Kolmogorov-Smirnov test. By removing one or two outliers (m±3s method) and 

data transformations (Ln, Cos and Tan), a normal distribution was obtained for 

each variable. 

For each plot and sampling time, the mean soil water content of all 100 

points was used to delineate a soil moisture class. Three classes were defined 

as wet (>20%), medium (10-20%) and dry (

cross-semivariograms (factorial kriging analysis). The LMC is a set of auto- and 

cross-semivariogram models in which all its semivariograms are linear 

combinations of the same set of elementary structures. A LMC with k = 1, . . ., q 

structures may be written as: 

g (h) 

k 

1 1 k k 

q q 

γ h) 

= b g ( h) 

+ b g ( h) 

+ ⋅⋅ 

⋅ + b g ( h) 

(3) 

i , j ( i, 

j 

i, 

j 

i, 

j 

k 

b i, 

j 

where is the partial sill for the i,jth semivariogram for structure k, while 

represents the type of semivariogram model (i.e. spherical, exponential, 

1 

etc.) for structure k. The first structure g ( h) 

represents the nugget effect model. 

The type of semivariogram model used in Eq. 3 was based on the experimental 

semivariograms of the variables (standardized to unit variance and zero mean) 

and knowledge of the main geological and anthropogenic factors. Modified 

kriging programs in ArcGIS were used to fit the LMC, in which semivariograms 

were fitted with appropriate model functions using the maximum likelihood 

cross-validation method (Samper and Carrera, 1990). 

The scale-dependent correlations between water content and other 

properties was determined from the structural correlation coefficients, ρ k i.j 

k 

bi 

, j 

k k 

i , ib 

j , j 

k 

ρ i, 

j = 

b 

(4) 

which are calculated from the partial sill value, , of the 

cross-semivariogram model between i and j and the two partial sill values, 

k 

b j, 

j 

and , of semivariogram models for i and j, respectively (Casa and 

Castrignanò, 2007). Furthermore, constructed variation contributions of each 

component to the total variation were calculated based on Eq. 3. 

3. Results and discussion 

Grazing intensity influenced a range of soil properties listed in Table 1. The 

heavily grazed plot (HG) was the most dense, had the least carbon, and the 

greatest shear strength. In contrast, the plot protected from grazing for 25 yr (UG 

79) was the least dense, had the most carbon and the smallest shear strength. 

There appears to be consistent trends between grazing intensity and many of 

46 

k 

b i, 

j 

k 

b i, 

i


the properties measured, thus providing valuable field plots for further analysis of 

the data collected. 

3.1. Spatio-temporal variations of topsoil moisture 

Table 3.1. The main characteristics of the different grazing intensity plots sampled: UG 79: 

ungrazed since 1979, UG 99: ungrazed since 1999, WG: winter grazed, CG: continuous 

grazed, and HG: heavily grazed 

Parameters UG 79 UG 99 WG CG HG 

asl. 1258.8 1273.7 1273.1 1256.9 1211.1 

Sand 48.6 46.1 47.1 44.6 68.1 

Silt 35.2 37.4 35.2 36.6 20.9 

Clay 16.3 16.5 17.8 18.7 11.0 

SOC 31.0 25.5 25.0 22.2 17.0 

BD 0.9 1.1 1.1 1.2 1.3 

SS 29.3 32.2 35.1 36.6 43.6 

WDPT 16.4 12.1 6.0 5.2 1.7 

K 23.3 52.6 38.1 42.0 49.5 

Air capacity 25.6 12.5 20.1 16.4 7.0 

SWC (pF=1.8) 40.4 46.0 38.4 38.3 43.9 

SWC (pF=4.2) 11.5 11.0 10.3 10.8 5.3 

VC 76.4 68.0 69.3 54.5 69.3 

AGB 313.1 404.6 160.8 142.6 131.3 

asl.: above sea level (m); Sand: sand content (%); Clay: clay content (%); SOC: soil organic 

carbon (g kg -1 ); BD: bulk density (g cm -3 ); SS: shear strength (kPa); WDPT: water drop 

penetration time (s); K: hydraulic conductivity (cm d -1 ); AC: air capacity (%); SWC: soil water 

content (%); pF: log matric potential in hPa; VC: vegetation coverage (%); and AGB: 

Aboveground biomass (g m -2 ). 

Temporal changes of mean soil water content (MSWC), spatial variance and 

range for the different grazing intensity plots are shown in Fig. 2. There was 

strong seasonal fluctuation in MSWC, ranging from 2.0 to 31.9% for UG 79, 2.1 

to 36.4% for UG 99, 2.4 to 34.9% for WG, 2.2 to 33.8% for CG, and 1.1 to 28.0% 

for HG (Fig. 2a). The variance followed a similar trend to MSWC (Fig. 2b), as 

indicated by the significant correlation (P0.05). 

47

Range (10 2 m) 

Variance (% 2 ) 

Mean (%) 

2.4 

2.0 

1.6 

1.2 

0.8 

0.4 

0.0 

15 

12 

9 

6 

3 

0 

36 

30 

24 

18 

12 

6 

0 

c 

b 

a 

UG79 UG99 WG CG HG 

2004-8-12 2005-6-16 2005-7-23 2006-5-28 2006-8-4 2006-9-5 

2004-8-9 2005-6-11 2005-7-31 2005-9-19 2006-6-27 2006-8-16 

2004-8-12 2005-6-16 2005-7-23 2006-5-28 2006-8-4 2006-9-5 

2004-8-12 2005-6-16 2005-7-23 2006-5-28 2006-8-4 2006-9-5 

Time (days) 

Rainfall 

Fig. 3.2. Time series of (a) mean soil moisture content, (b) soil moisture variance and (c) soil 

moisture range for the different grazing intensities during the sampling period 2004-2006. 

Although several investigators have also noted that variance increased with 

increasing MSWC (Hills and Reynolds, 1969; Reynolds, 1970; Henninger et al., 

1976; Famiglietti et al., 1998; Parent et al., 2006), other studies have not 

observed this trend (Hawley et al., 1983; Charpentier et al., 1992). From Fig. 3, it 

appears that environmental disturbance, contributed by grazing intensity, could 

influence the relationship. The HG plot showed the greatest trend, probably due 

to the legacy of sheep trampling. In contrast to it, plots with longer recovery time 

since last grazing (UG 79) have less of a positive relationship between MSWC 

and variance. Our results indicated that variance of soil moisture in this semi-arid 

region was higher under wet conditions since effects of main factors controlling 

hydrological processes (e.g. soil heterogeneity) would be at a maximum under 

wet conditions. Whereas it was lower under dry conditions due to a large water 

deficit and a low evapotranspiration rate. 

Table 2 lists the descriptive statistics of soil moisture for the different plots 

following grouping into dry, medium and wet water statuses. Soil water content 

(SWC) was greatest in UG 99, followed by UG 79, WG and CG, and smallest in 

48 

52 

39 

26 

13 

0 

Rainfall (mm)


HG, suggesting that water regimes were changed by different grazing intensities. 

SWC was always greater in UG 99 than in UG 79, which can be explained by the 

larger leaf area of shrubs increasing rainfall interception and the thicker litter 

layer reducing water infiltration in UG 79. Compared to the two plots grazed at 

moderate intensity (WG and CG), SWC in UG 79 was greater after long dryness 

but smaller under wet and medium conditions. This could be due to the 

pronounced accumulation of organic matter observed, which would impede 

evaporation under dry water status (Zhao et al., 2007) and inhibit water 

infiltration under medium and wet water statuses, as indicated by the WDPT 

(Table 1). The smallest SWC was always observed in HG where scarce vegetal 

cover could have increased water evaporation. 

Table 3.2. Descriptive statistics of soil moisture for three moisture conditions measured during 

2004-2006 

Treatment Sampling soil water statuses Min. Max. Mean CV Skewness Kurtosis 

UG 79 20% (Wet) 21 28 24 5 0.50 -0. 05 

UG 99 20% (Wet) 25 30 27 4 -0.38 0. 33 

WG 20% (Wet) 23 27 25 3 -0.38 -0. 35 

CG 20% (Wet) 22 28 25 5 0.50 0. 15 

HG 20% (Wet) 17 26 21 9 -0.21 0. 31 

Min.: Minimum (%); Max.: Maximum (%); CV: Coefficient of variation (%). 

There was a weak to moderate spatial dependency of soil water content 

varying from 13 to 75% (Table 3), according to the classification of spatial 

49

Table 3.3. Isotropic variogram parameters of soil moisture for three moisture conditions 

measured during 2004-2006 

Treatment 

Sampling soil 

water statuses 

Nugget 

Co 

Cs: structural variance; R 2 : determination coefficient; RSS: reduced sum of squares. 

Sill 

Co+Cs 

Range 

A0 (m) 

Proportion 

Cs/(Co+Cs) R2 RSS 

UG 79 20% (Wet) 0.90 1.46 47 0.38 0.95 0.01 

UG 99 20% (Wet) 0.69 1.54 111 0.55 0.89 0.12 

WG 20% (Wet) 0.58 0.72 105 0.19 0.65 0.02 

CG 20% (Wet) 0.61 1.93 170 0.68 0.92 0.16 

HG 20% (Wet) 1.04 4.32 96 0.75 0.85 1.60 

dependency by Cambardella et al. (1994). In most cases, the variograms of soil 

moisture were expressed well by a spherical or exponential model, as indicated 

by the high coefficients of determination (R 2 ). The smallest sill was found for the 

WG plot. For most soil water statuses, CG and HG plots had the largest sills, 

although the sill in UG 99 becomes greater at medium water status. A similar 

trend was also observed for the proportion of spatial structure, indicating a 

stronger spatial dependency with increasing grazing intensity (Table 3). In the 

two ungrazed plots, UG 79 had greater spatial variability than UG 99 for medium 

and wet water statuses, potentially due to greater plant heterogeneity with a 

spotty distribution of shrubs (Caragana microphylla) after 25 yr protection from 

grazing. The range did not show clear trend with increasing grazing intensity. 

The smaller range in HG compared to CG and UG 99 could be due to more 

partial and continuous compaction. 

50


Variance (%) 2 

18 

16 

14 

12 

10 

8 

6 

4 

2 

0 

UG 79 UG 99 WG CG HG 

R=0.34 R=0.67 R=0.79 R=0.46 R=0.78 

0 5 10 15 20 25 30 

Mean (%) 

Fig. 3.3. Correlations of variance to mean soil moisture content for five plots. 

35 40 

3.2. Factors controlling the Spatio-temporal patterns of soil moisture 

A correlation matrix between SWC and potentially associated properties for 

the different soil water statuses (wet, medium and dry) is shown in Table 4. For 

most grazing intensities and water statuses, a significant relationship between 

SWC and associated properties (e.g. soil texture, soil organic carbon (SOC) and 

bulk density) was found. For UG 99, CG and HG, there was a significant 

negative correlation between either sand content or bulk density and SWC. 

These results are supported by other studies that also found SWC to be 

correlated to soil texture and bulk density (Henninger et al., 1976; Gomez-Plaza 

et al., 2001). A similar trend was not found for WG and UG 79, however, 

probably because of the weak spatial dependency of SWC (Table 3). 

51

52 

SLO: slope (degree); ASP: aspect; CUR: curvature; CAR: upslope contribution area (m 2 ); and WI: wetness index. 

WDPT: water drop penetration time (s); K: hydraulic conductivity (cm d -1 ); VC: vegetation coverage (%); AGB: above ground biomass (g m -2 ); RE: relative elevation (m); 

D: dry; M: medium; W: wet; Sand: sand content (%); Clay: clay content (%); SOC: soil organic carbon (g kg -1 ); BD: bulk density (g cm -3 ); SS: shear strength (kPa); 

Tan(CAR) -0.09 -0.08 0.03 -0.19 -0.03 0.10 -0.02 0.13 -0.07 0.01 -0.10 -0.10 0.01 0.08 0.02 

WI 0.05 -0.14 -0.20* -0.12 -0.08 -0.01 -0.06 0.11 -0.07 0.08 0.03 0.1 0.04 0.01 0.03 

Cos(ASP) 0.01 0.14 0.03 -0.18 -0.12 0.12 0.30** 0.13 0.00 0.37** 0.40** 0.30** -0.13 -0.07 -0.09 

CUR 0.09 0.01 0.00 0.13 0.18 -0.08 0.18 -0.01 0.05 -0.05 -0.02 -0.01 -0.01 -0.06 0.00 

Topo RE 0.02 -0.13 -0.20* 0.05 0.18 0.19 -0.26* -0.19 -0.05 -0.25* -0.23* -0.08 0.21* 0.10 0.15 

-graphy Tan(SLO) -0.01 0.21* 0.32** 0.06 0.12 0.20* 0.03 -0.01 0.06 -0.08 -0.09 -0.14 0.09 0.18 0.11 

Plant VC -0.01 -0.01 0.22* -0.07 0.05 0 -0.12 -0.04 -0.11 -0.25* -0.22* -0.34** 0.26** 0.22* 0.23* 

AGB 0.04 0.03 0.08 0.1 0.26* 0.33** 0.21* 0.14 0.04 -0.25* -0.26** -0.30** -0.22* -0.22* -0.33** 

Ln(K) -0.31** -0.12 -0.07 -0.08 -0.07 -0.12 -0.09 -0.17 -0.14 0.01 -0.02 -0.04 0.13 0.00 0.14 

SOC 0.27* 0.19 0.13 0.29** 0.23* 0.50** 0.02 0 0.07 0.41** 0.35** 0.39** 0.55** 0.52** 0.52** 

BD -0.24* -0.19 -0.12 -0.30** -0.23* -0.45** 0.08 0.16 -0.06 -0.32** -0.22* -0.24* -0.38** -0.39** -0.34** 

SS 0.42** 0.20* 0.1 0.15 0.11 0.27** 0.13 -0.08 -0.02 0.14 0 -0.05 -0.32** -0.19 -0.21* 

Ln(WDPT) 0.33** 0.18 0.03 0.21* 0.07 0.19 -0.06 0.03 -0.06 0.05 0.06 0.09 -0.35** -0.32** -0.39** 

Soil Sand -0.14 -0.08 -0.21* -0.28** -0.26** -0.50** 0.07 0.17 -0.05 -0.50** -0.44** -0.47** -0.73** -0.73** -0.83** 

Clay 0.16 0.09 0.19 0.25* 0.24* 0.55** -0.07 -0.18 0.05 0.55** 0.49** 0.55** 0.62** 0.65** 0.65** 

D M W D M W D M W D M W D M W 

Factors Parameters 

UG 79 UG 99 WG CG HG 

Table 3.4. Correlation matrix between soil moisture and water-related factors for three soil water statuses for different grazing intensities (P

Chapter 3 Factors controlling spatio-temporal variability of soil moisture using multivariate geostatistics 

Many studies envisioned soil moisture variability should be strongly 

controlled by variations in hydraulic conductivity (K), especially under wet 

conditions (Hawley et al., 1983; Chanzy and Bruckler, 1993; Hébrard et al., 

2006). Unexpectedly, we did not detect a correlation between K and SWC (Table 

4). K depends strongly on pore size and continuity, wetting properties and SWC. 

Considerable temporal variability in all of these properties, combined with 

different impacts on K depending on antecedent environmental conditions, may 

have confounded any clear trend being observed. The only significant 

relationship between K and SWC was found for UG 79 when dry. This trend 

could be caused by the influence of SOC on wettability when the soil was dry, as 

indicated by the strong relationship between K and WDPT (Zhao et al., 2007). 

Animal trampling in HG may have caused the negative correlation between 

shear strength and SWC. The recovery of soil physical structure, reinforcement 

by plant roots and cohesion by SOC in UG 79 may have reversed this trend as a 

positive correlation was found. 

Plant heterogeneity may also control soil moisture patterns as rooting depth, 

surface cover, and species composition affect evapotranspiration processes. 

SWC was significantly correlated with vegetation coverage (VG) and 

aboveground biomass (AGB) in the continuously (CG) and heavily grazed plots 

(HG), irrespective of moisture condition (Table 4). However, in the ungrazed 

plots, correlations were only observed under a wet water status in UG 79, and 

under medium and wet water statuses in UG 99. Reynolds (1970), Hawley et al. 

(1983) and Francis et al. (1986) found that vegetation coverage was a major 

factor influencing soil moisture variability. Hawley et al. (1983) further noted that 

differences were often greater in wet conditions than in dry conditions, but our 

study did not support this finding. In the HG plot, the negative correlation 

between AGB and SWC showed the influence of plant transpiration. However, 

VC was positively correlated to SWC in the HG plot, illustrating the deleterious 

impact of grazing intensity on both plant coverage and soil physical structure. 

With minor exceptions, neither surface factors, e.g. relative elevation, slope and 

aspect (local control) nor subsurface factors, e.g. curvature, upslope contributing 

53

area (CAR) and wetness index (WI) (unlocal control) exhibited a significant 

correlation with soil moisture for any soil water statuses (Table 4). The aspect, 

and therefore the influence of solar radiation on soil moisture, was significant 

only for the CG plot, where the topography was relatively flat. The relative 

elevation was important at the grazed plots under dry conditions, while the slope 

was a key factor at the ungrazed plots under wet conditions. Under wet 

conditions, unlocal subsurface flow may mostly influence SWC (Grayson et al., 

1997), but this was not found for any grazing intensity. Thus topography had only 

a minor effect on soil moisture distribution, which agrees with the findings of 

Hébrard et al. (2006). This might relate to the local topographical conditions, 

which is a smooth wide plain (relative height of 20-30 m with slopes


water status was shown to have limited influence on the controlling factors, as 

discussed previously, all SWC data were averaged for individual plots. 

Semivariance 

Semivariance 

2.5x10 -3 

2.5x10 -3 

2.0x10 -3 

2.0x10 -3 

1.5x10 -3 

1.5x10 -3 

1.0x10 -3 

1.0x10 -3 

7.0x10 -3 

6.0x10 -3 

5.0x10 -3 

4.0x10 -3 

3.0x10 -3 

2.0x10 -3 

1.0x10 -3 

Soil moisture 

0 20 40 60 80 100 120 140 

Soil moisture vs SOC 

0 20 40 60 80 100 120 140 

Distance(m) 

Semivariance 

0.0 0.0 

-8.0x10 -4 

-8.0x10 -4 

-1.6x10 -3 

-1.6x10 -3 

-2.4x10 -3 

-2.4x10 -3 

-3.2x10 -3 

-3.2x10 -3 

0.0 

-8.0x10 -4 

-1.6x10 -3 

-2.4x10 -3 

-3.2x10 -3 

Soil moisture vs sand content 

0 20 40 60 80 100 120 140 

0 20 40 60 80 100 120 140 

Distance(m) 

Soil moisture vs BD 

Fig. 3.4. Semiariograms and cross-semivariograms maps of soil variables and models for the 

linear Model of Coregionalization exemplified by UG 99. SOC: soil organic carbon (g kg -1 ), BD: 

bulk density (g cm -3 ). 

The coregionalization modeling for the soil variables was carried out for 

sand content, SOC and bulk density only, since they showed the strongest 

correlations with SWC compared to other soil variables (Table 4). The LMC was 

fitted using three spatial structures: nugget effect (random;

corresponding to the different scales (Table 5), we found that the total spatial 

variation of soil moisture was mostly dominated by variations of soil variables 

within a short-range scale in UG 99, CG and HG, but a long-range scale in UG 

79 and WG. SWC had a positive correlation with organic carbon, whereas a 

negative correlation with either sand content or bulk density, which was 

consistent with single scale analysis. Noted that sand content was the main soil 

variable controlling SWC, as shown by the high correlation coefficient at different 

spatial scales. 

Table 3.5. Structural correlation coefficients between soil moisture and soil and plant 

properties at short- and long-range scales for different grazing intensities 

Treat- Scale Soil Plant 

ment Sand SOC BD VC AGB 

UG79 Short -0.59 0.18 -0.16 -1.00 -1.00 

Long 0.53 0.27 -0.32 0.84 0.53 

UG99 

WG 

CG 

HG 

Short -1.00 1.00 -1.00 1.00 1.00 

Long 0.26 -0.22 0.23 -0.45 -0.77 

Short -0.02 0.10 0.08 0.05 0.99 

Long -0.51 -0.98 -0.24 -0.53 -0.99 

Short -0.98 0.32 -0.28 -1.00 -1.00 

Long 1.00 0.18 -0.06 0.46 0.41 

Short -1.00 1.00 -1.00 0.81 -0.64 

Long 0.52 -0.39 0.12 -0.86 0.76 

Sand: sand content (%); SOC: soil organic carbon (g kg -1 ); BD: bulk density (g cm -3 ); VC: 

vegetation coverage (%); AGB: above ground biomass (g m -2 ). 

The coregionalization analysis of plant variables, i.e. VG and AGB, also 

included three basic spatial structures: nugget effect (


water redistribution are scale-dependent. Therefore, it is critical to define the 

spatial scale of a simulation model when devising a sampling scheme. 

Table 3.6. Structural variation contributions of each component to the total variation (%) at 

different spatial scales for different grazing intensities 

Treat- Scale Soil Plant 

ment 

Moisture 

Sand SOC BD VC AGB 

UG79 Nugget 73% 0% 62% 62% 93% 56% 

Short 7% 100% 0% 0% 0% 44% 

Long 20% 0% 38% 38% 7% 0% 

UG99 

WG 

CG 

HG 

Nugget 31% 0% 20% 31% 69% 72% 

Short 69% 3% 80% 69% 0% 0% 

Long 0% 97% 0% 0% 31% 28% 

Nugget 82% 0% 21% 14% 33% 0% 

Short 18% 0% 0% 0% 67% 0% 

Long 0% 100% 79% 86% 0% 100% 

Nugget 39% 0% 65% 44% 0% 22% 

Short 0% 0% 6% 56% 0% 0% 

Long 61% 100% 29% 0% 100% 78% 

Nugget 30% 8% 51% 84% 67% 61% 

Short 34% 39% 49% 16% 33% 39% 

Long 37% 53% 0% 0% 0% 0% 

Sand: sand content (%); SOC: soil organic carbon (g kg -1 ); BD: bulk density (g cm -3 ); VC: 

vegetation coverage (%); AGB: above ground biomass (g m -2 ). 

The structural variation contributions of each component to the total variation 

are shown in Table 6. The variation of soil moisture was dominated by nugget 

effects in UG 79 and WG, the short-range scale component in UG 99, long-range 

scale component in CG, and averaged random, short- and long-range scale 

components in HG. As found for the structural correlation coefficients, the 

variations of soil and plant variables were basically consistent with that of SWC 

at different spatial scales. The weak correlations between soil moisture and plant 

properties in the two ungrazed sites may be further explained by the high nugget 

variations of plant properties. This difference could be due to grazing intensity as 

the regionalized variables had a more heterogeneous spatial distribution in the 

57

ungrazed plots (Table 3). WG differed from the other grazing intensities as SWC 

evolved in the micro- (random;


heterogeneity of soil moisture. As grazing intensity increased, animal trampling 

and grazing decreased vegetation coverage and biomass, leading to increased 

bulk density and decreased soil organic carbon. With heavy grazing this results 

in a large proportion of bare soil and highly deteriorated soil properties. Under 

such conditions, high evaporation and albedo are expected, leading to a trend 

towards lower soil moisture and a more homogeneous moisture distribution at 

the long-range scale. 

To completely understand the controlling factors of soil moisture from these 

data, several more considerations are needed. For example, correlations 

derived from a single measurement, i.e. plant properties, normally omit seasonal 

variations. Another deficiency is the lack information of soil moisture for deeper 

soil, and potential shifts in sampling position and frequency. Although trends 

between grazing intensity and soil moisture distribution were found, uncertainties 

in the grazing gradient and its management history may have confounded the 

results. Nevertheless, a conceptual model of the mechanistic controls on surface 

soil moisture variability affected by grazing intensity could be derived. 

4. Conclusions 

In this paper we identified the main factors controlling the spatio-temporal 

patterns of soil moisture as a function of grazing intensity in a semi-arid 

grassland. Our results revealed that the variability of soil moisture was 

characterized by weak to moderate spatial structures depending on sampling 

time and grazing intensity. The spatial variance and, to a lesser extent, the 

correlation length were found to be related to mean soil water content. The 

correlation analysis showed that soil and plant properties, especially soil 

physical properties that were impacted by grazing intensity, were important in 

controlling the distribution of soil moisture. To adequately describe spatial 

patterns of soil moisture, we need to consider the dependency of controlling 

processes on the local conditions. 

Using factorial kriging analysis, scale-dependency correlations between soil 

moisture and its controlling factors were detected. The soil and plant properties 

strongly controlled the variation of soil moisture in UG 99 at short-scale (90 m), 

59

in CG at long-scale (165 m) and in HG at averaged random, short-scale and 

long-scale components; but weakly controlled in UG 79 at micro-scale (random; 


Braun-Blanquet, J., 1964. Pflanzensoziologie. Wien-New York: Springer Verlag,. 

Buttafuoco, G., Castrignano, A., Busoni, E., Dimase, A.C., 2005. Studying the 

spatial structure evolution of soil water content using multivariate 

geostatistics. Journal of Hydrology 311, 202–218. 

Cambardella, C.A., Moorman, T.B., Parkin, T.B., Karlen, D.L., Turco, R.F., 

Konopka, A.E., 1994. Field scale variability of soil properties in Central Iowa 

soils. Soil Science Society of America Journal 58, 1501-1511. 

Cantόn, Y., Solé-Benet, A., Domingo, F., 2004. Temporal and spatial patterns of 

soil moisture in semiarid badlands of SE Spain. Journal of Hydrology 285, 

199–214. 

Casa, R., Castrignanò, A., 2007. Analysis of spatial relationships between soil 

and crop variables in a durum wheat field using a multivariate geostatistical 

approach. Eur. J. Agron. doi:10.1016/j.eja.2007.10.001. 

Castrignano, A., Giugliarini, L., Risaliti, R., Martinelli, N., 2000. Study of spatial 

relationships among some soil physico-chemical properties of a field in 

central Italy using multivariate geostatistics. Geoderma 97, 39–60. 

Chanzy, A., Bruckler, L., 1993. Significance of soil surface moisture with respect 

to bare soil evaporation. Water Resources Research 29, 1113–1125. 

Charpentier, M.A., Groffman, P.M., 1992. Soil moisture variability within remote 

sensing pixels, Journal of Geophysical Research 97, 18987–18995. 

Crave, A., Gascuel-Odoux, C., 1997. The influence of topography on time and 

space distribution of soil surface water content. Hydrological Processes 11, 

203–210. 

Entin, J.K., Robock, A., Vinnikov, K.Y., Hollinger, S.E., Liu, S.X., Namkhai, A., 

2000. Temporal and spatial scales of observed soil moisture variations in 

the extratropics. Journal of Geophysical Research 105 (D9), 11865–11877. 

Evenari, M., Shanan, L., Tadmor, N., 1971. Landform and landscapes, in the 

Negev. the Challenge of a Desert, Harvard University Press, Cambridge, pp. 

39–75. 

Famiglietti, J.S., Rudnicki, J.W., Rodell, M., 1998. Variability in surface moisture 

content along a hillslope transect: Rattlesnake Hill, Texas. Journal of 

61

62 

Hydrology 210, 259–281. 

Francis, C.F., Thornes, J.B., Romero Diaz, A., Lopez Bermudez, F., Fisher, G.C., 

1986. Topographic control of soil moisture, vegetation cover and land 

degradation in a moisture stressed Mediterranean environment. Catena 13, 

211–225. 

Gao, Q., Ci, L., Yu, M., 2002. Modeling wind and water erosion in northern China 

under climate and land use changes. Journal of Soil Water Conservation 57, 

46–55. 

Giacomelli, A., Bacchiega, U., Troch, P.A., Mancini, M., 1995. Evaluation of 

surface soil moisture distribution by means of SAR remote sensing 

techniques and conceptual hydrological modelling. Journal of Hydrology 

166, 445–459. 

Gómez-Plaza, A., Martínez-Mena, M., Albaladejo, J., Castillo, V.M., 2001. 

Factors regulating spatial distribution of soil water content in small semiarid 

catchments. Journal of Hydrology 253, 211–226. 

Goovaerts P., 1992. Factorial kriging analysis–a useful tool for exploring the 

structure of multivariate spatial soil information. Journal of Soil Science 143, 

597–619. 

Grayson, R.B., Western, A.W., Chiew, F.H.S., 1997. Preferred states in spatial 

soil moisture patterns: Local and nonlocal controls. Water Resources 

Research 33, 2897–2908. 

Hawley, M.E., Jackson, T.J., McCuen, R.H., 1983. Surface soil moisture 

variation on small agricultural watersheds. Journal of Hydrology 62, 

179–200. 

Hébrard, O., Voltz, M., Andrieux, P., Moussa, R., 2006. Spatio-temporal 

distribution of soil surface moisture in a heterogeneously farmed 

Mediterranean catchment. Journal of Hydrology 329, 110–121. 

Herbst, M., Diekkrüger, B., 2003. Modelling the spatial variability of soil moisture 

in a micro-scale catchment and comparison with field data using 

geostatistics. Physics and Chemistry of the Earth 28, 239–245. 

Henninger, D.L., Peterson, G.W., Engman, E.T., 1976. Surface soil moisture


within a watershed: Variations, factors influencing, and relationships to 

surface runoff. Soil Science Society of America Journal 40, 773–776. 

Hills, T.C., Reynolds, S.G., 1969. Illustrations of soil moisture variability in 

selected areas and plots of different sizes. Journal of Hydrology 8, 27–47. 

Johnston, K., Ver Hoef, J.M., Krivoruchko, K., Lucas, N., 2001. Using ArcGIS 

Geostatistical Analyst. GIS by ESRI. ESRI, New York. 

Krümmelbein, J., Wang, Z., Zhao, Y., Peth, S., Horn, R., 2006. Influence of 


of grassland soils in Inner Mongolia, P. R. China. Advances in Geoecology 

38, 93-101. 

Li, S.G., Harazono, Y., Oikawa, T., Zhao, H.L., He, Z.Y., Chang, X.L., 2000. 

Grassland desertification by grazing and the resulting micrometeorological 

changes in Inner Mongolia. Agricultural Forest and Meteorology 102, 

125–37. 

Qiu, Y., Fu, B., Wang, J., Chen, L., 2001. Soil moisture variation in relation to 

topography and land use in a hillslope catchment of the Loess Plateau, 

China. Journal of Hydrology 240, 243–263. 

Parent, A.C., Anctil, F., Parent, L.E., 2006. Characterization of temporal 

variability in near surface soil moisture at scales from 1 h to 2 weeks. 

Journal of Hydrology 325(1-4), 56–66. 

Reynolds, S.G., 1970. The gravimetric method of soil moisture determination, III: 

An examination of factors influencing soil moisture variability. Journal of 

Hydrology 11, 288–300. 

Samper, F.J., Carrera, J., 1990, Geoestadística-Aplicaciones a la hidrogeología 

subterránea: CIMNE. Barcelona, Spain, pp. 484. 

Wackernagel, H., 1998. Multivariate geostatistics: An introduction with 

applications. Springer, Berlin. 

Webster, R., Oliver, M.A., 2001. Geostatistics for Environmental Scientists. John 

Wiley & Sons, New York. p. 271. 

Western, A.W., Blöschl, G., Grayson, R.B., 1998. Geostatistical characterisation 

of soil moisture patterns in the Tarrawarra Catchment. Journal of Hydrology 

63

64 

205, 20–37. 

Western, A.W., Grayson, R.B., Blöschl, G., Willgoose, G.R., McMahon, T.A., 

1999. Observed spatial organisation of soil moisture and its relation to 

terrain indices. Water Resources Research 35(3), 797–810. 

Western, A.W., Zhou, S.L., Grayson, R.B., McMahon, T.A., Blöschl, G., Wilson, 

D.J., 2004. Spatial correlation of soil moisture in small catchments and its 

relationship to dominant spatial hydrological processes. Journal of 

Hydrology 286, 113–134. 

Williams, A.G., Ternan, J.L., Fitzjohn, C., de Alba, S., Perez-Gonzalez, A., 2003. 

Soil moisture variability and land use in a seasonally arid environment. 

Hydrological Processes 17, 225–235. 

Zeleke, T.B., Si, B.C., 2006. Characterizing scale-dependent spatial 

relationships between soil properties using multifractal techniques. 

Geoderma 134, 440–452. 

Zhao, Y., Peth, S., Krümmelbein, J., Horn, R., Wang, Z., Steffens, M., Hoffmann, 

C., Peng, X., 2007. Spatial variability of soil properties affected by grazing 

intensity in Inner Mongolia grassland. Ecological Modelling 205, 241–254.

Chapter 4 Temporal stability of soil moisture and its application in model result validation 

4. Temporal stability of soil moisture in a semi-arid 

steppe and its application in model result validation 

Ying Zhao, Stephan Peth, Xiaoyan Wang, Katrin Schneider, and Rainer Horn 

In preparation for Hydrological Processes. 

Abstract 

Temporal stability of soil moisture may have profound implications for water 

management and for reliable hydraulic model. The aim of this study is to 

investigate whether the time stability concept can be applied in model approach 

to locate the representative site and to make reliable estimates of soil moisture. 

In 2004-2006, in situ field monitoring soil moisture data were collected under a 

semi-arid steppe ecosystem (North China) at four plots representing different 

grazing intensities. While for the comparison with the moisture data of the 

observed point, a sampling grid of 100 points at each plot were arranged to 

identify spatio-temporal patterns of soil moisture. The results showed that soil 

moisture pointed to a considerable temporal stability patterns. The degree of 

persistence varied with grazing intensity, which partly was related to 

grazing-induced differences in soil and plant properties. The spatial distribution 

of soil moisture was more stable under wet conditions, but less stable under dry 

or medium conditions. Time stable points (TSPs) with low mean relative 

differences (MRD) (

tool for sampling strategies and for verifications of hydraulic process. 

Keywords: Time stability; Soil moisture; Hydraulic model; Sample strategy 


66 

Temporal stability of soil moisture is essential to understand the variability of 

structures and functions over an area, as well as the underlying hydrological 

processes (Grayson et al., 2002; Lin et al., 2006). It is characterized by the 

moisture status of soils which in some areas may be consistently wetter or drier 

than in other areas, or compared with the average moisture content across the 

whole area (Vachaud et al., 1985). Therefore, the temporal stability suggests 

that the field mean moisture could be determined from few point measurements 

of time-stable sites. This aids to reduce the sampling number and frequency, 

thus increase sampling efficiency without a significant loss in accuracy. 

Using the temporal stability concept, the sampling strategy is to investigate 

if some locations can represent the field mean moisture and if this character can 

conserve during the considered period. This has been proofed by many reports 

(Kachanoski and de Jong, 1988; Lin et al., 2006; Parent et al., 2006), which 

showed that certain sampling locations represented the mean of the whole study 

area reasonably well. Moreover, Vachaud et al. (1985) found that some locations 

could represent the mean field water content at any time of the year. Therefore, if 

the time stability concept can be successfully applied, the selected points may 

be assumed to accurately represent the mean soil moisture beyond the 

measured period. However, the selection of the time-stable sampling locations is 

based on a posteriori information derived from previous data analysis, thus it is 

not really practical. Hence, an apriori approach to estimate time stable locations 

is of more interest from a practical view. Time-stable locations have been found 

to be related to the averaged soil and plant properties, e.g. soil particle size and 

vegetation coverage (Vachaud et al., 1985; Hupet and Vanclooster, 2004; Starr, 

2005). Thus it should be possible to choose some parameters as a covariance of 

soil moisture for an apriori procedure to estimate time-stable locations. 

The high spatial variability of soil moisture and the small measurement


support requires therefore appropriate sampling strategies and monitoring sites 

(Kamgar et al., 1993). Due to the high cost and time-consuming of long-term soil 

moisture monitoring, it is rare that the monitoring sites are uniformly distributed 

in the entire studied area. On the contrary, the design of monitoring sites is 

usually irregular (Gomez-Plaza et al., 2000; Martinez-Fernandez and Ceballos, 

2005; Lin, 2006). Consequently, selected monitoring sites may not represent the 

field mean moisture in terms of the temporal stability concept, especially in an 

area without grid samplings but under heterogeneous conditions of soil, plant 

and topography. The alternatives to direct measurement are estimations by 

remote sensing data or use of hydraulic models (Albertson and Kiely, 2001; 

Cosh et al., 2004; Martinez-Fernandez and Ceballos, 2005). Both methods, 

however, require in situ measurements for the calibration and validation steps. In 

addition, in the study of hydraulic model, normally the observation or modeled 

points are selected without the prior analysis of representativeness of the 

selected points. To account for this uncertainty, the temporal stability concept, 

combined with a hydraulic model applied in the derived time stability point (TSP), 

should be valuable. However, till now this kind of study is still lack (Hupet and 

Vanclooster, 2004). Moreover, selected monitoring sites according to the 

temporal stability of soil surface moisture may not represent time-stable 

conditions for the deep soil (Lin, 2006). As Martinez-Fernandez and Ceballos 

(2003) and Pachepsky et al. (2005) pointed out, relatively less is known about 

the temporal stability of soil moisture as a function of depth. Thus, the use of 

hydraulic models is a very promising alternative to obtain soil moisture not only 

for the topsoil but also for the subsoil. 

The temporal stability possibly helps to provide valid data for hydraulic 

models by which applied in the responsible points. However, until now this has 

not been used and tested to estimate the field water content and flux for a given 

probability level (Seyfried, 1998). Here we analyze the temporal stability of soil 

moisture using data from four plots under different grazing intensities during 3-yr 

measurement periods. At the same time, we run a hydraulic model HYDRUS-1D, 

combined with in situ field monitoring soil moisture data, to evaluate the validate 

67

from the view of model. The purposes of this study are: i) to investigate the 

temporal stability of soil moisture under different water conditions, ii) to explore 

the grazing impact on temporal stability of soil moisture, and iii) to infer whether 

time stability concept can be availably applied for the hydraulic model to make 

reliable estimations. 

2. Material and Methods 

Research area and field measurements 

68 

The study was carried out in a semi-arid steppe ecosystem of Inner 

Mongolia (North China). The region is marked by a continental climate with cold 

winter and warm summer (Chen and Wang, 2000). Mean annual precipitation is 

about 350 mm, 85% of which is happened during the growing period (May to 

September). The soils are sandy loamy Calcic Chemozems according to IUSS 

Working Group (2006). Field measurements were carried out on experimental 

areas of the Inner Mongolia Grassland Ecosystem Research Station (IMGERS, 

43 o 37′50′′N, 116 o 42′18′′E), where four plots characterized by different grazing 

intensities were investigated. Two plots were protected from grazing since 1979 

(denoted as UG 79, 24 ha) and 1999 (UG 99, 35 ha). The other two plots were 

grazed: one was grazed only during winter time with 0.5 sheep units (ewe and 

lamb) ha -1 yr -1 (WG, 40 ha) and the other was heavily grazed with 2 sheep units 

ha -1 yr -1 (HG, 100 ha) during the whole year. Before fencing, all plots were 

moderately grazed. 

On the geostatistical area consisted of 80 points on a regular 15 m x 15 m 

grid and additional 20 points nested into the grid at 5 m distance, topsoil 

moisture content (0-6 cm) was measured at each of the four plots (Fig. 1). Soil 

moisture was measured with the Theta-probe (Type ML2x, Delta-T Devices, 

Cambridge, UK) at regular time intervals (biweekly) during the growing period of 

2004-2006. Additionally, we conducted in situ measurements for hydraulic 

conductivity (Mini-disk Infiltrometer, Decagon devices, USA), bulk density, and 

soil texture at each grid point. Time stability of soil moisture was analysed for 12 

sampling dates in 2004, for 8 sampling dates in 2005, and for 10 sampling dates 

in 2006.

Y orientation 


4822400 

4822380 

4822360 

4822340 

4822320 

4822300 

4822280 

4822260 

4822240 

4822220 

4822200 

473580 

1 

2 

71 

61 

51 

93 

72 

41 

94 62 

31 

21 

42 

52 

73 

11 

63 

32 

53 

22 

74 

43 

12 

64 

33 

54 

75 

23 

13 

88 

44 

65 

87 

55 

3 

34 

76 

14 

24 

45 

66 

35 

56 

4 

25 

77 

46 

67 

15 

36 

57 

5 

26 

78 

47 

81 

16 

68 

37 

82 

58 

27 

79 

6 83 

48 

17 

69 

38 

7 

59 

28 

89 97 

90 

80 

49 

18 

98 

8 

60 70 

29 

39 

50 

19 

40 

9 

30 

20 

10 

84 

85 86 

96 95 

92 91 

99 100 

473600 

473620 

473640 

473660 

473680 

X orientation 

473700 

473720 

473740 

N 

473760 

Fig. 4.1. Demonstration of sample grids in the experimental site (example for UG 79). Each 

site has an area of 105 m×135 m. Sampling points are spaced every 15 m with a subgrid 

spacing of 5 m (Hollow circles donate the field long-term monitoring positions). 

Time stability 

1985): 

δ 

δ 

i j 

i j = , 

, 

The mean relative difference (MRD) is expressed as follows (Vachaud et al., 

S − S 

1 

S 

m 

i = ∑ m j= 

1 

j 

δ 

j 

i, 

j 

where S i, 

j is soil moisture at location i (1-100) and time j, and S j is spatially 

mean soil moisture at the same time, and δ i, j is the relative difference of the 

spatially mean soil moisture,. Subsequently, the MRD δ i are calculated by 

averaged δ i, j over the m sampling dates during the considered period for each 

location. Following Grayson and Western (1998), sampling locations are 

considered to be time stable when δ i is close to zero and the standard deviation 

(1) 

(2) 

69

of δ i reaches a minimum value. In this study, all sampling locations with a MRD 

smaller than 5% and a low standard deviation


and further select a modeled location (model certainty) with a prior time-stable 

analysis. Simply, we will therefore focus on the following steps: at first, we detect 

the best TSP in each plot, and then run the HYDRUS-1D model with this point to 

get the water dynamics throughout the whole monitoring time. Finally, we 

compare the model results applied in TSP with the measured value both from 

TSP and field fixed monitoring site to address the validity of this method. The 

van Genuchten hydraulic parameter of TSP is derived from data of soil texture 

and bulk density by the neural network prediction tool ROSETTA (Schaap et al., 

2001). The calculated parameter is further optimized by in situ measurements of 

hydraulic conductivity using a Mini-disk Infiltrometer in 0.5 cm suction (Zhang, 

1997). 

To link the spatial patterns of soil moisture and provide a validating data for 

a hydraulic model, in situ filed monitoring soil moisture was measured using 

horizontally inserted Theta-probes in three depths at 5, 20 and 40 cm. At each 

plot three replicate profiles were installed and connected to one solar powered 

automatic data-logger (type DL2e, Delta-T Devices, Cambridge, UK), and the 

depth-averaged soil moisture at 30-min intervals was used. During the 

experimental period, the in situ weather station recorded the precipitation and 

other variables necessary to estimate the reference evapotranspiration from 

FAO Penman-Monteith equation (Allen et al., 1998). Plant parameters, such as 

vegetation coverage, leaf area index, and plant height, were taken from 

vegetation measurements. Root samples were taken with soil root auger and soil 

cores were separated into five depths of 0-10, 10-20, 20-50, 50-70 and 70-100 

cm (see details in Zhao et al., 2008b). 

3. Results and discussion 

Temporal stability of soil moisture 

To investigate whether temporally stable locations exist within the field, we 

perform a visual inspection of the graphs presenting the ranked mean relative 

differences (MRD) with their errors (Figs. 2-5). In general, the MRD shows a 

strong temporal stability for most points at each plot during the period studied. 

The MRD ranges from –15.9 to +11.9% at UG 79, from –10.8 to +14.7% at UG 

71

99, from –10.1 to +10.0% at WG, and from –20.1 to +27.4% at HG. Apparently, 

the absolute span of the MRD is smaller than 15% except for HG, where it 

reaches 28%. Compared with the results from previous temporal stability 

experiments (Vachaud et al., 1985; Martinez-Fernandez and Ceballos, 2003), 

our results have much smaller MRD and also smaller deviations, which may be 

explained by the smaller spatial area in our study. Due to comparable 

topographical attributes at the four plots, we ascribed the site-specific 

differences of MRD to differences of soil and plant properties affected by grazing 

effects (Zhao et al., 2008a). Compared with the other three plots, the heavily 

grazed plot (HG) has very sparse vegetation and a weakly structured soil, which 

accelerates the dynamics of soil moisture and thus decreases the temporal 

stability. This is consistent with Hupet and Vanclooster (2002), who concluded 

that the phenology of the vegetation played an important role in the temporal 

stability of the spatial patterns. Mohanty and Skaggs (2001) found that fields with 

silt loam soils showed poor temporal stability than those containing sandy loam 

soils. However, this can not be proofed in our study as sandy soil in HG 

expresses low temporal stability. This indicates that there should be other 

contributor to time stability of soil moisture, e.g. low water content itself in HG 

results in intensive and fast water dynamics, i.e. a low temporal stability. 

72 

Similarly, the temporal stability of soil moisture is moisture-dependent, that 

is, the MRD and its deviation for all plots are larger under dry and medium 

conditions, but smaller under wet conditions (Figs. 2-5). This is consistent with 

Lin (2006), who indicated that temporal stability is higher when the soil is wetter. 

It is possibly associated with that during the wet conditions water supply from the 

subsoil to the topsoil aids to keep the soil from temporal variability. Again some 

time stability points (TSPs) under wet conditions obviously lose time-stable 

characteristics under dry conditions. For example, sampling point 34 (denoted 

as p34) in UG 79 is temporally stable under wet, medium and all conditions, but 

it is not stable under dry conditions (Fig. 2). However, although the time stability 

characteristics of some points vary with water conditions, most points were 

independent on water conditions.

MRD(%) MRD(%) 

100 

80 

60 

40 

20 

0 

-20 

-40 

-60 

-80 

-100 

100 

80 

60 

40 

20 

0 

-20 

-40 

-60 

-80 

-100 


D 

W 

98 72 34 

34 

72 98 

0 20 40 60 80 100 

Ranked 

M 

ALL 

72 34 

98 72 

0 20 40 60 80 100 

Ranked 

Fig. 4.2. Ranked MRD of soil moisture in UG 79 in different water conditions during 3-yr 

measurement period (All: all measurements, D: dry, M: medium, W: wet). Vertical bars 

correspond to associated time standard deviations, labeled numbers are the time stability 

points that at least appear three times from four soil water conditions. 

MRD(%) 

MRD(%) 

100 

80 

60 

40 

20 

0 

-20 

-40 

-60 

-80 

-100 

100 

80 

60 

40 

20 

0 

-20 

-40 

-60 

-80 

-100 

D 

W 

12 37 

44 46 

12 37 41 

41 

46 72 

0 20 40 60 80 100 

Ranked 

M 

ALL 

37 46 

1244 

41 

0 20 40 60 80 100 

Ranked 

Fig. 4.3. Ranked MRD of soil moisture in UG 99 in different water conditions during 3-yr 



points that at least appear three times from four soil water conditions. 

73

74 

MRD(%) 

MRD(%) 

100 

80 

60 

40 

20 

0 

-20 

-40 

-60 

-80 

-100 

100 

80 

60 

40 

20 

0 

-20 

-40 

-60 

-80 

-100 

D 

W 

78 49 87 

90 11 

78 49 87 

90 11 

0 20 40 60 80 100 

Ranked 

M 

ALL 

78 11 95 

87 90 49 

0 20 40 60 80 100 

Ranked 

Fig. 4.4. Ranked MRD of soil moisture in WG in different water conditions during 3-yr 

measurement period (All: all conditions, D: dry, M: medium, W: wet). Vertical bars correspond 

to associated time standard deviations, labeled numbers are the time stability points that at 

least appear three times from four soil water conditions. 

MRD(%) 

MRD(%) 

100 

80 

60 

40 

20 

0 

-20 

-40 

-60 

-80 

-100 

100 

80 

60 

40 

20 

0 

-20 

-40 

-60 

-80 

-100 

D 

W 

35 22 

35 

22 

0 20 40 60 80 100 

Ranked 

M 

ALL 

35 22 

22 

0 20 40 60 80 100 

Ranked 

Fig. 4.5. Ranked MRD of soil moisture in HG in different water conditions during 3-yr 



points that at least appear three times from four soil water conditions.


According to the definition of MRD (Eq. 1), the site where values of MRD 

less or more than 0 indicate that it is drier or wetter than the field average 

moisture content. The results show a strong temporal persistence for any water 

conditions at each plot, e.g. slightly drier points could be tracked throughout the 

whole sampling period. This is exemplified by WG (Fig. 6), where the dry points 

with an MRD from –10.1 to 0% maintain nearly constant for any water conditions. 

That is, 40 points (i.e. the dry sampling locations) are systematically below the 

mean value. This is in accordance with Martínez-Fernández and Ceballos (2003) 

who also reported that the overall characteristics of the time stable points, i.e. 

either a wet, average or dry location, persisted for almost all sampling sites. In 

general, the points selected in dry conditions still retain their time-stable 

characteristics in wet conditions. Normally, such points are those that can be 

used to validate the hydraulic model according to time stability concept. 

Y orientation 

Y orientation 

4821960 

4821940 

4821920 

4821900 

4821880 

4821860 

4821840 

4821820 

4821800 

4821780 

4821960 

4821940 

4821920 

4821900 

4821880 

4821860 

4821840 

4821820 

4821800 

4821780 

11 

62 

52 

63 

53 74 

12 

2 

3 

44 

8634 15 36 

5 26 

81 

82838416 

6 27 

17 

7 28 

18 

8 

9 

10 

65 

55 

56 

76 

77 

67 

78 

68 

58 

69 

59 

70 

50 

87 

94 

88 

89 

100 

71 

93 72 

95 73 

1 

2 

22 

12 

13 

33 

53 

64 

15 

5 26 

81 

8283 6 

7 

18 

8 

9 

56 

46 

47 68 

58 

89 

90 59 

91 

50 

78 

80 

99 

W 

D 

71 

M 

72 

95 62 

73 

1 

2 

21 

32 

22 

12 

23 

13 85 

56 77 

4 

15 

5 

8283 

16 

6 

26 

38 

68 

58 

90 59 

78 

80 

99 

7 

8 

9 

19 

29 

50 

1 

2 

22 

12 

13 

15 

81 

5 

8283 

16 

6 

26 

7 

8 

18 

9 

71 

72 

95 62 

73 

63 

53 

28 

29 

56 77 

67 

78 

68 

58 

69 

89 59 

80 

90 70 

473080 

473100 

473120 

473140 

473160 

473180 

473200 

473220 

473240 

473260 

473280 

473080 

473100 

473120 

473140 

473160 

473180 

473200 

473220 

473240 

473260 

473280 

X orientation 

Fig. 4.6. Distribution of drier points than overall average moisture content in WG under 

different water conditions during 3-yr measurement period (All: all measurements, D: dry, M: 

medium, W: wet; labeled numbers are the points that soil moisture drier than field average 

moisture content). 

To choose the representative of TSPs for the model validation, we identified 

those points closest to the zero MRD and simultaneously had the lowest 

86 

50 

All 

75

standard deviation. In Figs. 2-5, the labeled points represent TSPs that at least 

appear three times under the four water conditions. Regardless of water 

condition, a common TSP can be found for p72 at UG 79 and for p22 at HG. 

There are no common points at UG 99 and WG that can represent all water 

conditions. However, except for the medium conditions, few more points are 

included. We alternatively choose the common point under the three conditions 

(all, wet and dry) for p41 at UG 99 and for p49 at WG. Thus there is at least one 

point which can be selected at each plot under various water conditions. This 

implies that even when data of several hydrological years or any water 

conditions are included, a suitable soil moisture predictor from TSP can still be 

obtained, which provides a basis to validate a hydraulic model. 

76 

Table 4.1. Correlation matrix between soil moistures of the sampling grid for different water 

conditions (P


determined with Pearson’s correlation coefficient (Table 1). There is a close 

correlation between MRD under different water conditions, indicating high 

temporal persistence of moisture patterns in the whole investigated area. The 

correlation coefficient is persistently higher on the heavily grazed plot (HG) than 

on the other plots. This might relate to relatively small differences in soil moisture 

under different water conditions, and spatial homogeneity in HG (Zhao et al., 

2008a). 

Time stable location soil moisture (%) 

40 

35 

30 

25 

20 

15 

10 

5 

40 

35 

30 

25 

20 

15 

10 

5 

Fit line; 

y = 0.9527x + 0.0076 (R 2 = 0.98) 

Upper 95% confidence level 

Lower 95% confidence level 

Fit line; 

y = 0.9993x + 0.0008 (R 2 = 0.98) 



0 

0 5 10 15 20 25 30 35 40 

a 

c 

Field mean soil moisture (%) 

Fit line; 

y = 0.9387x + 0.0164 (R 2 = 0.97) 



Fit line; 

y = 0.9717x + 0.0033 (R 2 = 0.98) 



5 10 15 20 25 30 35 40 

Fig. 4.7. Field mean soil moisture versus soil moisture of time stable point in four sites during 

the period of 2004-2006 (a:UG 79, b: UG 99, c: WG, and d: HG). 

Implications for sampling strategies 

The former results indicate that the spatial pattern of soil moisture slightly 

varies over water conditions. Temporal stability is relatively high during the wet 

conditions in this small catchment. This will be useful for guiding our further field 

sampling strategies as we may not need to collect data during wet seasons as 

frequently as during dry seasons. Furthermore, errors related to measurement 

inaccuracies will introduce a basic error of MRD and thus limits the 

representativeness of time stable points (TSPs). Averaging several TSPs could 

b 

d 

77

compensate the error of small scale variability at single sampling location and 

lead to more reliable estimates. Our study shows that only one location is 

needed when estimates of field mean moisture with 1% accuracy are warranted 

at each plot (Fig. 7). For instance, the mean water content estimated from p72 

(UG 79) corresponding to each sampling date is plotted against the field mean 

water content. A very good linear regression is established between them 

(r 2 =0.98**) with a relative precision of


indicate that these additional sampling locations are also close to the field mean 

water content. Therefore, this suggests that the locations with sand contents 

representative of the field mean sand contents can be served as time stable 

locations for estimations of soil moisture beforehand. 

Table 4.3. Mean values of soil texture, bulk density (BD) and hydraulic conductivity (K) in time 

stability point (TSP) for four investigated sites. 

Parameter UG 79 UG 99 WG HG 

Point number 72 41 49 22 

Sand (%) 49.0 43.2 50.0 67.8 

Silt (%) 34.5 40.5 32.4 21.5 

Clay (%) 16.5 16.3 17.2 10.7 

BD (g cm -3 ) 0.92 0.94 1.11 1.19 

K (cm d -1 ) 164.6 63.6 106.8 141.8 

Application of the time stability concept to a hydraulic model 

In this part we explore whether time stable points (TSPs) could be used to 

identify a certain modeled location. The selected TSP in the previous section is 

used to test by the HYDRUS-1D model. This is exemplified by UG 79. Firstly, we 

derived the van Genuchten parameters of p72 by ROSETTA, that is, θr=6.3%, 

θs=52.3%, α=0.009 cm -1 , n=1.509, Ks=149.9 cm d -1 , and L=0.5, respectively. 

Secondly, the Ks from ROSETTA is optimized by the scaled-predictive method. 

Here the predictive unsaturated hydraulic conductivity is scaled using a single 

measurement at a given water suction, from which Ks is replaced by the 

matching water suction (Thomasson et al., 2006). As shown in Fig. 8, there is a 

very good match between the measured infiltration rates and the simulated 

values via a two-term numerical conic function. According to this, the Ks is 

scaled to 164.6 cm d -1 (Table 3). Furthermore, the hydraulic parameters in the 

subsoil layer of p72 are also optimized based on the field measured value (i.e. 

soil texture and bulk density) in the soil profile combined with the 

scaled-calibration method (Thomasson et al., 2006). Remarkably, the modeled 

results from TSP match well the measured soil surface moisture in the same 

point (Fig. 9). Furthermore, the simulated result from p72 also shows a good 

agreement with measured water dynamics in another long-term monitoring 

79

location (labeled in Fig. 1) for the three soil depths (Fig. 9). On the one hand, this 

indicates that, given that accurate weather data and basic soil properties are 

available, the HYDRUS-1D model result applied in time-stable location can 

match well the measured water content both in TSP and in the whole area at 

each sampled date. On the other hand, it can also express well the field water 

dynamics monitored in another location. Conversely, our monitoring position can 

be viewed as kind of location that can represent the field mean moisture content 

(i.e. representative position) since it matches well with measured water content 

in TSP. From this aspect, temporal stability is a bridge connected between 

representative of monitoring and field mean value. Moreover, the model result 

can express field water dynamics in both topsoil and subsoil. This may 

circumvent the challenge of representative sites needed for multiple depths 

monitoring by a hydraulic model. Therefore, we consider that the temporal 

stability concept, linked with a hydraulic model, provides a useful tool for 

verifications of hydraulic process. 

80 

Cumulative infiltration (cm) 

4.0 

3.5 

3.0 

2.5 

2.0 

1.5 

1.0 

0.5 

0.0 

y = 0.0101x 2 + 0.0453x 

R 2 = 0.999 

0 5 10 15 20 

Square root of time (s 1/2 ) 

Fig. 4.8. Cumulative infiltration and fitting curve for the p72 location in UG 79.

Soil water content (%) 

40 

30 

20 

10 

0 

30 

20 

10 

0 

30 

20 

10 

0 


5 cm 

24-May 18-Jun 13-Jul 7-Aug 1-Sep 26-Sep 

Measured Modeled TSP Rainfall 


20 cm 


40 cm 


Time (days) 

Fig. 4.9. Soil moisture comparison between measured and simulated results in UG 79 during 

the growing period in 2006 (Measured: measured value in field long-term monitoring site; 

Modeled: modeled result based HYDRUS-1D; TSP: measured soil moisture in time stability 

point). 

4. Conclusions 

The obtained results show that it is possible to reliably select a location to 

represent the field mean soil moisture in a given area. The temporal patterns of 

soil moisture were stable irrespective of water conditions. Temporal stability of 

soil moisture was slightly different at the four plots due to the difference in soil 

and plant properties associated with grazing effects. For the benefit of the 

sampling strategy or field monitoring, the number and frequency of field 

measurements could be reduced using the time stability concept. Moreover, the 

choice of representativeness was verified with a covariance analysis, e.g., sand 

content. This provides a method to decide one representative location with a 

priori knowledge. Finally, the application of a hydraulic model (HYDRUS-1D) in 

time-stable points can express measured water content both in TSP and field 

water dynamics in topsoil and even in subsoil. This provides a basis for sample 

strategy and verifications of hydraulic process. 

40 

30 

20 

10 

0 

Rainfall (mm) 

81

Acknowledgements 

82 

This work was done with the financial support of the German Research 

Council (DFG) for a research grant of the DFG RU #536 MAGIM “Matter fluxes 

in Inner Mongolia as influenced by stocking rate”. 

References 

Albertson, J.D., and G., Kiely. 2001. On the structure of soil moisture time series 

in the context of land surface models. J. Hydrol. 243:101–119. 

Allen, R.G., L.S. Pereira, D. Raes, and M. Smith. 1998. Crop evapotranspiration. 

Guidelines for Computing Crop Water Requirements. Irrigation and 

Drainage Paper No. 56, FAO, Rome, p. 300. 



Cosh, M.H., T.J. Jackson, R. Bindlish, and J.H. Prueger. 2004. Watershed scale 

temporal and spatial stability of soil moisture and its role in validating 

satellite estimates. Remote Sens. Environ. 92(4):427–435. 

Gómez-Plaza, A., J. Alvarez-Rogel, J. Albaladejo, and V.M. Castillo. 2000. 

Spatial patterns and temporal stability of soil moisture across a range of 

scales in a semi-arid environment. Hydrol. Proc. 14:1261–1277. 

Grayson, R.B., and A.W. Western. 1998. Towards areal estimation of soil water 

content from point measurements: time and space stability of mean 

response. J. Hydrol. 207:68–82. 

Grayson, R., G. Blöschl, A.W. Western, and T.A. McMahon. 2002. Advances in 

the use of observed spatial patterns of catchment hydrological response. 

Adv. Water Resour. 25:1313–1334. 

Hupet, F., and M. Vanclooster. 2002. Intraseasonal dynamics of soil moisture 

variability within a small agricultural maize cropped field. J. Hydrol. 

261:86–101. 

Hupet, F., and M. Vanclooster. 2004. Sampling strategies to estimate field areal 

evapotranspiration fluxes with a soil water balance approach. J. Hydrol. 

292:262–280. 

Kachanoski, R.G., and E. de Jong. 1988. Scale dependence and the temporal


persistence of spatial patterns of soil water storage. Water Resour. Res. 

24(1):85–91. 

Kamgar, A., J.W. Hopmans, W.W. Wallender, and O. Wendroth. 1993. Plot size 

and sample number for neutron probe measurements in small field trials. 

Soil Sci. 156(4):213–224. 


preferential flow pathways in the Shale Hills catchment. Vadose Zone J. 

5:317–340. 

Martínez-Fernández, J. and A. Ceballos. 2003. Temporal Stability of Soil 

Moisture in a Large-Field Experiment in Spain. Soil Sci. Soc. Am. J. 

67(6):1647–1656. 

Martínez-Fernández, J. and A. Ceballos. 2005. Mean soil moisture estimation 

using temporal stability analysis. J. Hydrol. 312(1-4):28–38. 

Mohanty, B.P. and T.H. Skaggs. 2001. Spatio-temporal evolution and time-stable 

characteristics of soil moisture within remote sensing footprints with varying 

soil, slope, and vegetation. Adv. Water Resour. 24:1051–1067. 

Pachepsky, Ya.A., A.K. Guber, and D. Jacques. 2005. Temporal persistence in 

vertical distribution of soil moisture contents. Soil Sci. Soc. Am. J. 

69:347–352. 

Parent, A.C., F. Anctil, and L.E. Parent. 2006. Characterization of temporal 

variability in near-surface soil moisture at scales from 1 h to 2 weeks. J. 

Hydrol. 325(1-4):56–66. 

Schaap, M.G., F.J. Leij, and M.T. van Genuchten. 2001. Rosetta: A computer 

program for estimating soil hydraulic parameters with hierarchical 

pedotransfer functions. J. Hydrol. 251:163–176. 

Seyfried, M. 1998. Spatial variability constraints to modeling soil water at 

different scales. Geoderma 85(2-3):231–254. 





83

84 


Starr, G.C. 2005. Assessing temporal stability and spatial variability of soil water 

patterns with implications for precision water management, Agr. Water 

Manage. 72:223–243. 

Thomasson, M.J., P.J. Wierenga, T.P.A. Ferre. 2006 A Field Application of the 

Scaled-Predictive Method for Unsaturated Soil. Vadose Zone J. 

5:1093–1109. 

Vachaud,G., A. Passerat de Silans, P. Balabanis, and M. Vauclin. 1985. 

Temporal stability of spatially measured soil water probability density 

function. Soil Sci. Soc. Am. J. 49:822–828. 

Zhang, R. 1997. Determination of soil sorptivity and hydraulic conductivity from 

the disk infiltrometer. Soil Sci. Soc. Am. J. 61:1024–1030. 

Zhao, Y., Peth, S., Horn, R., Wang, X.Y., Giese, M., Steffens, M., Hoffmann, C., 

Gao. Y.Z. 2008a. Spatio-temporal variability of soil moisture in grazed 

steppe areas by multivariate geostatistics. (Revision for Journal of 

Hydrology) 

Zhao, Y., S. Peth, J. Krümmelbein, B. Ketzer, Y.Z. Gao, X.H. Peng, C. Bernhofer, 

and R. Horn. 2008b. Modeling grazing effects on coupled water and heat 

fluxes in Inner Mongolia grassland. (Submitted to Vadose Zone Journal)

Chapter 5 Modeling Grazing Effects on Coupled Water and Heat Fluxes in Inner Mongolia Grassland 

5. Modeling Grazing Effects on Coupled Water and Heat 

Fluxes in Inner Mongolia Grassland 

Ying Zhao, Stephan Peth, Julia Krümmelbein, Bettina Ketzer, Yingzhi Gao, 

Xinhua Peng, Rainer Horn, and Christian Bernhofer 

Submitted to: Vadose Zone Journal. 

ABSTRACT 

Over-grazing is regarded as a main cause for grassland degradation in 

semi-arid regions. To evaluate how soil water and heat fluxes respond to grazing, 

investigations on soil, plant and meteorological parameters were conducted at 

four sites with different grazing intensities (i.e., ungrazed since 1979, ungrazed 

since 1999, winter grazed, and heavily grazed), through three growing periods 

(2004-2006) in a steppe ecosystem of Inner Mongolia. Our results showed that 

heavy grazing resulted in an increased meso-pore volume and a decreased 

total- and macro-pore volume, accompanied by a reduction of saturated 

hydraulic conductivity. These soil structural changes were parameterized by 

Laboratory-derived hydraulic parameters (LDP model) based on HYDRUS-1D. 

In addition, to account for effects of the site-specific plant on the boundary 

condition, we used the model SHAW to estimate interception and partitioned 

evapotranspiration via a weighing lysimeter. On the basis of former calibration, 

HYDRUS-1D was verified with a good agreement between modeled and 

measured soil moisture and temperature. The modeled results indicated that soil 

heat fluxes increased with increasing grazing intensity. In comparison with the 

two ungrazed sites, winter grazing did not show clear effects on the water 

household components, while heavy grazing remarkably decreased interception 

by 50-55% and transpiration by 20-30%, and increased evaporation by 25-40%. 

We conclude that intensive grazing deteriorated soil functions and reduced plant 

available water, consequently reduced grassland productivity and enhanced the 

soil risks for wind and water erosion. 

85

Abbreviations: ET, evapotranspiration; Tp, transpiration; E, evaporation; Ks, 

saturated hydraulic conductivity; RMSE, root mean square error; AWHC, 

available water holding capacity. 

Movement of soil water and heat plays a key role in the whole water and energy 

balance in many agricultural issues, especially in arid or semi-arid regions. Since 

the 1970s, many numerical solutions for the water and heat flow equations were 

increasingly used (Feddes et al., 1988; de Jong and Bootsma, 1996; Saito et al., 

2006). A vast variety of soil water models exist ranging from simple water 

balance calculations to complex mechanistic models based on Richards’ 

equation. Richards’ equation, which usually describes unsaturated water flow, 

requires the two basic hydraulic functions: water retention curve θ(h) and 

hydraulic conductivity K(h) (Butters and Duchateau, 2002). This approach is only 

valid for rigid and non-interacting pore walls if laminar flow is assumed, thus 

detailed and in situ measurements are normally necessary for accurate 

simulation considering the spatial variability of soil properties. 

86 

Land management is important to heat and water transfer by altering plant 

or residue architecture and soil functions (Flerchinger et al., 2003). It is essential 

to quantify and predict management effects on soil properties in order to assess 

their consequences on plant production and the environment. Many studies 

have addressed the effects of land management (tillage, wheel-traffic 

compaction, grazing, and crop residue management) on soil properties (Green 

et al., 2003). However, few studies have dealt with the consequences of these 

practices on water and heat fluxes (Chung and Horton, 1987; Ndiaye et al., 

2007). Especially, considering the stress-dependent changes of the environment, 

an adequate account of water flow processes based on hydrological models is 

lacking due to the complexity of the soil system (Peth and Horn, 2006; Ndiaye et 

al., 2007), although there are widely applicable hydrological models like SHAW 

(Flerchinger and Saxton, 1989), HYDRUS (Šimůnek et al., 1998) and SWAT 

(Neitsch et al., 2002). 

In the steppe of Inner Mongolia, grazing is a widely used land management


system. In recent years, this steppe has been reported to be severely 

deteriorated because of heavy grazing (Chen and Wang, 2000). Heavy grazing 

accompanied with animal trampling normally has detrimental effects on soil 

properties (Greenwood and McKenzie, 2001; Zhao et al., 2007). Particularly in 

the topsoil, soil deformation is characterized by a decrease of pore volume and 

altered pore size distribution, which both affect water and air conductivities 

(Willat and Pullar, 1983; Krümmelbein et al., 2006), and soil water retention 

characteristics (Martinez and Zinck, 2004; Kutilek et al., 2006). The disturbance 

of soil structure also leads to lower water infiltrability and thus increases the risks 

for soil erosion and nutrient depletion. Therefore, these processes may cause a 

low water storage capacity and the loss of soil fertility, consequently decrease 

the grassland productivity (Christensen et al., 2004). 

Under the prevailing semi-arid climatic condition in Inner Mongolia, plant 

available water plays a key role for the sustainable development of steppe 

ecosystems. Therefore, it is necessary to understand the effects of grazing on 

water-related mechanisms and water budgets. Some studies have shown that 

grazing increases evaporation and decreases transpiration caused by an 

increase in bare ground area and a decrease in biomass production (Bremer et 

al., 2001; Chen et al., 2007). However, to which extent grazing affects 

evapotranspiration and its partitioning into transpiration and evaporation is still 

unclear (Leenhardt et al., 1995). Another deficiency is that interception by the 

plant canopy and residue is normally ignored despite its potentially large 

contribution to the water budget. Moreover, it is widely recognized that the 

movement of water and heat is closely coupled, but their mutual interactions are 

rarely considered in practical applications (Saito et al., 2006). For instance, the 

effect of heat transport on water flow is often neglected mainly because of a lack 

of comprehensive data to fully parameterize coupled water and heat flow 

models. 

In this paper, a comprehensive dataset including soil, plant and 

meteorological measurements from the project “MAGIM” (Matter fluxes in 

grasslands of Inner Mongolia as influenced by stocking rate; 

87

http://www.magim.com) are used to parameterize the model HYDRUS-1D. 

Conversely, modeling results are compared with in situ soil water and 

temperature measurements to verify the model parameterization. The objectives 

of this study are: (i) to quantify the grazing effects on water and heat fluxes by a 

Richards’ based hydraulic model, and (ii) to identify water use mechanisms in a 

typical Inner Mongolia grassland ecosystem. 

Study Site Descriptions 

88 

MATERIALS AND METHODS 

Experimental Site and Measurements 

This study was performed on long-term experimental sites of the Inner 

Mongolia Grassland Ecosystem Research Station (IMGERS; 43 o 37′50′′N, 

116 o 42′18′′E) situated in the Xilin River catchment (3650 km 2 ). The vegetation is 

characterized by the perennial rhizome grass Leymus chinensis and bunch 

grass Stipa grandis, respectively. Soils are classified as Mollisols according to 

USDA (1999). In the last two decades, annual mean air temperature was 0.7°C 

and precipitation was 343 mm, of which more than 85% fell during the growing 

period. Although rainfall season occurs during the vegetation period, plant water 

uptake is limited because of strong seasonal and inter-annual rainfall 

fluctuations (Chen and Wang, 2000). 

Four sites with different grazing intensities were investigated. Two sites 

were ungrazed since 1979 (UG 79, 24 ha) and 1999 (UG 99, 35 ha). The other 

two sites were grazed: one was grazed only during winter time with 0.5 sheep 

units (ewe and lamb) ha -1 yr -1 (WG, 40 ha) and the other was heavily grazed with 

2 sheep units ha -1 yr -1 (HG, 100 ha) during the whole year. Before fencing, all 

sites were moderately grazed. 

Field and Laboratory Measurements 

Soil moisture and temperature were measured at 30-min intervals using 

horizontally inserted Theta-probes (Type ML2x, Delta-T Devices, Cambridge, UK) 

and Platinum ground temperature probes (Pt-100), respectively. Theta-probes 

were calibrated soil specific and installed in three depths at 5, 20, and 40 cm. 

Soil temperature was measured in five depths at 2, 8, 20, 40, and 100 cm. At


each site three replicate profiles were installed and connected to one solar 

powered automatic data-logger. Moreover, an in situ micrometeorological station 

was installed to record precipitation and other variables necessary to estimate 

the reference evapotranspiration (ET) from the FAO Penman-Monteith equation 

(Allen et al., 1998). Some missing data, especially in 2004 when our 

experiments began, were filled with data from the weather station of IMGERS (3 

km away) and calibrated to the local conditions. To calibrate the modeled 

boundary condition, weighing mini-lysimeter experiments were conducted to 

partition soil evaporation (E) from ET. Lysimeters were designed by inserting 

PVC tubes (20.0 cm long, 5.4 cm diameter) into the ground and sealing them at 

the bottom. At each site, three replicates were installed at selected locations for 

both representative grass cover and bare soil surface, respectively. Weighing 

was conducted bihourly by an electronic balance where the corresponding water 

loss is referred to as ET for grass-covered ground and as E for bare ground, 

respectively. The difference between ET and E is consequently calculated 

considering the vegetal coverage and referred to as transpiration (Tp). Crop 

parameters, such as vegetation coverage, leaf area index (LAI), and plant height, 

were taken from vegetation measurements (Gao, 2007). In addition, also the 

residue coverage and weight were recorded. To determine the root length 

density, root samples were taken with a soil root auger and soil cores were 

separated into five depths of 0-10, 10-20, 20-50, 50-70, and 70-100 cm. 

Undisturbed soil samples were taken from four layers at 4-8, 18-22, 30-34, 

and 40-44 cm depths (Table 1). Soil texture was determined by pipette method, 

and total C-content was measured coulometrically. Water retention functions 

were determined with a ceramic pressure plate assembly by stepwise 

desaturating initially saturated samples at equilibrium matric potentials of -1, -3, 

-6, -15, -30, and -1500 kPa. Soil bulk density was determined after oven-drying. 

Finally, saturated hydraulic conductivity (Ks) was determined with a falling head 

permeameter (Hartge and Horn, 1999). 

Water and heat flow equations 

Numerical Modeling 

89

Simulations of soil water and heat flow were performed with HYDRUS-1D 

(Šimůnek et al., 1998). HYDRUS-1D is a finite element model for simulating the 

one-dimensional movement of water, heat, and multiple solutes in variably 

saturated media. The program numerically solves the Richards’ equation for 

saturated and unsaturated water flow and Fickian-based advection dispersion 

equations for heat and solute transport. Variably saturated water flow is 

described by the modified Richards’ equation: 

∂h 

∂T 

[ ( h) 

+ K ( h) 

+ K ( h) 

S 

∂ ( h) 

∂ = K ∂t 

∂z 

h ∂z 

h 

T ∂z 

θ w ] − 

[1] 

where θw is the volumetric liquid water content, t is time, z is the spatial 

coordinate positive upward, h is the pressure head, T is the temperature, and S 

is a sink term usually accounting for root water uptake. Kh and KT donate 

hydraulic conductivity due to a gradient in h and T, respectively. Soil hydraulic 

properties are described by the following expressions (van Genuchten, 1980): 

S ( h) 

e 

θ ( h) 

−θ 

1 

w r = = 

[2] 

n m 

θs 

−θ 

r [ 1+ 

αh 

] 

K ( h) 

− 

h 

l 

1/ 

m m 2 

= K s × Se 

( 1− 

( 1 Se 

) ) 

[3] 

where Se is the effective saturation, θs and θr are the saturated and residual 

water contents (L 3 L -3 ), respectively; the symbols α (L -1 ), n, and m=1-1/n are 

empirical shape parameters, and the inverse of α is often referred to as the air 

entry value or bubbling pressure; Ks is the saturated hydraulic conductivity (L 

T -1 ), and L is a pore connectivity parameter which normally is set to 0.5. The Eq. 

[2] was fitted to the measured water retention data (drying branch) using the 

RETC code (van Genuchten et al., 1991). 

The soil thermal regime is modeled with the conduction–convection heat 

flow equation (e.g., Nassar et al., 1992): 

∂T 

∂T 

[ λ( 

] − C q C ST 

∂ T ∂ C( ) = θ 

t z w) 

z w − 

∂ ∂ 

∂ 

∂z 

w 

θ [4] 

where C(θ) and Cw denote the volumetric heat capacity of the bulk soil and 

liquid phase, respectively. C(θ) is determined according to De Vries (1963): 

90 

C( θ × 

6 

θ ) ≈ ( 1. 

92θ 

m + 2. 

51θ 

o + 4. 

18 w ) 10 

[5]


where θm, θo, and θw represent the volume fractions of the mineral, organic 

matter, and water in soil, respectively. The terms on the right-hand side of Eq. [4] 

represent soil heat flow by conduction, convection of sensible heat with flowing 

water, and uptake of energy associated with root water uptake, respectively. 

Transfer of latent heat by vapor movement is ignored. λ(θw) denotes the soil 

thermal conductivity and q the water flux density while T is the soil temperature. 

The λ(θw) is described with a simple equation given by Chung and Horton 

(1987): 

1 

2 

3 

0. 

5 

λ ( θ w ) = b + b θ w + b θ w 

[6] 

where b1, b2, and b3 are empirical regression parameters. 

Initial and Boundary Conditions 

Initial conditions were set in the model in terms of measured water 

contents at the beginning of the simulation period. Value-specified boundary 

conditions were used for the top and bottom boundary of the flow domain. At the 

soil surface, an atmospheric boundary condition was imposed using daily data of 

precipitation, soil surface temperature, potential evaporation and transpiration. 

Water fluxes at the soil-atmosphere interface were corrected accounting for daily 

water interception by the canopy and litter using the SHAW model (Flerchinger 

and Saxton, 1989). Actual evaporation equaled the sum of intercepted 

evaporation and soil evaporation. Furthermore, the ratio between potential 

transpiration and soil evaporation was based on radiation partitioning according 

to Beer’s law as a function of LAI (Ritchie, 1972). The partitioning results were 

subsequently calibrated using in situ measurements from the weighing 

mini-lysimeter experiments. A free drainage condition and the measured soil 

temperature at 100 cm depth were used as bottom boundary conditions 

assuming that the water table is located far below the domain of interest and that 

heat transfer across the lower boundary occurs only by convection of liquid water. 

Soil surface temperature Ts ( 0 C) was estimated from the soil heat flux G (J m -2 

h -1 ) as follows (Chung and Horton, 1987): 

T λ + 

[7] 

G 

s = − ( θ ) Tz* 

91

cm). 

92 

where Tz* is the temperature measured at the uppermost soil depth, z* (2 

Root water uptake was simulated using the model of Feddes et al. (1978). 

According to the root measurements, the maximum rooting depth was 

considered to be 1 m, with the highest root density in the upper 30 cm. The 

rooting depth increased linearly from 0 cm at the beginning of simulation to a 

maximum value at the date of “full cover” or harvest. The critical pressure heads 

in the water stress response function of Feddes were adapted from grass 

(Wesseling, 1991), and adjusted for the local conditions with a value of -1500 

kPa for the wilting point p3. 

Water and Heat Flow Simulations 

The coupled water and heat transport module implemented in HYDURS-1D 

was tested using measured depth-averaged soil moisture and temperature (soil 

temperature at 5 cm depth was derived from Eq. [7] using the measured value in 

2 cm depth). According to phenological data from IMGERS (Chen and Wang, 

2000), the growing season starts around late April and ends around early 

October. According to our measurements (Gao, 2007), LAI increased rapidly 

from early May to late July, reached a peak in late August, and then decreased 

precipitously. Thus we assigned the growing season, i.e., simulation time, to the 

period from 1 May to 30 September. The simulated soil profile was considered to 

be 100 cm deep with observation nodes located at 5, 20, and 40 cm depths. 

In this study, we used five model approaches to explore the influence of soil 

hydraulic parameter availability on the quality of the model results (Table 1). 

Simulation 1 was conducted as a direct approach, using laboratory-derived 

hydraulic parameters (LDP) fitted by RETC code without any optimization. In 

simulation 2, hydraulic parameters were derived by a neural network (NN) 

prediction tool ROSETTA (Schaap et al., 2001) based on data of soil texture and 

bulk density. In simulation 3, an inverse model (Inverse) was used to estimate 

hydraulic parameters using a Levenberg–Marquardt parameter optimization 

algorithm. A layered soil profile with different parameters for each layer is 

assumed. The hydraulic functions described in Eqs. [1] and [2] may contain up to


six unknown parameters (θr, θs, α, n, Ks, and L). Hence, our three-layer soil 

profile (second and third depths were combined in one layer) may have as many 

as 18 unknown variables. To improve the optimization efficiency, the parameters 

θs, α, n, and Ks were simultaneously optimized in all layers with other 

parameters (i.e., θr and L) constant using time series of observations of θw for 

each depth. For a preliminary sensitive analysis, we also mixed the simulation 1 

and simulation 2, i.e., combined θr, θs, α, and n from LDP but Ks from NN 

(LDP-NN; simulation 4) and combined Ks from LDP but θr, θs, α, and n from NN 

(LDP+NN; simulation 5), respectively. The model was calibrated in 2004 and 

validated in 2005 and 2006. Optimizing efficiency, as well as model efficiency 

was evaluated by the root mean square errors (RMSE): 

RMSE 

= 

N 

1 2 

N ∑ Pi 

− Oi 

) 

i= 

1 

( [8] 

where N is the number of observations and Pi and Oi are the simulated and 

measured values, respectively. 

Table 5.1. Summary of simulation approaches used in model. 

Simulation 

approach 

Abbreviation 

Description 

Simulation 1 LDP 

Laboratory-derived hydraulic parameter from soil core, 

fitted by RETC software 

Simulation 2 NN 

Neutral network values based on soil texture 

and bulk density analysis, predicted by ROSETTA 

Simulation 3 Inverse Simulation with inverse model 

Simulation 4 LDP-NN Simulation with Ks from NN and measured θr, θs, α, n 

Simulation 5 LDP+NN Simulation with θr, θs, α, n from NN and measured Ks 

RESULTS 

Soil Texture, Bulk Density and Carbon Content 

At all four sites soil texture is coarser in the subsoil (18-22 cm) than in the 

topmost soil layer (4-8 cm) (Table 2). Also bulk density increases from the first to 

the second depth, while total C-content generally decreases (except at the 

second layer of UG 99) with increasing soil depth. Sand content is the highest in 

HG and the lowest in WG. Compared to UG 99 and WG, a higher sand content in 

93

UG 79 is assumed to be the result of a sandier parent material. However, the 

coarse-textured topsoil in HG might be partly attributed to erosion of fine 

particles by wind from the predominantly bare ground. In the topsoil, bulk density 

increases with increasing grazing intensity, indicating the effects of mechanical 

stresses on soil structure induced by animal trampling. At the ungrazed sites, 

bulk density is much higher in the second soil layer compared to the topsoil. This 

is a sign that the upper soil layer has obviously recovered from previous sheep 

trampling. Compared to the other three sites, the total C-content (4-8 cm) in HG 

is reduced by about 50%, which may be due to both wind erosion of particulate 

organic matter and higher mineralization rates. 

94 

Table 5.2. Soil texture, bulk density and organic carbon content of the four sites. 

Site 

Soil depth Sand Silt Clay 

Bulk 

density 

Organic 

carbon 

(cm) (%) (%) (%) (g cm -3 ) (g kg -1 ) 

UG 79 4-8 59.0 27.5 13.5 1.14 22.0 

18-22 66.1 21.2 12.7 1.39 16.5 

30-34 75.3 14.6 10.2 1.42 10.1 

40-44 78.2 12.5 9.3 1.40 8. 2 

UG 99 4-8 52.5 31.0 16.5 1.11 19.6 

18-22 59.1 26.9 14.0 1.33 23.1 

30-34 54.8 29.9 15.3 1.34 15.9 

40-44 56.5 28.1 15.4 1.27 15.0 

WG 4-8 49.8 32.2 18.0 1.28 22.2 

18-22 54.9 28.0 17.1 1.29 13.2 

30-34 54.7 29.7 16.6 1.24 11.4 

40-44 54.9 27.9 17.3 1.25 11.6 

HG 4-8 65.8 21.6 12.6 1.30 11.8 

18-22 73.2 16.0 10.7 1.42 7. 2 

30-34 71.3 17.2 11.6 1.43 6. 5 

40-44 72.8 16.6 10.6 1.45 6. 0 

Water Retention Characteristics and Hydraulic Conductivity 

Total pore volume generally decreases from the topsoil layer to the subsoil 

for all sites (Fig. 1). In the topsoil, the total and large pore (>50 μm) volumes 

significantly decrease with increasing grazing intensity. Especially, HG shows a


clear reduction in the fraction of large pores which is accompanied by an 

increasing fraction of medium pore sizes (0.2-50 μm), indicating strong soil 

compaction effects due to animal trampling. Note that owing to the 

coarse-textured soil in HG, the fraction of larger pores should be higher than that 

in the other fine-textured sites. Moreover, the volume fraction of fine pores (50 μm), pF

causes a platy soil structure (Krümmelbein et al., 2006). The relatively high Ks in 

HG is again associated with the coarser soil texture. However, despite the 

similar sandy texture, Ks is nearly twice higher in UG 79 than in HG, which is 

attributed to not only the higher fraction of large pores but an improved pore 

continuity of the more structured soil in the former. 

Table 5.3. The laboratory-derived hydraulic parameters (θr=residual water content, 

θs=saturated water content, α=reciprocal value of air entry pressure, n=the 

smoothness of pore size distribution and m=1-1/n), and Ks=saturated hydraulic 

conductivity) for the four sites (R 2 is determination coefficient fitted by RETC 

software). 

Treat- θr θs α n R 2 Ks 

Soil depth 

ment (cm) (cm 3 cm -3 ) (cm 3 cm -3 ) (cm -1 (cm 

) (-) 

day -1 ) 

UG 79 4-8 0.075 0.572 0.026 1.766 0.983 165.0 

18-22 0.086 0.472 0.019 2.199 0.985 133.3 

30-34 0.079 0.453 0.015 2.549 0.986 113.7 

40-44 0.071 0.452 0.016 2.120 0.985 71.9 

UG 99 4-8 0.086 0.577 0.022 1.628 0.983 82.3 

18-22 0.096 0.526 0.020 1.750 0.988 66.7 

30-34 0.102 0.506 0.017 1.940 0.996 57.1 

40-44 0.107 0.502 0.016 2.241 0.992 54.3 

WG 4-8 0.075 0.570 0.021 1.636 0.996 54.7 

18-22 0.050 0.526 0.023 1.616 0.994 69.8 

30-34 0.072 0.529 0.012 1.791 0.969 72.7 

40-44 0.079 0.517 0.014 1.742 0.981 61.3 

HG 4-8 0.057 0.523 0.014 1.718 0.982 93.0 

18-22 0.049 0.473 0.013 2.030 0.983 92.7 

30-34 0.060 0.479 0.012 2.130 0.982 75.9 

40-44 0.058 0.453 0.013 2.108 0.987 79.4 

Soil Moisture and Temperature Regimes 

Fig. 2 shows contour plots of depth-averaged soil moisture and temperature 

in the four sites during the growing period in 2006. Soil moisture decreases with 

increasing soil depth from May to July for all sites, indicating low water storage in 

deeper soil layers. However, a reverse trend is observed from late July to 

September except for HG, indicating high plant water demands. Compared with 

96


the ungrazed sites, the grazed sites generally have lower soil water storage. In 

August, the driest soil conditions occur at HG, where water contents approach 

the permanent wilting point due to both strong evaporation losses and low water 

storage in the subsoil. In contrast to this, rainfall water can percolate into deeper 

soil layers (>40 cm) in the two ungrazed sites. At the topsoil (e.g., 20 cm), UG 79 

is drier than UG 99 probably due to higher water withdrawal at the more 

productive site UG 79. 

Soil temperature shows typical changes with diurnal weather condition (Fig. 

2). With increasing soil depth, soil temperature decreases, indicating energy 

transfer from the topsoil to the subsoil during the growing period. With increasing 

grazing intensity, soil temperature is higher, especially in the topsoil. In general, 

soils seem to warm up or cool down quicker in the grazed sites than the 

ungrazed sites. However, compared to the other sites, the soil in HG seems to 

buffer the abrupt drop in temperature as early autumn snow occurred on 7 

September, which might relate to the higher soil temperature and the lower heat 

conductivity of the drier soil at HG. 

a 

b 

Soil depth (cm) 

c 

d 

-20 

-40 

-20 

-40 

0 

-20 

-40 

0 

-20 

-40 

0 

0 

0 

20-May 9-Jun 29-Jun 19-Jul 8-Aug 28-Aug 17-Sep 

Time (days) 

0.25 

0.21 

0.17 

0.15 

0.13 

0.11 

0.09 

0.07 

0.05 

Soil moisture (cm 3 cm -3 ) 

a 

b 

Soil depth (cm) 

c 

d 

-20 

-40 

0 

0 

0 

-20 

-40 

0 

-20 

-40 

0 

-20 

-40 


Time (days) 

Fig. 5.2. Contour plots of soil moisture (left side) and soil temperature (right side) for the 

four grazing intensities from May to Sept. 2006 (a is UG 79, b is UG 99, c is WG, and d is 

HG). 

Model Calibration and Validation 

25 

22 

19 

16 

13 

10 

7 

4 

1 

Soil temperature ( 0 C) 

97

Model calibration is done separately for each site using site specific upper 

boundary conditions for partitioned evapotranspiration (Fig. 3) and interception 

(Fig. 4), as well as the different model approaches (Fig. 5). Evapotranspiration 

(ET) in average decreases with increasing grazing intensity. For the grazed sites 

we observe higher E/ET ratios than for the ungrazed sites indicating the strong 

influence of the plant canopy on the upper boundary. Diurnal variations in E/ET 

ratios show that great differences between Tp and E occur in the midday and 

vapor is condensated during the night. Water interception decreases with 

increasing grazing intensity with significantly lower values at the heavily grazed 

site (Fig. 4) due to lower canopy and residue coverage. Note that the SHAW 

model calculates intercepted evaporation as a function of LAI, plant biomass, 

amount of residue, precipitation, and previously intercepted water and therefore 

accounts for temporal dynamics of intercepted water. 

Evapotranspiration rate (mm h -1 ) 

0.5 

0.4 

0.3 

0.2 

0.1 

0.0 

-0.1 

-0.2 

-0.3 

0.5 

0.4 

0.3 

0.2 

0.1 

0.0 

-0.1 

-0.2 

-0.3 

WG 

0.5 

UG 79 UG 99 

0.4 

ETa E Ep 

21-Aug 22-Aug 22-Aug 22-Aug 23-Aug 24-Aug 


Time (day/hour) 

0.3 

0.2 

0.1 

0.0 

-0.1 

-0.2 

-0.3 

0.5 

0.4 

0.3 

0.2 

0.1 

0.0 

-0.1 

-0.2 

-0.3 

HG 


21-Aug 22-Aug 22-Aug 22-Aug 23-Aug 23-Aug 23-Aug 24-Aug 

Time (day/hour) 

Fig 5.4. Curve of partitioning evapotranspiration (ET) into transpiration (Tp) and 

evaporation (E) in August 2006 determined by weighing lysimeter experiments for four 

sites (error bar represents standard deviation of three replicates). 

98


Accumulative 

interception (mm) 

Accumulative 

rainfall (mm) 

20 

16 

12 

8 

4 

0 

240 

200 

160 

120 

80 

40 

0 

UG 79 UG 99 WG HG 



Time (days) 

Fig. 5.4. Cumulative canopy and residue interception calculated by SHAW model for the 

four sites during the growing period in 2006 (accumulative rainfall as a reference). 

Model parameterization was conducted from 1 Aug. to 31 Aug. 2004. The 

general trend in terms of model approaches is the same in different sites, this is 

exemplarily demonstrated for the treatment UG 79 at 5, 20, and 40 cm depths 

(Fig. 5). Except for the inverse model, all model approaches underestimate soil 

moisture at 40 cm depth, which might be associated with a high unsaturated 

hydraulic conductivity in this layer. Model accuracy among the five model 

approaches is compared in terms of RMSE (Table 4). The RMSE is the lowest 

for inverse model exhibiting the best goodness-of-fit, followed by LDP and then 

NN. This proofs that the calibrated and measured parameters are more accurate 

than the statistical estimations based on pedotransfer functions (PTFs). The 

weakest predictions are found for LDP-NN and LDP+NN, i.e., the approaches 

which measured and statistical parameters were mixed. However, the prediction 

by LDP-NN is closer to that by LDP, indicating that parameters θr, θs, α, and n 

are more sensitive than Ks, which is also supported by the fact that the 

estimation by LDP+NN is almost identical to that by NN. 

99

Soil water content (cm 3 cm -3 ) 

0.4 

0.3 

0.2 

0.1 

0.4 

0.3 

0.2 

0.1 

0.4 

0.3 

0.2 

0.1 

5 cm 

M LDP NN Inverse LDP-NN LDP+NN 

2 

20 cm 

4 6 8 10 12 

2 

40 cm 

4 6 8 10 12 

4-Aug 10-Aug 16-Aug 22-Aug 28-Aug 

Time (days) 

Fig. 5.5. Comparison of the five model parameterization procedures with measured soil 

moisture in 2004 for UG 79. M: measured, LDP: laboratory-derived hydraulic parameters 

NN: neural network, Inverse: inverse model, LDP-NN: simulation with Ks from NN and 

measured θr, θs, α, n, and LDP-NN: simulation with θr, θs, α, n from NN and measured 

Ks. 

Model validation is performed with measured soil moisture and temperature 

data during the growing period of 153 d from 1 May to 30 Sept. 2005 and 2006, 

respectively. In general, simulated and measured water contents are 

comparable during the entire simulation period (Fig. 6). The increases in water 

content after rainfall at 5 and 20 cm depths are predicted reasonably well. In 

contrast to this, the simulated water contents decreased quicker than the 

measured values during a prolonged drought period. However, except for the 

inverse model, simulated and measured water contents at 40 cm depth are not 

matching well. Consistent with the model calibration, the inverse model predicts 

water contents best with the lowest RMSE (Table 4). This is to be expected 

because it adjusts the hydraulic parameters to fit the observed data for each soil 

layer. The LDP does not yield satisfying predictions at the deeper soil mirroring 

transfer gaps from laboratory data to field conditions, such as the 

100


representativeness of sampling and errors of measurement. The weak prediction 

is obtained from the NN, which is likely due to the spatial variations of PTFs. 

Volumetric soil water content (cm 3 cm -3 ) 

0.4 

0.3 

0.2 

0.1 

0.0 

0.3 

0.2 

0.1 

0.0 

0.3 

0.2 

0.1 

0.0 

5 cm 

18 Jun 05 07 Aug 05 26 Sep 05 05 Jun 06 25 Jul 06 13 Sep 06 

18-Jun-05 

20 cm 

07-Aug-05 26-Sep-05 05-Jun-06 25-Jul-06 13-Sep-06 

18-Jun-05 07-Aug-05 26-Sep-05 05-Jun-06 25-Jul-06 13-Sep-06 

40 cm Measured LDP NN Inverse 


Time (days) 

Fig. 5.6. Soil moisture comparison between measured and simulated results in UG 79 

during the growing period in 2005-2006. LDP: laboratory-derived hydraulic parameters, 

NN: neural network, and Inverse: inverse model. 

Table 5.4. Root mean square errors (RMSE) of five simulation approaches for model 

calibration and validation in UG 79. 

Simulation time 

Depth 

(cm) 

40 

30 

20 

10 

0 

Rainfall (mm) 

LDP NN Inverse LDP-NN LDP+NN 

1 Aug. - 5 0.025 0.071 0.003 0.018 0.075 

31 Aug., 2004 20 0.011 0.041 0.003 0.032 0.045 

(calibration) 40 0.072 0.098 0.006 0.118 0.087 

1 May - 5 0.019 0.034 0.038 - - 

31 Sept., 2005 20 0.012 0.008 0.005 - - 

(validation) 40 0.066 0.062 0.006 - - 

1 May - 5 0.032 0.040 0.045 - - 

31 Sept., 2006 20 0.028 0.024 0.022 - - 

(validation) 40 0.041 0.048 0.008 - - 

-: Data not shown. 

101

Differences in soil temperature simulations by different model approaches 

are not as apparent as soil moisture simulations. All simulated temperatures 

match well with the measured values (Fig. 7), suggesting that the effects of 

hydraulic parameters on heat transfer are not present. The model 

underestimates soil temperature at 5 cm depth, which might be associated with 

the overestimation of heat conductivity in the topsoil. Heat flux into deeper soil 

layers is predicted well even at steep gradients like on 7 Sept. 2006, indicating 

that HYDRUS is also able to display heat transport in soils under abruptly 

changed boundary conditions. 

Soil temperature ( o C) 

30 

20 

10 

0 

30 

20 

10 

0 

30 

20 

10 

0 

5 cm 

18-Jun-05 

20 cm 

07-Aug-05 26-Sep-05 05-Jun-06 25-Jul-06 13-Sep-06 


40 cm 

Measured LDP NN Inverse 


Time (days) 

Fig. 5.7. Soil temperature comparison between measured and simulated results in UG 

79 during the growing period in 2005-2006. LDP: Laboratory-derived hydraulic 

parameters, NN: neural network, and Inverse: inverse model. 

Soil Water and Heat Fluxes 

After model calibration and validation soil water and heat fluxes, and water 

budget components are calculated from the LDP models for the four sites during 

the growing periods of 2004-2006. Despite identical net radiation are assumed 

for each site, daily soil heat fluxes increase with increase of grazing intensity (Fig. 

8), suggesting that partitioning of the surface energy is influenced by a change in 

102


Soil heat flux (MJ/m 2 /day) 

7 

6 

5 

4 

3 

2 

1 

0 

-1 

-2 



Time (days) 

Fig. 5.8. Simulated soil surface heat fluxes for the four sites during the growing period 

in 2006. 

vegetation cover and near surface soil moisture. Under the same input of rainfall, 

actual transpiration (Tp) decreases with increasing grazing intensity (Fig. 9), 

which is characterized by only minor differences between the two ungrazed and 

the moderately grazed sites, especially at the first stage of growing season, but 

significantly lower value for the HG site during the whole growing period. With 

increasing grazing intensity, ET slightly decreases because of increasing runoff 

and drainage (Table 5). Especially in HG, the annual runoff reaches to about 6 

mm in 2004 and 2006, respectively. Calculating mean Tp/ET ratios for the three 

growing seasons, we find that about 48-52% of ET for UG 79, UG 99 and WG 

are attributed to transpiration, and only 38% for HG. In comparison with the two 

ungrazed sites, winter grazing does not show clear effects on the water 

household components, while heavy grazing remarkably decrease water 

interception by 50-55% and transpiration by 20-30%, and increase evaporation 

by 25-40%. 

103

Accumulative actual transpiration (mm) 

140 

120 

100 

80 

60 

40 

20 

0 



Time (days) 

Fig. 5.9. Simulated accumulative actual transpiration for the four sites during the 

growing period in 2006. 

Table 5. Simulated (LDP model) water budget components from May to September 

during 2004-2006, RETC simulation result with Hydrus-1D (in mm, P: Precipitation, I: 

Interception, T: Transpiration, E: Evaporation, ΔS: Water storage change, D: Drainage, 

and R: Runoff; error is model errors). 

Year Treatment I T E ΔS D R error 

2004 UG 79 16.3 171.0 106.0 -23.6 1.1 1.7 2.5 

(P= UG 99 15.2 166.0 113.2 -20.1 3.2 1.6 -4.1 

275 mm) WG 14.7 154.0 115.3 -18.0 4.0 4.1 0.9 

HG 7.1 131.2 141.8 -21.1 6.5 5.5 4.0 

104 

2005 UG 79 14.8 90.4 73.8 -29.5 1.3 0.3 -4.1 

(P= UG 99 14.4 82.4 75.6 -23.4 1.7 0.2 -3.9 

147 mm) WG 13.2 79.7 86.1 -29.2 1.3 0.2 -4.3 

HG 6.6 65.3 104.6 -33.6 5.7 0.4 -2.0 

2006 UG 79 18.3 131.4 110.8 -19.5 1.8 1.0 -1.8 

(P= UG 99 17.4 117.5 109.3 -3.3 1.5 0.7 -1.1 

242 mm) WG 16.3 103.6 116.2 5.9 0 1.7 -1.7 

HG 8.2 71.9 157.7 1.8 1.0 5.1 -3.7


The decrease in soil water storage and the low drainage indicate that 

infiltrated water is mostly consumed by ET within a single growing period (Table 

5). Note that in a dry year (2005) ET decreases by 40% compared to a wet year 

(2004) due to water deficiency. This is in line with large differences between 

reference ET and Pan-evaporation in 2005, but minor differences between them 

in 2004 (Fig. 10). Differences in Tp related to grazing are also less pronounced 

in the wetter years (2004 and 2006) than in the drier year (2005), indicating the 

restriction on plant water uptake by low water availability under dry conditions. 

Evapotranspiration (mm) 

20 

15 

10 

5 

0 

20 

15 

10 

5 

0 

20 

15 

10 

5 

0 

ET FAO ET Pan ET Act 

25-May-06 19-Jun-06 14-Jul-06 08-Aug-06 02-Sep-06 27-Sep-06 



Time (days) 

Fig. 5.10. Evapotranspiration comparison from three methods in UG 79 during the 

growing period in 2004-2006 (ETFAO: FAO, ETPan: Pan evaporation, and ETAct: Modelled 

ET). 

DISCUSSION 

Grazing Effects on Soil Properties and Model Validation 

Our results showed a decrease of total- and macro- pore volume and an 

increase of meso-pore volume as a result of grazing. This was in agreement with 

Villamil et al. (2001), who proofed a change of water retention characteristics by 

105

grazing for a southern caldenal soil in Argentina. However, we also found that 

soil physical functions had been improved after fenced for a longer period (e.g., 

25 yr in UG 79), as mirrored by a decrease in bulk density accompanied by an 

increase of total pore volume. Processes contributing to the natural recovery of 

physically degraded soil in the ungrazed sites might include reduction of animal 

trampling, earthworm burrowing, root penetration and decay, wetting and drying 

cycles, and freezing and thawing cycles (Drewry et al., 2006). Our results 

therefore suggests that reduction of grazing intensity could improve soil 

functions again, which was in agreement with Proffitt et al. (1995), who found 

that natural recovery of soil physical properties under pasture happened 

relatively rapid. 

106 

Grazing influenced hydraulic parameters, i.e., decreasing θs, θr, α, and Ks 

due to compaction effects. These grazing-induced changes were parameterized 

by the coupled water and heat model HYDRUS-1D. To account for the special 

conditions in grazed semi-arid grassland systems we modified root growth 

dynamics and boundary conditions according to the different grazing intensities. 

Particularly, three main improvements were underlined. Firstly, we ran root 

growth model to reflect dynamics of plant water uptake, which was important 

because models designed to simulate agriculture management were normally 

limited to simulate crop growing. Secondly, estimate of the intercepted water 

provided an accurate boundary condition. Finally, the ET partitioning, based on 

the actual measurements, was essential for the prediction of plant transpiration 

since it was only determined by root water extraction function related with 

potential transpiration in HYDRUS code. 

After the model parameterization, there was a good matching between 

measured and simulated soil moisture and temperature at each site. This 

indicated that HYDRUS-1D model was suitable tool to reflect the grazing effects 

on water and heat fluxes. The LDP model was better than the NN model to 

predict the water fluxes as it reflected the soil structural changes affected by soil 

compaction. This result was consistent with Richard et al. (2001), who


distinguished textural and structural fractions for the soil pore space and found 

that soil compaction mainly affected the structural pore space. This can explain 

why the NN model, that only accounted for soil textural fraction, can not make an 

accurately predictions of compaction effects. Therefore, we consider that in situ 

measurements of θ(h) and K(h) are necessary to reflect the soil structural 

changes induced by land management when process-oriented modeling 

techniques is used to evaluate the influences of land management. Moreover, 

we showed both water and heat transports as a function of grazing 

managements and therefore coupled water flow with heat transport. However, 

we emphasized particularly on aspects of water fluxes (detailed in the next 

section) because the heat stress was weaker for grass growth than the water 

stress in this grassland (Zhang et al., 2005). 

Grazing Effects on Water Budget in Semi-arid Grassland 

The components of ET, i.e., interception, transpiration and evaporation 

significantly varied with grazing intensity, although ET itself was only slightly 

influenced by grazing as shown by other studies (e.g., Bremer et al., 2001; Chen 

et al., 2007). Compared with the grazed sites, interception and transpiration 

increased and soil evaporation decreased in the ungrazed sites where soil was 

covered more completely by live or dead plant materials. In contrast to this, 

evaporation in the grazed sites simultaneously increased due to larger bare 

ground areas. Consequently, soil water storage decreased with increasing 

grazing intensity. 

Parameters that influence the simulation of water budgets were soil 

hydraulic properties, plant characteristics and weather conditions, of which the 

latest one was assumed identical for all sites, while the other parameters were 

grazing-dependent. Because of the grazing-induced changes, overall, water 

budget in Inner Mongolia grassland was significantly influenced by grazing. 

Firstly, grazing decreased hydraulic conductivity because of animals trampling, 

especially in the topsoil (Wang and Ripley, 1997; Krümmelbein et al., 2006). 

Combined with weakly intercepted water, the infiltration excess runoff was most 

107

likely to occur in the heavily grazed site after strong rainfall events. Secondly, 

although grass growing in the heavily grazed site can experience water stress at 

a smaller water deficit than that in the ungrazed sites because the former 

resulted in an increase of available water holding capacity (AWHC) 

accompanied by increase of meso-pore fraction (Fig. 1). However, grazing 

abruptly decreased soil moisture and thus grass in the grazed sites reached 

water stress point much earlier and easier (Fig. 2), which was in accordance to 

findings by Snyman (2005). Moreover, the root distribution was more shallow in 

HG than the other sites (Gao, 2007), which also contributed to the more 

susceptible plant water stress in the former. These findings were in agreement 

with Rietkerk and van de Koppel (1997), who found that heavy grazing resulted 

in a decrease in soil moisture, and was vulnerable to threshold effects and thus 

fragile. Therefore, HG was more susceptible to drought, and prone to soil 

degradation, especially in the drier years. As shown in Fig. 10, ET differences 

calculated from Pan, FAO, and modeled ones were smaller in the wetter year 

(2004) than that in the medium year (2006) and the drier year (2005). This 

revealed that ET is mainly constrained by biotic factors (LAI) in the wetter years; 

while in the drier years abiotic factors (atmospheric evaporative demand and soil 

moisture condition) are more important. 

108 

In summary, our results clearly reflected that grazing influenced soil physical 

and hydraulic properties. Heavy grazing was in general detrimental to soil 

functions, while moderate to light grazing was less harmful (Gifford and Hawkins, 

1978; Wang and Ripley, 1997; Greenwood and McKenzie, 2001; Martinez and 

Zinck, 2004). Heavy grazing resulted in the dry soil water regime. Particularly, a 

diminished amount of plant available water obviously resulted in a decreased 

rate of plant growth. Reduction of grazing intensity will possibly increase soil 

water household, and consequently improve the system’s productivity within a 

few years (Drewry, 2006). In order to protect and restore degraded soils from 

intense grazing, future land use in Inner Mongolia needs to focus on reducing 

trampling intensity (e.g., rotational grazing), and even animal exclusion.


CONCLUSIONS 

Water and heat fluxes were highly associated with precipitation and solar 

radiation, which was further influenced by grazing via altering soil and vegetation 

conditions, consequently soil hydraulic and thermal properties. Especially, the 

heavy grazing deteriorated soil functions, as indicated by decreased total and 

larger porosity accompanied by decreased saturated hydraulic conductivity and 

increased bulk density. By using multiyear growing season measurements of 

hydro-meteorological and energy elements at the experimental sites of IMGERS, 

model HYDRUS-1D was parameterized and verified with a reasonable 

agreement between simulated and measured data. The simulation could be 

improved by a more precise representation of soil structure (i.e., LDP model). 

The model results showed that soil heat fluxes were increased with increase of 

grazing intensity. In comparison with the two ungrazed sites, winter grazing did 

not show clear effects on the water household, while heavy grazing led to a 

completely different situation: decreased interception and transpiration, and 

increased evaporation, runoff and drainage. As a consequence, intense grazing 

not only reduces the amount of plant available water thus grassland productivity 

but possibly increases soil risks for wind and water erosion. 

ACKNOWLEDGEMENTS 


Council (DFG) for a research grant of the DFG RU #536 MAGIM. Mr. Yujin Wen 

and Mr. Klaus Erdle are acknowledged for the manuscript improvement and 

co-operated field work, respectively. We also thank Prof. J. Šimůnek for his helps 

on the HYDRUS code and Prof. G.N. Flerchinger for his helps on the SHAW 

code. 

REFERENCES 




Bremer, D.J., L.M. Auen, J.M. Ham, and C.E. Owensby. 2001. 

109

110 

Evapotranspiration in a prairie ecosystem: Effects of grazing by cattle. 

Agono. J. 93 (2):338–348. 

Butters, G.L., and P. Duchateau. 2002. Continuous flow method for rapid 

measurement of soil hydraulic properties; I. Experimental considerations. 

Vadose Zone J. 1:239–251. 



Chen, Y., G. Lee, P. Lee, and T. Oikawa. 2007. Model analysis of grazing effect 

on above-ground biomass and above-ground net primary production of a 

Mongolian grassland ecosystem. J. Hydrol. 333 (1):155–164. 

Christensen, L., M.B., Coughenour, J.E., Ellis, and Z.Z. Chen. 2004. Vulnerability 

of the Asian typical steppe to grazing and climate change. Climatic Change 

63: 351–368. 


mulch. Water Resour. Res. 23 (12):2175–2186. 

Coronato, F.R., and M.B. Bertiller. 1996. Precipitation and landscape related 

effects on soil moisture in semi-arid rangelands of Patagonia. J. Arid 

Environ. 34:1–9. 

de Jong, R., and A.Bootsma. 1996. Review of recent developments in soil water 

simulation models. Can. J. Soil Sci. 76:263–273. 

de Vries, D.A. 1963. The thermal properties of soils. P. 210–235. In R.W. van 

Wijk (ed.) Physics of plant environment. North Holland, Amsterdam. 

Drewry, J.J. 2006. Natural recovery of soil physical properties from treading 

damage of pastoral soils in New Zealand and Australia: a review. Agr. 

Ecosyst. Environ. 114:159–169. 

Feddes, R.A., P.J. Kowalik, and H. Zaradny. 1978. Simulation of field water use 

and crop yield, John Wiley and Sons, New York, NY. 

Feddes, R.A., P. Kabat, P.J.T. van Bakel, J.J.B. Bronswijk, and J. Halbertsma. 

1988. Modelling soil water dynamics in the unsaturated zone-State of the art. 

J. Hydrol. 100:69–111.




ASAE 32:565–571. 

Flerchinger, G.N., T.J. Sauer, and R.A. Aiken. 2003. Effects of crop residue cover 

and architecture on heat and water transfer. Geoderma 116:217–233. 

Gao, Y.Z. 2007. Influences of different land use management on net primary 

productivity and belowground carbon allocation in a semi-arid Inner 

Mongolia steppe, PhD Thesis. Kiel University, Germany. 

Gifford, G.F., and R.H. Hawkins. 1978. Hydrologic impact of grazing on infiltration: 

a critical review. Water Resour. Res. 14:305–313. 

Green, T.R., L.R. Ahuja, and J.G. Benjamin. 2003. Advances and challenges in 

predicting agricultural management effects on soil hydraulic properties. 

Geoderma 116:3–27. 

Greenwood, K.L., and B.M. McKenzie. 2001. Grazing effects on soil physical 

properties and the consequences for pastures: a review. Austr. J. Exp. Agr. 

41:1231–1250. 

Hartge, K.H., and R. Horn. 1992. Die physikalische Untersuchung von Böden. 

Ferdinand Enke Verlag Stuttgart. 





Kutilek, M., L. Jendele, and K.P. Panayiotopoulos. 2006. The influence of 

uniaxial compression upon pore size distribution in bi-modal soils. Soil Till. 

Res. 86 (1):27–37. 

Leenhardt, D., M. Voltz., and S. Rambal. 1995. A survey of several agroclimatic 

soil water balance models with reference to their spatial application. Eur. J. 

Agron. 4:1–14. 

Martinez, L.J., and J.A. Zinck. 2004. Temporal variation of soil compaction and 

deterioration of soil quality in pasture areas of Colombian Amazonia. Soil Till. 

111

112 

Res., 75:3–17. 

Nassar, I.N., and R. Horton. 1992. Simultaneous transfer of heat, water, and 

solute in porous media: I. Theoretical development. Soil Sci. Soc. Am. J. 

56:1350–1356. 

Ndiaye, B., J. Molénat, V. Hallaire, C. Gascuel, and Y. Hamon. 2007. Effects of 

agricultural practices on hydraulic properties and water movement in soils 

in Brittany (France). Soil Till. Res. 93:251–263. 

Neitsch, S.L., J.G. Arnold, J.R. Kiniry, J.R. Williams, and K.W. King. 2002. Soil 



Texas. 





Proffitt, A.P.B., R.C. Jarvis, and S. Bendotti. 1995. The impact of sheep trampling 

and stocking rate on the physical properties of a Red Duplex soil with two 

initially different structures. Aust. J. Agric. Res. 46:733–47. 

Richard G., I. Cousin, J.F. Sillon, A. Bruand, and J. Guérif. 2001. Effect of 


properties. Eur. J. Soil Sci. 52:49–58. 

Rietkerk, M., and J. van de Koppel. 1997. Alternate stable states and threshold 

effects in Semi-Arid grazing systems, OIKOS 79:69–76. 

Ritchie, J.T. 1972. Model for predicting evaporation from a row crop with 

incomplete cover. Water Resour. Res. 8(5):1204–1213. 

Saito, H., J. Šimůnek, and B.P. Mohanty. 2006. Numerical analysis of coupled 

water, vapor, and heat transport in the vadose zone. Vadose Zone J. 

5:784–800. 

Schaap, M.G., F.J. Leij, and M.T. van Genuchten. 2001. Rosetta: A computer


program for estimating soil hydraulic parameters with hierarchical 

pedotransfer functions. J. Hydrol. 251:163–176. 






Snyman, H.A. 2005. Rangeland degradation in a semi-arid South Africa—I: 



van Genuchten, M.Th. 1980. A closed-form equation for predicting the hydraulic 

conductivity of unsaturated soils. Soil Sci. Soc. Am. J. 44:892–898. 

van Genuchten, M.Th., F.J. Leij, and S.R. Yates. 1991. The RETC code for 

quantifying the hydraulic functions of unsaturated Soils. Ver. 1.0 EPA Rep. 

600/2-91/065. U.S. Salinity Lab., USDA-ARS, Riverside, CA. 

Villamil, M.B., N.M. Amiotti, and N. peinemann. 2001. Soil degradation related to 

overgrazing in the semi-arid Southern Caldenal area of Argentina. Soil Sci. 

166:441–452. 

Willat, S.T., and D.M. Pullar. 1983. Changes in soil properties under grazed 

pastures. Austr. J. Soil Res. 22:343–348. 

Wang, R.Z., and E.A. Ripley. 1997. Effects of grazing on a Leymus chinensis 

grassland on the Songnen plain, north-eastern China. J. Arid Environ. 

36:307–318. 

Wesseling, J.G. 1991. Meerjarige simulaties van grondwateronttrekking voor 

verschillende bodemprofielen, grondwatertrappen en gewassen met het 

model SWATRE, Report 152, Winand Staring Centre, Wageningen (In 

Dutch). 

Woodward, S.J.R., D.J. Barker, and R.F. Zyskowski. 2001. A practical model for 

predicting soil water deficit in New Zealand. New Zeal. J. Agr. Res. 

44:91–109. 

113

Zhang, Y., E. Munkhtsetseg, T. Kadota, and T. Ohata. 2005. An observational 

114 

study of ecohydrology of a sparse grassland at the edge of the Eurasian 

cryosphere in Mongolia. J. Geophys. Res. 110, D14103. 

doi:10.1029/2004JD005474. 

Zhao, Y., S. Peth, J. Krümmelbein, R. Horn, Z. Wang, M. Steffens, C. Hoffmann, 

and X. Peng. 2007. Spatial variability of soil properties affected by grazing 

intensity in Inner Mongolia grassland. Ecol. Modell. 205:241–254.

Chapter 6 Modeling of Coupled Water and Heat Transfer in Freezing and Thawing Soil 

6. Modeling of Coupled Water and Heat Transfer in 

Freezing and Thawing Soil 

Ying Zhao, Stephan Peth, Jirka Šimůnek, Rainer Horn 

In Preparation for Vadose Zone Journal. 

ABSTRACT 

Accurate simulation of soil freezing and thawing behavior is critical to 

understand hydraulic processes in the vadose zone under cold and arid 

climatic conditions. Using an extended freezing code incorporated in the 

HYDRUS-1D model, this study was conducted 1) to verify the freezing 

model using field soil water and temperature data collected in Inner 

Mongolia grassland, 2) to simulate grazing effects on soil water and heat 

fluxes under both frozen and unfrozen conditions, and 3) to investigate the 

contributions of snowmelt and soil thawing to the seasonal water balance. 

The results showed that both the freezing model and the snow routine 

matched well with the measured soil water and temperature under 

unfrozen conditions. However, under frozen conditions, the freezing model 

substantially improved the simulation results than the snow routine. 

Compared with the two treatments (ungrazed since 1979=UG 79 and winter 

grazing=WG), the freezing model could express well grazing effects on soil 

water and heat fluxes under unfrozen conditions. This confirmed that it 

was able to accurately predict behaviors of soil freezing and thawing, as 

well as the effects of land management. The weak prediction of soil 

moisture in WG might relate with weak parameterization of hydraulic 

properties, e.g., platy structure. In addition, the freezing model did not 

obviously produce surface runoff generated by snowmelt or soil thawing 

from frozen soil layers. Instead, the freezing model overestimated water 

content and thus underestimated surface runoff after spring snowmelt. We 

suggest that a detailed knowledge of the soil-atmosphere processes is 

needed to improve the surface runoff algorithm in the freezing code. 

115

Coupled water, vapor, and heat movement in the vadose zone is a central 

process in many agricultural and engineering issues (Saito et al., 2006). 

Especially, in cold and arid regions, snowmelt or lateral water movement on 

frozen soil layers have a non-negligible influence on seasonal water balances. 

Frozen soil normally reduces infiltration capacity dramatically due to the blocking 

effects of ice lenses in soil pores. Consequently, lateral flow and snow melt may 

release huge quantities of water in spring and early summer, and cause 

substantial surface runoff (Lewkowicz and Kokelj, 2002; Bayard, et al., 2005). 

Furthermore, this will increase the risk for soil erosion and nutrient loss. However, 

although the former processes are recognized widely, the simulations of snow 

hydrology and soil freezing and thawing are rarely done due to limited data 

availability to parameterize or validate such models and lack of suitable models 

that describe the complex processes during phase changes. 

Generally, frozen soils are characterized by the coexistence of ice and water. 

Thus it is necessary to simulate the freezing and thawing process by specifically 

considering the phase change of ice and water at variable freezing points. Over 

the last few decades, two prevalent classes of soil freezing and thawing models, 

namely hydrodynamic models (e.g., Harlan, 1973; Flerchinger and Saxton, 1989) 

and frost heave models (e.g., Miller, 1980) have been developed. They mainly 

differed in the treatment of ice pressure, which is assumed to remain constant in 

hydrodynamic models and variable in the frost heave models (Hansson et al., 

2004). As a result, the former offers accurate predictions under saturated 

conditions but limits to frost heave, which is reverse for the latter. Regarding the 

model’s availability in the unsaturated zone, it is customary to use the 

hydrodynamic model in soil science. 

Generally, water and energy balances are linked at the soil-atmosphere 

interface and controlled by climatic conditions and soil properties. In winter, 

snow cover results in low thermal conductivity and high albedo of soil surface, 

which greatly impacts soil thermal regime and microclimate. Within the soil, 

thermal gradient induces moisture transfer which, in turn, affects heat flow. 

Furthermore, soil heat and water regimes may be modified by different land uses 

116


due to differences in vegetation cover and soil conditions. Both result in altered 

heat and water transfer rates, which are further complicated by snow melt and 

soil freezing and thawing behaviors. Gerasimova et al. (1996) found that soil 

frost might create platy aggregates in loamy soils due to soil freezing and 

thawing processes. Further, a platy soil structure can also be generated by 

animal trampling during grazing in winter (Krümmelbein et al., 2006). Thus, 

lateral water movement on the upper frozen layer is expected to be high during 

soil thawing, as a frozen and platy-structured underlying soil horizon prevents 

infiltration (Lewkowicz and Kokelj, 2002). 

These established facts emphasize the need to better understand snowmelt 

and freezing and thawing processes. Currently, there are several numerical 

codes, e.g., SVAT (Jansson and Halldin, 1980), SHAW (Flerchinger and Saxton, 

1989) and HYDRUS (Šimůnek et al., 1998), developed for simulating coupled 

water and heat flow in frozen soils. However, until now, the mutual interactions of 

water and heat flows in the frozen soil are limited in laboratory observation and 

theoretical analysis, and rarely considered in field applications. The present 

study addresses the field application of the hydrodynamic model HYDRUS-1D. 

In this new program, an extended freezing code is incorporated, which 

numerically solves coupled equations governing phase changes between water 

and ice and heat transport with a mass- and energy-conservative method 

(Hansson et al., 2004). Specifically, we will focus on following questions: (i) How 

well does HYDRUS-1D simulate soil water and temperature with and without 

“frozen soil module”? 2) How does the frozen soil module affect soil temperature, 

soil moisture and runoff simulations? 3) What is model sensitivity to reflect the 

role of land uses in soil water and heat fluxes? 

MATERIALS AND METHODS 

Experimental Site and Measurements 

Fieldwork was conducted at the grassland catchment of Xilin River (43 o 37′N, 

116 o 42′E). Detailed descriptions of the experimental area were given in Zhao et 

al. (2008). The vegetation is a perennial rhizome grass, Leymus chinensis and 

Stipa gradis steppe. The mean annual precipitation is 343 mm. The annual mean 

117

air temperature is 0.7°C, with the highest monthly mean of 19.0°C in July and the 

lowest -21.2°C in January. Soil begins freezing at first November and melting 

around the following April. The investigated soils are classified as Mollisols 

according to USDA (1999). 

Four plots with different grazing intensities were investigated. Two plots 

were protected from grazing since 1979 (denoted as UG 79, 24 ha) and 1999 

(UG 99, 35 ha). The other two plots were grazed: one was grazed only during 

winter time with 0.5 sheep units (ewe and lamb) ha -1 yr -1 (WG, 40 ha) and the 

other was heavily grazed with 2 sheep units ha -1 yr -1 (HG, 100 ha) during the 

whole year. In each plot, three profiles along transect were installed to measure 

soil moisture and temperature from June 2004 to June 2007. Data are recorded 

automatically by a data-logger at 30-min intervals in summer and at 1-h intervals 

in winter, respectively. Soil water content was measured at 5, 20, and 40 cm 

depth by theta-probes (Type ML2x). It refers to volumetric unfrozen water 

content in case the soil is frozen. To describe the detailed course of soil freezing 

and thawing profile, soil temperature was measured in five depths at 2, 8, 20, 40, 

and 100 cm using Platinum ground temperature probes (Pt-100). In 2004, 

undisturbed soil samples were taken at four layers of 4-8, 18-22, 30-34, and 

40-44 cm to determine soil texture, bulk density, total C-content, water retention 

characteristics, and saturated hydraulic conductivity (Zhao et al., 2008). 

From the beginning of the experiment in 2004, an in situ automatic 

micrometeorological station was installed to record the precipitation and other 

variables necessary to estimate the reference FAO evapotranspiration (Allen et 

al., 1998). Snow is treated as snow-water equivalent when air temperature is 

lower than 0°C in the measurement or model. Plant parameters, such as 

vegetation coverage, leaf area index, and plant height were taken from 

vegetation measurements (Gao, 2007). The residue coverage and weight were 

also additionally recorded. Root samples were taken with a soil root auger in five 

depths of 0-10, 10-20, 20-50, 50-70, and 70-100 cm. 

Liquid and Ice Water Flow 

Variably saturated water flow for above- and sub-zero temperatures is 

118


described using the modified Richards equation (e.g., Hansson et al., 2004): 

∂h 

∂T 

∂h 

∂T 

[ ( h) 

+ K ( h) 

+ K ( h) 

+ K ( θ ) + K ( θ ) ] S 

∂θ 

( h) pi 

∂θi 

( T ) ∂ 

∂t 

+ p t = z K 

w ∂ ∂ Lh ∂z 

Lh 

LT ∂z 

vh ∂z 

vT ∂z 

u − 

where θu [L 3 L -3 ] is the volumetric unfrozen water content (=θ+θv), θ is the 

volumetric liquid water content, θv is the volumetric vapor content, θi is the 

volumetric ice content, t is time, z is soil depth, pw and pi is the density of liquid 

and ice water, respectively, h is the pressure head, T is the temperature, and S is 

a sink/source term usually accounting for root water uptake. 

In Eq. 1, the first five terms on the right-hand side represent liquid flows due 

to gradient in pressure head (KLh, [L T -1 ]), gravity, and temperature (KLT, [L 2 T -1 

K -1 ]), and vapor flows due to gradient in pressure head (Kvh) and temperature 

gradients (KvT), respectively. Eq. 1 is nonlinear, namely dependencies of the 

water content and the hydraulic conductivity on the gradient of pressure head 

and temperature, i.e. θ(h), θi(T), KLh(h), and KLT(h). The KLh is described by the 

following expressions (van Genuchten, 1980): 

θ ( h) 

−θ 

r 1 

Se 

( h) 

= = 

n 

θ −θ 

[ 1+ 

αh 

] 

s 

K ( h) 

− 

Lh 

r 

m 

l 

1/ 

m m 2 

= Ks 

× Se 

( 1− 

( 1 Se 

) ) 

[3] 

where Se is the effective saturation, θs and θr are the saturated and residual 

water contents, respectively; the symbols α [L -1 ], n, and m=1-1/n are empirical 

shape parameters, and the inverse of α is often referred to as the air entry value 

or bubbling pressure; Ks is the saturated hydraulic conductivity [L T -1 ], and L is a 

pore connectivity parameter, which normally a value of 0.5 is given. 

The hydraulic conductivity of frozen soil is significantly reduced by ice 

lenses, blocking parts of the pore space. To account for this blocking effect, the 

hydraulic conductivity in our study is reduced by an impedance factor, Ω (Lundin, 

1990), which is multiplied by Q. Namely, the Ω reduces the hydraulic conductivity 

of the partially frozen soil, KfLh, as follows: 

KfLh =10 -ΩQ KLh [4] 

the parameter Q is the ratio of the ice content to the total water content, 

which accounts for the more significant blocking effects with the increase in ice 

[1] 

[2] 

119

content. 

The hydraulic conductivity of temperature driven, KLT is defined as follows 

(e.g., Noborio et al., 1996): 

K 

dγ 

= K hG ) 

[5] 

1 

Lh ( wT γ dT 

LT o 

where GwT is the gain factor (unitless), which quantifies the temperature 

dependence of the soil water retention curve (Nimmo and Miller, 1986), r is the 

surface tension of soil water (= 75.6 - 0.1425T - 2.38×10 -4 T 2 g s -2 ), and r0 is the 

surface tension at 25°C (=71.89 g s -2 ). 

Soil Heat Transport 

The governing equation for the movement of energy in a variably saturated 

rigid porous medium is given by the following conduction–convection heat flow 

equation (e.g., Nassar and Horton, 1992): 

∂T 

∂qlT 

∂qvT 

∂q 

[ λ( 

θ ) ] − C − C + L ( T ) − C ST 

∂C 

pT 

v 

∂t 

− f ∂t 

∂t 

∂z 

∂z 

w ∂z 

v ∂z 

∂z 

∂θ 

i 

∂θv( 

T ) ∂ 

L ρi 

+ L 0( 

T ) = 

[6] 

0 

w 

where L0 is the volumetric latent heat of vaporization of water (L0=Lw × pw, 

Lw is the latent heat of vaporization of water (= 2.501×10 6 - 2369.2 T J kg -1 ), and 

Lf is the latent heat of freezing (approximately 3.34×10 5 J kg -1 ). In Eq. 6, the 

three terms on the left-hand side represent changes in the energy content, the 

latent heat of the frozen and vapor phases, respectively. The terms on the 

right-hand side represent, respectively, soil heat flow by conduction, convection 

of sensible heat with flowing water, transfer of sensitive heat by diffusion of water 

vapor, transfer of latent heat by diffusion of water vapor, and uptake/provision of 

energy associated with the sink/source term. Cp, volumetric heat capacity of the 

bulk soil, is determined as the sum of the volumetric heat capacities including 

solid (Cn), organic (Co), liquid (Cw), ice (Ci), and vapor (Cv) phase multiplied by 

their respective volumetric fractions θ as following (De Vries, 1963): 

Cp ≈ ( C 

θv 

× 

6 

nθ 

n + Coθ 

o + Cwθ 

w + Ciθi 

+ Cv 

) 10 

[7] 

The symbol λ(θ) in Eq. 6 denotes the apparent thermal conductivity and q 

the water flux density while T is the soil temperature. The apparent thermal 

conductivity combines the thermal conductivity of the porous medium in the 

absence of flow and the macro-dispersivity, which is assumed to be a linear 

120


function of velocity and modified as followings: 

λ + 

w L q C β θ λ θ = ) ( ) ( 0 [8] 

λ 0 ( θ ) 

+ i 

0. 

5 

= b1 + b2 

( θ w + Fθi 

) + b3( 

θ w Fθ 

) 

[9] 

where β is the thermal dispersivity (m). The thermal conductivity, λ0(θ), is 

described with a simple equation given by Chung and Horton (1987). The ice 

enhancement factor F (-), compensating for the higher thermal conductivity of 

ice with respect to water, was defined by Hansson et al. (2004) as following: 

F θ 

2 = 1 + F 

[10] 

1 

F 

i 

where F1 and F2 are empirical parameters. 

The phase change between water and ice is controlled by the generalized 

Clapeyron equation, which defines a relationship between the liquid pressure 

head and temperature when ice is present in the porous material. Hence, the 

unfrozen water content can be derived from the liquid pressure head as a 

function of temperature alone when ice and pure water co-exist in the soil. The 

form of the generalized Clapeyron equation was: 

L f 

g 

T h = ln T 

[11] 

0 

where g is the gravitational acceleration [L T -2 ], T is the temperature [K], and 

T0=273.15 K. 

Initial and Boundary Conditions 

Initial conditions were obtained by linear interpolation of pressure heads at 

5, 20, and 40 cm depths measured at the beginning of the simulation period. 

Value-specified boundary conditions were used for the top and bottom boundary 

of the flow domain. At the soil surface, an atmospheric boundary condition was 

imposed accounting for daily data of precipitation, soil surface temperature, 

potential evaporation and transpiration, and minimum allowed pressure head 

(-50000 kPa). Free drainage condition and the measured temperature at 100 cm 

depth were used as bottom boundary, assuming that the water table is located 

far below the domain of interest and that heat transfer across the lower boundary 

occurs only by convection of water and vapor. Soil surface temperature Ts [ 0 C] 

121

was estimated from the soil heat flux G [J m -2 h -1 ] as follows (Chung and Horton, 

1987): 

cm). 

G Ts = + T 

λ ( θ ) z * 

[12] 

where Tz* is the temperature measured at the uppermost soil depth, z* (2 

Root water uptake was simulated using the model of Feddes et al. (1978). 

The maximum rooting depth was considered to be 100 cm, with the highest root 

density in the upper 30 cm. The critical pressure heads in the 

water-stress-response function of Feddes et al. (1978) were design to be similar 

to the ones for grass based on Wesseling (1991), and adjusted for the local 

conditions with a value of -1500 kPa for the permanent wilting point. 

Water and Heat Flow Simulations 

In Zhao et al. (2008), grazing effects on soil hydraulic and thermal 

properties have been parameterized by calibrating the model HYDURS-1D, 

based on data for evaporation ratio (weighing lysimeter experiments), water 

interception (obtained by the model SHAW), and hydraulic parameters 

(laboratory-derived hydraulic properties). However, this work was limited to show 

the effects of grazing on coupled water and heat fluxes during the growing period, 

i.e. unfrozen condition. Here we particularly attend to the snow melt and soil 

freezing and thawing processes. For this aim, the current HYDRUS-1D version 

including soil frozen module (denoted as freezing model) was modified to 

consider the phase changes between water and ice, as well as the surface 

energy and water balances. The algorithms in HYDRUS-1D for the description of 

soil temperature and water movement during the freezing and thawing process 

will be tested and adjusted to the local conditions. To verify the function of 

freezing model, the “normal” version without soil frozen module (snow routine) 

was also run as a reference. In the algorithm of snow hydrology it was assumed 

that all precipitation is in the form of snow when the air temperature (soil surface) 

is below -2°C, and in the form of liquid when the air temperature is above 2°C. 

There is a linear interpolation between -2 and 2°C, i.e. 50% of precipitation is 

122


snow and 50% water for temperature equal to zero (Jarvis, 1989). The soil 

profile was considered to be 100 cm deep, with observation nodes located at 0, 

5, 20, 40, and 100 cm depths. Three soil layers were defined according to 

measured and calibrated layered-soil parameters (Zhao et al., 2008). Model 

efficiency is evaluated in terms of root mean square error (RMSE), which is 

defined as: 

RMSE 

= 

N 

1 2 

N ∑ Pi 

− Oi 

) 

i= 

1 

( [13] 

where N is the number of observations and Pi and Oi are the simulated and 

measured values, respectively. 

RESULTS AND DISCUSSION 

Soil Moisture and Temperature Regimes 

Based on in situ soil water measurements (down to1 m depth), soil water 

storage was calculated for the four sites from August 2005 to July 2006 (Fig. 1). 

Soil water storage (mm) 

140 

120 

100 

80 

60 

40 

20 

0 

9-Sep 19-Oct 28-Nov 7-Jan 16-Feb 28-Mar 7-May 16-Jun 26-Jul 

0 

Rainfall 


60 


Time (days) 

Fig. 6.1. Soil water storage as a function of grazing intensity from August 2005 to July 

2006. 

Liquid water storage is high in summer but low in winter, which coincides with 

rainfall and temperature. Compared with the grazed sites, soil water storage is 

higher in the ungrazed sites, indicating grazing reduced the water stored in soil. 

Even under soil frozen conditions, it is higher in the ungrazed sites than the 

grazed sites, indicating that the volume of unfrozen water is also affected by the 

15 

30 

45 

Rainfall (mm) 

123

water content prior soil freezing. In comparison of the two ungrazed sites, soil 

water storage is identical in summer but different in winter. In winter, higher liquid 

stored water in UG 79 indicates that the long-term ungrazed site is more 

effective in preventing soil from freezing. 

Amplitude ( o C) 


15 

12 

9 

6 

3 

0 

30 

20 

10 

0 

-10 

-20 

WG 

2 cm 8 cm 20 cm 40 cm 100 cm 


9-Sep 19-Oct28-Nov 7-Jan 16-Feb28-Mar 7-May 16-Jun 26-Jul 

Time (days) 

Fig. 6.2. Soil temperature and amplitude in 2, 8, 20, 40, and 100 cm depths in WG 

from August 2005 to July 2006. 

Fig. 2 shows soil temperature and its amplitude as a function of soil depth 

exemplified by the winter grazed site (WG). It is observed that depth variations of 

soil temperature are season-dependent. During the summer time, soil 

temperature decreases with increasing soil depth, indicating energy transfer 

from soil surface to deeper soil layers. On the contrary, during the winter time, 

soil temperature increases with increasing soil depth. In contrast to this, during 

the whole year, the amplitude decreases with increasing soil depth. Especially at 

the 100 cm depth, it keeps constant due to weak energy transfer. Expectedly, the 

amplitude fluctuations are largest in the transition period from freezing to 

thawing, but relatively constant in other periods. In 2005, the topsoil (2 cm) 

began to freeze in 6 th November, i.e. the date from which the daily mean 

temperature became successively negative. With increasing soil depth, the 

freezing process started later. For instance, the soil at 100 cm depth did not 

freeze until 12 th December. Also, the thawing process shows a similar trend with 

depth, i.e. the deeper soil layers begin to thaw later. This is consistent with the 

124


duration of ground freezing (in average about 140 to 150 days), which decreases 

with increasing soil depth (Table 1). 

Amplitude ( o C) 


15 

12 

9 

6 

3 

0 

30 

20 

10 

0 

-10 

-20 

2 cm 

UG 99 WG 


9-Sep 19-Oct28-Nov 7-Jan 16-Feb28-Mar 7-May 16-Jun 26-Jul 

Time (days) 

Fig. 6.3. Soil temperature and amplitude in topsoil (2 cm) as a function of grazing 

intensity (UG 99 and WG) from August 2005 to July 2006. 

Furthermore, soil temperature and amplitude are affected by grazing 

intensity, which is exemplarily shown for the ungrazed site UG 99 and the winter 

grazed site WG at the topsoil (2 cm) (Fig. 3). Compared with UG 99, WG has a 

higher temperature in summer but a lower value in winter, indicating that soil 

thermal properties are not only seasonal specific but also treatment-dependent. 

We observe that the monthly mean of soil temperature decreases in winter and 

increases in summer with increasing grazing intensity (Table 1). It can be 

ascribed to the differences in insulating effect of vegetation cover. The dense 

vegetation at the ungrazed sites (Gao, 2007) protects soil from receiving strong 

radiation in summer and from releasing energy quickly in winter. Consequently, 

the temperature amplitude happened higher in the grazed sites than the 

ungrazed sites (Fig. 3), indicating that grazing decreases the thermal diffusivity 

by increasing thermal conductivity in the soil surface. 

125

126 

40 20.0 15.4 8.1 -0.7 -8.7 -11.2 -10.5 -5.6 1.7 8.9 14.8 18.6 -5.8 144 

100 17.3 13.7 9.9 4.4 -1.5 -5.0 -6.4 -4.7 -1.1 2.9 9.0 13.3 -2.4 141 

HG 

20 21.1 15.0 6.3 -3.8 -12.4 -14.1 -12.5 -6.0 2.8 11.4 17.2 20.3 -7.7 146 

2 22.1 15.9 4.6 -6.6 -15.8 -16.4 -13.8 -5.8 3.9 15.0 19.0 21.5 -9.1 150 

8 21.7 15.4 5.2 -5.7 -14.6 -15.7 -13.5 -5.9 3.6 13.6 19.5 21.7 -8.6 147 

100 14.0 13.4 9.8 4.7 -1.0 -4.7 -5.5 -3.8 -0.8 3.4 8.1 12.0 -1.9 136 

40 19.2 15.1 7.2 -1.6 -9.6 -12.1 -10.7 -5.3 2.1 9.9 15 18.3 -6.2 144 

WG 

20 19.8 15.2 6.8 -2.5 -11.1 -13.2 -11.4 -5.3 2.5 10.9 15.9 19.0 -6.8 145 

2 21.7 16.0 4.9 -5.5 -14.8 -15.5 -12.6 -4.5 4.7 15.9 19.0 21.1 -8.0 143 

8 20.9 15.3 4.9 -5.2 -14.3 -15.3 -12.8 -5.2 3.8 14.0 17.8 20.4 -8.2 146 

100 13.0 12.3 8.7 3.0 -1.2 -4.3 -5.2 -4.5 -1.6 2.9 7.7 11.6 -2.3 141 

40 16.8 13.5 6.9 -0.7 -6.1 -8.5 -8.2 -5.4 0 7.3 12.8 16.3 -4.8 157 

UG 99 

20 18.3 13.8 5.8 -3.1 -9.9 -11.4 -10.3 -5.9 1.1 9.6 14.9 17.9 -6.6 157 

2 19.5 14.4 4.7 -4.7 -12.7 -13.2 -11.6 -5.4 2.6 12.4 16.5 19.1 -7.5 154 

8 19.1 14.1 5.0 -3.9 -10.4 -11.4 -10.2 -5.2 2.2 11.4 16 18.8 -6.5 150 

100 15.0 14.1 10.2 4.6 -1.0 -4.2 -5.2 -4.0 -0.9 4.3 9.4 13.1 -1.8 136 

40 18.4 14.9 8.0 -0.1 -6.9 -9.4 -8.9 -5.0 0.9 8.4 13.9 17.1 -4.9 148 

UG 79 

20 20.0 15.2 6.7 -2.5 -9.8 -11.5 -10.3 -4.9 2.4 11.2 16.1 19.0 -6.1 145 

8 21.1 15.7 5.7 -4.3 -11.8 -12.9 -11.2 -4.6 3.5 13.8 17.6 20.2 -6.9 146 

2 21.9 16.1 5.1 -5.3 -12.9 -13.6 -11.5 -4.3 4.2 15.5 18.5 20.8 -7.2 144 

Treat- Depth Month (August 2005 – July 2006) Average Period of ground 

ment (cm) 8 9 10 11 12 1 2 3 4 5 6 7 in winter freezing (day) 

Table 6.1. Monthly-averaged soil temperatures at 2, 8, 20, 40, and 100 cm depth in four sites from August 2005 to July 2006.


Simulated Soil Temperatures and Water Contents 

Soil water and heat fluxes are numerically simulated for the whole 

hydrological year in 2006. This is exemplified by two sites of UG 79 (Fig. 4) and 

WG (Fig. 5). In general, the simulated and measured soil water contents (SWC) 

are comparable during the studied periods (Figs. 4-5; RMSE of 0.02-0.07 cm 3 

cm -3 for UG 79 and 0.02-0.08 cm 3 cm -3 for WG). Particularly, the freezing model 

simulates well the diurnal water dynamics caused by phase changes between 

water and ice, i.e. soil moisture increases with increasing soil temperature and 

vice versa (Figs. 4b and 5b). This suggests the freezing and thawing processes 

can be accurately simulated by the applied freezing code. However, the snow 

routine is only available for the simulation of soil water contents under unfrozen 

condition. Under frozen condition, there is a clear disparity between the liquid 

water content line simulated by the freezing model and the total water content 

line simulated by the snow routine (Fig. 4a and 5a). This discrepancy is 

apparently caused by the function of the two models, and may be used to 

approximate the ice content in the soil. Thus, the comparison between the two 

model approaches just provides a way to calculate the ice content. For instance, 

the SWC simulated by the freezing model drops shortly after 6 th November 

associated with soil freezing in UG 79 (Fig. 4a). However, the SWC simulated by 

snow routine keeps constant. From which, an ice content of 0.07 cm 3 cm -3 can 

be estimated. Similarly, the ice content in WG is 0.05 cm 3 cm -3 (Fig. 5a). This 

also gives an evidence of grazing effects on soil freezing point and ice content. 

In late March, coinciding with soil temperature increasing above 0°C, both 

simulated (freezing model) and measured SWC raise, while simulated SWC by 

snow routine keeps constant. This again proofs that the freezing model can 

predict well the increase in SWC by soil thawing. An increase in SWC in the 

deeper soil layers (20 and 40 cm) accompanied by soil thawing is also clearly 

predicted (Figs. 4-5). However, there is no indication of vertical water movement 

since soil water content is low and it can be held by soil. 

127


Soil moisture(cm 3 cm -3 ) 





30 

20 

10 

0 

-10 

-20 

a 

0.00 

0.05 

0.10 

0.15 

0.20 

0.25 

0.30 

0.35 

20 

10 

0 

-10 

-20 

0.20 

0.15 

0.10 

0.05 

0.00 

20 

10 

0 

-10 

-20 

0.20 

0.15 

0.10 

0.05 

0.00 

b 

c 

9-Feb 21-Mar 30-Apr 9-Jun 19-Jul 28-Aug 7-Oct 16-Nov 26-Dec 


M S F Rainfall 40 

0 

9-Feb 21-Mar 30-Apr 9-Jun 19-Jul 28-Aug 7-Oct 16-Nov26-Dec 

Time (days) 

20-Mar 25-Mar 31-Mar 5-Apr 11-Apr 16-Apr 21-Apr 27-Apr 

M S F 


Time (hours) 

2 4 6 8 10 12 14 16 18 20 22 24 

M S F 

2 4 6 8 10 12 14 16 18 20 22 24 

Time (hours) 

20 

Rainfall (mm) 





30 

20 

10 

0 

-10 

-20 

0.30 

0.35 

20 cm 

0.00 

0.05 

0.10 

0.15 

0.20 

0.25 

30 

20 

10 

0 

-10 

-20 

40 cm 

0.05 

0.00 

0.10 

0.15 

0.20 

0.25 

0.30 

0.35 

30 

20 

10 

0 

-10 

-20 

0.05 

0.00 

0.10 

0.20 

0.15 

0.25 

0.35 

0.30 



M S F Rainfall 

40 

0 


Time (days) 



M S F Rainfall 

40 

0 


Time (days) 



0 


Time (days) 

20 

20 

S F Rainfall 

40 

Fig. 6.4. Measured and simulated soil moisture and temperature at 5 cm (a, b, c), 20, 40, 

and 100 cm depth in UG 79 during the whole year of 2006 (M: Measured liquid water 

content; S: Simulated total water content running snow routine; and F: Simulated liquid 

water content running freezing model). 



In contrast to the soil water simulations, soil temperature is simulated well 

by both models with minor differences (Figs. 4-5). However, regarding the 

detailed diurnal variations of soil temperature (Figs. 4c and 5c), the freezing 

model matches measured values better than the snow routine. This gives an 

evidence of the impact of the function of the frozen soil module on soil 

temperature simulations. In a real world, when soil becomes freezing (i.e. the 

phase change from water to ice), soil temperature decreases and energy is 

released which is used to warm up soil from a cold. However, given the same 

total water content, the thermal conductivity of frozen soil is higher than that of 

128 

100 cm 

20 

Rainfall (mm) 

Rainfall (mm) 

Rainfall (mm)


unfrozen soil due to present of ice. Consequently, the upward soil heat flux 

becomes higher when the soil gets frozen, and thus it tends to cool the soil 

(Smirnova et al. 2000). But the effect of thermal conductivity is probably smaller 

than that of water phase change. Consequently, the freezing model, which 

considers the both processes, provides a more reasonably and realistic 

simulations of soil temperature than the snow routine. 







30 

20 

10 

0 

-10 

-20 

a 

0.05 

0.00 

0.10 

0.15 

0.20 

0.25 

0.30 

0.35 

20 

10 

0 

-10 

-20 

0.20 

0.15 

0.10 

0.05 

0.00 

20 

10 

0 

-10 

-20 

0.20 

0.15 

0.10 

0.05 

0.00 

b 

c 




0 


Time (days) 


M S F 


Time (hours) 

2 4 6 8 10 12 14 16 18 20 22 24 

M S F 

2 4 6 8 10 12 14 16 18 20 22 24 

Time (hours) 

20 

Rainfall (mm) 






Soil moisture (cm 3 cm -3 ) 

30 

20 

10 

0 

-10 

-20 

0.30 

0.35 

20 cm 

0.05 

0.00 

0.10 

0.15 

0.20 

0.25 

30 

20 

10 

0 

-10 

-20 

40 cm 

0.00 

0.05 

0.10 

0.15 

0.20 

0.25 

0.30 

0.35 

30 

20 

10 

0 

-10 

-20 

0.05 

0.00 

0.10 

0.15 

0.20 

0.25 

0.30 

0.35 




0 


Time (days) 




0 


Time (days) 

100 cm 



0 


Time (days) 

20 

20 

S F Rainfall 40 

Fig. 6.5. Measured and simulated soil moisture and temperature at 5 cm (a, b, c), 20, 40, 

and 100 cm depth in WG during the whole year of 2006 (M: Measured liquid water 

content; S: Simulated total water content running snow routine; and F: Simulated liquid 

water content running freezing model). 

Hydrological Effect of Frozen Soil 

The temperature threshold at which precipitation is considered to fall as snow 

is specified as -2°C in the model, suggesting that precipitation in winter is falling 

20 

Rainfall (mm) 

Rainfall (mm) 

Rainfall (mm) 

129

as snow. The snow routine predicts up to 15 mm snow depth (Fig. 6a), however, 

the simulated runoff after air temperature increasing above 0°C is negligible (Fig. 

6c). This might relate with that the snow routine does not account for surface 

runoff from the frozen soil layer since the code cannot consider the subsurface 

soil freezing and thawing process. In fact, it is likely that surface runoff is 

generated during snowmelt while soil is fully or at least partially frozen (Fig. 6b). 

Unexpectedly, the freezing model, which can account for the subsurface freezing 

and thawing processes, also does not compute surface runoff during winter (Fig. 

2c). Instead, we found that the simulated SWC by freezing model is higher than 

the measured values in the transition time when soil begins to thaw (Figs. 4a and 

5a). This implies that the freezing model might overestimate water content and 

underestimate surface runoff after spring snowmelt. Therefore, the freezing 

model seems still not sensitive enough to estimate surface runoff accompanied 

with snowmelt from the soil frozen layer. This might relate to the fact that the 

freezing model we applied adopts soil surface temperature as the atmospheric 

boundary condition instead of air temperature, which neglects the lag-effects of 

energy transfer. Consequently, the freezing model may incorrectly partition all 

the snowmelt into infiltration as both soil thawing and snow melting happen 

simultaneously. Therefore, to solve this, a transferable and double-layered 

boundary condition (e.g., one accounting for air temperature and other 

accounting for considering soil temperature) should be introduced. Additionally, 

the gradual release of water from the frozen soil profile also might reduce the 

maximum rate of runoff. 

130 

In contrast to the freezing model, an ‘‘implicit’’ frozen soil module that used 

in other studies might totally stop water infiltration inside the soil (Mitchell and 

Warrilow, 1987) by specifying that the meltwater has to run off when snow melts 

while soil is still frozen. However, the soil ice is subsequently melting when the 

snow is melting (Luo et al., 2003). Thus the frozen soil layer moves downward 

and leaves the upper soil layer available for infiltration. In contrast to this, an 

‘‘explicit’’ frozen soil module like the freezing model that we used is theoretically 

reasonable as it only reduces the infiltration rate based on soil temperature, soil


ice content, and soil hydraulic properties (Luo et al., 2003). However, the 

reduction in infiltration capacity owing to the blocking effects of ice is possibly 

underestimated by the current freezing model. 

Snow depth (mm) 

Temperature ( o C) 

Runoff (mm) 

16 

14 

4 

2 

0 

30 

20 

10 

0 

-10 

-20 

-30 

0.24 

0.20 

0.16 

0.12 

0.08 

0.04 

0.00 

a 

b 

c 


Air 

Soil 


Snow routine Freezing model 

9-Feb 21-Mar30-Apr 9-Jun 19-Jul28-Aug 7-Oct 16-Nov26-Dec 

Time (days) 

Rainfall 

Fig. 6.6. Rainfall, measured air and soil temperature, simulated snow depth and 

runoff during the whole year of 2006. 

60 

40 

20 

0 

Rainfall (mm) 

Frozen soil was expected to have a great influence on the runoff simulation 

(Mitchell and Warrilow, 1987; Pitman et al., 1999). But in agreement with our 

study, Pitman et al. (1999) did not obtain the improvement of runoff simulation 

when soil ice was included in their model. Cherkauer and Lettenmaier (1999) 

also found that the frozen soil module had a relatively small effect on runoff. 

131

They explained that the meltwater percolating could occur on the unfrozen 

topsoil even though snow was present. In fact, Hortonian surface runoff can 

always find its way to infiltrate somewhere over a large region because soil might 

be never frozen homogeneously considering spatial variability of soil properties. 

Although the current freezing soil module has little effect on the simulations of 

surface runoff, we expect a detailed study of the soil-atmosphere processes and 

effects of boundary conditions to improve the surface runoff algorithm in the 

freezing code. 

132 

Simulation of Grazing Effects on Soil Freezing and Thawing 

In addition to the effect of a frozen soil layer on the water and heat transfer 

rates, land management also plays a key role in modifying soil hydraulic and 

thermal parameters through changing soil structure (Hillel, 1998). Freezing 

normally reduces the permeability of soils owing to the impeding effect of ice 

lenses as well as structural changes. In the grazed sites with a poor-structured 

soil (Zhao et al., 2008), freezing often leads to a higher level of aggregation 

induced by dehydration and the pressure of ice crystals. In contrast to this, in the 

ungrazed sites with a well-structured soil, expansion of the freezing soil may 

cause the aggregates to break down. 

In Zhao et al. (2008), soil hydraulic and thermal parameters were proofed to 

be a function of grazing intensity under unfrozen conditions. For instance, the 

van Genuchten hydraulic parameters θs, α, and Ks are smaller at the grazed 

sites than those at the ungrazed sites due to the grazing-induced soil 

compaction. However, it is possible that model parameterizations regarding 

grazing effects differ between frozen and unfrozen conditions. As for the change 

of hydraulic parameters under frozen condition, the blocking effect of hydraulic 

conductivity is considered (Eq. 4). However, we can not modify water retention 

parameters under frozen conditions based on the limited knowledge on how to 

measure and finally parameterize this effect. It is known that soil heat capacity is 

not sensitive to the modeling result given that it is straightforward calculated from 

the material constituents and its volumetric heat capacity. Therefore, our method 

is assumed as accurate parameterization of heat capacity. That is, soil heat


capacity decreases with increasing grazing intensity due to high water content at 

the ungrazed sites and similar soil textural classes for each site (Eqs. 7-10). 

Thermal conductivity, which shows a non-linear dependency on, e.g., water 

content and geometrical arrangement of particles, has also been included in the 

current freezing model (e.g., Eq. 9). 

Comparing the two treatments of UG 79 and WG, soil water and heat fluxes 

are clearly affected by grazing management. The relatively weak prediction in 

WG might ascribe to the weak parameterization of hydraulic parameters, which 

is associated with a platy structure created by the growth of ice segregation 

blades due to winter grazing (Krümmelbein et al., 2006). Furthermore, 

desiccation also results in the formation of vertical shrinkage fissures and a 

prismatic structure in deeper soil depths in WG. This can be observed by that a 

decrease in soil moisture at 20 cm depth is slower in WG than in UG 79 because 

of relatively low evaporation by the less continued pore system in WG (Figs. 4-5). 

The decrease in soil moisture at 5 cm depth after rainfall is slower in WG than in 

UG 79 since water supplied to the topsoil layer lasts longer in WG due to a lower 

hydraulic conductivity (Zhao et al., 2008). However, the constant water contents 

at 40 cm depth in both sites show that residual water content is reached and 

maintained in the deeper layers due to slight influences either by water 

infiltration or by soil evaporation. Moreover, soil water dynamics in WG is less 

responsive to freezing and thawing processes than in UG 79. Due to the lower 

unsaturated hydraulic conductivity in WG, there is much less moisture migration 

to the freezing front than in UG 79. Consequently, water content increases much 

smaller in WG than in UG 79 (Figs. 4-5). 

CONCLUSION 

We used an extended frozen soil module of HYDRUS-1D to govern coupled 

flow equation which solves water and heat transport under both frozen and 

unfrozen conditions simultaneously. The model was evaluated using field data of 

soil water and temperature at a long-term experimental site in Inner Mongolia 

grassland (North China). The results showed that both freezing model and snow 

routine reflected well the measured soil water and temperature under unfrozen 

133

condition, whereas the freezing model substantially improved the simulation 

results under frozen condition. Compared with two treatments (UG 79 and WG), 

the freezing model could express well grazing effects on soil water and heat 

fluxes under unfrozen conditions. This confirmed the freezing model could 

predict the behavior of soil freezing and thawing, as well as the effects of land 

management. The weak prediction of soil moisture in WG might relate with weak 

parameterization of hydraulic properties, e.g. platy structure. In addition, the 

freezing model did not obviously produce surface runoff generated by snowmelt 

or soil thawing from frozen soil layer. We suggest that seasonal water balance, 

especially considering rainfall water stored as snow, snow drift and the lateral 

water flow on frozen soil layers need to be investigated further because of the 

complicated interactions at the soil-atmosphere interface and thus effects of 

boundary conditions on the simulation. 

134 

ACKNOWLEDGEMENTS 


Council (DFG) for a research grant of the DFG RU #536 MAGIM. We thank Prof. 

Dr. Xinguo Han, Prof. Dr. Yongfei Bai and the Institute of Botany (Chinese 

Academy of Sciences) for the opportunity to work at IMGERS. 

REFERENCES 




Bayard, D., Stahli, M., Parriaux, A., Fluhler, H. 2005. The influence of seasonally 

frozen soil on the snowmelt runoff at two Alpine sites in southern 

Switzerland. J. Hydrol. 309:66-84. 

Cherkauer, K. A., and D. P. Lettenmaier. 1999. Hydrologic effects of frozen soils 

in the upper Mississippi River basin. J. Geophys.Res. 104:19599–19610. 


mulch. Water Resour. Res. 23(12):2175–2186. 

de Vries, D.A. 1963. The thermal properties of soils. p. 210–235. In R.W. van


Wijk (ed.) Physics of plant environment. North Holland, Amsterdam. 

Feddes, R.A., P.J. Kowalik, and H. Zaradny. 1978. Simulation of Field Water Use 

and Crop Yield, John Wiley and Sons, New York, NY. 



ASAE 32:565–571. 

Gao, Y.Z., 2007. Influences of different land use management on net primary 

productivity and belowground carbon allocation in a semi-arid Inner 

Mongolia steppe, PhD Thesis. Kiel University, Germany. 

Gerasimova, M.I., S.V. Gubin, and S.A. Shoba. 1996. Soils of Russia and 

Adjacent Countries: Geography and Micromorphology. Moscow State 

University—Wageningen Agricultural University, p. 204. 

Harlan, R.L., 1973. Analysis of coupled heat-fluid transport in partially frozen soil. 

Water Resour. Res. 9:1314–1323. 

Hansson, K., J., Šimůnek, M. Mizoguchi, L.-C., Lundin, and M.T., van Genuchten. 

2004. Water flow and heat transport in frozen soil: numerical solution and 

freeze-thaw applications. Vadose Zone J. 3:693–704. 

Hillel, D., 1998. Environmental Soil Physics. Academic Press, 771 pp. 

Jansson, P.-E., and S. Halldin. 1980. SOIL water and heat model: Technical 

description. Swedish Coniferous Forest Proj. Tech. Rep. 26. Swedish 

University of Agricultural Sciences, Uppsala, Sweden. 

Kane, D.L. 1980. Snowmelt infiltration into seasonally frozen soils. Cold Regions 

Science and Technology, 3, 153-161. 

Kay, B.D., and P.H. Groenevelt. 1974. On the interaction of water and heat 

transport in frozen and unfrozen soils: I. Basic theory; the vapor phase. Soil 

Sci. Soc. Am. Proc. 38:395–400. 

Koopmans, R.W.R., and R.D. Miller. 1966. Soil freezing and soil water 

characteristic curves. Soil Sci. Soc. Am. J. 30:680–685. 

135


136 




Lewkowicz, A.G., and S. V. Kokelj. 2002. Slope sediment yield in. arid lowland 

continuous permafrost environments, Canadian Arctic. Archipelago. 

Lundin, L.-C. 1990. Hydraulic properties in an operational model of frozen soil. J. 

Hydrol. 118:289–310. 

Luo, L.F., A. Robock, K.Y. Vinnikov, C. Adam Schlosser, A.G. Slater, et al.. 2003. 

Effects of Frozen Soil on Soil Temperature, Spring Infiltration, and Runoff: 

Results from the PILPS 2(d) Experiment at Valdai, Russia. Journal of 

Hydrometeorology 4:334–351. 

Miller, R.D., 1980. Freezing phenomena in soils. In: Hillel, D. (Ed.), Applications 

of Soil Physics. Academic Press, pp. 254– 299. 

Mitchell, J.F.B., and D.A. Warrilow. 1987. Summer dryness in northern 

mid-latitudes due to increased CO2. Nature 341:132–134. 

Nassar, I.N., and R. Horton. 1992. Simultaneous transfer of heat, water, and 

solute in porous media: I. Theoretical development. Soil Sci. Soc. Am. J. 

56:1350–1356. 

Nimmo, J.R., and E.E. Miller. 1986. The temperature dependence of isothermal 

moisture vs. potential characteristics of soils. Soil Sci. Soc. Am. J. 

50:1105–1113. 

Noborio, K., K.J. McInnes, and J.L. Heilman. 1996. Two-dimensional model for 

water, heat and solute transport in furrow-irrigated soil: I. Theory. Soil Sci. 

Soc. Am. J. 60:1001–1009. 

Pitman, A.J., A.G. Slater, C.E. Desborough, and M. Zhao. 1999: Uncertainty in 

the simulation of runoff due to the parameterization of frozen soil moisture 

using the Global Soil Wetness Project methodology. J. Geophys. Res. 

104:16879–16888.


Saito, H., J. Šimůnek, and B.P. Mohanty. 2006. Numerical analysis of coupled 

water, vapor, and heat transport in the vadose zone. Vadose Zone J. 

5:784–800. 

Smirnova, T.G., J.M. Brown, and S.G. Benjamin. 1997. Performance of different 

soil model configurations in simulating ground surface temperature and 

surface fluxes. Mon. Wea. Rev. 125:1870–1884. 






van Genuchten, M.Th. 1980. A closed-form equation for predicting the hydraulic 

conductivity of unsaturated soils. Soil Sci. Soc. Am. J. 44:892–898. 

Wesseling, J.G., 1991. Meerjarige simulaties van grondwateronttrekking voor 

verschillende bodemprofielen, grondwatertrappen en gewassen met het 

model SWATRE, Report 152, Winand Staring Centre, Wageningen (In 

Dutch). 

Zhao, Y., S. Peth, J. Krümmelbein, B. Ketzer, Y.Z. Gao, X.H. Peng, R. Horn, and 

C. Bernhofer. 2008. Modelling grazing effects on coupled water and heat 

fluxes in Inner Mongolia grassland. (Submitted to Vadose Zone Journal). 

137

138

Chapter 7 General discussion and conclusions 

7. General discussion and conclusions 

In Inner Mongolia grassland, grazing-induced changes in the productivity and 

stability of grassland ecosystems have been reported recently. However, the 

environmental impact of grazing and especially the role of different grazing 

intensities are not well be understood. This study therefore concentrates on the 

effect of grazing on soils, plant ecosystem and surface fluxes at multiple scales 

in general, and on modeling grazing effects on coupled water and heat fluxes at 

the plot scale in particular. 

Spatio-temporal variability of water-related variables affected by grazing 

intensity (Chapters 2, 3 and 4) 

The characterization of the spatial variability of water-related variables is 

essential to achieve a better understanding of eco-hydrological processes 

accompanied by environmental changes. In agreement with other studies (e.g. 

Shouse et al., 1995; Lophaven et al., 2006), we also found that soil properties 

exhibited a moderate to strong spatial dependence. Furthermore, this spatial 

dependence was modified by grazing management, e.g. heavy grazing resulted 

in a more homogenous spatial distribution of soil properties. We partly attributed 

it to soil compaction effects accompanied by animal trampling. Multiple 

regression analysis showed significant correlations among various soil variables, 

this is in agreement with our assumption that regionalized variables are 

correlated each other. Especially, multivariate geostatistical analysis further 

revealed a scale-dependent correlation between controlling parameters affected 

by grazing intensity. To our knowledge, there are few studies describing the 

interactions of spatial variability of soil properties affected by land management 

at multiple spatial scales (Williams et al., 2003). Recently, Casa and Castrignanò 

(2007) applied factorial kriging analysis to clarify the spatial relationships 

between the different variables acting at the different scales. Our studies are in 

agreement with it, we considered this approach will be helpful to explain the 

139

mechanisms of environmental changes in response to grazing, which is the 

basis for making a more sustainable land management. 

140 

Although a large amount of data is required for this kind of analysis we 

consider it is a prerequisite when matter fluxes on a regional scale are 

investigated. Spatio-temporal distribution of water-related variables is valuable 

information not only for deriving a conceptual understanding of landscape fluxes 

but also for the spatial discretization and parameter estimation in the modeled 

domain. For instance, the analysis of correlation length is very useful to 

determine appropriate model element sizes, and to extrapolate and upscale 

processes with the aid of physically-based hydrological models from plot (e.g. 

Hydrus-1D) to catena (e.g. Hydrus-2D) and finally to a regional scale (e.g. SWAT) 

(Western et al. 1999). Currently, the estimation of water fluxes at different spatial 

scales remains a challenging task in hydrology. Indeed, the scale of the 

hydrological measurement technique is generally much smaller than the scale at 

which the predictions are required. The knowledge of scaling (regionalization, 

transferability and upscaling) is therefore needed to estimate water fluxes at 

large scales based on a series of local measurements. We expect that further 

studies will have to concentrate on questions like what size model elements 

should have at different spatial scales. 

In addition, except for the spatial dependence of water-related variables 

aforementioned, time stability analysis of soil moisture is also important. In fact, 

due to the high costs of long-term soil moisture monitoring, it is rare that the 

monitoring sites are uniformly distributed in the entire studied area 

(Gomez-Plaza et al., 2000; Martinez-Fernandez and Ceballos, 2005; Lin, 2006). 

Consequently, selected monitoring sites may not represent the true field mean 

water content. In addition, in the study of hydraulic model, normally the 

observation or modeled points are selected without the prior analysis so that 

representative of the selected points is also uncertain. To account for this 

uncertainty, we combined the temporal stability concept with a hydraulic model 

(HYDRUS-1D) applied in the derived time stability point (TSP). Helpfully, we 

proofed it to be a suitable method to reduce the sampling number of soil


moisture. Conversely, our monitoring position can be considered as kind of 

location that can represent the field mean moisture content (i.e. representative 

position) since it also matched with measured water content in TSP. From this 

aspect, we set up a bridge connected the representative of monitoring and field 

mean value in term of temporal stability concept. 

Grazing effects on soil properties and functions (Chapters 2, 3 and 5) 

Grazing has an influence on soil hydraulic, thermal and mechanical properties 

and functions, predominantly in the upper layers (10-15 cm). Furthermore, those 

properties are interlinked with each other. For instance, the mechanical results 

revealed that grazing increased the precompression stress on the grazed sites 

(Krümmelbein et al., 2006), and further increased bulk density accompanied with 

changes in soil structure as indicated by Peth and Horn (2006). The structural 

change due to grazing is also reflected by a decrease of Ks, total- and macro- 

pore volume and an increase of meso-pore volume. This is in agreement with 

Villamil et al. (2001), who proofed a change of water retention characteristics by 

grazing in Argentina. Furthermore, grazing decreased hydraulic conductivity 

because of animals trampling as indicated by Wang and Ripley (1997). 

Therefore, we considered that heavy grazing particularly deteriorated soil 

physical and hydraulic properties, while moderate to light grazing was less 

harmful. 

Conversely, we also noticed that soil physical functions had recovered to 

some extent after being fenced for a long-term period (e.g. 25 yr in UG 79). 

Drewry et al. (2006) summarized that processes contributing to the natural 

recovery of physically degraded soil might include reduction of soil compaction 

(e.g. tillage, wheel-traffic compaction), earthworm burrowing, root penetration 

and decay, wetting and drying cycles, and freezing and thawing cycles (Drewry 

et al., 2006). Owing to exclusion of animal trampling, we mainly ascribed the soil 

recovery in the ungrazed sites to the regeneration of soil structure by root 

penetration and decay, wetting and drying cycles, and freezing and thawing 

cycles. In agreement with Proffitt et al. (1995), our results proofed that reduction 

141

of grazing intensity could improve soil functions again under pasture. In order to 

protect and restore degraded soils from intense grazing, we suggest future land 

use in Inner Mongolia needs to focus on reducing trampling intensity and animal 

exclusion. 

Modeling grazing effects on coupled water and heat fluxes (Chapters 4, 5 

and 6) 

142 

Grazing-induced changes in soil properties were parameterized by the 

coupled water and heat model HYDRUS-1D (Šimůnek et al., 1998). In addition, 

we introduced the different root densities and boundary conditions for the 

different grazing intensities. Especially, three improvements is needed to 

underline when we calibrated the model. Firstly, we worked with the root growth 

model to consider dynamics of plant water uptake. We considered it important 

because models designed to simulate agricultural managements are normally 

limited to simulate crop growing thus keep root constant, e.g. SWAT (Neitsch et 

al., 2002). Secondly, interception, which can not be calculated by HYDRUS-1D, 

was estimated by the model SHAW (Flerchinger and Saxton, 1989) since it 

might occupy a large component of the water budget in our case. Finally, we 

partitioned evapotranspiration (ET) based on the field measurements, which was 

particularly essential for the prediction of plant transpiration since it is only 

determined by root water extraction function related with potential transpiration 

in HYDRUS code. The modifications in boundary conditions resulted in a 

significant improvement of the simulation accuracy. 

We showed that HYDRUS-1D model is capable to reflect the grazing effects 

on water and heat budgets, and it can be used for the analysis of land 

management scenarios. Apart from the inverse model, the Laboratory-derived 

hydraulic parameter (LDP) model showed the best match between simulated 

and measured water contents as it accounts for soil structural changes resulted 

from grazing. This result is consistent with the description by Richard et al. 

(2001), who distinguished textural and structural fractions for the soil pore space 

and found that soil compaction mainly affected the structural pore space. Our


results therefore suggest that the detailed laboratory measurements of soil 

hydraulic properties from undisturbed soil sampling are necessary to reflect the 

effect of land managements on water flux. 

Under the prevailing semi-arid climatic conditions in Inner Mongolia, plant 

available water plays the most key role for the sustainable development of 

steppe ecosystems. At this moment, some studies have shown that grazing 

affects water budget (Bremer et al., 2001). However, to which extent grazing 

affects evapotranspiration, and how far it is partitioned into transpiration and 

evaporation is unclear. We proofed that the water budget in Inner Mongolia 

grassland is significantly influenced by grazing. Although there was no apparent 

difference in evapotranspiration (ET) among different grazing intensities, the 

components of ET, i.e. interception, transpiration and evaporation significantly 

varied with grazing intensity. It is obvious that, compared with the grazed sites, 

interception and transpiration increased and soil evaporation decreased in the 

ungrazed sites. In contrast to this, evaporation in the grazed sites simultaneously 

increased. We deem it important information for judgments of water use 

efficiency in this region. 

Currently, although snowmelt or lateral soil water movement on frozen soil 

layers is recognized as an important part for the seasonal water balance, 

simulations of snow hydrology and soil freezing and thawing are rarely done due 

to a lack of suitable models that describe the complex processes during phase 

changes and limited data availability to parameterize or validate such models 

(Hansson et al., 2004). Especially, the mutual interactions of water and heat 

flows in frozen soil are limited to laboratory observation and theoretical analysis, 

but rarely conducted in field applications. Based on this, an extended freezing 

code was incorporated into HYDRUS-1D to numerically solve coupled equations 

governing phase changes between water and ice and heat transport (Hansson 

et al., 2004). The freezing model was proofed to predict well the soil water and 

heat fluxes under both unfrozen and frozen conditions. This provided a basis for 

the studies of soil freezing and thawing behavior. It has been found that soil 

143

temperature is a main factor in determining biological processes, e.g. annual 

carbon uptake in the grassland ecosystem (Liu et al., 2007). As land surface 

schemes begin to simulate carbon and nitrogen fluxes from biological processes, 

they need to get soil freezing and soil temperature correctly. Our results suggest 

that it is possible to do that since the freezing model can provide an accurate 

temperature simulation. 

144 

However, the freezing model seems to overestimate water content after 

spring snowmelt and thus underestimate runoff. We considered that it is 

reasoned that the freezing model adopted soil surface temperature instead of air 

temperature as the atmospheric boundary condition in the numerical algorithm of 

the freezing model. Consequently, the model incorrectly partitions all the 

snowmelt into infiltration as soil thawing and snow melting happen 

simultaneously. At this moment, the complicated interactions between the soil 

surface microclimate and physical processes is not solved. Obviously the 

surface runoff algorithm in the freezing model is not sensitive enough. In addition, 

the Theta-probe that we used cannot measure the ice content under frozen 

condition, which also limited the verification of modeled ice content in this study. 

Moreover, the rain gauge just can record the snow-water equivalent, which 

obviously omitted the time of snow falling and melting. We suggests that the 

seasonal water balance, especially considering rainfall water stored as snow, 

snow drift and the lateral flow on frozen soil layers need to be investigated 

further. 

Conclusions 

1. Soil hydraulic, thermal and mechanical properties were interrelated and 

modified by grazing. Especially, multivariate geostatistical analysis revealed 

scale-dependent correlations among them at multiple spatial scales. 

2. Soil compaction by sheep trampling resulted in a homogenous spatial 

distribution of soil properties, which increased soil vulnerability against water and 

wind erosion.


3. After the long-term exclosure (~ 25 yr), grassland soils showed distinct signs 

of recovery from grazing-caused damages in hydraulic as well as mechanical 

aspects. 

4. The temporal stability concept, linked with a hydraulic model, provided a 

useful tool for sampling strategies and verifications of hydraulic process. 

5. The modified hydraulic model HYDRUS was successful in simulating the 

coupled transport of water and heat in the investigated rangeland ecosystems. 

The model results showed that intense grazing deteriorated soil functions, 

consequently reduced the plant available water and thus grassland productivity. 

6. An extended freezing model is validated with measured data, which provide a 

basis for simulations of snow hydrology and soil freezing and thawing 

processes. 

Outlook 

It is still a challenge to estimate fluxes at different spatial scales. Especially, how 

hydraulic model, e.g. HYDRUS-1D/2D can be applied to a large area, e.g. Xinlin 

River catchment where the MAGIM project took place with the knowledge of 

scaling (regionalization, transferability and upscaling). 

The monitoring (e.g. soil hydraulic properties, water content) in the deep soil 

must be determined not only for water use efficiency considering water use by 

roots deep penetrating into deeper soil layers (e.g. hydraulic lift) but also for the 

verification of model. 

Improvement of boundary conditions, respectively, an assessment of regional 

heterogeneity of rainfall is vital for model validations. 

145

Under frozen conditions, the differences between simulated and measured water 

contents could be also related to soil structural changes. Further work should 

establish a relation between the impedance factor and soil hydraulic properties 

that can be measured in the field. It is possible that measurements of soil 

moisture characteristic curves under both frozen and unfrozen conditions are 

helpful to provide a basis for this aspect. 

A detailed knowledge of the soil-atmosphere interface is needed to improve the 

surface runoff algorithm in the “freezing module”. 

Our contributions to MAGIM project 

146 

We mostly explored the spatial distribution of various regionalized variables, 

which not only helped to quantify the environmental changes due to the grazing 

effects at various scales, but also provided a basis for the regionalization, 

interpolation, transferability and model upscaling from local scale to a large area 

where the MAGIM project took place. Especially, we deliver a better process 

understanding for grazing-induced changes in physical and hydraulic soil 

properties and their effects on local and regional water and heat fluxes in 

semi-arid regions. This should improve the assessment and prediction of 

ecosystem changes caused by grazing management. 

For further studies and possibly future prediction, the hydrological model 

HYDRUS including freezing and thawing processes was parameterized and 

validated. This is an important part for the integrated model network of MAGIM. 

The calibrated hydraulic and thermal parameters is useful to other models, e.g. 

SWAT model that will be used to calculate C, N and water fluxes in the Xilin 

River catchment. The modeling products are also essential for inter-verification 

between the related models. In addition, the modeling products are essential as 

the additional information, which provide some data for the time and place which 

sampling is not available. Certainly, the model will definitely serve to characterize 

the present conditions, and further to predict the future conditions.

References 


Bremer, D.J., L.M. Auen, J.M. Ham, and C.E. Owensby. 2001. 

Evapotranspiration in a prairie ecosystem: Effects of grazing by cattle. 

Agono. J. 93(2):338–348. 

Casa, R., and A. Castrignanò. 2007. Analysis of spatial relationships between 

soil and crop variables in a durum wheat field using a multivariate 

geostatistical approach. Eur. J. Agron. doi:10.1016/j.eja.2007.10.001. 

Christensen, L., M.B. Coughenour, J.E. Ellis, and Z.Z. Chen. 2004. Vulnerability 

of the Asian typical steppe to grazing and climate change. Climatic Change 

63:351–368. 

Drewry, J.J., 2006. Natural recovery of soil physical properties from treading 

damage of pastoral soils in New Zealand and Australia: a review. Agr. 

Ecosyst. Environ. 114:159–169. 



ASAE 32:565–571. 

Gifford, G.F., and R.H. Hawkins. 1978. Hydrologic impact of grazing on infiltration: 

a critical review. Water Resour. Res. 14:305–313. 

Gómez-Plaza, A., J. Alvarez-Rogel, J. Albaladejo, and V.M. Castillo. 2000. 

Spatial patterns and temporal stability of soil moisture across a range of 

scales in a semi-arid environment. Hydrol. Proc. 14:1261-1277. 

Greenwood, K.L., and B.M. McKenzie. 2001. Grazing effects on soil physical 

properties and the consequences for pastures: a review. Austr. J. Exp. Agr. 

41:1231–1250. 

Hansson, K., J., Šimůnek, M. Mizoguchi, L.-C., Lundin, and M.T., van Genuchten. 

2004. Water flow and heat transport in frozen soil: numerical solution and 

freeze-thaw applications. Vadose Zone J. 3:693–704. 





147

Lophaven, S., J. Carstensen, and H. Rootzéna. 2006. Stochastic modelling of 

148 

dissolved inorganic nitrogen in space and time. Ecol. Modell. 193, 467-478. 


preferential flow pathways in the Shale Hills catchment, Vadose Zone 

Journal, 5:317–340. 

Liu, C.Y. J. Holst, N. Brüggemann, K. Butterbach-Bahl, Z.S. Yao, J. Yue, S.H. 

Han, X.G. Han, J. Krümmelbein, R. Horn, and X.H Zheng. 2007. 

Winter-grazing reduces methane uptake by soils of a typical. Atmospheric 

Environment. 41(28):5948–5958. 

Neitsch, S.L., J.G. Arnold, J.R. Kiniry, J.R. Williams, and K.W. King, 2002. Soil 



Texas. 





Proffitt, A.P.B., R.C. Jarvis, and S. Bendotti. 1995. The impact of sheep trampling 

and stocking rate on the physical properties of a Red Duplex soil with two 

initially different structures. Aust. J. Agric. Res. 46:733–47. 

Richard G., I. Cousin, J.F. Sillon, A. Bruand, and J. Guérif. 2001. Effect of 


properties. Eur. J. Soil Sci. 52:49–58. 

Rietkerk, M., and J. van de Koppel.1997. Alternate stable states and threshold 

effects in Semi-Arid grazing systems, OIKOS 79:69–76. 

Shouse, P.J., W.B. Russell, D.S. Burden, H.M. Swlim, J.B. Sisson, and van 

Genuchten, M.Th., 1995. Spatial variability of soil water retention functions 

in a silt loam soil. Soil Sci. 159(1):1-12. 


software package for simulating the one dimensional movement of water,





Snyman, H.A. 2005. Rangeland degradation in a semi-arid South Africa - I: 



Villamil, M.B., N.M. Amiotti, and N. peinemann. 2001. Soil degradation related to 

overgrazing in the semi-arid Southern Caldenal area of Argentina. Soil Sci. 

166:441–452. 

Wang, R.Z., and E.A. Ripley. 1997. Effects of grazing on a Leymus chinensis 

grassland on the Songnen plain, north-eastern China. J. Arid Environ. 

36:307–318. 

Western, A.W., and G. Blöschl. 1999. On the spatial scaling of soil moisture. J. 

Hydrol. 217:203-224. 

Williams, A.G., J.L. Ternan. C. Fitzjohn, S. de Alba, A. and Perez-Gonzalez. 

2003. Soil moisture variability and land use in a seasonally arid environment. 

Hydrological Processes 17:225–235. 

149

150

8. ACKNOWLEDGMENTS 

Firstly I would like to thank my supervisors Prof. Rainer Horn and Dr. Stephan 

Peth for their guidance and support continuously. Additional thanks to Prof. Bin 

Zhang for his recommend in the MAGIM project, and Dr. Xinhua Peng for his 

support in the data analysis and paper work. 

I acknowledged the fund support of Deutsche Forschungsgemeinschaft (DFG 

#536) and the colleagues of MAGIM project. Julia Krümmelbein, Zhongyan 

Wang, Xingzhi Gao, Marcus Giese, Markus Steffens, Carsten Hoffmann, Katrin 

Schneider, Yujin Wen and Klaus Erdle are acknowledged for the co-operated 

field work, data support and manuscript improvement. 

I thank colleagues of Kiel University who provided good study environments, e.g. 

Julia Krumbelein who helps me in German material things, Pia Luttith who helps 

me for many staffs and Jens Rostek for computer things. I like Kiel summer: 

beach, forest, sunshine and flower, warm-hearted persons and clear 

environment. I will miss Kiel and my colleagues in Kiel. 

“Sonntag” dinner in Tuanbu and barbecue in the ecological station (IMGERS) 

were also indispensable to this work. I will miss my drinking partner, e.g. 

Chengjie Wang and many others. Most especially, to Inner Mongolia grassland, 

thank you for rustic people, blue sky, sunshine, delicious sheep meat “hot pot”, 

and strong wind and wine, and wish you “healthiness” forever! 

Thanks to my family for the consistent faith, care and support in study and life – 

my parents, who always understand me although they don’t know what I on 

earth did and do, and – my wife Xiaoyan Wang, who gives the huge contribution 

to this work and definitely know what my work is. 

Ying Zhao 

16 Dec. 2007 

Kiel University, Germany 

151

152

Curriculum Vitae 

M.Sc. Ying Zhao 

Office Address: Institute of Plant Nutrition and Soil Science, CAU Kiel, 

Olshausenstr. 40. 24118 Kiel. Germany 

E-mail: y.zhao@soils.uni-kiel.de or swa_eag@yahoo.com.cn 

Tel: +49-431-8804078; Fax: +49-431-8802940 

PERSONAL DETAILS 

Nationality: Chinese 

Sex: Male 

Martal status: Married 

Place of birth: Gansu, China 

Date of birth: 04. 07. 1979 

EDUCATION 

Ph.D study, 05/2005-02/2008, Institute of Plant Nutrition and Soil Science, CAU 

Kiel, Germany (IPNSS) 

Supervisor: Prof. Dr. R. Horn 

Master degree, 09/2002-07/2005, Institute of Soil Science, Chinese Academy of 

Sciences, Nanjing (ISSSAS) 

Supervisor: Prof. Dr. Bin Zhang 

Bachelor degree, 09/1998-07/2002, College of Geography and Environment; 

Northwest Normal University (NWNU) 

153

154

Schriftenreihe des Instituts 

Neuere vorrätige Titel 

Band 61 Götz Reimer (2003): Spektrale Naherkundung und Ertragskartierung als Basis von 

digitalen Hof-Bodenkarten im Präzisen Landbau 

Band 62 Thomas Baumgartl (2003): Kopplung von mechanischen und hydraulischen 

Bodenzustandsfunktionen zur Bestimmung und Modellierung von Zugspannungen 

und Volumenänderungen in porösen Medien 

Band 63 Frank Nahrwold (2004): Bodenerosionsstudien und Möglichkeiten der 

Reliefmodellierung zur Reduzierung des Abtrages 

Band 64 Stephan Peth (2004): Bodenphysikalische Untersuchungen zur Trittbelastung von 

Böden bei der Rentierweidewirtschaft an borealen Wald- und subarktisch-alpinen 

Tundrenstandorten 

Band 65 Jörg Voßbrink (2005): Bodenspannungen und Deformationen in Waldböden durch 

Waldmaschinen (http://e-diss.uni-kiel.de/diss_1434) 

Band 66 Frank-Helge Richter (2005): Vergesellschaftung und Eigenschaften von Böden 

unterschiedlicher geomorpher Einheiten einer Jungmoränenlandschaft des 

ostholsteinischen Hügellandes 

Band 67 Orsolya Fazekas (2005): Bedeutung von Bodenstruktur und Wasserspannung als 

stabilisierende Kenngrößen gegen intensive mechanische Belastungen in einer 

Parabraunerde aus Löss unter Pflug- und Mulchsaat 

Band 68 José Dörner (2005): Anisotropie von Bodenstrukturen und Porenfunktionen in 

Böden und deren Auswirkungen auf Transportprozesse im gesättigten und 

ungesättigten Zustand 

Band 69 Wibke Markgraf (2006): Microstructural changes in soils rheological 

investigations in soil mechanics. (//e-diss.uni-kiel.de/diss_1882) 

Band 70 Dorota Agnieszka Dec (2006): Thermal properties in Luvisols under conventional 

and conservation tillage treatment (//e-diss.uni-kiel.de/diss_1904) 

Band 71 Emilia Jasinska (2006): Management effects on carbon distribution in soil 

aggregates and its consequences on water repellency and mechanical strenght 

(//e-diss.uni-kiel.de/ diss_1706) 

Band 72 Rainer Horn, Heiner Fleige, Stephan Peth (2006): Soils and Landuse 

Management Systems in Schleswig-Holstein (Germany) - Guide of ISTRO 

Excursion 2006 

Band 73 Jürgen Lamp, Hans-Jürgen Hess und Götz Reimer (2007): Technik- und 

Feldführer zur Veranstaltung „Präziser Landbau / Precision Farming im Dienst für 

Landwirtschaft und Umwelt 

Band 74 Julia Krümmelbein: (2007): Influence of various grazing intensities on soil 

stability and water balance of a steppe soil in Inner Mongolia, P.R. China 

Band 75 Stephan Gebhardt (2007): Wasserhaushalt und Funktionen der Böden im 

Grundwasserabsenkbereich des Wasserwerkes Wacken in Schleswig-Holstein 

Band 76 Imke Janßen (2008): Landnutzungsabhängige Dynamik hydraulischer und 

mechanischer Bodenstrukturfunktionen in Nassreisböden

SCHRIFTENREIHE Institut für Pflanzenernährung und Bodenkunde ...

You also want an ePaper? Increase the reach of your titles

Delete template?

Save as template?