1 Spatial Modelling of the Terrestrial Environment - Georeferencial

Coupled Land Surface and Microwave Emission Models 61
ensures consistency between the parameterizations used within the modelling system. Two
coupled models are commonly discussed in the literature, MICRO-SWEAT (Burke et al.,
1997; 1998; 2003) and LSP/R (Judge et al., 1999, 2001; Liou et al., 1998; Liou and
England, 1998). Both of these models produce excellent agreement with measurements:
MICRO-SWEAT has been successfully compared with 1.4 GHz (L-band) truck-based data
and LSP/R with 19 GHz tower-based data. To date, most research has focused on running
these models at point scale, but Burke et al. (2003) used data collected during the Southern
Great Plains 1997 experiment (SGP97 homepage) to investigate the effectiveness of using
a 2-D array of MICRO-SWEAT models to predict distributed L-band microwave brightness
temperatures measured by an aircraft. They found discrepancies between the spatial
distributions of modelled and measured brightness temperature which likely result from
imperfect knowledge of the spatial distributions of soil properties, precipitation, and the
representation of vegetation.
In theory, any land-surface model [e.g. BATS (Dickinson et al., 1986); CLM (Zeng
et al., 2002); LSM (Bonan, 1996)] could be coupled with a microwave emission model.
However, in practice, because the microwave brightness temperature is mainly sensitive to
near-surface soil moisture (in top 2–5 cm) and decreases with depth in the soil, the most
accurate predictions of microwave brightness temperature result when using a land surface
model that represents very fine near-surface soil layers. For example, in their studies, Burke
et al. (1997; 1998) used SWEAT (Daamen and Simmonds, 1996), a land surface model
that has 8 layers in the top 5 cm, with layer thickness increasing with depth from a 1 mm
thick top layer.
In general, the microwave emission from the soil can be described using coherent
(Wilheit, 1978; Njoku and Kong, 1977; England, 1976) or non-coherent models (Burke
et al., 1979; England, 1975). Schmugge and Choudhury (1981) and Ulaby et al. (1986)
compared these two types of models, as follows. Non-coherent models estimate emissivity
using the dielectric contrast at the air/soil interface and are accurate only when the (variable)
sampling depth within the soil is well known. Most retrieval algorithms are based on
non-coherent models, with the near-surface soil water content assumed to be uniform to
a specified depth. In the case of L-band, this is usually 5 cm (e.g. Jackson et al., 1999).
In coherent models, the value of the emissivity at the surface is coupled to the dielectric
properties of the soil below the surface; consequently, they provide better estimates of the
emissivity when the soil water content profile is not uniform. One significant disadvantage
of coherent models is that interference effects can occur, but these have only rarely been
witnessed in nature (Schmugge et al., 1998).
The emission component of MICRO-SWEAT is based on the Wilheit (1978) coherent
model of electromagnetic propagation through a plain stratified medium. The Wilheit (1978)
model relates the brightness temperature emergent at the soil surface to the dielectric
properties and temperature of the underlying soil layers. For radiation with polarization, p,
the microwave brightness temperature, T Bp (K), is given (Wilheit, 1978) by:
T Bp (θ) =
N∑
f ip (θ) T i (1)
i=1
where T i is the temperature of the ith layer (K), f ip is the fraction of energy absorbed from
an incident microwave with polarization p by the ith layer (a function of the dielectric