Relationship between leaching in lysimeter studies and ... - pfmodels

**Relationship** **between** **leach ing**

**lysimeter** **studies** **and** **in** FOCUS

Hamburg groundwater scenario

Jos Boesten

Outl**in**e

• Introduction

• Effect of **lysimeter** boundary condition

• Effect of experimental design of **lysimeter** study

• Effect of uncerta**in**ty **in** soil/pesticide properties of **lysimeter**

soil

• Conclusions

Introduction

• FOCUS Groundwater Workgroup:

• Tiered approaches for EU **and** MS registration

• Tier 1: st**and**ard scenario

• Tier 2: **lysimeter** study (German guidel**in**e)

• Discussion on relationship **between** tier 1 **and** 2

• My role: scientist present**in**g his personal view

• appreciat**in**g your feedback as **in**put to further discussion **in** FOCUS

GW

(next meet**in**g 20 December)

Introduction

• Regulatory assumption: protection goal for

groundwater is 90 th percentile **in** space **and**

time of annual average concentration **in**

groundwater for agricultural use area of

pesticide

• Not **lysimeter** is protection goal but field

Introduction

Comparison of endpo**in**ts:

• Tier 1: FOCUS **leach ing** concentration (80 th percentile of 20 periods)

• Tier 2: **lysimeter**: maximum annual average **leach ing** concentration

• Consensus to correct l**in**early for pesticide dose (assum**in**g same application period)

• FOCUS **leach ing** concentration closer to protection goal than

concentration

• Most simple approach: assume that both endpo**in**ts can be directly compared

• Implicit assumptions of direct comparison:

• **lysimeter** boundary condition has no significant **in**fluence = H1

• experimental design of **lysimeter** has no significant **in**fluence = H2

• uncerta**in**ty **in** system parameters has no significant **in**fluence = H3

Introduction

German **lysimeter** guidel**in**e:

• Use soil with low organic matter

• Ensure annual average ra**in**fall of 800 mm (via irrigation)

• Vulnerability concept based on chromatographic flow

(otherwise emphasis on events **and** pore structure)

• My procedure consistent with guidel**in**e assumptions:

• simulations with chromatographic model that h**and**les both

**lysimeter** **and** field boundary conditions

• PEARL (MACRO **and** HYDRUS would be possible as well)

Effect of **lysimeter** boundary condition

• Effect on water flow

• Effect on pesticide **leach ing**

Effect of **lysimeter** boundary condition

PEARL simulations:

• FOCUS Hamburg scenario

• Lysimeter system: 1-m

deep **lysimeter**

• Field system: 4.5-m deep

soil profile with

groundwater at average

depth of 1.8 m (range 0.7-

3 m)

• Same soil, crop **and**

weather

**lysimeter**

field

Effect of **lysimeter** boundary condition

Scenario characteristics:

• FOCUS runs for applications every year or other year

• Summer wheat (often used **in** **lysimeter**s)

• Application of 1 kg/ha one day before emergence

• Endpo**in**ts: FOCUS **leach ing** concentration for both

systems

• A few model compounds (us**in**g PEARL default

parameters for long-term sorption k**in**etics)

Effect of **lysimeter** boundary condition on water flow

Field system

Water flow rate at 1 m

depth (mm/d)

Groundwater level

(m below soil surface)

Effect of **lysimeter** boundary condition on water flow

Water flow rate at 1 m depth

Field system

Lysimeter system

Water flow k**in**etics differ

Effect of **lysimeter** boundary condition on water flow

Volume fraction of water at 30 cm depth

Field system

Lysimeter system

Differences **in** top soil limited

Effect of **lysimeter** boundary condition on water flow

Volume fraction of water at 1 m depth

Field system

Lysimeter system

At 1 m depth the **lysimeter** is wetter

Effect of **lysimeter** boundary condition on **leach ing**

1. field BC gives higher **leach ing** concentration than

BC

2. **in**dication: effect larger for mobile, rapidly degrad**in**g

pesticides

Effect of **lysimeter** boundary condition on **leach ing**

P1

P2

P3

Results **in** terms of numbers

Effect of **lysimeter** boundary condition on **leach ing**

Possible explanation for lower **lysimeter** concentrations:

- Lysimeter boundary leads to higher volume fraction of

water **in** bottom part of **lysimeter**

- So longer residence time **in** **lysimeter** so more time for

pesticide degradation

Consequence: boundary condition has effect so **lysimeter**

endpo**in**t cannot be used directly **in** the risk assessment

Effect of experimental design of **lysimeter**

• Typical design of **lysimeter** experiment:

• 2 years with 800 mm annual average ra**in**fall/irrigation

• Clean soil profile at start

• Endpo**in**t: maximum annual average **leach ing**

concentration

• Design of FOCUS scenarios:

• Run 20-40-60 y after 6 y warm-up

Effect of experimental design of **lysimeter**

Procedure: simulation of **lysimeter** experiment

• 1-m deep **lysimeter** with Hamburg soil **and** weather

• Period: 31 March 1913 – 31 March 1915

• Summer wheat emerg**in**g 1 April

• Pesticide dose of 1 kg/ha at 31 March 1913

• Annual ra**in**fall was 800 **and** 826 mm **in** 2 years after application

• Analysis restricted to 2 years **and** one application

• Comparison with FOCUS Hamburg scenario assum**in**g

**lysimeter** boundary condition

• not with field boundary condition to keep comparison as simple as

possible

• application every year (20 y) or every two years (40 y)

Effect of experimental design of **lysimeter**

P-1

P-2

P-3

**lysimeter** **leach ing** concentration may be much

lower than FOCUS **leach ing** concentration

Effect of experimental design of **lysimeter**

FOCUS/**lysimeter**

P-1 : zero sorption so leaches **in** first w**in**ter

P-3 : moderate sorption: 2 y **lysimeter** is too short for maximum

Effect of experimental design of **lysimeter**

P-3: simulated **leach ing** from

time

Two years with

800 mm ra**in**fall

not enough for

moderately

sorb**in**g pesticide

**lysimeter** period

(K OM = 90 L/kg)

Effect of experimental design of **lysimeter**

Effect of experimental design depends strongly on

pesticide properties:

for mobile pesticides **lysimeter** higher concentrations

for sorb**in**g pesticides FOCUS scenario higher

concentrato**in**s

- difference **in** endpo**in**ts may be a factor 100

- also experimental design has effect, so **lysimeter**

endpo**in**t cannot be directly used **in** risk assessment

Effect of uncerta**in**ty **in** soil/pesticide

properties of **lysimeter** soil

Introduction

Assumptions **in** tier 1:

• Hamburg soil has dispersion length of 5 cm for PEARL

• Average DT50 or K OM

If **lysimeter** would have been conducted with soil with

• shorter dispersion length

• shorter DT50 or higher K OM

then **lysimeter** endpo**in**t lower by co**in**cidence: undesirable for higher tier.

Effect of uncerta**in**ty **in** soil/pesticide properties of **lysimeter**

soil

Example of DT50/K OM

• Mean DT50 of 30 d based on four values: 22, 27, 35, 36

• Mean K OM of 35 L/kg based on four values: 28, 34, 38, 40

• Lysimeter endpo**in**t (PEARL)

• mean values: 0.38 μg/L

• most favorable comb**in**ation (DT50=22 d, K OM =40 L/kg): 0.03 μg/L

• Uncerta**in**ty **in** DT50 **and** K OM can lead to wrong risk

assessment conclusion

Effect of uncerta**in**ty **in** soil/pesticide properties of **lysimeter**

soil

• What is probability of obta**in****in**g **lysimeter** endpo**in**t

concentration of below 0.1 μg/L by co**in**cidence ?

• Important parameters that may be uncerta**in**:

• DT50 **and** K OM

• Freundlich exponent (curvature of sorption isotherm)

• Dispersion length

• Exploratory calculations on this probability via Monte

Carlo

Effect of uncerta**in**ty **in** soil/pesticide properties of **lysimeter**

soil

• Stochastic variables **in** Monte Carlo simulations:

• Normally distributed DT50 **and** K OM with CV of 25%

• Uniformly distributed Freundlich exponent rang**in**g from 0.8 to 1.0

• Lognormally distributed dispersion length

Effect of uncerta**in**ty **in** soil/pesticide properties of **lysimeter**

soil

• Justification of probability density functions used **in**

Monte Carlo simulations

Effect of uncerta**in**ty **in** soil/pesticide properties of **lysimeter**

soil

- **studies** with 18 UK soils (>15% clay)

Allen & Walker (1987) + Walker & Thompson (1977)

assumption: variation coefficient of 25% for DT50 **and** K OM

Effect of uncerta**in**ty **in** soil/pesticide properties of **lysimeter**

soil

- **studies** with 18 UK soils (> 15% clay) Allen & Walker (1987)

assumption: Freundlich exponent uniformly distributed

**between** 0.8 **and** 1.0

Effect of uncerta**in**ty **in** soil/pesticide properties of **lysimeter**

soil

dispersion length: lognormal distribution with median of 5 cm **and**

st.dev. of 0.87 ln(cm) ;

based on pooled distribution of column **and** field by Jan

V**and**erborght [median 6.2 cm **and** st.dev. of 0.87 ln(cm) ]

Effect of uncerta**in**ty **in** soil/pesticide properties of **lysimeter**

soil

• Background of output quantity

Effect of uncerta**in**ty **in** soil/pesticide properties of **lysimeter**

soil

Example distribution

with low % below 0.1

μg/L

Most relevant output for risk assessment:

probability that **lysimeter** gives less than 0.1 μg/L by

co**in**cidence, so % of runs below 0.1 μg/L.

Effect of uncerta**in**ty **in** soil/pesticide properties of **lysimeter**

soil

Example distribution

with high % below

0.1 μg/L

Effect of uncerta**in**ty **in** soil/pesticide properties of **lysimeter**

soil

• Procedural aspects

Effect of uncerta**in**ty **in** soil/pesticide properties of **lysimeter** soil

Procedure: runs

along two l**in**es **in**

the DT50-K OM

plane

constant K OM of 35 L/kg **and**

variable DT50

Effect of uncerta**in**ty **in** soil/pesticide properties of **lysimeter**

soil

constant DT50 of 20 d **and** variable K OM

Effect of uncerta**in**ty **in** soil/pesticide properties of **lysimeter**

soil

DT50 = 75 d

K OM = 35 L/kg

500 Monte

Carlo runs is

enough

different seeds

**in** all runs

Effect of uncerta**in**ty **in** soil/pesticide properties of **lysimeter**

soil

• Results

Effect of uncerta**in**ty **in** soil/pesticide properties of **lysimeter**

soil

- % plotted aga**in**st concentration for median parameter values

-0.1 μg/L no unique function of concentration level for median values

- smaller effect of uncerta**in**ty for low K OM values

Effect of uncerta**in**ty **in** soil/pesticide properties of **lysimeter** soil

- at 2 to 5 μg/L still 10% of runs below 0.1 μg/L so very wide

distributions of calculated concentrations

- effect of uncerta**in**ty **in** these properties may be very large

Effect of uncerta**in**ty **in** soil/pesticide properties of **lysimeter**

soil

0.5% below 0.1 μg/L 9% below 0.1 g/L

Examples of simulated probability density functions

Effect of uncerta**in**ty **in** soil/pesticide properties of **lysimeter**

soil

DT50 = 35 d

K OM

= 35 L/kg

• Large effect of dispersion length on **leach ing**:

• 0.7 μg/L for 5 cm

• 0.2 μg/L for 2 cm

Effect of uncerta**in**ty **in** soil/pesticide properties of **lysimeter**

soil

How large is effect of

dispersion length

**and** how large effect of

sorption/degr. parameters ?

Effect of uncerta**in**ty **in** soil/pesticide properties of **lysimeter**

soil

disp.var.

allvar.

sorp.deg.var.

Prelim**in**ary

conclusions:

sorption **and**

degradation larger

effect than dispersion

effect more than

additive; cause

unknown

Conclusions

• Use of **lysimeter** endpo**in**t directly for risk assessment

unacceptable because significant effects of

• Lysimeter boundary condition

• Lysimeter experimental procedure

• Uncerta**in**ty **in** soil or pesticide properties

• Recommendation: **in**terpret results of **lysimeter** **studies** with

methods that account for such effects

• e.g. Verschoor et al. (2001) presented at 1 st EU Modell**in**g Workshop

End

© Wagen**in**gen UR