3 Leaching models

The description of vertical flow of water and pesticides in soils not only accounts for water percolation and downward flux, but also for chemical and biological pesticide attenuation processes, such as sorption and microbial degradation. Numerous 1D and 2D models describe pesticide leaching from soils to the phreatic zone, such as EXSOL (Matthies and Behrendt, 1991*), LEACHP (Hutson and Wagenet, 1993*), VARLEACH (Walker et al., 1996*), and MACRO (Jarvis et al., 1994*). Table 2 provides an overview of existing numerical leaching models.

Differences between these models arise from diverse approaches to describe water flow. PRZM (Lorber and Offutt, 1986; Donigian and Carsel, 1987), PELMO (Klein, 1994), PLM (Hall, 1994), and VARLEACH use a capacity approach to describe water flow, whereby water in excess of field capacity in any layer moves down to the next layer within the same time step. Capacity based approaches generally have a low computational demand, but they reveal some significant deficiencies. For example, the upward movement of water due to evaporation cannot be simulated. PESTLA (Brouwer, 1994), LEACHP, and MACRO use the Richards equation, whereby water flow is determined by differences in water potential and soil hydraulic conductivity. All leaching models describe plant uptake of water, and convective transport is described equally, whereas diffusive and dispersive fluxes are handled differently: PRZM, VARLEACH, EXSOL, and PLM adopt numerical procedures based on soil layer thickness; MACRO, PESTLA, and LEACHP calculate these fluxes based on user-specified diffusivity and dispersivity parameters, and PELMO uses a combination of both approaches. Most of the models are able to simulate time-dependent changes in sorption: LEACHP, PESTLA, PELMO, and MACRO are able to describe non-linear sorption according to the Freundlich-isotherm. PELMO can also simulate higher order degradation pathways. Except for PRZM, all models consider temperature and moisture on microbial degradation rates. Most models can simulate pesticide uptake by plants. In addition, PRZM, PELMO, EXSOL, and LEACHP consider volatization of pesticides.

Compared to LEACHP, HYDRUS (Persicani, 1993) represents the modern type of leaching models, which use Richards equation to calculate water flow and which describe solute transport by a convective-dispersive equation. HYDRUS was found to be sensitive to the K_D value employed to describe partitioning, but delivered robust results in a comparative study (Persicani, 1996). Gärdenäs et al. (2006) used HYDRUS-2D to account for preferential flow. A dual-permeability approach was found to accurately simulate preferential drainage flow, while equilibrium and mobile-immobile approaches largely failed to capture the preferential flow process.

Leaching models are deterministic or hybrid and describe the same processes of attenuation. However, only few of them consider horizontal runoff as a loss term. GLEAMS (Sabbagh et al., 1993), PRZM and PELMO can calculate water and pesticide runoff and thus may provide simulated pesticide loads to a 2D transport model to simulate diffuse emissions on larger scales. GLEAMS uses a curve number approach driven by daily rainfall and relates the runoff curve number to daily soil water content in the root zone, while PRZM uses the same (SCS) approach relating curve number to soil moisture limits in the surface zone. PELMO uses the same approach as does PRZM. In these three models, loads in runoff are calculated from edge-to-field water runoff volumes, empirical extraction coefficients and sediment concentrations, assuming linear sorption isotherms and a constant mixing depth at the surface.

Ma et al. (1999) compared the runoff components of PRZM and GLEAMS and found good correlations of simulated loads to empirical results measured on a field scale, but both models failed to calculate pesticide concentrations in runoff water and underestimated pesticide loads. It remains doubtful, if both models would exhibit a reasonable performance on even larger scales, because PRZM and GLEAMS do not consider any retention above soil surface during horizontal transport and thus may overestimate actual pesticide loads.

Gottesbüren et al. (2000) compared MACRO, LEACHP, GLEAMS, PELMO, and further leaching models by simulating the leaching of the herbicide isoproturon and the water tracer bromide in a profile of a silty loam soil. The blind test employing eight persons who applied the models exhibited an overwhelming influence of the individual users on the simulation results. Herbst et al. (2005a) executed a test of the models MARTHE (Thiéry and Amraoui, 2001*), TRACE (Herbst et al., 2005b*), ANSWERS (Park et al., 1982), and MACRO. These authors found that the Richards equation-based models MARTHE, TRACE, and MACRO performed better for water flow predictions than the capacity-based model ANSWERS. Preferential flow implemented in the models MARTHE, TRACE, and ANSWERS did not influence the simulation of water flow significantly, but had great influence on the simulated pesticide concentrations. In another study, Vanclooster et al. (2000) not only tested several leaching models, but also gave recommendations how to improve them.

Generally, leaching models were made to predict pesticide concentrations in groundwater and therefore, horizontal transport was not implemented or was described only in a rudimentary way. Hence, these models need to be combined with surface transport models, if runoff to surface waters shall be predicted accurately.