Calibration and validation of the EPIC model to predict glyphosate movement with different agronomic practices under shallow water table conditions

Glyphosate (GLP) has been the most frequently used herbicide worldwide, including Europe. Due to its systemic, post-emergence, and non-selective characteristics, it offers optimal weed control without the need of mechanical treatments. Therefore, it is widely used in no-till practices. However, increasing awareness of its potential harmful effect to human health and ecosystems has led numerous countries to restrict or, even, ban its use. The EPIC (Environmental Policy Integrated Climate mode) model has been selected as a screening tool to evaluate the vulnerability of groundwater to glyphosate contamination under different pedo-climatic and agronomic conditions across the Veneto Region (NE Italy), an area where the interaction of different pedo-climatic and agronomic conditions makes it difficult to predict site-specific GLP movement. The aim of this study was to evaluate the performance of a modified version of EPIC that includes a fast solution of Richards’ equation to predict GLP dynamics under shallow water table conditions. The experimental site was in Northeastern Italy and consisted of eight drainable lysimeters; 4 treatments, replicated twice, in factorial combination of two management practices (conventional -CV- and conservation -CA- agriculture) and two water table levels (60 and 120 cm). Degradation and movement of GLP in the soil profile were monitored in 2019 from May to September. The herbicide (144 mg m^-2) was applied on bare soil in CV and on the cover crop (Secale cereale) in CA. Water samples were systematically collected at 15, 30 and 60 cm depth using suction cups, whose suction was regulated by an automated system that combined matric potential readings, provided by electronic tensiometers, with a vacuum regulator. Water samplings from groundwater were also performed. Soil samples were collected at 0-5 and 5-15 cm depth every other week. Weather and soil data were used as input to EPIC, while the GLP experimental results, along with yields, soil water content, evapotranspiration and water percolation data, were used to calibrate (from 2011 to 2017) and validate (from 2018 to 2019) the model. In all lysimeters, GLP reached the groundwater the day after the first irrigation event, with higher leaching in CV than in CA and at 120 than at 60 cm. After 40 days, GLP was almost completely dissipated in CA soil, while it was still detected in CV. EPIC was able to acceptably reproduce evapotranspiration (R2=0.76), yields (R2=0.74) and water percolation (R2= 0.59-0.90). In general, GLP predictions compared well with observations but the predictions in CV treatments were closer to observations than in CA treatments. This work showed the robustness of the modified EPIC, suggesting its use as a tool to assess the potential vulnerability of the groundwater under different management scenarios and water table levels.