Metamodels and nonpoint pollution policy in agriculture

Abstract

Complex mathematical simulation models are generally used for quantitative measurement of the fate of agricultural chemicals in soil. But it is less efficient to use them directly for regional water quality assessments because of the large number of simulations required to cover the entire region and because the entire set of simulation runs must be repeated for each new policy. To make regional water quality impact assessment on a timely basis, a simplified technique called metamodeling is suggested. A metamodel summarizes the input‐output relationships in a complex simulation model designed to mimic actual processes such as groundwater leaching. Metamodels are constructed and validated to predict groundwater and surface water concentrations of major corn and sorghum herbicides in the Corn Belt and Lake States regions of the United States. The usefulness of metamodeling in the evaluation of agricultural nonpoint pollution policies is illustrated using an integrated environmental economic modeling system. For the baseline scenario, we estimate that 1.2% of the regional soils will lead to groundwater detection of atrazine exceeding 0.12 Mg/L, which compares well with the findings of an Environmental Protection Agency monitoring survey. The results suggest no‐till practices could significantly reduce surface water concentration and a water quality policy, such as an atrazine ban, could increase soil erosion despite the conservation compliance provisions.