Application of Sample Selection Model in Estimating Response Functions for Nitrate Percolation

The Seymour aquifer region of Texas has been identified as containing elevated levels of nitrate in ground water. Various state and federal agencies are currently studying policy options for the region by gathering more site-specific information. However, because of lack of sufficient information, cause and effect relationships between water quality and agricultural practices have not been well established for the region. Some recently available biophysical simulation models have impressive capabilities in generating large amounts of data on environmental pollution resulting from agricultural production practices. In this study, the data generated by a biophysical simulation model were used to estimate the nitrate percolation response functions for the Seymour aquifer region. Interestingly, nitrate percolation values obtained from simulation models often comprise acensoredsample because the non-zero percolation values are only observed under certain climatic events and input levels. It has been shown in the econometric literature that the use of Ordinary Least Squares (OLS) on censored sample data produces biased and inconsistent parameter estimates. Thus, a sample selection model was used in this study to estimate the response functions for nitrate percolation. The study provides some insight into the relationship between nitrate percolation and agricultural production practices. In particular, the study demonstrates the potential of selected design standards in minimizing agricultural nonpoint-source (NPS) pollution for the study area.