Using RZWQM2-PEST for Model Calibration, Data Worth, and Uncertainty: Examples from USGS Field Studies.

The Root Zone Water Quality Model (RZWQM2) is a widely used agricultural systems model that the U.S.  Geological Survey’s National Water Quality Assessment project has used to predict water, nutrient, and pesticide fluxes to the water table.  For example, accurate estimation of N flux to the water table can improve numerical groundwater models that simulate N transport to public-supply wells and streams.  RZWQM2 simulates water quantity and quality, crop growth, and management practices, and therefore has numerous parameters that require estimation.  Recent incorporation of PEST Parameter Estimation Software into RZWQM2’s user interface provides an efficient means of model calibration and sensitivity and uncertainty analyses.  In particular, linear sensitivity and uncertainty analyses can help users identify which parameters are well informed by the data, which are important to the predictions, and which observations most reduce prediction uncertainty.  Examples from field studies in Nebraska and Maryland show that soil hydraulic parameters were more sensitive in RZWQM2 calibration than pesticide properties or macropore parameters.  Further, the calibrated RZWQM2 model reduced the prediction uncertainty of maximum metolachlor and metolachlor ethane sulfonic acid (ESA) concentrations by up to 99%.  Data-worth analysis showed that measured metolachlor, ESA, metolachlor oxanilic acid, soil-moisture content, soil-water tension, and nitrate reduced the pre-calibration prediction uncertainty by 28 – 96% overall, even though the pesticide/degradate data were comparatively sparse.  Although models are essential tools to better understand complex systems and to extrapolate findings in space and time, the data required for calibration can be costly to obtain.  The findings show that the available data substantially reduced prediction uncertainty in the unsampled regions of pesticide/degradate breakthrough curves.