EFFECT OF PARAMETER UNCERTAINTY ON DRAINMOD PREDICTIONS: I. HYDROLOGY AND YIELD

Abstract

The computerbased hydrologic model DRAINMOD can be used to predict the effect of drainage design on therate of subsurface drainage and on crop yield. An uncertainty analysis was conducted to quantitatively assess the variabilityin model outputs caused by parameter uncertainty. The analysis was based on an experimental field at the Tidewater ResearchStation in Plymouth, North Carolina. As a first step in the uncertainty analysis, a sensitivity test was conducted to determinewhich parameters in the model have the most influence on the model objective functions. Firstorder approximation andMonte Carlo simulation were used to determine the effect of the uncertainty in the most sensitive parameters on theuncertainty in the model objective functions. Objective functions evaluated were: average annual subsurface drainagevolume; SEW30 (a measure of stress caused by excessive soil water in the top 30 cm) during the growing season; and relativeyield for both conventional and controlled drainage. Nine parameters found to significantly affect model output were usedin the uncertainty analysis. The firstorder approximation showed that in the case of conventional drainage, lateral saturatedhydraulic conductivity accounted for 81% of the uncertainty in terms of variance for predicted annual subsurface drainagevolume, 81% for growing season SEW30, and 71% for relative yield. For controlled drainage, lateral saturated hydraulicconductivity contributed 62% of the uncertainty in terms of the variance in predicted annual subsurface drainage volume,69% in growing season SEW30, and 62% in relative yield. The Monte Carlo simulation showed similar results. Improving theknowledge of these most influential parameters will help to reduce the uncertainty in DRAINMOD predictions for theseobjective functions.