The effect of parameter uncertainty on DRAINMOD predictions for hydrology, yield and water quality.

摘要

Reducing the nutrient levels in sensitive coastal waters has become a national priority. Agriculture has been targeted as a significant contributor to the nutrient problem. Controlled drainage has been recognized as one way to reduce nitrate losses from agricultural fields requiring subsurface drainage. The computer model DRAINMOD can be used to determine the effect of drainage design on the rate of subsurface drainage, crop yield and water quality. An uncertainty analysis was conducted to quantitatively assess the variability in model results caused by parameter uncertainty. The methodology included a sensitivity analysis, First Order Approximation and Monte Carlo Simulation. The objective functions evaluated were average annual subsurface drainage, SEW30, relative yield, NO3-N loss via subsurface drainage, NO3-N loss via surface runoff, and NO3-N transformation by denitrification for conventional and controlled drainage. Base values for the parameters were taken from an experimental site in Plymouth, North Carolina. The predominant soil type at the site was a poorly drained Portsmouth sandy loam. Results of the analysis showed that the majority of the uncertainty in the hydrology/yield objective functions was caused by the uncertainty in the horizontal saturated hydraulic conductivity and the maximum surface storage. The majority of the uncertainty in the water quality objective functions was caused by the denitrification rate constant. Despite the uncertainty in the model objective functions a clear difference between controlled drainage and conventional drainage was found for all of the objective functions. Monte Carlo Simulations for conventional and controlled drainage showed significant (alpha = 0,05) reductions of NO3-N (29%) at the field edge.