SENSITIVITY ANALYSES OF THE NITROGEN SIMULATION MODEL, DRAINMOD-N II

Abstract

A two-step global sensitivity analysis was conducted for the nitrogen simulation model DRAINMOD-N II to assessthe sensitivity of model predictions of NO3-N losses with drainage water to various model inputs. Factors screening usingthe LH-OAT (Latin hypercube sampling - one at a time) sensitivity analysis method was performed as a first step considering48 model parameters; then a variance-based sensitivity analysis was conducted for 20 model parameters, which were theparameters ranked 1 to 14 by the LH-OAT method, five organic carbon (OC) decomposition parameters, and the empiricalshape factor of the temperature response function for the nitrification process. DRAINMOD-N II simulated a continuous cornproduction on a subsurface drained sandy loam soil using a 40-year (1951-1990) eastern North Carolina climatologicalrecord. Results from the first 20-year period of the simulations were used to initialize the soil organic matter pools, and resultsfrom the last 20-year period of the simulations were considered for the sensitivity analyses. Both yearly and 20-year averagemodel predictions of NO3-N losses through drainage flow were used in the analyses. Both sensitivity analysis methodsindicated that DRAINMOD-N II is most sensitive to denitrification parameters, especially those controlling temperatureeffect on process rate. Results also indicated that the model is mildly sensitive to the parameters controlling OC decompositionand associated N mineralization/immobilization. The use of different sensitivity analysis methods with dissimilar theoreticalfoundations increases the confidence in key parameters identification. More efforts should be focused on quantifying keyparameters for more accurate model predictions.