Global sensitivity and uncertainty analysis of nitrate leaching and crop yield simulation under different water and nitrogen management practices

Abstract

Assessing the sensitivity and uncertainty of soil-crop models is beneficial to model calibration and development of best water and N management practices. This study adopted the Morris screening method and the Sobol’ variance-based method, combined with an agricultural system model (WHCNS), to analyze the global sensitivity and uncertainty of nitrate leaching and crop yield to model input parameters under different water and N management practices. A two-year field experiment was conducted in a desert oasis of Inner Mongolia, China using a factorial combination of standard (Istd, standard, 750mm per season; Nstd, standard, 138kgNha−1) and conservation (Icsv, conservation, 570mm per season; Ncsv, conservation, 92kgNha−1) levels of irrigation and N fertilization: IstdNstd, IstdNcsv, IcsvNstd and IcsvNcsv. Sensitivity analysis (SA) based on this experiment showed that nitrate leaching demonstrated significant sensitivity to soil hydraulic and crop parameters, but generally low sensitivity to N transformation parameters. Based on Sobol’ SA, crop parameters accounted for 64.3%, 63.2%, 39.2% and 39.2% of simulated nitrate leaching variability for the IstdNstd, IstdNcsv, IcsvNstd and IcsvNcsv treatments, respectively. The greater the crop water and N stress, the stronger the parameters interaction. Uncertainty analysis showed the average amount of nitrate leaching under Istd (135.3kgNha−1) to be 2.3 times greater than under Icsv (58.0kgNha−1); however, the distributions of yield between the four treatment combinations did not show significant differences. Overall, irrigation practice was the main factor influencing the parameter sensitivities and the uncertainty of nitrate leaching and crop yield simulation.