Global sensitivity and uncertainty analysis of a sugarcane model considering the trash blanket effect

Abstract

The deterministic approach in crop modeling simplifies uncertainty present in the environment using a unique parameter set. In practice, this uncertainty is seen in the variability of data collected in a field experiment. One way to exploit this uncertainty is to use the stochastic approach, by inserting the range of plausible variability into the simulation’s parameters and inputs. This study aims to evaluate the ability of a process-based crop model to simulate the uncertainty of a sugarcane field. We employed the recently updated version of SAMUCA model to simulate the sugarcane growth and development in a 4-year field experiment, where the crop was grown under the effect of green cane trash blanket (GCTB) and bare soil (Bare). To analyze the effect of genotype and soil variability on output variables, a stochastic approach was applied to the corresponding parameters of the SAMUCA model. A global sensitivity analysis was utilized to prioritize and identify the most important parameters to explain the model uncertainty. Then, the uncertainty was analyzed in three different ways: uncertainty analysis only for genotype parameters (UG), uncertainty analysis only for soil parameters (US) and the analysis of both soil and genotype parameters (UGS). We quantified the variability of the stochastic simulation by the ratio between the average of the standard deviation of the simulations and the average of the standard deviation of the observed data. The variability observed in the field is not fully explained by the hydraulic parameters of the soil, possibly due to irrigation and good rainfall distribution in the area. Furthermore, the variability in US simulations were higher for GCTB than in Bare treatment, suggesting that the GCTB has a larger influence in SAMUCA’s variability than for the hydraulic parameters in the conditions of this study. The UG and UGS had the same capacity to quantify the variability present in the environment for the treatments Bare and GCTB. In this case, sensitivity to soil parameters can simply be ignored and genotype parameters can be chosen as the only source of variability for practical applications. Our suggestion for future work is to explore environments without irrigation, different amounts of GCTB and other soil parameters present in the model.