Emulator based Global Sensitivity Analysis of Apsim-Sugar Model: Emulator Accuracy Assessment
Global sensitivity analysis of process-based crop models is laborious and computationally expensive. Hence, the less computationally, expensive, and reliable approach is required. Gaussian Emulator Machine for Sensitivity Analysis (GEM-SA) use an emulator-based approach to conduct sensitivity analysis. Hence it is a computationally less expensive way to conduct global sensitivity analysis. It is tested with some process-based models, however rarely tested for crop models. This study aimed to assess the applicability of GEM-SA to conduct a sensitivity analysis of the APSIM crop model. Above-ground biomass, fresh cane yield, and sucrose weight of sugarcane were simulated for 300 input combinations using apsimr package of R 3.5 software. Thirty emulators (10 years x 3 outputs) were developed to estimate parameter sensitivity of 26 plant parameters using GEM-SA software. Results revealed that developed emulators are accurate. Hence the emulator-based approach is confident enough to conduct the sensitivity analysis of APSIM. GEM-SA provide a computationally less expensive and reliable platform to conduct a sensitivity analysis of the APSIM crop model. Keywords - Apsimr, Emulator, GEM-SA, Global Sensitivity Analysis, Sugarcane.