A Gaussian-Process-Based Global Sensitivity Analysis of Cultivar Trait Parameters in APSIM-Sugar Model: Special Reference to Environmental and Management Conditions in Thailand

Abstract

<jats:p>Process-based crop models are advantageous for the identification of management strategies to cope with both temporal and spatial variability of sugarcane yield. However, global optimization of such models is often computationally expensive. Therefore, we performed global sensitivity analysis based on Gaussian process emulation to evaluate the sensitivity of cane dry weight to trait parameters implemented in the Agricultural Productions System Simulator (APSIM)-Sugar model under selected environmental and management conditions in Khon Kaen (KK), Thailand. Emulators modeled 30 years, three soil types and irrigated or rainfed conditions, and emulator performance was investigated. rue, green_leaf_no, transp_eff_cf, tt_emerg_to_begcane and cane_fraction were identified as the most influential parameters and together they explained more than 90% of total variance on the simulator output. Moreover, results indicate that the sensitivity of sugarcane yield to the most influential parameters is affected by water stress conditions and nitrogen stress. Our findings can be used to improve the efficiency and accuracy of modeling and to identify appropriate management strategies to address temporal and spatial variability of sugarcane yield in KK.</jats:p>

Global Optimization of Cultivar Trait Parameters in the Simulation of Sugarcane Phenology Using Gaussian Process Emulation

“…Although the APSIM-Sugarcane model does not contain rue and transp_eff_cf among the cultivar-specific parameters, they were included in the current study because Sexton et al [29] have identified rue as a highly sensitive parameter for biomass yield estimation and transp_eff_cf as a highly sensitive parameter for biomass yield estimation under water stress conditions. Bandara et al [27] have obtained similar results related to rue and transp_eff_cf for estimation of CDW. If the supply of soil water is not a growth-limiting factor, the APSIM-Sugarcane model controls dry matter assimilation via radiation interception and rue.…”

“…Parameter ensembles were prepared by using 500 random numbers which are distributed uniformly between the maximum and minimum values of each parameter space (Table 3: Parameter space) using functions in R [38] software. Reasonable values for the parameter spaces were selected based on the descriptions of previous studies of Bandara et al [27] and Sexton et al [22,29].…”

“…The optimized parameter ensembles for experiment B1 (P3 and P4) resulted in comparatively low NRMSE val values in the estimates of the observed sugarcane yields for experiment A. Because in APSIM-Sugarcane, parameter rue is previously identified as a highly sensitive parameter for the estimation of biomass and CDW of sugarcane [27][28][29][30]. Therefore, when optimizing parameters under low water stress conditions (B1), the estimated value of rue could be increased to estimate higher yields.…”

“…An emulator of sufficient accuracy can be used as a substitute for the original simulator (APSIM), and parametrization can be based on the emulator. The feasibility of substituting APSIM-Sugarcane for sensitivity analysis based on GP emulation has previously been evaluated in several studies [27][28][29][30]. However, there has not yet been any study of the use of the GP emulators to optimize parameters in the APSIM-Sugarcane model nor an examination of the accuracy of the optimization method.…”

Improving the Performance of Index Insurance Using Crop Models and Phenological Monitoring

“…In an effort to address these challenges, we analysed to what extent the integration of crop models and phenological monitoring can help reduce these design and temporal basis risks, respectively. Biophysical crop simulation models can be leveraged to generate larger synthetic yield datasets, which can then be used to train weather-or satellite-based index models [18][19][20] or support spatial targeting of limited numbers of CCEs that can be conducted as part of area-yield insurance products. However, to date, this approach has not been widely applied in the context of index insurance design, with limited evidence about its performance at spatial scales relevant for insurance applications (e.g., field, farm or village) or in comparison with index models derived empirically from available observational yield datasets.…”