Uncertainty analysis of DSSAT plant model parameters in estimating cotton yield using GLUE

Crop growth simulation models are extensively used for various agricultural studies, including optimal crop selection, irrigation management, and assessing climate change impacts. Among these models, the DSSAT (Decision Support System for Agrotechnology Transfer) is particularly prominent for its ability to simulate growth, yield, and other dynamics for 34 different crops. The DSSAT model integrates various components such as soil, weather, crop management, and genetic factors to provide comprehensive insights into crop performance (Jones et al., 2003). Accurate parameter calibration in this model is crucial for reliable simulations. However, the inherent variability and uncertainty in parameter values pose significant challenges. Uncertainty can arise from various sources, including measurement errors, spatial and temporal variability, and model structure. Addressing these uncertainties is essential to enhance the reliability and accuracy of the model predictions. The Generalized Likelihood Uncertainty Estimation (GLUE) algorithm offers a robust framework for quantifying and incorporating parameter uncertainty into model simulations (Beven & Binley, 1992).In this study, we focus on the application of the GLUE algorithm to the DSSAT model for cotton, aiming to improve the model's predictive accuracy by accounting for parameter uncertainty. We utilize observational data from different irrigation treatments to calibrate the model and evaluate the posterior probability distributions of the parameters.

The study used data from a 2009 experiment conducted at the Birjand University research farm. The DSSAT v4.5 model was employed, requiring inputs such as weather, soil properties, and crop management data. Four irrigation treatments (50%, 75%, 100%, and 125% of crop water requirement) were tested to evaluate the GLUE algorithm’s performance in estimating model parameters.

The results demonstrated that the GLUE algorithm effectively estimated the probability distributions of the DSSAT model parameters for cotton. The algorithm’s performance was compared with previous models lacking uncertainty assessments, showing significant improvements in simulation accuracy (Qasemi et al., 2019). The findings highlighted the importance of considering parameter uncertainty for better predictive accuracy and model reliability.

The GLUE algorithm, through Monte Carlo simulations, provides a robust method for assessing and incorporating parameter uncertainty in crop growth models like DSSAT. This approach enhances the model's reliability in predicting crop performance under varying conditions, which is crucial for agricultural planning and management.