Applicability of Different Ensemble Techniques in Enhancing Water Requirement Simulations

Securing accurate insights into crop water requirements is crucial, particularly in arid regions and for key strategic crops. This study presents a novel applied methodology to evaluate the efficacy of ensemble modeling in determining the Saffron Water Requirement (SWR). The research aims to generate more reliable insights and enhance the accuracy of water requirement estimations for saffron, a vital crop in eastern Iran. The proposed plan includes establishing a rigorous testing process to evaluate the efficiency of various ensemble methods. Three significant ensemble classes— Ensemble Learning Machine (ELM), combination, and averaging techniques — were addressed to produce the new prediction of SWR. As such, a Decision Tree Regression (DTR) tool and six different experimental methods, serving as base models, were initially applied. The effectiveness of various ensemble methods was evaluated using statistical and qualitative tests. This plan included time series comparisons, key diagnostic indices such as RMSE and NSE, Absolute Error Decomposition (AED) analysis, and the Rate of Improvement (ROI). Results showed that applied ensemble systems are not only of high quality but also capable of presenting a skillful prediction of SWR compared to base models. Results revealed that the boosting procedure had a beneficial effect on DTR simulations, increasing them by more than ۷۴ percent. Additionally, combining methods could enhance the base prediction by more than ۷۵%.

کلیدواژه ها:

Absolute Error Decomposition ، AdaBoost ، Bias and Variance Errors ، Collinearity ، Inter-dependency