| Summary: | Accurate estimation of reference evapotranspiration (ET0) is a crucial parameter in implementing precise irrigation strategies and managing regional water resources effectively. While various methods have been proposed to obtain accurate ET0, the conventional approach is complex, uneconomical, and unable to contend with the rising variability and unpredictable weather patterns. Meanwhile, the lack of meteorological data limits the accurate estimation of ET0 via empirical models. Considering the recent approach in coupling ML techniques with optimisation algorithms to enhance the accuracy and robustness of ET0 estimation, this study was conducted to explore the performance of optimised hybrid Support Vector Regression (SVR) models integrated with meta-heuristic algorithms for daily ET0 estimation in Malaysia. Four hybrid SVR models, including SVR-Particle Swarm Optimisation (SVR-PSO), SVR-Whale Optimisation Algorithm (SVR-WOA), SVR-Differential Evolution (SVR-DE), and SVR-Covariance Matrix Adaptation Evolution Strategy (SVR-CMAES), were developed and assessed using the Root Mean Square Error (RMSE), Mean Absolute Error (MAE), coefficient of determination (R2), and Global Performance Index (GPI). The accuracy of the hybrid SVR models was then compared against standalone Machine Learning (ML) and empirical models using limited meteorological data. Accordingly, the findings highlighted the superior accuracy of the SVR-PSO model in estimating ET0, followed closely by the SVR-DE and SVR-CMAES models. The outstanding performance of the SVR-PSO model was attributed to the inherent versatility and robustness of PSO, as well as its core social behaviour and swarm intelligence principles that allow for an exhaustive exploration of the solution space, thus enhancing the model's accuracy and reliability. In conclusion, the integration of SVR with the meta-heuristics algorithm represents a significant advancement in ET0 estimation models with enhanced accuracy. The study underlines the critical role of advanced hybrid models in enhancing ET0 prediction accuracy, thereby supporting the implementation of efficient water resource management and strategic planning across Malaysia. © The Author(s), under exclusive licence to Springer Nature B.V. 2024.
|
|---|