A review of the Artificial Intelligence (AI) based techniques for estimating reference evapotranspiration: Current trends and future perspectives

Reference Evapotranspiration (ET_o) is a complex, dynamic and non-linear hydrological process. Accurate estimation of ET_o has long been an eminent topic of interest in the research community for its importance in effective planning and sustainable water resource management. Although the FAO-56 Penman-Monteith (PM) equation has been accepted as a standard equation for ET_o measurement, the primary concern that inhibits the applicability of this equation is the requirement for all the climatological variables, which might not be available at a given location. Owning to the remarkable success and accuracy achieved by Artificial Intelligence (AI) in almost every sphere, scientists have proposed the usage AI models for ET_o prediction as an alternate to the conventional methods. This comprehensive review will serve to raise awareness regarding the various state-of-the-art standalone AI frameworks, along with capturing the intriguing developments in the advanced AI space such as the hybrid and ensemble models, evolutionary models and a range of optimization techniques. The results from the publications published over the last 15 years (2007-2022) for ET_o prediction using AI under varied agro-climatic scenarios have been analysed and synthesized. The advantages and disadvantages of the established AI techniques have been discussed in each subsection. Some of the derived insights and major findings are discussed along with the future research recommendations. This review will not only provide a research vision for the novice researchers in the applicability of the aforementioned techniques, in context of ET_o prediction, but also be helpful as a compilation of the AI modelling studies for ET_o prediction for the established water resource engineers and hydrologists.