Reference evapotranspiration estimation in Benin-Owena River Basin using deep learning approaches: A review

Reference evapotranspiration (ETo) is essential for water resource management. Yet, its accurate estimation in the Benin-Owena River Basin Development Authority (BORBDA) is constrained by sparse meteorological networks and incomplete data records. While the FAO Penman-Monteith method remains the standard, its application is limited by data scarcity. This literature review synthesizes existing research on ETo estimation within the BORB and examines the emerging potential of deep learning (DL) approaches. Conventional studies in the basin have primarily relied on empirical and basic machine learning models, revealing systematic data gaps and limited spatial coverage. In contrast, recent DL applications in similar humid tropical regions-such as the Amazon and Southeast Asia-have demonstrated superior performance, with architectures like Long Short-Term Memory (LSTM) and hybrid models effectively capturing complex temporal dependencies and achieving high accuracy even with limited inputs. Despite this success, a significant research gap remains: no comprehensive study has applied advanced DL techniques to ETo modelling in the BORB. This review highlights the critical opportunity to develop robust, DL-based tools tailored to the basin's unique climatic and data conditions. Advancing such methodologies could significantly enhance irrigation scheduling, agricultural water management, and climate adaptation strategies across one of Nigeria's most vital agricultural regions.