Performance evaluation of AI and hybrid-AI models for estimation of evaporation in Lesser Himalayan Valley
摘要
Evaporation holds a significant position in the global hydrological cycle and is one of the intricate phenomena widely affected by various hydrometeorological parameters. Evaporation accounts for 66% of global precipitation losses, profoundly influencing surface and rainfall losses, necessitating its meticulous quantification. Its estimation is resource-intensive, time-consuming, costly and sensitive to climatic and spatial variability. Over 22 numerical-physical methods are available, affected by time, data availability and climatic conditions. Data-driven artificial intelligence (AI) and machine learning (ML) models can be useful where it is very difficult to estimate the evaporation spatially and precisely. The current study uses 10 daily hydrometeorological in situ parameters: temperature (maximum, minimum and mean), vapour pressure (7.19 h, 14.19 h), relative humidity (7.19 h, 14.19 h), rainfall, bright sunshine hours and mean wind velocity for 23 years (2001:2023) for a Lesser Himalayan Valley (Doon Valley), India. The models developed for the estimation are ANN-SGD, ANN-LM, ANN-Adam, SVM and LSTM in addition to two hybrid models;