Development and evaluation of hybrid deep learning long short-term memory network model for pan evaporation estimation trained with satellite and ground-based data

• We built hybrid Long Short-Term Memory model for Pan Evaporation (E p ). adopt Neighbourhood Component Analysis feature selection. predict E in drought-prone Queensland with satellite predictor variables. Our NCA-LSTM is compared competitive benchmark models. useful estimation of evaporative water losses. Evaporation, as a core process within the global hydrological cycle, requires reliable methods to monitor its variation, decision-making agriculture, irrigation systems and dam operations, also other areas hydrology resource management. Accurate monitoring pan evaporation ( ) one most popular approaches understand process. This work aims construct (LSTM) predictive that coupled selection regions Queensland, Australia (Amberley, Gatton, Oakey, & Townsville). Utilizing daily-scale dataset [31 August 2002 22 September 2020], performance proposed deep learning (DL) model, denoted NCA-LSTM, models, i.e ., standalone LSTM, types DL, single hidden layer neuronal architecture decision tree-based method. The testing results reveal lowest Relative Root Mean Square Error ≤ 20 % , Absolute Percentage Bias 14.5 highest Kling-Gupta Efficiency ≥ 87 attained by (relative models) tested Amberley, Oakey sites. In respect efficiency, improved selection, outperforms all indicating future utility prediction daily . practical sense, developed provides an accurate loss cycle therefore, can be implemented management, planning irrigation-based mitigation financial losses agricultural related sectors where, regular forecasting resources are vital part sustainable livelihood business.