Evaporation Estimation in the Muko Reservoir, Songwe Region, Tanzania: A Multi-Model Framework Using Temperature, Radiation, and Combined Approaches

This study presents the development and evaluation of a multi-model framework (MMF) embedded in a Decision Support System (DSS) for estimating evaporation in data-scarce environments. Focusing on the Muko Reservoir catchment in the semi-arid Songwe Region of southwestern Tanzania, the framework integrates four widely recognised reference evapotranspiration (ET₀) models: Hargreaves-Samani, Jensen-Haise, Priestley-Taylor, and Penman. These models represent three methodological categories: temperature-based, radiation-based, and combined approaches. A 10-year dataset of monthly pan evaporation and meteorological variables was used to compare model performance through descriptive statistics, time series analysis, and correlation metrics. Results show that while all models reasonably capture seasonal evaporation patterns, significant differences exist in magnitude and variability. The Priestley-Taylor and Penman models consistently overestimated ET₀, while the Jensen-Haise model underestimated it. The Hargreaves-Samani model showed the closest agreement with observed values. The study highlights the value of a DSS-driven multi-model approach in improving evaporation estimation accuracy, especially in regions with variable data availability. The findings have significant implications for water balance modelling, reservoir operation, and climate-resilient water resource management.