A multi-scale evaluation of 30 empirical models for reference evapotranspiration estimation in a data-scarce semi-arid region: A case study of Jhansi, India

Study region: Semi-arid agro-climatic zone of central India. Study focus: This study evaluated the performance of 30 empirical ET0 models; ten each from the temperature-based, mass-transfer-based and radiation-based categories, against the FAO-PM standard using 21 years (2000–2020) of daily meteorological observations at Jhansi, India. The analysis is framed as a benchmarking of empirical models against FAO-PM. Model accuracy was quantified at daily, monthly, seasonal and annual scales using four statistical indices (NSE, RMSE, MBE and r), and an integrated model ranking was derived using the entropy-weighted TOPSIS multi-criteria decision-making framework. New hydrological insights for the region: The 21-year FAO-PM record reveals a persistent annual atmospheric water deficit of ~731 mm (ET₀ = 1571 mm yr⁻¹ versus rainfall = 840 mm yr⁻¹). This deficit is highly asymmetric: the summer season alone (May–June) accounts for 440 mm of ET0 against negligible rainfall, defining the period of maximum irrigation demand, while the monsoon surplus is offset within weeks by rapid soil moisture depletion. Pre-winter and winter seasons contribute a further 295 mm of ET0. The leading models are wind-driven mass-transfer models, showing that aerodynamic demand rather than solar radiation is the primary control on daily ET₀ variability in this climate. Category alone, however, poorly predicts performance across the models. The Penman model (TOPSIS rank 1, closeness = 0.953) and WMO model (rank 2), requiring only wind speed and vapour pressure inputs, are identified as most accurate FAO-PM substitutes at the daily and seasonal scale. The findings of the study are location-specific and need validation before applying to other locations in the semi-arid region.