Comparative evaluation of eight simplified potential evapotranspiration methods against the FAO-56 penman–monteith benchmark in the Jayakwadi Command Area, Maharashtra, India: a seasonal and spatial analysis

Abstract

Accurate estimation of reference evapotranspiration (ET 0 ) is fundamental to irrigation scheduling, crop water requirement computation, and hydrological modeling in semi-arid agricultural regions. The FAO-56 Penman–Monteith (FAO-56 PM) method, the globally recommended standard, demands complete meteorological datasets that are rarely available across India's irrigation command areas. This study evaluates whether simplified empirical ET 0 methods can provide operationally reliable estimates in place of FAO-56 PM under the semi-arid, monsoonal conditions of the Jayakwadi Command Area (JCA), Maharashtra, India. A 34-year monthly synthetic meteorological dataset (1990–2024) was constructed for five representative virtual stations distributed across the head reach zone of JCA, using interpolated India Meteorological Department (IMD) climatological statistics and NASA-POWER reanalysis data, with independent validation against records from the Maharashtra Water Resources Department. Eight empirical ET 0 methods were evaluated against the FAO-56 PM benchmark: three radiation-based methods [Makkink (Mak), Abtew (Abt), Priestley–Taylor (PT)] and five temperature-based methods [Hargreaves–Samani (HS), Thornthwaite (Tho), Hamon (Ham), Linacre (Lin), and Blaney–Criddle (BC)]. Model performance was assessed across annual and four seasonal time scales (spring, summer, autumn, winter) using Nash–Sutcliffe Efficiency (NSE), Relative Error (Re), Normalized Root Mean Square Error (NRMSE), coefficient of determination ( R 2 ), and Percent Bias (PBIAS). Radiation-based models consistently outperformed temperature-based models. On a seasonal scale, Abtew achieved the highest summer performance (NSE = 0.937, Re = 0.007, NRMSE = 0.076) and strong spring accuracy (NSE = 0.664). Makkink performed best in winter (NSE = 0.886) and autumn (NSE = 0.630). Temperature-based models produced consistently negative NSE values at the annual scale, rendering them unreliable for annual ET 0 estimation. All eight models showed negative NSE annually, confirming that seasonal disaggregation is essential for empirical model selection in this region. High R 2 values for Priestley–Taylor (up to 0.986) did not translate to low bias or error, underscoring that correlation alone is an insufficient performance criterion. Makkink and Abtew, both requiring only solar radiation as their primary input, offer practical, data-light alternatives to FAO-56 PM for operational ET 0 estimation across the JCA. Model selection should be guided by season and available data rather than applied uniformly. Future work should incorporate field-measured ET data and explore machine learning-based residual correction to further enhance estimation accuracy.