The Spatio-temporal variability of reference evapotranspiration (ET0) was analysed using different estimation techniques with limited data over western Maharashtra, India. Sixteen ET0 models were compared with FAO 56 Penman-Monteith (P-M) method, out of that eight climate-based methods viz., Soil Conservation Service Blaney-Criddle, Thornthwaite, Hargreaves-Samani, Pan evaporation, Jensen-Haise, Priestly-Taylor, Turc, and Radiation were selected, and four linear regression (LR) and four artificial neural networks (ANN) models were developed. The LR and ANN models viz., Model1 (evaporation), Model2 (maximum and minimum temperature), Model3 (maximum and minimum temperature, bright sunshine hours), Model4 (maximum and minimum temperature, maximum and minimum humidity, bright sunshine hours). Resulting LR and ANN models showed satisfactory performance and can be accepted to predict ET0 values. The inverse distance weighting technique used for mapping, and the Mann-Kendall and linear regression methods are used for analysing trends of ET0. The map of P-M method showed that the values of ET0 increase from central part towards south-east and relatively more ards north-east of study area. Among climate based methods, it was observed that none of methods shown close in spatial distribution pattern of ET0 with P-M method, whereas the maps of LR and ANN models showed close approach to P-M method. It was observed that there was decreasing trend of ET0 using P-M method for all stations as well as the more complex methods which require temperature, radiation and humidity data showed decreasing trends of ET0 for most of the stations Whereas the methods which requiring only temperature data showed no or increasing trends of ET0. Hence, it was concluded that trends of ET0 depends on type of meteorological data used for calculation under current global scenario.