This study employs a methodological approach for estimating long-term series of monthly reference evapotranspiration (ETo) from historical data. To carry it out a regionally calibrated version of the Hargreaves equation was applied at old ordinary weather stations which only provide data of air temperature and precipitation. The proposed approach was based on the analysis of: (1) the Hargreaves coefficient obtained by local calibration from data of 66 modern automatic weather stations; (2) the regional characterization of the spatial variability of that coefficient by means of a “regional function”; and (3) the final application of this function to the old ordinary weather stations. This approach was assessed under the semiarid conditions of the Segura River Basin (south-eastern Spain) by comparing ETo estimates against those obtained with the Penman-Monteith method, which was used as reference. Spatial variability of the Hargreaves coefficient was well correlated with the annual and monthly means of daily temperature range, so they were selected as explanatory variables for the regionalization of the Hargreaves coefficient following two approaches: a global regional function and monthly regional functions. The regionally calibrated version of the Hargreaves equation by monthly functions clearly improved the performance of its original parameterization (average relative error decreased from 19.8% to 10.1%) although, as expected, estimates were not as good as those obtained with the local calibration (average relative error=7.7%).