Global changes have largely altered the magnitude and distribution of evapotranspiration (ET), which in turn significantly impacts regional water availability and water partitioning, particularly in semi-arid areas, such as the Agricultural-Pastoral Ecotone in Northwest China (APENC). In this study, the Budyko framework was revised and applied at a finer spatio-temporal scale for the simulation and attribution of monthly ET variability in dry regions. As a holistic approach for modeling water partitioning, the revised Budyko framework (RBF) is expected to achieve parsimony of data input and model parameters. Application of the Budyko-based framework in the APENC showed an accurate ET simulation (R2 = 0.96, RMSE = 3.45 mm/mon, NSE = 0.95), which is a better performance compared with two classical models, FAO crop coefficient model and Complimentary Relationship (CR) model with in situ observations. For the attribution of monthly ET variability, vegetation has the highest sensitivity coefficient, while the precipitation shows the largest actual contribution because of its larger variability. Subsequently, the RBF was assessed with two different groups of in situ observations in semi-arid areas with different environment and vegetation conditions, and showed credible results on ET simulation. This shows that the RBF with parsimony of model parameters is highly applicable in data lacking dry regions.