*4.1. Influence of the ET0 Model*

The models commonly used internationally for *ET*<sup>0</sup> estimation can generally be divided into four major categories [54]: temperature methods, such as the Hargreaves [50] and Blaney–Criddle models [55]; mass transfer methods, such as the Rohwer [56] and Penman models [57]; radiation methods, such as the Priestley–Taylor [58] and Jensen–Haise models [59]; and synthesis methods, such as the Penman–Monteith model [4]. Although the Penman–Monteith model is considered the standard method for calculating *ET*0, accurate predictions of future *ET*<sup>0</sup> in watersheds require more reliable meteorological data, and the outputs of existing GCMs and downscaling methods typically have low modeling accuracy for meteorological parameters such as wind speed, relative humidity, and radiation. Compared to other models, temperature-based models have lower data requirements and are computationally simple, and downscaling methods have a clear advantage in modeling temperature [60,61], with correlation coefficients generally above 0.90 (Figure 6), making them more widely applicable [31]. In addition, several studies have revealed strong relationships between future *ET*<sup>0</sup> and temperature in other basins or regions. Xing et al. [27] attributed the increase in *ET*<sup>0</sup> in the 21st century to an increase in temperature, and Ding and Peng [31] found that global warming led to a change in the main sensitive factor for potential evapotranspiration in the Loess Plateau from the average temperature in the historical period to the maximum temperature in the future period. The results of both these studies theoretically supported the use of the temperature-based *ET*<sup>0</sup> model. Ahmadi and Baaghideh [25] explored the effect of climate change on *ET*<sup>0</sup> in pistachio cultivation areas in Iran using the Hargreaves model. Yan and Mohammadian [47] evaluated the performance of the evaporation model based on the Hargreaves formulation using various fitting methods, and the results showed that the simulation was satisfactory. Among the many *ET*<sup>0</sup> simplification methods, this study also selected the Hargreaves model, a temperature method recommended by the FAO [4] and researchers [62,63], to estimate the spatial and temporal characteristics of *ET*<sup>0</sup> in the YRB under four future emission scenarios. The fitted scatter density plot had an *R*<sup>2</sup> of 0.8523 at the monthly scale (Figure 7). The Hargreaves model can effectively simulate basin *ET*0; reflecting the rationality of the formula selection in this study. Notably, this study only considered the effect of temperature on *ET*0, thus, there may be some bias in the estimation results. The applicability of temperature models in specific regions [64,65] should be improved, and the effect of other meteorological elements on future *ET*<sup>0</sup> [66] should be explored in future studies.
