摘要:WXGEN, a weather generator model based on stochastic process theory and mathematical statistics, was widely used in hydrological monitoring models, crop yield estimation models and derived fields of meteorological data. In this study, we used WXGEN to evaluate the simulation accuracy of hydrothermal conditions in the highlands of the Qinghai–Tibet Plateau. Results showed that: (1) The Markovian chain transfer parameters P(W|D) and P(W|W) of each station were between 0.03–0.30 and 0.12–0.74, which was basically consistent with the temporal and spatial distribution of actual precipitation; (2) In the thermal data simulation, more than 96% of the meteorological stations passed the 0.05 level in three different significance tests of monthly mean minimum and maximum temperature and solar radiation, and the measured deviations of simulated annual mean temperature and solar radiation were 0.686 °C and 1.65 MJ/m2, respectively. In all, 94% of the stations in the hydrological simulation passed the monthly precipitation significance test; (3) The simulated vs. measured deviations of annual precipitation, heavy rain days and wet days were 8.04 mm, 1.023 d and 8.374 d, respectively; (4) The simulation of extreme hydrothermal conditions that may affect the yield of highland barley was very close to the measured situation, and the R2 of simulation and measured value was all above 0.85. The simulation of freezing damage was less accurate, but also higher than 0.85.