To evaluate the effect of soil moisture (SMOIS) initializations on simulated short-range high-temperature weather in Chuanyu Area, Southwest China, we conducted Advanced Research WRF simulations, and found that results are sensitive to the SMOISs (i.e. the FNL, CFSR, ERA, and JRA data) and perturbation (i.e. Experiments WET30 and DRY30). Specifically, results of root-mean-squared error and threat score of ERA and JRA for surface air temperature (SAT) are better than those of FNL and CFSR. Additionally, we defined a conditioned correlation coefficient (CCC) that is suitable for evaluating high-temperature distribution, showing the highest CCC by JRA and large differences among the experiments. Despite of the differences and the model-dependent nature of SMOIS, the ranges in the simulated high temperature indicate the general reasonability of the data in terms of area-averaged values. This reasonability and the large CCC differences highlight the importance of accurate SMOIS data in model-forecasted high-temperature weather. Considering the assimilation systems in which FNL and CFSR were produced by the Noah land surface model (same as in this study), the results also suggest that the SMOIS accuracy or the model configuration is more important than the consistency between assimilation systems and the model. Moreover, the mechanism for the SMOIS-temperature change was explored. Using the equation for temperature change, the relative importance of the physical processes was quantified, which also shows large SMOIS-induced differences over sub-daily periods.