Abstract: The goal of this article is to evaluate land surface models Noah,SHAW and CLM4.0 in arid regions (desert-grassland transition),using data observed at Microclimate and Evaporation Experiment in Inner Mongolia,with an emphasis on comparison of the models performance before and after a rare rainstorm which occurred during the simulated period.The results show that all of the three models can simulate the trend of turbulence flux,radiation flux,and soil moisture and soil temperature before and after the rare rainstorm.Except the simulation of soil temperature and soil water content,the performance of Noah land surface model is the best,especially the simulation of latent heat flux of Noah LSM,with the 0.99 correlation coefficient and 7.89 W·m^-2 root mean square error.The performance of three models in sensible heat flux,upward long wave radiation and net radiation is better after the rainstorm,and the models performance in soil temperature is better before the rainstorm,this shows that all of the three models are more appropriate to simulate sensible heat flux,upward long wave radiation and net radiation in moister land after rainstorm and more appropriate to simulate soil temperature without rainfall.For the three models,the influence period of rainstorm on soil temperature simulation is too short.

Key words: rainstorm, land surface process model, turbulence flux, radiation flux, soil temperature, desert-grassland transition area, arid region