The study was conducted in a typical warm desert area of Alashan to study the lagging response of rodent communities to climate change,the Shannon-Wiener index and meteorological factors in 2006~2014 were used to establish the BP neural network model under the different grazing disturbance from 2009 to 2014.The results indicated that 1)The effect of the prediction of the BP neural network model were different under different grazing type(goodness of fit values were 0.9499,0.9442 and 0.8678, respectively),and rotational grazing was better than that of grazing enclosure and overgrazing.2)Shannon- Wiener index of the rodent communities which was affected by climate change had obvious hysteresis effect,and lagged for 3 months,3 months and 1 month for grazing enclosure,rotational grazing,and overgrazing,respectively.3)According to the lag effect and the lag time,it will be predicted in advance for the change trend of Shannon-Wiener index under different grazing disturbance,which can provide theoretical guidance for prevention and control of rodents.