Irrigation is the only method to sustain oasis agriculture in northwest China. In this study, the effects of soil moisture perturbations on the oasis self‐maintenance mechanism were investigated using the mesoscale model MM5 with satellite observations of land‐surface parameters from MODIS data. Twelve simulations were performed, among which CTL (default simulation) and MOD (with parameters replaced with MODIS data) were used to evaluate the model results. Another 10 simulations with soil moisture perturbations (DRY and WET) assessed the influence of soil moisture on the oasis self‐maintenance mechanism. Independent of the initial soil moisture content, all simulations showed dramatic reductions of soil moisture in the initial several days of the simulation, due to the strong evapotranspiration in this arid environment. Both the DRY and WET perturbations led to substantial changes in surface heat flux partitioning and the oasis ‘cold‐wet’ effect, which were also verified by an analysis of the secondary circulation and moisture‐inversion level. Drying tends to have more influence than wetting, which indicates that irrigation should be performed more frequently than intensively to protect the oasis self‐maintenance mechanism.