Differences between simulated and observed data often occur when the watershed model is applied under extreme climate. It is necessary to assess the stability of hydrological models in a wide range of climate variation. A case study was conducted in Fuhe basin of Poyang Lake, China using the Soil and Water Assessment Tool (SWAT) model, which was calibrated under different climates in average years, dry and wet years, high and low temperature years, respectively. The model was first calibrated with dataset in average years, and the validation in the whole period showed results agreed well with the observed stream flow. The well-parameterized model calibrated under extreme climate was used to simulate hydrological responses in different climate years. All simulations generated results closely matching observed data with R2 and ENS greater than 0.88, although the model was likely to slightly overestimate stream flow in average and dry years, and underestimate in wet years and high temperature years. In addition, each simulation was independent to other simulations with different parameters calibrated in different climate periods tested by Student's T-test. Therefore, the model has the potential probability to accommodate a large range of climate variation to predict hydrological responses to climate change.