In order to predict long-term flooding under extreme weather conditions in central Asia, an energy balance-based distributed snowmelt runoff model was developed and coupled with the Soil and Water Assessment Tool(SWAT) model. The model was tested at the Juntanghu watershed on the northern slope of the Tian Shan Mountains, Xinjiang,China. We compared the performances of temperature-index method and energy balanced method in SWAT model by taking Juntanghu river basin as an application example(as the simulation experiment was conducted in Juntanghu River, we call the energy balanced method as SWAT-JTH). The results suggest that the SWAT snowmelt model had overall Nash-Sutcliffe efficiency(NSE) coefficients ranging from 0.61 to 0.85 while the physical based approach had NSE coefficients ranging from 0.58 to0.69. Overall, on monthly scale, the SWAT model provides better results than that from the SWAT-JTH model. However, results generated from both methods seem to be fairly close at a daily scale. Thestructure of the temperature-index method is simple and produces reasonable simulation results if the parameters are well within empirical ranges. Although the data requirement for the energy balance method in current observation is difficult to meet and the existence of uncertainty is associated with the experimental approaches of physical processes, the SWAT-JTH model still produced a reasonably high NSE. We conclude that using temperature-index methods to simulate the snowmelt process is sufficient, but the energy balance-based model is still a good choice to simulate extreme weather conditions especially when the required data input for the model is acquired. In order to predict long-term flooding under extreme weather conditions in central Asia, an energy balance-based distributed snowmelt runoff model was developed and coupled with the Soil and Water Assessment Tool (SWAT) model. The model was tested at the Juntanghu watershed on the northern slope of the Tian Shan Mountains, Xinjiang, China. We compared the performances of temperature-index method and energy balanced method in SWAT model by taking Juntanghu river basin as an application example (as the simulation experiment was conducted in Juntanghu River, we call the energy balanced method as SWAT-JTH). The results suggest that the SWAT snowmelt model had overall overall Nash-Sutcliffe efficiency (NSE) coefficients ranging from 0.61 to 0.85 while the physical based approach had NSE coefficients ranging from 0.58 to 0.69. Overall , on monthly scale, the SWAT model provides better results than that from that from the SWAT-JTH model. However, results generated from both methods seem to be fairly close at a daily scale. The structure of the temperature-index method is simple and produces reasonable simulation results if the parameters requirement for the energy balance method in current observation is difficult to meet and the existence of uncertainty is associated with the experimental approaches of physical processes, the SWAT-JTH model still produced a reasonably high NSE. We conclude that using temperature-index methods to simulate the snowmelt process is sufficient, but the energy balance-based model is still a good choice to simulate extreme weather conditions especially when the required data input for the model is acquired.