Abstract: Abstract：Commonly low standards of flood control for medium and small rivers and frequently occurred flood disaster usually lead to extremely serious economic loss in China. Two typical medium and small rivers, the Jingou River and Kuytun River in alpine areas in Xinjiang, were selected in this study. Based on snowmelt runoff model, stepwise multivariate regression and nearest neighbor bootstrap regression, the daily runoff volumes of the two catchments were simulated after collecting the hydrological data, meteorological data and the snow cover data extracted from MODIS remote sensing imags. The results showed that the three models could be used to well simulate the runoff process of the Jingou River and Kuytun River catchments, Nash efficiency coefficients in both calibration and validation periods were higher than 0.7, and each model had its own advantages and disadvantages, which means that the snowmelt runoff model, stepwise multivariate regression and nearest neighbor bootstrap regression could be applied to predict flood in the Jingou River and Kuytun River catchments. The study results are of some guiding significance for preventing and earlywarning flood of the medium and small rivers in alpine area.

Key words: medium and small rivers, snowmelt runoff model, stepwise multivariate regression, nearest neighbor bootstrap regression, flood prediction, Xinjiang