Probabilistic forecasting model for annual runoff based on bootstrap

Annual runoff is influenced by many factors, and its forecasting value has some uncertainty. Probabilistic forecasting of annual runoff is more scientific and reasonable. A similarity forecasting model for annual runoff was established first, and then the bootstrap method was used to generate annual runoff sample sets on random re-sampling of model error. Based on the forecasting model and the re-sampling sample sets of annual runoff, the probabilistic forecasting model for annual runoff on bootstrap method was proposed, and named PFAR-B. The application results of the proposed model to forecast annual runoff of Yili River basin in Xinjiang show that the proposed model is of high precision, the observed value of annual runoff is contained in the 90% confidence interval of forecasting value of annual runoff, and the estimate of confidence interval is also robust. In view of its clear physical concept, easy implementation, and high prediction accuracy, the proposed model has application value in the hydrology and water resources prediction.