Abstract: Precipitation forecasting in arid region is of great significance for water resources utilization and drought disaster reduction. A precipitation forecasting model in yearly and monthly scales based on particle swarm algorithm (PSO) and least squares support vector machine (LSSVM) model was established using the annual precipitation sample of a seven-year cycle and the monthly precipitation sample of seasonal characteristics. The applicability of the model was verified through the measured precipitation sequence from 1960 to 2013 in Altay region. Results show that the model based on PSO and LSSVM could effectively forecast the annual and monthly precipitation in Altay region, hence is of high precision and strong generalization ability. It offers a reliable research idea and method for medium and long-term precipitation forecast in arid areas.

Key words: PSO, LS-SVM, arid areas, Altay region, precipitation forecasting