关键词: 水文学, 气候变化, 高寒区水资源, 核主成分支持向量机模型

Abstract: In this paper, time series of 49 years monthly runoff, temperature, precipitation and part of the permafrost monitoring data recorded at Jimai control hydrometrical station and Maduo and Dari weather stations in the upper Yellow River high cold region were studied. The annual variation, interannual variation, main influencing factors of river water resources, were investigated using peak pattern analytical method, rate change of wetness-dryness analytical method, Mann-Kendall test, wavelet analysis and grey relational analysis, etc. The river water resources regime in 2008-2014 were obtained by Kernel Principle Component Analysis and Least Square-Support Vector Machines prediction model. The results showed that river runoff demonstrates obvious evolutionary characteristics, the year 1961 is the first catastrophe point at which the runoff begins an upward tendency, it begins to reduce after 1982, until 1999 the second catastrophe point which brought a significantly downward tendency. The results of wavelet analysis and the trend prediction analyses show that the hydrological regime in this area will be little more than the 1990s which will last 3-9 years; after 1986, the maximum seasonal frozen depth reduced, which exacerbated the total annual runoff reduction in upper Yellow River. The prediction results indicate that the river water resources in 2011-2014 will reduce by 25.3% the annual average water resources.

Key words: hydrology, climate change, water resource in high cold region, KPCA-LSSVM model