关键词: 气候变化, 克里格插值, Hopfield神经网络, 方差分析, 中国

Abstract: Using the daily precipitation, temperature and relative humidity data of 194 meteorological stations from 1961 to 2000 in China (not including Hong Kong, Macao, Taiwan and Hainan), based on the Spatial Kriging interpolation method in ArcGIS, and through the cluster of Hopfield neural networks as well as analysis of variance, the spatial and temporal characteristics of climate change in China.The results show that a dominant trend of warming occurrs in China, followed by the wet trend, drying trend occurrs in the eastern region, while the western region is gradually humidifying.Judging from the decadal results, we got that there was an equal trend between the trend of dry cooling and wet humid heat from 1960 to 1969; dry-type distribution from 1970 to 1979 was very broad, almost throughout the country; climate change types presented a layered distribution from 1980 to 1989, and according to the main types, warming trend was in the leading position; from 1990 to 2000, warming trend still dominated, followed by drying trend.We conducted clustering analysis using Hopfield neural network method, and test whether differences among the three kinds of climatic factors are significant using analysis of variance.In the final, according to the clustering results, we have analyzed the trend of precipitation, temperature and relative humidity change, results in line with interpolation show the space Kriging credibility.The results show that, driven by global warming, patterns of climate change in China is in a state of adjustment, and humid regions become arid, whereas arid regions humid.

Key words: climate change, Kriging interpolation, Hopfield neural network, analysis of variance, China