中国台站冬夏季气温、降水的气候变化特征及其显著性检验

利用1951—2010年中国160站气温、降水资料，分析中国代表性台站冬季和夏季气温、降水的气候值及气候变率在前后30 a的差异，并对结果使用不同方法进行显著性检验。结果表明，季气温气候平均值的变化总体与全球增暖一致，以升温为主，但夏季在秦岭以南及长江中游地区出现显著局部变冷现象；季气温气候变率的变化相对较小，冬季总体不显著，夏季仅有少数台站显著。降水的气候变化总体不明显，季降水气候值变化的空间分布复杂，冬季南方地区、夏季东部地区总体增加，冬、夏季降水气候变率的变化均不显著。理论检验方法（t检验、F检验）与随机模拟方法（EMC法）的显著性检验结果，对气温的差别较小、对降水的差别较大，这与样本距平序列是否服从正态分布有关。EMC法可在确保样本统计特征不变的情况下，通过多次随机模拟，无需考虑其理论统计分布特征，使检验结果更为可靠。

Climate change characteristics and significance tests of temperature and precipitation of China weather stations in winter and summer

Based on the temperature and precipitation data of 160 stations in China from 1951 to 2010,this paper analyzed the climate average and climate variability of temperature and precipitation in winter and summer at the representative stations in China,as well as their differences between the former and later 30 years.Furthermore,the differences are tested by two different types of significance test methods.Results show that the change of climate average of seasonal temperature is generally consistent with the global warming,mainly warming up,but a significant local cooling phenomenon occurs in the south of Qinling Mountains and the middle reaches of the Yangtze River in summer.The change of climate variability of seasonal temperature is relatively small,and the overall change in winter is not significant,and only a few stations change significantly in summer.The climate change of seasonal total precipitation is not obvious on the whole.The spatial distribution of seasonal precipitation climate value changes is complex,and it increases in southern China in winter and eastern China in summer,while the changes of climate variability of winter and summer precipitation are not significant.For the differences of significance test results between the theoretical test method (t-test and F-test) and the stochastic simulation method (EMC (empirical Monte Carlo) method),they are small for the temperature and relatively large for the precipitation,which is related to whether the sample anomaly series obeys the normal distribution.Under the condition that the statistical characteristics of the samples remain unchanged,the EMC method can make the test results more reliable by multiple random simulations without considering the theoretical statistical distribution characteristics of the samples.