塔里木河流域气候变化及未来趋势预估

Observed and Projected Climate Changes in Tarim River Basin

根据塔里木河流域1961—2008年39个气象站观测气温和降水量数据，对流域近50a气候变化进行了分析，并对参与IPCCAR4的17个气候模式在塔里木河流域气候模拟能力进行了评估．结果表明：近50a气温和降水时间序列分别在1986和1996年发生突变．1996年后塔里木河流域四季均呈现变暖趋势，最为明显的是夏季气温变化．从突变前后的降水概率分布变化来看，除秋季无明显变化外，其他3个季节均有所变化，其中，夏季表现为平均降水量增加较为明显，而春季和冬季降水变化则主要表现在方差增加．气候模式在塔里木河流域模拟能力评估表明，所有模式对降水量的模拟均偏高，而对气温模拟均偏低．各模式可以再现塔里木河流域气温年内变化，但普遍对降水的年内分布模拟有所提前，月降水峰值出现在4月和5月．采用模式加权平均数据对2011—2050年的气温变化进行预估，结果表明塔里木河流域年平均气温在不同情景下均呈上升趋势，上升幅度在0．5～2．4℃之间．

Climate changes in the Tarim River basin during 1961--2008 were studied by using the mete- orological data from 39 stations. Annual and monthly precipitation and temperature series derived from 17 climate models were compared with available observations. Abrupt change points were detected in the observed precipitation and temperature series in 1986 and 1996, respectively. According to the probability variation of different seasons, a warming tendency can be seen for the four seasons since 1996, especially in summer. With regard to the precipitation probability variation before and after the abrupt change, variation can be seen in spring, summer and winter, but no evident variation in autumn. In summer, the probability variation is characterized by distinct increase in average precipitation, but by increase in variance in spring and winter. The result of model assessment shows that there is obviously correlation between observed and simulated annual mean temperature. However, only five models predicting annual mean air temperature and one model predicting annual mean precipitation able to reach 95% confidence level. Most of the models predict precipitation on the high side and predict air temperature on the low side. Excellent reproducibility of the inner cy- cle of air temperature can be found in every model, but the modeled monthly precipitation distribution is in advance, with the top precipitation mostly in April and May. A general air temperature increasing tendency can be seen for all scenarios during 2011- 2050, projected by weighted average of modeled results.