摘要
研究和探讨了基于ANN的统计降尺度法,通过ANN建立大尺度气候观测资料和实测降水之间的统计关系,并同多元线性回归降尺度法进行了比较.结果表明,基于人工神经网络的统计降尺度法模拟精度优于多元线性回归法,可以应用其研究未来气候情景下汉江流域降水变化情况.通过对A2气候情景下全球气候模式HadCM3的尺度降解,预测未来2011-2100年汉江流域降水变化情况,最终发现汉江上游未来降水在2020s(2011-2040年)和2050s(2041-2070年)时期比基准年减少,2080s(2071-2100年)时期则比基准年增加;中游未来降水在2020s时期比基准年减少,2050s和2080s时期比基准年增加;下游未来降水变化趋势不明显.
To establish the statistical relationship between the larger scale climate predictors and observed precipitation in the Hanjiang River basin,a statistical downscaling method based on artificial neural network（ANN） was discussed and studied by comparing with multilinear regression（MLR）.It can be seen that ANN is superior to MLR and it is suitable for predicting the change of precipitation in the Hanjiang River basin in the future.Finally,the changes of precipitation,which projected from HadCM3 for A2 scenario were predicted during 2011 to 2100 by ANN.The results show that the precipitation will be reduced in 2020s（2011-2040） and 2050s（2041-2070）;but increased in 2080s（2071-2100） in the upper basin.In the middle basin,the precipitation will be decreased in 2020s,while increased in 2050s and 2080s.However,in the lower basin,the precipitation will be no significantly changed in these three periods compared with resent.
出处
《武汉大学学报（工学版）》
CAS
CSCD
北大核心
2010年第2期148-152,共5页
Engineering Journal of Wuhan University
基金
国家自然科学基金项目(编号:50679063
50809049)
教育部高等学校博士学科点专项科研基金项目(编号:200804861062)
关键词
气候变化
统计降尺度法
人工神经网络
降水
汉江流域
climate change
statistical downscaling method
artificial neural network
precipitation
Hanjiang River basin
