利用共同经验正交函数(EOF)分解和逐步线性回归相结合的统计降尺度方法,研究了1月和7月华东地区41个气象观测站2070—2100年未来月平均温度变化情景的集合预估。同时采用850 hPa温度场、850 hPa位势高度和温度的联合场以及海平面气压和850 hPa温度的联合场作为预报因子变量场,对于两个场联合的预报因子变量场,采用的是两个变量场空间联合的EOF分解的方法。同时通过改变统计降尺度过程中输入的预报因子变量场、预报因子变量取值的区域,以及输入逐步线性回归方程的主分量个数共建立27种统计降尺度模型,并把它们应用于2种全球气候模式(GCMs):Echam5和HadCM3 IPCC AR4 20C3M和A1b情景,从而每个站点均生成1950—2099年(HadCM3)或1951—2100年(Echam5)1月和7月共54个IPCC TR4 A1b温度变化情景,然后对54种预估情景进行集合分析。多个温度变化情景的集合预估采用它们的中位数来表示。结果表明:(1)当前气候条件下,多个统计降尺度结果的集合预报如采用箱线图的中位数能够在一定程度上提高统计降尺度方法的模拟性能;(2)2070—2100年1月和7月未来气温情景相比当前气候条件的增温约3~4℃,7月与1月相比不确定性增大。
This study introduced how to use stepwise linear regressions with common empirical orthogonal functions(EOF) to statistically downscale temperature in January and July of Eastern China.The predictor fields included temperature on 850 hPa(T 850),and the combination of geo-potential heights on 850 hPa(T 850+ T 850),and the combination of sea level pressure and temperature on 850 hPa(SLP+ T 850).For the combined predictors,EOF analysis with the two fields combined was used.IPCC AR4 HadCM3 and Echam5 IPCC AR420C3M and A1b scenarios were applied to the statistical downscaling models to construct local present and future climate change scenarios for each station during 1950 —2099(HadCM3) or 1951 —2100(Echam5).The 54 present and future climate scenarios have been obtained due to the 27 different predictor choices used.The results showed that(1) By the use of climate model ensemble predictions,it may therefore be possible to improve the skill of prediction to some degree,(2) Under IPCC AR4 A1b scenario,compared to the present climate,there will be an increase of 3~4℃ in temperature change in January and July in 2070 —2100.