中国区域地理、地形因子对降水分布影响的估算和分析

Estimation and analysis for geographic and orographic influences on precipitation distribution in China

不同与以往基于最小二乘的多元线性回归方法,本文首次尝试将新型的第二代回归分析方法——偏最小二乘回归分析方法应用到中国区域的降水建模中.利用区域内394个气象观测站建站到2000年45年(及以上)的降水资料,建立了一个简单的年、季降水量和地理、地形因子(包括纬度、经度、地形高程、坡度、坡向和遮蔽度)的关系模型,估算了区域降水量中地理、地形的影响部分,并分析了这种影响的特征.结果表明,用此方法建立的模型能够解释70%以上的因变量的变异,相关系数基本都在0.84以上,经交叉有效性检验,模型的回归效果较显著.分析表明,在多元线性回归不适用的情况下,本文基于偏最小二乘法的简单模型能够比较准确地定性、定量地再现实际降水分布.

Different from the framework of multiple linear regression based on least square method, tries to apply a novel second-generation regression method based on partial least-squares to p estimation in China for the first time. The 45-yr precipitation data from 394 meteorological stations in are used. Several simple formulae used to estimate the annual mean and seasonal precipitation obtained, and the characteristics of the this paper recipitation study area have been c effects have been presented. The impact factors include longitude, latitude, height, slope, sloping direction and close limit. The results show that the fraction of the variation of response explained by the model is above 70%, and the average correlation coefficients are nearly all above 0.84. The results are satisfied through the test of cross-validation. Through it is not appropriate to set up multiple linear regression models, the estimated precipitation based on partial leastsquares regression correctly replicates real spatial distribution of precipitation qualitatively and quantitatively.