地理和地形影响下我国区域的气温空间分布  

Spatial distribution of temperature in China:Geographic and orographic influences

作　　者：舒守娟[1] 王元[1] 储惠芸[1] 

机构地区：[1]南京大学大气科学系中尺度灾害性天气教育部重点实验室,南京210093

出　　处：《南京大学学报（自然科学版）》2009年第3期334-342,共9页Journal of Nanjing University（Natural Science）

基　　金：国家科技基础条件平台建设项目"气象科学数据共享中心"(2005DKA31700);国家自然科学基金(40575017);国家重点基础研究发展规划项目(973(2004CB418301));江苏省自然科学基金(BK2005081)

摘　　要：不同于以往基于最小二乘的多元线性回归方法,本文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气温空间分布的建模中.利用区域内394个气象观测站建站到2000年45年(及以上)的逐日气温资料,建立了一个简洁的年、季平均气温和地理、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,估算了区域地理、地形影响下气温的空间分布,并分析了这种影响的特征.结果表明,此模型具有88%的平均拟合能力,平均相关系数达到0.94,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准确地再现我国实际气温的空间分布,具有一定的使用价值.Different from the previous framework of multiple linear regression based on least square method, this paper tries to apply a novel second-generation regression method based on partial least-squares to temperature estimation in China. The 45-yr temperature data from 394 meteorological stations in study area are used. Several simple formulae relating impact factors of longitude, latitude, heigh, slope, slope direction and close limit to the annual mean and seasonal temperature have been obtained, and the characteristics of the geographic or topographic effects have been presented. The results show that the model has an average fitting ability of 88%, and the average correlation coefficient comes to 0.94. The results are satisfied through the test of cross-validation. Through it is not appropriate to set up multiple linear regression model, the estimated temperature based on partial least-squares regression correctly replicates real spatial distribution of temperature.

关 键 词：气温空间分布 地形因子 地理因子 偏最小二乘回归 

分 类 号：P4[天文地球—大气科学及气象学]