基于气象要素空间分布模拟优化的中国草地综合顺序分类

摘要

依据草地综合顺序分类系统(CSCS)原理,利用中国区域1961-2004年的气象信息数据,以及1km分辨率的DEM(高程)数据,借助ArcGIS平台,采用优化的气象要素模拟方法——多元回归+残差分析(AMMRR),以及3种传统插值方法,模拟了中国多年平均的＞0℃年积温(∑θ)和年均降水量(r)的空间分布,并利用AMMRR法对中国草地进行分类.研究结果表明:1)AMMRR模拟得到的∑θ和,与实测样本的相关系数分别为0.976和0.974,极显著相关(p＜0.01)；相对平均误差(RME)、平均绝对误差(MAE)和均方根误差(RMSE)部小于3种传统插值方法.2)AMMRR通过经纬度、海拔高度与气象要素间的多元回归以及残差分析修正,不仅弥补了原始站点不足且分布不均匀的缺点,而且也充分体现了海拔落差较大区域气象要素的垂直变化.气象站点分布密集的区域,4种方法模拟的效果较一致,其空间格局与真实地理环境相似；但站点稀疏而且分布不均匀的区域,只有AMMRR的模拟结果才能反映出小尺度空间分布中地形的空间分异作用,结果比较理想.3)依据CSCS,中国草地包括除炎热极干热带荒漠类(ⅦA7)的其他41个地带性草地类.从南到北,Σθ减小,依次分布着炎热潮湿雨林类(ⅦF42) -亚热潮湿常绿阔叶林类(ⅥF41) -暖热潮湿落叶、常绿阔叶林类(ⅤF40) -暖温潮湿落叶阔叶林类(ⅣF39):从东向西,r降低,依次分布着微温潮湿针叶阔叶混交林类(ⅢF38) -微温湿润森林草原、落叶阔叶林类(ⅢE31) -微温微润草甸草原类(ⅢD24) -微温微干温带典型草原类(ⅢC17),地带性规律明显.%According to Comprehensive and Sequential Classification System of Rangeland (CSCS) and ArcGIS platform, an improved interpolation method named analytic method based on multiple regression and residues (AMMRR), and 3 traditional interpolation methods, in conjunction with 1 km resolution DEM and meteorological data from 1961 to 2004 in China as main data sources, were applied in simulating the spatial distribution pattern of annual accumulative temperature (>0°C Iff) and annual precipitation (r), and carrying out classification for Chinese rangeland with AMMRR after the comparison among 4 interpolation methods. The results showed that: 1) The correlation coefficents between simulative and observational value of Σθ and r were 0.976 and 0.974 with highly significant correlation (p<0.01) based on AMMRR, respectively. Its relative mean error (RME), mean absolute error (MAE) and root-mean-square error (RMSE) were entirely lower than traditional interpolation methods. 2) AMMRR method can remedy the shortcoming and heterogeneous distribution of meteorological stations, and fully reflect the vertical variation of meteorological factors in the regions of big net height via modifying residual errors between observational and simulative value and implementing multiple regression among meteorological factors and longitude, latitude, elevation. In the region with densely meteorological stations, the simulative effects of 4 simulative methods were highly agreement with each other, and the spatial distribution pattern of Z8 and r were similar with their actual geographical conditions. Compared to the region of sparsely and heterogeneously meteorological stations, only AMMRR method with ideal interpolation effect can reflect the spatial heterogeneity of topography on the spatial distribution pattern of SO and r in the small scale regions. 3) Associate with CSCS, Chinese rangeland could be classified into 41 classes exception of Tropical-extrarid tropical desert(ⅦA7). Subsequently Tropical-perhumid rain forest(ⅦF42), Sub-tropical perhumid evergreen broad leaved forest (VIF41), Warm-perhumid deciduous-evergreen broad leaved forest (VF40), Warm temperate perhumid deciduous broad leaved forest (IVF39) and Cool temperate perhumid mixed coniferous broad leaved forest (ⅢF38), Cool temperate-humid forest steppe, deciduous broad leaved forest (ⅢE31), Cool temperate-subhumid meadow steppe(ⅢD24), Cool temperate semiarid temperate typical steppe (ⅢC17) distributed with the decreasing Σθ and r from south to north and from east to west, which indicated significantly zonality patterns.s