基于气候的植被空间分布的数字模拟——以内蒙古为例

Climate-based digital simulation on spatial distribution of vegetation——A case study in Inner Mongolia

介绍了一种在 GIS支持下 ,基于气候的植被空间分布的数字模拟方法。研究以内蒙古为例 ,结果表明 ,内蒙古植被的空间分布与气候存在明显的相关性。以气候因子为变量 ,对植被类型的判别较好 ,地带性植被、山地植被、沙地植被和低湿地植被的判对率分别为 61 .82 %、64.4 4%、82 .72 %和 77.4 4% ,数字模拟的结果较好 ,k值为 0 .50。对植被地带的判别和模拟的准确性明显提高 ,判对率达 86.84 % ,k值为 0 .57,模拟结果属很好。判别和模拟的错误主要出现在因基质、地形、人类干扰等因素而产生的非地带性和次生、退化的植被中。该方法适于分析大尺度空间中植被地带与气候的关系。与传统的植被 -气候研究方法相比 ,所使用的植被和气候因子数字图象具备空间和数值上的连续性 ,结果客观且可以重复。使用精准的专题数据如数字遥感信息 ,以及引入更多的专题数据如土壤、植被生产力、样地资料等 。

In Inner Mongolia, previous researches showed correlation between vegetation zones and climate in early 1960s. Later, humidity data was used to describe zonation in detail. Since 1980s, vegetation climate relationship was studied deeply with the successful application of remote sensing on vegetation survey, and the range of humidity for main vegetation zones was modified. Moreover, multivariate analysis was also introduced to explain vegetation attributes quantitatively in climate space. The objective of this paper is to model vegetation type and zonation with climate factors by means of discrimination analysis, and to simulate spatial distribution of them digitally based on climate data under the support of GIS. Two sets of data, climate data and thematic maps, were used. Climate data was cited from 30 years observations of 156 meteorological stations in Inner Mongolia Autonomous Region. Among the 23 selected climate factors, 14 belonged to simple climate data (e.g. annual mean temperature). Others were indices derived from integrating two or more climate factors (e.g. Penman's aridity). Thematic maps were vegetation type map and vegetation zonation map. Firstly, to set up climate model with geographic elements. Correlative models of the 23 climate factors with longitude, latitude, and altitude were generated by mean values. Factors related to temperature changed linearly with geographic elements, and showed significant correlation. Factors concerning precipitation had nonlinear relationships with geographic elements, and models were more complicated. Digital images of climate factors were produced by integrating these models to images of geographic elements through GIS operation. Secondly, to recognize climate pattern of vegetation type and zonation. Climate data sets for vegetation types and zones, at first, were extracted by overlaying images of vegetation type and zonation with climate images. Then, these sets were re sampled by pseudo freedom sampling. Finally, Fisher functions using climate factors as variables were generated by stepwise discrimination analysis based on re sampled data. These functions were models of pattern recognition. Simulated spatial distribution of vegetation type and zonation was produced by integrating these models and related climate images. The agreement of actual and simulated distributions was identified by Kappa Statistic ( k ). Vegetation types were divided into 4 groups, i.e., zonation, mountain, sandy land and wetland type, and analyzed group by group. Total correct recognition rates for them were 61.82%, 64.44%, 82.72% and 77.44% respectively. Among the low rated types, some of the mis recognition was caused by changes of soil or local topography. In zonal vegetation, for example, correct recognition rate for rocky desert in low mountains and hills was 26.6%, and for Haloxylon ammodendron desert was 32.5%. In sandy land, only two types showed less than 90% of correct rate, i.e., Ulmus pumila open forest (39 8%), Caragana spp. and Salix spp. shrubs (34.3%). In wetland, mis recognition occurred in two types, i.e., swamp meadow (grass and sedge) and valley meadow (mesic forb and grass). Others were degraded or secondary types developed under human disturbances. They had the same climate condition as native ones, so it was difficult to distinguish them only by climate based pattern recognition. For example, Cleistogenes squarrosa and micro bunchgrasses steppe induced by overgrazing had low correct rate of 26 7%. In Daxinganling Mountains, most of Betula platyphylla and Populus davidiana forest was wrongly recognized as Larix gmelini forest (27.2%), Corylus spp. and Lespedeza bicolor shrub (25.3%), or Quercus mongolicus forest (15.1%). 33.8% of Larix gmelini forest was wrongly recognized as Corylus spp. and Lespedeza bicolor shrub. Results of discrimination on vegetation zonation were much better than that on vegetation type, Total correct recognition rate was 86.84%. Zones with correct rate over 90% accounted for 6