Chengde Bureau of Environmental Protection of Hebei Chengde,Chengde Bureau of Environmental Protection of Hebei Chengde,Key Laboratory of Forest Ecology and Environment of State Forestry Adinistration Institute of Forest Ecology,Environment and Protection,Chinese Academy of Forestry Beijing
Biodiversity is the basis of ecosystem functioning. Species richness has been widely used in biodiversity studies. Understanding the spatial distribution of species richness at landscape scale is vitally important in biodiversity conservation and natural resources management. Predicting species richness on large scale could help managers to rationally conserve and utilize natural resources. With the availability of remotely sensed data and the development of geographical information system (GIS) techniques spatial analysis on large scale has been possible. Integrating field sample plot investigation, remote sensing (RS), and geographic information system is a novel way to explore the distribution of species richness at macro spatial scales. The Altai Mountains is one of the magnificent Mountains of Asia, which distributes across Mongolia, China, Kazakhstan, and Russia. In this study, we adopted the above mentioned approach to predict the spatial distribution of tree species richness in Xiaodonggou forest region of the Altai Mountains in Xinjiang, Northwest China. In the south and north slope of the Xiaodonggou forest region, a investigation transect was selected respectively. In each of the transect, we set investigation plots (each was 20 m×20 m in size) at intervals of 50 m along the altitude. All woody plants in the plots with diameter at breast height (DBH)≥1cm were identified and measured. The species richness in each plot was calculated. Normalized difference vegetation index (NDVI) was obtained from ETM+ image. In order to overlay ETM+ image and topographic factor maps, we selected ETM+ image in size of 30 m×30m. The predictor variables include climate, topography, and NDVI. Principle component analysis (PCA) and multiple linear regression were firstly utilized for obtaining the environmental factors and developing the predictive model of species richness distribution. Annual minimum temperature, annual average relative humidity, aspect, slope and NDVI were selected into the predictive model. Tree species richness distribution map was produced by GIS. The residual map was produced by the inverse distance weighted interpolation (IDW) method. The residual map was used to evaluate the validity of the model. In order to analyze variation of species richness with the topographic factors, the spatial distribution map of species richness was overlaid with the slope, aspect and elevation maps, respectively. The results showed that the areas with 3-4 tree species occupied 70.08% of the total study region. In slopes of 0- 5°, the areas with tree species richness of 3 had the highest presence frequency, while in slopes of other ranges, the areas with tree species richness of 4 had the highest presence frequency. In west and northwest aspects, the areas with three tree species had the highest presence frequency, in the other aspects, however, the areas contained four tree species had the highest presence frequency. Along with the altitudinal gradient, the tree species richness showed a unimodal distribution pattern, which is consistent with the hypothesis of mid-domain effect. The statistic results of residual types showed that strongly predicting area and moderately predicting area together reached 94.62% of the total study area, which implied that our predictive model was robust and could be successfully implemented in this forest region.