Spatial Distribution Of Forest Vegetation And Its Impact Factors In Heilongjiang Province

The forest vegetation is an important part of terrestrial ecosystem, with a variety of ecological functions of soil and water conservation, climate regulation, wind and sand protected plays an important role in the regional and global balance of energy metabolism and material circulation. So how fastly, accurately, high efficiently and real-timely to achieve forest resources information and the spatial distribution rules is beneficial to make scientific and effective protection also utilization of forest resources plans. It also has important significance to promote the sustainable development of the ecological system and maintain national and regional ecological security. With the development of "3S" technology, especially the appearance of high spatial resolution remote sensing image and new classification algorithm, it is possible to use remote sensing image to classify vegetation and study its spatial distribution. Taking Heilongjiang Province as the main research region, mainly using remote sensing (RS) and geographic information system (GIS) technology to classify the main vegetation types in Heilongjiang Province, and analysis the spatial distribution features of main forest vegetation types and how the topographic factors and meteorological factors influence the spatial distribution. The main research contents are as follows:(1)Taking landsat8 remote sensing images as the data source, fixed sample data and DEM data as ancillary data, through extracting the feature of spectral characteristics and the characteristics of the terrain to build random forest model classify the vegetation types of Heilongjiang province. Making comparison of classification accuracy with traditional maximum likelihood classification, then analyzes the applicability of the two kinds of classification methods for classification of forest vegetation. At the same time, to make choice of random forest model parameters and characteristic variables to realize the optimization of classification accuracy and classification speed.(2) Using ArcGIS as a platform, based on map of vegetation types after classification, overlay the map of elevation, slope, aspect, annual average precipitation and annual average temperature with the map of vegetation types to do spatial analysis. Then analyzing of spatial distribution characteristics of different vegetation types and how the terrain and climate factors influence the vegetation types.The results show that compared with the traditional maximum likelihood method, the overall classification accuracy of the random forest method is increased by 4.33%, and the kappa coefficient is improved by 0.8%. With producer’s accuracy coniferous forest, broad-leaved forest, coniferous and broadleaved mixed forest were improved by 2.04%,6.27%4.34%, user accuracy were increased by 0.92%,7.08%,5.62%. Thus, the random forests model based on vegetation index and terrain factor classification accuracy is significantly improved. So we know that random forest has better classification performance than the maximum likelihood classification method for vegetation types.Random forests classification process needs to set a few parameters artificially, only need to set the number of tree (ntree) and the number of variable (mtry). Through analysis the influence of classification accuracy, the parameters of ntree is set to 100 and the mtry is set to 3 has better classification accuracy. Then analysis the feature model of variable importance then the Band4, Elevation, NDVI, band5, the Band6 of the, band7, RVI, band3 and slope are selected as characteristic variables to construct random forest model is the best, not only improve the classification accuracy, shortens the time of operation model.Based on spatial overlay analysis:with increasing of elevation, the different types of forest vegetation coverage area have obvious increase, within a range of 400 is the most suitable for vegetation growth; with the increase of slope, vegetation coverage area is slightly rising trend, forest vegetation is concentrated on the flat and gentle slope, when the slope is greater than 45 degrees, vegetation coverage area is sharply reduced; aspect was negatively correlated to vegetation coverage area, vegetation growth in the sunny and semi sunny slope are more than in shady and half shady slope. The forest vegetation concentrated in the range 400-700mm of the precipitation, especially in the 500-600mm is optimal interval; the average temperature Heilongjiang Province was about zero degrees with the increase in mean annual temperature the vegetation coverage area increase too, with the range 0-4 is most suitable for vegetation growth.