| Abstract: |
Mapping is an important tool in studying and monitoring vegetation cover, particularly its spatial inventory, dynamics and biodiversity estimation. Vegetation maps are actually spatial vegetation cover models. But maps built using traditional methods become out of date very soon. Accurate vegetation maps are missing for the most part of Siberia. Nowadays, vegetation maps are successfully built and effectively updated using GIS technologies and remote sensing data.
A technique of automated mapping of forest regeneration dynamics was developed and applied to south central Siberia. This technique is based on GIS-technology involving spatial analysis of multi-band satellite data, digital elevation model (DEM), and ground observation data.
Automated mapping of forest cover, particularly its regeneration dynamics, is a challenging problem covering a number of tasks, which can be accomplished only through interdisciplinary research efforts. In this work, we have solved several key tasks in a stepwise manner to resolve this problem.
The first step was to develop an appropriate vegetation classification, upon which the map legend would be based. Classification of forest communities was performed using a topogenetic approach. According to the approach of Russian forest scientist Kolesnikov (1956), forest ecosystem diversity is formed by forest development stages present at the same time in a given area. The entire diversity of vegetation communities was classified on the basis of site condition similarity, and not on the basis of continuously changing outward characteristics (e.g., species composition). A preliminary classification of forest chronosequences with an account of site conditions was developed by analyzing thematic maps, literature and field data.
The second step was an automated classification and mapping of potential site conditions of a test area. Two-layer DEM-composite (elevation above sea level and slopes) was classified using ISODATA. Terrain roughness classes relatively similar in morphometric relief parameters were identified and interpreted with respect to geomorphology, zonal soil types, and vegetation. The electronic layer of potential site conditions was created as a basis of forest regeneration dynamics mapping.
The next step was an automated classification of multi-band satellite images. Landsat 5-TM images were classified by the method of maximum likelihood to identify land cover classes based on spectral characteristics. 50 classes of land cover were obtained and 20 were interpreted as forest cover classes and forest regeneration stages of different age. The forest vegetation chronosequences were formed from their age stages in the range of site conditions using their preliminary classification.
On the final step, expert system for classification and mapping of forest site conditions and forest regeneration dynamics was developed using Knowledge Engineer module / ERDAS Imagine. As a result, raster and vector polygonal layers were built. The maps incorporated in the GIS database reflected the distribution of forest vegetation chronosequences and regeneration stages in the range of site conditions. They can be used to model forest succession in similar landscape types within central Siberia. |