Study On Estimating Methods Of Forest Stem Volume Based On Multi-source Remote Sensing Image

Forest land area accounts for approximately one third of global land area. As the most abundant organic carbon sink on earth, forest is one of the most important part controlling the energy transmission of terrestrial biosphere. Accounting for about 90 percent of the terrestrial ecosystem biomass, forest biomass is not only an important standard of carbon fixation, but also an important parameter of assessing carbon budget. Additionally, forest biomass plays a role in global carbon cycle. Remote sensing is the small or large-scale acquisition of information of an object or phenomenon, by the use of either recording or real-time sensing device(s) that are wireless, or not in physical or intimate contact with the object. Nowadays, remote sensing has become a main means of forest management. Remote sensing can be classified into passive and active modes which are both important to inversion of each forest stand parameter. By acquiring the spectral reflectance information from forest stand and using the information to identify the forest or extract the physical parameters (e.g. Vegetation coverage, LAI) and chemical parameters (e.g. Chlorophyll a, N/P, APAR), optical remote sensing can be well used in vegetation study. Compared with optical remote sensing, microwave remote sensing, longer wavelength microwave radiation can penetrate through cloud cover, haze, dust, and all but the heaviest rainfall as the longer wavelengths are not susceptible to atmospheric scattering which affects shorter optical wavelengths. This property allows detection of microwave energy under almost all weather and environmental conditions so that data can be collected at any time. To forest, longer wavelength microwave radiation can penetrate through the vegetation deeper into canopy and trunk. So the information acquired by microwave remote sensors is surface and volumetric scattering information, which has more advantages in research on forest biomass inversion. As it is difficult to measure the large scale forest biomass, the biomass data is often acquired from National Forest Inventory by regression relationship between biomass and forest stem volume.Based on the optical images and SAR (synthetic aperture radar) images, the thesis studies several methods to inverse forest stem volume which is located in Wangqing forestry bureau. The main works and results are as follows: 1. Forest stem volume estimation based on LANDSAT ETM+and SPOT-5 images using multi-regression model. Combining with the fixed plot data, the study uses the spectral reflectance 16f each band and one vegetation index to produce a multi-regression model of each dominant tree species of each optical image. The results show that the correlation coefficients of each multi-regression model are higher than that of each model produced by the single reflectance of each band. As the time of getting the fixed plot data and LANDSAT ETM+ image is a little different, the estimating results of LANDSAT ETM+image is not well as that of SPOT-5, image.2.Forest stem volume estimation based on two PALSAR images using logarithm regression model. SAR signal has a large range in plot level at each stem volume level. The study produces logarithmic models of each dominant tree species between backscatter coefficients and forest stem volume. The results show that HV coefficient has better correlation relationship with forest stem volume than HH coefficient.3. Forest stem volume estimation using k-NN algorithm (non-parametric algorithm). To the two optical images, the estimating bias results of all species are under 10m3/ha on the pixel-lever but the Fraxinus mandshurica Rupr and Populus ussuriensis. Compared with the relative root mean square error(RMSE') from multi-regression models, the RMSE'from k-NN is lower. The results based on optical images are the same with that based on two PALSAR images.4. Forest stem volume estimation based on two PALSAR images which has not been topographic correction pre-processed. Compared with the RMSE'from the two PALSAR images (which has been topographic correction pre-processed), the RMSE'from the two PALSAR images which are not topographic correction pre-processed is higher at forest farm scale, except two forest farms. The reason for this is because the area of the two forest farms covering by the two PALSAR images is very small.5. Forest stem volume synergy estimation based on SPOT-5 and PALSAR images. Combining with the fixed plot data, the study uses the spectral reflectance of every band, one vegetation index and HV backscatter coefficient to produce a multi-regression model of each dominant tree species. The results show that the correlation coefficients of each multi-regression model are higher than that of each model based on a single image. The RMSE' of the synergy model is lower than that of the each model based on a single image.