Study On Forest Soil Fertility By Remote Sensing Estimation In Yongan City

Woodland soil fertility is the foundation of forestry productivity. The monitoring ofwoodland soil fertility has great significance for the management of forest resources andthe enhancing of forest land productivity.This paper took the forest land in Yongan city asthe research object. And studied on the basic characteristics of the woodland soil fertilityby field investigation and soil samples for the area together with the soil physical andchemical experiments. This paper used principal component analysis and regressionanalysis method to evaluate the soil fertility and calculates the soil fertility comprehensivevalue by106research sample points. With3S technology and the soil survey of the ALOSremote sensing image in the same period, we selected evaluation factors on the basis ofgray relation theory. Then finally used the general regression neural network (GRNN) toconstruct soil fertility remote sensing estimation model and simulation output woodlandsoil fertility composite index figure in Yongan. Combined with the factors of altitude,aspect, slope, the forest types and so on, the analyses with dimensional overlay and thespatial distribution regularities were discussed in this paper. This dissertation’s maincontext are as follows:(1)This paper utilized ENVI ZOOM module to extract thematic information offorestry resources types in Yongan city based on the object-oriented. The results showedthat: The classification accuracy of Chinese fir, horsetail pines, eucalyptus, bamboo grove,economical forest, broad-leaved forest, shrubbery, open woodland were93.75%,83.75%,92.5%,90.28%,84.09%,80.00%,93.75%,100%respectively, the overall accuracy was86.54%, Kappa coefficient was0.8419.(2)The soil physical and chemical experiments of106research sample points weredetermined based on the soil samples from the field investigation during2008. Using ofcomprehensive loading method to determine the evaluation indexes:organic matter, total P,capillary water capacity, the K value, thickness of the soil layer, available p, soil bulkdensity. The soil fertility comprehensive value was measured based on comprehensiveindex measuring method in the paper.(3) The remote sensing image of band1, band2, band3, ratio vegetation index, theslope steepness factor and slope were taken as the independent variable with gray relationtheory factor. The soil fertility comprehensive value was taken as the dependent variable. Then we used the general regression neural network (GRNN) to construct Yongan city soilfertility remote sensing estimation model and take the corresponding remote sensingfactors as the simulation output to get the woodland soil fertility distribution graph ofYongan city. Utilizing the site quality level to verify the overall accuracy, which attained to72%.(4)Some results would come out after the analysis of Yongan city woodland soilfertility distribution. Results showed: the woodland soil fertility decreases with the increaseof elevation, decreases with the increase of gradient, slope had less influence on soilfertility. Woodland soil fertility had great difference between different tree species. Theminimum value of the average soil fertility composite index of economic forest was0.351,while the maximum value of the average soil fertility composite index of broad-leavedforest was0.383.