The Research On The Spatial Variability Of Forest Soil Nutrients

This research analyzed the spatial variability of forest soil nutrients and forecasted forest soil nutrient distribution in Yuncheng and Yunan's district. Using traditional statistics and geostatistics method combined with GIS technology to predict the distribution patterns of forest soil nutrients spatial variation. The main traditional statistics were described from three different point of view including descriptive statistics, soil nutrients' classification and soil nutrients background; geostatistics mainly described the multiple linear regression model, spatial interpolation model and back-propagation artificial neural network model. Besides, the root mean square error(RMSE) and the mean error(ME)were also used to assess the model accuracy. The model with higher correlation coefficients between prediction data and reference, and lower RMSE and ME was consider to be a successful model. As a result, BP-ANN was the best model to describe spatial distribution characteristics of soil nutrient content and nutrients' classification. This study aimed to improve the entire Yunfu City forest soil nutrient utilization efficiency, and to improve fertilization management, which are important to the soil nutrients' variation quantitative research and the implementation of precision forestry. The conclusions were as followed:(1) Coefficient of variation(CV) of all index belonged to medium variation. CV of soil nutrients in Yuncheng and Yunan districts were as followed: soil organic carbon(SOC)>total potassium(TK)>(total nitrogen)TN>total phosphorus(TP). The content of SOC in Yuncheng and Yunan districts was twice as much as the background value of SOC, while the content of TN and TP were far lower than the background value of nitrogen and phosphorus, and the little difference observed between TK and background value. Generally, soil nutrients' level in Yuncheng and Yunan districts were in the middle. The content of SOC and TK were both in middle level while the content of TN and TP were both at the state of deficiency.(2) IDW interpolation's correlation coefficient reached an extremely significant level(P<0.01) in the calibration data, which was also the best in spatial interpolation model.However, the correlation coefficient of spatial interpolation model in the validation data were all much lower than that in the calibration data, which indicated that the spatial interpolation model were not perform well at predicting distribution of soil nutrients.Further more, the average accuracy fell between the calibration data and validation data were(Universal Kriging) UK(3.9%) <SPLINE(80.21%) <(Inverse Distance Weighted)IDW(81.96%), while for the BP-ANN model, average accuracy increased 35.78%, and that could also explain why BP-ANN model was the only model which can better predict the spatial distribution of soil nutrients.(3) From the view of the predicted distribution of soil nutrients content, the content of SOC increased from the northeastern to southwestern area of Yuncheng and Yunan districts,of which mainly classified as three or four level in nutrient. At the same time, the trend of TN distribution was similar with SOC distribution, but it was mainly classified as four or five levels, relatively low in nutrient. Besides,the content of TP was mainly classified as five levels which was in a poor state of nutrient, with the tendency of being high in the northeastern area and southwestern area, whereas rather low in the middle area. The content of soil TK was increased from the northeastern to southwestern area, which was classified as 3 or 4 levels, relatively moderate in nutrient, and it was rich in the southwestern part of the forest.