运用地理加权泊松模型估测天然次生林枯损量分布

运用帽儿山实验林场2004年、2016年2期固定样地(共101块)数据,采用全局泊松模型和地理加权泊松模型(GWPR)对天然次生林枯损株数的空间分布进行了研究,利用Arc GIS软件绘制2种模型关于天然次生林枯损株数的空间分布图。结果表明:局域模型(GWPR)的赤池信息准则(AIC)值和均方误差(MSE)值显著小于全局泊松模型,拥有较高的拟合精度;局域模型残差的全局莫兰指数值在0上下浮动,变动程度明显小于全局模型,全局空间自相关性大大下降;局域模型的局域空间自相关性显著减小,形成了不同观测值少量聚类这一理想的空间分布模式;在对枯损株数空间分布的模拟上,局域模型的拟合偏差要小于全局模型。

Distribution of Natural Secondary Forest Mortality by Geographically Weighted Poisson Model

With 101 permanent sample plots of Maoershan Experimental Forest Farm from 2004 to 2016, the spatial distribution of natural secondary forest mortality were studied by using the global Poisson model and Geographically Weighted Poisson Regression ( GWPR) , the fitting effect and spatial autocorrelation of the two models were analyzed, and finally based on ArcGIS software, the global Poisson model and the GWPR model were used to simulate the spatial distribution of the num-ber of natural secondary forest mortality.The AIC value and MSE value of local GWPR model were significantly less than those of global Poisson model, the GWPR model had a high accuracy of model fitting, the global Moran I value of the local model residuals was floated up and down at 0, and the degree of change was also smaller than that of overall Poisson mod-el, that was the spatial autocorrelation of model residuals was greatly decreased.The local spatial auto-correlation of GWPR model was significantly reduced, which formed an ideal spatial distribution pattern of small clusters with different observa-tions.In the simulation of the spatial distribution of the number of forest mortality , the fitting error of the local model was smaller than that of the global model.