威宁县松林生物量遥感估测研究

Remote Sensing Estimation of Pine Forest Biomass in Weining County

基于威宁县2017年森林资源规划调查数据与Landsat 8 OLI遥感影像对威宁县云南松林和华山松林分别构建BP神经网络和支持向量机(SVM)模型,并选择最优模型对威宁县2种松林进行地上生物量反演。结果表明:2种模型中,SVM模型有着最好的估测效果,云南松模型决定系数(R^(2))系数为0.409,均方根误差(RMSE)为39.04,云南松林单位生物量主要分布在3~30 t/hm^(2),其次在30~120 t/hm^(2),集中分布于威宁县西南部。华山松模型决定系数(R^(2))为0.35,RMSE为47.6,华山松林单位生物量主要分布在2~30 t/hm^(2),其次在30~150 t/hm^(2),集中分布在威宁县北部。

Based on the survey data of forest resource planning in 2017 and Landsat 8 OLI remote sensing images in Weining County,BP neural network and support vector machine(SVM)models were constructed for Pinus yunnanensis and Pinus armandii in Weining County,respectively,and the optimal models were selected to invert the above ground biomass of the two pine forests in Weining County.The results show that:Among the two models,the support vector machine(SVM)model had the best estimation effect.The R^(2) coefficient of model was 0.409,and the root mean square error(RMSE)was 39.04.The unit biomass of the Pinus yunnanensis was mainly distributed in the range of 3~30 t/hm^(2),followed by the range of 30~120 t/hm^(2).It is concentrated in the southwest of Weining County.The R^(2) coefficient of the model was 0.35,and the root mean square error(RMSE)was 47.6.The unit biomass of Pinus armandii was mainly distributed in 2~30 t/hm^(2),followed by 30~150 t/hm^(2),and concentrated in the northern part of Weining County.