基于不同立地质量的香格里拉高山松林郁闭度遥感反演

摘要

以香格里拉市为研究区,基于2006年TM遥感影像和2006年森林资源规划设计调查数据,以小班各遥感因子为自变量,每个小班的郁闭度为因变量,对香格里拉高山松林按区分立地质量等级和不区分地位级2种策略建立郁闭度神经网络遥感反演模型,并进行精度评价.基于小班平均高和平均年龄建立的地位级表将立地质量等级划分为好、中、差3种类型.研究结果表明,红外、近红外、植被指数和第一主成分等遥感因子对郁闭度的解释能力较强;从模型独立样本验证结果得出,不区分地位级郁闭度估测精度为67.64%,区分地位级后,好、中、差3种立地类型的郁闭度估测精度分别为74.14%、75.32%、72.38%,区分立地质量类型模型的精度优于不区分地位级模型的精度.%Taking Shangri-La City as the study area,the remote sensing factors of subcompartment as the independent variables,the canopy density as the dependent variable,this paper classified the site quali-ties and made indiscriminate site class to respectively establish the neural network inversion models for the canopy density of Pinus densat based on TM remote sensing image and forest resource planning and design data in 2006,as well as evaluated the accuracy of the models.The site qualities were divided into three levels of good,medium and poor in site class table according to the average height and the average age.The results showed that remote sensing factors such as infrared,near infrared,vegetation index,and first principal component,had stronger ability to explain canopy density.According to the validation re-sults of independent sample,the estimation accuracy was 67.64%regardless of site class,the estimation accuracy of good, medium and poor sites were 74.14%、75.32%、72.38% as differentiating site class,which indicated that the accuracy was better than that of the model regardless of site class.