基于混合像元分解的香格里拉市高山松空间分布变化研究

利用香格里拉市1997-2017年每5 a的Landsat遥感影像、2018年外业调查数据、森林资源二类调查数据和DEM数据为数据源，以高山松为对象，综合运用混合像元分解技术、决策树分类和GIS技术对其空间分布变化进行分析。结果表明，1）分类中，高山松归一化多分量指数阈值为0.333，云南松归一化多分量指数阈值为0.208；云冷杉归一化多分量指数阈值为0.362。2）各年分类结果总体精度分别为69.42%、76.73%、81.07%、78.90%和76.53%。3）香格里拉市高山松覆盖面积2002年比1997年减少了13.40%，2007年比2002年减少了2.47%，2012年比2007增加了8.96%，2017年比2012年增加了4.06%，呈现出先减少后增加的趋势。4）研究区内高山松主要分布在海拔高度2 800~3 800 m，1 800~2 800 m高山松面积总体呈现为轻微下降趋势；2 800~3 300 m海拔区间1997-2007年呈现下降趋势，2007-2017年呈现上升趋势，3 300~3 800 m区间内1997-2002年高山松面积呈现下降趋势，2002-2017年高山松面积逐年上升，3 800~4 800 m区间内1997-2017年高山松面积逐年下降。利用混合像元分解构建归一化多分量指数结合决策树分类对树种分类具有一定的参考价值，高山松时空变化结果对森林资源管理和后续研究可提供科学数据支撑。

Spatial Distribution Changes of the Pinus densata Forests in Shangri-La City Based on Mixed Pixel Decomposition

In this study,several data resources were adopted:Landsat remote sensing images of the Shangri-La City collected every five years from 1997 to 2017,field survey data in 2018,forest resource second-class survey data,and DEM data.Pinus densata forests occurring in Shangri-La was taken as the research objects.Different technologies were comprehensively used to extract relative data of P.densata to conduct the special distribution variations investigation,including mixed pixel decomposition technology,decision tree classification,and GIS technology.The following results were obtained:1)in the view of classification,the threshold of normalized multicomponent index of P.densata was 0.333,while it was 0.208 for P.yunnanensis,and 0.362 for spruce-fir.2)The values of overall classification accuracy of the study years,i.e.,1997,2002,2007,2012,and 2017 were 69.42%,76.73%,81.07%,78.90%and 76.53%,respectively.3)The coverage area of P.densata in Shangri-La decreased by 13.40%in 2002 compared with 1997,decreased by 2.47%in 2007 compared with 2002,increased by 8.96%in 2012 compared with 2007,and increased by 4.06%in 2017 compared with 2012.4)P.densata was mainly distributed in the areas with altitudes of 2800 to 3800 m,and the area with altitudes of 1800 to 2800 m decreased slightly from 1997 to 2017.The area with altitudes of 2800 to 3300 m showed a downward trend from 1997 to 2007,increased from 2007 to 2017.The area with altitudes of 3300 to 3800 m showed a downward trend from 1997 to 2002,and increased from 2002 to 2017.The area with altitudes of 3800 to 4800 m decreased year by year from 1997 to 2017.In a word,using mixed pixel decomposition to construct normalized multi-component index combined with decision tree classification demonstrated certain referential values for the classification of tree species.The spatiotemporal changes of P.densata provided scientific data support for forest resource management and follow-up research.