基于Landsat时间序列数据的祁连山区域土地利用变化

发布时间：2018-05-09 07:53

本文选题：Landsat时序数据影像 + 长时间序列　；参考：《北京工业大学学报》2017年05期

【摘要】：为了研究祁连山地区土地利用变化情况,基于祁连山区域1986—2015年的Landsat时间序列数据,通过相对辐射校正获取时序地表反射率数据.采用光谱扩展与基于回归树的决策树分类(CART)获取规则的决策树分类方法,应用于长时间序列卫星影像,对各类土地利用类型近30 a的变化情况进行分析.结果表明:相对辐射归一化能有效减少时间序列数据之间光谱值差异,基于CART获取规则的决策树分类方法具有较高的分类精度.以2012年分类结果为例,总体分类精度为88.72%,Kappa系数为0.86,并分析了可能存在的误差.研究区耕地、林地和草地面积总体呈下降趋势发展,并且草地破碎化程度加剧,戈壁面积增多,植被退化导致土地荒漠化问题更加严重.最后,根据研究区土地利用变化情况进行讨论,并针对该情况提出建议.
[Abstract]:In order to study the land use change in Qilian Mountains, based on the Landsat time series data from 1986 to 2015, the time series surface reflectance data were obtained by relative radiation correction. The decision tree classification method based on spectral expansion and decision tree classification based on regression tree is applied to long-time series satellite images to analyze the changes of various land use types in the past 30 years. The results show that relative radiation normalization can effectively reduce the spectral difference between time series data, and the decision tree classification method based on CART rules has higher classification accuracy. Taking the result of classification in 2012 as an example, the overall classification accuracy is 88.72 and Kappa coefficient is 0.86, and the possible errors are analyzed. The area of cultivated land, woodland and grassland in the study area generally showed a downward trend, and the degree of grassland fragmentation increased, the Gobi area increased, and the degradation of vegetation led to a more serious problem of land desertification. Finally, according to the situation of land use change in the study area, some suggestions are put forward.
【作者单位】：重庆交通大学建筑与城市规划学院;中国科学院遥感与数字地球研究所数字地球重点实验室;重庆交通大学水利水运工程教育部重点实验室;中国科学院遥感与数字地球研究所国家遥感应用工程技术研究中心;长丰县杨庙中学;
【基金】：国家自然科学基金资助项目(41571423) 国家重点研发计划资助项目(2016YFB0501501) 中国科学院青年创新促进会资助项目国土资源部地学空间信息技术重点实验室开放基金资助项目(KLGSIT2015-07) 重庆交通大学国家内河航道整治工程技术研究中心暨水利水运工程教育部重点实验室开放基金资助项目(SLK2014B04)
【分类号】：F301.2;TP79

【相似文献】