摘要
针对高山湿地遥感研究的复杂性,本文提出了一种基于气候分区、面向对象分层分割的高山湿地遥感提取方法。利用MODIS归一化植被指数(NDVI)时间序列及数字高程模型(DEM)数据,在气候分区的基础上采用面向对象分类方法,进行大兴安岭地区的湿地遥感提取,并分析了2001—2013年大兴安岭湿地的时空动态变化及成因。结果表明:基于MODIS影像综合特征的面向对象方法,比基于像元的最大似然法和Logistic模型法提取湿地信息的精度更高,面向对象方法提取大兴安岭湿地制图精度为82%,Kappa系数为0.74;2001年大兴安岭地区湿地面积为414.89×104hm^2,2013年湿地面积为340.39×104hm^2,13年间湿地面积总体呈减少趋势;不同地区湿地增减趋势不同,湿地增加区域主要分布在塔河县和呼玛县,是由于2013年降雨量高于2001年降雨量引起的,而湿地农田化导致分布在甘南县和龙江县的湿地减少。研究成果将为高山湿地时空动态信息提取提供技术支持。
As to the complexity of wetland remote sensing on high mountain wetlands, an objectoriented hierarchical segmentation based on climate division was proposed in this study to extract information of high mountain wetlands. Time series of MODIS normalized difference vegetation index （NDVI） and the data of Digital Elevation Model （DEM） were deployed to extract wetlands on the basis of climate division, and the temporalspatial changes and causes of wetlands in Daxing’an Mountains during 2001-2013 were quantitatively analyzed. The objectoriented method with characteristics of MODIS images showed better accuracy than maximum likelihood method and Logistic model with pixels in wetland extraction. The cartographic accuracy of Daxing’an Mountains reached 82% by the objectoriented method, and the Kappa coefficient was 0.74. The Daxing’an Mountains wetland area was 414.89 hm2 in 2001 and then decreased to 340.39 hm2 in 2013. The area of wetlands increased and decreased differently in various regions. Increased wetland area mainly distributed in Tahe and Huma counties due to higher annual rainfall in 2013, while decreased wetland area mainly concentrated in Gannan and Longjiang counties due to conversion of wetlands into cultivated lands. This study provides technical support for information extraction of spatial and temporal dynamics of mountain wetlands.
出处
《生态学杂志》
CAS
CSCD
北大核心
2017年第7期2068-2076,共9页
Chinese Journal of Ecology
基金
国家自然科学基金项目(41165005)
黑龙江省自然科学基金项目(D201414)
中国气象局气象关键技术集成与应用项目(CMAGJ2015M18)资助
关键词
大兴安岭湿地
NDVI时间序列
面向对象分类
空间插值
时空变化
Daxing＇ an Mountains wetland
NDVI time series
object-oriented classification
spatial interpolation
spatial and temporal variation.
