基于Sentinel-2数据的火烧迹地提取研究

Study on burned area extraction based on Sentinel-2 data

昆明为云南省的政治经济文化中心,森林火灾多发给人民生产生活带来了极大的威胁。准确提取火烧迹地有利于掌握林火规律,加强森林火险区划,通过合理配备资源提高森林防火效率。以昆明地区2016—2020年Sentinel-2数据、高温火点数据和土地利用分类数据为基础,基于差分归一化火烧比率(dNBR)提取该地区火烧迹地面积,并对其空间分布特征进行分析。结果表明:土地利用数据有利于提高火烧迹地提取效率,但高温火点数据错漏较多,对火烧迹地提取贡献不大。Sentinel-2数据与dNBR指数的结合在单火场可以达到平均89%以上的提取精度,有较高适用性。昆明地区火灾高发区为安宁市、晋宁区等地,过火区域集中在高程1700~2300 m、坡度6°~25°的林地耕地混杂地区,阴坡过火面积略大于阳坡。

Kunming is the political,economic and cultural center of Yunnan Province,and forest fires have occured frequently in Kunming and brought great threats to people’s production and life.Accurate extraction of burned area is beneficial to master forest fire law,strengthen forest fire risk zoning,and improve forest fire prevention efficiency through rational allocation of resources.Based on Sentinel-2 data,high temperature fire point data and land use classification data from 2016 to 2020 in Kunming,this paper extracted the burned area based on differential normalized fire ratio(dNBR)and analyzed its spatial distribution characteristics.The results showed that the land use data was beneficial to improve the extraction efficiency of fire site,but the data of high temperature fire site had many errors and omissions,which made little contribution to the extraction of burned area.The combination of Sentinel-2 data and dNBR index can achieve an average extraction accuracy of more than 89%in a single fire site,showing high applicability.The high-incidence areas of fire in Kunming were Anning City and Jinning District,and the fire areas were concentrated in the mixed areas of forestland and farmland with elevation of 1700~2300 m and slope of 6~25 degrees,the area of fire on shady slope was slightly larger than that on sunny slope.