Determining the use of Sentinel-2A MSI for wildfire burning & severity detection
Determining the use of Sentinel-2A MSI for wildfire burning & severity detection
复制标题
DOI：
10.1080/01431161.2018.1519284
复制
发表时间：
2018-10
影响因子：
3.4
通讯作者：
C. Amos;G. Petropoulos;K. P. Ferentinos
中科院分区：
文献类型：
--
作者：
C. Amos;G. Petropoulos;K. P. Ferentinos
准确、可靠、及时的烧伤严重程度地图对于野火后的规划、管理和恢复是必要的。这项研究的目的是评估哨兵-2A卫星探测燃烧区和区分燃烧严重程度的能力。它还试图测量不同波段的光谱可分离性，并推导出通常用于探测燃烧区的指数。还对烧毁景观中存在的相关环境变量进行了简短的调查，以探讨是否存在任何相关性。作为一个案例研究，野火发生在塞拉利昂德加塔地区的卡塞雷斯省在西班牙东北部被使用。计算了一系列光谱指数，包括归一化植被指数（NDVI）和归一化燃烧率（NBR）。还使用了Sentinel-2A MSI传感器附带的三个新Red Edge波段的潜在附加值。坡度、坡向、植被覆盖度和地形粗糙度都被用来产生环境变量。使用光谱角映射器（SAM）分类器测试燃烧严重度。欧洲环境署的CORINE土地覆盖图也被用来制作被烧毁地区的土地覆盖类型。哥白尼应急管理服务使用0.5米分辨率的昂宿星图像制作了火灾分级图，用作参考。结果表明，燃烧的严重程度和测试的光谱指数之间的相关性程度可变。电磁波谱的可见光部分不太适合区分燃烧和未燃烧的土地覆盖。NBRb 12（短波红外2 -SWIR 2）在检测烧伤区域方面效果最佳。SAM在专题制图方面的总体准确率为73%。没有一个环境变量似乎对燃烧的严重程度有显着的影响。总之，我们的研究结果表明，Sentinel-2 MSI传感器可以用于识别烧伤面积和烧伤严重程度。然而，在推广目前研究的结果之前，应在不同地区使用相同的数据集类型和方法进行进一步的研究。
ABSTRACT Accurate, reliable, and timely burn severity maps are necessary for planning, managing and rehabilitation after wildfires. This study aimed at assessing the ability of the Sentinel-2A satellite to detect burnt areas and separate burning severity levels. It also attempted to measure the spectral separability of the different bands and derived indices commonly used to detect burnt areas. A short investigation into the associated environmental variables present in the burnt landscape was also performed to explore the presence of any correlation. As a case study, a wildfire occurred in the Sierra de Gata region of the province of Caceres in North-Eastern Spain was used. A range of spectral indices was computed, including the Normalized Difference Vegetation Index (NDVI) and the Normalized Burn Ratio (NBR). The potential added value of the three new Red Edge bands that come with the Sentinel-2A MSI sensor was also used. The slope, aspect, fractional vegetation cover and terrain roughness were all derived to produce environmental variables. The burning severity was tested using the Spectral Angle Mapper (SAM) classifier. European Environment Agency’s CORINE land cover map was also used to produce the land cover types found in the burned area. The Copernicus Emergency Management Service have produced a grading map for the fire using 0.5 m resolution Pleiades imagery, that was used as reference. Results showed a variable degree of correlation between the burning severity and the tested herein spectral indices. The visible part of the electromagnetic spectrum was not well suited to discern burned from unburned land cover. The NBRb12 (short-wave infrared 2 – SWIR2) produced the best results for detecting burnt areas. SAM resulted in a 73% overall accuracy in thematic mapping. None of the environmental variables appeared to have a significant impact on the burning severity. All in all, our study result showed that Sentinel-2 MSI sensor can be used to discern burnt areas and burning severity. However, further studies in different regions using the same dataset types and methods should be implemented before generalizing the results of the current study.