Giglio Louis, Schroeder Wilfrid, Justice Christopher O
Department of Geographical Sciences, University of Maryland, College Park, MD, USA.
Remote Sens Environ. 2016 Jun 1;178:31-41. doi: 10.1016/j.rse.2016.02.054. Epub 2016 Mar 11.
The two Moderate Resolution Imaging Spectroradiometer (MODIS) instruments, on-board NASA's Terra and Aqua satellites, have provided more than a decade of global fire data. Here we describe improvements made to the fire detection algorithm and swath-level product that were implemented as part of the Collection 6 land-product reprocessing, which commenced in May 2015. The updated algorithm is intended to address limitations observed with the previous Collection 5 fire product, notably the occurrence of false alarms caused by small forest clearings, and the omission of large fires obscured by thick smoke. Processing was also expanded to oceans and other large water bodies to facilitate monitoring of offshore gas flaring. Additionally, fire radiative power (FRP) is now retrieved using a radiance-based approach, generally decreasing FRP for all but the comparatively small fraction of high intensity fire pixels. We performed a Stage-3 validation of the Collection 5 and Collection 6 Terra MODIS fire products using reference fire maps derived from more than 2500 high-resolution Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) images. Our results indicated targeted improvements in the performance of the Collection 6 active fire detection algorithm compared to Collection 5, with reduced omission errors over large fires, and reduced false alarm rates in tropical ecosystems. Overall, the MOD14 Collection 6 daytime global commission error was 1.2%, compared to 2.4% in Collection 5. Regionally, the probability of detection for Collection 6 exhibited a ~3% absolute increase in Boreal North America and Boreal Asia compared to Collection 5, a ~1% absolute increase in Equatorial Asia and Central Asia, a ~1% absolute decrease in South America above the Equator, and little or no change in the remaining regions considered. Not unexpectedly, the observed variability in the probability of detection was strongly driven by regional differences in fire size. Overall, there was a net improvement in Collection 6 algorithm performance globally.
搭载在美国国家航空航天局（NASA）的Terra和Aqua卫星上的两台中等分辨率成像光谱仪（MODIS），已经提供了超过十年的全球火灾数据。在此，我们描述了作为2015年5月开始的第6版陆地产品再处理工作的一部分，对火灾探测算法和扫描带级产品所做的改进。更新后的算法旨在解决在之前第5版火灾产品中观察到的局限性，特别是由小片森林砍伐导致的误报，以及被浓烟遮蔽的大火被遗漏的问题。处理范围还扩展到了海洋和其他大型水体，以方便对近海天然气燃烧进行监测。此外，现在使用基于辐射率的方法来反演火灾辐射功率（FRP），除了相对较小比例的高强度火灾像素外，这通常会降低所有火灾像素的FRP。我们使用从2500多张高分辨率先进星载热发射和反射辐射仪（ASTER）图像中得出的参考火灾地图，对第5版和第6版Terra MODIS火灾产品进行了第3阶段验证。我们的结果表明，与第5版相比，第6版有源火灾探测算法的性能有针对性地得到了改进，在大型火灾上的遗漏误差减少，热带生态系统中的误报率降低。总体而言，第6版MOD14白天全球委托误差为1.2%，而第5版为2.4%。在区域上，与第5版相比，第6版在北美北部寒带和亚洲北部寒带的探测概率绝对增加了约3%，在赤道亚洲和中亚绝对增加了约1%，在赤道以南的南美洲绝对减少了约1%，在其他考虑的区域几乎没有变化或没有变化。不出所料，探测概率的观测变化主要由火灾规模的区域差异驱动。总体而言，全球第6版算法性能有了净提升。
