ABSTRACT
Given the plethora of free satellite images, either those coming from medium to high spatial resolution sensors, e.g. Landsat and Sentinel-2 (that provide satellite images continuously every 16 [8 if the two Landsat are concerned] or 5 days [for both Sentinel-2 A and B]) or from moderate resolution sensors, e.g. MODIS (that provide satellite images continuously every day), time series methods are becoming very popular approaches for their processing. The time series approach is useful to create temporal spectral profiles for characterizing vegetation phenology—an important element of vegetation characteristics that is important in vegetation monitoring, especially when dense satellite remote sensing observations are available. Time series methods can be used not only to study post-fire recovery of burned areas but also to map the actual burned areas by considering abruptions of phenological cycles of the vegetation caused by fire. Techniques to analyze the time series data are very popular and come either from time series statistics, like those that characterize trends and assess break points, or from other fields like the dynamic time warping approach originally developed for speech recognition and then integrated into time series statistics.
Apart of the time series approach, the existence of historical archives of satellite images, e.g. Landsat or nowadays Sentinel-2, that are freely available to the public and cover large spatial and temporal extents at continental scale provide a unique opportunity to overcome cost constraints when reconstructing recent fire history, e.g. 1984 (systematically after the launch of Landsat TM), globally at low to high spatial resolution.
In this chapter, details about two hot topics in satellite remote sensing technology of forest fires are presented, the first concerns the reconstruction of fire history by using mainly Landsat images and the second the monitoring of fire-affected areas by using mainly Sentinel-2 satellite data. Some of the analysis have been implemented in R, and parts of the scripts are also provided.
