Mapping Wildfires in Southern Belize

Mapping Wildfires in Southern Belize – the case for analysis ready data (ARD)

Savanna based ecosystems across the world are suffering increasingly from wildfires due to climate change and illegal human activities. In many regions, this puts the fragile ecosystems under threat, so mapping the extent of wildfires becomes important to enable the organisation of mitigation measures.

Recently, Environment Systems supported a Masters student from the University of Edinburgh School of GeoSciences, Chris Halliday, in a project that sought to investigate a new approach to mapping savannas. An 1,800 km2 area of Southern Belize was chosen as the area of study. The area was chosen because it suffers from extensive wildfires, which destroy saplings, the habitats of nesting birds, and cause a general decline in biodiversity. Three ‘Protected Areas’ within this area are designated to protect key savanna species.

Currently, the burnt areas of savanna are mapped annually at the end of the dry season in May by visual interpretation of Sentinel-2 optical imagery. This method requires cloud-free imagery, which is not always available. In addition, the timing is not optimal due to rapid savanna regrowth. Radar data, which can penetrate cloud, is not generally used to map burnt areas of savanna as few land managers have the required expertise to handle this data source.

Burn areas in Southern Belize
Burnt areas mapped from Sentinel-2 (left) compared to burn areas mapped using a time-series of Sentinel-1 indices
Step-in Sentinel-1 analysis ready data (ARD) from Environment Systems Data Services. The project investigated pairs of radar images before and after a fire. The physical basis for detecting burnt areas using radar relies on being able to observe changes in backscatter over time. With imagery captured from January to December 2019, object-based image analysis was used to compare radar- based methods with the visual analysis of Sentinel-2 imagery obtained for the nearest dates. The radar-based method detected 87.6 % of the burnt areas compared to the visual analysis, but was also able to reveal more about fire evolution over the season due to the increased frequency of the data capture, and its ability to see through the cloud.