Mapping spatial patterns of burned areas using Landsat images across Peruvian biomes

Abstract:

Fire events caused by agricultural and land-use practices pose severe issues to Peruvian biomes; therefore, quantifying spatial patterns of burned areas is fundamental for fire prevention and monitoring under climate change scenarios and REDD+ initiatives. Previous studies have recorded fire occurrences based on heat points with 375-500 m spatial resolution using the VIIRS-MODIS sensors, but these studies have failed to accurately identify small-scale burned areas. This study focused on quantifying burned areas during 2016 using Landsat images (ETM+ and OLI) with 30 m resolution from the three most important Peruvian biomes: the Costa, Sierra and Amazon, and then conducting an accuracy assessment using field data to estimate an adjusted area to eliminate bias accompanied by confidence intervals. We used annual (pre-fire) and semestral composite images (post-fire) were computed with Earth Engine (GEE) for 2015 and 2016, respectively. Multi-temporal changes were measured by spectral indexes (dNBR and RdNBR) and then processed using NDVI, NDWI, AFRI, DEM variables. These results show that burned areas reached ca. 233,087.85 ha ± 46,617.4 ha (95% confidence interval) in the Costa-Sierra biome and 133,553.07 ha ± 26,690 ha (95% confidence interval) in the Amazon biome, while MODIS estimated 103,664.70 ha and 53,300.29 ha, respectively. Our results suggest that Landsat images can be used to detect fires on a 0,5-ha scale under the forest definition of Peru’s National Forest Inventory, and can serve as a baseline to aid the country's forest monitoring system while being incorporated into a free access platform in GEE. Landsat images yield a great potential for accurately quantifying burned areas, particularly in countries where Forest Reference Emissions Levels are fundamental to implement REDD+ and report emissions to the UNFCCC. Further studies are needed to overcome difficulties related to cloud cover by using Radar over extensive areas as the Amazon biome.