An improved spatio-temporal clustering method for extracting fire footprints based on MCD64A1 in the Daxing’anling Area of north-eastern China

Abstract

Background Understanding the spatio-temporal dynamics associated with a wildfire event is essential for projecting a clear profile of its potential ecological influences. Aims To develop a reliable framework to extract fire footprints from MODIS-based burn products to facilitate the understanding of fire event evolution. Methods This study integrated the Jenks natural breaks classification method and the density-based spatial clustering of applications with noise (DBSCAN) algorithm to extract the fire footprints in Daxing’anling region of China between 2001 and 2006 from MCD64A1 burned area data. Key results The results showed that the fire footprints extracted by the model gained an overall accuracy of 80% in spatial and temporal domains after an intensive validation by using the historical fire records provided by the local agency. The agreement of burned area between the extracted fire patches and the historical fire records for those matched fire points was characterised by an overall determination coefficient R2 at 0.91. Conclusions The proposed framework serves as an efficient and convenient wildfire management tool for areas requiring large-scale and long-term wildfire monitoring. Implications The current framework can be used to create a reliable large-scale fire event database by providing an important alternative for the improvement of field investigation.