Estimation and evaluation of multi-decadal fire severity patterns using Landsat sensors

• Published in: Remote sensing of environment Vol. 170; pp. 340 - 349
• Main Authors: Parker, Brett M., Lewis, Tom, Srivastava, Sanjeev K.
• Format: Journal Article
• Language: English
• Published: Elsevier Inc 01-12-2015
• Subjects: Classification; dNBR; Ecosystems; Eucalyptus; Fire history; Fires; GeoCBI; Heath; Ignition; Melaleuca quinquenervia; Remote sensing; Satellite imagery; Sclerophyll vegetation; Temporal change; Vegetation; Fire history; GeoCBI; dNBR; Temporal change; Sclerophyll vegetation; Heath
• ISSN: 0034-4257; 1879-0704
• DOI: 10.1016/j.rse.2015.09.014

Retrospective identification of fire severity can improve our understanding of fire behaviour and ecological responses. However, burnt area records for many ecosystems are non-existent or incomplete, and those that are documented rarely include fire severity data. Retrospective analysis using satellite remote sensing data captured over extended periods can provide better estimates of fire history. This study aimed to assess the relationship between the Landsat differenced normalised burn ratio (dNBR) and field measured geometrically structured composite burn index (GeoCBI) for retrospective analysis of fire severity over a 23year period in sclerophyll woodland and heath ecosystems. Further, we assessed for reduced dNBR fire severity classification accuracies associated with vegetation regrowth at increasing time between ignition and image capture. This was achieved by assessing four Landsat images captured at increasing time since ignition of the most recent burnt area. We found significant linear GeoCBI–dNBR relationships (R2=0.81 and 0.71) for data collected across ecosystems and for Eucalyptus racemosa ecosystems, respectively. Non-significant and weak linear relationships were observed for heath and Melaleuca quinquenervia ecosystems, suggesting that GeoCBI–dNBR was not appropriate for fire severity classification in specific ecosystems. Therefore, retrospective fire severity was classified across ecosystems. Landsat images captured within ~30days after fire events were minimally affected by post burn vegetation regrowth.
•We apply methods to estimate fire severity patterns over multi-decadal periods.•Model effects of image capture with increasing time following fire event•Assess accuracy of Landsat dNBR fire severity in differing vegetation types•Identified utility of Landsat fire severity for multi-decadal studies