Remote Sensing of Burn Severity Using Coupled Radiative Transfer Model: A Case Study on Chinese Qinyuan Pine Fires

• Main Authors: Changming Yin, Binbin He, Xingwen Quan, Marta Yebra, Gengke Lai
• Format: Article
• Language: English
• Published: MDPI AG 2020-11-01
• Series: Remote Sensing
• Subjects: burn severity; composite burn index; Tree canopy cover; RTM; Sentinel-2A
• Online Access: https://www.mdpi.com/2072-4292/12/21/3590

Burn severity mapping is critical to quantifying fire impact on key ecological processes and post-fire forest management. Satellite remote sensing has the advantages of high spatial-temporal resolution and large-scale monitoring and provides a more efficient way to evaluate forest fire burn severity than traditional field or aerial surveys. However, the proportion of tree canopy cover (TCC) affects the spectral signal received by remote sensing sensors from the background charcoal and ash. Consequently, not considering this factor normally leads a spectral confusion in burn severity retrieval. In this study, the burn severity of two Qinyuan forest fires was estimated using a coupled Radiative Transfer Model (RTM) and Sentinel-2A Multi-Spectral Instrument (MSI) reflectance data. A two-layer Canopy Reflectance Model (ACRM) RTM was coupled with the GeoSail RTM by replacing the spectra of the background input of GeoSail RTM to simulate the spectra of the three-layered forests for burn severity retrieval measured as the Composite Burn Index (CBI). The TCC data was then served to RTM parameterization and constrain the backward inversion procedure of the coupled RTM to alleviate spectral confusion. Finally, the inversion retrievals were evaluated using 163 field measured CBI. The coupled RTM can simulate the radiative transfer characteristics of three-layer vegetation and has greater potential to accurately estimate burn severity worldwide. To evaluate the merit of our proposed method, the CBI was estimated through coupled RTM inversion with TCC constraint (CP_RTM+TCC), coupled RTM inversion with global optimal search (CP-RTM+GOS), Forest Reflectance and Transmittance (FRT) RTM inversion with TCC constraint (FRT+TCC), and random forest (RF) algorithm. The results showed that the method proposed in this study (CP_RTM+TCC) yielded the highest estimation accuracy (R² = 0.92, RMSE = 0.2) among the four methods used as benchmark, indicating its reasonable ability to assist forest managers to better understand post-fire vegetation regeneration and forest management.