ABSTRACT In Fennoscandia, wildfires are typically low-intensity surface fires that burn vegetation on the forest floor. Multispectral remote sensing, particularly the normalized burn ratio (NBR) index, is widely used to assess burn severity, but its effectiveness in identifying surface fires under dense canopies is less studied. This study examined the characteristics of surface fires across seven one-hectare test sites in Scots pine-dominated boreal forests in Finland. Fire-induced spectral changes, measured as difference normalized burn ratio (dNBR) values from Sentinel-2 data, were compared with structural changes on the forest floor measured using terrestrial laser scanning (TLS). Breakpoint analysis of NBR time series identified most surface fires, with a noticeable decline in NBR values. Fires not identified with the breakpoint analysis had less burned ground vegetation and denser canopy cover. A moderate correlation (r = –0.5) was observed between spectral and volumetric changes, with higher dNBR values corresponding to greater vegetation reduction. However, dense canopies lowered burn severity despite similar ground vegetation volume changes. NBR changes were explained by vegetation volume, canopy cover, and site conditions (R² = 84%). These findings improve understanding of the applicability of multispectral remote sensing in detecting fires, assessing burn area, and estimating burn severity in boreal forests.
Read full abstract