Burned area mapping time series in Canada (1984–1999) from NOAA-AVHRR LTDR: A comparison with other remote sensing products and fire perimeters

A new algorithm for mapping burned areas in boreal forest using AVHRR archival data Long Term Data Record (LTDR) (0.05°, ca. 5 km, version 3) was developed in Canada using burn records for the period between 1984 and 1999 and evaluated against AVHRR 1 km and AVHRR-PAL 8 km burned area map products. The algorithm combined 1) absolute and relative radiometric thresholds, 2) a Bayesian network classifier, and 3) neighborhood analysis for spatial fire coherence. Fire event records from Canadian Forest Service National Fire Database (CFSNFD) for western Canada were used to train the algorithm. LTDR and AVHRR 1 km burned area mapping were similar for the same area, and correlated well to CFSNFD annual fire event records for western Canada, r2 = 0.72 and 0.77, respectively. In addition, the LTDR mapping correlated well with fires for all of Canada in the CFSNFD database (r2 = 0.65). This mapping product was a significant improvement over an 8 km AVHRR-PAL burned area map product. For mapping boreal forests burned areas globally, this study demonstrates the potential accuracy for where LTDR represents the highest spatial and temporal resolution of daily images available since the 1980s.

► We generated annual burned area maps from LTDR for Canada from 1984 to 1999.
► Our algorithm used band thresholds, a Bayesian network and neighborhood analysis.
► LTDR correlated well with wildland fires for all of Canada (r2 = 0.65).
► Significant improvement over an 8 km AVHRR-PAL burned area map product.