Comparing methods of reconstructing fire history using fire scars in a southwestern United States ponderosa pine forest

Fire scars have been used to understand the historical role of fire in ponderosa pine (Pinus ponderosa Dougl. ex P. & C. Laws.) ecosystems, but sampling methods and interpretation of results have been criticized for being statistically invalid and biased and for leading to exaggerated estimates of fire frequency. We compared "targeted" sampling, random sampling, and grid-based sampling to a census of all 1479 fire-scarred trees in a 1 km ² study site in northern Arizona. Of these trees, 1246 were sufficiently intact to collect cross-sections; of these, 648 had fire scars that could be cross-dated to the year of occurrence in the 200-year analysis period. Given a sufficient sample size (approximately n ≥ 50), we concluded that all tested sampling methods resulted in accurate estimates of the census fire frequency, with mean fire intervals within 1 year of the census mean. We also assessed three analytical techniques: (1) fire intervals from individual trees, (2) the interval between the tree origin and the first scar, and (3) proportional filtering. "Bracketing" fire regime statistics to account for purported uncertainty associated with targeted sampling was not useful. Quantifying differences in sampling approaches cannot resolve all the limitations of fire-scar methods, but does strengthen interpretation of these data.