This article has been peer reviewed and accepted for publication. It is in production and has not been edited, so may differ from the final published form.
Higher sensitivity and lower specificity in post-fire mortality model validation of eighteen western U.S. tree species
Managers require accurate models to predict post-fire tree mortality to plan prescribed fire treatments and examine their effectiveness. Here we assess the performance of a common post-fire tree mortality model with an independent dataset of 11 tree species from 13 National Park Service units in the western U.S. Overall model discrimination was generally strong, but varied among all species. Generally, most models had higher sensitivity (proportion of correctly classified dead trees) and lower specificity (proportion of correctly classified live trees). Variation in model accuracy (% of live and dead trees correctly classified) among species was not related to sample size or percent observed mortality. However, specificity was positively related to bark thickness, where mortality was commonly overestimated in thin barked species. Accuracy was also quite low for many thinner bark classes (< 1 cm) for many species, generally leading to poorer performing models. Our results indicate that the commonly used post-fire mortality model generally performs well; however, some thin-barked species and size classes would benefit from further model development to improve model specificity.
WF16081 Accepted 28 February 2017
© CSIRO 2017