Fig. 4

A Results of four post-fire stem density generalized linear mixed models (i.e., coefficients) and modeled stem density (line) with varying B distance to seed source and C percent grass cover, assuming median values for other model predictors in 2011. Separate models were created for (1) all species, (2) grand fir, (3) Douglas fir, and (4) ponderosa pine. Final models of post-fire stem density included the following predictor variables: distance to the seed source, percent grass cover, transformed topographic aspect, salvage logging (presence/absence), and year of survey (see Table 1 for all variables considered); symbol shape in A indicates whether each predictor variable improved model fit (i.e., change in AIC > 2pts). A Predictor variables were standardized to allow comparison and error bars are 95% confidence intervals on coefficient estimates. B, C Note the y-axis is on a log scale (i.e., models appear linear), and points are the observed values in 2011. Data were collected in 2011 and 2021 in the School Fire, Blue Mountains, WA, USA