Effective fire suppression in boreal forests

Fire suppression is (functionally) effective insofar as it reduces area burned. In North American boreal forests, fire regimes and historical records are such that this effect cannot be detected or estimated directly. I present an indirect approach, proceeding from the practice of initial attack (IA), which is intended to limit the proportion of "large" fires. I analysed IAʹs (operational) effectiveness by a controlled retrospective study of fire-history data for an approximately 86 000 km2 region of boreal forest in northeastern Alberta, Canada, from 1968 to 1998 (31 years). Over this interval, various improvements to IA practice, including a 1983 change in management strategy, created a natural experiment. I tested the results with multiple logistic regression models of the annual probabilities of a fire becoming larger than 3 and 200 ha. Annual fire counts (Nt) were a surrogate for fire weather and peak daily counts within years (arrival load). Measured by odds ratios, mean IA effectiveness against 3- and 200-ha fires increased in 1983 by factors of 2.02 (95% CI = 1.70-2.40) and 2.41 (95% CI = 1.69-3.45), respectively. Prior to 1983, the functional response to Nt was consistent with saturation of IA capacity at high arrival loads. From 1983-1998, effectiveness was independent of Nt. I introduce the proportional reduction in area burned (impact) as a measure of functional effectiveness and state conditions under which it can be estimated from the regression models. Over 1983-1998, if suppressed and actual fires were comparable, relative IA impact (r) was 0.58 (95% CI = 0.34-0.74) and area burned was reduced by 457 500 ha. If fires larger than 1 × 10^5, 1 × 10^4, or 1 × 10^3 ha are assumed to be unpreventable, r declines to 0.46, 025, or 0.08, respectively, but there is no evidence this is the case.