Fuel treatment effectiveness in the context of landform, vegetation, and large, wind-driven wildfires.

Autor: Prichard SJ; School of Environmental and Forest Sciences, University of Washington, Box 352100, Seattle, Washington, 98195-2100, USA., Povak NA; USDA Forest Service, Pacific Northwest Research Station, Wenatchee Forestry Sciences Lab, Wenatchee, Washington, 98801, USA.; Oak Ridge Institute for Science and Education (ORISE), Oak Ridge, Tennessee, 37830, USA., Kennedy MC; Sciences and Mathematics, Division of the School of Interdisciplinary Arts and Sciences, University of Washington - Tacoma, Tacoma, Washington, 98801, USA., Peterson DW; USDA Forest Service, Pacific Northwest Research Station, Wenatchee Forestry Sciences Lab, Wenatchee, Washington, 98801, USA.
Jazyk: angličtina
Zdroj: Ecological applications : a publication of the Ecological Society of America [Ecol Appl] 2020 Jul; Vol. 30 (5), pp. e02104. Date of Electronic Publication: 2020 Apr 02.
DOI: 10.1002/eap.2104
Abstrakt: Large wildfires (>50,000 ha) are becoming increasingly common in semiarid landscapes of the western United States. Although fuel reduction treatments are used to mitigate potential wildfire effects, they can be overwhelmed in wind-driven wildfire events with extreme fire behavior. We evaluated drivers of fire severity and fuel treatment effectiveness in the 2014 Carlton Complex, a record-setting complex of wildfires in north-central Washington State. Across varied topography, vegetation, and distinct fire progressions, we used a combination of simultaneous autoregression (SAR) and random forest (RF) approaches to model drivers of fire severity and evaluated how fuel treatments mitigated fire severity. Predictor variables included fuel treatment type, time since treatment, topographic indices, vegetation and fuels, and weather summarized by progression interval. We found that the two spatial regression methods are generally complementary and are instructive as a combined approach for landscape analyses of fire severity. Simultaneous autoregression improves upon traditional linear models by incorporating information about neighboring pixel burn severity, which avoids type I errors in coefficient estimates and incorrect inferences. Random forest modeling provides a flexible modeling environment capable of capturing complex interactions and nonlinearities while still accounting for spatial autocorrelation through the use of spatially explicit predictor variables. All treatment areas burned with higher proportions of moderate and high-severity fire during early fire progressions, but thin and underburn, underburn only, and past wildfires were more effective than thin-only and thin and pile burn treatments. Treatment units had much greater percentages of unburned and low severity area in later progressions that burned under milder fire weather conditions, and differences between treatments were less pronounced. Our results provide evidence that strategic placement of fuels reduction treatments can effectively reduce localized fire spread and severity even under severe fire weather. During wind-driven fire spread progressions, fuel treatments that were located on leeward slopes tended to have lower fire severity than treatments located on windward slopes. As fire and fuels managers evaluate options for increasing landscape resilience to future climate change and wildfires, strategic placement of fuel treatments may be guided by retrospective studies of past large wildfire events.
(Published 2020. This article is a U.S. Government work and is in the public domain in the USA.)
Databáze: MEDLINE
Externí odkaz: