Disentangling Post-Fire Logging and High-Severity Fire Effects for Spotted Owls

Simple Summary: High-severity ﬁre transforms Western U.S. conifer forests into a unique forest type known as “snag forest habitat”, which spotted owls actively use to hunt for small mammal prey. This snag forest habitat is heavily targeted by post-ﬁre logging projects. Studies have shown that post-ﬁre logging signiﬁcantly reduces spotted owl occupancy, but efforts have generally not been made to separate the effects of such logging from the inﬂuence of high-severity ﬁre alone on spotted owls. We document that articles reporting negative effects of high-severity ﬁre on spotted owls were pervasively confounded by post-ﬁre logging. Our results indicate a need to approach analyses of high-severity ﬁre and spotted owls differently in future research. Abstract: The spotted owl is a rare and declining raptor inhabiting low/middle-elevation forests of the Paciﬁc Northwest, California, and the Southwest in the USA. It is well established that spotted owls select dense, mature, or old forests for nesting and roosting. High-severity ﬁre transforms such forests into a unique forest type known as “snag forest habitat”, which the owls select for foraging. This habitat is disproportionately targeted by post-ﬁre logging projects. Numerous recent articles have explored the inﬂuence of high-severity ﬁre and post-ﬁre logging on this species. Studies have shown that post-ﬁre logging signiﬁcantly reduces spotted owl occupancy, but efforts have generally not been made to disentangle the effects of such logging from the inﬂuence of high-severity ﬁre alone on spotted owls. We conducted an assessment of published, peer-reviewed articles reporting adverse impacts of high-severity ﬁre on spotted owls, exploring the extent to which there may have been confounding factors, such as post-ﬁre logging. We found that articles reporting adverse impacts of high-severity ﬁre on spotted owls were pervasively confounded by post-ﬁre logging, and in some cases by a methodological bias. Our results indicate a need to approach analyses of high-severity ﬁre and spotted owls differently in future research.


Introduction
Patches of high-severity fire, in which most or all trees are killed [1], are a natural component of mixed-conifer and mixed-evergreen forest types commonly inhabited by the spotted owl (Strix occidentalis) across most of its range, including the Eastern Cascades, Klamath-Siskiyou region of Northwestern California, the Sierra Nevada, and the Southwestern USA [2][3][4][5][6][7]. The spotted owl is one of the most-studied raptors in the world, with decades of intensive research [8]. Populations have been declining in managed forests that were largely unaffected by recent wildfires [9,10] while remaining stable in unmanaged forests that experienced large fires [11]. Despite this, it remains a commonly held belief that large fires are a primary threat [12].
It is well documented that spotted owls select stands of dense, mature, or old forest for nesting and roosting [13,14]. High-severity fire transforms such stands into "snag forest habitat", also known as "complex early seral forest" [15], immediately after fire. Complex early seral snag forests contain many standing fire-killed trees, coarse woody debris, and regrowing shrubs and seedlings, which provide habitat for many fire-adapted taxa [15]. Spotted owls have been found to actively use snag forest habitat areas for foraging from one to more than 12 years post-fire [16][17][18][19], likely due to an ample small mammal prey base [20], which can be 2-6 times greater in such post-fire habitat compared to unburned old forest [21]. Conversely, post-fire logging has been found to adversely affect spotted owl occupancy and reproduction, by removing and degrading the snag forest habitat used by the owl's prey species and by removing nearly all the snags that spotted owls depend upon as a "perch and pounce" predator [22,23]. Seemingly minor amounts of post-fire logging (as little as 5%) significantly reduce spotted owl occupancy [22]. Although the complex, but generally neutral or positive, effects of high-severity fire for spotted owls is established [23,24], debate remains about the extent to which factors such as post-fire logging may confound attempts to better understand effects of high-severity fire [22].
In November of 2019, the U.S. Fish and Wildlife Service denied a petition to list the California spotted owl subspecies (S. o. occidentalis) as threatened under the Endangered Species Act, asserting that high-severity fire is the primary threat to the owl [12]. The decision did not mention post-fire logging as a threat, and encouraged increased logging of all types ostensibly to curb high-severity fire. The decision also did not address research that found increased logging, including post-fire logging, is often associated with higher, not lower, fire severity in subsequent wildland fires [25][26][27][28].
Here, we investigate published articles that have reported adverse effects of highseverity fire on spotted owls, exploring the extent to which confounding factors such as post-fire logging may have affected results and conclusions. This information has important implications for forest management that impacts populations of this imperiled raptor.

Materials and Methods
For our evaluation of the extent to which confounding effects, such as post-fire logging, have occurred in existing studies on high-severity fire, we assessed all articles reporting adverse effects of forest fire on spotted owls. The search encompassed articles on all three subspecies, the northern spotted owl (S. o. caurina), California spotted owl, and Mexican spotted owl (S. o. lucida), inhabiting low/middle-elevation conifer forests in the Pacific Northwest, California, and the Southwestern USA.
Our approach was to conduct a literature search for all peer-reviewed, published articles that reported one or more adverse impacts from high-severity fire on spotted owl occupancy, survival, reproduction, recruitment, or foraging. We first queried the Google Scholar database using the keywords "high severity fire" and "spotted owl" to find any suitable articles. We then evaluated each article to determine whether a confounding influence, including post-fire logging, was present in the analysis of high-severity fire effects. We extracted and tabulated data from each paper regarding: subspecies, location, number of burned breeding sites, occurrence of post-fire logging in the study area (yes or no), whether logging and fire were conflated (yes or no), whether distance from breeding site center was included in foraging selection analysis (yes or no), the type of effect, and whether there was a statistically significant (α = 0.05) high-severity fire or post-fire logging effect (positive = +, no effect = 0, and negative = −). In our assessment, we did not include review papers without primary data, nor did we include papers with anecdotal references to wildland fire effects on spotted owls.

Results
We found 13 published articles reporting adverse high-severity fire effects on spotted owls (Table 1). Every study that showed a significant negative effect of "high-severity fire" was also confounded by post-fire logging, except for one study [29] that was not confounded by post-fire logging within the nest core but could have been influenced by post-fire logging outside the core [22]. Table 1. Summary results of 13 spotted owl studies reporting one or more negative effects of high-severity fire (this included studies reporting mixed negative/positive effects). Data include article reference; subspecies of spotted owls (MSO = Mexican spotted owl, NSO = northern spotted owl, CSO = California spotted owl); geographic location (AZ = Arizona, NM = New Mexico, So. OR = Southern Oregon, So. CA = Southern California, No. CA = Northern California); number of burned spotted owl breeding sites in the study; whether post-fire logging occurred; whether post-fire logging and high-severity fire were conflated in the study; whether distance from site center was analyzed in foraging selection studies; effects analyzed; statistically significant result as stated in the article (0 = no effect, + = positive effect, -= negative effect).  [30]-this study reported no significant effect of fire on occupancy and reproduction of Mexican spotted owls in fire areas relative to unburned forests in Southwestern USA. They described a statistically non-significant adverse impact (p = 0.11) and attributed this to high-severity fire. Bond (2016) [31] reported that post-fire logging was minor in most of the fire areas in this study. The amount of post-fire logging was not quantified.  [32], and post-fire occupancy of northern spotted owls in the Quartz, Timbered Rock, and Biscuit fires of Southwestern Oregon [33], these studies reported adverse impacts on spotted owls from the combined effects of high-severity fire and post-fire logging. In both studies, the area of high-severity fire and post-fire logging were included as a single variable and their effects were therefore conflated. Lee et al. (2013) [34]-for fire areas in the San Bernardino and San Jacinto Mountains of Southern California, this study reported lower California spotted owl site occupancy with higher levels of high-severity fire and post-fire logging. Both variables were included and analyzed independently. The 95% confidence intervals of beta coefficients for all fire and post-fire logging effects overlapped zero indicating the effects were not statistically significant. The analysis examined habitat variables only within the 203-ha nesting core area and did not include impacts from post-fire logging outside these innermost core areas.

Reference
Tempel et al. (2014) [35]-this study reported reduced California spotted owl site colonization due to high-severity fire, based on four owl sites in the 2001 Star fire in the central Sierra Nevada. However, nearly all of the high-severity fire areas in these four owl sites were post-fire logged shortly after the Star fire occurred [22].
Lee and Bond (2015a) [29]-this study found previous occupancy state (reproductive or not) was an important predictor of occupancy within a California spotted owl site's 203-ha site in the San Bernardino and San Jacinto Mountains of Southern California. Highseverity fire and post-fire logging covariates were both negatively correlated with the probability of site occupancy. The occupancy of sites that supported reproductive owls the previous year was reduced 0.02 by high-severity fire and 0.03 by post-fire logging. The occupancy of sites that supported non-reproductive owls the previous year was reduced by 0.19 by severe fire and 0.26 by post-fire logging. Neither high-severity fire nor post-fire logging significantly influenced reproduction. The analysis examined habitat variables only within the 203-ha nesting core area and did not include impacts from post-fire logging outside these innermost core areas. Comfort et al. (2016) [36]-in a radio-telemetry study of northern spotted owl foraging in the 2002 Timbered Rock fire in Southwestern Oregon, the owls generally avoided "hard edges" defined as abrupt changes in fire severity. The study noted that post-fire logging had been extensive and pervasive in the high-severity fire areas in the Timbered Rock fire and that hard edges tended to result from logging, and their analysis did not distinguish between unlogged high-severity fire areas and post-fire logged sites. Therefore we considered this study as showing adverse effects of post-fire logging. Jones et al. (2016) [37]-this study investigated the effect of the King fire of 2014 in the central Sierra Nevada on California spotted owl occupancy and foraging. The study reported reduced occupancy in sites with >50% high-severity fire, and a pattern of avoidance of high-severity fire areas for foraging. The authors stated that some post-fire logging had occurred but described it as inconsequential (<5%). A subsequent quantitative assessment showed that a mean of 12% of the area within spotted owl sites in the King fire had been post-fire logged-higher than the level (5%) above which significant reductions in occupancy occur [22].
The foraging analysis in Jones et al. (2016) did not distinguish between snag forest habitat (high-severity fire occurring in mature forest, with no post-fire logging) versus high-severity fire areas in pre-fire clearcuts (young plantations), or high-severity fire areas that had been post-fire logged. Therefore the foraging analysis was confounded by both pre-fire clearcutting and post-fire logging [22].
A further problem is that Jones et al. (2016) did not analyze use of high-severity fire areas as a function of distance from spotted owl site centers. Spotted owls have a higher probability of foraging in stands nearest to site centers [16,17,[38][39][40][41][42][43][44]. Differential availability of high-severity fire areas near site centers created a bias in foraging habitat selection results because the effect of distance was not explicitly analyzed. Eyes et al. (2017) [18]-this study reported a mix of non-significant positive and negative effects on California spotted owl foraging in high-severity fire areas in the absence of post-fire logging. Eyes et al. (2017) found that the odds of spotted owls using "high contrast edges" created by high-severity fire for foraging "were 2.8 times greater than odds for the low contrast edge type and 3.5 times greater than for the no edge type". The authors reported a reduction (4%) in the probability of use from the lowest to the highest fire severity index, but there was nearly equivalent evidence (weight) for models that had no fire effect relative to models that included a fire effect. However, Eyes et al. did not analyze the interaction of distance with fire severity. Figure 1 of Eyes et al. shows that, in most of the owl sites analyzed by the authors, high-severity fire patches were several hundred meters to >1 km from site centers. In the three spotted owl sites in Eyes et al. for which the owls had access to high-severity fire areas in reasonable proximity to site centers, the owls actively foraged in the high-severity fire patches (Eyes et al. 2017).
Moreover, Eyes et al. under-reported foraging in snag forest habitat created by highseverity fire in two of these three territories. Numerous foraging locations were in highseverity fire patches created by an earlier fire (the 1990 A-Rock fire) that subsequently reburned at low/moderate-severity (compare Figure 1  If territory fitness (apparent survival + recruitment) reflects habitat quality for northern spotted owl territories, Rockweit et al. (2017) indicates that 36% of territories analyzed showed no fitness consequences, 14% showed decreased fitness, and 50% showed increased fitness after being burned. Thus, high-severity fire resulted in more benefits than costs to northern spotted owls, which might have been even more evident if Rockweit et al. had accounted for post-fire logging in the Sims fire.
Jones et al. (2020a) [19]-this study reported that, in the King fire, California spotted owls avoided post-fire logged areas, and selected high-severity fire patches where the average patch size was up to 115 ha, but avoided making "deep forays" into larger highseverity fire patches. One important limitation of this study is the fact that most of the post-fire salvage logging that had occurred by the time they conducted their research (2015-2017) was not included in their analysis. This is illustrated by comparing  [48]. The omitted areas of postfire logging were almost entirely within the largest high-severity fire patches. Therefore, the conclusions in Jones et al. (2020a) regarding large high-severity fire patches were-similar to their earlier study in the same fire-confounded by post-fire logging that was not fully accounted for in their analysis. Additionally, in their analysis of spotted owl selection/avoidance of high-severity fire patches, Jones et al. (2020a) did not exclude recent numerous pre-fire clearcuts that burned at high severity in the King fire [22]. As a result, their analysis of selection/avoidance of high-severity fire areas was not specific to snag forest habitat but, rather, was confounded by extensive pre-fire and post-fire clearcutting.
Schofield et al. (2020) [11]-this study investigated California spotted owl occupancy before and after the Rim fire of 2013 in Yosemite National Park within areas protected from logging (no pre-fire or post-fire logging). They found 11 occupied spotted owl territories during a ten-year survey period before the Rim fire (2004-2013) and 12 occupied territories during a three-year survey period in the same area after the Rim fire (Schofield et al. 2020, Figure 1). Based on this, the authors concluded that, in these areas that were not subjected to post-fire logging, this large mixed-intensity fire did not adversely affect spotted owl occupancy.
Schofield et al. also assessed the percent high-severity fire within the 153-ha nest core, and found that all of the sites occupied after the Rim fire had <30% high-severity fire within nest cores, which might be perceived as a negative effect of large areas of severe fire.
Kramer et al. (2021) [49]-the authors of this study investigated foraging habitat selection of California spotted owls in mixed-severity fire areas in three national parks (no pre-or post-fire logging). Kramer et al. (2021) found that spotted owls neither selected nor avoided high-severity fire patches overall, and reported reduced use of the largest high-severity fire patches. The findings of Kramer et al. (2021) were confounded by the fact that they did not analyze use of resources as a function of distance from spotted owl site centers (no analysis of distance), which creates a bias since spotted owls are central-place foragers and select lower-severity areas for their nest cores.

General Observations
We found that articles reporting adverse impacts of high-severity fire on spotted owls are widely confounded by reported and unreported post-fire logging. Two articles [18,49] did not have post-fire logging because they occurred in national parks, but were problematic for a different reason-a failure to analyze distance from site center in foraging habitat selection. One study that did explicitly disentangle fire and post-fire logging effects found adverse effects of both types of disturbances on California spotted owls [29], but was conducted in one region, Southern California, where owl sites had substantially lower amounts of suitable old-forest habitat before fire compared with owl sites in other regions, and post-fire logging had occurred outside of the nest core areas analyzed [22]. This suggests that studies should analyze effects of post-fire logging at a larger scale around owl core areas.
We did not include in our assessment anecdotal published reports, such as the brief mention in Gutiérrez and Pritchard (1990) [50] that they did not find spotted owls in a 1987 fire area west of Palomar Mountain in Southern California. Notably, however, the first available post-fire satellite imagery following this 1987 fire, taken in 1994 and 1995, reveals logging roads, landings, skid trails, and cleared areas consistent with post-fire logging in numerous areas northwest, west, and southwest of Palomar Mountain, but no data were provided on pre-fire owl locations in the fire area [50]. Additionally, we did not include published reports or presentations that were not subjected to the peer-review process, such as Gaines et al. (1997) [51], King et al. (1998) [52], and Keane et al. (2010) [53]. However, King et al. (1998) discussed extensive post-fire logging in their study area, and though Gaines et al. (1998) and Keane et al. (2010) did not specifically mention post-fire logging, subsequent research documented that widespread post-fire logging had indeed occurred in both study areas by the time of surveys [22,54].
In the absence of post-fire logging-or prior to post-fire logging-in the Sierra Nevada, spotted owl occupancy has been found to be high, and often increasing, in fire areas with substantial high-severity fire effects [22,55], including in very large forest fires [11,56]. However, high-severity fire and post-fire logging are pervasively entangled in existing scientific literature regarding spotted owls. In this context, federal wildlife and land managers are using existing confounded studies to promote the view that only highseverity fire, not post-fire logging, is a threat to spotted owls and, ironically, are promoting increased post-fire logging and tree plantation establishment in spotted owl habitat on public lands ostensibly to curb future fires [12,57]. This is in stark contrast to research finding that such management not only harms spotted owls [22,23] but also tends to increase future fire severity [25,27]. This suggests a need for more rigorous efforts to be made in future research and forest management decisions to disentangle the impacts of post-fire logging from the effects of high-severity fire.
Several rebuttals have been published attempting to counter findings that post-fire logging was harmful to spotted owls. However, these rebuttals also failed to properly disentangle post-fire logging from severe fire effects. For example, Jones et al. (2019) [58] presented a critique of Hanson et al. (2018) [22], which found a large adverse impact of post-fire logging on California spotted owl occupancy, but no effect of high-severity fire when sites without the confounding effect of post-fire logging were analyzed. Hanson et al.  [37] did not analyze distance from the center of spotted owl sites in their foraging habitat selection analysis, and that this created a methodological bias against high-severity fire locations, as discussed above regarding Eyes et al. (2017) [18] and Kramer et al. (2021) [49]. In response, Jones et al. (2019) [58] note that they excluded the farthest 5% of foraging locations in Jones et al. (2016) [37]. However, this does not remedy the methodological bias against high-severity fire locations. Analyzing foraging habitat selection as a function of distance means that distance from the center is a covariate of every available and used habitat element [16]. The farthest locations are an important part of the dataset when distance from the site center is specifically incorporated into resource selection analyses, and should not have been excluded.

Observations on Opposing Studies
Peery et al. (2019) [59] also mentioned that Hanson et al. (2018) [22] excluded four owl sites in the King fire with >80% high-severity fire from the analysis of high-severity fire effects, and suggested that these sites indicated negative impacts of high-severity fire on spotted owl occupancy. As discussed above regarding Jones et al. (2019) [58], the fundamental purpose of the high-severity fire analysis in Hanson et al. (2018) was to avoid the confounding effect of post-fire logging, which is why this analysis was restricted to sites with <5% post-fire logging. None of the four King fire sites noted by Peery et al. (2019) met that criterion; all had substantial post-fire logging, as discussed above. Therefore, the exclusion of these sites from the high-severity fire analysis in Hanson [59], and arguing that negative effects of high-severity fire might possibly emerge if the data were analyzed differently. In response, Lee (2020) [24] reanalyzed the data according to their suggestions and found that the original results and conclusions in Lee (2018) [23] were robust to Peery et al.'s technical criticisms. Notably, removing studies confounded by post-fire logging resulted in an overall positive effect of mixed-severity fires on spotted owl occupancy, and increasing percentage of high-severity fire was associated with increased spotted owl reproduction.

Management Implications
Based on the current state of knowledge indicating significant adverse effects of post-fire logging on this imperiled forest raptor, we recommend that post-fire logging be avoided within at least a 2000-m radius around spotted owl site centers. If we are to better understand the response of spotted owls to high levels of high-severity fire, and make forest management and wildlife protection decisions based on evidence rather than assumptions, we must investigate this issue without the confounding effect of post-fire logging. Further, we recommend that future research funded by land management agencies regarding high-severity fire effects on spotted owls exclude sites with >5% post-fire logging within a 2000-m radius of the site center (see Hanson et al. 2018). Third, we recommend that such future research always analyze distance from site center as a covariate when spotted owl habitat selection is being analyzed with regard to fire severity, and that foraging habitat analyses separate pre-fire and post-fire clearcut areas that burned at high-severity from snag forest habitat created by high-severity fire in mature/old forest.