Jones et al. "Megafire" paper is bad science

(Post Updated Jan 2019.)
A paper by Gavin Jones et al. "Megafires: an emerging threat to old-forest
species" claims to describe negative effects of the 2014 King fire on California spotted owl occupancy and space use. There are several papers by other authors documenting how severe fire has little or no effect on spotted owl occupancy, and that they forage in high severity patches preferentially or in proportion to availability (Lee 2018). Jones et al. (2016) claim their data describe a strong negative impact of severe fire on spotted owls. The paper has fatal flaws in their analyses that render their results and discussion unreliable.

• Only 1 year of post-fire data which was erroneously compared with the mean occupancy rate of all previous years.
• Very small sample sizes. Authors of "Megafires: an emerging threat to old-forest
  species" are making a claim of large extinction effects from 8 severely burned, previously occupied sites, versus 9 unburned sites. Jones et al. fail to mention the high levels of post-fire logging that occurred in half of the burned sites. 4 of these 8 sites (those with the greatest amount of high-severity fire) had an average of 22% post-fire logging.
• Excluding known temporal trends from the occupancy analysis is the most onerous fatal flaw. Jones et al. neglected to include the temporal trend effect in colonization and extinction rates that were described in Tempel & Gutierrez 2013 using the same data up to 2010 as were used by Jones et al. (2016).
  

Figure 1 from Tempel & Gutierrez 2013 showing annual rates of occupancy, colonization, and extinction probabilities. Temporal trends existed in both colonization and extinction probabilities, but Jones et al. (2016) neglected to include these trends in their analysis of fire effects in 2015.

The population is in freefall, yet the authors compared their 1 year of post fire data against the mean of all previous years without accounting for the temporal trend in colonization and extinction. The observed long-term decline in annual site occupancy probability was the result of documented long-term trends of decreasing site colonization probability and increasing site extinction probability unassociated with fire (Tempel and Gutiérrez 2013). Jones et al. (2016) clearly described their method of partitioning their data into before-after and control-impact groups using only two dummy variables, and the first paragraph of their results and discussion compared estimates of extinction probability and occupancy probability from 2015 (one year post-fire) against the mean probability for all previous years.
       Clearly, Jones et al. (2016) did not account for the significant pre-fire temporal trends in colonization and extinction rates during their site occupancy analyses. This omission of temporal trends means their result for 2015 is due to the fact that this single year of post-fire data was the last year in the dataset, and should not be attributed unequivocally to the King fire. A simple linear trend of occupancy probabilities from 1993–2014 intersected the +1 SE error bar of the 2015 estimate (see below). This indicates that site occupancy in 2015 was not significantly different from the occupancy probability predicted for that year by the observed declining occupancy trend during the 22 pre-fire years.
     Jones et al. did a separate post-hoc analysis of trend in occupancy, but that was not included in the dynamic occupancy modelling. Including a pre-fire trend in the analysis would swamp the effect the authors attributed to fire. The temporal trend graphs Jones et al. should have presented and included in their dynamic occupancy analysis were the linear or quadratic trend (most likely the quadratic trend based on WebTable4)

Annual spotted owl site occupancy data from El Dorado density study area with linear trend.

Annual spotted owl site occupancy data from El Dorado density study area with a quadratic trend.

• Compositional analysis of foraging habitat as done by the Jones et al. 2016 Megafire paper is inappropriate for central place foragers like spotted owls. By omitting the distance to center effect and analyzing simplistic ratios rather than resource selection functions, radiotelemetry results and discussion are invalid. Jones et al. (2016) did not account for distance from the center of spotted owl sites in their foraging habitat selection analysis. In response, Jones et al. suggest that they accounted for distance by merely excluding the outer 5% of foraging locations in Jones et al. (2016). Jones et al. misunderstand the problem. Since spotted owls are central-place foragers during the breeding season, analyzing foraging habitat selection as a function of distance is critical because spotted owl foraging habitat selection studies routinely find distance from the site center significantly affects probability of use for all habitat types (Carey et al. 1992, Carey and Peeler 1995, Rosenberg and McKelvey 1999, Glenn et al. 2004, Irwin et al. 2007, Irwin et al. 2012, Bond et al. 2009, Bond et al. 2016). Thus, researchers can avoid an erroneous conclusion that owls are avoiding a particular habitat type when, in reality, that habitat type is simply distributed farther from the site center, on average, relative to other habitat types. This is precisely the error that Jones et al. (2016) committed, thus their results are intrinsically biased towards selection of habitat types near the nest (Rosenberg and McKelvey 1999), and their error is actually made worse because they omitted the farthest foraging points that reached into the largest high-severity burned area. Jones et al. point to several other published studies that did not account for distance from center, but pointing to numerous wrong analyses does not make the method correct.
         Of the spotted owl territories that Jones et al. (2016) chose for their foraging habitat selection, 5 out of 6 of the territories near the large tract of high-severity fire in the north-central portion of the King fire have site centers that are approximately 1500 to 3500 meters from the nearest edge of this large tract of high-severity fire, and the sixth has a site center approximately 1200 meters from the nearest edge of this large high-severity fire tract. Jones et al. (2016) state that they used an 1100-meter radius around site centers to spatially define spotted owl territories. Jones et al.’s claim that these spotted owls were avoiding this large high-severity fire tract is baseless, and is no more credible or logical than claiming that the owls were “avoiding” unburned old-growth forest 1200 to 3500 meters away from the site centers in the opposite direction of the King fire. Moreover, the portions of the large high-severity fire tract near the 3 owl sites close to the northwestern boundary of the King fire had been heavily post-fire logged at the time of the Jones et al. (2016) research.
  
  
• Inaccurate scholarship was evident in the text as well. e.g.,
  • p. 300 “In the western US, major reforms in forest fire management have been proposed to restore low-and moderate-severity fire regimes through forest tree thinning (North et al. 2015).” This citation neglects the literature supporting a natural mixed-severity fire regime in the Sierra Nevada.
  • pg. 304 "The observation that lower-severity fire is benign, and perhaps even moderately beneficial, to spotted owls is consistent with previous studies (Roberts et al. 2011; Lee et al. 2012)." Both those studies found no effect on occupancy from mixed-severity fire, as is common in the Sierra Nevada with a mix of low, moderate, and high severity burn.
  • p. 305, “because owls were not individually marked in the Rim Fire study, some detections at “occupied” sites may have involved individuals from neighboring territories or non-territorial “floaters” (Lee and Bond 2015), both of which may have contributed to inflated estimates of territory occupancy.” This exact same situation exists in the data analysed by Jones et al., data were collected as described in Tempel & Gutierrez 2013, “We included both nocturnal and diurnal surveys in our occupancy analyses.”  During nocturnal surveys leg bands are usually not resighted, therefore detections at occupied sites in Jones et al. would have been similarly inflated by individuals from neighboring territories or non-territorial floaters.
    

Wildfire has more benefits than costs for spotted owl (Strix occidentalis) populations, according to a quantitative meta-analysis of forest wildfires, including “megafires” (Lee 2018). The best available summary of research on spotted owls and fire is Lee (2018)—the only systematic review and quantitative meta-analysis to date regarding spotted owls and mixed-severity fire (5–70% of burned area was comprised of high-severity fire patches with >75% mortality of dominant vegetation). Lee (2018) analyzed all studies of spotted owls (all three subspecies) in relation to mixed-severity fires that had not been subjected to post-fire logging. Fifteen papers reported 50 effects from fire that could be differentiated from post-fire logging. Meta-analysis of mean standardized effects (Hedge’s d) found only one parameter was significantly different from zero, a significant positive foraging habitat selection for low-severity burned forest. Multi-level mixed effects meta-regressions (hierarchical models) of Hedge’s d against percent of study area burned at high severity and time since fire found: a negative correlation of occupancy with time since fire; a positive effect on recruitment immediately after the fire, with the effect diminishing with time since fire; reproduction was positively correlated with the percent of high-severity fire in owl territories; and positive selection for foraging in low- and moderate-severity burned forest, with high-severity burned forest used in proportion to its availability, but not avoided. Spotted owls were usually not significantly affected by mixed-severity fire, as 83% of all studies and 60% of all effects found no significant impact of fire on mean owl parameters. It is important to clarify that instances of upwardly biased occupancy rates from false positives (e.g. Berigan et al. 2019) do not have any bearing on the effect size as analyzed in Lee (2018). False positives occur at all sites, regardless of level of burn severity (Berigan et al. reported no difference in false positive rates by burn severity), thus the effect size remains the same even if absolute occupancy rates might be upwardly biased. Contrary to current perceptions and recovery efforts for the spotted owl, mixed-severity fire does not appear to be a threat to owl populations; rather, wildfires as they have been burning in recent years appear to have more benefits than costs for spotted owls.

The Jones et al. study is one of only 3 examples of severe fire having a strong negative effect on spotted owls. The study was funded by the US Forest Service, and California Department of Forestry and Fire Protection, groups with a financial interest in promoting thinning logging of our forests in the name of protecting them from severe fire.