Background

Patterns of wildfire in the southwestern United States have been dramatically altered in the last 150 years. Many southwestern mixed-conifer forests have been shaped by low-severity, high-frequency fire regime with a minimum fire return interval of 4–6 years and maximum fire return interval of 18–32 years, only rarely burning at higher severity. In addition to natural ignitions from lightning strikes, indigenous use of fire shaped patterns of burning at high frequency over thousands of years (Allen 2002). These fire regimes were interrupted around the 1850s with the Euro-American settlement (Allen et al. 2002; Brown et al. 2001; Coop et al. 2019; Swetnam and Baisan 1994). Post-settlement anthropogenic forces, namely climate change, fire suppression, and land-use practices, have shifted the historic fire regimes of these semi-arid forest systems towards uncharacteristic fire severity and frequency (Fulé et al. 2009; Singleton et al. 2019).

Due to anthropogenic fire regime alteration, many southwestern mixed-conifer forests are now vulnerable to uncharacteristically large, high-severity fires, leading to altered forest structure and a risk of conversion in vegetation community composition, including possible shifts to non-forest composition (Coop et al. 2019; Fornwalt et al. 2016). Fire size, severity, and frequency have increased throughout southwestern mixed-conifer forests in response to suppression-induced structural change (increased density and fuel accumulation; Fulé et al. 2009) and climate change (hotter, drier conditions; Westerling et al. 2006). Suitable conditions for fire in these forests are expected to continue increasing with climate change (Hurteau et al. 2014; Singleton et al. 2019) and have already led to return intervals as short as 3–14 years, including some high-severity reburns (Holden et al. 2010). Such changes have uncertain consequences for vegetation dynamics and wildlife habitat, having already led to the possible decline of many native vertebrates (Allen et al. 2002).

A species of particular concern in light of this fire regime change is the Mexican spotted owl (Strix occidentalis lucida), a recognized subspecies of the spotted owl, which commonly uses dense canopies in mature mixed-conifer forests within the Southwest (U.S. Fish and Wildlife Service 2012; 1995). This subspecies was listed as federally threatened under the Endangered Species Act in 1993 (58 Federal Register 14248) due to the threat of historic and continued habitat alteration from timber harvest practices (U.S. Fish and Wildlife Service 1995). Due to its protected status and overall population decline observed in its westerly distributed subspecies, the California spotted owl (Gutiérrez et al. 2017), the spotted owl has been the center of much controversy (Jones et al. 2020), initially raising conflict between timber harvest activities (a job-creating and economically profitable activity) and the conservation of old-growth forest valuable for the owl species (Simberloff 1987). The controversy has since shifted to conflicting views on the impacts of both high-severity fire and severe fire-mitigating activities (e.g., fuel reduction treatments) on all three subspecies (Ganey et al. 2017; Peery et al. 2019).

Since its initial listing, protective restrictions on timber harvesting near nest/roost locations and an increase of larger, more severe wildfires within the Mexican spotted owl’s range have shifted the recognized primary threat to this protected owl. Now, rather than habitat loss from timber harvest, the primary threat is identified as the increased risk of stand-replacing fires throughout the geographic distribution of the owl (U.S. Fish and Wildlife Service 1995; 2012; Wan et al. 2018). A stand-replacing fire is one that kills all or most of the living overstory trees in a forest stand, potentially initiating a pulse of regeneration. As the occurrence of larger, more severe fires continues to increase in southwestern mixed-conifer forests (Fulé et al. 2009; Hurteau et al. 2014; Singleton et al. 2019; Westerling et al. 2006), the effects on Mexican spotted owl habitat remain in question.

Despite the fact that stand-replacing wildfire is recognized as the primary threat to the Mexican spotted owl, and the fact that the area burned in its range is projected to increase by 13-fold in the next six decades (U.S. Fish and Wildlife Service 2012; Wan et al. 2019), uncertainties still surround the impacts of both high-severity fire and fire-mitigating activities (i.e., fuel reduction treatments) on the owl (Ganey et al. 2017; Peery et al. 2019). In comparison to other spotted owl subspecies, fewer studies have investigated the relationship between the Mexican spotted owl and fire (Ganey et al. 2017; U.S. Fish and Wildlife Service 2012), and of those limited studies, results focus on short-term effects and typically fail to account for fire severity, making it somewhat difficult to reconcile findings (Wan et al. 2018). One to four years after fire, Bond et al. (2002) found little impact of fire on owl survival, reproduction, and mate/site fidelity while Jenness et al. (2004) found occupancy/reproduction in burned areas was slightly lower than in unburned areas, but different severities across burned areas did not have a significant effect. Ganey et al. (2014a) found that owls foraged disproportionately in burned areas during winter, favoring post-fire habitat in their ecosystem. These three studies examined short-term (≤ 6 years) data after fire, and only limited analyses of the impact of severity were considered. Lommler (2019) observed that Mexican spotted owl occupancy within the perimeter of the Rodeo-Chediski fire, a fire that burned 187,000 ha in east-central Arizona in 2002, was lower than occupancy outside of the perimeter more than a decade after the fire. Mexican spotted owls avoided nesting/roosting in areas with ≥ 33% fire-killed canopy (moderate- to high-severity fire) 13–15 years following the fire, and a negative relationship between owl occupancy and salvage logging was also found, suggesting that such practices degrade owl habitat (Lommler 2019). Wan et al. (2020) found differences in the effects of fire severity on the Mexican spotted owl habitat; high-severity fire reduced nesting habitat suitability at broad scales more than lower severity fire when habitat suitability was modeled in the first 3 years after fire. Furthermore, Wan et al. (2019) found that stand-replacing fire is expected to increase dramatically within the range of the Mexican spotted owl over the long term (by 2080s), highlighting the need to understand the long-lasting effects of high-severity fire on forest structure.

We designed the current study to address the long-term (100-year timeframe) effects of fires of different severities on elements of mixed-conifer forest structure important for Mexican spotted owl nesting habitat—canopy cover, basal area, large/medium tree basal area, and large tree density (detailed in Table 1). This approach is based on nesting habitat being the assumed limiting factor for Mexican spotted owl populations (U.S. Fish and Wildlife Service 2012; Wan et al. 2020). We did this by quantifying whether forest structure met those conditions desired for nest/roost habitat (Table 1) at randomly placed points within the mixed-conifer forest that did not burn (control) and burned between 1921 and 2020 in the Lincoln National Forest, New Mexico.

Table 1 Minimum desired conditions, outlined in the first revision of the Mexican spotted owl recovery plan, for the management of Mexican spotted owl forested recovery nesting and roosting habitat in mixed-conifer forests of the Basin and Range East EMU, in which the study area falls (Tables C.2 and C.3 in U.S. Fish and Wildlife Service 2012)
Full size table

Methods

Study area

Our observational study was conducted in the Sacramento (222,117 ha) and Smokey Bear (171,350 ha) Ranger Districts of the Lincoln National Forest in the Sacramento Mountains of south-central New Mexico. The study area is contained within the Basin and Range East Ecological Management Unit (EMU), one of the five EMUs within the US range of the Mexican spotted owl (U.S. Fish and Wildlife Service 2012). Mexican spotted owls occur in high-elevation forested “sky islands” (isolated mountain ranges capped with forests and surrounded by low-elevation desert and grasslands) and deep, intertwined canyons within this EMU (U.S. Fish and Wildlife Service 2012). In the Lincoln National Forest, Mexican spotted owls primarily nest in the mixed-conifer forest (92% of nests; Ganey et al. 2013), so we restricted the study area to this forest type (Fig. 1) and above 2438 m (Ronco et al. 1983). Average precipitation was 76 cm/year at Cloudcroft, New Mexico (2640 m elevation), within the Sacramento Ranger District (from 1981 to 2010 30-year normal; Western Regional Climate Center 2021); 60–70% of precipitation comes during summer (July–September) monsoons, and the remainder is largely snowfall (Brown et al. 2001; Kauffman et al. 1998).

Fig. 1
figure 1

Map of the mixed-conifer (2438-m lower-elevation limit) study area within the Lincoln National Forest boundaries, New Mexico

Full size image

Mixed-conifer forests in the study area are dominated by white fir (Abies concolor) and/or Douglas fir (Pseudotsuga menziesii) with a common occurrence of southwestern white pine (Pinus strobiformis), ponderosa pine (Pinus ponderosa), and quaking aspen (Populus tremuloides; Brown et al. 2001; Ganey and Vojta 2011; Kauffman et al. 1998). Gambel’s oak (Quercus gambelii) is prevalent in much of the midstory, and the most frequently occurring shrub is mountain spray (Holodiscus dumosus). Mixed-conifer forest within the defined elevation range comprises 26% (132930 ha) of the Lincoln National Forest, and for comparison, roughly 8.5% of the Lincoln National Forest is designated as Mexican spotted owl protected activity centers (PACs; Ganey et al. 2014b). PACs are established to protect areas used by resident Mexican spotted owl(s) and represent a minimum of 243 protected hectares delineated around known nests/roosts—a 40-ha core area centered on a nest tree and an additional ≥ 203 ha potentially being used for foraging (U.S. Fish and Wildlife Service 2012).

The primary disturbances that historically shaped diversity in the mixed-conifer forests of the Lincoln National Forest included fire (Brown et al. 2001), but following the Euro-American settlement of the region in the mid-1800s, forest structure was expansively homogenized through activities such as logging, grazing, agriculture, and fire suppression (Ganey et al. 2013; Kauffman et al. 1998), similar to much of the mixed-conifer forest of the region. Prevalent commodity-oriented logging from the 1880s to the 1940s removed much of the heterogeneous, old-growth mixed-conifer forest on this landscape, and logging, although not at the same magnitude, still occurs throughout the study area with the intent of reducing fire danger levels (Kauffman et al. 1998; Ganey et al. 2013; USDA Forest Service). The percent of vegetation treatments that timber sales comprised on the Smokey Bear and Sacramento districts was 31% from 1980 to 1999 and reduced to just 6% from 2000 to 2010 (U.S.D.A. Forest Service n.d.). As a result, while 10–26% of mixed-conifer stands were estimated to be in the old-growth condition in 1880 prior to the Euro-American settlement, less than 5% of old-growth mixed-conifer stands in the Sacramento Mountains remained as of the late twentieth century (Kauffman et al. 1998). More recently, the Lincoln National Forest, along with six other southwestern national forests, underwent an approximately 1-year injunction that suspended timber harvesting due to the deleterious effects of such practices, namely even-aged silviculture, on Mexican spotted owl habitat (WildEarth Guardians v. United States Fish and Wildlife Service n.d.) until its dismissal in October of 2020 (U.S.D.A. Forest Service 2020).

Plot placement/stratification

Spatial information and severity data for fires within our study area dating from 1984 to the present (hereafter “recent fires”) were obtained from the Monitoring Trends in Burn Severity (MTBS) project (MTBS 2019), and spatial information for fires within our study area dating from 1921 to 1983 (hereafter “older fires”) was obtained from the National Wildfire Coordinating Group (NWCG 2020). It is important to note that historical reconstructions like the NWCG spatial information may underestimate the area burned on a landscape (Syphard and Keeley 2016). However, in an effort to ensure plots were defined by the correct disturbance category, we looked for visual evidence of historic fire (or lack thereof for the control plots) such as fire scars and char. To ensure that we captured variability in habitat that burned at different severities, we initially stratified plots in the recent fires by differenced normalized burn ratio (dNBR; MTBS 2019); however, final fire severity classes for analysis were not determined until after data collection from the field to improve classification accuracy (see the “Defining fire severity” section). Using Geographic Information Systems (GIS; ArcGIS version 10.7), we randomly generated 125 survey plots stratified by time of fire (2010s, 1990s–2000s, 1950s–1960s, 1920s, or control—no fire in the last > 100 years) in the 100-year timeframe and, for recent fires, by fire severity (high, medium, low). Additionally, all plots were above 2438-m elevation, were within 500 m of a trail or road, and fell outside of Lincoln National Forest’s documented logged areas and areas defoliated by insects within the last 21 years (temporal extent of data). No sites were located in areas that burned more than once in the last 100 years, and all plots were separated by > 100 m from other plots of the same age-severity class. Plots were located in distinct patches of burned forest and were not spatially autocorrelated with other plots of the same category of fire severity and decade (Figure S1; Table S1). Occasionally, plots that had initially been selected failed to meet these conditions due to deficiencies in the reference data used. In these cases, in-field adjustments were made to relocate plots away from disqualifying characteristics, or alternate random plots were generated in their place if necessary. Following data collection, 12 plots were permanently removed from the study for various discrepancies (e.g., error in sampling methods, or fire severity could not be determined—see the “Defining fire severity” section), resulting in further analysis of 113 total burned and control plots.

Sampling design

We conducted forest structure surveys during the summer of 2020. At the 113 plots, data were collected in a 0.1-ha circular plot centered on the nearest tree or snag of sufficient diameter at breast height (dbh) to approximate the minimum sized tree for Mexican spotted owl nesting (27 cm dbh; Seamans and Gutierrez 1995). This protocol maintained realistic potential nesting habitat condition by avoiding small trees and treeless areas (Ganey et al. 2013). If the plot had no tree or snag ≥ 25 cm, we centered the plot on the largest tree—this happened only once at a plot that burned moderately in the 1950s; this plot was centered on a tree 24 cm in dbh. Within the 0.1-ha circular plot, we established three bisecting 36-m transects for measuring canopy cover. We also established a nested 0.03-ha circular plot for measuring other elements of forest structure (see below). We modeled our plots after the sampling design of Ganey et al. (2013; Fig. 2).

Fig. 2
figure 2

Schematic diagram of the study's sampling design. The 0.1-ha circular plot contains a nested 0.03-ha circular plot centered on a tree or snag ≥ 25 cm in diameter at breast height. The density and status of trees, shrubs, and snags of different size classes were measured in these plots. The canopy cover was measured at each point every 2 m along the three transects that bisect the plot. The area shaded in darker blue-gray indicates the predetermined 0.067-ha portion from which tree-ring samples were taken to estimate fire severity at plots that burned > 50 years ago

Full size image

At the plot center, we recorded the aspect, slope, elevation, and Universal Transverse Mercator (UTM) coordinate pair. Aspect was defined as the cardinal direction of downhill, and the slope was measured using a Laster Technology, Inc. TruPulse 200 hypsometer. Elevation and geographic coordinates were recorded using a DeLorme Earthmate PN-60w. In the entire 0.1-ha plot, we counted, documented the status [alive, fire-killed, dead before fire, or dead-unknown; determined based on tree/snag characteristics described by Harvey et al. (2014)], and measured the dbh of all trees and snags ≥ 25 cm in dbh. Within a predetermined 0.067-ha section of the 0.1-ha plot (see Fig. 2) at older fire plots (n = 45), we used an increment borer to take tree core samples from as close to the tree base as possible from all live conifers ≥ 25 cm in dbh. If fewer than 10 trees occurred within this area, every live conifer ≥ 25 cm dbh in the entire 0.1-ha plot was sampled to assure sufficient sample depth. In the 0.03-ha circular nested plot, we counted, documented the status (fire-killed, dead before fire, dead after fire, established after fire, survived fire), and measured the dbh of small trees/shrubs (standing woody growth > 1.4 m tall and < 25 cm in dbh) and stumps/log bases at the older fire plots. The three 36-m transects bisecting the plot were randomly oriented and spaced evenly by 60°. We classified canopy cover as “open” or “closed” using a moosehorn densitometer along these transects at 2-m intervals, skip** the middle 4 m on the second and third transect to avoid oversampling at the plot center (total n = 50 two-measurements per plot).

Defining fire severity

Two different approaches had to be taken to define fire severity using field measurements based on the longevity of fire evidence over time—one approach for recent (later than 1984) fires and another, more time-intensive approach for older (before 1984) fires (Fig. 3). The recent fires were initially stratified by dNBR-derived severity classes, but these classes (yielding only one severity value per 900 m2) proved inconsistent with field observations at the spatial scale relevant to spotted owl nesting habitat (Figure S2). Thus, we recategorized the plot severity by percent fire-killed basal area of dominant trees (≥ 25 cm dbh) within 0.1-ha plots. Thresholds for the three severity classes were set at < 20% fire-killed dominant tree basal area (low severity) and > 70% fire-killed dominant tree basal area (high severity). These thresholds coincided with natural breaks in the distribution of the data (Figure S2) and severity classes defined in other studies (Agee 1996; Hessburg et al. 2016; Odion et al. 2014). Due to uncertainty in timing and agent of mortality, 4.7% of recently burned plots could not be classified into a severity class and were excluded from further analysis (Figure S3). Forty-one recently burned plots remained in the dataset after these exclusions (Table 2).

Fig. 3
figure 3

Workflow diagram for the two different methods of defining fire severity: one for recent (1984–2020) fires and another for older (before 1984) fires. Defining the severity of older fires necessitated a separate workflow because the status of trees as surviving a fire or establishing afterward could not be visually discerned in the field with the same level of confidence as recently burned plots. Furthermore, in older fire plots, there was no clear threshold at 70% fire-killed basal area, but a clear distinction existed between plots that had no evidence of fire-surviving trees and those that clearly did have fire-surviving trees. Thus, a slightly different classification scheme was used for older fires, designed to keep the distinctions of low-, moderate-, and high-severity fires as consistent as possible between old and recent fires

Full size image
Table 2 Plots (n = 113) were categorized by disturbance history, including the decade of the fire and severity. The time period 2010s consisted of fires in 2012; the time period 1990s–2000s consisted of fires in 2004 and 1994; the time period 1950s–1960s consisted of fires in 1967, 1956, 1953, and 1951; and the time period 1920s consisted of fires in 1927 and 1921
Full size table

The older burn plots required additional steps to define severity in order to avoid misclassification into higher-than-actual severity classes due to loss of evidence over time. Tree core samples enabled us to determine the tree establishment year and relate it to the plot burn year. We prepared and dated tree-ring samples according to standard dendrochronological methods outlined by Stokes and Smiley (1968) to determine individual recruitment years and, ultimately, inform if a tree was established before or after the fire. Along with the status of trees/snags ≥ 25 cm in dbh, we used observations of density and status (fire-killed, established after the fire, etc.) of stumps and log bases ≥ 25 cm in basal diameter that occurred within the sampling plot to determine each plot’s dominant tree density at the time of the fire. We determined severity classes from the percent density of trees ≥ 25 cm in dbh (dominant trees) that were fire-killed. Because this definition of severity class differs from our definition used for recent fire data, the high-severity class threshold differed also—only plots that may have had stand replacement (i.e., the highest extreme was 100% mortality) were considered high severity (Figure S4). This decreased our chance of misclassification of these plots where evidence of less-than-stand-replacing fire has deteriorated over many decades, while still allowing for three distinct classes of severity. Data from older fire plots exhibited more error in severity classification than plots from more recent fires. Due to this uncertainty, 12.8% of older burn plots could not be classified into a severity class and were excluded from further analysis (Figure S4). Thirty-four older fire plots remained in the dataset after these exclusions. Following stratification of the in situ measurements of severity, we had 113 total plots representing 13 different age-severity classes of fire history (Table 1).

Statistical analysis

After collecting forest structure data, we compared our measurements with the minimum desired conditions outlined in the first revision of the Mexican spotted owl recovery plan (Table 1; Ganey et al. 2016; U.S. Fish and Wildlife Service 2012) and then evaluated differences between stands of different severities and time since fire. All data processing, statistical analysis, and figure generation were performed in R statistical software (R Core Team 2020).

To detect structural differences between forest stands of different fire severity classes, we conducted a permutational multivariate analysis of variance (PERMANOVA) on the structural attributes. PERMANOVA tests and pairwise post hoc tests were conducted in the vegan and the pairwiseAdonis packages, respectively (Martinez Arbizu 2017; Oksanen et al. 2020).

To address how the effects of severity changed with time since fire, we performed three separate rounds of PERMANOVA, comparing control plots that had not burned in the last century (n = 38) with three different subsets of the data: (1) all burned plots (n = 75) regardless of time since fire, (2) only more recent fire plots that burned later than 1984 (n = 41), and (3) only older fire plots that burned earlier than 1984 (n = 34).

To further explore the effects of time since fire on habitat structure, we performed linear regression characterizing the relationship of each structural attribute with time since fire, stratified by severity class.

Results

Summary statistics

Only nine plots (7.96%) met all five desired conditions for nesting habitat simultaneously (Table 3). Of those nine plots, all were either unburned control plots or varying-aged low-severity plots. Still, in each of those categories, only up to 25% of plots of a given category satisfied all five desired conditions (13.2% of unburned, 25% of 1990s–2000s, 20% of 1950s–1960s, and 10% of 1920s). There was a wide spread of slope and aspect within the plots of all disturbance categories, and no predominant trends in slope or aspect were detected (Figure S5 and Figure S6).

Table 3 Distribution of the plots meeting all criteria simultaneously across their disturbance history categories (total basal area ≥ 33.3 m2/ha, percent medium tree basal area > 30%, percent large tree basal area > 30%, large tree density ≥ 37 trees/ha, and canopy cover ≥ 60%). Categories not listed (plots that burned at medium or high severity in any year) had 0% of plots that met all criteria)
Full size table

Relationship between fire severity and structural attributes

Across all plots, regardless of time since fire, the results revealed compelling evidence that high-severity plots were different in structure from every other disturbance category in every metric, with the exception of canopy cover between moderate- and high-severity fire plots (Table 4). When we compared only the control and recent fire plots, the data revealed strong evidence that high-severity plots differed in every structural metric from every other disturbance category, except the percent large tree basal area, for which only control and high-severity plots differed (Table 5). Additionally, data revealed strong evidence that recent moderate-severity fires also differed structurally in both canopy cover and total basal area from control plots (Table 5). In our third PERMANOVA analysis, in which we compared only control and older fire plots, differences in structure were muted. The only compelling evidence for differences in structure detected in this comparison was between high-severity plots and control or low-severity plots, and even these differences were only detected for the percent large tree basal area, large tree density, and total basal area (Table 6).

Table 4 Outputs from the Bonferroni-adjusted post hoc tests for pairwise comparisons between control and severity classes across all plots, regardless of time since fire. The “overall” category refers to a model that evaluates the five considered structure attributes together
Full size table
Table 5 Outputs from the Bonferroni-adjusted post hoc tests for pairwise comparisons between control and severity classes of recent fires (1984–2020). The “overall” category refers to a model that evaluates the five considered structure attributes together
Full size table
Table 6 Outputs from the Bonferroni-adjusted post hoc tests for pairwise comparisons between control and severity classes of older fires (1921–1983). The “overall” category refers to a model that evaluates the five considered structure attributes together
Full size table

 

Temporal trends in structural attributes with time since fire

Canopy cover is believed to be the most important single aspect of forest structure for the Mexican spotted owl nesting habitat (Ganey et al. 2016; Wan et al. 2017). Distinction in canopy cover between fire severity classes diminished with increased time since fire, demonstrated by low, moderate, and high severity categories all having medians from 62 to 67% canopy cover (all above the minimum desired condition of ≥ 60%) within 100 years following fire or sooner (Fig. 4). Following the 2010s fires, canopy cover decreased with increased fire severity and the medians of high- and low-severity classes varied by as much as 30% (and 50% in the 2000s and 1990s group).

Fig. 4
figure 4

Boxplots of canopy cover at each plot (n = 113) across severity and fire decade

Full size image

Another important metric of nesting habitat structure, the percent large tree basal area (trees > 46 cm dbh; Table 1), remained low (median < 15%) in high-severity plots relative to the other severity classes and well below the minimum desired condition of 30%, even a century after fire (Fig. 5). For moderate-severity plots, this metric varied greatly across the different fire decades—but the median was only below the minimum desired condition threshold of 30% for fires 50–70 years ago. For low-severity plots, the percent large tree basal area was lowest (and below the desired threshold) after the most recent fires, but median values rose above the threshold after 20–70 years of recovery and increased even higher by 100 years following fire.

Fig. 5
figure 5

Boxplots of percent large tree basal area at each plot (n = 113) across severity and fire decade

Full size image

The percent medium tree basal area (trees 30–46 cm dbh; Table 1) of high-severity plots increased over time with the median satisfying the minimum desired condition of 30% by a century following fire (Fig. 6). The median of this metric for low- and moderate-severity fire plots reached the minimum desired condition threshold sooner—low-severity plots exceeding 30% in less than 10 years and moderate exceeding 30% in 20–30 years following fire. However, the median percent medium tree basal area of both low- and moderate-severity plots decreased with time since fire in the absence of repeated fire.

Fig. 6
figure 6

Boxplots of percent medium tree basal area at each plot (n = 113) across severity and fire decade

Full size image

Regarding large tree density (trees > 46 cm dbh; Table 1), plots that burned even at low severity tended to fall short of meeting the minimum desired condition of 37 large trees/ha until 50–70 years following fire, when the median large tree density was 50 trees/ha (Fig. 7). In contrast, median large tree density in moderate-severity plots did not reach the desired threshold until 90–100 years following fire, and median high-severity plots failed to reach the large tree density threshold within the 100-year timeframe (Fig. 7). However, even these plots exhibit a trend in a positive direction (albeit slowly); while medians of sites that burned at high severity less than 90 years ago remain at zero large trees/ha, by 90–100 years following fire, a few high-severity plots surpassed the large tree density threshold for the first time.

Fig. 7
figure 7

Boxplots of large tree density at each plot (n = 113) across severity and fire decade

Full size image

The total basal area of control plots ranged widely, demonstrating the heterogeneity of these forests, but the median value was 37.6%, above the minimum desired total basal area of 30 m2/ha (Fig. 8). After a century of post-fire regeneration, the median values of this metric for low- and moderate-severity fire plots were above the desired threshold, but only one out of eight high-severity plots satisfied that criterion (with the median still below it by 2.5 m2/ha).

Fig. 8
figure 8

Boxplots of the total basal area at each plot (n = 113) across severity and fire decade

Full size image

Discussion

In the Lincoln National Forest, high-severity fire alters forest structure important for the Mexican spotted owl nest/roost habitat differently than lower severity fire, but this distinction diminishes with time (Fig. 9). Even so, across the study’s 100-year chronosequence, none of the high-severity plots met the desired levels (U.S. Fish and Wildlife Service 2012) of all the structural attributes simultaneously. Forest structure important for the Mexican spotted owl nest/roost habitat is most negatively affected by high-severity fire, and each structural element takes 70 years or longer to recover to a level that provides sufficient nesting habitat structure whereas thresholds for some of the attributes were often met within a single decade in recovery after lower severity fire.

Fig. 9
figure 9

Nine photographs visually demonstrating the post-fire recovery of mixed-conifer forest burned at different severity through a space-for-time substitution. Photographs were taken by S. Battye, T. Durboraw, N. Gill, and S. Iida

Full size image

In recent fires, important measures of forest structure at high-severity plots were significantly lower, and therefore less adequate for Mexican spotted owl nesting, than structure at plots of all other severity categories, with few significant differences among the other categories (low severity, moderate severity, and control plots). Sixty or more years after fire, there were no significant differences in the forest structure between moderate-severity classes and high-severity classes—only a distinction between high-severity and low-severity/control plots—suggesting that high-severity plots that were once different in structure than moderate-severity plots had achieved comparable structure within the timeframe. Canopy cover, the most important of the five attributes we investigated in predicting nesting habitat (Ganey et al. 2016), recovered from stand-replacing fires to levels matching those that are known to support Mexican spotted owls, but only after 90–100 years.

Ganey et al. (2016) found the desired minimum structural thresholds to be much greater than the structure typically observed at successful Mexican spotted owl nest sites in the Sacramento Mountains. However, even if the thresholds were hypothetically set 50% lower, the high-severity burned areas would still take ≥ 40–50 years to recover, a period of time that is well above the length of the historic mean interval between fires. While the impacts of high-severity fire on forest structure create long-lasting conditions unsuitable for Mexican spotted owls, management actions can be taken to reduce, though not eliminate, the potential for large, predominantly high-severity fires in the future (Evans et al. 2011; Hurteau et al. 2014). Additionally, structural recovery following low- and, to a lesser degree, moderate-severity fire is more amenable to the nesting of Mexican spotted owls.

While it is encouraging that forest stands burned at highest severity are eventually recovering to sufficient forest structure to support Mexican spotted owl nesting habitat, the results of the current study suggest the following question: is recovery occurring quickly enough? Before the Euro-American settlement, the fire regime of the study area was characterized by frequent, low-to-moderate-severity fires with a mean return interval of 8–16 years (Azpeleta Tarancón et al. 2018), but the trends we found suggest for the five structural attributes important to Mexican spotted owl nesting habitat, it will take ≥ 80–100 years for sufficient nesting structure in high-severity burned areas to develop Mexican spotted owl nesting structure comparable to that of more recent lower severity burns. Given the historic mean fire return interval in dry, mixed-conifer forests and the reality that the status quo of aggressive fire suppression is not sustainable (Dunn et al. 2017; Smith et al. 2018), it is possible that the high-severity burned areas will burn again before adequate recovery to nest habitat structural levels is achieved (Holden et al. 2010). Due to the minimal negative response of nesting habitat structure to low- and moderate-severity fires, setbacks to habitat recovery may be dampened if subsequent fires are of lesser severity than stand-replacing. However, Holden et al. (2010) demonstrated that on a similar landscape, reburns occurring 3–14 years after severe burns have a high probability of reburning at high severity. The study of habitat structure in reburned areas was beyond the scope of the current study, but this represents an important area for future research as stand-replacing fires are expected to increase. Also, fire size and severity have increased throughout southwestern forests in response to fire suppression and climate change, and conditions for severe fire in these forests are only expected to increase (Fulé et al. 2003; Hurteau et al. 2014; Singleton et al. 2019). Indeed, current predictions are for a predicted 13-fold increase in burned area over time due to climate change alone in the range of the Mexican spotted owl (Wan et al. 2019). These trends, paired with our findings of negative structural effects and slow recovery of severely burned areas, demonstrate an imminent threat to Mexican spotted owls through the reduction of forest with suitable structure for nesting habitat over time.

Mexican spotted owls evolved with and can likely persist alongside wildfire if a fire regime similar to historic patterns can be restored. However, restoration of stand structure is not sufficient for restoring fire regimes, given the rapidly warming climate. If low-severity fire can be promoted through prescribed burning or other means, it can reduce the risk of stand-replacing wildfire (Pollet and Omi 2002; Stevens-Rumann et al. 2016) while both minimizing timelines of nesting habitat recovery and creating ecological diversity that benefits Mexican spotted owls and numerous other species. Restoration practices that are intentionally maintaining large, old trees may allow spotted owls to persist while increasing the resilience of the system under a changing climate (Jones et al. 2021a, b).

Fuel reduction efforts that remove fuel ladders and aim to improve the survival of medium and large trees through the next fire are effective in reducing fire hazard and are thus important to broadly maintain the forest structure that provides habitat for nesting Mexican spotted owls and other wildlife (Agee and Skinner 2005; Jones et al. 2022). Restoration activities that focus on plant species composition of the canopy and understory as well as spatial patterns of open patches and grouped and scattered trees support biodiversity and are expected to improve forest resilience (Reynolds et al. 2013). These restoration activities may include post-fire reforestation (via seeding and/or outplanting), which can accelerate timelines to structural conditions that are suitable for wildlife species (Stevens et al. 2021) including Mexican spotted owls.

Severely burned areas are not completely unusable by Mexican spotted owls; in fact, on the nest site scale, increased pyrodiversity—namely, landscape variation in fire severities—leads to increased persistence of spotted owls (California subspecies; Jones et al. 2021a, b). More open canopy might increase foraging habitat (Bond et al. 2016; Hanson et al. 2018), but our current study shows that nesting habitat will likely be more limited than foraging habitat given the long period of time associated with post-fire recovery of stand structures. High-quality spotted owl nesting habitat may be found in isolated patches of unburned forest within burn perimeters (Andrus et al. 2021). Such patches may be protected from future high-severity fire (Meddens et al. 2018) and may continue to provide post-fire habitat for Mexican spotted owls, which have high site fidelity (Ganey et al. 2014b). Oftentimes, these patches persist through multiple fires (i.e., fire refugia) because of topographic position and/or high moisture content (Meddens et al. 2018), which also tend to make areas suitable for spotted owl nests. Also, Kramer et al. (2021) found that spotted owls do not avoid high-severity burned areas for foraging per se, but rather avoid larger patches of severely burned areas for foraging. However, as fires increase in severity and size (Singleton et al. 2019), our research demonstrates that the landscape will face an overall decrease in the structural conditions needed for nesting habitat, the abundance of which is identified as the primary limiting factor for Mexican spotted owls (U.S. Fish and Wildlife Service 2012). Perhaps the largest threat to the Mexican spotted owl is not necessarily the occurrence of stand-replacing fire, but rather, the root cause: the absence of lower severity fire and pyrodiversity.

Conclusion

By quantifying how forest recovers following low-, moderate-, and high-severity fires over a century, we were able to show that forest structure important for the Mexican spotted owl nest/roost habitat is most negatively affected by a high-severity fire. After a high-severity fire, each structural element took 70 years or longer to recover to a level that provides the most suitable nesting habitat structure. As fires increase in frequency, severity, and size, the Lincoln National Forest will experience an overall decrease in the structural conditions needed for Mexican spotted owl nesting habitat. Particularly, uncharacteristic high-severity fire, exacerbated by a lack of lower severity fire and pyrodiversity, poses an imminent threat to nesting habitat far into the future.