An Ecoregional Conservation Assessment for the Klamath-Siskiyou Ecoregion and Proposed Siskiyou Crest Climate Refuge, Southwest Oregon and Northern California, USA

Abstract

The Klamath-Siskiyou Ecoregion (KSE) of southwest Oregon–northern California, USA, has globally exceptional biodiversity but is experiencing mounting pressures from climate change and land uses. We conducted an ecoregional conservation assessment of the KSE and the Siskiyou Crest subregion (SCS), a proposed climate refugium within the KSE. We integrated protected area priorities based on established conservation targets with climate change planning and fire risk reduction for communities. Both areas contained very low levels (<30%) of protection (GAP status 1, 2) for nearly all land cover types (n = 17), including serpentine substrates where endemic plants are highly concentrated, older forests with potential refugia properties, and habitat for Northern Spotted Owl (Strix occidentalis caurina) and Pacific fisher (Pekania pennanti). At the ecoregional scale, high-severity fire levels were proportionately similar across GAP land-use status (“managed” vs. protected). However, high-severity fire was lowest for protected areas at the subregional scale, reflective of potential refugium properties. Most fuel treatments by federal agencies were >1 km from nearest structures, far removed from effective community fire protection in both locales. The relatively higher-elevation SCS is projected to maintain refugia properties (cooler, wetter) for longer periods than the KSE; however, that function may dissipate toward the end of the century and under a higher emissions scenario. We recommend increased protections of potential refugia combined with fire risk reduction of the built environment to more effectively maintain unique biota and prepare communities for increased likelihood of wildfire spillover events.

1. Introduction

Reserve design is the hallmark of conservation biology approaches that typically include representation of landforms, focal species, or vegetation types in large, interconnected protected areas [1]. To slow the extinction crisis, global conservation targets using reserve design approaches have been proposed such as 30 percent of lands and waters in protected areas by 2030, “30 × 30” [2].

Ecoregional conservation assessments (ECAs) are a means for meeting conservation targets by integrating strategies to address climate change and land-use impacts together [3]. ECAs include spatially explicit analyses of select ecosystems and focal species overlaid on land-use categories using the U.S Geological Survey Protected Areas Database (PAD) to determine gaps in protected area coverages (i.e., the USGS Gap Analysis Project or GAP). They include climate change downscaling to assess future risks to biodiversity and identify potential refugia for possible protection. Such an ECA methodology was piloted for the Mogollon Highlands Ecoregion of Arizona to New Mexico, including identifying the Gila Bioregion as potential refugium [4]; the Southern Rockies Ecoregion of Wyoming to New Mexico, including Santa Fe, New Mexico, watersheds as potential refugia [5]; and the Northern Rockies Ecoregion of Washington to Montana, including the Yaak Valley watershed as potential refugium [6]. ECA approaches can assist governments and conservation groups in prioritizing protected area additions that might provide time for species to adapt to the unprecedented pressures of land-use changes and climate change. Additionally, such multi-scaled spatial analyses that integrate land management and ecological data can broadly inform short- and long-term environmental and land-use planning [7].

The Klamath-Siskiyou Ecoregion (KSE) of southwest Oregon and northern California has a history of conservation interests due to its widely recognized biodiversity importance [1,8,9,10], including recognition as a World Wildlife Fund “Global 200 Ecoregion” [11]. The KSE is considered a “center of floristic diversity and narrow endemism” [9]. The ecoregion supports a continental maximum of conifer species [12] along with exceptional mollusk (Mollusca); butterfly (Lepidoptera); bee (Apidae) [13]; and amphibian richness and endemism ([10,14] for reviews). Notably, the KSE has the most species-rich herpetofauna of any similarly sized mountain range in the Pacific Northwest, in part due to eight endemic species [15]. Plant endemism on ultramafic substrates is exceptional, promoted by the combination of extensive serpentine areas and a moderate inland to coastal climate [16].

The KSE also has presumed refugia properties related to its moderate climate [8,10,17] and the lack of volcanic activity and glaciation events. In particular, old forests on north-facing slopes [18] have been proposed as potential refugia in this ecoregion presumably because of favorable microclimatic properties (cool, wet). The primarily east–west running, high-elevation Siskiyou Crest Subregion (SCS) within the KSE is also noted for its high biodiversity and forest carbon importance that is greatest in older forests [19], along with its land-bridge functions connecting the Cascade and Coast ranges [18]. This subregion has been proposed for protection by conservation groups (https://siskiyoucrestcoalition.org/; accessed on 8 May 2026) that requested an ECA to help align protected area strategies with presumed refugium properties of the Crest. Moreover, our ECA updates prior conservation planning in the KSE [1].

Main climate stressors in the KSE include increases in extreme temperatures, more frequent droughts, insect outbreaks, and wildfires [20,21]. Such changes may impact moisture dependent taxa (e.g., salamanders, mollusks) [10,18], endemic plants with restricted ranges and “relict” populations [22], and species that use older forests as wildfire [23] and climate refugia [24]. Additionally, logging, roads, mining, off-highway vehicles, and livestock grazing are cumulative stressors impacting large portions of the KSE [10,25,26]. Such stressors may interact with climate change to accelerate biodiversity loss (e.g., logging x extreme fire weather [27]).

We provide a multi-scaled ECA for the KSE and the SCS that is based on representation analyses of select ecosystems and focal species in relation to recognized conservation targets, including 30% protected by 2030 (i.e., “30 × 30” [2], herein low target), 50% protected by 2050 (i.e., “50 × 50” [28], herein intermediate target), and 100% protected (no timeline, herein upper target) of select types. The upper bound (100%) target is focused on older forests [29] and federally Inventoried Roadless Areas (IRAs) due to their conservation importance [30]. We also chose the Pacific fisher (Pekania pennanti) and Northern Spotted Owl (Strix occidentalis caurina, a federally threatened species; NSO) as focal forest vertebrates because of their association with older forests [31,32] and the availability of datasets met the spatial analysis requirements of our study.

We also examine wildfire severity in relation to land-use categories (i.e., GAP status, see below) mainly because protected area proposals have frequently been blocked by decision makers over concerns about wildfires spilling into urban areas with management responses aimed at unprecedented increases in fire suppression and various forms of logging, road building, and prescribed burning (slash pile and prescribed fire) (e.g., the Fix Our Forests Act H.R. 471, see https://www.congress.gov/bill/119th-congress/house-bill/471, accessed on 24 April 2026, and numerous land management forest planning documents). Thus, we integrate conservation priorities with wildfire risk reduction for communities living in the so-called wildland–urban interface (WUI) as defined by federal agencies. Like in the other ECAs, our findings may be exportable to other ecoregions facing similar stressors that require integrated conservation strategies.

2. Materials and Methods

2.1. Study Area

2.1.1. Ecoregion

The ~4.83 M ha KSE spans the Klamath and Siskiyou Mountains of southwest Oregon and northern California (Figure 1). We used the U.S. EPA Level III Ecoregion (#78, Klamath Mountains/California High North Coast Range, L3 Code:6.2.11) to define the study area. This ecoregion dataset was originally derived from Omernik [33] and from mapping done in collaboration with U.S. EPA regional offices, federal agencies, and state resource management agencies [34]. There are some boundary differences with the WWF Global 200 Ecoregion #39 (Klamath-Siskiyou Forests), which spills over into portions of coast redwood (Sequoia sempervirens) and includes more watersheds at the northern boundary of the KSE. Our KSE map is also like Bailey’s [35] Klamath Mountains and Southern Cascades (M261A, M261D); however, Bailey also included portions of the Cascades not included in our map (see [11] for comparisons of ecoregional boundaries).

The KSE is considered to have “central significance” [9] in adjoining the Coast Range Ecoregion to the west, the Southern and Central California Chaparral and Oak Woodlands Ecoregion to the south, the Cascades and Eastern Cascades Slopes and Foothills Ecoregions to the east, and Willamette Valley Ecoregion to the north [36].

Using a continental dataset of multiple taxa, Ricketts et al. [11] (Appendix C) reported that the KSE supports 1859 vascular plant species with up to 150 plant endemics. DellaSala et al. [10] identified several distinctive, rare and imperiled taxa in the KSE. Kauffmann and Garwood [12] listed 35 conifer species, which is considered globally significant among temperate conifer forest regions [10]. Several conifers are considered to have relict populations (i.e., more common in a different geological period but restricted today) [37]. Some conifers also have range extensions mainly on cool, moist slopes including Alaska yellow cedar (Callitropsis nootkatensis, relict populations), Pacific silver fir (Abies amabilis, relict populations), subalpine fir (Abies lasiocarpa, relict populations), Engelman spruce (Picea engelmannii, relict populations), and noble fir (Abies procera). Others conifers have range extensions on hot, dry slopes, including western juniper (Juniperus occidentalis), foxtail pine (Pinus balfouriana, relict species), and gray pine (Pinus sabiniana) [12,38]. Additionally, three conifer species are endemic to the KSE, including Brewer spruce (Picea breweriana) (relict populations), Baker’s cypress (Hesperocyparis bakeri), and Port Orford cedar (Chamaecyparis lawsoniana) [12]. The KSE also overlaps with the northern extension of the California Floristic Province, a recognized global biodiversity hotspot [39]. In sum, the diverse geological, topographic, and climatic processes in the KSE have resulted in zonal vegetation considered more complex than the nearby Sierra Nevada and Cascade ranges [17].

Older forests of the KSE have multi-layered canopies characterized by tall evergreen needleleaf trees overtopping evergreen broadleaf and sclerophyllous vegetation [9] that ostensibly confer fire-resistant properties as refugia [40]. Xeric sites are dominated by Douglas-fir (Pseudotsuga menziesii) and ponderosa pine (Pinus ponderosa) mixing with a rich assortment of oak (Quercus spp.) woodland and chaparral [12]. Upper elevations are dominated by firs (Abies spp.).

2.1.2. Subregion

The 697,755 ha SCS (14.5% of the KSE) is the only high-elevation land bridge (east–west) in the ecoregion (Figure 1). Original mapping of the SCS was provided by the Siskiyou Crest Coalition (local conservation groups) that identified the upper-elevation land bridge importance. Their mapping included watersheds intersecting the Siskiyou Crest and the geographically distinct land masses and boundaries that note the convergence of the Klamath-Siskiyou Mountains with the Cascade Mountains and Coast Ranges of California and Oregon (https://siskiyoucrestcoalition.org/; accessed 4 May 2026). The SCS also contains the watershed divide between the Klamath and Smith watersheds and the Klamath and Rogue watersheds (see Figure 1). The area has local significance as the 1850s proposed “State of Jefferson” that demarked the large mountain range dividing Siskiyou County, California, from southwestern Oregon. The boundaries of the region thus include the Coast Range and Middle Fork Smith River to the west, the Cascade Mountains to the east of Siskiyou Summit, the Rogue and Illinois river valleys to the north in Southwest Oregon and the north bank of the Klamath River from Cottonwood Creek downstream to Blue Creek.

2.1.3. Climate and Wildfire

Climatic conditions in the KSE are generally Mediterranean with wet, cool winters and dry, warm summers (e.g., Koppen climate classification Csb [41]) characterized by steep climatic gradients from the coast (wet, fog, cool) to inland (dry, warm) and from low (seasonal rain) to upper elevations (snow) (see Table 9.1 in Skinner et al. [17] for temperature and precipitation figures). Wet and dry alternating climatic cycles, and therefore fire activity, are on long timelines in relation to the Pacific Decadal Oscillation [42]. Fires tend to produce mixed severity effects on plant communities characterized by large and small patches of low, moderate, and high severity [43] that support exceptional biodiversity. Fires were more prevalent ecoregionally during the pre-fire suppression era and were often set by Indigenous peoples at local scales [17]. Fires also tend to be more naturally (unaided) prevalent at low-elevation, xeric sites (especially south slopes) vs. longer fire return intervals in mesic and upper elevations [17].

Contemporary wildfires are increasingly driven by extreme fire weather [21] interacting with heavily logged landscapes [27] that result in fast-moving wildfires that can spill over into towns juxtaposed with wildlands [44]. High road densities in the KSE [25] add to wildfire risks from unwanted human-caused ignitions [45,46]. Additionally, reoccurring high-severity fires have functioned as a self-perpetuating process for rejuvenating fire-dependent sclerophyllous plants mainly on south-facing and xeric environments [47]. The KSE has also experienced overlapping short-interval (<20 years), high-severity fires in forests, which despite repeat stand-replacing events, still contain plant species characteristic of pre-fire events [48]. There is also high taxa richness in the ensuing complex early seral forests [49], including high bird [50] and small mammals [51] richness with the latter providing prey for species that hunt in open, shrubby areas and nest in unburned (or low-severity) old forest patches (fire refugia) such as NSO [52]. However, others have reported the potential for type conversions of conifers to hardwoods if the interval between high-severity fires is further reduced by climate change [53] or unplanned and frequent wildfire ignitions [46].

2.2. Mapping Methods

We generally followed the methods outlined in DellaSala et al. [6]. We obtained GIS data from a variety of sources (Figure S1), reprojected to a CONUS Albers projection (EPSG:5070) and clipped geospatial datasets to the study area using QGIS version 3.44. For any of our GAP representation analyses, we rasterized vector datasets using a 30 m grid aligned with LANDFIRE (2024, LF 2.5.0) rasters. All clipped and reprojected rasters we used in our GAP representation analyses therefore had the same parameters except for pixel values. We stacked these rasters into a multiband virtual raster using the Geospatial Data Abstraction Library (GDAL) in Python 3.11.14. We then used the rasterio, pandas, and numpy Python libraries to parse the virtual raster data, calculate the area for each unique combination of stacked pixel values, and store results in a SQLite database. We used pandas and the sqlite3 library to perform a series of SQL queries on our SQLite database from within Python and create summary tables across all our metrics except those used in our fuels and WUI analyses and our climate change analyses. All maps, except those in our climate change analyses, were created in QGIS 3.44.

2.3. Topography and Watershed Boundaries

We obtained elevation and aspect data from LANDFIRE. For elevation, we created six 500 m bins, with the highest elevation bin being >2500 m. We grouped degrees aspect into five standard categories: flat, north, east, south, and west. For watersheds, we obtained Hydrologic Unit Code (HUC) 6 and 8 polygons from the United States Geological Survey (USGS) Watershed Boundary Dataset. While we conducted a GAP distribution analysis across all watersheds at both scales, we were most interested in analyzing HUC6 watersheds at the KSE scale and HUC8 watersheds at the SCS scale.

2.4. Land Ownership/Management and GAP Status

We used the PAD (PAD-US 4.0, [54]) to determine land ownership or management entity across the study area and consolidated land management entities into seven groups (Figure S2). Our methods were similar to DellaSala et al. [4,5,6] who included extracted USGS GAP status codes 1–4 from the PAD. GAP 1 is assigned to lands managed for biodiversity, where disturbance events proceed or are mimicked (e.g., federally designated Wilderness areas). GAP 2 is assigned to lands managed for biodiversity, but where disturbance events are suppressed (e.g., national monuments). GAP 3 is assigned to lands managed for multiple uses and may be subject to extraction such as mining or logging as well as livestock grazing and off-highway vehicle use (e.g., non-Wilderness national forest lands). GAP 4 is assigned to lands with no known mandate for biodiversity protection (e.g., private lands). We considered both USDA Forest Service-designated and -managed Inventoried Roadless Areas (IRAs) and U.S. Department of Interior Bureau of Land Management (no date) “Lands with Wilderness Characteristics” (LWCs) that the agency has identified as managed “to protect” as having enhanced protection beyond what is usually assigned GAP 3 status. We therefore split GAP 3 into GAP 3a and GAP 3b, assigning IRAs and select LWCs (that are managed “to protect”) a status of 3a and all other GAP 3 lands a status of 3b. We consider lands with a status of either GAP 1 or 2 to be “protected” and lands with a status of either GAP 1, 2, or 3a to be “protected+” to identify landscapes that can help meet 30 × 30 or 50 × 50 targets both with and without IRAs and select LWCs. Notably, while “protected+” is not a legal designation, we created this category to demonstrate the relative contributions of GAP3a lands to the conservation targets if they received formal protections. IRA polygons were obtained from the PAD while LWC polygons were obtained for the Oregon portion of the KSE from the U.S. Department of Interior Bureau of Land Management [55]. We did not include LWC polygons from the California portion of the KSE as those GIS data, if they exist, are not publicly available. However, ~80% and 99% of U.S. Department of Interior Bureau of Land Management lands in the KSE and SCS, respectively, are in Oregon, where LWC data were available. And only 6.7% of these lands in Oregon are LWCs managed “to protect,” thus we expect the overall contribution of such LWCs in California, if they exist, to GAP 3a to be relatively minor. Modified GAP status is depicted for the entire KSE and SCS in Figure S3.

2.5. LANDFIRE Biophysical Settings and Existing Vegetation Types

For vegetation types, we used LANDFIRE Biophysical Settings (BPS) combined with agricultural and developed vegetation types from the LANDFIRE (2024, LF 2.5.0) Existing Vegetation Type (EVT) dataset. Based on initial comparisons of EVT data to high-resolution satellite imagery across post-fire landscapes in our study area, we determined that BPS would be a better representation of vegetation types in the study area. Disturbances such as moderate- and high-severity fire can significantly change EVT classification for a given area. For example, many mixed-conifer forests that experience high-severity fire are subsequently classified as shrubland in the EVT dataset, likely due to a change in canopy structure. However, these areas may still be naturally regenerating as part of a post-fire succession process that will eventually result in a return of mature mixed-conifer canopy. We therefore did not want to classify complex early seral forest (CESF) [49] as non-forest vegetation types when analyzing protection levels. BPS is not sensitive to natural disturbances like EVT, but it also does not account for areas that have been converted to agricultural use or any type of development. To effectively leverage the long-term natural vegetation classification stability of BPS while still accounting for agricultural or developed lands, we combined the two datasets and kept BPS vegetation types except where they intersected agricultural and developed EVTs, in which case we retained the EVTs. We then organized these modified BPS vegetation types into 17 broader categories (see Table S1) that represent consolidated forest types, shrubland, grassland, and other land covers similarly to DellaSala et al. [5,6]. This allowed us to simplify our GAP representation analyses across vegetation types.

2.6. Late-Successional and Old-Growth Forest Distribution

We obtained late-successional and old-growth forest distribution spatial data from the Northwest Forest Plan Monitoring Program’s annual old-growth structure index (OGSI) [56] and GIS datasets [57]. OGSI data use two age thresholds: ≥80 years but <200 years (herein, OGSI 80) indicates late-successional (mature) forests; ≥200 years (herein, OGSI 200) indicates old-growth forests. Annual data for the entire KSE were available from 1986 to 2024. These raster datasets included four contiguity categories for each OGSI age threshold: core, edge, finger, and scatter. For our analyses, we consolidated these contiguity categories into a single class.

We only used data from 1986 and 2024 for our analyses. The 2024 data were used for GAP analyses of “current” OGSI distribution across the study area. We also analyzed GAP representation across each OGSI age threshold that persisted from 1986 to 2024 despite experiencing at least one fire (see Section 2.8 below for methods on fire data collection). That is, if a cell in our raster grid was identified as OGSI 80 in both 1986 and 2024, and it burned in at least one wildfire from 1986 to 2023, it was considered to have “persisted.” We also analyzed GAP representation of areas that shifted from OGSI 80 to OGSI 200 between 1986 and 2024 (“promoted”) despite burning in at least one fire. We did not consider fires that burned in 2024 for these persistence and promotion analyses as our inspection of the spatial data indicated that most fires did not affect the OGSI data until the year after the fire. For example, a large high-severity fire patch occurring in 2020 would not appear as a loss of OGSI 80 or 200 until the 2021 OGSI datasets. Therefore, 2024 OGSI data would only reflect fires that occurred up to 2023.

2.7. Focal Species Suitable Habitat and Serpentine Areas

We held an experts workshop in August 2024 to identify conservation priorities for analysis and compile information on taxa that might benefit from presumed refugia in the KSE and the SCS. At the workshop, we chose focal taxa based on habitat associations with old forests and availability of published datasets that fit the multi-scaled analysis. Nearly all available taxa datasets were not at the scale of our analysis (ecoregion, subregion) and some had inherent sampling biases, for instance, with data points concentrated along roads. Thus, our focal species analysis is limited to 2 forest vertebrate species that use old forests and have conservation relevance.

For NSO, we collected habitat suitability data from the Northwest Forest Plan Monitoring Program (similarly to OGSI data), which were available as annual rasters from 1986 to 2024 [58]. NSO habitat data were classified into unsuitable, marginal, suitable, and highly suitable. We analyzed only the GAP status distribution for suitable and highly suitable habitat in 1986 and 2024 that persisted despite experiencing at least one fire through 2023.

For the Pacific fisher, we analyzed GAP status distribution for the entire study area for the Connectivity Conservation Priority Areas [59], which used habitat suitability models to delineate core areas and least-cost corridor models to identify linkages among them.

We also used a serpentine geology dataset that was manually digitized by Noss et al. [1] using 1:500,000 USGS geology maps for both Oregon and California and spanning our entire study area.

2.8. Monitoring Trends in Burn Severity

We collected fire perimeter and fire severity data from the Monitoring Trends in Burn Severity (MTBS) database only available from 1985 through 2023. We limited our analyses to fires categorized as wildfire, wildland fire use, and unknown—we did not include prescribed fire. We used the relativized delta normalized burn ratio (RdNBR) dataset for each fire and classified RdNBR into unprocessed (i.e., areas that had no RdNBR values due to cloud cover or satellite image collection issues), unchanged, low-, moderate-, and high-severity fire according to established thresholds [60]. We then mosaicked these RdNBR class data by year. These annual mosaics allowed us to calculate the number of fires a given location in our study area experienced over the study period. We also combined all annual mosaics into a single fire severity footprint dataset, where the highest fire severity a cell in our raster grid experienced over the study period was retained. We then analyzed the GAP status distribution across each fire severity class footprint. We conducted this analysis once for all land cover categories combined and then again for only the more-frequent-fire, drier-forest types found in the KSE: dry Douglas-fir forest and woodland, dry mixed-conifer forest and woodland, mesic mixed-conifer forest and woodland, and ponderosa pine forest, woodland and savanna.

2.9. Fuel Treatments and the Wildland-Urban Interface

Most public land in the study area is managed by either the USDA Forest Service or the Bureau of Land Management (Figure S2). As in our previous ECAs [5,6], we analyzed the location of fuel treatments in relation to the wildland-urban interface (WUI). We obtained WUI data from Radeloff et al. [61] by extracting low-, medium-, and high-density wildland–urban interface and intermix polygons similar to [5,6]. We considered any land within these six categories to be WUI. We also created buffers around WUI areas in 250 m increments up 1000 m. We collected fuel treatment data from the USDA Forest Service’s Hazardous Fuel Treatment Reduction (polygon) dataset [62] and the Bureau of Land Management’s Historical Fuels Treatment Polys dataset (itself pulled from the National Fire Plan Operations and Reporting System) [63] from the last 20 years of our study period (2005–2024). We selected this date range for our analysis as we were primarily interested in contemporary fuel treatment patterns and because pre-2005 fuel treatment data were limited. We filtered the large, national datasets for each agency to only treatments with actual completion dates between 2005 and 2024 and with an activity or type name of chipping, crushing, mastication, mastication/mowing, broadcast burn, jackpot burn, machine pile burn, and thinning, and then clipped them to the KSE. Each agency’s dataset also has information about whether the agency classified the treatment polygon as WUI (i.e., a true/false field called “IsWui” or “ISWUI”). For each agency we dissolved all treatments by the agency’s WUI classification and then intersected these dissolved treatment polygons with the 250 m WUI buffers. Any treatment polygons or portions of polygons outside of the WUI buffers were assigned a WUI distance of >1000 m. We did this at both the KSE and SCS scale and calculated the total hectares and percentages of treatments at each buffer distance and the proportions of these areas considered WUI or not WUI by the respective agency.

2.10. Climate Change Downscaling

We used the global climate model (GCM) output from phases 5 and 6 of the Coupled Model Intercomparison Project (CMIP) for climate projections [64,65]. These two modeling efforts established common future climate scenarios under which all global climate models were forced. Climate scenarios for CMIP5 are called representative concentration pathways, or RCPs, while for CMIP6 we updated the scenarios to shared socio-economic pathways, or SSPs. RCPs from the CMIP5 models are roughly comparable to the SSPs from CMIP6 except they now have socio-political narratives attached to each scenario. For both CMIP5- and CMIP6-based future climate projection data, we selected both an “intermediate” future scenario (RCP 4.5 and SSP245) and a “very high” future scenario (RCP 8.5 and SSP585). We refer to RCP 4.5 and SSP245 scenarios as “intermediate” and the RCP 8.5 and SSP585 scenarios as “very high” due to their relative amounts of greenhouse gas emissions and to maintain consistency with international climate reports [66,67].

Raw GCM projection output from the CMIP models has resolutions that are typically greater than 1° latitude x 1° longitude. Because the data are so coarse, it is difficult to interrogate the data for meaningful information at regional scales, like ecoregions. To make the data more relevant for regional-scale assessments, we used climate model projections that were “downscaled” to a finer resolution through either statistical or dynamical downscaling. Statistical downscaling uses empirical relationships and statistical methods to interpolate data to a higher resolution, and dynamical downscaling uses a regional climate model forced by the GCM data to create a dataset with higher resolution.

For several of the climate indicators assessed in this study, we used GCM data that have been statistically downscaled by the localized constructed analogs method version 2, which we hereafter refer to as LOCA2 data [68]. The LOCA2 method uses empirical relationships between large-scale atmospheric climate patterns and on-the-ground conditions to bias-correct the global climate model output. Other indicators require dynamical downscaling as they are not a direct output from global climate models; namely, indicators that are related to hydrology, like snowpack. To identify the projected change in April 1st snow–water equivalent, we used hydrologic projections from the Integrated Scenarios Dataset Collection [69]. That model used the Variable Infiltration Capacity (VIC; [70]) hydrologic model forced by a statistically downscaled global climate model dataset: the Multivariate Adaptive Constructed Analogs version 2—Livneh dataset [71]. This statistically downscaled dataset uses data from Livneh et al. [72] to train the statistical downscaling method. Finally, to better understand future wildfire conditions, we used output from the modeling effort conducted by Sheehan et al. [73], who used statistically downscaled global climate model projections to run the MC2 dynamic global vegetation model for multiple future greenhouse gas scenarios both with and without fire suppression. The LOCA2 data uses the CMIP6 future climate scenarios (SSPs), and the MACAv2-Livneh and MC2 wildfire projections use the CMIP5 future scenarios (RCPs). Information on the specific models used for each climate indicator can be found in Table S7.

3. Results

3.1. Representation Analyses

3.1.1. Topography and Hydrography

Protected areas are highly skewed toward upper (>1500 m) elevations in the KSE but not as much as in the SCS that had low levels of protection across all elevations (Figure 2). Notably, IRAs and select LWCs are also skewed toward upper elevations (>1500 m) in both areas (orange color).

There are 35 HUC8 watersheds within the KSE and seven HUC8 watersheds within the SCS (Table S2). For the KSE, six HUC8 watersheds met at least the lowest target (Chetco, Middle Fork Eel, Salmon, Smith, Trinity, Upper Cache) with four additional watersheds (Cottonwood Creek, Illinois, Lower Klamath, and McCloud) meeting at least the lowest target with the inclusion of protected+. Notably, the Salmon HUC8 watershed approaches 70% under protected+. For the SCS, only one of the HUC8 watersheds (Smith HUC8) met the lowest target with the addition of the Lower Klamath meeting the target if protected+ is included (Table S2).

3.1.2. Land Ownership/Management

Approximately one-half to two-thirds of the KSE and SCS, respectively, are under the management of the USDA Forest Service where there are opportunities for large-scale conservation (

Table 1. Landownership and GAP categories for the KSE and SCS. IRAs and select LWCs (managed “to protect”) are GAP 3a. “Other” includes lands owned or managed by local jurisdictions (e.g., city or county) and non-governmental organizations.

Klamath-Siskiyou Ecoregion
Land Management Entity	GAP ha
	(%)					Total ha
	1	2	3a	3b	4	(%)
U.S. National Park Service	417	17,219	271	7	908	18,822
	(2.2)	(91.5)	(1.4)	(0.0)	(4.8)	(0.4)
USDA Forest Service	601,364	108,472	412,844	1,280,587	0	2,403,267
	(25.0)	(4.5)	(17.2)	(53.3)	(0.0)	(49.7)
U.S. Bureau of Land Management	21,169	17,493	25,650	399,271	0	463,582
	(4.6)	(3.8)	(5.5)	(86.1)	(0.0)	(9.6)
Tribal	0	0	0	0	58,943	58,943
	(0.0)	(0.0)	(0.0)	(0.0)	(100.0)	(1.2)
State	524	2940	4	16,320	2109	21,897
	(2.4)	(13.4)	(0.0)	(74.5)	(9.6)	(0.5)
Other	619	5583	0	4441	3974	14,617
	(4.2)	(38.2)	(0.0)	(30.4)	(27.2)	(0.3)
Private	111	1309	1	2235	1,851,023	1,854,677
	(0.0)	(0.1)	(0.0)	(0.1)	(99.8)	(38.4)
Total ha	624,204	153,016	438,769	1,702,860	1,916,957	4,835,805
%	(12.9)	(3.2)	(9.1)	(35.2)	(39.6)	(100.0)
Siskiyou Crest Subregion
U.S. National Park Service	406	647	271	7	907	2237
	(18.1)	(28.9)	(12.1)	(0.3)	(40.5)	(0.3)
USDA Forest Service	80,885	19,572	78,210	278,706	0	457,373
	(17.7)	(4.3)	(17.1)	(60.9)	(0.0)	(65.5)
U.S. Bureau of Land Management	0	186	4536	84,746	0	89,469
	(0.0)	(0.2)	(5.1)	(94.7)	(0.0)	(12.8)
Tribal	0	0	0	0	123	123
	(0.0)	(0.0)	(0.0)	(0.0)	(100.0)	(0.0)
State	0	37	0	1491	322	1850
	(0.0)	(2.0)	(0.0)	(80.6)	(17.4)	(0.3)
Other	0	42	0	436	40	517
	(0.0)	(8.1)	(0.0)	(84.3)	(7.7)	(0.1)
Private	0	143	0	693	146,351	147,187
	(0.0)	(0.1)	(0.0)	(0.5)	(99.4)	(21.1)
Total ha	81,291	20,626	83,017	366,079	147,742	698,755
%	(11.6)	(3.0)	(11.9)	(52.4)	(21.1)	(100.0)

). There are some large contiguous GAP 1 areas across the region (Figure S3), but overall protection levels are well below the lowest target with only 15% and 14.5% protected for the KSE and SCS, respectively. Protected+ would approach 25% protected for both areas, which is still below the lowest target.

3.1.3. Land Cover

We mapped 71 modified BPS land cover types in the KSE and 44 in the SCS (Table S1) that were then consolidated into 17 broader categories for the representation analysis (

Table 2. Modified BPS subgroups for the KSE and SCS. Green color indicates a conservation target was met in GAP 1 and 2 coverages. Blue color indicates the added contribution to meeting protection targets if GAP 3a is included with GAP 1 and 2.

Klamath-Siskiyou Ecoregion
Modified BPS (Group)	GAP ha
	(%)					Total ha
	1	2	3a	3b	4	(%)
Agricultural	147	931	15	1406	137,265	139,764
	(0.1)	(0.7)	(0.0)	(1.0)	(98.2)	(2.9)
Barren and Sparsely Vegetated	3137	646	750	3635	4316	12,486
	(25.1)	(5.2)	(6.0)	(29.1)	(34.6)	(0.3)
California Mixed-Evergreen Forest and Woodland	100,945	29,121	109,457	330,734	247,774	818,031
	(12.3)	(3.6)	(13.4)	(40.4)	(30.3)	(16.9)
Chaparral	46,633	8150	11,473	36,481	60,933	163,671
	(28.5)	(5.0)	(7.0)	(22.3)	(37.2)	(3.4)
Developed	8849	7035	4253	93,184	141,865	255,186
	(3.5)	(2.8)	(1.7)	(36.5)	(55.6)	(5.3)
Dry Douglas-fir Forest and Woodland	27	277	753	6578	10,225	17,860
	(0.2)	(1.6)	(4.2)	(36.8)	(57.3)	(0.4)
Dry Mixed-Conifer Forest and Woodland	138,566	50,137	150,153	586,799	578,633	1,504,287
	(9.2)	(3.3)	(10.0)	(39.0)	(38.5)	(31.1)
Grassland	1314	1370	1308	4491	29,053	37,535
	(3.5)	(3.7)	(3.5)	(12.0)	(77.4)	(0.8)
Mesic Mixed-Conifer Forest and Woodland	206,416	18,240	95,800	336,187	176,468	833,111
	(24.8)	(2.2)	(11.5)	(40.4)	(21.2)	(17.2)
Oak Woodland and Savanna	6311	9640	8902	33,487	105,642	163,982
	(3.8)	(5.9)	(5.4)	(20.4)	(64.4)	(3.4)
Open Water and Perennial Ice/Snow	2815	4289	163	19,700	11,452	38,419
	(7.3)	(11.2)	(0.4)	(51.3)	(29.8)	(0.8)
Other Forest, Woodland, Savanna, and Shrubland	737	189	252	2230	4871	8278
	(8.9)	(2.3)	(3.0)	(26.9)	(58.8)	(0.2)
Pacific Northwest Coastal Rainforest	849	2828	1508	37,295	131,097	173,578
	(0.5)	(1.6)	(0.9)	(21.5)	(75.5)	(3.6)
Ponderosa Pine Forest, Woodland and Savanna	3669	9716	19,722	112,983	193,346	339,436
	(1.1)	(2.9)	(5.8)	(33.3)	(57.0)	(7.0)
Red Fir Forest	70,839	1155	17,541	33,879	9424	132,839
	(53.3)	(0.9)	(13.2)	(25.5)	(7.1)	(2.8)
Riparian and Wetland	23,475	9008	15,088	59,328	73,477	180,376
	(13.0)	(5.0)	(8.4)	(32.9)	(40.7)	(3.7)
Subalpine Forest and Woodland	9474	283	1630	4462	1117	16,966
	(55.8)	(1.7)	(9.6)	(26.3)	(6.6)	(0.4)
Total ha	624,204	153,016	438,769	1,702,860	1,916,957	4,835,805
%	(12.9)	(3.2)	(9.1)	(35.2)	(39.6)	(100.0)
Siskiyou Crest Subregion
Agricultural	1	24	12	61	5473	5571
	(0.0)	(0.4)	(0.2)	(1.1)	(98.2)	(0.8)
Barren and Sparsely Vegetated	73	75	37	114	73	371
	(19.6)	(20.3)	(10.0)	(30.6)	(19.5)	(0.1)
California Mixed-Evergreen Forest and Woodland	24,617	10,144	17,561	110,245	20,590	183,157
	(13.4)	(5.5)	(9.6)	(60.2)	(11.2)	(26.2)
Chaparral	5565	231	2078	3867	2210	13,951
	(39.9)	(1.7)	(14.9)	(27.7)	(15.8)	(2.0)
Developed	636	1032	466	23,269	13,005	38,408
	(1.7)	(2.7)	(1.2)	(60.6)	(33.9)	(5.5)
Dry Douglas-fir Forest and Woodland	0	14	72	2421	3913	6419
	(0.0)	(0.2)	(1.1)	(37.7)	(61.0)	(0.9)
Dry Mixed-Conifer Forest and Woodland	35,931	4672	26,901	119,017	61,214	247,735
	(14.5)	(1.9)	(10.9)	(48.0)	(24.7)	(35.5)
Grassland	1	1	0	23	57	83
	(0.7)	(1.3)	(0.5)	(28.3)	(69.2)	(0.0)
Mesic Mixed-Conifer Forest and Woodland	6809	1192	21,619	56,548	16,055	102,224
	(6.7)	(1.2)	(21.2)	(55.3)	(15.7)	(14.6)
Oak Woodland and Savanna	1	66	10	1692	1321	3090
	(0.0)	(2.1)	(0.3)	(54.7)	(42.8)	(0.4)
Open Water and Perennial Ice/Snow	40	549	7	441	340	1377
	(2.9)	(39.9)	(0.5)	(32.0)	(24.7)	(0.2)
Other Forest, Woodland, Savanna, and Shrubland	0	16	0	39	101	156
	(0.0)	(10.3)	(0.0)	(24.8)	(64.9)	(0.0)
Pacific Northwest Coastal Rainforest	114	109	469	3023	898	4613
	(2.5)	(2.4)	(10.2)	(65.5)	(19.5)	(0.7)
Ponderosa Pine Forest, Woodland and Savanna	95	537	2852	24,185	17,080	44,749
	(0.2)	(1.2)	(6.4)	(54.0)	(38.2)	(6.4)
Red Fir Forest	3593	55	8354	11,504	1922	25,426
	(14.1)	(0.2)	(32.9)	(45.2)	(7.6)	(3.6)
Riparian and Wetland	3644	1908	2358	9271	3423	20,604
	(17.7)	(9.3)	(11.5)	(45.0)	(16.6)	(3.0)
Subalpine Forest and Woodland	172	0	220	362	67	821
	(20.9)	(0.0)	(26.8)	(44.1)	(8.1)	(0.1)
Total ha	81,291	20,626	83,017	366,079	147,742	698,755
%	(11.6)	(3.0)	(11.9)	(52.4)	(21.1)	(100.0)

and Figure S4). Mixed-conifer forest and woodland (both dry and mesic) comprise 48.3% of the KSE and 50.1% of the SCS, and California mixed-evergreen forest and woodland account for 16.9% and 26.2% of the KSE and SCS, respectively.

Of the 17 land cover categories, only four met the 30% conservation target in the KSE (Table 2). Upper elevation categories, including red fir forest and subalpine forest and woodland are more than 50% protected (Table 2) due to substantial proportions of these landscapes being within federally designated Wilderness areas across the KSE. Including protected+ would boost protection levels across most categories; however, only mesic mixed-conifer forest and woodland would be elevated enough to meet the 30% conservation target (shifting from 27% to 38.5%; Table 2).

For the SCS, only three of the 17 subgroups met the conservation target (Table 2). Protected+ would improve representation levels for most categories, elevating three categories to the 30% conservation target: red fir forest (shifting from 14.3% to 47.2%), riparian and wetland (27% to 38.4%), and subalpine forest and woodland (20.9% to 47.7%).

Notably, there are 580,507 ha of serpentine areas, where plant endemism is exceptionally high, within the KSE; however, only 24% of it is protected (see Figure S5 and Table S3). The addition of protected+ would increase representation targets to >40% (Table S3). The SCS includes 130,649 ha of serpentine with only 26% protected; protected+ would boost overall protection levels to 40% as in the KSE (Table S3).

3.1.4. Late-Successional and Old-Growth Forest (LSOG)

The KSE contains ~1.85 M ha of OGSI 80 and ~896 K ha of OGSI 200, while the SCS contains ~348 K ha of OGSI 80 and ~195 K ha of OGSI 200 (

Table 3. Old Growth Structural Index 80 and 200 persistence for the KSE and SCS, 1986–2024. Structural indices are based on USDA Forest Service mapping (Northwest Forest Plan Monitoring 2025) and persistence is based on areas that missed fires during this period.

Klamath-Siskiyou Ecoregion
Old Growth Structure Index 200 (USDA Forest Service, 2025)	GAP ha					Total ha
	(%)
	1	2	3a	3b	4
OGSI 80	284,335	54,527	209,308	823,534	480,542	1,852,246
	(15.4)	(2.9)	(11.3)	(44.5)	(25.9)
OGSI 80 Persistence	181,351	22,756	110,705	252,564	35,640	603,016
	(30.1)	(3.8)	(18.4)	(41.9)	(5.9)
OGSI 80 Promotion to 200	111,048	13,414	68,857	156,032	14,780	364,131
	(30.5)	(3.7)	(18.9)	(42.9)	(4.1)
OGSI 200	160,824	26,223	121,645	436,594	150,365	895,651
	(18.0)	(2.9)	(13.6)	(48.8)	(16.8)
OGSI 200 Persistence	90,032	11,195	55,662	129,504	10,575	296,969
	(30.3)	(3.8)	(18.7)	(43.6)	(3.6)
Siskiyou Crest Subregion
OGSI 80	50,702	10,693	48,501	191,597	46,613	348,105
	(14.6)	(3.1)	(13.9)	(55.0)	(13.4)
OGSI 80 Persistence	33,419	1987	21,270	41,841	2365	100,882
	(33.1)	(2.0)	(21.1)	(41.5)	(2.3)
OGSI 80 Promotion to 200	21,649	1171	14,458	26,973	1160	65,411
	(33.1)	(1.8)	(22.1)	(41.2)	(1.8)
OGSI 200	31,390	5490	32,534	107,857	17,902	195,173
	(16.1)	(2.8)	(16.7)	(55.3)	(9.2)
OGSI 200 Persistence	18,269	980	12,628	22,818	764	55,459
	(32.9)	(1.8)	(22.8)	(41.1)	(1.4)

, Figure 3). Notably, the SCS stands out as having a high concentration of OGSI 80 and 200, particularly in the southern portion (Figure 3). In fact, while the SCS accounts for only 14.5% of the KSE land area, it contains 18.8% and 21.8% of all OGSI 80 and 200, respectively, in the ecoregion.

We also analyzed the subset of late-seral forests that have persisted from 1986 to 2024 despite burning in at least one fire event from 1985 to 2023. The KSE and SCS respectively contain 603 K ha and ~101 K ha of persisted OGSI 80 (Table 3). And while only 18.3% of OGSI 80 in the KSE was in protected areas, these GAP 1 or 2 areas contained almost 34% of persisted OGSI 80. Similarly, 17.6% of OGSI 80 in the SCS was in protected areas while these areas contained 35.1% of persisted OGSI 80 (Table 3).

There are nearly 297 K ha and 55,459 ha of late-seral forest that were OGSI 200 in 1986 and stayed OGSI 200 in 2024 despite burning in at least one fire in the KSE and SCS, respectively (Table 3). About 21% of OGSI 200 in the KSE was found in protected areas compared to 34.1% of persisted OGSI 200 found in these areas. And nearly 19% of OGSI 200 in the SCS was in protected areas compared to 34.7% of persisted OGSI 200.

Lastly, there are 364 K ha and 65,411 ha of late-seral forest that were OGSI 80 in 1986 and developed (promoted) into OGSI 200 by 2024 despite experiencing at least one fire from 1985 to 2023 in the KSE and SCS, respectively (Table 3). These areas accounted for 40.7% of all OGSI 200 in the KSE and 33.5% in the SCS. And 34.2% of these promoted areas were in protected areas in the KSE, about 35% in protected areas in the SCS. Notably, a substantial proportion of OGSI 80, persisted OGSI 80, promoted OGSI, OGSI 200, and persisted OGSI 200 was in GAP 3a areas (protected+, ranging from 11.3 to 18.9%) in the KSE, a pattern that was even more prevalent in the SCS (13.9–22.8%; Table 3).

3.1.5. Forest Vertebrate Focal Species

The KSE contains 1.23 M ha of NSO suitable (~40%) and highly suitable (~60%) habitat while the SCS contains 291 K of suitable (39%) and highly suitable (61%) (i.e., similar proportions; see

Table 4. GAP status distribution across NSO current suitable and highly suitable habitat (2024) and suitable and highly suitable habitat that persisted from 1986 to 2024 despite burning at least once (persistence) for the KSE and SCS. Protected+ (GAP 3a) consists of IRAs and select LWCs.

Klamath-Siskiyou Ecoregion
Northern Spotted Owl Habitat (USDA Forest Service, 2025)	GAP ha					Total ha
	(%)
	1	2	3a	3b	4
NSO Suitable/Highly Suitable (Combined) Habitat (2024)	132,215	35,342	136,286	617,113	310,018	1,230,974
	(10.7)	(2.9)	(11.1)	(50.1)	(25.2)
Suitable	57,247	13,437	47,674	215,831	153,309	487,498
	(11.7)	(2.8)	(9.8)	(44.3)	(31.5)
Highly Suitable	74,968	21,905	88,612	401,282	156,710	743,476
	(10.1)	(3.0)	(11.9)	(54.0)	(21.1)
NSO Suitable/Highly Suitable Habitat Persistence	77,441	11,690	63,581	142,813	12,684	308,208
	(25.1)	(3.8)	(20.6)	(46.3)	(4.1)
Suitable	29,694	4003	20,489	44,669	5257	104,112
	(28.5)	(3.9)	(19.7)	(42.9)	(5.1)
Highly Suitable	47,747	7686	43,092	98,144	7427	204,096
	(23.4)	(3.8)	(21.1)	(48.1)	(3.6)
Siskiyou Crest Subregion
NSO Suitable/Highly Suitable (Combined) Habitat (2024)	32,259	11,486	31,627	176,107	40,028	291,507
	(11.1)	(3.9)	(10.9)	(60.4)	(13.7)
Suitable	16,004	3767	12,750	61,728	19,355	113,603
	(14.1)	(3.3)	(11.2)	(54.3)	(17.0)
Highly Suitable	16,255	7720	18,878	114,379	20,673	177,904
	(9.1)	(4.3)	(10.6)	(64.3)	(11.6)
NSO Suitable/Highly Suitable Habitat Persistence	17,687	1798	12,523	27,315	1055	60,378
	(29.3)	(3.0)	(20.7)	(45.2)	(1.8)
Suitable	8123	459	4375	8983	510	22,449
	(36.2)	(2.0)	(19.5)	(40.0)	(2.3)
Highly Suitable	9564	1339	8148	18,332	545	37,929
	(25.2)	(3.5)	(21.5)	(48.3)	(1.4)

and Figure S6). The KSE also contains 308 K ha of NSO habitat that persisted (as either suitable or highly suitable) from 1986 to 2024 despite burning at least once between 1985 and 2023, with 28.9% of this persisted habitat in protected areas and nearly 50% within protected+ (Table 4). Another 60 K ha of NSO habitat persistence occurs in the SCS with most (62%) in the highly suitable layer and just over 30% of that protected. Protected+ would boost levels to near the intermediate target for the KSE and slightly above this target for the SCS.

For the Pacific fisher, there were >1.6 M ha and >380 K ha of connectivity habitat within the KSE and SCS, respectively, with comparably low protection levels in both areas (Table S4, Figure 4). Notably, the SCS stands out as having important fisher habitat connectivity (Figure 4). In both areas, protected+ would still leave fisher with very low (<20%) levels of protected connectivity habitat (Table S4). Thus, a substantial portion of the land-bridge and focal species representing its connectivity importance remains outside protected areas.

3.1.6. Fire Severity

From 1985 to 2023, slightly more than 2 M ha (42%) of the 4.83 M ha KSE and nearly 268 K ha (38%) of the 697 K ha SCS experienced at least one wildfire (

Table 5. Fire severity (1985–2023) across all land cover categories for the KSE and SCS in relation to GAP categories. ¹ The percentages of each fire severity class for each GAP status are the percentage of a particular GAP status that experienced a particular fire severity (the sum of percentages for each GAP status column will equal 100%). ² The percentages in the last column are the percentage of the total fire footprint in each severity class, irrespective of GAP status. ³ The percentages in the last row for the ecoregion and subregion represent the percentage of the total fire footprint across the study period that was in a particular GAP status.

Klamath-Siskiyou Ecoregion
Fire Severity	GAP ha					Total ha (%) 2
	(%) 1
	1	2	3a	3b	4
Unprocessed	1111	1540	1045	530	315	4541
	(0.2)	(1.5)	(0.3)	(0.1)	(0.1)	(0.2)
Unchanged	41,330	7188	20,555	62,254	30,599	161,925
	(8.3)	(7.1)	(6.3)	(7.7)	(10.3)	(8.0)
Low	142,740	25,006	87,761	230,354	69,727	555,587
	(28.6)	(24.8)	(26.9)	(28.4)	(23.4)	(27.3)
Moderate	137,790	26,728	93,729	215,764	84,758	558,769
	(27.6)	(26.6)	(28.7)	(26.6)	(28.4)	(27.5)
High	176,156	40,201	123,244	300,958	112,756	753,314
	(35.3)	(39.9)	(37.8)	(37.2)	(37.8)	(37.0)
Total ha	499,127	100,661	326,333	809,860	298,154	2,034,136
(%) 3	(24.5)	(4.9)	(16.0)	(39.8)	(14.7)
Siskiyou Crest Subregion
Unprocessed	56	10	854	145	41	1106
	(0.1)	(0.2)	(1.7)	(0.1)	(0.2)	(0.4)
Unchanged	8148	1040	3507	11,616	2416	26,726
	(13.0)	(16.9)	(7.0)	(9.2)	(10.9)	(10.0)
Low	21,155	2495	15,056	34,556	4808	78,071
	(33.8)	(40.5)	(29.9)	(27.3)	(21.8)	(29.2)
Moderate	16,008	1401	14,424	27,353	5769	64,955
	(25.6)	(22.7)	(28.6)	(21.6)	(26.1)	(24.3)
High	17,170	1214	16,524	52,889	9067	96,865
	(27.5)	(19.7)	(32.8)	(41.8)	(41.0)	(36.2)
Total ha	62,538	6159	50,365	126,559	22,101	267,722
(%) 3	(23.4)	(2.3)	(18.8)	(47.3)	(8.3)

), with 680 K ha of the KSE experiencing up to five fires (Table S5). During this period, 78,104 ha (1.62%) burned on average each year. More than 5% (nearly 242 K ha) of the KSE burned in 1987, 2008, 2018, 2020, and 2021 (Figure S7). The 2020 fire season resulted in ~530 K ha burned (almost 11% of the KSE), substantially more than in the next largest fire year (2018, nearly 341 K ha). Additionally, 9342 ha (1.34%) burned on average each year in the SCS, which also saw burning rates greater than 5% in 1987, 2008, 2017, 2020, and 2023 (Figure S7). For most years (26 of 36), the burn rate was greater at the KSE scale than in the SCS. Notable years where the reverse was true include 1994, 2017, and 2023 (Figure S7).

There was a substantial amount of high-severity fire across both the KSE (37%) and SCS (36.2%), with high-severity fire rates similar in GAP 3a, 3b, and 4 across the KSE (37.2–37.8%) (Table 5). High-severity fire rates were 35.3% in GAP 1 and 39.9% in GAP 2, though the latter also represents the smallest amount of land area in general and therefore represented a much smaller proportion of the total burned area (4.9%). The area-weighted mean of high-severity fire in GAP 1, 2, and 3a was 36.7%, only slightly lower than the area-weighted mean in GAP 3b and 4 (37.3%), indicating that fire severity was not particularly influenced by GAP categories at the ecoregional scale. While the high-severity fire rate across the SCS was similar (36.2%), there were notable differences among GAP status. High-severity fire rates in GAP 3b and 4 ranged from 41 to 41.8% (area-weighted mean of 41.7%) while GAP 1, 2, and 3a lands had substantially lower rates (19.7–32.8%) of high-severity fire (Table 5), with an area-weighted mean of 29.3%.

When we analyzed fire severity across the more-frequent-fire, drier-forest types at the ecoregion scale (i.e., dry and mesic mixed-conifer, dry Douglas-fir, and ponderosa pine), we found a similar pattern. About 38.3% and 38.9% of these forests in the KSE and SCS, respectively, burned at high severity (

Table 6. Fire severity (1985–2023) across dry and mesic mixed-conifer forest and woodland, dry Douglas-fir forest and woodland, and ponderosa pine forest, woodland and savanna land cover categories for the KSE and SCS in relation to GAP categories. ¹ The percentages of each fire severity class for each GAP status are the percentage of a particular GAP status that experienced a particular fire severity (the sum of percentages for each GAP status column will equal 100%). ² The percentages in the last column are the percentage of the total fire footprint in each severity class, irrespective of GAP status. ³ The percentages in the last row for the ecoregion and subregion represent the percentage of the total fire footprint across the study period that was in a particular GAP status.

Klamath-Siskiyou Ecoregion
Fire Severity	GAP ha					Total ha (%) 2
	(%) 1
	1	2	3a	3b	4
Unprocessed	401	54	171	266	159	1051
	(0.1)	(0.1)	(0.1)	(0.0)	(0.1)	(0.1)
Unchanged	20,752	3264	12,539	36,136	16,555	89,245
	(6.9)	(5.6)	(6.0)	(6.7)	(8.4)	(6.8)
Low	84,342	14,934	52,887	148,118	44,693	344,974
	(28.1)	(25.8)	(25.4)	(27.3)	(22.6)	(26.4)
Moderate	87,472	15,792	61,317	149,053	58,105	371,740
	(29.1)	(27.3)	(29.4)	(27.5)	(29.3)	(28.4)
High	107,353	23,881	81,480	208,685	78,513	499,911
	(35.7)	(41.2)	(39.1)	(38.5)	(39.6)	(38.3)
Total ha	300,320	57,924	208,393	542,258	198,025	1,306,921
(%) 3	(23.0)	(4.4)	(15.9)	(41.5)	(15.2)
Siskiyou Crest Subregion
Unprocessed	17	2	85	57	4	1106
	(0.1)	(0.1)	(0.3)	(0.1)	(0.0)	(0.7)
Unchanged	3447	186	2398	5880	1694	26,726
	(10.3)	(10.2)	(7.4)	(8.4)	(10.2)	(17.3)
Low	9906	631	9385	17,680	3495	78,071
	(29.5)	(34.7)	(28.9)	(25.1)	(21.1)	(50.4)
Moderate	9483	447	9185	16,034	4553	64,955
	(28.2)	(24.5)	(28.3)	(22.8)	(27.5)	(42.0)
High	10,735	555	11,378	30,746	6812	96,865
	(32.0)	(30.5)	(35.1)	(43.7)	(41.1)	(62.6)
Total ha	33,590	1819	32,430	70,398	16,557	154,795
(%) 3	(21.7)	(1.2)	(21.0)	(45.5)	(10.7)

). The area-weighted mean fire severity rate was 37.5% across GAP 1, 2, and 3a compared to 38.8% across GAP 3b and 4 at the KSE scale. However, at the subregion scale, an area-weighted mean of 33.4% of dry forests in GAP 1, 2, and 3a represents lower levels of high severity compared to 43.2% in GAP 3b and 4. Notably, these differences generally held across elevation bins, especially in the SCS (Table S6).

We also analyzed a subset of two of the largest fires during the study period in the KSE and SCS, respectively, in relation to GAP status (Figure 5). The August Complex fire of 2020 affected 381,261 ha of the KSE (7.9% of the ecoregion in this one event) and the Slater fire of 2020 affected 61,746 ha of the SCS (8.8%), with the highest amounts of high-severity fire in GAP 3b for both fires (Figure 5).

3.2. Fuel Treatments and the Built Environment

Nearly all fires in the KSE and SCS were outside the WUI boundary (Figure S8). Additionally, most fuel treatments of federal agencies (Forest Service and Bureau of Land Management combined) were >1000 m from the WUI boundary (Figure 6, Figure S9), a pattern observed at both spatial scales. Interestingly, 57.4% and 74.2% of the combined agency treatment footprint was >1000 m from the WUI boundary across the KSE and SCS, respectively, but was considered within the WUI by either agency according to agency activities database. The Bureau of Land Management tended to classify treatment areas beyond 1000 m from the WUI as within the WUI more than the USDA Forest Service across the KSE (86.5% vs. 52.3%) and the SCS (100% vs. 69.2%).

3.3. Climate Change

Future climate impacts in the KSE depend on atmospheric greenhouse gas concentrations both now and in the future. Up until the 2040–2069 period, however, the climate impacts in the KSE are relatively similar in the intermediate and very high climate scenarios (Table S7). Growing degree-days are projected to steadily increase throughout the next decade under an intermediate greenhouse gas scenario (SSP245), with most increases in areas to the north and southeast of the SCS. As expected, the very high scenario shows much greater increases in growing degree-days at the end of the century (2070–2099) than the intermediate scenario, with nearly twice the increase as the intermediate scenario for the same period (Figure 7, Table S7).

Summer mean maximum temperature for the KSE shows similar increases to growing-degree days with projected increases throughout the ecoregion and small areas in the northwest and southwest warming slightly less (Figure 8). Consistent with the increase in growing degree days, the intermediate and very high climate scenarios diverge in their projected warming in the 2040–2069 period, with the very high climate scenario showing much more pronounced warming at the end of the 21st century (5.6 °C compared to 3.3 °C, Table S7). Increased heat can influence more plant growth and yet can also indicate the potential for heat stress and reduced water availability of the potential SCS refugium, especially with such large increases.

Some areas in the SCS are cooler historically than the ecoregion writ large, due in part to relatively high elevation (Figure 8 and Figure S11; also see Figure 1 for elevation). This may indicate some refugium conditions are available for longer periods than the rest of the ecoregion in both intermediate and very high greenhouse gas scenarios. However, the higher elevation locations and their relatively cooler conditions are still projected to have large increases in mean summer maximum temperature at the end of this century, especially under the very high greenhouse gas.

Total summer precipitation is projected to decline for the KSE for all time periods and future climate scenarios, excluding the very high scenario (Figure 9). However, the amount of decrease for each scenario differs. The intermediate climate scenario suggests a larger decrease in total summer precipitation for the near-term period (2010–2039) and long-term (2070–2099), and the very high scenario projects a larger decrease in the mid-term (2040–2069). Though there is large variation in magnitude of the projected change, most models show decreases in total summer precipitation for the KSE, meaning less water will be available during this critical seasonal period when temperatures are highest (i.e., drought stress to vegetation is particularly prevalent in hot summer months in this ecoregion, pers. observations). Notably, the SCS, particularly the eastern edge, appears to maintain some summer precipitation potential through the different time periods and both emission scenarios, reflecting longer “hang-times” for refugia persistence (Figure 9 and Figure S12). Large uncertainties exist for precipitation projections, and especially for summer precipitation, due to model limitations in resolving clouds. This is reflected in the wide range of model results for the SCS summaries in Table S7.

The historical and future projected change in the April 1st snow–water equivalent (SWE) for the KSE shows decline in the region’s snowpack (Figure 10). Within the SCS, the historical climate model simulation shows that the higher elevations have historically had >300-mm of SWE on April 1st, which is able to supply the local refugium with water into the spring and early summer. As temperatures increase, this snowpack is projected to decrease until there are only a few very localized areas in the SCS with April 1st snowpack for the 2070–2099 period. The shift to less snowpack is primarily due to more winter precipitation falling as rain instead of snow due to warming temperatures and thus more surface water runoff into streams during the winter instead of being stored in snowpack. Because of less spring snowpack, water availability later in the spring and summer seasons when most needed is expected to decline, likely stressing aquatic organisms and especially those requiring cold water (e.g., salmonids and stream amphibians).

Increasing temperatures, decreasing summer precipitation, and less water available via snowmelt all contribute to the propagation of wildfires across much of the KSE (Figure 11). Historically, the SCS had less favorable wildfire conditions (i.e., longer fire-return intervals) than lower elevations in the KSE. Under both future climate scenarios, the SCS is projected to have conditions that are more favorable for wildfires with future warming, though less so than much of the ecoregion. Interestingly, towards the end of the 21st century, the westernmost areas of the KSE (nearest the coast) show decreasing likelihood of favorable wildfire conditions, possibly due to a reduction in vegetation available to burn.

4. Discussion

4.1. Ecoregional Comparisons and Representation Analyses

The Klamath-Siskiyou Ecoregion has long received scientific attention dating back to pioneering botanists of the late 1800s, for which many plant species bear their names, and the seminal work of early ecologists that elevated the global status of the ecoregion [8,9]. As such, the KSE is considered a Proposed Area of Global Botanical Significance, Proposed World Heritage Site and International Biosphere Reserve, and Global Centre of Plant Diversity [10]. Despite this recognition and decades of conservation planning (e.g., [1]), the KSE and SCS have progressed little in meeting conservation targets that are well below the lowest targets for most representation analyses. Meanwhile, pressures are mounting from climate change and extensive land uses that have already degraded substantial portions of the ecoregion [10].

A comparison of ECAs conducted in four ecoregions we have investigated thus far (see [4,5,6]) shows common patterns among areas, including each ecoregion and its associated subregion were well-below even the lowest conservation target, ranging from 0.25% to 18.4% protected (

Table 7. Ecoregional conservation assessments for the Klamath-Siskiyou/Siskiyou Crest (this study), Northern Rockies/Yaak Valley [6], Southern Rockies/Santa Fe [5], and Mogollon Highlands [4]. Note: the Mogollon Highlands did not include a specific subregional analysis.

Location	Area (M ha)	% Protected
Klamath-Siskiyou Ecoregion	4.83	15
Siskiyou Crest Subregion of KSE	0.68	14.5
Northern Rockies Ecoregion (NRE)	8.19	2.2
Yaak Valley Subregion of NRE	0.16	0.25
Southern Rockies Ecoregion (SRE)	14.5	18.4
Santa Fe Watershed Subregion of SRE	2.2	12.1
Mogollon Highlands Ecoregion	11.3	9

). Some additional common patterns include priority areas that met conservation targets were largely at upper elevations, important as high-elevation potential refugia, but missing most other priorities at lower elevations (e.g., BPS subgroups). For instance, there is poor representation of low–mid elevation forested areas and habitat for focal species reflective of older forests and landscape connectivity for all ecoregions. Additionally, across ecoregions a common conservation strategy is to bolster protections for IRAs, Lands with Wilderness Characteristics, and Wilderness Study Areas (protected+) which, if fully protected, can help close gaps in protected area coverages as proposed by local conservation groups. In our prior ECAs, IRAs were considered GAP 2.5 to reflect an intermediate protection level from most forms of logging but were not protected enough to be “permanent” and inviolate (GAP 1 or 2). However, the USDA Forest Service is in the process of rescinding the national conservation roadless rule on ~18 M ha (https://www.federalregister.gov/documents/2025/08/29/2025-16581/special-areas-roadless-area-conservation-national-forest-system-lands; accessed on 18 June 2026), including within the KSE, placing these potential refugia in jeopardy from mining, logging, and road building.

More specific to the KSE is the importance of the BPS representation analysis because of the variety of plant communities present within those types that did not meet any conservation target. This is particularly the case for serpentinites where endemic plants are highly concentrated [73]. We suspect that the low protection for serpentine areas is due to mining interests in nickel deposits in some areas, along with historic gold mining along creeks [10].

There are six HUC8 watersheds in the KSE and one in the SCS that met the low or intermediate conservation targets. Boosting protections for IRAs and select LWCs (protect +) would improve representation and add a few more watersheds to the lowest targets. Importantly, the recent removal of four mainstem dams on the Klamath River, the nation’s largest dam removal to date, is having restorative properties throughout the watershed even with the low levels of protection, thereby improving potential refugia for aquatic species and restoring reciprocal relations with regional tribes that include ceremonial practices and the return of anadromous fish [74].

4.2. Focal Forest Vertebrate Species

The KSE and SCS contain similar proportions of NSO habitat poorly protected and an important subset of NSO habitat that has persisted over multiple fires during the multi-decadal period of analysis. Notably, protected areas containing late-seral forests generally and NSO habitat specifically have fared better overall in fire persistence over this time frame, indicating that these areas may provide refugia for NSO and old-forest species as shown by other researchers [23,40]. With the increasing number of growing-degree days and summer temperatures documented herein, refugia are more critical for NSO given that their physiological and behavioral responses to heat stress are greatest when temperatures exceed their thermoneutral envelope [75,76]. Due to interference competition from Barred Owls (Strix varia), NSO is forced to forage farther from streams and nest sites and on hotter slopes [77], further illustrating the importance of potential refugia for persistence of NSOs in the presence of Barred Owls.

Federal agencies also have focused on reducing wildfire severity in NSO habitat areas through intensive thinning operations [78]. However, the degree to which NSO are negatively impacted by severe fire is equivocated by nearly all known NSO territories have experienced multiple logging entries before and after severe fires [79]. NSO also are known to use severe burn patches for foraging that can positively affect fecundity rates [52], and in fire-adapted systems of northern California the species selects territories with a mixture of burned and unburned patches [80]. NSO even have been documented nesting successfully in severely burned patches [81], indicating that these burned patches provide necessary habitat requirements. Our study supports the refugia concept for NSO regarding LSOG and habitat persistence levels; however, it is not known if a tipping point/threshold exists regarding high-severity fire at NSO territory or larger scales, a factor that is concerning given the climate-related projections for increased wildfire activity.

For the Pacific fisher, the KSE and SCS contain important connectivity habitat but at very low (<10%) levels of protection. Notably, however, the SCS may function as fisher refugia due to high proportions of connectivity habitat. Like NSO, managers have been focused on reducing severe fires via fuel reduction projects. However, this focal species also is known to persist in moderate- to high-severity burn patches within the southern Sierra Nevada mountains of California, USA, and is adversely impacted by logging [82]. Like NSO, the projected increases in wildfire activity could reduce connectivity habitat and refugia over time; however, it is not known whether there is a threshold in wildfire activity that would impact this focal species.

Although not a focal taxon in our study due to few studies in the KSE, amphibians may be vulnerable to projected increases in wildfires. For instance, evidence in the KSE suggests wildfire impacts to amphibians occur one year post-fire but the decrease in stream amphibians appears temporary [83]. However, studies nearby in the Oregon Cascade Mountains indicate little or no decrease in amphibians post-fire in stream-dependent [84] or terrestrial [85] taxa. Monitoring of wildfire- and land-use-sensitive species is essential in developing appropriate conservation strategies to avoid projected extirpation rates that may be as high as 10% of all taxa in the KSE in the coming decades [18].

4.3. Climate Change and Refugia Properties

Our climate findings show how the degree of climatic change for the KSE depends on current and future greenhouse gas emissions, particularly toward the end of the 21st century as the intermediate and very high scenarios diverge more. In the near term (through ~2050), climate impacts to both the KSE and SCS are similar between emissions scenarios with projected increases for the KSE in warming (+1.5 °C and +1.8 °C for the intermediate and very high scenario, respectively), declines in summer precipitation (−21% and −19%), and declines in the April 1st SWE (also see [22]). Such conditions will impact water delivery to streams and agriculture, increase tree stress from drought, insect and pathogen infestations, and create conditions for wildfire propagation. Notably, our analysis suggests that due to the higher elevation position of the SCS it may maintain refugium properties for longer periods, providing some “hang time” for species to adapt to the changes. However, those properties are predicted to break down toward the end of the century and under the highest emissions scenario. As such, the degree to which this ecoregion, especially the higher elevation subregion, maintains refugia properties is highly dependent upon the combination of global emissions and avoidance of land-use stressors within old forests and intact watersheds. Such findings likewise have been noted in each of the other ecoregions examined [4,5,6].

Additionally, many older forests in this fire-prone ecoregion have escaped recurring severe fires by chance or because of their cooler/wetter microclimates acting as long-standing refugia (persistence) for focal species like the Northern Spotted Owl and Pacific fisher. Thus, it is not just the fire refugia that are important to focal species but the distribution of patches of varying burn severities that provide a mix of nesting and foraging opportunities (e.g., [80]). Nonetheless, the identification of older forests as refugia is generally supported by other studies documenting the superior climate [24] and fire refugia properties of these forests [40]. Follow-up studies focused on microclimatic properties of older forests (e.g., as in [24]) and the Crest in general are warranted to tease out refugia properties.

For both the ecoregion and the subregion, older forests, IRAs, LWCs, and protected watersheds collectively can serve as refugia, as also presumed by others [18,23,24,40,86,87]. Interestingly, our findings show that land-use status (GAP status) had no apparent effect at the ecoregional scale on wildfire severity and hence refugia properties despite claims in agency forest planning documents and proposed wildfire legislation that active management is needed to reduce wildfire severity in most land uses. However, at the SCS scale, protected areas and protected+ had substantially lower levels of wildfire severity, which was reflected also in the largest recent fire that had higher severity levels in “managed areas” (GAP3b). We were unable to tease out specific factors involved in this relationship; however, we note that protected areas in general were biased toward upper elevations where fires are expected to be more severe and on longer return intervals. Our findings for the SCS showed the opposite (lower levels of high severity) and point to the differences in land-use status as a contributing factor in whether refugia are maintained (protected areas) or degraded (land uses) at this scale and especially for dry forest areas. Further, in Oregon, high-severity fire is influenced by a combination of extreme fire weather interacting with heavily logged areas [27], with most large fires spilling over into the built environment originating on private lands where logging is greatest [44].

In general, the SCS exhibits refugium properties due to its upper elevation distribution, high proportion of late-seral and late-seral persistence forests, land-bridge features, and lower levels of high-severity fire in protected areas. Climate downscaling showed it is likely to retain moister, cooler conditions for longer periods. Follow-up research is needed to determine site-specific factors that might have micro-refugia properties within the SCS and the KSE in general [18]. Additionally, like the KSE, the SCS did not meet most of the conservation targets and is under pressure mainly from logging, mining, road building, and other uses (e.g., livestock grazing, off-highway vehicles; see [10]). Notably, land-use pressures are known to amplify climate-related impacts [27], speeding up the degradation of refugia properties without stepped up protections.

4.4. Fuel Treatments and the Built Environment

Our findings show that most fuel treatments by the two federal agencies with the most land in the KSE (BLM and USFS) are far removed from the built environment as was the case for the other ecoregions investigated (

Table 8. Results of fuel treatment distance to WUI analyses in ecoregional conservation assessments for the Klamath-Siskiyou/Siskiyou Crest (this study), Northern Rockies/Yaak Valley [6], and Southern Rockies/Santa Fe [5].

Location	Fuel Treatment Distance to WUI (m, % values)
	0–250	251–500	501–750	751–1000	>1000
Klamath-Siskiyou Ecoregion	10.9	6.6	5.6	4.9	72.0
Siskiyou Crest Subregion of KSE	14.5	9.0	7.3	5.7	63.5
Northern Rockies Ecoregion (NRE)	5.7	4.1	3.9	3.7	82.6
Yaak Valley Subregion of NRE	8.5	10.1	8.6	7.4	65.5
Southern Rockies Ecoregion (SRE)	8.1	6.9	5.8	5.7	73.5
Santa Fe Watershed Subregion of SRE	4.4	5.2	4.9	5.7	79.7

). One limitation of this comparison is that the categories of fuel treatments differed in one of our previous ECAs. DellaSala et al. [5] (Southern Rockies) only included USDA Forest Service activities categorized by the agency as some type of thinning (from the hazardous fuel treatment dataset as well as from the timber harvest and timber stand improvement datasets), whereas in the current study we included only activities from the hazardous fuel treatment dataset that included additional fuel treatments such as mastication and prescribed fire. The method we used in the current study was more like that of DellaSala et al. [6] (Northern Rockies). Regardless, our results here and in other ecoregions support calls for a more strategic focus on community protection within a narrow defensible space area around structures themselves [88,89] along with a redirect of wildfire spending to home hardening (e.g., H.R 582 (119th Congress)—Community Protection and Wildfire Resilience Act; https://www.congress.gov/bill/119th-congress/house-bill/582/text; accessed on 18 June 2026). Moreover, human-caused fire ignitions are a major factor that make up a large portion of wildfires nationwide [45], especially in areas with high road densities [46], and road density reduction and forest use along roads need attention in agency planning.

5. Data Limitations and Uncertainties

Our study provides a conservation reserve design at two spatial scales: the KSE and the SCS. However, most published taxa datasets were not available at these scales. Therefore, we had to resort to using habitat suitability models that met the criteria of the study design in selecting the two focal species for our analysis (Pacific fisher, Northern Spotted Owl). Importantly, spotted owl habitat suitability models do not reflect actual occupancy or owl demographics and the availability of suitable habitat may not reflect spotted owl occupancy due to competition with Barred Owls. Fuel treatments may also interact with Barred Owl encroachment to limit spotted owl occupancy even in high-suitability spotted owl habitat [79]. Demographic monitoring is needed to validate refugium functions.

We could not determine if the refugia properties of the SCS are due to its cooler/moister upper-elevation microclimate or because it has greater climatic stability in general. The latter case makes it refugium for existing cool-adapted species, whereas the first case makes it refugium for specific species from warmer/drier environments. Follow-up analysis of site-specific attributes would be helpful in deciphering refugium mechanisms tied to the SCS and potentially other refugia within the KSE [18]. Specifically, a site-specific study of microclimate conditions (ambient and soil temperatures, humidity levels, wind speeds) in potential refugia (e.g., old forests e.g., see [24]) compared to surroundings (e.g., clearcuts) would provide further refinement of the refugia concept.

Our study determined that at the scale of the ecoregion there was no apparent effect of land-use status (GAP status) on fire severity but that at the subregion scale, protected and protected+ areas had lower levels of high-severity fire compared to “managed areas” (GAP status 3 and 4) and the pattern was apparent even in drier forest types, which has been noted by other researchers and warrants further investigation to determine the causal factors involved [27,90].

Climate model projections include uncertainty from two main sources: natural variability and model bias. We limited uncertainty from natural variability by averaging the climate projections over 30-year periods. Averaging over 30 years of data isolates the climate signal by averaging out noise from random natural variations. Without focused analysis on how individual downscaled climate model datasets perform for the SCS and KSE compared to observations, it is impossible to know which model is the least biased. Thus, we chose to average the climate indicator results over all available climate model simulations in each dataset, as studies have suggested that using an ensemble mean or median from climate model data is often a better estimate than any individual model [91].

Regardless of our efforts to reduce uncertainties in the climate model projections, each climate model projects different future conditions under the same greenhouse gas scenario as bias cannot be eliminated. We have included the range of model results for growing degree days, summer precipitation, and summer maximum temperature in the Supplemental Materials as an estimate of uncertainty for our results (Table S7). An additional uncertainty with climate projections is what future greenhouse gas concentrations to choose. Though we can argue for the likelihood of scenarios, we cannot know the trajectory of greenhouse gas emissions with absolute certainty. We chose to display two greenhouse gas scenarios to demonstrate a range of possible futures: an intermediate scenario (RCP 4.5, SSP 245) and a very high scenario (RCP 8.5, SSP 585).

6. Conclusions

The KSE is a globally significant area of biodiversity that is under increased land-use and climate stressors, while progressing minimally in meeting long-standing conservation targets over the past two decades [1,10,18]. The ecoregion is now facing increased logging levels proposed to reduce wildfires (e.g., The “Fix Our Forests Act” H.R. 471; https://www.congress.gov/bill/119th-congress/house-bill/471; accessed on 30 June 2026) and elevated domestic timber targets on federal lands. There have been calls for expanding treatments in wildlands to reduce fire intensity in frequent-fire forests that have a history of fire suppression (e.g., [92]). However, there are significant financial and ecological costs associated with wildland treatments that typically remove large, commercially valuable trees and increase road densities for access [93,94]. Wildland treatments away from structures also are inconsistent with calls by researchers to redirect community wildlife risk reduction to the structure itself [88,89].

Our findings also point to the key potential refugia properties of IRAs, old forests and select LWCs that tend to burn at lower severities (as in [23,40]), and intact watersheds for focal species connectivity (e.g., Pacific fisher). If protected, these areas offer the best hope for extending the “hang time” for plants and wildlife to adapt while providing invaluable ecosystem services (e.g., clean water, clean air, recreation, fire refugia) to communities. Restorative efforts that re-wild areas can slow degradation losses especially those broadly supported by communities such as the dam removals along the mainstem Klamath and its bio-cultural restorative powers for tribes [74]. Demonstrating the compatibility of reserve design, refugia protection, effective wildfire risk reduction for communities, and restoration as a means for helping both nature and people in a world where nature is under increasing pressures is needed locally and to comply with global targets such as those in the Montreal-Kunming Biodiversity Framework. The KSE can contribute to global targets through increased local conservation that effectively moves the ecoregion and subregion into the upper protection levels while restoring degraded portions of the ecoregion. Our results span multiple ecoregions facing mounting pressures that overall would benefit from increased protection of locally proposed protected areas and refugia and a greater emphasis on strategic risk reduction of the built environment to build public support for biodiversity conservation.