Evaluating ecosystem protection and fragmentation of the world's major mountain regions

Conserving mountains is important for protecting biodiversity because they have high beta diversity and endemicity, facilitate species movement, and provide numerous ecosystem benefits for people. Mountains are often thought to have lower levels of human modification and contain more protected area than surrounding lowlands. To examine this, we compared biogeographic attributes of the largest, contiguous, mountainous region on each continent. In each region, we generated detailed ecosystems based on Köppen−Geiger climate regions, ecoregions, and detailed landforms. We quantified anthropogenic fragmentation of these ecosystems based on human modification classes of large wild areas, shared lands, and cities and farms. Human modification for half the mountainous regions approached the global average, and fragmentation reduced the ecological integrity of mountain ecosystems up to 40%. Only one‐third of the major mountainous regions currently meet the Kunming‐Montreal Global Biodiversity Framework target of 30% coverage for all protected areas; furthermore, the vast majority of ecosystem types present in mountains were underrepresented in protected areas. By measuring ecological integrity and human‐caused fragmentation with a detailed representation of mountain ecosystems, our approach facilitates tracking progress toward achieving conservation goals and better informs mountain conservation.

. Although mountains are considered to have lower levels of human development than other landscapes (Weber, 2004), they are nevertheless home to 1−2 billion people (Körner et al., 2017;Thornton et al., 2022) and provide freshwater to billions more (Viviroli et al., 2020).Mountains are not disproportionately protected globally; nearly 40% lack any form of protection and fall short of internationally recognized protection targets (Elsen et al., 2018).They also face pronounced pressures from increasing human developments and activities (Thornton et al., 2022) that disrupt species movements and connectivity among habitats.Increasing protection and connectivity across mountains is critical to ensure species can adapt as climates change (Parks et al., 2023).
Establishing protected areas (PAs) that better achieve ecosystem representation in mountains is needed to meet global conservation goals (e.g., Sustainable Development Goal 15.4), but it remains challenging in practice.For example, reporting on biodiversity in mountain ecosystems requires subnational data and analyses to inform policy-making, planning, and conservation at scales relevant to mountain ecosystems (Ly et al., 2023).Recent work by the International Union for the Conservation of Nature (IUCN) Mountain Ecosystem Specialist Group has provided initial guidance to inform the placement of new mountain PAs (Jacobs et al., 2021).Yet, efforts to conserve mountains are complicated by their transboundary nature and the diverse political, social, and cultural systems they typically span (Baldwin et al., 2018;Lemieux et al., 2022).As a result, large-landscape conservation managers need detailed, management-relevant information to track progress toward protecting biodiversity and nature's contributions to people (Hebblewhite et al., 2022).
Mountain ecosystems are particularly sensitive to human modification (HM) that increases habitat fragmentation, soil erosion, and alters natural processes (Chester et al., 2013).As ecosystems and species shift their distributions to track changing climates, sharp topographic constraints and bottlenecks from human-induced habitat fragmentation will decrease species' abilities to persist and adapt.High physiographic and habitat diversity in mountains also amplifies the inadequacy of spatial prioritization and planning for use with relatively coarse analytical units, like ecoregions (e.g., Dinerstein et al., 2017), because they fail to capture critical features, such as landforms characteristic of mountains (Sultaire et al., 2022).
We assessed current levels of HM, ecosystem fragmentation, and PA coverage in mountains.We selected a prominent, contiguous mountainous region representative of each continent (excluding Antarctica), which reflects the broad landscape extents needed to retain ecological processes, such as migration.We developed a fine-grained, thematically detailed classification of mountain ecosystems based on climate and abiotic factors.We then quantified the degree of human-caused modifications and fragmentation-both critical to measuring the structural element of ecological integrity.Finally, we assessed the proportion of mountain regions protected to conserve biodiversity relative to the target 3 goal of protecting 30% of the landscape by 2030 (i.e., Kunming-Montreal Global Biodiversity Framework [GBF]) (United Nations, 2022).
We addressed the following questions: How much HM occurs across each mountainous region, and how does this level compare to the global average, how fragmented are ecosystems due to HM, how much of each mountainous region is protected relative to the target of 30% protection by 2030, and do PAs represent the diversity of ecosystems in the mountainous regions?

METHODS
The mountainous regions we examined were Albertine Rift (Africa), Alps (Europe), Andes (South America), Great Eastern Ranges (Australia), Himalayas (Asia), and the Yellowstone to Yukon (Y2Y) area (North America) (Figure 1 & Appendix S1).We delineated mountain-range boundaries based on study areas defined by conservation organizations for the Albertine Rift (Plumptre et al., 2016) and Y2Y regions (Databasin, 2023).For the remaining regions, we selected and merged mountain units defined in the Global Mountain Biodiversity Assessment (GMBA) inventory database 1.2 (Körner et al., 2017) (Appendix S2).We used data from the GMBA 1.2 rather than 2 (Snethlage et al., 2022a(Snethlage et al., , 2022b) because 1.2 units are more inclusive of valley bottoms and, therefore, better approximate a landscapelevel context and are generally more consistent with the Y2Y and Albertine Rift delineations (Appendix S3).We also examined how our results varied when incorporating lands within 10 and 30 km of each region (Table 2 & Appendix S4).We contextualized our results with particular reference to the Y2Y because this area has been the focus of one of the betterknown mountainous conservation efforts (Hebblewhite et al., 2022).
We used the degree of HM data set (Theobald et al., 2020) to quantify anthropogenic pressures on each mountainous region in our study because it provides globally consistent, comprehensive, and recent (circa 2017) data.The HM values (H) range from 0.0 (no modification) to 1.0 (complete modification).We categorized HM into 3 classes consistent with the "3 conditions" framework (Locke et al., 2019): large wild areas (H ≤ 0.1), shared lands (0.1< H ≤ 0.4), and cities and farms (0.4< H ≤ 1.0) (hereafter wild, shared, and developed, respectively).To enable adequate resolution to calculate fragmentation in complex topographic environments, we reran the model of Theobald et al. (2023) to produce a 90-m data set of H values.
To more adequately reflect the topographic diversity that characterizes mountains, we generated biogeographic units (hereafter ecosystems) in each mountainous region defined by climatic, abiotic, and biotic factors.We created these ecosystems by spatially intersecting 3 data layers: Köppen−Geiger climate regions, which are consistent with but more detailed than global ecological zones (FAO, 2012); delineations of terrestrial realms and biomes (Dinerstein et al., 2017) to accommodate intercontinental variation and ecoregional variation; and highresolution (90 m) representations of landforms (Theobald et al., 2015).
We grouped the original 15 landform classes (Appendix S6) into valley bottoms and cool and neutral lower slopes; warm lower slopes and cool and neutral upper slopes; and ridges, mountains, cliffs, and warm upper slopes.Our final mountain ecosystem classification system produced 22−38 climate regions, 14 biomes, and 3 landform classes.We recognize that subjective thresholds of temperature and precipitation variables were used to generate climate regions but chose to adopt the Köppen−Geiger classification because it is commonly used to examine global ecological patterns (e.g., Elsen et al., 2022;Sayre et al., 2020) and is valuable in presenting, translating, and communicating findings to conservation practitioners, land managers, and decision-makers (Sayre et al., 2020).We excluded ecosystems that constituted < 1% of a region.
To measure the fragmentation of ecosystems resulting from HM, we used distance from the nearest edge (Crooks et al., 2017;Kennedy et al., 2019;Ripple et al., 1991) calculated from each contiguous area for each unique ecosystem class (i.e., a patch).This metric is appropriate for quantifying fragmentation patterns in mountains because it also accounts for patch shape, which is especially important for mountains because narrow landform patterns commonly are not resolved at the coarser spatial resolution of 1 km.
To measure the fragmentation of ecosystem patches, f is calculated as: where d is the inward distance from the patch edge to each pixel i of ecosystem type j and f is the summation of the distances over all pixels.Squaring d places greater emphasis on the infrequent but disproportionately important large patches (Li & Archer, 1997).To account for fragmentation due to natural edges of ecosystems and focus only on fragmentation due to HM, f is normalized by the natural fragmentation of the ecosystems.Thus, F is calculated, for each of the 3 HM classes as: where f n is the sum of the inward squared-distance-from-edge values for all pixels in each contiguous (8-neighbor) patch for all patches in an ecosystem class.The f is similar to f n , but patches can also be fragmented by HM where H exceeds a threshold: H > 0.1 or H > 0.4 (and H ≥ 0.0 for f n ).Consequently, F varies from 0.0 (no fragmentation) to 1.0 (highly fragmented).
To assess the coverage of PAs relative to the 2010 Aichi Biodiversity Target 11 (17% protection) and 2030 GBF Target 3 (30% protection), we reported the proportion of ecosystem classes in each mountain region that were ≥ 17% and ≥ 30% protected, respectively (following Aycrigg et al., 2013).We obtained PAs (polygons and points) from the World Database of Protected Areas (UNEP-WCMC and IUCN, 2021) and included refinements for the Y2Y region from Hebblewhite

RESULTS
The average H for each mountainous region varied from 0.02 to 0.19 (Table 1 & Appendix S4).Compared with the global average H of 0.10, H for the Alps was nearly double (0.19), whereas the Y2Y was about half (0.05) and the Himalayas was about one-quarter (0.02), respectively.The Y2Y had the lowest H (0.02) and the highest percentage (95%) of lands classified as wild (Figure 2 & Appendix S7).The Albertine Rift, Alps, and Andes had at least triple the area of shared lands (17−33%) as compared with the Y2Y (5%).The Alps had 15% of lands classified as developed.The percent wild in the mountainous regions declined strongly as a function of increasing HM (R 2 = 0.98; Appendix S8) and for shared lands increased moderately (R 2 = 0.90) because HM included aspects of low-intensity agricultural activities, such as pasture and grazing.Mountainous regions with high proportions of land in PAs did not tend to have low degrees of HM (i.e., average H values, Pearson's r = −0.17).
In addition to the loss of wild lands (defined by summed percentages of shared and developed classes), ecosystems were fragmented 1.3−27.3%across mountainous regions when fragmented by developed land only and 3.9−43.0%when fragmented jointly by shared and developed lands (Table 2 & Appendix S9).The Y2Y had fragmentation values of 1.7% and 4.6% (developed vs. jointly, respectively), whereas the Alps had the highest levels of fragmentation at 26.8% and 43.0% for developed versus jointly, respectively.Across all regions, wild lands were moderately fragmented by developed and shared lands (R 2 = 0.81; Appendix S10), and wild and shared lands were moderately fragmented by developed lands (R 2 = 0.83).
Overall protection coverage ranged from 7.2% (the Himalayas) to 36.2% (Great Eastern Ranges) of the mountainous regions (Table 1).The number of ecosystems in each region ranged from 10 (Great Eastern Ranges) to 48 (Andes; Table 3 & Appendix S11).In each region, the representation of ecosystems (i.e., the proportion of the ecosystems that exceeded a specific proportion of PA) ranged from 25.0% to 81.0% at the 10% threshold and declined 11.1% to 70.0% and 0% to 70.0% at the respective thresholds of 17% and 30%.The Y2Y had the highest representation of ecosystems: 81% for the 10% threshold, declining to 33% for the 17% protection threshold and to 14% for the 30% threshold.Four of the 6 regions had <50% of their ecosystems represented at the 10% protection threshold, <33% at the 17% threshold, and <15% at the 30% threshold.

DISCUSSION
Our results showed that the major mountainous regions varied in their degrees of HM, human-caused fragmentation, and PA coverage.In general, the level of HM in the major mountainous ranges was lower than the global average, although in the Alps it was nearly twice the global average.Only the Himalayas and Y2Y had levels of HM that were less than half the global average-perhaps better aligned with expectations that mountain regions would be substantially lower than the global average.
Although mountain ecosystems are naturally fragmented to some degree, we found they were subject to substantial additional fragmentation caused by HM.Under the assumption that developed areas fragment natural ecosystems, we found minimal (<3%) fragmentation in the Albertine Rift, Himalayas, and Y2Y but substantial fragmentation (>25%) in the Alps.That is, the average geographical distance in an ecosystem caused by an edge formed by developed land was reduced by >25%.Fragmentation caused by shared versus developed lands was markedly greater for all mountainous regions: 1.7−4.6%for the Y2Y and >25 to >40% for the Alps.This suggests that fragmentation from shared landscapes, which is more prevalent in all mountainous regions, may affect biotas and processes sensitive to fragmentation.
Protection levels for regions other than the Himalayas exceeded the global average (7.2%) for all PAs, consistent with

TABLE 2
The average percent fragmentation (F) for ecosystems in each of 6 mountain regions and percent F calculated as fragmentation due to human modification at greater than developed (F developed ) or shared (F shared ) classes (H > 0.1 and ≥ 0.4, respectively).findings of global biases toward disproportionately protecting high-elevation areas (Elsen et al., 2018;Joppa & Pfaff, 2009).Nevertheless, each mountainous region we assessed, except the Great Eastern Ranges, will require additional protection to reach the 30% target by 2030.Our results strongly support the need to expand PAs in these mountainous regions to reach this area target and to be strategic in additional protection placements to enhance ecosystem representation.For example, although overall the Y2Y is currently 15.6% protected, two-thirds of ecosystems were not protected at levels that meet the 17% Aichi target, and only one-sixth currently meet the 30% GBF target.Importantly, mountainous regions with higher proportions of PAs in our analysis generally did not have lower percentages of HM, suggesting that their ecosystems continue to face threats from human activities, despite their perceived remoteness and level of protection.These findings underscore the critical need to enhance ecosystem representation-concomitant with ensuring a well-connected, effectively managed, and equitably governed PA networkwhen establishing new PAs to fully achieve Target 3.Although area-based targets may incentivize countries to advance conservation, a challenge is to ensure that increasing PA coverage actually achieves the GBF purpose of halting and reversing biodiversity loss (Maxwell et al., 2020).

Lands within 10 km added
Using ecosystem units based on climatic and fine-grained abiotic factors overcomes limitations of coarser ecosystem surrogates that are often too large and poorly aligned with mountains.Moreover, detailed maps of ecosystems enable more ecologically relevant assessments of ecosystem representation in mountains to inform reporting toward targets.Our work thus provides a more ecologically informed and detailed assessment of protection and representation in mountainous regions.
Our approach also provides a strong foundation for extending this analysis to mountainous regions globally (e.g., Jacobs et al., 2023).Previous assessments show large regional variation in patterns of protection (Joppa & Pfaff, 2009), human pressure (Plumptre et al., 2016), and forest loss (He et al., 2023), which suggests that fragmentation patterns are also likely to be regional and context specific.We also suggest that future assessments consider the nuanced but important differences between mountain delineations that prioritize geomorphology (e.g., GMBA 2) versus socioecological context.We note that HM is likely underrepresented in our analyses due to the limits of global data sets; thus, additional efforts might also examine fragmentation due to land-use change and address climate adaptation (Elsen et al., 2023;Thurman et al., 2022).
We speculate that HM and climate change will interact in potentially unanticipated ways.For example, fragmentation outside wild areas may decrease as they merge with nearby ecosystems in response to climate or become eliminated altogether (Elsen et al., 2020;Rhemtulla et al., 2002).Or, fragmentation may increase from accelerated upslope forest loss observed in many mountains globally (Feng et al., 2021;He et al., 2023).Sociopolitically, adaptation in mountain ranges may be challenging because equitable governance at the community level differs by local context (Tucker et al., 2021), whereas complications of governance are likely at national and broader scales due to transnational dynamics (Titley et al., 2021).
We anticipate our work can also be useful to inform conservation planning and decision support.For example, prioritization of protected lands could be refined by identifying minimally fragmented, unrepresented ecosystems, or high-diversity locations; connections among ecosystems could be maintained by targeting protection of unfragmented lands linking PAs; and adaptation strategies could be informed by mapping potential ecosystem shifts by leveraging Köppen−Geiger maps reflecting climate forecasts (from Brun et al., 2022).
The combined impacts of current land use and climate change on mountains, coupled with conservation through PAs that overwhelmingly falls short of global targets currently, highlights the magnitude of the threats and the urgency for action needed to achieve these targets.This also reinforces that additional protection required to meet area targets must simultaneously enhance connectivity and representation to ensure long-term biodiversity conservation.
Mountains are critical for conserving biodiversity and ensuring human well-being.Understanding how human activities fragment mountain ecosystems is also important for their conservation.Our analysis of global patterns of HM, fragmentation, and landscape protection in 6 large mountainous regions underscores pressing concerns about the conservation of mountainous regions more broadly.Although mountains are commonly perceived as remote and disproportionately protected, we found most of the mountainous regions exceeded half the global average of HM.Protection levels for all regions but the Himalayas exceeded the global average, yet ecosystem representation under protection was low: less than one-third and approximately one-fifth of ecosystems currently meet the 17% and 30% protection targets, respectively.Although the Himalayas showed low levels of HM, they are vulnerable to rapid human-caused habitat loss and fragmentation due to their current low levels of protection.By contrast, the Y2Y region has the lowest level of HM, the highest proportion of wild landscapes, and a high level of protection; yet, it too is characterized by only moderate ecosystem representation that is under protection.Across all mountainous regions, some ecosystems experience human-caused fragmentation levels near 40%.To achieve 30% protection and global protected-area goals, mountainous regions will require efforts to better protect the diversity of ecosystems.Our results underscore that protecting and restoring intact representative ecosystems, minimizing fragmentation, and connecting ecosystems across mountainous regions remain urgent conservation priorities.

FIGURE 1
FIGURE 1 The 6 major mountainous regions analyzed in the study of protection, fragmentation, and representation of ecosystems defined by aggregating mountainous regions from the Global Mountain Biodiversity Assessment (Snethlage et al. 2022a) (detailed cross-walk in Appendix S2), from Plumptre et al. (2016) for the Albertine Rift, and from Databasin (2023) for Yellowstone to Yukon (ellipses, map scale at 2000 km diameter).

FIGURE 2
FIGURE 2 Levels of human modification (H) (left or top) in 6 major mountainous regions (H < 0.01, large wild areas; 0.1 ≤ H <0.4, shared lands; 0.4 ≤ H <1.0, cities and farms) and the region's climate-abiotic ecosystems used in the fragmentation analysis (right or below): (a) Albertine Rift, central Africa, (b) Alps, western Europe, (c) Andes, western South America, (d) Great Eastern Ranges, eastern Australia, (e) Yellowstone to Yukon, northern North America, and (f) Himalayas, northern South Asia (light gray, countries).

TABLE 1
For the world's major mountain regions (1 from each continent excluding Antarctica), the average area, average human modification value (H), percent protected area, and percentage of region in 3 human modification classes.

millions of km 2 ) H (average) Protected (%) b Wild (%) Shared (%) c Developed (%)
(Gorelick et al., 2017)ected results are in Appendix S4.et al. (2022)and PAs for China from April 2018.We excluded PAs not considered primarily terrestrial and those designated as biosphere reserves and proposed or not reported.Protection amounts were calculated based on all PAs (including not reported, assigned, or applicable) because PA status categories are not compulsory or consistently reported across all countries.We performed all analyses with Google Earth Engine(Gorelick et al., 2017)software in the WGS84 coordinate system.The code and data sets generated are available at https://zenodo.org/records/10136285. b

TABLE 3
The number of distinct ecosystems in each mountain region for which the ecosystem area exceeds 1% of the region and the percentage of ecosystems that meet the specified threshold of the percentage of protected area (i.e., ≥10% related to the 2010 CBD target, ≥17% related to the 2020 Aichi target, and ≥30% related to the 2030 Kunming-Montreal target).* *Results are provided for all PAs (including not reported, assigned, or applicable) because IUCN protected-area status categories are not compulsory or consistently reported across all countries.