Climate change is creating a mismatch between protected areas and suitable habitats for frogs and birds in Puerto Rico

Changes in species distributions are one of the most evident fingerprints of climate change. Species range shifts towards higher elevations and latitudes have been documented for a variety of functional and taxonomic groups in different regions of the world (Parmesan and Yohe 2003; Chen et al. 2011, 2012). Changes in species distribution due to climate change have been linked to population declines (Rosenberg et al. 2019; Sánchez-Bayo and Wyckhuys 2019), local extirpation (Wiens 2016; Campos-Cerqueira and Aide 2017) and extinctions (Monzón et al. 2011). In addition, projections based on meta-analyses indicate that this trend will continue in the future, threatening the conservation of fauna worldwide (Nunez et al. 2019). Reflecting the global nature of climate change, population declines have been documented even in pristine and protected areas (Blake and Loiselle 2015). As climate changes, species that presently occur in protected areas may also shift their ranges outside the boundaries of the protected areas, increasing their vulnerability to other human impacts (Monzón et al. 2011). There is thus an urgent need to understand how climate change is altering species distributions relative to distributions of protected areas.

While protected areas are still our primary strategy to conserve species, communities, and ecosystems (Rodrigues et al. 2004; Saout et al. 2013), the recognition that our current network of protected areas may not efficiently protect some species under climate change scenarios dates back to the late 1980s (Hannah 2008). Despite an increase in studies testing and quantifying the role and effectiveness of protected areas in conserving species under future scenarios (e.g., Araújo et al. 2004, 2011; Welch 2005; Hannah et al. 2007; Mackey et al. 2008; Hole et al. 2011; Alagador et al. 2014; Gaüzère et al. 2016), much debate still exists regarding the ability of protected areas to mitigate the effect of climate change on species distributions. Some studies have shown that existing protected areas will be able to mitigate the effects of climate change on bird species distributions in France (Gaüzère et al. 2016), to offer native species refuge from invasive species under climate change worldwide (Gallardo et al. 2017), to facilitate adjustments by bird species to rapid climate conditions in Finland (Lehikoinen et al. 2019), to highly overlap (> 45%) with high diversity of endangered anuran species in Costa Rica (García‐Rodríguez et al. 2012), and to slow bird species decline in Finland (Virkkala et al. 2019). Nevertheless, species occurrence inside protected areas are expected to change because protected areas worldwide will fail in retaining more than 90% of their available climate (based on mean annual temperature and annual precipitation) at current levels (Elsen et al. 2020). In Europe, an in-depth study encompassing a network of protected areas and 75% of terrestrial vertebrates (n = 585) found that current unprotected areas will be more effective in retaining climate suitability for species than existing protected areas (Araújo et al. 2011). An assessment of the distribution of 1200 plant species in Europe found that 6–11% of species would be absent in the existing reserves in the future (Araújo et al. 2004). In the tropics, the distribution and suitable habitat of bird and frog species in the Cerrado and Atlantic Forest (Lemes et al. 2014; Vale et al. 2018; Borges et al. 2019; Velazco et al. 2019) are projected to decline inside protected areas. Given that the population of many species are expected to decline inside protected areas in the future, studies assessing long-term suitable areas for species occurrence outside the boundaries of protected areas are urgently needed. By identifying these suitable areas that are stable through time for different species and taxonomic groups, we can adapt our conservation strategy by creating, expanding, moving, and connecting protected areas to accommodate population range shifts in the landscape.

To assess the role and the effectiveness of protected areas in the future, studies have applied species distributions models (SDM) to evaluate the effect of climate change on species distributions and conservation strategies. Use of SDM is the most common approach used to evaluate the effectiveness of protected areas in conserving species under climate change scenarios (Seo et al. 2009; Araújo et al. 2011; Hannah et al. 2020), and the implementations of these models, as well as their uncertainties, have been extensively reviewed (Araújo and Guisan 2006; Heikkinen et al. 2006; Elith and Leathwick 2009). One common approach to assess the impact of climate change on species distribution is identifying suitable areas in both present and future scenarios (Araújo et al. 2004; Wang et al. 2017; Báez et al. 2019; Beaumont et al. 2019). Nevertheless, the current distribution of a species is only a snapshot of a dynamic process that holds a historical legacy (Maguire et al. 2015). By identifying areas that remained suitable from past climate change, and that will stay suitable in the future, we can prioritize areas for conservation that are more likely to stay stable against climatic fluctuations. Unfortunately, few of these studies have included historical projections to assess habitats that maintain suitable conditions for the persistence of species in the long-term. Moreover, most studies rely on presence-only data (Hao et al. 2019) and often focus on a single species hindering our understanding of community response to climate change. Presence-only data do not provide information about the environmental variables associated with areas where the species are absent, cannot estimate prevalence (proportion of occupied sites) and suffer from sampling bias (Guillera-Arroita et al. 2015). Thus, the dynamic response of species to climate change still poses a challenge for the establishment of successful long-term protected areas networks.

This study focuses on areas that maintain suitable areas for several species of birds and frogs over past, current, and future climate regime periods. These always-suitable areas (ASA) may facilitate species' persistence and should be prioritized for conservation. Given that most protected areas were not established considering climate change or non-static species ranges, we asked the following question: Do the distributions of always-suitable areas for species of birds and frogs overlap with the distribution of existing protected areas? To answer this question, we determined the species distributions of eight frog and 13 bird species based on historical, current, and future climate models and detection/non-detection data from ~ 700 sites and > 900,000 1-min audio recordings in Puerto Rico.

The distributions of 21 bird and frog species were based on detection/non-detection data from > 900,000 1-min recordings from ~ 700 sites across Puerto Rico. Model performance assessed by average AUC varied from 0.63 (Icterus portoricensis) to 0.94 (Eleutherodactylus unicolor) (Fig. 2, Table S1). Model performance accessed by averaged TSS varied from 0.31 (Melanerpes portoricensis) to 0.89 (E. unicolor).

Precipitation in the warmest quarter was the most important variable affecting frog distributions whereas temperature seasonality and precipitation in the warmest quarter were the most important variable affecting bird distribution (Table S2). Precipitation in the warmest quarter increased from the historical (1980–1989) to current time period (2005–2014) across the whole island (Fig. S1). The increase was less than 25 mm for most of the island, but there were areas in the northern karst region and the Central Cordillera where the increase was greater. In contrast, there was an overall drying trend from current conditions to the future (2040–2060). The models suggest that precipitation in the warmest quarter will decline for much of the island by 150 mm or more, and in parts of western Puerto Rico precipitation could decline by at least 325 mm. An exception to this trend is a portion of the Luquillo Mountains, which is projected to have a slight increase in precipitation (Fig. S1).

The distributions of birds varied greatly (Fig. S2). Vireo altiloquus was the most widespread species in all periods, occurring in ~ 80% of Puerto Rico, followed by Loxigilla portoricensis and Melanerpes portoricensis. All other species had limited distributions, occurring in less than 30% of Puerto Rico in any period. Three species (Geotrygon montana, Nesospingus speculiferus, and Setophaga angelae) had very limited distributions occurring in less than 10% of Puerto Rico. ASA were very small (less than 10% of Puerto Rico) for seven bird species (Buteo platypterus, Antrostomus noctitherus, Geotrygon montana, Icterus portoricensis, Nesospingus speculiferus, Setophaga angelae, and Vireo latimeri). The stable areas of three of these seven species (Antrostomus noctitherus, Vireo latimeri, Icterus portoricensis) occurred mostly in the west portion of Puerto Rico. In contrast, the ASA for the other four species were mainly concentrated in montane rainforest habitats.

Most anuran species had limited distributions occurring in less than 10% of Puerto Rico (Fig. S2). In addition, ASA were extremally small (< 2% of Puerto Rico) for the majority of frog species. The grass coqui (Eleutherodactylus brittoni) was the most widespread species occurring in over 39% of Puerto Rico. Eleutherodactylus unicolor was detected only in the Luquillo Mountains in both historical and current time, and its distribution in all scenarios is restricted to this region. Eleutherodactylus gryllus, and E. portoricensis are both currently restricted to the Luquillo and Cayey Mountains. Although historically these species occurred in other regions (e.g., Central Cordillera), the Luquillo Mountains are the only ASA for these species. Eleutherodactylus locustus was detected in the Luquillo and Cayey Mountains in both historical and current time periods, and the always-suitable areas encompass both regions and a small portion of the Toro Negro State Forest. Eleutherodactylus hedricki was detected in the Luquillo, Cayey, and Toro Negro Mountains, both in historical and current time periods, but always-suitable areas encompass only the Luquillo Mountains. Although historically E. richmondi occurred in several locations in Puerto Rico, currently, this species is mainly restricted to the Luquillo and Cayey Mountains and Maricao State Forest. The ASA for this species encompasses a small portion of the Cayey Mountains and several small localities along the Central Cordillera.

A large portion of ASA is outside of protected areas in Puerto Rico (≥ 80%) for birds, frogs, and birds and frogs combined (Table 4). Despite the large extension of ASA outside protected areas, approximately 85% of protected areas overlap within ASA for at least one species (Table 5). These results show that the existing protected area network is located in appropriate areas for species protection, but the extent of these areas captures < 20% of the ASA. Furthermore, the protected area network captures a greater percent of the ASA of birds (78%) than for frogs (39%).

In general, the relationship between the accumulative number of overlapping distribution of species ASA and protected areas was the same for birds and frogs (Fig. 3). As the number of overlapping ASA species distribution increased there was a dramatic reduction in the extent of the area occupied, meaning that areas with a high number of overlapping species ASA distributions are relatively small compared with areas with just a few overlapping species ASA distributions. In contrast, the area with the greatest number of overlapping species had the highest level of protection. The area of overlap of ASA for more than six species was ~ 96 km² for frogs and 388 km² for birds. Areas with a high number of overlapping ASA for frog species (> 6 species) were restricted to few sites in the east (Luquillo Mountains), southeast (Cayey Mountains), and the Central Cordillera (Toro Negro) (Fig. 4). In contrast, areas with a high number of overlapping ASA for bird species were restricted to the west (Maricao State Forest), northwest (Karst Mountains), and to a few sites in the east (Luquillo Mountains) (Fig. 4).

The Maricao State Forest appears as an always-suitable area for the majority of bird species (12 species), followed by the northwest Karst mountains (9 species) and the Luquillo Mountains (7 species) (Fig. 4). In contrast, the Luquillo Mountains appear as an always-suitable area for the majority of frog species (8 species), followed by a portion of Toro Negro (6 species) and Cayey Mountains (6 species) (Fig. 4).

Our findings indicate that existing protected areas in Puerto Rico will not suffice to safeguard Species of Greatest Conservation Need in the context of projected climatic changes that could occur by mid-century under a very high greenhouse gas emissions scenario. A larger portion of the always-suitable areas for several species are outside protected areas, which may compromise the long-term survival of these species. In addition, protected areas are relatively small and sparse, and this can hinder species movement and increase local extinction risk. Therefore, the establishment of new protected areas, buffer zones, and natural corridors can substantially improve the likelihood of species conservation and persistence even in an ‘extreme’ mid-century climate scenario. By combining species distribution projections from different time periods (i.e., historical, current, and future) and identifying the always-suitable area (i.e., ASA) for each species, we can assess habitat stability for species and identify and prioritize areas that are more likely to promote long-term survival of multiple species to mitigate the impacts of climate change.

Despite more than 80% of the always-suitable areas for several species are outside protected areas, three protected areas (El Yunque National Forest, Maricao State Forest and Carite State Forest) had high ASA overlap. Although these areas may play an important role in preserving bird and frog populations, they are currently isolated from other protected areas and land use intensity in the surrounding matrix may not support the well-being of remaining bird and frog population inside the protected areas. In addition, the extension of these areas captures < 20% of the ASA.

Our study and others focused on the effects of climate change on species distributions (Lemes et al. 2014; Noce et al. 2017; Vasconcelos et al. 2018; Báez et al. 2019; Velazco et al. 2019) are limited by the assumption that the current species distribution is at equilibrium with the environment. In contrast, the current distribution may reflect past biotic (e.g., competition, predation, diseases, dispersal) or anthropogenic (e.g., urbanization, habitat degradation) effects, (Nogués-Bravo 2009; Boulangeat et al. 2012; Maguire et al. 2015). In addition, species niches are not static, and current distributions may reflect the adaptation of species to past and current climate conditions (Nogués-Bravo 2009). There is also much discussion about using binarization to determine suitable and unsuitable areas. Although some studies have advocated against this approach (Guillera-Arroita et al. 2015; Santini et al. 2020), there are few alternative solutions. To address these limitations, we provide a conservative approach that requires suitable areas to agree in the three time periods and include several species from two taxonomic groups.

Another limitation of our study is a sampling gap in part of the Central Mountains (Fig. 1). Nevertheless, our work provides the most comprehensive effort in acquiring detection/non-detection data using a standardized methodology, and current species distribution projections are in agreement with our experience in the field. Furthermore, there are few species records for this region in other databases (e.g., Global Biodiversity Information Facility—GBIF).

Overall, the distributions of individual frog species were more limited than those of the bird species, and consequently, patterns of occurrence in the protected areas and ASA are different for frogs and birds. Frogs had lower representativeness in the protected areas than bird species, a pattern also observed at the global scale (Rodrigues et al. 2004), and unfortunately most of the always-suitable areas for both frogs and bird species occurred outside of protected areas. Always-suitable areas for frogs are mostly concentrated in the eastern and central mountains of the island, which are more humid and receive more precipitation than the western part of the island (Daly et al. 2003). In contrast, always-suitable areas for birds are mainly concentrated in the western part of the island. The limited distribution of the frog species associated with a small overlap with protected areas and poor dispersion abilities may increase frog species vulnerability to extinction due to natural and human disturbances. Given that almost all frog species in Puerto Rico are already under a threat category (IUCN 2020), expansion of protected areas to incorporate the present and future distributions of these species may be necessary to ensure continued persistence.

Although birds have higher dispersal abilities than frogs, increasing connectivity between protected areas may also benefit bird populations. For instance, two endemic and forest specialist bird species (Nesospingus speculiferus and Setophaga angelae) currently occur only in a few isolated populations in montane forests suggesting either poor dispersal abilities or specific habitat requirements. Expanding protected areas and connectivity can also mitigate the effects of natural disturbances such as hurricanes and droughts on fauna (Campos-Cerqueira and Aide 2021; Wunderle 1995) by providing a better representation of diverse landscapes. Alternatively, and complementary, the development of buffer zones around protected areas may offer a cheaper solution to increase connectivity. Several studies have shown that an increase in forested coverage around protected areas can increase their capacity to conserve viable populations (DeFries et al. 2005; Hull et al. 2011; Xu et al. 2016).

Most of the always-suitable areas are in higher elevations (> 400 m), with a clear spatial west–east pattern along the central cordillera. It is not a surprise that most of the overlap of always-suitable areas occur in regions at high elevations because these areas coincide with forest refugia during the intense agricultural phase that occurred in Puerto Rico between 1800 and 1940. In addition, many species are moving toward high elevation areas in Puerto Rico (Campos-Cerqueira and Aide 2017; Campos-Cerqueira et al. 2017) and other regions of the world (Chen et al. 2011) due to climate change. Therefore, given the essential role that high elevation areas play in species distribution any strategic plan to improve the conservation of SCGN will need to focus on the protection of areas around the Central Cordillera.

Secondary forest with different ages now covers more than 50% of the island. However, these new forests are still spatially fragmented and mainly concentrated in mountainous regions, whereas high-density urban areas and the few remaining agropastoral activities remain concentrated in the lowlands (Gould et al. 2012). Despite having suitable climatic conditions for the occurrence of species in the long-term, the ability of these secondary forest to provide essential resources (e.g., nesting cavities) to long-term species occurrence remain to be tested. Although forest cover has remained stable since 2004, the human population is still declining, and the likelihood that forest cover will increase is high. In this optimistic scenario, always-suitable areas outside protected areas will remain in land cover states less modified by humans and will be able to provide stable habitats for SGCN, provided that these secondary forests allow long-term species occurrence. Nevertheless, a recent study shows a 90% increase in residential development around protected areas in Puerto Rico between 2000 and 2010 (Castro-Prieto et al. 2017), despite the human population decline, which may compromise the future of always-suitable areas outside protected areas. In this scenario, the establishment of new protected areas in locations with a high number of overlapping ASA from species of great conservation need may be the best conservation strategy. Alternatively, and complementarily, assisted migration, captive breeding, and reintroduction may also be part of future conservation planning (Loss et al. 2011), especially with species with low dispersal abilities like frogs.

The approach presented in this study to assess the effectiveness of protected areas to promote species persistence can be easily applied in other parts of the world to identify always-suitable areas for several taxonomic groups and evaluate the role of protected areas in the context of climate change: (1) historical, current and future climate data from around the world are currently available online (e.g., Worldclim; Fick and Hijmans 2017), (2) audio recorders are more accessible and affordable (e.g. Audiomoth; Hill et al. 2019), and (3) accessible analytical platforms to store, manage and analyses audio recordings and extract data for species distribution models already exist (i.e., RFCx-ARBIMON; LeBien et al. 2020). Furthermore, several user-friendly R packages are available to create species distribution models. Given the speed of projected climatic changes, the variability in species-level responses, and the fact that only 12% of the Earth’s surface (and 16% in Puerto Rico; Castro-Prieto et al. 2019) is currently under some form of legal protection (Jenkins and Joppa 2009; Joppa and Pfaff 2011), effective conservation actions will also likely require efforts to systematically improve the extent, frequency, and duration of long-term monitoring of fauna.

Some protected areas in Puerto Rico provide always-suitable habitats for many bird and frog species, helping to conserve these species in a future, which may include extreme climate change. Nevertheless, long-term protection of these species may not be successful given that (1) regions with a high number of overlapping ASA species are relatively small, (2) protected areas are small and sparse (Castro et al. 2019), and (3) large portion (~ 80%) of always-suitable areas is currently outside protected areas. Focusing special attention to the large proportion of always suitable areas currently outside protected areas may be effective in planning investments for regional conservation. To our knowledge, no new protected areas have been established in Puerto Rico to address the impact of climate on species distributions; nevertheless, we already have all the tools to identify and prioritize areas that are more likely to safeguard species in a global change scenario.