Introduction
Global environmental change is receiving growing attention in the scientific literature, as mounting evidence of its presence and impacts continues to emerge across diverse regions worldwide. ‘Global environmental change’ refers to the cumulative impact of human activities, driven by socioeconomic pressures, that result in climate change, land-cover change, biodiversity loss and alterations in atmospheric chemistry (Steffen et al. Reference Steffen, Sanderson, Tyson, Jäger, Matson and Moore2005), disrupting Earth system dynamics and threatening planetary stability (Freedman Reference Freedman2014). Climate change and land-cover change are critical drivers of these global environmental shifts, with climate change altering weather patterns and reducing agricultural yields, while land-cover change contributes to deforestation and carbon emissions (Houghton & Castanho Reference Houghton and Castanho2023). Together, these types of change disrupt ecosystems and nature’s contributions to people (Díaz et al. Reference Díaz, Pascual, Stenseke, Martín-López, Watson and Molnár2018), driving species extinction and biodiversity loss. Addressing these challenges requires localized conservation strategies that incorporate Indigenous knowledge systems and community participation, which are essential for achieving the global targets outlined in the Kunming–Montreal Global Biodiversity Framework (Saunders et al. Reference Saunders, Grand, Bateman, Meek, Wilsey and Forstenhaeusler2023). The Global Biodiversity Framework consists of four long-term goals for 2050 and 23 targets to be achieved by 2030 aiming to reduce threats to biodiversity, promote sustainable use and ensure equitable benefit-sharing, while also developing tools for conservation implementation.
Tropical forests represent 45% of the world’s forested area (Hartshorn Reference Hartshorn and SA2013), with tropical dry forests (TDFs) constituting 40% of these (Miles et al. Reference Miles, Newton, DeFries, Ravilious, May and Blyth2006). The Neotropics host >50% of the total TDF (Dinerstein et al. Reference Dinerstein, Olson, Joshi, Vynne, Burgess and Wikramanayake2017). However, previous studies have erroneously included African savannahs in global estimates, underreporting that the Neotropics contain c. 24% of the global TDF (Rivero-Villar et al. Reference Rivero-Villar, de la Peña-Domene, Rodríguez-Tapia, Giardina and Campo2022), distorting understanding of the Neotropics and neglecting their critical role in TDF conservation. Mexico contains the largest global extent of TDFs, yet 73% of these forests are disturbed, mainly due to farming (Mendoza-Ponce et al. Reference Mendoza-Ponce, Corona-Núñez, Kraxner, Leduc and Patrizio2018) and fires (Corona-Núñez et al. Reference Corona-Núñez, Li and Campo2020). With only 6.7% of these forests protected nationally compared to 13.4% in the Neotropics (Comer et al. Reference Comer, Hak, Josse and Smyth2020), their biodiversity and ecosystem services are under significant threat, impacting nature’s contributions to people and human well-being. TDFs harbour large Indigenous communities (Stoner & Sánchez-Azofeifa Reference Stoner and Sánchez-Azofeifa2009) that play a vital role in biodiversity conservation due to their close relationships with local ecosystems and their sustainable use of natural resources (Simón-Salvador et al. Reference Simón-Salvador, Arreortúa, Flores, Santiago-Dionicio and González-Bernal2021). TDFs also support high biodiversity and rich species endemism (Dirzo and Raven Reference Dirzo and Raven2003). Despite their ecological and socioeconomic importance, these forests have received less attention and protection compared to humid tropical forests (Siyum Reference Siyum2020).
This study aims to evaluate the localized impacts of global environmental changes on a Mexican TDF. We focus on the influence of local human activities driven by socioeconomic pressures, biophysical factors and climate change impacts on land-cover change and on biodiversity loss on the southern Pacific coast of Oaxaca. The research: (1) evaluates future land-cover change under different global environmental change scenarios; (2) identifies terrestrial vertebrate species threatened by land-cover change and the consequences of their loss for nature’s contribution to people; and (3) proposes alternative conservation management practices, aligned with the goals of the Global Biodiversity Framework.
Materials and methods
Study site
This study focuses on the southern Pacific in Oaxaca, Mexico, encompassing 1471 km2 of mature and secondary TDF (Corona-Núñez et al. Reference Corona-Núñez, Mendoza-Ponce and López-Martínez2017) stretching between 15.6° and 16.1°N latitude and between 96.5° and 95.9°W longitude (Fig. 1). The climate in the region is mainly hot-subhumid with a prolonged dry season. This region is recognized for its rich diversity of plant and animal life, and its biodiversity is classified as endangered, making the region a priority for conservation (Bezaury Reference Bezaury, Ceballos G, Martínez, García, Espinoza, Creel and Dirzo2010). The study region encompasses three adjacent municipalities with similar socioeconomic characteristics (Table S1).
Figure 1. Location of the study region on the Pacific coast of Oaxaca, Mexico. Each of the three municipalities studied is delimited by its political borders.
In the study region, local people primarily utilize resources for subsistence, including fuel-wood extraction (Corona-Núñez et al. Reference Corona-Núñez, Mendoza-Ponce and Campo2021), shifting cultivation and bushmeat hunting (Robinson & Bennett Reference Robinson and Bennett2006). However, the communities – mainly the Indigenous Zapotec – play a vital role in biodiversity conservation. A significant portion of the territory is conserved within their community-managed areas, locally known as the Sistema Comunal de Áreas Protegidas (Communal System of Protected Areas), within which community-led initiatives such as reforestation and sustainable resource management contribute to biodiversity conservation and enhance ecosystem resilience (González & Miranda Reference González and Miranda2003).
Drivers of land-cover change
To identify the drivers of land‑cover change, we evaluated land‑cover dynamics using aerial photographs (1996, 2006, 2011) and Sentinel imagery (2021), from which we produced annual land‑cover maps. Each land-cover map comprised seven classes (Table S2). The land-cover classes were contrasted with a range of socioeconomic, biophysical and climatic explanatory variables (Table S3).
To address collinearity, we used Spearman rank‑order correlations and removed one variable from any pair with a value of >|0.70|. We retained the variable prioritizing predictors with clear mechanistic links to land‑cover processes. Then we used the DALEX library to identify and to assess the drivers of the land-cover classes (Biecek et al. Reference Biecek, Maksymiuk and Baniecki2023); the importance of each variable was assessed by applying the ceteris paribus principle, which here examined the effect of one variable on land-cover allocation while keeping all other variables constant. Finally, ceteris paribus profiles were created to visualize these effects (Figs S1–S5).
Spatial model calibration and model validation
To build the land‑cover change model, we calibrated the allocation of each land‑cover class using a three‑step process. First, we randomly sampled c. 0.2% of all pixels (18 500 points) to characterize relationships between anthropogenic land covers and their drivers. This sample size captured spatial variability while minimizing duplicate selections, reducing overfitting and improving model robustness. Second, we tested two modelling approaches using the R package PSIG (Corona-Núñez Reference Corona-Núñez2023): Random Forest machine learning (Breiman Reference Breiman2001) and a General Linear Model (Appendix S1 & Supplementary Method 1). Random Forest was selected for its ability to model complex interactions and non-linear relationships, while the General Linear Model framework was used for its flexibility in handling different types of response variables. Both approaches were applied to generate binomial probability surfaces representing the likelihood of land-cover transitions. For both modelling approaches, we generated an ensemble by averaging the multiple predictions produced by each method. Combining models trained on different data subsets improved overall performance. Third, to allocate future land‑cover, we applied four approaches that weighted historical land‑cover change trajectories (temporal weighting; Supplementary Method 2 & Table S4):
-
• Approach 1 assigned maximum weight (1.0) to the permanence of a land-cover class and penalized its absence across temporal units by c. 0.25.
-
• Approaches 2–4 explored alternative combinations of land-cover class presence/absence across years, allowing for variation in land-cover change temporal sequences differing in penalties for discontinuity in land-cover change trajectories.
-
◦ In Approach 2, higher weights indicated sustained patterns (1.0), while lower weights reflected intermittent (>0.0 and <1.0) or absent patterns (0.0).
-
◦ Approach 3 followed a similar principle but applied stronger penalties for discontinuity in land-cover change trajectories than Approach 2.
-
◦ Approach 4 applied a uniform temporal weight (0.25) for the presence of a land-cover class in each individual year.
-
All model approaches were blindly validated for both quantity and allocation using accuracy, F1 scores and agreement–disagreement components against the 2021 land‑cover map. Accuracy was calculated per class and for the entire landscape as the proportion of correctly classified pixels. The F1 score, computed as the mean accuracy across predictions, provided a balanced measure of model performance. Agreement and disagreement components were derived from a three‑map comparison (1996 observed, 2021 observed and 2021 simulated), evaluating gains, losses and permanence across land‑cover classes at multiple resolutions (Pontius et al. Reference Pontius, Peethambaram and Castella2011). Gains represented transitions into a focal class, losses transitions out of it and permanence unchanged areas. Quantity disagreement captured errors in the total amount of predicted change, whereas allocation disagreement reflected spatial misplacement of change (Pontius et al. Reference Pontius, Peethambaram and Castella2011). All validations were conducted in R using the lulcc package (Moulds et al. Reference Moulds, Buytaert and Mijic2015).
Land-cover change scenarios and threatened terrestrial vertebrates
After identifying the best‑performing model, we projected land‑cover change under climate and socioeconomic scenarios for 2050 and 2080. These horizons align with international climate and biodiversity frameworks and capture mid‑ and long‑term ecological responses (Bufford et al. Reference Bufford, Brandt, Ausseil, Black, Case and Forbes2024). Climate change and human pressures were treated as external drivers of forest transformation, while changes in TDF extent and biodiversity represented system‑level responses within a spatially explicit modelling framework. This structure supported the inference of cause–effect relationships through a spatially explicit modelling approach, linking global drivers to localized ecological outcomes.
To generate scenarios, we dynamically updated climatic and socioeconomic variables (Table S3), assuming that historical relationships between land‑cover change and its drivers persisted into the future. We developed three scenarios – Business as Usual (BAU), Pessimistic and Optimistic – based on Shared Socioeconomic Pathways (SSPs). These scenarios represent plausible trajectories of land‑use decisions and TDF deforestation. Scenario directionality reflected projected changes in TDF extent, with the Optimistic scenario favouring regeneration and the Pessimistic scenario emphasizing forest loss.
The BAU scenario followed observed trends from 1996 to 2021 and aligned with SSP2‑4.5: a ‘middle‑of‑the‑road’ pathway with moderate demographic and economic growth. The Pessimistic scenario followed SSP5‑8.5: a ‘high‑emissions’ pathway characterized by intensive economic expansion and strong pressure on natural land covers, producing the lowest projected forest extent. The Optimistic scenario followed SSP1‑2.6: a ‘sustainability‑orientated’ pathway with reduced population growth, improved land‑use efficiency and lower emissions; here, we assumed reduced deforestation based on the upper confidence interval of projected forest cover.
To characterize future climate, we used a multi‑model mean from eight CMIP6 (Coupled Model Intercomparison Project, Phase 6) General Circulation Models (ACCESS‑CM2, CMCC‑ESM2, EC‑Earth3‑Veg, GISS‑E2‑1‑G, HadGEM3‑GC31‑LL, INM‑CM5‑0, IPSL‑CM6A‑LR and MPI‑ESM1‑2‑HR), which provides a more robust representation of climate trends than individual models (Altamirano del Carmen et al. Reference Altamirano del Carmen, Estrada and Gay-García2021). Historical and future climatic data were obtained from WorldClim (Fick & Hijmans Reference Fick and Hijmans2017).
Spatially explicit land‑cover change projections were used as proxies for habitat availability, assuming that forest-cover change is the primary mechanism through which global environmental change affects biodiversity in this region. By overlaying species occurrences with projected land‑cover maps, we identified areas of potential habitat loss and species exposure under each scenario. We focused on native forest‑specialist terrestrial vertebrates, which depend on mature forest structure and typically exhibit limited dispersal, narrow dietary niches and low reproductive rates (Munstermann et al. Reference Munstermann, Heim, McCauley, Payne, Upham, Wang and Knope2022). These traits increase vulnerability to rapid environmental change, forest disturbance and human pressures (Kadoya et al. Reference Kadoya, Takeuchi, Shinoda and Nansai2022).
The threat level for each species was assessed using the International Union for Conservation of Nature (IUCN) Red List (IUCN 2022), considering their endemicity to Mexico or the region. Focusing on TDF indicator species avoided overestimating extinction risks for species with broader ranges and tolerances to human disturbances (Alroy Reference Alroy2017). The risk was evaluated by comparing endemic species distributions with future land-cover change, identifying species that are most vulnerable due to restricted ranges and high endemism (Mendoza-Ponce et al. Reference Mendoza-Ponce, Corona-Núñez, Kraxner and Estrada2020).
Species distribution data were obtained from the National Biodiversity Information System of Mexico (SNIB 2024). Geographical information on species distributions and levels of threat was obtained from the IUCN (2022) and BirdLife International (2019). These maps are based on minimum convex polygons around known occurrences and may include unsuitable habitat, potentially overestimating distributions and underestimating extinction risk (Ramesh et al. Reference Ramesh, Gopalakrishna, Barve and Melnick2017).
Results
Historical tropical dry forest deforestation dynamics
During the period 1996–2021, the TDF showed deforestation in 3.7% (31 km2) of the landscape with an annual rate of 0.15% year–1 (1.2 km2 year–1; Fig. 2a & Tables S5–S8). The main drivers of the TDF loss were related to the expansion of rainfed agriculture and cattle ranching (Figs 2 & 3), explaining 55.9% of the total forest loss. Rainfed agriculture contributed both to deforestation (30.5 km2) and to regrowth (23.4 km2) due to the shifting cultivation practices. The expansion of rural and urban settlements explained 18.0% and 21.9% of the total deforestation, respectively.
Figure 2. Historical (1996–2021) and projected (2050 and 2080; based on historical trajectories) land-cover change trajectories (in km2): (a) total area covered by tropical dry forest; (b–f) total area covered by each anthropogenic cover type. BAU = Business as Usual.
Figure 3. Past (1996, 2006, 2011 and 2021) and future (2050 and 2080) anthropogenic drivers of deforestation: (a) the total area covered by each anthropogenic cover type; (b) the proportion relative to the total anthropogenic area. BAU = Business as Usual; Opt = Optimistic; Pes = Pessimistic.
The TDF showed a consistent deforestation trend, with projected reductions of 10.4% and 14.4% by 2080 compared to its extent in 1996 under the Optimistic and Pessimistic scenarios, respectively (Fig. 2a). Anthropogenic land covers generally showed growth rates in all scenarios (Fig. 2b–f). In the Optimistic scenario, agricultural land covers would decline in their total area, while minimal changes were expected in the BAU scenario, and in the Pessimistic scenario all anthropogenic land covers would expand, primarily driven by farming (Fig. 2b–f & Supplementary Note 1).
Projections of the land-cover changes
Overall, Random Forest outperformed the General Linear Model in terms of accuracy, especially when combined with Approach 2 of the weighting method, demonstrating the highest performance by accurately allocating 93% of the land-cover classes (Fig. 4 & Supplementary Note 2). The model had a ‘miss’ rate of 3–4% (simulated as persistence), a ‘wrong hit’ rate of ≤1% (simulated changes to the wrong category), a ‘false alarm’ rate of 2–3% (persistence simulated as change), a ‘hit’ rate of 13–15% and a ‘correct rejection’ rate of c. 80% (Fig. 4). All scenarios indicated that, by 2050, the TDF would experience a reduction in extent (7–9% compared to 1996), with an annual deforestation rate of 0.13–0.17%.
Figure 4. Spatial distribution of real (1996, 2006, 2011 and 2021) and modelled (2021) land covers, including the category losses and gains over 1996–2021, and analysis of the agreement and disagreement between the observed and simulated land-cover map in 2021. The graph shows the proportion of each component of agreement at different resolutions for the period of 1996–2021. The model is based on Random Forest modelling Approach 2.
Deforestation was primarily driven by the expansion of rainfed agriculture, rural areas and cattle ranching. Rainfed agriculture is projected to increase by 16% under the BAU scenario and to double under the Pessimistic scenario, whereas it is projected to decrease by 80% in the Optimistic scenario (Fig. 3). These transitions are expected in the northern part of the region (Fig. 5). In the south, rainfed and irrigated agriculture as well as urban expansion drive deforestation and forest degradation in the transition zone between TDF and riparian vegetation. Under the Pessimistic scenario, rainfed agriculture would expand but would be allocated differently than in the BAU scenario (Fig. 5). The Optimistic scenario produces the lowest deforestation due to the near absence of rainfed agriculture, with remaining changes being driven mainly by urban growth and cattle ranching (Supplementary Note 3).
Figure 5. Future (2050 and 2080) land-cover class distributions under different socioeconomic and climate change scenarios, based on Random Forest modelling Approach 2. BAU = Business as Usual.
Drivers of land-cover changes
Ceteris paribus profiles from the Random Forest model showed that biophysical and socioeconomic factors influenced land‑cover classes differently. Urban settlements (Fig. S1) were driven mainly by accessibility, proximity to roads and highways, low elevation and flat terrain. Rural settlements (Fig. S2) were associated with roads, highways and nearby farms. Irrigated agriculture (Fig. S3) occurred mostly below 100 m and was shaped by soil physical and chemical properties and water stress. Cattle ranching (Fig. S4) was linked to distance from rural settlements and transportation networks and favoured drier environments. Rainfed agriculture (Fig. S5) was constrained by climate, soil characteristics, lowland areas and gentle slopes.
Under climate change, water stress is projected to increase across the region by 2080 in all scenarios. Aridity is expected to rise by 8.9% ± 1.5% to 19.4% ± 1.1%, reaching up to 23.9% in the north-east. Mean temperature is projected to increase by 2.3°C ± 0.3°C to 3.7°C ± 0.1°C (8.9% ± 1.4% to 14.3% ± 1.1%), while precipitation is expected to decline by 0.7% ± 0.2% to 7.9% ± 0.5%, corresponding to annual reductions of 8.6 ± 2.2 mm to 94.4 ± 8.0 mm, depending on the scenario.
Identification of threatened vertebrate species
Our analysis identified 99 native vertebrate species in the region: 12 amphibians, 44 birds, 10 mammals and 33 reptiles. Of these, 44.4% are endemic to Mexico and 22.2% are regional endemics restricted to Oaxaca and Guerrero. Amphibians represent the highest number of endemic species (Table S9) and the highest proportion of threatened species (83.3%), followed by reptiles (36.4%, excluding Data Deficient species), mammals (30.0%) and birds (22.7%).
Across all scenarios, 43 endemic and threatened vertebrate species are locally affected by land‑cover change. Reptiles are the most impacted group (20 species), followed by amphibians (10 species), birds (10 species) and mammals (3 species; Fig. 6a). All endemic mammals are classified as Vulnerable or Endangered. Among amphibians and birds, 30% of endemic species are Vulnerable, Endangered or Critically Endangered, while for reptiles this proportion is 10% (Fig. 6b).
Figure 6. Classification of threat of the 35 endemic vertebrate species in the study region: (a) numbers of species and (b) percentages of species. CR = Critically Endangered; DD = Data Deficient; E = Endangered; LC = Least Concern; NT = Near Threatened; Vu = Vulnerable.
The expansion of farming activities not only anticipates that 43 endemic species would be negatively affected, but also that 27 species would face local extinction. The species that could become extinct are seven species of amphibians (Charadrahyla juanitae, Craugastor uno, Dermophis oaxacae, Eleutherodactylus syristes, Exerodonta juanitae, Exerodonta melanoma and Megastomatohyla pellita), five species of birds (Amazona finschi, Cyanolyca mirabilis, Eupherusa cyanophrys, Peucaea sumichrasti and Ramosomyia viridifrons), three species of mammals (Peromyscus melanurus, Sigmodon planifrons and Spilogale pygmaea) and 12 species of reptiles (Anolis boulengerianus, Anolis nebuloides, Ctenosaura oaxacana, Geophis sallaei, Kinosternon oaxacae, Lepidophy malineri, Micrurus bogerti, Ophryacus undulatus, Phyllodactylus muralis, Porthidium dunni, Sceloporus tanneri and Tantilla oaxacae).
Discussion
Historical dynamics of land-cover change
Deforestation rates in the region align with the national average of 0.22% year–1 (Mendoza-Ponce et al. Reference Mendoza-Ponce, Corona-Núñez, Kraxner, Leduc and Patrizio2018). The main drivers of TDF loss are the expansion of rainfed agriculture and cattle ranching. Rainfed agriculture is largely subsistence‑based and is characterized by low productivity: maize yields range from 1.0 ± 0.1 t ha–1 year–1 to 1.3 ± 0.2 t ha–1 year–1 (SIAP 2020), far below the global average of 5.8 t ha–1 year–1 (Erenstein et al. Reference Erenstein, Jaleta, Sonder, Mottaleb and Prasanna2022). In contrast, more profitable activities attract labour migration by offering greater employment opportunities (Corona et al. Reference Corona, Galicia, Palacio-Prieto, Bürgi and Hersperger2016). These economic differences explain the concentration of rainfed agriculture in the northern part of the region. Under the Pessimistic scenario, rural settlements, irrigated agriculture and cattle ranching are projected to expand mainly in the more accessible southern portion of the landscape.
Implications of climate change for agricultural practices
Water availability is a critical constraint for agriculture, and projected warmer, drier conditions suggest that agricultural practices may expand as farmers attempt to offset declining yields (Haro et al. Reference Haro, Mendoza-Ponce, Calderón-Bustamante, Velasco and Estrada2021, North et al. Reference North, Franke, Ouweneel and Tristos2023). Moreover, historically, fallow periods lasted c. 20 years, but they have recently declined to 6–8 years (Corona Reference Corona2012). Shorter fallows alter forest structure, species composition and seed‑bank dynamics (Dalle & de Blois Reference Dalle and de Blois2006), ultimately increasing deforestation rates and reducing vertebrate diversity and abundance. Increases in secondary vegetation may also reflect rural abandonment and the erosion of biocultural landscapes. Interpreting land‑cover outcomes therefore requires considering social and cultural processes, especially where conservation depends on the continuity of traditional livelihoods.
Impacts on the extirpation of endemic and threatened vertebrates
The region’s vertebrate fauna faces substantial risk from land‑cover change, wildfires and human‑induced burning associated with shifting cultivation and grassland creation (Montoya et al. Reference Montoya, Corona-Núñez and Campo2023). These threats are expected to intensify under climate change (IPCC 2021), further limiting natural regeneration and increasing the need for active restoration to recover plant species assemblages (Jasso-Flores & Corona-Núñez Reference Jasso-Flores and Corona-Núñez2025).
Although the effects of land management practices such as fire use and selective logging on species distributions remain poorly documented, we estimate that 27 Mexican endemic species in the region are at risk of global extinction. Three rodent species (P. melanurus, S. planifrons and S. pygmaea) are particularly vulnerable due to microhabitat specialization and restricted ranges (Cantú-Salazar et al. Reference Cantú-Salazar, Hidalgo-Mihart, López-González and González-Romero2009, Ramirez & Quintero Reference Ramirez and Quintero2016). Rising temperatures and declining precipitation further increase their risk of local extinction (Rodríguez-Ruiz et al. Reference Rodríguez-Ruiz, Juárez-Agis, Sánchez, Salome and Hugo2022).
Bird populations are also expected to decline and to experience range contractions. On average, Mexican bird species may lose 25% of their geographical range under climate change (Sierra-Morales et al. Reference Sierra-Morales, Rojas-Soto, Ríos-Muñoz, Ochoa-Ochoa, Flores-Rodríguez and Almazán-Núñez2021). Although C. mirabilis may expand northwards under future climatic conditions, deforestation limits its ability to establish new populations.
Many amphibians and reptiles in the region are classified as Data Deficient (Simón-Salvador et al. Reference Simón-Salvador, Arreortúa, Flores, Santiago-Dionicio and González-Bernal2021), leaving them largely excluded from conservation planning despite evidence that many may be highly threatened with extinction. Up to 85% of Data Deficient amphibians may face extinction risk (Borgelt et al. Reference Borgelt, Dorber, Høiberg and Verones2022). Strengthening conservation in ecologically important areas, particularly transition zones between TDF and riparian vegetation, is essential.
Amphibians and reptiles are especially sensitive to habitat loss, with their population trends closely tied to habitat use. Declines in amphibian populations across Mexico have been linked to land‑cover change, climate change, contamination and disease (Lips et al. Reference Lips, Mendelson, Muñoz-Alonso, Canseco-Márquez and Mulcahy2004). Four endemic snake species in the region are threatened by both habitat loss and negative human perceptions, which often lead to intentional killing (Pandey et al. Reference Pandey, Subedi Pandey, Devkota and Goode2016, Kido Cruz et al. Reference Kido Cruz, Zuñiga Marroquín and Kido Cruz2019). Similar attitudes in the study area contribute to declines in rare and endangered species, increasing the risk of local extirpation. Additionally, Geophis species are highly sensitive to climate change, with extinction risk rising as similar habitats disappear (Archis et al. Reference Archis, Akcali, Stuart, Kikuchi and Chunco2018).
Implications of species extinctions for nature’s contributions to people
The loss of 27 Mexican endemic vertebrate species carries major ecological and socio‑environmental consequences, threatening the integrity of the TDF. Endemic vertebrates perform irreplaceable functional roles, supporting ecosystem resilience and the regeneration capacity of these highly threatened forests (Rubalcava-Castillo et al. Reference Rubalcava-Castillo, Sosa-Ramírez, Luna-Ruíz, Valdivia-Flores and Íñiguez-Dávalos2021). Their extirpation disrupts key ecological processes, reduces resilience to climate change and diminishes essential services such as pollination, water regulation and soil fertility (Pearson et al. Reference Pearson, Martínez-Meyer, Velázquez, Caron, Corona-Núñez and Davis2019, Zambrano et al. Reference Zambrano, Fernandez Vargas, González, Mendoza-Ponce, Vazquez Prada and Flores Lot2025). Rodents and birds, for example, are critical seed dispersers (Dickman Reference Dickman1999), while small carnivores regulate pest populations and interact with other herpetofaunal species (Williams et al. Reference Williams, Maree, Taylor, Belmain, Keith and Swanepoel2018). Biodiversity loss also erodes the biocultural heritage of Indigenous and local communities, for whom endemic species are central to traditional knowledge, cultural identity and livelihoods grounded in the sustainable use of forest resources (Alcántara-Salinas et al. Reference Alcántara-Salinas, Hunn, Ibáñez-Bravo, Aldasoro-Maya, Flores-Hernández and Pérez-Sato2022).
Integrated conservation strategies for threatened vertebrates
TDF regions and their threatened species face interconnected pressures. Protected areas remain insufficient, while deforestation and frequent fires continue to degrade habitat (Corona-Núñez Reference Corona-Núñez2022). Under the Optimistic scenario, deforestation would decline by 23–38% relative to BAU and Pessimistic outcomes, largely due to reduced rainfed agriculture. This highlights the need for strategic conservation actions that address deforestation, forest degradation and their socioeconomic drivers. Effective biodiversity strategies must integrate local development, sustainable resource use, landscape planning and the protection of Indigenous rights. We propose five priority targets:
-
(1) Ensure food sovereignty through sustainable agroecological practices that reduce pressure on biodiversity, improve land management and support long‑term agricultural sustainability.
-
(2) Diversify local economies, particularly in the primary sector, through commercial forestry, agroecoforestry and REDD+ initiatives that promote sustainable livelihoods and reduce deforestation. These actions can enhance forest restoration, limit fragmentation and improve biodiversity, especially in the northern part of these municipalities where conditions favour forest recovery. Socioeconomic resilience and cultural continuity must be central, recognizing that effective stewardship depends on maintaining community‑based land management.
-
(3) Minimizing the use of fire in land management is essential; fire harms soil quality, biodiversity and ecosystem services, leading to reduced agricultural yields and slower forest regeneration (Jasso-Flores & Corona-Núñez Reference Jasso-Flores, Escobar-Chanona and Corona-Núñez2025).
-
(4) Developing compact and better-connected human settlements can help reduce deforestation and forest degradation (Jasso-Flores et al. Reference Jasso-Flores, Escobar-Chanona and Corona-Núñez2025a). The current pattern of scattered and poorly connected settlements contributes to ecosystem fragmentation and the loss of urban and peri-urban habitats.
-
(5) Conserve the transition zones between TDF and riparian vegetation. These zones are vital for amphibians such as E. syristes and D. oaxacae, which thrive in riparian environments. The southern Pacific coast, which is home to many endangered TDF species, should be prioritized for conservation efforts.
Study uncertainties and future challenges: implications for conservation planning
Several limitations should be acknowledged. First, the absence of local monitoring restricts our understanding of species distributions and population densities, limiting our ability to assess how landscape fragmentation influences extirpation risk and shifts in species composition. Second, the effects of land management practices such as fire use and selective logging on species distributions remain poorly documented in this region (Corona-Núñez et al. Reference Corona-Núñez, Mendoza-Ponce and Campo2021). Third, the shortening of fallow periods from c. 20 years to 6–8 years (Corona Reference Corona2012) alters forest structure and composition, modifies seed‑bank survival and affects soil properties, including nutrient availability (Jasso-Flores et al. Reference Jasso-Flores, Smallman and Corona-Núñez2025b). These shifts influence vertebrate composition and abundance, suggesting that our estimates of species extirpation may be conservative.
Despite these limitations, this long‑term study provides a robust assessment of species exposed to land‑cover change and an initial framework for implementing COP15 and Global Biodiversity Framework objectives. This approach can be replicated to prioritize conservation strategies in Mexico and other data‑deficient regions where habitat loss remains the most immediate risk.
Conclusions
We emphasize the urgent need in this region for proactive conservation measures and call for comprehensive policies and practices to protect TDF biodiversity and nature’s contributions to people in the face of global change. Rainfed agriculture is identified as the primary driver of deforestation, highlighting the necessity of effective climate change mitigation practices and policies to prevent significant losses in nature’s contributions to people and biodiversity. Implementing landscape planning and sustainable agricultural practices to increase agricultural yields can contribute to biodiversity conservation and promote forest regeneration. Furthermore, diversifying economic opportunities, especially within Indigenous rural communities, can strengthen conservation efforts. Overcoming challenges in land-use planning is crucial for biodiversity conservation under global change scenarios. Integrated biodiversity spatial planning should be feasible in diverse landscapes such as the TDF, offering a pathway to achieving the objectives of COP15 and the Global Biodiversity Framework. This study offers local recommendations based on the Global Biodiversity Framework to mitigate human activities and prevent species extinction.
Supplementary material
To view supplementary material for this article, please visit https://doi.org/10.1017/S0376892925100313.
Data availability
Data sources are cited within the text in the ‘Materials and methods’ section and included in the ‘References’ section, as well as in Appendix S1 and Table S3.
Acknowledgements
We thank JC from the IE-UNAM and AVMP from the ICA-UNAM for their fruitful early inputs to the manuscript. We thank Equipo Verde Huatulco, AC, for feedback on the land-use planning proposals to achieve the goal of biodiversity conservation. We are also grateful to Dr Nicholas Polunin for recommendations that have significantly enhanced the manuscript.
Financial support
We thank the National Council of Science and Technology (CONACyT), now the Secretariat of Science, Humanities, Technology, and Innovation (SECIHTI), for postdoctoral funding to IJ-F and ROC-N.
Competing interests
The authors declare that there are no known competing financial interests or personal relationships that could have appeared to influence the work reported.
Ethical standards
Not applicable.
Open access
