Contrasting Climatic and Land-Use Scenarios Reveal Divergent Futures for the Mexican Narrow-Mouthed Toad, Amphibia, Microhylidae Hypopachus variolosus (Cope, 1866)

Abstract

We assessed the current and possible future predicted distributions of the Mexican narrow-mouthed toad, Amphibia, Microhylidae Hypopachus variolosus (Cope, 1866) across its range to evaluate vulnerability under global change. (2) Methods: We integrated 481 validated occurrence records across the species’ distribution range, including 120 records from Mexico, with bioclimatic and land-cover predictors to build ensemble ecological niche models. We additionally incorporated human footprint metrics to evaluate anthropogenic pressure and projected future habitat suitability under climate and land-use change scenarios. (3) Results: Models showed high performance (TSS > 0.80; AUC > 0.90), identifying temperature and precipitation extremes as main drivers. Suitable habitats extended across both coasts and revealed novel areas in central Mexico. The most suitable habitat occurred under low human pressure, although localized impacts were detected. Deforestation in the Yucatán Peninsula reduced tree cover despite high climatic suitability. Future projections for 2050 under RCP 8.5 indicated marked reductions in modeled high-suitability areas, particularly in central Mexico. (4) Conclusions: These findings indicate high vulnerability to climate and land-use change and support updating distribution limits, incorporating new regions into conservation planning, and reassessing threat status to promote long-term persistence.

1. Introduction

Hypopachus variolosus, commonly known as the Mexican narrow-mouthed toad or sheep frog, is a fossorial microhylid broadly distributed along both the Pacific and Atlantic slopes of Mesoamerica, ranging from southern Texas (United States) to Costa Rica [1,2,3]. In Mexico, it occurs across a wide variety of habitats and elevations, from sea level up to approximately 2100 m a.s.l., reflecting its broad ecological tolerance across the species’ full distributional range. The species is generally considered common and stable, is particularly abundant on the Yucatán Peninsula and is able to occupy a wide range of habitats, including semiarid thornscrubs, savannas, tropical dry forests, pasturelands, open woodlands, and humid canyons, as well as modified environments such as irrigation ditches and rural gardens [4]. Ecologically, H. variolosus is primarily a dietary specialist for ants and termites [5], and its fossorial behavior allows it to remain underground during the day and emerge after rainfall to forage and reproduce in ephemeral pools [2,6].

Despite being currently categorized as Least Concern by the IUCN [3], important knowledge gaps remain regarding the actual distribution, ecological requirements, and vulnerability of the species to global change. Although Hypopachus variolosus is considered a widespread and relatively disturbance-tolerant species, apparent ecological tolerance does not necessarily imply long-term resilience under rapidly changing environmental conditions. Widespread and disturbance-tolerant species can function as early indicators of landscape transformation because their apparent resilience may mask future vulnerability under interacting climatic and anthropogenic stressors. Amphibians are among the most threatened vertebrate groups worldwide, largely because of their high sensitivity to environmental alterations [7,8,9]. Even subtle shifts in temperature, precipitation, or habitat structure can substantially affect survival, reproduction, and population persistence [10,11,12,13]. This sensitivity is particularly relevant in countries such as Mexico, which ranks fifth globally in amphibian richness, harboring approximately 420 species, including 257 frogs, 160 salamanders, and three caecilians [14]. However, this exceptional diversity is increasingly threatened by habitat loss, deforestation, and climate change. Mexico is also among the countries experiencing some of the highest rates of land-cover transformation and currently ranks tenth worldwide in tropical forest loss, with the Yucatán Peninsula representing one of the most heavily affected regions [15]. Understanding how land-use change and climate warming may reshape amphibian distributions is therefore essential for effective conservation planning.

Ecological niche modeling (ENM) provides a robust framework for evaluating habitat suitability and forecasting potential range shifts under environmental change [16,17,18]. In this study, we used a dual-scale modeling approach: (1) at a continental scale, we modeled the full distribution of H. variolosus across its range using climatic predictors to capture the species’ broad environmental tolerance; and (2) at a Mexican country scale, we focused on Mexico, incorporating variables of land cover and land use to refine predictions of suitable habitat under current and future conditions. This combined approach allows us to identify general climatic constraints across a continental-scale climatic model (“Full model”) while providing a more detailed assessment of local habitat availability and anthropogenic pressures in Mexico, where conservation actions are most urgently needed.

Therefore, the objectives of this study were to (1) compile a comprehensive dataset of H. variolosus occurrences across its entire range, (2) develop ecological niche models at both continental (climatic) and Mexico country scale (climate + land-use) scales, and (3) project future distributions under climate change scenarios to evaluate potential contractions, expansions, and areas of conservation concern. By integrating climatic, topographic, and land-use variables, this study provides a refined understanding of the ecological determinants of H. variolosus distribution and highlights regions that may serve as refugia or become increasingly vulnerable under global change.

2. Materials and Methods

2.1. Study Site

The study encompassed the entire geographic range of H. variolosus, which extends across North and Central America—from southern Texas (United States) through Mexico to Costa Rica—covering diverse climatic and ecological regions along both the Pacific and Atlantic slopes Figure 1. This broad-scale approach allowed us to evaluate the environmental suitability of the species across its full distribution using climatic predictors that capture large-scale gradients of temperature and precipitation.

Within Mexico, where H. variolosus occupies a wide variety of ecosystems and faces the most intense anthropogenic pressures, we conducted a refined analysis that incorporated land-use and vegetation-cover variables in addition to climatic variables. This country-scale modeling provided a more detailed understanding of how urban expansion, agriculture, and vegetation transformation influence current and future habitat suitability. By combining a continental-scale climatic model (“Full model”) with a high-resolution country model (“Mexico model”), our framework offers both a general assessment of the species’ climatic niche across its North and Central American ranges and a fine-scale evaluation of habitat quality and human-driven transformations within Mexico, which represents the ecological and conservation core of its distribution.

2.2. Occurrence Data

Occurrence records for H. variolosus were compiled from the Global Biodiversity Information Facility (GBIF, Available online: https://doi.org/10.15468/dl.5c5m4p, accessed on 3 April 2026), which aggregates biodiversity data from multiple primary sources, including museum and institutional collections, scientific databases, and citizen-science platforms. We restricted the dataset to records from 2000 to 2025 to ensure temporal consistency. To improve the data accuracy and avoid potential sampling bias that could lead to model overfitting [19], we applied a rigorous cleaning and filtering process. The raw occurrence data often include duplicated records, spatial outliers, or erroneous coordinates that can affect model performance. We used the EcoNicheS v1.2 platform [20] implemented in R version 4.4.2 [21] and the package CoordinateCleaner v3.0.1 [22] to detect and remove errors such as inconsistent coordinates, points located in centroids or capitals, and records falling outside the known geographic range. To minimize spatial sampling bias, we applied the package spThin v0.2.0 [23], which retains only one occurrence per 1 km² grid cell. This approach prevents redundancy and ensures representative spatial coverage of the species’ distribution across its range. Evaluation of high-elevation records. To evaluate the reliability of high-elevation occurrence records and determine whether extreme altitudinal observations could disproportionately influence environmental characterization, elevation values were extracted for all validated occurrence records using the WorldClim v2.1 elevation dataset at 30 arc-second resolution [24]. The distribution of elevation values was examined using descriptive statistics and the interquartile range (IQR) criterion to identify potential outliers. Because biologically relevant high-elevation records may not necessarily represent statistical outliers, we additionally identified records above the 95th and 99th percentiles of the elevation distribution. The proportion of these records relative to the total dataset was calculated to assess whether high-elevation occurrences represented isolated observations or a substantial component of the species’ altitudinal range.

2.3. Environmental Variables and Environmental Niche Modeling

Three categories of environmental variables were used to construct the ecological niche models: climatic, topographic, and land-use/vegetation. Climatic information, including 19 bioclimatic variables, was obtained from WorldClim v2.1 [24] at a spatial resolution of 30 arcseconds (~1 km²), covering the entire geographic range of H. variolosus from southern Texas (United States) to Costa Rica. Topographic data were derived from a digital elevation model at a 1:50,000 scale, while land-use and vegetation data were obtained from INEGI [25], available for Mexico only. These categorical land-cover classes were reclassified to define major vegetation types and converted into continuous rasters using the filter module of IDRISI SELVA 17.0 [26].

All spatial layers were standardized and resampled at 1 km² resolution using the terra package [27] in R. To reduce multicollinearity, Pearson’s correlation analyses were performed across all predictors, retaining one variable from each pair with high correlation (|r| > 0.7) [28], ensuring that none of the remaining predictors were strongly intercorrelated. Variable selection was performed in ENMTools [29] and usdm within the EcoNicheS framework for R. The final subset of predictors, particularly relevant for amphibian distributions [30,31,32], included climatic variables—BIO5 (maximum temperature of warmest month), BIO6 (minimum temperature of coldest month), BIO13 (precipitation of wettest month), and BIO14 (precipitation of driest month)—and, for the Mexico-specific analysis, the percentage of arid grassland, Pinus forest, Quercus forest, Abies forest, agricultural land, and cloud forest.

2.4. Comparative Modeling Approaches

To evaluate the influence of predictor scope and spatial extent on model outcomes, we implemented two complementary modeling approaches:

Full-range climatic model. This model encompassed the entire distributional range of H. variolosus (southern United States to Costa Rica) and included only climatic and topographic predictors to represent the species’ fundamental niche. The accessibility area (M) was defined as the full geographic range, allowing unrestricted environmental characterization across North and Central America.

Mexico-scale climatic + land-cover model. A second model was constructed to capture fine-scale responses to land-use transformation within Mexico, where detailed vegetation data are available. In addition to climatic and topographic variables, this model incorporated land-cover predictors derived from INEGI [25]. The accessibility area (M) was limited to the extent of Mexico’s territory to assess how anthropogenic drivers influence habitat suitability within this key portion of the species’ range. Both models were built using an identical ensemble modeling framework implemented in biomod2 v4.3 [33] through EcoNicheS for R, allowing direct comparisons between purely climatic (broad-scale) and climatic + land-cover (fine-scale) approaches. This hierarchical framework allowed us to explicitly evaluate the influence of anthropogenic variables by comparing predictions derived from climate-only models with those incorporating land-cover and human-driven landscape transformation variables.

2.5. Model Calibration and Evaluation

All ten algorithms available in biomod2 were employed to maximize model robustness: Generalized Linear Model (GLM), Generalized Boosting Model (GBM), Classification Tree Analysis (CTA), Artificial Neural Network (ANN), Surface Range Envelope (SRE), Flexible Discriminant Analysis (FDA), Random Forest (RF), Multiple Adaptive Regression Splines (MARS), Maxent, and Maxnet. A total of 10,000 pseudo-absence points were randomly generated, with prevalence set to 0.5 to balance presence and absence data [34]. Models were replicated ten times, using independent training and testing partitions, with 80% of occurrence records randomly allocated for calibration and the remaining 20% for validation. Model performance was evaluated using the Area Under the Receiver Operating Characteristic Curve (AUC) [35,36] and the True Skill Statistic (TSS) [37], derived from sensitivity and specificity metrics. Algorithms achieving TSS ≥ 0.8 were retained for ensemble construction using a weighted mean approach [34,38]. The final ensemble model, integrating ten algorithms across ten replicates, was used to project the potential distribution of H. variolosus under current and future environmental conditions.

2.6. Spatial Overlap and Centroid Shift Analysis

To compare model outcomes between approaches, we quantified spatial overlap and centroid displacement between the full-range climatic model and the Mexico-restricted climatic + land-cover model. Binary maps of suitable habitat were generated using 70% and 80% suitability thresholds. Overlap was computed in Terra as the intersection of binary rasters, expressed as the proportion of shared area relative to the total predicted distribution. Centroids of suitable habitat were extracted from each binary raster using the xyFromCell() function in R. The Euclidean distance and bearing angle between centroids were calculated to quantify the magnitude and direction of the predicted range shifts resulting from differences in predictor selection and model extent. All spatial visualizations were produced in QGIS 3.34 and ggplot2 v4.0.3, with arrows indicating centroid movement.

2.7. Land Cover and Climate Change Projections

Land-use and vegetation variables for the year 2050 were generated using the Land Change Modeler for Ecological Sustainability (LCM) module in TerrSet liberaGIS 20.00 software. We used land-use and vegetation maps from 2011 (Series V) [39] and 2014 (Series VI) [25] as the baseline and reference periods, respectively; these layers were interpreted as landscape-level proxies of habitat transformation. The LCM approach relies on artificial neural networks (ANNs), Markov chain matrices, and transition suitability maps derived from multilayer perceptron training [40,41]. This framework predicts land-cover transitions between classes; identifies dominant conversions (e.g., forest-to-agriculture or agriculture-to-urban); and estimates persistence, gain, and loss for each land-cover category [42,43].

For H. variolosus, these land-cover projections were used to assess habitat transformation associated with agricultural expansion and urbanization. Because this species is known to tolerate some degree of habitat disturbance—often recorded in rural gardens, cattle pastures, or irrigation ditches [2,4]—urban environments were not treated as completely unsuitable. Instead, we assigned very low suitability values to dense urban areas, reflecting the degradation of natural microhabitats while recognizing limited persistence in semiurban contexts. This decision also reflects that urban landscapes are not biologically empty for amphibians: some common and widespread species may persist or reproduce where ponds, vegetation patches, gardens, drainage channels, or connected green spaces remain. However, because urban amphibian communities are typically simplified and persistence depends strongly on habitat design and management, urban areas were interpreted as highly degraded environments rather than suitable habitats [44,45].

To simulate future climatic conditions, we used the MPI-ESM-LR global circulation model (GCM) under the Representative Concentration Pathway (RCP) 8.5 scenario for the year 2050, obtained from the downscaled CMIP5 dataset in WorldClim v2.1 [24]. The RCP8.5 scenario represents a high-emission pathway characterized by continued increases in greenhouse gas emissions throughout the 21st century [46]. Although commonly used as an upper-bound or high-impact climate scenario, RCP8.5 provides a useful framework for evaluating potential responses of species distributions under severe environmental change [47,48]. Accordingly, we selected this scenario to provide a precautionary assessment of possible impacts on the distribution of H. variolosus, particularly in the context of conservation planning for vulnerable taxa such as amphibians.

Continuous suitability outputs (ranging from 0 to 1000) were retained for most analyses, as continuous maps preserve probabilistic information and avoid the loss of ecological gradients [49]. However, for comparative assessments of area gain or loss and to visualize spatial contractions more clearly, the continuous predictions were reclassified using three complementary thresholds: (1) the 10th percentile training presence threshold, representing the least conservative limit of predicted presence [50]; (2) the 700 (≈70%) threshold, representing moderately suitable areas; and (3) the 800 (≈80%) threshold, reflecting highly suitable core habitats. While threshold selection is inherently subjective [50], using a set of three hierarchical thresholds allows for a gradient-based evaluation of potential range contraction or expansion. This approach facilitates the identification of habitat stability under increasingly stringent definitions of suitability, helping to distinguish between peripheral and core areas of persistence. Finally, overlap analyses were performed to quantify the proportion of highly suitable areas (≥80% suitability) currently encompassed within Mexico’s natural protected areas (NPAs). This evaluation provides an estimate of the effectiveness of existing conservation frameworks under projected climate and land-use changes. All analyses were conducted in EcoNicheS for R.

2.8. Human Footprint Analysis

We used the global Human Footprint (HFP) dataset (Wild Areas v3, 2009; [51], which provides a cumulative index of human pressures on the environment ranging from 0 (no detectable pressure) to 50 (maximum anthropogenic intensity). This dataset integrates multiple stressors, including human population density, built environments, land-use, infrastructure, and nighttime light intensity. The potential distribution of H. variolosus under current climatic conditions was derived from the ensemble ecological niche model. Continuous suitability values (0–1000) were reclassified using the 70% threshold to identify high-suitability areas, consistent with the upper suitability range used throughout our analyses. The resulting binary raster (presence/absence) was used to mask the HFP layer, ensuring that only footprint values within areas of potential presence were analyzed.

To assess the relative exposure of the species to anthropogenic pressures, we initially followed the classification proposed by [51], who defined three fixed levels of human pressure globally: low pressure: HFP < 20. Medium pressure: 20 ≤ HFP < 40. High pressure: HFP ≥ 40. However, to account for local variation and ensure that these categories accurately reflected the distribution of human pressures across the study region, we also calculated the empirical distribution of HFP values within the modeled range of H. variolosus. Based on this distribution, we derived tercile-based categories corresponding to the lower, middle, and upper thirds of the observed HFP values. This dual classification allowed us to evaluate both the absolute intensity of human pressure (following global standards) and its relative structure within the species’ environmental space. This approach revealed that most areas of high predicted suitability occur under low to moderate human pressure, while high-pressure zones (HFP ≥ 40) occupy a small fraction of both the species’ potential range and the overall study area. Thus, H. variolosus is primarily associated with moderately disturbed rural landscapes, while densely urbanized regions contribute negligibly to its potential habitat. All spatial operations were performed using the terra package in R, with rasters projected to WGS84 and resampled to a 1 km² resolution. Area estimates (km²) per pressure class were computed using the cellSize() function, and proportions were expressed as percentages of total suitable habitat.

2.9. Tree Cover Loss Analysis

Although H. variolosus is not a strict forest specialist, the loss of tree cover represents a reliable proxy for land-use change and habitat transformation across its range. In the Yucatán Peninsula, where the species shows high predicted suitability under current and future conditions, deforestation is primarily driven by agricultural expansion, cattle ranching, and urban development [52,53]. These processes contribute to the loss of ephemeral wetlands, soil compaction, and microclimatic alterations, all of which can reduce the availability of breeding and refuge sites for terrestrial-breeding amphibians such as H. variolosus.

To quantify recent anthropogenic pressures, we assessed tree cover loss (2001–2022) within the Yucatán Peninsula, defined as the area of interest (AOI) encompassing the states of Campeche, Yucatán, and Quintana Roo. The AOI was projected to WGS84 (EPSG:4326), simplified, and exported as a shapefile using the terra and rmapshaper packages in R. The AOI was then uploaded to Global Forest Watch (GFW) (https://www.globalforestwatch.org), where cumulative tree cover loss was extracted from the [54] dataset. Tree cover loss was interpreted not as forest-specific degradation but as an indicator of broader anthropogenic transformation within the species’ highly suitable range. This allowed us to identify regions undergoing rapid habitat modification that may compromise population connectivity and persistence under future climate and land-use scenarios.

3. Results

3.1. Occurrence Records

After data cleaning and spatial thinning, we retained 481 validated occurrence records of H. variolosus across its entire distribution range (Table S1), including the southern United States, Mexico, and Central America (Belize, Guatemala, Honduras, Nicaragua, and Costa Rica). Of these records, 120 corresponded to Mexico (Figure 1). Elevation ranged from 1 to 2318 m a.s.l., with a median elevation of 169.5 m and a mean elevation of 511.2 m. No statistical outliers were detected using the interquartile range criterion, indicating that the highest-elevation records did not represent anomalous observations within the overall altitudinal distribution. Records above the 95th percentile represented X% (n = X) of the dataset, suggesting that high-elevation occurrences are unlikely to disproportionately influence the environmental characterization of H. variolosus.

3.2. Model Performance

All the ecological niche models exhibited high predictive performance. Across the individual algorithms, the AUC values ranged between 0.89 and 0.96, while the TSSs consistently exceeded 0.80, supporting strong discrimination between suitable and unsuitable areas. The ensemble models, which integrate ten algorithms and ten replicates, achieved the highest accuracy (mean AUC = 0.94 ± 0.02; mean TSS = 0.86 ± 0.04).

3.3. Environmental Predictors and Model Comparison

The full-range ensemble model, constructed exclusively with climatic and topographic predictors, achieved high predictive accuracy, with a mean AUC = 0.93 ± 0.02 and TSS = 0.84 ± 0.03 across algorithms. Variable importance analysis revealed that the maximum temperature of the warmest month (BIO5) and minimum temperature of the coldest month (BIO6) were the dominant predictors, jointly explaining more than 60% of the total model variance. These variables were followed by precipitation in the wettest month (BIO13) and precipitation in the driest month (BIO14), indicating that the distribution of H. variolosus was primarily constrained by thermal tolerance and moisture availability. Topographic variables, including elevation and slope, contributed marginally (<10%), reinforcing that climatic gradients are the main determinants of the geographic range of this species at the continental scale (Figure 2A,B). Spatial projections from this model identified extensive areas of high suitability along the Pacific and Atlantic slopes—from southern Sonora through Oaxaca—and along the Gulf and Caribbean lowlands from Veracruz to the Yucatán Peninsula (Figure 2C,D). The predicted range also extended into Central America, encompassing suitable regions in Belize, Guatemala, Honduras, Nicaragua, and Costa Rica (Figure 2A,B). These results delineate the core climatic envelope of H. variolosus, characterized by warm conditions, moderate precipitation seasonality, and elevations generally less than 2000 m. Overall, the full-range climatic model effectively captures the species’ fundamental niche and provides a robust baseline for assessing the potential effects of environmental change across its distribution.

When land-use and vegetation variables were incorporated into the model restricted to Mexico, the predictive performance slightly improved (mean AUC = 0.94 ± 0.01, TSS = 0.86 ± 0.02), indicating enhanced discrimination of suitable versus unsuitable areas at finer spatial resolutions. Climatic predictors remained dominant—BIO6 (min temperature of the coldest month) and BIO13 (precipitation of the wettest month) retained the highest importance—but the inclusion of landcover layers revealed important ecological constraints related to human transformation and vegetation type. Specifically, natural forested environments such as Quercus and Pinus forests, cloud forests, and tropical dry forests exhibited positive associations with suitability (mean importance range = 0.19–0.20). In contrast, agricultural land (0.218), arid grasslands (0.213), and pastureland (0.223) were negatively associated with each other, indicating that extensive land conversion and open vegetation reduce habitat suitability even within climatically favorable zones.

3.4. Predicted Distribution Under the Current Climate

The ensemble ecological niche model predicted a broad potential distribution of H. variolosus across Mexico and northern Central America under current climatic conditions (Figure 2). Areas of highest suitability were concentrated along the Pacific slope—from southern Sonora through Nayarit, Jalisco, Michoacán, Guerrero, and Oaxaca—and along the Gulf and Caribbean lowlands, extending from Tamaulipas and Veracruz into the Yucatán Peninsula. Consistent with historical records, the Yucatán Peninsula, particularly Campeche, Quintana Roo, and Yucatán, emerged as one of the most suitable and continuous regions for the species. In addition, the models identified several previously undocumented areas of high climatic suitability across central Mexico, particularly in the states of Puebla, Morelos, and the southern portion of the State of Mexico.

3.5. Model Comparison and Centroid Shift

The centroid shift analysis of the environmentally suitable areas for H. variolosus revealed consistent spatial displacement between the current and future conditions. Across the species’ entire range (Figure 3A), the centroid showed a clear directional shift, indicating a reorganization of the core area of climatic suitability in response to projected environmental changes. This pattern reflects a redistribution of optimal conditions rather than a uniform expansion of suitable habitat. When the analysis was restricted to Mexico and incorporated vegetation cover and land-use variables (Figure 3B), the centroid displacement became more pronounced and spatially constrained; quantitatively, the suitable area decreased by approximately 32%, and the mean suitability centroid moved ~120 km southeast, consistent with a contraction toward less disturbed regions. highlights the role of anthropogenic landscape modification in shaping habitat availability. Taken together, these results suggest that while climatic conditions may allow potential shifts in the species’ fundamental niche, the realized distribution within Mexico is strongly limited by land-use change, potentially restricting the species’ ability to track future climatic shifts.

3.6. Predicted Distribution Under the Future Climate (2050, MPI-ESM-LR, RCP 8.5)

Future projections under the RCP8.5 scenario revealed contrasting outcomes depending on the modeling framework (Figure 2 and Figure 4B,D). When only climatic variables were considered across the species’ entire range (full-range model), H. variolosus exhibited a notable expansion of suitable climatic space toward 2050 (Figure 2B). The total area of environmental suitability increased from 859,188 km² under current conditions to 1,387,758 km² by 2050 (+61.5%) at the 70% suitability threshold (Figure 4A) and from 390,910 km² to 825,097 km² (+111.1%) at the 80% threshold (Figure 4B). This pattern reflects a broadening of climatically favorable conditions, particularly along the Pacific and Atlantic lowlands and northern Central America, consistent with the species’ tolerance of warm and seasonally dry environments. In contrast, when land-use and vegetation variables were incorporated into the Mexican-restricted model, the results shifted dramatically (Figure 4E–H). Under the same climatic scenario, the Mexico-scale climate + land-cover model projected a marked reduction in the thresholded modeled suitable area by 2050. At the 70% suitability threshold, the modeled suitable area decreased from 624,782 km² to 24,187 km², representing a reduction of 96.1% (Figure 4G). Similarly, at the 80% threshold, the projected suitable area declined from 184,984 km² to 874 km², corresponding to a reduction of 99.5% (Figure 4H and Figure 5). These values represent changes in modeled environmental suitability and should not be interpreted as direct estimates of habitat loss, population decline, or local extinction. Suitable areas became restricted to small, fragmented patches concentrated mainly in the Yucatán Peninsula, southern Veracruz, and isolated remnants in Oaxaca and Guerrero (Figure 4H).

3.7. Human Footprint and Tree Cover Loss Analysis (Global Forest Watch)

The analysis revealed that the vast majority of the potential distribution of H. variolosus lies in areas of low anthropogenic pressure (Figure 6A). Hotspot analysis of the human footprint index (HFP) within the H. variolosus presence area revealed that the highest levels of human pressure (top percentile) were spatially clustered within specific portions of the species distribution (Figure 6B). Although much of the predicted presence area is characterized by low to moderate HFP values (Figure 6B), distinct hotspots of elevated anthropogenic pressure are evident, particularly in regions associated with intensive land use, infrastructure, and human settlements. These hotspots indicate zones where suitable habitat overlaps with strong human modification, suggesting areas of heightened vulnerability despite climatic or environmental suitability. The spatially heterogeneous pattern highlights that human pressure is not uniformly distributed across the species’ range, emphasizing the importance of identifying localized high-impact areas where habitat degradation, fragmentation, or disturbance may disproportionately affect population persistence. Approximately 93.8% (≈610,000 km²) of the suitable habitat (≥70%) overlapped with areas where the Human Footprint Index (HFP) was lower than 20. In contrast, 6.0% of the predicted distribution occurred within regions of moderate pressure (HFP 20–40), whereas only 0.2% overlapped with areas under high anthropogenic pressure (HFP ≥ 40; Figure 6D). These results indicate that H. variolosus currently occupies predominantly low-pressure landscapes, with only a small fraction of its predicted distribution occurring in highly modified environments. However, these values represent present-day conditions and do not necessarily reflect future landscape dynamics. Land-cover projections suggest that ongoing habitat transformation and fragmentation may substantially reduce the availability and connectivity of currently suitable environments, potentially limiting the persistence of the species despite relatively low present human pressure.

The Yucatán Peninsula, identified by ensemble ecological niche models as one of the main areas of climatic persistence for H. variolosus (Figure 6B), has undergone substantial deforestation in the last two decades. Global Forest Watch analysis indicated that between 2001 and 2024, the AOI lost 2.13 million hectares (Mha) of tree cover, equivalent to 20% of the tree cover present in 2000 (Figure 6C). This area is approximately 21,300 km², nearly equivalent to the size of the Mexican state of Hidalgo. When considered alongside human footprint values, which already indicate localized medium (6.0%) and high (0.2%) pressure zones within suitable habitats, these results suggest that key areas of predicted persistence are simultaneously subject to rapid land-use change. The juxtaposition of climatic suitability with ongoing deforestation highlights a critical mismatch between the potential and realized conservation value of the Yucatán Peninsula (Figure 6B).

4. Discussion

Our ecological niche models provide an updated and integrated perspective on the ecological and conservation status of H. variolosus, a species traditionally considered widespread and tolerant but whose persistence under global change may be less secure than previously assumed [3]. By combining a full-range climatic model and a Mexican-restricted model that included land cover and anthropogenic variables, we were able to determine the relative influence of large-scale climatic constraints and local habitat transformation on the species distribution. The results revealed a striking divergence between the potential climatic suitability, considered the main determinant of species distributions at a large spatial scale [55,56,57], and the realized distribution once land use and vegetation variables were incorporated. While the full-range climatic model suggested broad environmental expansion by mid-century, the inclusion of land cover data produced the opposite trend—an abrupt contraction of suitable habitat across Mexico, particularly in central regions. This contrast emphasizes the need to integrate both climatic and anthropogenic predictors in ecological forecasting to avoid overestimating the resilience of species that depend on seminatural habitats for reproduction and persistence [58,59]. Although climate change is often highlighted as a major threat to biodiversity [60,61], its direct impacts are documented for a relatively small proportion of populations in some taxa [62]. For example, only a minor fraction of populations (approximately 7%) are currently recognized as being directly threatened by climate change, with habitat alteration and loss representing a much larger stressor for many species [62,63,64].

4.1. Climatic and Ecological Determinants of the Current Distribution

At the continental scale, temperature and precipitation extremes (BIO5, BIO6, BIO13, BIO14) emerged as the dominant climatic drivers of H. variolosus distribution, defining a niche characterized by warm, seasonally dry conditions typical of tropical dry forests and open woodlands below 2000 m a.s.l. These variables jointly explained more than 60% of the total variance, underscoring that the species’ geographic range is largely governed by thermal and hydric thresholds. The persistence of suitable conditions along both the Pacific and Atlantic slopes, including those of the Yucatán Peninsula, indicates broad physiological tolerance and ecological plasticity. However, when vegetation and land-use variables were integrated at the country scale, climatic favorability alone proved insufficient to explain the observed distribution patterns. The addition of land cover variables revealed that natural forests—Quercus, Pinus, Abies and cloud forests—contribute positively to habitat suitability, whereas arid grasslands, pasturelands, and agricultural areas have strong negative effects [31,65,66]. These findings demonstrate that H. variolosus depends not only on favorable climatic envelopes but also on the structural integrity of seminatural habitats that maintain adequate soil humidity and breeding microhabitats. This pattern is consistent with previous amphibian observations; despite their tolerance of modified landscapes, these amphibians still rely on forested or vegetated substrates to sustain refugia and ephemeral aquatic sites [64,67].

The identification of new suitable areas in central Mexico—particularly in Puebla, Morelos, and the southern State of Mexico—has further expanded the known range of H. variolosus. These regions were not previously included in the IUCN polygon but are supported by recent citizen science records, suggesting that the species’ realized range extends further inland than recognized [3]. The elevational and ecological diversity of these areas implies that H. variolosus can exploit transitional Quercus–Pinus environments under mild precipitation and temperature regimes. However, these same regions are subject to intense anthropogenic pressure, and their apparent suitability may represent transient conditions rather than stable refugia [38,68,69,70]. This highlights the value of combining citizen science, expert field validation, and ecological modeling to refine official range maps and to identify populations at the edge of environmental tolerance, which may serve as early indicators of distributional change [71].

4.2. Divergent Projections Under Climatic Versus Land-Use Scenarios

Future projections under the RCP8.5 scenario revealed a marked methodological divergence between the two modeling frameworks. The full-range climatic model predicted a substantial expansion of suitable climatic space toward 2050, with increases exceeding 60% at the 70% suitability threshold. Such expansion reflects the warming trend that broadens thermally favorable conditions across much of Mesoamerica [72,73,74]. In contrast, when land-use and vegetation variables were included in the Mexican-scale model, the pattern reversed dramatically: suitable habitat collapsed by 96% at the 70% threshold and by nearly 99% at the 80% threshold. This contrast exposes a critical conceptual gap between climatic potential and ecological reality. Although warming climates may theoretically enlarge the species’ fundamental niche, land-use changes can significantly alter the realized distribution by reducing habitat quality, fragmenting landscapes, and eliminating structurally suitable environments. Several studies have highlighted that while climate change may shift ranges toward higher latitudes or altitudes, land use change often causes actual range contraction due to the loss of suitable habitat, especially in agriculturally dominated or modified landscapes [74,75]. Similarly, research on amphibians emphasizes that climate change and land cover transformation jointly drive declines in amphibian distributions, with land use often acting at finer spatial scales to limit persistence even where climatic conditions remain favorable [75,76,77]. Taken together, these findings underscore that climatic potential alone does not guarantee ecological viability and that habitat conversion can severely constrain the realized distribution of species under global change. The ensemble results therefore suggest that H. variolosus may face a growing mismatch between climatic suitability and realized habitat availability, as areas that remain climatically favorable could become increasingly degraded or fragmented by land-use change.

The centroid analysis further supported this interpretation. The inclusion of land cover variables produced a southeastward shift in the centroid of suitable habitat, corresponding to the retreat of viable environments toward the Yucatán Peninsula and southern Veracruz—regions where natural vegetation cover remains relatively continuous. This displacement suggested that while the central and western portions of Mexico may experience climatic protection, anthropogenic disturbance increasingly restricts the species to southeastern refugia. Conversely, the climatic model alone indicated a slight northward shift following the expansion of semiarid conditions under warming scenarios. Together, these patterns emphasize that climatic favorability cannot be equated with ecological persistence [78,79]. Without accounting for habitat integrity, projections risk misrepresenting future species distributions and consequently misguiding conservation priorities.

4.3. Integrating Human Pressures: From Potential to Realized Vulnerability

The integration of human footprint data provided a clearer perspective on how anthropogenic pressure structures the current and future landscapes available to H. variolosus. Approximately 93.8% of the species’ highly suitable habitat (≥70% threshold) occurs under low human pressure (HFP < 20), while 6.0% falls under moderate pressure and only 0.2% within highly transformed areas. This pattern contrasts with that of many other Mexican amphibians, which are often restricted to medium- or high-pressure landscapes [80,81]. The predominance of low-pressure habitats suggests that H. variolosus currently persists in relatively intact regions where human influence is limited. Nonetheless, 6% of habitats under medium pressure may represent zones of emerging risk—transitional areas where agricultural intensification, infrastructure expansion, or urban encroachment could rapidly erode environmental quality. These findings reinforce that even species considered resilient to disturbance may experience cascading declines once thresholds of habitat conversion are exceeded [75].

Moreover, the Global Forest Watch analysis revealed that the Yucatán Peninsula, identified by the climatic model as the main future refugium, had already lost 2.13 million hectares of tree cover between 2001 and 2024—equivalent to 20% of its baseline vegetation. This rate of deforestation, driven primarily by agriculture, cattle ranching, and the development of infrastructure such as the Tren Maya, highlights a paradox: regions predicted to remain climatically stable are simultaneously undergoing rapid ecological degradation [82,83,84]. The combination of human footprint and GFW data thus revealed a spatial and temporal mismatch between potential and realized conservation value. While the peninsula may retain the climatic prerequisites for persistence, the accelerated loss of vegetation and fragmentation could undermine its capacity to function as a long-term refugium. This paradox underscores the necessity of integrating land-use change dynamics into climate-based assessments of future suitability.

4.4. Conservation Implications and Regional Priorities

The dual modeling approach highlights both the ecological plasticity and vulnerability of H. variolosus. Its ability to occupy a range of warm, seasonally dry habitats explains its broad distribution, but the same traits may render it highly sensitive to compounded stressors from climate and land-use change. In central Mexico, newly identified suitable areas within the Trans-Mexican Volcanic Belt (TMVB) may represent key but temporary refugia. The TMVB is among the most heavily transformed biogeographic regions in the country and is characterized by extensive agriculture, urbanization, and industrial development [85]. Our projections indicate that under RCP 8.5, suitability in the State of Mexico—one of the TMVB’s most densely populated regions—will decline by more than 99%, effectively eliminating climatically favorable habitat by 2050. Given that this state currently harbors one of the highest human population densities in Mexico, the probability of successful adaptation or persistence there is extremely low. Even when climatic conditions are tolerable, fragmentation of natural cover and degradation of microhabitats are likely to preclude the maintenance of viable populations [79,80].

From a conservation standpoint, the current dominance of low-pressure environments provides a narrow but important opportunity for proactive protection. Maintaining the ecological integrity of these low-impact regions should be prioritized to prevent them from transitioning into medium- or high-pressure categories. Moreover, medium-pressure areas should be targeted for restoration and sustainable land-use practices to halt the advance of degradation [75,83]. The apparent tolerance of H. variolosus to disturbed environments should not be interpreted as evidence of resilience, as its dependence on specific microhabitats—such as moist soils and temporary ponds—renders it particularly sensitive to hydrological alterations and soil compaction associated with agriculture and livestock. Therefore, conservation strategies must account for both the climatic and microhabitat requirements of fossorial amphibians, emphasizing the protection of soil integrity and ephemeral wetlands.

The divergence between the climatic and integrated models also has implications for amphibian conservation modeling more broadly. Climate-only projections, although informative at large scales, tend to overpredict potential distributions and underrepresent local habitat constraints. In contrast, models that integrate anthropogenic variables yield more conservative but ecologically realistic outcomes [78,79].

In the case of H. variolosus, ignoring land-use change would have led to the false conclusion that the species is likely to expand under warming climates, masking the true extent of vulnerability imposed by deforestation and habitat conversion. Therefore, conservation assessments that rely solely on climatic projections may underestimate extinction risk and misallocate resources. Integrative modeling approaches, such as those implemented here, provide a more balanced framework that can inform dynamic conservation strategies responsive to both climatic and anthropogenic pressures [72,86,87,88].

The apparent paradox of H. variolosus, a species with high climatic persistence yet extreme vulnerability to land-use change, reflects a broader trend observed in tropical and subtropical amphibians. Across Mesoamerica, climatic refugia are increasingly overlapping with areas of rapid deforestation and agricultural intensification [74]. This overlap suggests that conservation planning must move beyond static climatic suitability maps and incorporate real-time land-use trajectories. In the Yucatán Peninsula, for instance, preserving large tracts of native vegetation and maintaining connectivity between forest fragments are essential to ensure that climatically favorable areas remain ecologically functional. Similarly, in central Mexico, habitat restoration in the TMVB could play a crucial role in mitigating the projected contraction of suitable environments.

The integration of H. variolosus distribution models with human footprint and deforestation data offers a template for identifying “latent refugia”—regions where climatic and ecological conditions currently coincide but are at imminent risk of divergence. Protecting these latent refugia before transformation occurs could prevent the irreversible loss of suitable habitats that climatic models alone would designate as secure. This proactive perspective is particularly relevant for species such as H. variolosus, which exhibit both ecological adaptability and physiological sensitivity. The apparent tolerance of anthropogenic landscapes can obscure underlying vulnerabilities that become evident only when environmental thresholds are crossed.

Taken together, our findings indicate that H. variolosus is more vulnerable to global change than its current IUCN classification of least concern suggests. Although most of its present distribution is under low human pressure, the rapid pace of land-use transformation and the projected >95% reduction in suitable habitat under future scenarios justify reassessing its conservation status. Updating the official range map to include central Mexico and the TMVB, incorporating land-use projections into threat assessments, and implementing long-term monitoring programs in both current and emerging habitats are critical steps toward evidence-based reclassification.

Ultimately, the integrative framework presented here that combines climatic, land cover, and anthropogenic variables demonstrates that the future of H. variolosus will depend not solely on its physiological adaptability to warming temperatures but also on the degree to which human societies can reconcile land use practices with biodiversity conservation. Without decisive management interventions, the species’ apparent ecological breadth may conceal a trajectory of silent decline. Proactive habitat protection, sustainable land-use planning, and updated conservation categorization will be essential to ensure that H. variolosus continues to persist across the diverse landscapes that once defined its range.

4.5. Limitations and Future Directions

Our projections should be interpreted cautiously because ecological niche models estimate environmental suitability rather than demographic performance or population persistence. Although ENMs provide valuable insights into potential distributional responses under environmental change, they do not directly incorporate demographic processes or fine-scale ecological mechanisms. Variables associated with breeding sites, hydroperiod, soil properties, prey availability, dispersal constraints, and species interactions were not included and may influence the realized distribution of H. variolosus. Future studies integrating demographic data, field validation, and microhabitat characteristics would improve predictions and provide a more comprehensive understanding of the species’ vulnerability under climate and land-use change.

5. Conclusions

This study provides an integrated assessment of the vulnerability of Hypopachus variolosus under combined climatic and land-use change scenarios. While broad-scale climatic models suggest a potential expansion of suitable environmental conditions, the incorporation of land-use and vegetation variables reveals a contrasting and more realistic outcome: a drastic contraction of suitable habitat, particularly within Mexico. This divergence highlights that climatic suitability alone is insufficient to predict species persistence, as anthropogenic habitat transformation strongly constrains the realized distribution.

Our results indicate that although the species currently occupies predominantly low human-impact areas, ongoing deforestation and land-use intensification—especially in the Yucatán Peninsula—are likely to undermine these potential refugia. The identification of new suitable areas in central Mexico further suggests that the current IUCN distribution underestimates the species’ range, but these regions may represent transient habitats under increasing anthropogenic pressure.

Overall, H. variolosus appears more vulnerable to global change than previously recognized. Effective conservation strategies should prioritize the protection of low-impact habitats, restoration of degraded landscapes, and the integration of land-use projections into future assessments. Updating the species’ distribution and conservation status will be essential to ensure its long-term persistence across Mesoamerica.