The ongoing biodiversity crisis, fueled by climate change and human activities like land conversion, is causing a sixth mass extinction. Madagascar, with its diverse fauna, is severely impacted. Predicting the future of its biodiversity is challenging due to factors like the expanding network of protected areas, non-additive effects of climate change and land use conversion, and the difficulty of modeling rare species. Many species, including chameleons, have narrow geographic ranges, making traditional species distribution models (SDMs) inadequate. The 'rare species modeling paradox' arises from the need for numerous data points for accurate modeling, which are often unavailable for rare species. To address this, the researchers utilized ENphylo, an algorithm that leverages phylogenetic signals in species' climatic preferences to infer the climatic niche of rare species, and then generates habitat suitability models using limited observational data. This study expanded ENphylo to model species with as few as two data points, making it suitable for analyzing Madagascar's diverse but often rare chameleon fauna. The study aimed to assess the combined effects of future climate and land-use change on Madagascar's chameleon diversity by modeling their potential future distributions under various scenarios of global warming and land-use change. The study examined how these factors interact to influence chameleon survival and distribution, considering additive, antagonistic, and synergistic interactions.

The introduction extensively cites existing research highlighting the global biodiversity crisis, the severity of the crisis in Madagascar, the challenges of modeling rare species, and previous attempts to address these challenges. The literature review emphasizes the combined and often non-additive effects of climate change and human land-use change on biodiversity. It highlights the limitations of existing SDM techniques when applied to species with limited occurrence data and the need for new approaches like ENphylo. Specific studies on Madagascar's biodiversity, deforestation rates, and the impact of protected areas are also referenced.

The study utilized occurrence data for 151 Chamaeleonidae species downloaded from the Global Biodiversity Information Facility (GBIF). Data were filtered to ensure accuracy and remove duplicates. Nineteen bioclimatic variables from the CHELSA database were used as environmental predictors, considering both current (1981–2010) and future (2071–2100) data under mild (SSP1-2.6) and severe (SSP5-8.5) Shared Socioeconomic Pathways (SSPs) scenarios, with three Global Circulation Models (GCMs). Land Use/Land Cover (LULC) data were also incorporated. Three different SDM approaches were employed depending on the number of occurrences per species: ENphylo for species with <15 occurrences, Ensemble of Small Models (ESMs) for species with 15-30 occurrences, and traditional SDMs (MaxEnt, RF, GLM) for species with >30 occurrences. ENphylo was expanded to handle species with as few as two data points, using pseudo-presences generated based on proximity to known occurrences and climatic/LULC similarity. Phylogenetic imputation of niche characteristics was done using a chameleon supertree. Model accuracy was assessed using AUC, TSS, and Boyce index. Future projections were created under three scenarios: dynamic climate (constant LULC), dynamic LULC (constant climate), and dynamic LULC and climate. Binarization of the maps used three thresholding methods (SensSpec, MaxSens+Spec, TenPerc). A dispersal constraint of 1 km/year was applied. Species richness, loss, and gain indices were calculated. Finally, interaction types (synergistic, additive, antagonistic, only climate, only LULC) were determined by comparing the changes across the three future scenarios.

The study modeled 134 chameleon species, including 56 from Madagascar. ENphylo performed well for rare species (AUC > 0.9), while ESMs and traditional SDMs also showed robust performances. Future projections revealed significant habitat loss and gain across Madagascar. Land use change (LULC) had the most substantial impact, exceeding the effects of climate change and their interactions, regardless of the climate change scenario or SDM method. The 'only LULC' effect had the highest percentage for both gain and loss. Antagonistic interactions were relevant for species loss only. Highest biodiversity loss is predicted in western and northwestern dry deciduous forests, while most species turnover is in eastern lowland forests. The study found that approximately 1 chameleon species per grid cell in Madagascar may experience unsuitable habitats in the future. The maximum potential loss ranged from 8 (mild scenario) to 11 (severe scenario) species, primarily due to LULC changes. For some 10-20 species, a >90% habitat loss is predicted, putting them at high risk of extinction. While some widespread species are projected to gain range, these gains are considered hypothetical due to chameleons' limited dispersal abilities and potential lack of habitat connectivity. There was no clear relationship between species commonness and habitat loss, especially in dry deciduous forests.

The findings strongly suggest that land conversion is the most significant threat to Madagascar's chameleons, outweighing the effects of climate change. The high predictive power of the ENphylo model, even with limited data, demonstrates its effectiveness in assessing the vulnerability of rare species. The predicted habitat loss in specific regions highlights the importance of targeted conservation efforts. The hypothetical nature of the predicted range gains underscores the need for considering habitat connectivity and dispersal limitations in conservation planning. The study's focus on multiple scenarios and the incorporation of both climate and LULC changes provide a more comprehensive understanding of the threats to chameleon biodiversity than previous studies. The results highlight the need for integrated conservation strategies addressing both habitat loss and climate change.

This study demonstrates the significant threat posed by land conversion to Madagascar's chameleon biodiversity. Land use change overrides climate change effects in shaping future chameleon distributions. The expanded ENphylo method effectively models rare species, enabling better assessments of extinction risk. Future research could focus on refining habitat suitability models by incorporating finer-scale data and investigating the effectiveness of specific conservation interventions.

The study relies on occurrence data from GBIF, which might have biases. The dispersal model uses a simplified approach with a fixed dispersal rate, neglecting potential variations across species and environmental conditions. The model does not fully account for potential human-mediated translocation or other factors impacting chameleon populations, such as disease or predation.