Global patterns of climate change impacts on desert bird communities

Climate change is causing major shifts in species distributions and abundances, yet the effects on warm desert ecosystems have been understudied relative to polar regions or a few well-studied countries. Warm deserts host unexpectedly high biodiversity, have already experienced substantial climate change impacts, are projected to warm considerably, and many of their species live near physiological limits, making them especially vulnerable to additional warming. Simply mapping changes in air temperature can misrepresent biological impacts because organisms experience microclimates and can buffer or amplify heat via behavior and physiology. The study therefore aims to integrate microclimate modeling with physiologically explicit biophysical models and biodiversity data to address four questions: (a) How will the world’s warm deserts be affected by climate change, and do projected impacts vary between and within major desert realms? (b) Do physiological models (based on TEWL and ADR) produce spatially different results than models based solely on air temperature (Tair)? (c) Which areas are likely to serve as climate change refugia for desert birds? (d) To what extent do these refugia overlap existing protected areas? Birds are the focal taxon due to their diurnality, limited access to thermally buffered microsites, and high mass-specific evaporative water loss rates, which together imply heightened sensitivity to extreme heat and dehydration.

Prior work shows global biodiversity responses to climate change but has paid limited attention to warm deserts. Existing studies often focus on polar deserts or select countries. Deserts support high biodiversity and endemism, have already experienced disproportionate climate impacts, and many desert species operate near physiological thresholds. Microclimate exposure and behavioral/physiological responses can decouple organismal thermal stress from coarse air temperature measures, motivating the use of physiologically explicit models. For birds, evaporative water loss and dehydration risk are key determinants of survival under chronic warming and extreme heat waves. Documented declines in avian abundance in some desert regions have been linked to climate change, underscoring the need for physiology-informed projections. The study builds on these insights by employing TEWL and ADR as physiological metrics and evaluating how they differ from Tair-based approaches for identifying refugia.

Study extent and deserts: Warm deserts were mapped at 50 km resolution by selecting desert-related IUCN habitat classes (hot and temperate desert, subtropical/tropical dry shrubland, dry lowland grassland, and dry savanna) from a global habitat map and restricting to areas with <500 mm annual precipitation (WorldClim v2.1, 1970–2000). Deserts were grouped into six zoogeographic realms using an updated Wallacean regionalization. Climate and microclimate data: Historical and projected monthly climate data were from TerraClimate (50 km; maximum/minimum temperature, precipitation, soil moisture, vapor pressure, shortwave radiation, wind). Two future scenarios were modeled: global mean warming of +2 °C (main) and +4 °C (supplementary) above pre-industrial, derived from 23 CMIP5 GCMs via pattern scaling (Qin et al. 2020) and downscaled. NicheMapR’s micro-terra function disaggregated monthly data to hourly, applying elevation/terrain corrections, daily spline interpolation, and diurnal cycles to generate hourly air temperature, relative humidity (from vapor pressure and diurnal T cycles), and radiation (via clear-sky fraction). The typically hottest month (July in the Northern Hemisphere, January in the Southern Hemisphere) was analyzed. Model species and physiology: Three model birds represented body-size classes of desert birds: 13 g (small, 0–33rd percentile), 39 g (medium; based on Cactus Wren Campylorhynchus brunneicapillus), and 185 g (large, 66–100th percentile). Size-related traits (plumage depths, feather lengths) scaled with mass^1/3. Other parameters came from the Cactus Wren and literature or best estimates. The customized NicheMapR endoR_devel model computed heat and mass exchange, minimal metabolic thermoregulation, and hourly evaporative water loss (cutaneous + respiratory). Thermoregulatory sequence under heat stress was modified to: reduce ptiloerection, stretch, increase flesh conductivity, then simultaneously increase core temperature (to 44 °C) and respiratory rate (up to 7.5× resting). Birds were assumed perched 1.5 m above ground and behaviorally shifted between open (0% shade) and shade (90%) each hour to minimize water loss; a no-shade scenario was also tested. Physiological metrics and impacts: Total evaporative water loss (TEWL) was computed as total daily water loss during the average day of the hottest month; acute dehydration risk (ADR) as the maximum 3-hour cumulative water loss as % of body mass during that day, with a lethal threshold near 15% within 3 hours. For each pixel, TEWL and ADR were averaged across the three size classes weighted by the number of species per size class in the local community. Quantifying climate change impacts: Overlap between current (1986–2015) and future (pseudo 1986–2015 consistent with +2 °C) distributions of TEWL and ADR was computed via kernel density overlap; lower overlap indicates higher impact. Parallel overlaps and mean changes were also calculated for air temperature (Tair). Kruskal–Wallis tests compared distributions among realms; Pearson correlations assessed relationships among TEWL/ADR/Tair impacts; spatial similarity was assessed via Weighted Jaccard Index. Bird diversity and refugia: Desert birds were defined as species with ≥90% of Area of Habitat (AOH) within warm deserts (n=152). Rarity-weighted richness (RWR) per grid cell was computed as the sum of inverse AOH for species present. Bivariate heatmaps placed each pixel by percentile ranks of physiological overlap (TEWL or ADR) and RWR within each realm. Refugia were defined as pixels with high diversity and low physiological impact. Two approaches: (1) fixed threshold at 75th percentile for diversity and physiological overlap; (2) floating threshold adjusted so that at least 5% of each realm’s area qualified as refugia, assuming each realm merits baseline protection irrespective of relative impact levels. Protected areas: Refugia were overlaid with the World Database on Protected Areas (IUCN categories Ia–IV) to estimate current PA coverage. Validation and sensitivity: Model predictions of body temperature and evaporative water loss across air temperatures were validated against laboratory data from nine bird species (four orders, nine families), showing good agreement (slight overestimation for some non-passerines due to using a passerine BMR equation; sensitivity analyses indicated minimal effect on safe-site identification). Historical validation related model-predicted changes in TEWL and ADR (1911–1940 vs 1971–2000) at a Mojave site to observed occupancy changes for 50 species, finding significant negative correlations (p=0.028 for TEWL; p=0.031 for ADR). Sensitivity analyses varied physiological and behavioral parameters (e.g., sitting height, basal heat generation for non-passerines, panting onset, parameter combinations minimizing/maximizing water loss) and found that ≥69.3% (often >90%) of safe sites were consistently identified.

• Climate change impacts on desert birds are highly heterogeneous between and within warm desert realms under +2 °C warming. • The Saharo–Arabian realm exhibits the largest changes in mean Tair and TEWL; ADR changes are of similar magnitude among realms. The Saharo–Arabian realm also shows the smallest current–future overlap (i.e., largest impact) for Tair, TEWL, and ADR (p<0.001). • Physiological-impact distributions (TEWL/ADR) are more positively skewed than Tair-only impacts, with more extreme high-impact pixels (Supplementary Fig. 4). • Spatial patterns of physiological impacts do not mirror air temperature changes; agreement is limited (Weighted Jaccard Index ≤75.5%). Using Tair as a proxy leads to both under-protection and over-protection of refugia, each up to ~60% of predicted refugia area in some realms. • Refugia (high diversity, low impact) are predicted in all realms, often concentrated near coastlines. With a fixed 75th-percentile threshold, the Australian realm had the largest refugia proportion (6.6% of its desert area) while the Neotropical realm had the smallest (1.7%). Using floating thresholds ensured at least 5% refugia per realm, requiring stricter thresholds (>75%) in Australian, Afrotropical, and Saharo–Arabian realms. • Existing protection of refugia is generally low: typically <20% of refugial areas fall within current protected areas, with the Neotropical realm having the lowest PA coverage among predicted refugia.

Integrating biodiversity patterns with physiological exposure offers a more realistic assessment of climate risks to desert birds than air temperature alone. The modeling indicates substantial heterogeneity of impacts across and within warm deserts, with areas of concentrated high risk and distinct refugia. Physiological models based on microclimate and organismal water balance produce spatially different impact maps than Tair-only approaches; relying on Tair would misidentify many refugia and allocate protection inefficiently, leading to both under- and over-protection. The concentration of predicted refugia near coastlines may reflect oceanic buffering of terrestrial warming, yet these areas face threats from sea-level rise and human disturbance, potentially eroding their refugial value. Conservation implications include prioritizing the expansion or designation of protected areas to encompass refugia, considering realm-specific thresholds or minimum area targets (e.g., at least 5% per realm). Even outside refugia, high-diversity areas projected to experience large physiological impacts warrant efforts to reduce additional anthropogenic stressors (e.g., land-use change). Findings are robust across more severe scenarios (+4 °C) and under pessimistic behavioral assumptions (no access to shade), underscoring their relevance for climate adaptation planning.

The study provides a physiologically grounded, microclimate-informed framework for assessing climate change impacts on desert bird communities and identifying climate refugia across the world’s warm deserts. It demonstrates that physiological metrics (TEWL, ADR) yield markedly different and more conservation-relevant patterns than air temperature alone, revealing heterogeneous impacts, coastal-skewed refugia, and substantial protection gaps (<20% PA coverage). These insights argue for prioritizing refugia in conservation planning, potentially within global targets such as protecting 30% of land by 2030, and for refining protection to avoid under- and over-protection when using Tair proxies. Future work should extend the framework to other taxa to test overlap of refugia, incorporate additional ecological traits and behaviors, improve parameterization for non-passerines, consider finer spatial resolution and non-summer stressors, and further refine models of microclimate and water balance to enhance predictive performance.

• Taxonomic scope: Analyses focus on birds; refugial patterns for other taxa may differ and need dedicated modeling. • Model species approach: Three size-based model species approximate community physiology; individual species’ traits and behaviors vary and could shift local risk rankings, though sensitivity analyses suggest relative patterns are robust. • Physiological parameterization: A passerine BMR equation was applied broadly, slightly overestimating some non-passerines; shade availability and behavioral thermoregulation were assumed (with an open-habitat sensitivity case), and birds were modeled at 1.5 m perch height. • Temporal and climatic scope: Impacts were assessed for the hottest month and +2 °C (main) and +4 °C (supplementary) scenarios; other seasons, extreme events beyond modeled diurnal cycles, and hydrological constraints outside the hottest month were not explicitly evaluated. • Spatial resolution and inputs: Core analyses at 50 km resolution may miss fine-scale microrefugia; pattern-scaled climate projections and downscaling choices introduce uncertainty. • Conservation overlays: PA coverage considered IUCN categories Ia–IV; management effectiveness, enforcement, and future land-use changes were not assessed. Coastal refugia may be compromised by sea-level rise and human pressures not explicitly modeled.