Data from: Maps and additional figures for climate change refugia hotspots for priority species: A case study in East Africa
Data files
Dec 11, 2025 version files 40.59 KB
-
README.md
24.29 KB
-
SampleTable_ITIS_Habitat.csv
16.30 KB
Abstract
Natural resource managers and policymakers need actionable climate data to guide conservation decisions. Conserving climate change refugia, areas relatively buffered from contemporary climate change, is increasingly considered an effective strategy for adaptation. Despite tropical species facing heightened vulnerability to climate change, the tropics remain underserved in climate adaptation research. We coproduced the first comprehensive assessment of climate change refugia across Tanzania with Tanzanian partners through extensive consultation, in-person conversations, and field visits to priority ecosystems, ensuring our analysis addressed local conservation needs and decision-making contexts. We developed species distribution models for 33 terrestrial animal species using maximum entropy and boosted regression tree algorithms. We projected future suitable habitats for SSP126 and SSP585, for 2011-2040 and 2071-2100, using GFDL Earth System and the U.K. Earth System models. More than half under SSP126 and 79% of the focal species under SSP585 lost their suitable habitat by 2100. Serengeti National Park, Northern Highlands Forest Reserve, and the Eastern Arc Mountains emerged as key climate change refugia, while other protected areas, including Kigosi and Ugalla River National Parks, had no climate change refugia. This assessment provides actionable insights for Tanzania’s conservation prioritization while identifying critical research gaps in western and montane ecosystems.
Dataset DOI: 10.5061/dryad.w3r228151
Description of the data and file structure
Species information
We acquired species occurrence data (Table S2) from the Global Biodiversity Information Facility (GBIF, www.gbif.org; 2024) by searching for occurrences using the scientific name and crosschecking historical and alternative names against Mammal Species of the World, 3rd edition (Wilson & Reeder, 2005) and the Integrated Taxonomic Information System (https://www.itis.gov/). TAWIRI contributed a dataset of mammal occurrences from their previous work, including data from camera traps and opportunistic encounters. Our analysis included 20,265 occurrence records across study species. (more details here SampleTable_ITIS_Habitat.csv)
Column header description:
IUCN_Common_name: The commonly used name of the species as listed by the IUCN (e.g., Cheetah, Lion).
Species: The scientific binomial name of the species (genus and species), e.g., Acinonyx jubatus.
ITIS taxonomy: The taxonomic name from ITIS (Integrated Taxonomic Information System), confirming species identification.
Gbif Data: Indicates if occurrence data for this species is available from GBIF (marked with an "x").
TAWIRI Data: Indicates if occurrence data is available from TAWIRI (marked with an "x").
Sample_size_cleaned: The total cleaned sample size or number of data records available for this species.
TAWIRI_SampleSize: Number of sample records specifically from TAWIRI.
GBIF_DIO: Citation or source of the GBIF data download, including date and DOI link.
Taxa_1: Broad taxonomic group the species belongs to (e.g., Mammal).
Taxa_2: More specific ecological or dietary classification (e.g., Carnivore, Herbivore, Primate).
Family: Taxonomic family of the species (e.g., Felidae, Bovidae).
Habitat_1: Primary or general habitat classification (e.g., Savanna, Forest).
IUCN_Habitat_and_ecology_notes: Detailed habitat description and ecological notes from the IUCN Red List, describing environments where the species occurs.
Range_Map_Citation: Full citation for the IUCN Red List species assessment, providing the official source for range and ecological information, often including a DOI link.
Climate data
Dossier_DD_1_Climate_Historical_Future_PAs_AllBios_SSP126_and_SSP585_20Oct2025_Complete.pdf
We obtained climate data from CHELSA version V2.1 (Climatologies at High resolution for the Earth’s Land Surface Areas), a downscaled model output temperature and precipitation estimates at a horizontal resolution of 30 arcsec for the present and future time periods under CMIP6 (Karger et al., 2017) (Table S3). We selected two atmospheric-ocean global circulation models: the U.K. Earth System Model (UKESM1-0-LL), which has higher climate sensitivity, indicating greater temperature response to CO2 changes and thus stronger climate change impacts (than other General Circulation Model (GCM) in the CHELSA dataset ( Sellar et al., 2019; Hausfather et al., 2022); and the GFDL Earth System Model Version 4.1 (GFDL-ESM 4.1), which has lower climate sensitivity (Dunne et al., 2020). We chose to bracket future climate change by selecting two future scenarios of climate change: SSP126, a low emissions pathway that projects future warming below 2°C and peak emissions before 2050; and SSP585, a high emissions pathway with the highest warming and continuous emission growth for mid-century 2040 and late-century 2100 (IPCC, 2023) (Figure 1, Figure S4 and DD1).
Environmental data
We selected 9 environmental variables known to influence target species distributions: crop probability (Pittman et al., 2010), soil type (FAO/UNESCO, 2003), contrast (a measure of habitat heterogeneity) (Tuanmu et al., 2015), stratified forest cover (Tyukavina et al., 2015), human impact (JRC & CIESIN, 2021), population density (CIESIN, 2018), terrain ruggedness (Sappington et al., 2007), slope (U.S. Geological Survey, 1996), and aspect transformed to “northness” and “eastness’ (U.S. Geological Survey, 1996) (Table S3). We converted categorical environmental variables to factors.
Modeling methods
Since the data lacked true absences, background sites were created using each species’ study area (Barbet-Massin et al., 2012; Liu et al., 2019). We randomly sampled 10,000 background sites from this study area and extracted environmental and climate data (Barbet-Massin et al., 2012; Liu et al., 2019).
We employed both Maximum Entropy (MaxEnt) and Boosted Regression Tree (BRT) modeling through the ‘enmSdmX’ package (Smith et al., 2023). Both models were trained using occurrence data and background sites as the response variable and the selected variables as predictors.
We modeled current habitat suitability by projecting MaxEnt and BRT models using CHELSA climate data for the baseline period (1981-2010). For future projections, we implemented these models under two contrasting emissions scenarios (SSP126 and SSP585) for both mid-century (2011-2040) and late-century (2071-2100) time periods, using climate projections from the GFDL Earth System Model Version 4.1 and U.K. Earth System Model. These projections provided habitat suitability estimates across all combinations of scenarios, timeframes, and climate models. Current values of non-climate predictors were used as projections were not available. We established thresholds using the Maximum Sum of Sensitivity and Specificity (Max SSS) method to convert continuous suitability predictions into binary presence-absence maps, as recommended by Liu et al. (2013).
Climate change refugia identification
- TIFs_Individual_species_in_situ_and_ex_situ_climate_change_refugia.zip
- TIFs_Individual_species_in_situ_climate_change_refugia_model_agreement.zip
- Dossier_DD_2__Individual_Species_In_situ__and_Ex_situ__Climate_Change_Refugia_n_33_20Oct2025_Complete.pdf
- Dossier_DD_3_Species_In_situ_Climate_Change_Refugia_Model_Agreement_map_n_33_20Oct2025_Complete.pdf
We identified two types of climate change refugia, in situ climate change refugia (in situ CCR) are restricted to locations that remain suitable for a species and ex situ climate change refugia (ex situ CCR) are found in previously unsuitable locations (Figure 2 B and DD 2).
Climate change refugia were calculated by overlaying current and future binary habitat suitability maps, where cells were classified as in situ CCR if they had suitable habitat in the current and future time periods and as ex situ CCR if they were not suitable habitat in current but contained suitable habitat in future. We also calculated loss of suitable habitat, cells that contained suitable habitat in the current but not future time period.
To compare habitat changes across climate change refugia scenarios and time periods, we calculated the percent of current suitable habitat projected to be lost under future climate conditions by dividing the number of cells classified as habitat loss by the total number of current suitable habitat cells (Table 3). We calculated the percent of current suitable habitat in in situ CCR and ex situ CCR using the same method (Table 3).
TIFs_Multi-species_in_situ_climate_change_refugia_hotspots_for_33_species.zip
We created individual in situ CCR maps for each unique combination of species, time period, General Circulation Model (GCM), Shared Socioeconomic Pathway (SSP), and statistical model, resulting in 528 distinct maps (16 variations/species for 33 species)(Figure 2 A). To assess the robustness of refugia predictions, we developed time period and SSP in situ CCR maps for each species by combining individual in situ CCR maps across modeling approaches. These composite maps integrated predictions from both GCMs (GFDL-ESM4 and UKESM1-0-LL) and statistical models (BRT and MaxEnt) for each time period (2011-2040 or 2071-2100) and scenario (SSP 126 or SSP 585). Cells were classified as refugia only when identified consistently across all four model combinations, generating 132 consolidated maps (4 maps/species). To identify areas of conservation significance across multiple species, we developed multi-species Climate Change Refugia “Hotspot” maps by aggregating the Year and SSP in situ climate change refugia maps across all 33 species. The resulting maps quantify the number of species for which each cell represents in situ CCR, with values ranging from 0 (no species have in situ CCR) to 33 (all study species have in situ CCR).
Code/software
You can view the raster files in R using library terra and the rast function.
Access information
Other publicly accessible locations of the data:
- None
Data was derived from the following sources:
| Comman name | Scientifc name | Source |
|---|---|---|
| IUCN_Common_name | Species | GBIF_DIO |
| Cheetah | Acinonyx jubatus | GBIF.org (16 April 2024) GBIF Occurrence Download https://doi.org/10.15468/dl.qazh4q |
| Lion | Panthera leo | GBIF.org (21 April 2024) GBIF Occurrence Download https://doi.org/10.15468/dl.xck4jz |
| Spotted hyena | Crocuta crocuta | GBIF.org (21 April 2024) GBIF Occurrence Download https://doi.org/10.15468/dl.ssvsf9 |
| African wild dog | Lycaon pictus | GBIF.org (09 May 2024) GBIF Occurrence Download https://doi.org/10.15468/dl.vjk85u |
| Giraffe | Giraffa camelopardalis | GBIF.org (13 December 2023) GBIF Occurrence Download https://doi.org/10.15468/dl.c8veb6 |
| Impala | Aepyceros melampus | GBIF.org (08 January 2024) GBIF Occurrence Download https://doi.org/10.15468/dl.qfdrut |
| African buffalo | Syncerus caffer | GBIF.org (21 April 2024) GBIF Occurrence Download https://doi.org/10.15468/dl.6sfxv5 |
| Elephant | Loxodonta africana | GBIF.org (21 April 2024) GBIF Occurrence Download https://doi.org/10.15468/dl.ey9buz |
| Olive baboon | Papio anubis | GBIF.org (10 January 2024) GBIF Occurrence Download https://doi.org/10.15468/dl.2r4h7u |
| Blue monkey | Cercopithecus mitis | GBIF.org (10 January 2024) GBIF Occurrence Download https://doi.org/10.15468/dl.357e83 |
| Angolan colobus | Colobus angolensis | GBIF.org (01 May 2024) GBIF Occurrence Download https://doi.org/10.15468/dl.r4cawm |
| African straw-coloured fruit-bat | Eidolon helvum | GBIF.org (01 May 2024) GBIF Occurrence Download https://doi.org/10.15468/dl.kutxtw |
| Guttural toad | Sclerophrys gutturalis | GBIF.org (09 May 2024) GBIF Occurrence Download https://doi.org/10.15468/dl.f5ewzh |
| Dune squeaker | Arthroleptis stenodactylus | GBIF.org (09 May 2024) GBIF Occurrence Download https://doi.org/10.15468/dl.c3vew7 |
| Nile monitor | Varanus niloticus | GBIF.org (09 May 2024) GBIF Occurrence Download https://doi.org/10.15468/dl.wbvn66 |
| Common African flap-necked chameleon | Chamaeleo dilepis | GBIF.org (09 May 2024) GBIF Occurrence Download https://doi.org/10.15468/dl.yy5q9f |
| Mwanza flat-headed rock agama | Agama mwanzae | GBIF.org (09 May 2024) GBIF Occurrence Download https://doi.org/10.15468/dl.nr3uf5 |
| Striped skink | Trachylepis striata | GBIF.org (09 May 2024) GBIF Occurrence Download https://doi.org/10.15468/dl.ap844k |
| Puff adder | Bitis arietans | GBIF.org (09 May 2024) GBIF Occurrence Download https://doi.org/10.15468/dl.rzj7u7 |
| Leopard tortoise | Stigmochelys pardalis | GBIF.org (09 May 2024) GBIF Occurrence Download https://doi.org/10.15468/dl.evf8z6 |
| Hamerkop | Scopus umbretta | GBIF.org (01 May 2024) GBIF Occurrence Download https://doi.org/10.15468/dl.d5r5ux |
| African wood-owl | Strix woodfordii | GBIF.org (01 May 2024) GBIF Occurrence Download https://doi.org/10.15468/dl.5wfymm |
| Silvery-cheeked hornbill | Bycanistes brevis | GBIF.org (01 May 2024) GBIF Occurrence Download https://doi.org/10.15468/dl.8s3797 |
| Evergreen-forest warbler | Bradypterus lopezi | GBIF.org (01 May 2024) GBIF Occurrence Download https://doi.org/10.15468/dl.2r2ha3 |
| Palm-nut vulture | Gypohierax angolensis | GBIF.org (10 May 2024) GBIF Occurrence Download https://doi.org/10.15468/dl.tjvuvt |
| Livingstone's turaco | Tauraco livingstonii | GBIF.org (01 May 2024) GBIF Occurrence Download https://doi.org/10.15468/dl.9r3cwk |
| Common ostrich | Struthio camelus | GBIF.org (10 May 2024) GBIF Occurrence Download https://doi.org/10.15468/dl.mrze9z |
| African fish-eagle | Haliaeetus vocifer | GBIF.org (13 May 2024) GBIF Occurrence Download https://doi.org/10.15468/dl.2ptyu4 |
| White-backed vulture | Gyps africanus | GBIF.org (13 May 2024) GBIF Occurrence Download https://doi.org/10.15468/dl.wj5k4b |
| Secretarybird | Sagittarius serpentarius | GBIF.org (13 May 2024) GBIF Occurrence Download https://doi.org/10.15468/dl.7mb66k |
| White-headed buffalo-weaver | Dinemellia dinemelli | GBIF.org (13 May 2024) GBIF Occurrence Download https://doi.org/10.15468/dl.nezdj9 |
| Red-necked francolin | Pternistis afer | GBIF.org (14 May 2024) GBIF Occurrence Download https://doi.org/10.15468/dl.6rqs8t |
| Yellow-necked spurfowl | Pternistis leucoscepus | GBIF.org (13 May 2024) GBIF Occurrence Download https://doi.org/10.15468/dl.f9apya |
1. Aspect "northness"
- Description: The compass direction or azimuth that a terrain surface faces, derived from a digital elevation map. Aspect was calculated using Arc GIS Pro 3.1.1. As aspect is a circular variable and to contrast northern versus southern exposition, it was transformed using sine "northness" (1 = north, -1 = south).
- Spatial precision: 1 km
- Hypothesized ecological role: Influences solar radiation and microclimate effects.
2. Aspect "eastness"
- Description: The compass direction or azimuth that a terrain surface faces, derived from a digital elevation map. Aspect was calculated using Arc GIS Pro 3.1.1. To contrast eastern versus western exposition, it was transformed using cosine "eastness" (1 = east, -1 = west).
- Spatial precision: 1 km
- Hypothesized ecological role: Influences solar radiation and microclimate effects.
3. Ruggedness
- Description: Estimate of evenness and steepness of terrain derived from a digital elevation map using Arc GIS Pro 3.1.1.
- Spatial precision: 1 km
- Hypothesized ecological role: Estimates traversal difficulty, affecting species movement, habitat suitability, and human development patterns.
4. Slope
- Description: Percent change in elevation over a certain distance derived from a digital elevation map (GTOPO30) using Arc GIS Pro 3.1.1.
- Spatial precision: 1 km
- Hypothesized ecological role: Influences species movement, precipitation dynamics, vegetation communities, soil development, and erosion.
5. Crop probability
- Description: Probability that a pixel is production cropland; values range from 1 to 100.
- Spatial precision: 1 km
- Hypothesized ecological role: Indicates human disturbance, simplified habitat structure, altered disturbance regimes, and edge effects.
6. Soil
- Description: Digital Soil Map of the World from the FAO and UNESCO.
- Spatial precision: 1 km
- Hypothesized ecological role: Affects primary productivity, water dynamics, and habitat structure.
7. Contrast
- Description: Measures heterogeneity by exponentially weighted differences in Enhanced Vegetation Index (EVI) between adjacent pixels.
- Spatial precision: 1 km
- Hypothesized ecological role: Spatial variation in biotic and abiotic conditions influences species and community diversity, disturbance, edge effects, and species movement.
8. Pan Tropical Forest Strata
- Description: Forest strata classification (1 = low cover, 2 = medium short, 3 = medium tall, 4 = dense short, 5 = dense short intact, 6 = dense tall, 7 = dense tall intact). A measure of landcover.
- Spatial precision: 1 km
- Hypothesized ecological role: Influences species movement, connectivity, microclimate effects, and habitat complexity.
9. Human impact
- Description: Population (POP), Built-Up Estimates (BUILT), and Degree of Urbanization Settlement Model Grid (SMOD), v1 (2014).
- Spatial precision: 1 km
- Hypothesized ecological role: Reflects habitat conversion and fragmentation, edge effects, connectivity, altered disturbance regimes, competition with livestock or invasive species, hunting pressure.
10. Population Density
- Description: UN WPP-Adjusted Population Density, v4.11 (2015).
- Spatial precision: 1 km
- Hypothesized ecological role: Habitat conversion and fragmentation, edge effects, connectivity, altered disturbance regimes, competition with livestock and invasive species.
11. Elevation
- Description: Global digital elevation model (DEM) with a horizontal grid spacing of 30 arc seconds (~1 km).
- Spatial precision: 1 km
- Hypothesized ecological role: Influences temperature gradients, precipitation patterns, growing season length, solar radiation, and microclimate effects.
12. bio1: Mean annual air temperature
- Description: Mean annual daily air temperature averaged over one year (°C).
- Spatial precision: 1 km
- Hypothesized ecological role: Thermal stress and stress on vegetation.
13. bio2: Mean diurnal air temperature range
- Description: Mean diurnal range of temperatures averaged over one year (°C).
- Spatial precision: 1 km
- Hypothesized ecological role: Average daily thermal experience.
14. bio3: Isothermality
- Description: Ratio of diurnal variation to annual variation in temperatures (°C).
- Spatial precision**:** 1 km
- Hypothesized ecological role: Thermal variation experience over the year.
15. bio4: Temperature seasonality
- Description: Standard deviation of the monthly mean temperatures (°C/100).
- Spatial precision: 1 km
- Hypothesized ecological role: Seasonal thermal variation experience.
16. bio5: Mean daily maximum air temperature of the warmest month
- Description: Highest monthly daily mean maximum temperature (°C).
- Spatial precision: 1 km
- Hypothesized ecological role: Thermal stress and stress on vegetation.
17. bio6: Mean daily minimum air temperature of the coldest month
- Description: Lowest monthly daily mean minimum temperature (°C).
- Spatial precision: 1 km
- Hypothesized ecological role: Thermal stress, stress on vegetation, thresholds for overwintering insects.
18. bio7: Annual range of air temperature
- Description: Difference between the maximum temperature of the warmest month and the minimum of the coldest month (°C).
- Spatial precision: 1 km
- Hypothesized ecological role: Range of thermal experience.
19. bio8: Mean daily mean air temperatures of the wettest quarter
- Description: Average temperature of the wettest quarter (°C).
- Spatial precision: 1 km
- Hypothesized ecological role: Thermal stress and stress on vegetation.
20. bio9: Mean daily mean air temperatures of the driest quarter
- Description: Average temperature of the driest quarter (°C).
- Spatial precision: 1 km
- Hypothesized ecological role: Thermal stress and stress on vegetation.
21. bio10: Mean daily mean air temperatures of the warmest quarter
- Description: Average temperature of the warmest quarter (°C).
- Spatial precision: 1 km
- Hypothesized ecological role: Thermal stress and stress on vegetation.
22. bio11: Mean daily mean air temperatures of the coldest quarter
- Description: Average temperature of the coldest quarter (°C).
- Spatial precision: 1 km
- Hypothesized ecological role: Thermal stress, stress on vegetation, thresholds for overwintering insects.
23. bio12: Annual precipitation amount
- Description: Total precipitation over one year (kg m⁻² year⁻¹).
- Spatial precision: 1 km
- Hypothesized ecological role: Forage, prey, and water availability.
24. bio13: Precipitation amount of the wettest month
- Description: Precipitation of the wettest month (kg m⁻² month⁻¹).
- Spatial precision: 1 km
- Hypothesized ecological role: Forage, prey, and water availability.
25. bio14: Precipitation amount of the driest month
- Description: Precipitation of the driest month (kg m⁻² month⁻¹).
- Spatial precision: 1 km
- Hypothesized ecological role: Forage, prey, and water availability.
26. bio15: Precipitation seasonality
- Description: Coefficient of variation of monthly precipitation estimates expressed as a percentage of the annual mean (kg m⁻²).
- Spatial precision: 1 km
- Hypothesized ecological role: Forage, prey, and water availability.
27. bio16: Mean monthly precipitation amount of the wettest quarter
- Description: Average precipitation of the wettest quarter (kg m⁻² month⁻¹).
- Spatial precision: 1 km
- Hypothesized ecological role**:** Forage, prey, and water availability.
28. bio17: Mean monthly precipitation amount of the driest quarter
- Description: Average precipitation of the driest quarter (kg m⁻² month⁻¹).
- Spatial precision: 1 km
- Hypothesized ecological role: Forage, prey, and water availability.
29. bio18: Mean monthly precipitation amount of the warmest quarter
- Description: Average precipitation of the warmest quarter (kg m⁻² month⁻¹).
- Spatial precision: 1 km
- Hypothesized ecological role: Forage, prey, and water availability.
30. bio19: Mean monthly precipitation amount of the coldest quarter
- Description: Average precipitation of the coldest quarter (kg m⁻² month⁻¹).
- Spatial precision: 1 km
- Hypothesized ecological role: Forage, prey, and water availability.
