The Refugia habitability index is a composite raster built by classifying six environmental factors into three-class suitability masks (suitable, marginal, unsuitable) and combining them multiplicatively. A single failing factor drives the composite toward zero, ensuring that regions must satisfy all criteria simultaneously. Below, each factor is shown as its continuous input dataset alongside its reclassified index mask. The flooding factor combines projected sea-level rise with WRI Aqueduct 100-year riverine flood depth, and applies to all three epochs (current, 2100, 2300). Habitability is then crossed with two asset surfaces — gridded population and the Biodiversity Intactness Index — to produce a pair of risk rasters: Human Risk (people in declining habitability) and Biodiversity Risk (intact ecosystems becoming climate refugia).
Slope is computed from the GMTED2010 1 km global mean elevation product (Danielson & Gesch 2011, USGS OFR 2011–1073) using Horn's algorithm (Horn 1981, Proc. IEEE 69:14) and aggregated as cell-mean to the 30 km analysis grid. The resulting continuous slope-in-degrees raster is reclassified following the FAO/UNESCO 3-class physiographic convention (Soils Bulletin 32; HWSD v1.2): cells with mean slope ≤ 8° are suitable ("level to gently undulating"), 8–16° are marginal ("rolling to hilly"), and ≥ 16° are unsuitable ("steeply dissected to mountainous"). The 8/16 thresholds are interpreted in degrees rather than percent grade (the FAO break points of 8% and 30% grade ≈ 4.6° and 16.7°) because cell-mean averaging at 30 km tends to mask the within-cell low-slope microsites where settlement and agriculture concentrate (Cohen & Small 1998, PNAS 95:14009). Topography is time-invariant across all epochs.
Soil quality classes from SoilGrids are reclassified into suitable (high fertility), marginal (moderate), and unsuitable (non-soil, water, permafrost). For 2100 and 2300 epochs, permafrost zones projected to thaw under CMIP6 SSP5-8.5 mean annual temperature are reclassified as marginal rather than unsuitable.
Annual runoff / water balance from the WRI Aqueduct dataset. Regions with runoff below 9 × 10⁶ m³ are unsuitable; above 5 × 10⁹ m³ are suitable. Water availability is held constant across all three epochs — the analysis isolates climate-driven changes to the other factors against a fixed hydrological baseline.
Wet-bulb temperature (Tw) is computed from CMIP6 daily maximum temperature and relative humidity using the Stull (2011) approximation. Thresholds are applied uniformly across all three epochs and anchored on the physiological literature: Tw ≤ 26 °C is suitable (sustainable for unacclimatised populations doing ordinary work; ISO 7243 moderate-work compensability ceiling); 26–30 °C is marginal (severe heat strain); Tw > 30 °C is unsuitable, matching the empirical compensability ceiling determined by the PSU HEAT Project (Vecellio et al. 2022, J. Appl. Physiol. 132:340; Vecellio et al. 2023, PNAS 120:e2305427120). Population-mortality events independently emerge at Tw ≈ 28 °C (Mora et al. 2017, Nat. Clim. Change). The theoretical 35 °C survivability ceiling (Sherwood & Huber 2010, PNAS) is now widely regarded as a thermodynamic upper bound rather than an operational habitability threshold and is not used here.
The maximum number of consecutive dry days (CDD) per year from CMIP6 projections under SSP5-8.5. Regions with ≤ 63 CDD are suitable; 64–171 are marginal; above 171 are classified as unsuitable, representing extreme drought stress incompatible with rainfed agriculture.
Two flood hazards are combined into a single 3-class suitability mask, taking the worst classification at each cell. Coastal sea-level rise is keyed to the SRTM DEM. For the 2100 epoch (5 m SLR scenario): cells with elevation ≤ 1 m are unsuitable (deep inundation), 1–5 m are marginal (high-tide and storm-surge exposure), and > 5 m are suitable. For the 2300 epoch (65 m total-ice-melt scenario: Antarctica ~58 m + Greenland ~7 m + other glaciers): ≤ 5 m are unsuitable, 5–65 m are marginal, and > 65 m are suitable. Riverine flooding uses the WRI Aqueduct 100-year return-period inundation depth: depth ≥ 1 m is unsuitable, 0.1–1 m is marginal, and < 0.1 m is suitable. The current epoch has no SLR component and applies only the riverine layer. Riverine depth at the future horizon is the per-pixel ensemble maximum across four CMIP5 GCMs (NorESM1-M, GFDL-ESM2M, HadGEM2-ES, IPSL-CM5A-LR) for the RCP 8.5 / 2080 horizon; the historical (WATCH) baseline is used for the current epoch, and the 2080 horizon is reused as a proxy for 2300 since no published Aqueduct projections run past 2080.
The Biodiversity Intactness Index (BII) from the PREDICTS database measures how much of each cell's original biodiversity remains, as a fraction of its pre-industrial baseline. BII is not a habitability factor — humans live readily in biodiverse landscapes — but it is a key asset class for the Biodiversity Risk raster downstream, which crosses habitability with intact-ecosystem extent to surface the places where displaced humans are most likely to encroach on remaining biodiversity.
The six factor masks are multiplied element-wise. Because each mask uses values of 0 (unsuitable), 1 (suitable), or 2 (marginal), the product encodes which combination of factors constrain each cell. A product of zero (any single unsuitable factor) maps to a habitability score of 0.0; all-suitable maps to 1.0. Intermediate products are mapped to a continuous red–yellow–green color ramp. Finally, ocean and inland water bodies are masked out using slope-derived water detection.
The composite habitability raster describes where the planet remains liveable, but not what stands to lose by its decline. We cross habitability with two static 2020 asset surfaces — gridded population and the Biodiversity Intactness Index — to produce a pair of risk rasters. Human Risk = (1 − habitability) × population, surfacing where many people live in places whose climate is becoming uninhabitable. Biodiversity Risk = habitability × BII, surfacing intact ecosystems whose climate is becoming a refugium for displaced humans, and which may therefore face encroachment. Both asset surfaces are held at their 2020 baseline; the evolution of risk through 2100 and 2300 is therefore driven entirely by climate-driven changes in habitability against fixed present-day human and ecological footprints.