Table 2.
Description of data layers used as covariates or masking layers during the downscaling process
| Description | Units | Use | Source | |
|---|---|---|---|---|
| Climate | ||||
| EARS | MOD16 dataset gap filled with Annual Actual Evaporations calculated as the sum of monthly EA derived using the Budkyo framework based on WorldClim climatic data, using PAWHC calculated from 1 km Soil Depth from www.soilgrids.org combined with AWC from the Harmonized World Soil Database | mm | Predictor | Hijmans et al. (2005); Mu et al. (2007); FAO/IIASA/ISRIC/ISSCAS/JRC (2012); Hengl et al. (2014) |
| MAT | Mean Annual Temperature with maximum and minimum temperature corrected for radiation differences due to variation in terrain based on Danielson and Dean (2011) following Wilson and Gallant (2000) | °C | Predictor | Wilson and Gallant (2000); Hijmans et al. (2005); Danielson and Dean (2011) |
| PTA | Annual precipitation. Sum of monthly precipitation from WorldClim | mm | Predictor | Hijmans et al. (2005) |
| TWI | Topographic Wetness Index. Calculated at 9″ and upscaled to 30″ | Predictor | Reuter and Hengl (2012) | |
| Landform/substrate | ||||
| ICE | Presence of permanent ice | Binary | Mask | Olson et al. (2001) |
| SLP | Slope calculated at 9″ and upscaled to 30″ | % | Predictor | Reuter and Hengl (2012) |
| SOC | Soil Organic Carbon content. Weighted average of all depth classes | g/kg | Predictor | Hengl et al. (2014) |
| WATER | Presence of permanent water bodies | Binary | Mask | Lehner and Doll (2004) |
| Anthropogenic | ||||
| ACC | Global Accessibility Index. The travel time to the nearest population center of 50,000 or more | Predictor | Uchida and Nelson (2010) | |
| POP | Population density. | People/km2 | Predictor | Balk et al. (2005) |
| Land cover | ||||
| CLC | Consensus land cover. 30″ land‐cover product made by harmonizing multiple products | % | Predictor | Tuanmu and Jetz (2014) |
PAWHC is the Plant Available Water Holding Capacity of soil. AWC is the Available Water Capacity of soil. A variable used as a predictor describes a fine‐grained covariate used in the regression model. A variable described as a mask describes a binary (0 or 1) variable used to determine whether values from a cell are included in the model or excluded.