The geospatial data set for wildfire simulation consists of raster layers that characterize the biophysical properties of the landscape. The description of the layers, as well as references to original data sources, are provided below.
Land cover. Spatial characterization of land cover types is based on a Landsat-derived map aggregated to 90-m spatial resolution. The layer includes eight major land cover types while the overall accuracy of the map (95% confidence interval) is within 0,877 ± 0,008 (Myroniuk et al., 2020)
| Land cover category | Land coversubcategory |
|---|---|
| Water bodies | River; lake; water reservoir |
| Wetlands | Seasonal lake; peatland; marshland |
| Settlements | Urban and suburban territory; infrastructure; highway |
| Other unproductive lands | Sand; rock; bare ground |
| Croplands | Fallow land; cropland; orchard |
| Grasslands | Meadow; grassland with trees; grassland with shrubs |
| Shrublands | Shrublands with and without trees; riparianvegetation |
| Forested areas | Coniferous forest (share of coniferous tree speciesgreater than 75%); mixed forest; deciduous forest (share of deciduous treespecies greater than 75%); windbreak; urban forest; forest regrowth on abandonedlands |
Elevation. Digital elevation model provided by the SRTM – Shuttle Radar Topography Mission, 90 m.
Slope and aspect. The slope and aspect features were calculated based on the elevation model.
Ignition probability grid. The map was generated using all available ignitions from MODIS MOD14/MYD14 fire thermal anomalies product aggregated for individual fire events between 2001–2016 and spatially smoothed within a search radius of 15 km (kernel density).
Burn probability grid The map indicates probability that a given point of the landscape will burn. The burn probability is calculated using fire simulation, e.g., FlamMap.
Fuel models An input data set that provides necessary parameters to Rothermel’s (1972) surface fire spread model. The layer was created after reclassifying the land cover map using the best association of land cover types on the site with the corresponding model from the fuel model atlas (Scott & Burgan, 2005).
| Code | Landcover class | Fuelmodel | Fuelmodel code | Fuel type |
|---|---|---|---|---|
| 1 | Water bodies | 98 | NB8 | Nonburnable |
| 2 | Wetlands | 121 | GS1 | Grass-Shrub |
| 3 | Settlements | 91 | NB1 | Nonburnable |
| 4 | Other unproductive lands | 99 | NB9 | Nonburnable |
| 5 | Croplands | 101 | GR1 | Grass |
| 6 | Grasslands | 102 | GR2 | Grass |
| 7 | Shrublands | 142 | SH2 | Shrub |
| 100 | Coniferous forest | 188 | TL8 | Timber litter |
| 200 | Deciduous forest | 182 | TL2 | Timber litter |
| 300 | Mixed forest | 162 | TU1 | Timber understory |
Fire spread over forested areas depends on canopy fuels data: 1) canopy cover; 2) canopy height, 3) canopy base height, and 4) canopy bulk density. The two last parameters are used to predict transition of surface fire into passive (torching) or active crown fire.
Canopy cover The layer characterizes canopy cover for a given pixel (100×100 m) based on the Global Forest Change map updated for 2015 (Hansen et al., 2013).
Canopy height Coarse-scale assessment of stand height (1 × 1 km) based on the GLAS (Geoscience Laser Altimeter System) RH100 metrics circa 2005 (Simard et al., 2011). The layer was harmonized with the land cover map so that unforested land cover classes were assigned canopy height = 0 m.
Canopy base height The layer was produced using Landsat 8 OLI imagery and machine learning. The information calculated for the Chornobyl Exclusion Zone (Ager et al., 2019) using the US Forest Service algorithms (FVS) was used as a reference
Canopy bulk density The layer represents an approximate estimation of canopy fuel density (in kg·m-3) predicted using the machine learning technique (see canopy base height)
References
