Supplementary Data: Global rise in forest fire emissions linked to climate change in the extratropics

Supplementary Data for the paper "Global rise in forest fire emissions linked to climate change in the extratropics" by Jones et al. (2024, Science).

The records include mapped pyromes and data and code used to delineate the pyromes.

Mapped Pyromes

The data records include mapped pyromes in three forms:

Shapefile (Jones_etal_2024_Global_Forest_Pyromes.shp.zip). Vector features in shapefile format containing data fields pyrome ID and pyrome name. The zipped file contains .shp, .dbf, .prj, .shx files.

Lower-resolution NetCDF (Jones_etal_2024_Global_Forest_Pyromes_Qdeg.nc). NetCDF version 4 file containing gridded values of pyrome ID at quarter-degree resolution.

Higher-resolution NetCDF (Jones_etal_2024_Global_Forest_Pyromes_005deg.nc). NetCDF version 4 file containing gridded values of pyrome ID at 0.05 degree resolution.

Shapefiles are accessible via GIS programmes such as QGIS or ArcGIS. All files .shp, .dbf, .prj, .shx files must be stored in a single directory

NetCDF files can be access by a variety of programming languages such as Python and R. For quick visualisations and access to the data structure, we suggest using the Panoply  tool https://www.giss.nasa.gov/tools/panoply/.

Correlation Data

The data records (Correlation_Qdeg.zip) include gridded quarter-degree correlations between forest burned area (BA) and each of the following variables:

Fire weather index

Atmospheric instability (continuous Haines index)

Lightning flash density

Soil moisture

Vegetation productivity (Normalised Difference Vegetation Index)

Population density

Cropland cover

Pasture cover

Road density

Potential fuel loads - surface fuels

Potential fuel loads - shrub fuels

Potential fuel loads - canopy and ladder fuels

Terrain ruggedness index

Forest area density

The BA data derive from MODIS MCD64A1 collection 6.1 (Giglio et al., 2018). BA data for forests is masked using the MODIS MOD44B product (DiMiceli et al., 2021) with a 30% tree cover threshold. The predictor data derive from multiple sources as desribed by Jones et al. (2024). See Supplementary Methods and Materials.

The gridded correlations data are provided in Hierarchical Data Format version 5 (.hdf5) files, zipped to Correlation_Qdeg.zip. File names describe the variables used. Cropland_Pasture_Qdeg.hdf5 contains data for both cropland and pasture. Each file contains layers describing the Spearman's rho (ρ) correlation coefficient and the related p-value.

As explained and justified by Jones et al. (2024), the correlation structure used depends on the variable (see Supplementary Methods and Materials) as per the following categories:

Fire Weather Index, Atmospheric Instability, and Lightning Flash Density: Monthly correlation between forest BA and each variable across all fire season months in the period 2001-2021 at the quarter-degree resolution.

Soil Moisture: Inter-annual correlation between (i) mean soil moisture during the fire season and (ii) accumulated forest BA during the fire season at quarter-degree resolution across years 2001-2021. 

Vegetation Productivity (NDVI): Inter-annual correlation between (i) mean NDVI during the prior growing season and (ii) accumulated forest BA during the fire season at quarter-degree resolution across years 2001-2021. 

Population Density, Cropland Cover, Pasture Cover, Road Density, Terrain Ruggedness Index, Forest Area Density: Spatial correlation between mean annual forest BA and each variable across the 0.05° cells within each quarter-degree cell during 2001-2021.

Note that these grids are provided for insights into spatial variation in the input correlation data. Pyromes are defined based on correlations fitted on the spatial scale of Olson ecoregions, not quarter-degree grid cells (see further details below).

Clustering Code

DEMO_Clustering.zip contains R Statistics code for clustering forest ecoregions into pyromes based on correlations observed between forest BA and 14 predictors at regional level. The Input directory contains a .RData data frame with correlations between forest BA and each predictor for ecoregions. For demonstrative purposes the code is applied to cluster forest ecorgions of North America into pyromes. The Regions directory contains ecoregions of North America in shapefile format. The Output directory contains output generated by M. Jones, which can be used for validation purposes once other users have trialled the code.