Metadata for the file:
Crezee_2022_Median_Peat_thickness_RF_100runs.tif

This is the file used to create Figure 3a of the paper:

Crezee et al. (2022). Nature Geoscience. Mapping peat thickness and carbon stocks of the central Congo Basin using field data. 

-------------------------------------------------------------------------------------
This is a raster file representing modelled peat thickness (in m below the surface)
across the central Congo Basin peatland complex. 

For each pixel, the median value of 100 Random Forest (RF) regressions is provided. 
Each RF regression was trained on 463 peat thickness measurements in the field, and
using 4 predictor variables: 
- distance from the peatland margin, 
- precipitation seasonality, 
- climatic water balance (precipitation minus potential evapotranspiration), 
- distance from the nearest drainage point. 
 
For each of the 100 regressions, a different input map of distance from the peatland
margin was constructed, by randomly selecting (with replacement) a minimum peat 
probability threshold > 0 and < 100 % and generating a binary peat/non-peat map from 
the peat probability map (combined hardwood + palm dominated peat swamp probabilities).
This binary map was then used to calculate the shortest distance to a non-peat pixel. 

----------------------------------------------------------------------------------------
Projection and file details:

The file is a GeoTIFF in a lat/long WGS-84 projection and a pixel size of 0.00044444 degrees (c. 50 m).

---------------------------------------------------------------------------------------
You are free to use these data for any purpose provided you cite the original paper.

The file was created by Bart Crezee and Edward Mitchard.
For questions about these data layers please contact Edward Mitchard 
at edward.mitchard@ed.ac.uk