Carbon map and uncertainty in forested areas of Canada, 250m spatial resolution
datasetposted on 18.05.2021, 06:21 by Camile Sothe, Alemu Gonsamu, James Snider, Joyce Arabian, Werner A. Kurz, Sarah Finkelstein
This project aimed to produce the first wall-to-wall estimate of C stocks in plants and soils of Canada at 250 m spatial resolution. This dataset contains the map with total C stored in plants of forested areas in Canada (AGB, BGB and dead plants) in kg/m² and C stock uncertainty. To estimate the C stored in plants of forest areas, we used 47,967 ground measurements of AGB measures and 58 covariates mainly composed of optical data, terrain parameters, structural parameters (e.g., SAR data, clump index, canopy heights – generated from satellite LiDAR- included in the other dataset), soil type map and radiation flux data. Different models were trained using a recursive feature elimination, random forest scheme and a 5-fold cross-validation assessment. The model with higher R² and lowest root mean square error (RMSE) was used for spatial prediction of AGB in forest areas while 1st and 3rd quantiles of RF quantile regression were used to build the uncertainty map. After generating the AGB map, the root biomass of forest areas was computed by its relationship to AGB according to forest type. The dead plant materials were computed by a linear regression between live and dead AGB defined with ground measurements. Ultimately, the AGB as well as dead plant materials and BGB were multiplied by 0.5 to provide the maps in kg C m-2.