Soil Organic Carbon Stock Estimates with Uncertainty across Latin America
This dataset provides 5 x 5 km gridded estimates of soil organic carbon (SOC) across Latin America that were derived from existing point soil characterization data and compiled environmental prediction factors for SOC. This dataset is representative for the period between 1980 to 2000s corresponding with the highest density of observations available in the WoSIS system and the covariates used as prediction factors for soil organic carbon across Latin America. SOC stocks (kg/m2) were estimated for the SOC and bulk density point measurements and a spatially explicit measure of the SOC estimation error was also calculated. A modeling ensemble, using a linear combination of five statistical methods (regression Kriging, random forest, kernel weighted nearest neighbors, partial least squared regression and support vector machines) was applied to the SOC stock data at (1) country-specific and (2) regional scales to develop gridded SOC estimates (kg/m2) for all of Latin America. Uncertainty estimates are provided for the two model predictions based on independent model residuals and their full conditional response to the SOC prediction factors. These SOC estimates provide a reproducible example, on country-specific and regional scales, for digital soil mapping across Latin America and contribute to reducing the uncertainty of SOC estimates and improving the parameterization of global models across Latin America. This dataset includes six data files in GeoTIFF (.tif) format at 5 km resolution across Latin America, including: (1) a mosaic of country-specific soil organic carbon estimates, (2) model uncertainty derived for the country-specific estimates, (3) a mosaic of the regional soil organic carbon estimates, (4) model uncertainty derived for the regional estimates, and (5-6) two trend maps of approximate errors associated with the SOC stock calculation method. There is one data file in comma-separated format (.csv) of the point soil characterization data with calculated SOC stock estimates. Four companion files include: a 133-band GeoTiff containing the environmental predictor variables for SOC across Latin America, a .csv file with descriptions of the environmental variables, a shapefile (.shp) of the point soil characterization data with SOC stock estimates and a *.kmz file to display the same.
Main Authors: | , , , , , , , , , , , , , , , , , , , |
---|---|
Format: | info:ar-repo/semantics/artículo biblioteca |
Language: | eng |
Published: |
ORNL-DAAC
2019-07-03
|
Subjects: | Carbono Orgánico del Suelo, América Latina, Soil Organic Carbon, Factores de Predicciones, Error de Estimación, Prediction Factors, Estimation Error, |
Online Access: | http://hdl.handle.net/20.500.12123/17694 https://daac.ornl.gov/CMS/guides/Country_SOC_Latin_America.html https://doi.org/10.3334/ORNLDAAC/1615 |
Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
Summary: | This dataset provides 5 x 5 km gridded estimates of soil organic carbon (SOC) across Latin America that were derived from existing point soil characterization data and compiled environmental prediction factors for SOC. This dataset is representative for the period between 1980 to 2000s corresponding with the highest density of observations available in the WoSIS system and the covariates used as prediction factors for soil organic carbon across Latin America. SOC stocks (kg/m2) were estimated for the SOC and bulk density point measurements and a spatially explicit measure of the SOC estimation error was also calculated. A modeling ensemble, using a linear combination of five statistical methods (regression Kriging, random forest, kernel weighted nearest neighbors, partial least
squared regression and support vector machines) was applied to the SOC stock data at (1) country-specific and (2) regional scales to develop gridded SOC estimates (kg/m2) for all of Latin America.
Uncertainty estimates are provided for the two model predictions based on independent model residuals and their full conditional response to the SOC prediction factors.
These SOC estimates provide a reproducible example, on country-specific and regional scales, for digital soil mapping across Latin America and contribute to reducing the uncertainty of SOC estimates and improving the parameterization of global models across Latin America.
This dataset includes six data files in GeoTIFF (.tif) format at 5 km resolution across Latin America, including: (1) a mosaic of country-specific soil organic carbon estimates, (2) model uncertainty derived for the country-specific estimates, (3) a mosaic of the regional soil organic carbon estimates, (4) model uncertainty derived for the regional estimates, and (5-6) two trend maps of approximate errors associated with the SOC stock calculation method. There is one data file in comma-separated format (.csv) of the point soil characterization data with calculated SOC stock estimates.
Four companion files include: a 133-band GeoTiff containing the environmental predictor variables for SOC across Latin America, a .csv file with descriptions of the environmental variables, a shapefile (.shp) of the point soil characterization data with SOC stock estimates and a *.kmz file to display the same. |
---|