National Soil Carbon Depth

Soil carbon sequestration has great, but untapped potential for mitigation of climate change. This is reflected in the share of carbon credits sold globally, and is due to the lack of a cheap, quick method for soil carbon sampling. The cost of soil carbon sampling increases with depth, so establishing a baseline depth for a given area can allow for more accurate and cheaper testing. The purpose of this dataset, maps and associated Google Earth Engine application is to provide a baseline for sampling depth. This project was completed in partnership with Kirsten McKnight and Jacob Penner from  Native , a Burlington-based public benefit corporation and B Corporation helping companies to find sustainable community scale investment to reduce emissions. We also partnered with Dan Kane, a soil scientist, agroecologist, and doctoral student at the Yale School of the Environment and lead researcher at  Quick Carbon .

We downloaded SoilGrids 100m data for bulk density and soil organic carbon (SOC), which was split into 7 layers across the contiguous United States at depths of 0cm, 5cm, 15cm, 30cm, 60cm, 100cm, and 200cm. In ArcGIS Pro we averaged the depths to create 6 ranged layers (0-5cm, 5-15cm, 15-30cm, 30-60cm, 60-100cm, 100-200cm). We created a soil profile by multiplying SOC, bulk density, and the depth of individual sections; we then added each profile to have a sum of all depths (0-200cm) and divided the 6 profiles by the total profile to create proportion layers displaying what percent of carbon from the profile was in each given range; we summed proportional layers create cumulative SOC layers (0-5cm, 0-15cm, 0-30cm, 0-60cm, 0-100cm, 0-200cm).

Using U.S. rooting depth data from gNATSGO layer "Depth to Any Soil Restrictive layer", we restricted the possible depth at which SOC could be located and recalculated the layers; this was done using raster calculator, with the condition that for any cumulative layer, if the rooting depth is less than the layer depth, then return 100% of SOC, if not, return the value from the cumulative layer.  

We exploded a US shapefile from the US Census Bureau in order to extract only the contiguous United States; during this process, smaller islands were omitted, excluding them from further analysis. Splitting the United States into four quadrants, we ran a per pixel linear trend analysis using R script written by Charlie Bettigole along with our cumulative restricted layers on four computers to calculate slope and intercept for the relationship between depth and proportion of soil. For our trend analysis the depth for each layer was assigned to the midpoint of it’s range (ie 2.5 for 0-5 range). In ArcGIS Pro we merged our four quadrants back together. Using the intercept and slope files we used raster calculator to solve for depth at 50% and 95% carbon. We excluded extraneous values by setting unusually high values to NoData and negative values to 0.

We observed a higher percentage of carbon located in both the southeast and southwest of the U.S. in the 0-30cm soil carbon depth map; in particular, the areas in the southwest are much higher. We suspect this to be due to the soil type (desert/sand) with less overall carbon and shallow rooting depths. This is likely why 100% of total carbon is present closer to the surface. The 0-100cm map follows these trends, with a universal increase in the proportion of total carbon across the country.

In the percentage total carbon maps (50% and 95%), the same trends appeared; the shallower depths on the 50% map are mostly located in the southeast, with some spots in the southwest of the U.S. 95% of total carbon is mostly found between 100 to 140cm.

Created for ID-351G: Advanced GIS and Modeling at Skidmore College. For more information contact gis@skidmore.edu.