Predictive Soil Carbon Maps

The Linear Stepwise Regression (SR) and Random Forest (RF) results indicated that growing season average of Normalized Difference Vegetation Index (NDVI), Elevation, and Mean Annual Precipitation (MAP) were the most useful factors in predicting SC and SOC in 2013 and 2014. Figure 1 displays the predicted SC values across BC grasslands based on the SLR results for 2013 fenced systems. Notice the distribution of higher SC values in upper grasslands, at high elevations which are associated with more moisture and vegetation.

Figure 1

To evaluate how carbon stores over time, we compared SC and SOC from 2013 to 2014 using a Paired Ttest. Significantly higher SC was recorded in 2014 as compared to 2013 (p=0.005).  Likewise, more SOC was recorded in 2014 as compared to 2013 (p=0.083).  The figure below displays the SR results for change in SC in grazed systems from 2013 to 2014 (Figure 2).

change in sc grazed


Based on comparisons of mean squared error, RF models created better predictions than SLR. See the maps below for a comparison of SR and RF based SC and SOC prediction for grazed and fenced systems in 2013 and 2014. Simply click on a map to enlarge it.

Linear Stepwise Regression Random Forest


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s