Table 5. Model predictors, results, and errors from linear mixed, random forest, and boosted regression models built to predict % organic carbon in randomly selected 20% proportion of the total data.
Model Predictor Variables % OC Estimates % OC Estimates % OC Estimates RMSE R2, modeled vs measured data
Deg Treat Ref
Measured Data NA 3.09 2.83 3.93 NA NA
Linear Mixed Model Site, reach sample hole, depth, mean annual temperature, NDVI, silt+clay, soil type 3.19 2.78 4.26 1.33 0.64
Random Forest Model treatment, depth, texture, elevation, NDVI, mean annual temperature, drainage area, land cover class, geology, ecoregion, silt+clay, vegetation 3.36 2.93 4.01 1.26 0.68
Boosted Regression Model treatment, depth, texture, elevation, NDVI, mean annual temperature, drainage area, land cover class, geology, ecoregion, silt+clay, vegetation 3.30 2.73 4.10 1.36 0.62