View Full Table | Close Full ViewTable 2.
Akaike Information Criterion (AIC) values for the best six models. AIC1, AIC2, and AIC3 use the measured soil organic carbon (SOC) and observed land-use classes (where appropriate). AIC4 uses the model predictions of Jones et al. (2005) to give the SOC and the Corine Land Cover (CLC) 2000 database (Heymann et al., 1994) to give the land use, AIC5 uses the SOC model predictions of Meersmans et al. (2012) and the CLC database for land use. AIC1, AIC2, AIC4, and AIC5 were computed using a simulating model with a constant mean and Gaussian plus nugget covariance function. AIC3 used the model selected by AIC1 (with SOC and population density as covariates and the Gaussian plus nugget covariance function) to give the simulating distribution. Simulating parameters were chosen for AIC1, AIC2, AIC4, and AIC5 based on a short initial run of the algorithm, and for AIC3 were the Monte Carlo maximum likelihood fitted parameters from AIC1.
|Population density, SOC
|Population density, SOC, altitude
|Population density, SOC, road density
|Population density, SOC, land use
†NA, not applicable (in case of neither SOC nor land use being covariates).