Assessing Uncertainty in Soil Carbon Assessment Using Bayesian Hierarchical Modeling.

Assessing the uncertainty of digital soil maps is critical when using them in the subsequent decision-making processes. In this study, we adopted the Bayesian hierarchical model to assess SOC and quantify the prediction uncertainty due to parameter uncertainty in the state of Florida, USA. The prediction accuracy and interval were validated using an independent dataset. Results showed that the Bayesian model incorporating six covariates representing soil moisture, vegetation and topography had better prediction accuracy than the Bayesian model with only spatial random effects. In addition, the former has a narrow prediction interval indicating it has better prediction precision. Generally, the size of Bayesian prediction intervals increased with the posterior mean SOC predictions. The validation of the prediction interval with the independent validation data also confirmed that the credible interval from Bayesian inference was statistically accurate in describing the prediction uncertainty. The prediction uncertainty stemmed from the uncertainty in the model parameter estimation.  Generally, the spatial random effect parameters from both models did not exhibit much uncertainty while most of the fixed effect parameters from the Bayesian model with covariates showed considerable uncertainty compared to those of the spatial random effect parameters, which can be attributed to the correlation among these covariates. Our findings are important to quantify the SOC in the Southeastern USA where a wide suite of carbon-poor and carbon-rich soils are found. The SOC assessment with uncertainty can be used not only in support of soil assessment, but also target soils for mitigation and adaptation to global climate change.