Soil organic carbon (SOC) or the carbon found in soil’s organic matter is an important component in the Earth’s carbon cycle and can help mitigate the negative consequences of climate change. Because of its broad relevance to climate and agriculture, the Intergovernmental Panel on Climate Change recommended that the spatial distribution of SOC be carefully monitored. Due to time and cost constraints, spatially-referenced measurements of SOC or “stocks”, e.g. the amount of SOC in a volume, are available only at limited locations. In this talk, I will present two Bayesian non-stationary statistical models for SOC with the goal of understanding spatial variability and dependence in SOC for sampling and spatial prediction of SOC. Both approaches model SOC by assuming that the process is globally non-stationary but locally stationary with regions of stationarity identified using different strategies. In one approach, the regions are defined as partitions of the spatial domain where spatially-varying covariates are more homogeneous; in the other, the regions are defined as subsets of the spatial domain with varying strength of spatial dependence.

I will show that inference and SOC predictions yielded by our models are valuable for sampling design and to decision makers, in that they can be used to better benchmark mechanistic models and identify targets for soil restoration projects.

 

 

The link Teams is available here.