NONSTATIONARY SPATIAL PREDICTION OF SOIL ORGANIC CARBON: IMPLICATIONS FOR STOCK ASSESSMENT DECISION MAKING.

Area of Science:

• Environmental Science
• Soil Science
• Geostatistics

Background:

• The Rapid Carbon Assessment (RaCA) project provided soil organic carbon (SOC) measurements across the US (2010-2012).
• Direct SOC observations lack the high spatial resolution required for detailed climate and agricultural studies.
• Predicting SOC in unmeasured locations is crucial for understanding carbon dynamics.

Purpose of the Study:

• To compare various statistical methods for spatial prediction of SOC.
• To introduce and evaluate a novel nonstationary spatial prediction method.
• To address challenges in domain partitioning, covariate incorporation, and uncertainty quantification.

Main Methods:

• Spatial prediction of SOC using a subset of RaCA data.
• Comparison of stationary and nonstationary geostatistical methods.
• Development of a novel nonstationary approach using domain partitioning and covariate analysis.

Main Results:

• Minimal differences in out-of-sample performance criteria across tested methods for this dataset.
• Significant variations in prediction uncertainty maps generated by different methods.
• The novel method produces spatially-varying uncertainty measures.

Conclusions:

• The novel nonstationary method offers valuable insights into prediction uncertainty for SOC.
• Spatially-varying uncertainty maps aid in benchmarking models and identifying areas for soil restoration and carbon sequestration projects.
• Accurate SOC prediction is vital for effective climate and agricultural management.