Estimating spatial distribution of soil organic carbon for the Midwestern United States using historical database

• Estimation of SOC at regional scale is important to address climate change issues.
• The GWRK approach was used in this study for SOC estimations.
• GWRK provided lower estimation errors compared to GWR for estimating SOC.
• GWRK performs better at regional scale as compared to the global regression.
• Local models provide better explanation of spatial distribution of SOC.

Soil organic carbon (SOC) is the most important parameter influencing soil health, global climate change, crop productivity, and various ecosystem services. Therefore, estimating SOC at larger scales is important. The present study was conducted to estimate the SOC pool at regional scale using the historical database gathered by the National Soil Survey Staff. Specific objectives of the study were to upscale the SOC density (kg C m−2) and total SOC pool (Pg C) across the Midwestern United States using the geographically weighted regression kriging (GWRK), and compare the results with those obtained from the geographically weighted regression (GWR) using the data for 3485 georeferenced profiles. Results from this study support the conclusion that the GWRK produced satisfactory predictions with lower root mean square error (5.60 kg m−2), mean estimation error (0.01 kg m−2) and mean absolute estimation error (4.30 kg m−2), and higher R2 (0.58) and goodness-of-prediction statistic (G = 0.59) values. The superiority of this approach is evident through a substantial increase in R2 (0.45) compared to that for the global regression (R2 = 0.28). Croplands of the region store 16.8 Pg SOC followed by shrubs (5.85 Pg) and forests (4.45 Pg). Total SOC pool for the Midwestern region ranges from 31.5 to 31.6 Pg. This study illustrates that the GWRK approach explicitly addresses the spatial dependency and spatial non-stationarity issues for interpolating SOC density across the regional scale.