On spatial scaling & environmental controls of soil organic carbon stocks

Tuesday, 16 December 2014
Umakant Mishra1, William J Riley2 and Charles D Koven2, (1)Argonne National Laboratory, Argonne, IL, United States, (2)Lawrence Berkeley Natl Lab, Berkeley, CA, United States
Spatial heterogeneity of terrestrial land surface modulates the fluxes of energy, moisture, and greenhouse gases. However, representing the terrestrial heterogeneity of biogeochemistry in earth system models (ESMs) remains a critical scientific challenge. We investigated the impact of spatial scaling on environmental controls and predicted soil organic carbon (SOC) stocks across the state of Alaska, USA. We used over 500 soil profile observations and environmental factors such as topography, climate, land cover types, and surficial geology to predict the SOC stocks at 50 m spatial resolution. We upscaled both the predicted SOC stocks and environmental variables from finer to coarser spatial scales (100 m, 200 m, 500 m, 1 km, 2 km, 5 km, and 10 km) and generated SOC stock estimates for each scale till the predicted variance of SOC stocks became constant. We found different environmental factors as statistically significant predictors at different spatial scales. Topographic attributes were important predictors at finer scales whereas surficial geology types became significant predictors at larger spatial scales. Only elevation, temperature, potential evapotranspiration, and barren land cover types were significant predictors at all scales. The controls (predictive power) of these environmental variables on SOC stocks decreased with upscaling. Highest and lowest decrease in predictive power was observed for potential evapotranspiration (55%) and elevation (25%). Similarly, intermediate decrease was observed for temperature (45%), and barren land cover types (45%). The predicted variance of SOC stocks decreased by 45% as the spatial scaling was increased from 50 m to 10km. We believe the statistical structure of the scaling behavior of SOC stocks can inform ESMs in appropriately representing the spatial heterogeneity of SOC stocks.