Deep Soil: Quantifying and Modeling Subsurface Carbon

Tuesday, 16 December 2014
Jason Nathaniel James, Warren Devine and Robert B Harrison, University of Washington Seattle Campus, Seattle, WA, United States
Some soil carbon datasets that are spatially rich, such as the USDA Forest Service Inventory and Analysis National Program dataset, sample soil to only 20 cm (8 inches), despite evidence that substantial stores of soil C can be found deeper in the soil profile. The maximum extent of tree rooting is typically many meters deep and provides: direct exchange with the soil solution; redistribution of water from deep horizons toward the surface during times of drought; resources for active microbial communities in deep soil around root channels; and direct carbon inputs through exudates and root turnover. This study examined soil carbon to a depth of 2.5 meters across 22 soils in Pacific Northwest Douglas-fir forests. Excavations at 20 additional sites took place in summer 2014, greatly expanding the spatial coverage and extent of the data set. Forest floor and mineral soil bulk density samples were collected at depths of 0.1, 0.5, 1.0, 1.5, 2.0 and 2.5 meters. Pool estimates from systematic sampling depths shallower than 1.5 m yielded significantly smaller estimates than the total soil stock to 2.5 meters (P<0.01). On average, only 5% of soil C was found in the litter layer, 35% was found below 0.5 meter, and 21% was found below 1.0 meter. Due to the difficulty of excavating and measuring deep soil carbon, a series of nonlinear mixed effect models were fit to the data to predict deep soil carbon stocks given sampling to 1.0 meter. A model using an inverse polynomial function predicted soil carbon to 2.5 meters with -5.6% mean error. The largest errors occurred in Andisols with non-crystalline minerals, which can adsorb large quantities of carbon on mineral surfaces and preserve it from decomposition. An accurate spatial dataset of soil depth to bedrock would be extremely useful to constrain models of the vertical distribution of soil carbon. Efforts to represent carbon in spatial models would benefit from considering the vertical distribution of carbon in soil. Sampling deep soil will avoid biased estimates of soil carbon and create a more complete picture of soil pools and changes over time.