The rate of gaseous diffusion in soils affects the exchange of gases between the soil and the atmosphere, thereby affecting rates of soil respiration and other soil microbial processes. Understanding the causes of spatial and temporal variation in soil diffusivity will help explain controls of soil sources and sinks of atmospheric gases. In a study of sources of CO₂ in deep soils of forests and pastures of the eastern Amazon, we estimated gaseous diffusivity from bulk density and volumetric water content using published equations that assume the soil to be either an aggregated or nonaggregated medium. The aggregated model requires differentiation of interand intra-aggregate pore space; we estimated intra-aggregate pore space from volumetric water content at field capacity. Steady state ²²²Rn profiles were predicted from a 1-D model using the diffusivities generated by both aggregated and nonaggregated models. Predicted values were compared with ²²²Rn activities measured to 5 m depth. While the models predict similar radon activities below about 1 m, large differences are predicted for the top 1 m of soil. The nonaggregated model underestimated diffusivity and overestimated ²²²Rn activities at 1 m and above, which is not surprising given that surface soils are usually well aggregated. Having validated the aggregated media model using the ²²²Rn profiles, estimates of diffusivity were combined with measured profiles of CO₂concentrations to estimate CO₂ production by depth. About 70-80% of the measured CO₂ flux from the soil surface was produced in the top 1 m of soil (including litter in the forest). The 20-30% produced below 1 m results from root respiration and microbial decay of root inputs at depth, indicating that deep soil processes are a non-trivial component of carbon cycling in these deep-rooting ecosystems. About 1% of the 20 kg C m^-2 stock of soil C found between 1 m and 8 m depths turns over annually, indicating that land-use changes that affect rooting depth could significantly affect deep soil C stocks over decades to centuries. Fully understanding the role of land-use change on the global carbon cycle will require consideration of these deep soil processes.

Boreal forests and wetlands are thought to be significant carbon sinks, and they could become net C sources as the Earth warms. Most of the C of boreal forest ecosystems is stored in the moss layer and in the soil. The objective of this study was to estimate soil C stocks (including moss layers) and rates of accumulation and loss for a 733 km² area of the BOReal Ecosystem-Atmosphere Study site in northern Manitoba, using data from smaller-scale intensive field studies. A simple process-based model developed from measurements of soil C inventories and radiocarbon was used to relate soil C storage and dynamics to soil drainage and forest stand age. Soil C stocks covary with soil drainage class, with the largest C stocks occurring in poorly drained sites. Estimated rates of soil C accumulation or loss are sensitive to the estimated decomposition constants for the large pool of deep soil C, and improved understanding of deep soil C decomposition is needed. While the upper moss layers regrow and accumulate C after fires, the deep C dynamics vary across the landscape, from a small net sink to a significant source. Estimated net soil C accumulation, averaged for the entire 733 km² area, was 20 g C m⁻² yr⁻¹ (28 g C m⁻² yr⁻¹ accumulation in surface mosses offset by 8 g C m⁻² yr⁻¹ lost from deep C pools) in a year with no fire. Most of the C accumulated in poorly and very poorly drained soils (peatlands and wetlands). Burning of the moss layer in only 1% of uplands would offset the C stored in the remaining 99% of the area. Significant interannual variability in C storage is expected because of the irregular occurrence of fire in space and time. The effects of climate change and management on fire frequency and on decomposition of immense deep soil C stocks are key to understanding future C budgets in boreal forests.

Forests in seasonally dry areas of eastern Amazonia near Paragominas, Pará, Brazil, maintain an evergreen forest canopy through an extended dry season by taking up soil water through deep (>1 m) roots. Belowground allocation of C in these deep-rooting forests is very large (1900 g C m⁻² yr⁻¹) relative to litterfall (460 g C m⁻² yr⁻¹). The presence of live roots drives an active carbon cycle deeper than l m in the soil. Although bulk C concentrations and ¹⁴C contents of soil organic matter at >l-m depths are low, estimates of turnover from fine-root inputs, CO₂ production, and the ¹⁴C content of CO₂ produced at depth show that up to 15% of the carbon inventory in the deep soil has turnover times of decades or less. Thus the amount of fast-cycling soil carbon between 1 and 8-m depths (2–3 kg C m⁻², out of 17–18 kg C m⁻²) is significant compared to the amount present in the upper meter of soil (3–4 kg C m⁻² out of 10–11 kg C m⁻²). A model of belowground carbon cycling derived from measurements of carbon stocks and fluxes, and constrained using carbon isotopes, is used to predict C fluxes associated with conversion of deep-rooting forests to pasture and subsequent pasture management. The relative proportions and turnover times of active (including detrital plant material; 1–3 year turnover), slow (decadal and shorter turnover), and passive (centennial to millennial turnover) soil organic matter pools are determined by depth for the forest soil, using constraints from measurements of C stocks, fluxes, and isotopic content. Reduced carbon inputs to the soil in degraded pastures, which are less productive than the forests they replace, lead to a reduction in soil carbon inventory and Δ¹⁴C, in accord with observations. Managed pastures, which have been fertilized with phosphorous and planted with more productive grasses, show increases in C and ¹⁴C over forest values. Carbon inventory increases in the upper meter of managed pasture soils are partially offset by predicted carbon losses due to death and decomposition of fine forest roots at depths >1 m in the soil. The major adjustments in soil carbon inventory in response to land management changes occur within the first decade after conversion. Carbon isotopes are shown to be more sensitive indicators of recent accumulation or loss of soil organic matter than direct measurement of soil C inventories.