Density fractionation of forest soils: methodological questions and interpretation of incubation results and turnover time in an ecosystem context

Soil organic matter (SOM) is often separated by physical means to simplify a complex matrix into discrete fractions. A frequent approach to isolating two or more fractions is based on differing particle densities and uses a high density liquid such as sodium polytungstate (SPT). Soil density fractions are often interpreted as organic matter pools with different carbon (C) turnover times, ranging from years to decades or centuries, and with different functional roles for C and nutrient dynamics. In this paper, we discuss the development and mechanistic basis of common density-based methods for dividing soil into distinct organic matter fractions. Further, we directly address the potential effects of dispersing soil in a high density salt solution on the recovered fractions and implications for data interpretation. Soil collected from forested sites at H. J. Andrews Experimental Forest, Oregon and Bousson Experimental Forest, Pennsylvania was separated into light and heavy fractions by floatation in a 1.6 g cm-³ solution of SPT. Mass balance calculations revealed that between 17% and 26% of the original bulk soil C and N content was mobilized and subsequently discarded during density fractionation for both soils. In some cases, the light isotope was preferentially mobilized during density fractionation. During a year-long incubation, mathematically recombined density fractions respired ~40% less than the bulk soil at both sites and light fraction (LF) did not always decompose more than the heavy fraction (HF). Residual amounts of tungsten (W) present even in well-rinsed fractions were enough to reduce microbial respiration by 27% compared to the control in a 90-day incubation of Oa material. However, residual W was nearly eliminated by repeated leaching over the year-long incubation, and is not likely the primary cause of the difference in respiration between summed fractions and bulk soil. Light fraction at Bousson, a deciduous site developed on Alfisols, had a radiocarbon-based mean residence time (MRT) of 2.7 or 89 years, depending on the interpretation of the radiocarbon model, while HF was 317 years. In contrast, both density fractions from H. J. Andrews, a coniferous site developed on andic soils, had approximately the same MRT (117 years and 93 years for LF and HF). At H. J. Andrews the organic matter lost during density separation had a short MRT (19 years) and can account for the difference in respired CO₂ between the summed fractions and the bulk soil. Recognition and consideration of the effects of the density separation procedure on the recovered fractions will help prevent misinterpretation and deepen our understanding of the specific role of the recovered organic matter fractions in the ecological context of the soil studied.