Developing a mechanistic understanding of belowground ecosystem dynamics is critical to understanding and forecasting whole-ecosystem behavior. Although significant advances have been made in belowground ecosystem ecology, a number of challenges remain. One particularly important problem is the ability to rigorously partition the effects of different belowground processes and to identify how they vary across space and time. This talk presents a data-model integration approach for reconstructing processes related to belowground carbon dynamics. The approach employs a hierarchical Bayesian (HB) framework that simultaneously analyzes diverse data sources (e.g., soil respiration, soil incubations, soil carbon, soil water, carbon isotopes) within the context of process-based models. Two interrelated examples will be presented related to soil carbon dynamics in a Sonoran desert site characterized by a mixture of mesquite shrubs, grasses, and bare ground. The first describes an HB analysis of a soil incubation experiment to infer depth-dependent patterns of microbial activity and carbon substrate availability. The second describes a HB framework for coupling diverse data sources, including the incubation experiment, with isotope mixing and soil respiration models to partition sources of soil carbon efflux. Both examples illustrate challenging problems that cannot be satisfactorily addressed without modern data-model integration approaches.
Results/Conclusions
The HB model used with the incubation study modeled all responses (e.g., carbon fluxes, soil carbon, microbial biomass) simultaneously, treated all observations as stochastic quantities, and allowed for the possibility of “missing” data. For example, as is typical of many field studies, the soil carbon flux, carbon content, and microbial biomass datasets were 93%, 93%, and 56% complete, respectively. The HB analysis produced estimates of microbial carbon-use efficiency, depth-dependent distributions of soil carbon and microbial biomass, and effects of microsite type on these quantities. The results suggest that carbon-use efficiency of soil microbes is 3x higher under large, N-fixing shrubs (mesquite) compared to under grasses and bare areas. This large difference is attributed to enhanced litter quality (higher N), greater soil carbon (ca. 2x higher), and higher microbial biomass (ca. 2x higher) under the large shrubs compared to grass and bare sites. This talk concludes by highlighting the second example, which describes a HB framework for partitioning the contributions of different heterotrophic and autotrophic sources to soil carbon flux and to learn how these contributions vary with soil depth and time.
