Forest Soils on the Edge: Partitioning and Modeling Drivers of Soil Respiration Near the Forest-Prairie Ecotone
Melanie Mayes1, Rose Abramoff2, Matthew Craig1* (firstname.lastname@example.org), Anthony Walker1, Jeff Wood3, Jana Phillips1, Sarah Ottinger1, Lianhong Gu1, Paul J. Hanson1
1Oak Ridge National Laboratory, Oak Ridge, TN; 2Lawrence Berkeley National Laboratory, Berkeley, CA; 3University of Missouri, Columbia, MO
The Missouri Ozarks AmeriFlux site (MOFLUX) is located on the eastern U.S. prairie-forest ecotone and is subject to periodic and seasonal droughts. Long-term measurements of total soil respiration were complemented with plots trenched to isolate heterotrophic respiration beginning in 2017, and quarterly measurements collected of texture, pH, moisture content, carbon, and nitrogen in bulk soils and in microbial biomass, and root length and mass. Scientists applied artificial intelligence and wavelet coherence analysis to determine the effects of environmental factors on 17 years of soil respiration data (2004 to 2021) including 4 years (2017 to 2021) of heterotrophic respiration. Random Forest models identified the relative importance of heterotrophic and autotrophic respiration, including soil temperature, soil moisture, leaf-area index (LAI), and time. Wavelet coherence analysis examined the timescales (e.g., daily, weekly, monthly, or seasonal) of key drivers. Results showed that heterotrophic respiration was most responsive to soil temperature at daily and seasonal timescales, while autotrophic respiration was most responsive to aboveground productivity (using LAI as a proxy) and the time of the year. Heterotrophic and autotrophic respiration had similar responses to temperature but not to LAI or time of year; therefore, these were most influential for partitioning between heterotrophic and autotrophic respiration. Soil moisture was most important to respiration on synoptic weekly-to-monthly timescales. Finally, respiration data was used to test implementations of conventional and microbially explicit soil carbon models with alternative soil moisture response functions. To do this, scientists used the multi-assumption soil carbon model developed in the multi-assumption architecture and testbed (MAAT). This model has been developed in response to the vast diversity of soil carbon models representing different combinations of process representations. MAAT is a highly modular modeling framework that can be used to probe the sources of structural and parametric model uncertainty. Using MAAT, scientists found that the choice of moisture response process representation can lead to simulated soil carbon differing by as much as 10% after a 4-year simulation.