2018 ESA Annual Meeting (August 5 -- 10)

COS 112-9 - Integrating -omics data into soil organic matter decomposition models

Thursday, August 9, 2018: 4:20 PM
338, New Orleans Ernest N. Morial Convention Center
Edward R. Brzostek1, Benjamin Sulman2, Ember Morrissey3, Joseph E. Carrara1, Nanette Raczka1 and Kara E. Allen4, (1)Department of Biology, West Virginia University, Morgantown, WV, (2)School of Natural Sciences, University of California, Merced, Merced, CA, (3)Plant and Soil Sciences, West Virginia University, Morgantown, WV, (4)Manaaki Whenua--Landcare Research, Lincoln, WV, New Zealand
Background/Question/Methods

Recent technological innovations in nucleic acid-based research methods are transforming our understanding of how microbes drive and respond to environmental change. However, finding ecologically meaningful simplifications that can leverage this data to improve soil organic matter decomposition models remains a key challenge. To meet this challenge, we used measurements of the composition, traits, and gene expression of microbial communities in forest soils to parameterize and validate new microbial processes and guilds in a state-of-the-art plant-microbial interactions model, Fixation and Uptake of Nitrogen – Carbon, Organisms, and Rhizosphere Processes in the Soil Environment (FUN-CORPSE). We iteratively added complexity to the plant and soil components of the model. First, we used data generated using stable isotope probing in soil mesocosms to improve parametrizations of how labile and recalcitrant litter inputs drive variability in microbial carbon use efficiency (CUE) and turnover. Second, we used transcriptomics data from a whole watershed N fertilization experiment to parameterize the abundance of microbial guilds (i.e., fungi vs. bacteria; oligotrophs vs. copiotrophs) that vary in CUE, turnover and enzyme synthesis to test whether this addition improved model performance. Finally, we compared each model iteration to the baseline model as well as assessed model sensitivity to key parameters.

Results/Conclusions

At the mesocosm level, the empirical constraints on microbial CUE and turnover enhanced the model’s ability to predict CO2 losses and litter decay rates over the baseline parameterization. Sensitivity analyses showed that the empirical parameterization of CUE drove the majority of model improvement. At the plot level in the N fertilization experiment, relative gains in model improvement declined with increasing model complexity. The simplest microbial guild model that encompassed only fungal and bacterial guilds predicted enzyme activity in control soils with about the same accuracy as complex guild arrays. This decline in relative model improvement with increasing complexity may reflect increasing empirical uncertainty in using bioinformatics to assign traits at lower taxonomic levels. When we ran the validated model in the fertilized plots, the model captured observed declines in enzyme activity. However, sensitivity analyses showed that reduced plant C allocation to the rhizosphere was a larger driver of these declines than shifts in microbial community abundance or traits. Given that -omics datasets and our bioinformatics capabilities are increasing rapidly, this modeling effort provides a preliminary framework for using -omics data in predictive soil organic matter decomposition models.