Litter decomposition is an important ecosystem process and global carbon flux that has been shown to be controlled by climate, litter quality, and microbial communities. Process-based ecosystem models are used to predict responses of litter decomposition to climate change. While these models represent climate and litter quality effects on litter decomposition, they have yet to integrate empirical microbial community data into their parameterizations for predicting litter decomposition. To fill this gap, our research used a comprehensive leaf litterbag decomposition experiment at 10 temperate forest U.S. National Ecological Observatory Network (NEON) sites to calibrate (7 sites) and validate (3 sites) the MIcrobial-MIneral Carbon Stabilization (MIMICS) model. MIMICS was calibrated to empirical decomposition rates and to their empirical drivers, including the microbial community (represented as the copiotroph-to-oligotroph ratio). We calibrate to empirical drivers, rather than solely rates or pool sizes, to improve the underlying drivers of modeled leaf litter decomposition. We then validated the calibrated model and evaluated the effects of calibration under climate change using the SSP 3–7.0 climate change scenario. We find that incorporating empirical drivers of litter decomposition provides similar, and sometimes better (in terms of goodness-of-fit metrics), predictions of leaf litter decomposition but with different underlying ecological dynamics. For some sites, calibration also increased climate change-induced leaf litter mass loss by up to 5%, with implications for carbon cycle-climate feedbacks. Our work also provides an example for integrating data on the relative abundance of bacterial functional groups into an ecosystem model using a novel calibration method to bridge empiricism and process-based modeling, answering a call for the use of empirical microbial community data in process-based ecosystem models. We highlight that incorporating mechanistic information into models, as done in this study, is important for improving confidence in model projections of ecological processes like litter decomposition under climate change.

We leveraged 16S rRNA gene amplicon sequence data from soil samples at experiment initiation (temporally matching the litter lignin and N measurements) in 5–12 plots (depending on sampling extent and data quality) at each of the seven sites to obtain bacterial copiotroph:oligotroph ratios to be used in our statistical model (Polussa and Oliverio 2025). Five plots were used at GRSM and HARV, 8 at BART, 10 at TREE, 11 at LENO and TALL, and 12 at SERC. We use copiotroph and oligotroph groupings to represent the microbial community because these groups are represented in the process-based model used in this study but acknowledge there are multiple ways to represent functional traits of microbial communities. In brief, DNA was extracted from 200 to 700 mg soil using the Zymo Quick-DNA Fecal/Soil Microbe DNA Miniprep Kit and then amplified using a 250-bp fragment of the V4–V5 region of the 16S rRNA gene. Sample concentrations were normalized and sequenced on the Illumina MiSeq platform with 2 × 150-bp paired-end chemistry at the University of Colorado Next Generation Sequencing Facility along with negative controls to check for possible contamination. We processed the raw reads to amplicon sequence variants (ASVs) with the DADA2 pipeline (Callahan et al. 2016) as per Shepherd and Oliverio (2024). Samples were rarefied to 2447 reads per sample and bacterial copiotroph: oligotroph was calculated as the sum of copiotroph ASV counts over the sum of oligotroph ASV counts per plot.