Abstract Forest fires alter soil organic carbon and suppress soil respiration for decades following disturbance. However, uncertainties in model parameterization and sensitivity hinder robust predictions of autotrophic and heterotrophic soil respiration responses. We addressed this challenge using a novel dataset from a fire chronosequence in the Yukon and Northwest Territories of Canada. The dataset included field measurements of total soil respiration at four sites with varying time since fire, supplemented by field measurements of soil temperature at two depths, remote sensing data on aboveground productivity, and soil moisture at two depths. We evaluated a suite of soil respiration models, ranging from exponential Q₁₀ formulations to heterotrophic respiration models using Michaelis–Menten kinetics. To estimate parameters efficiently, we (1) derived algebraic expressions for soil respiration components assuming quasi-steady state dynamics and (2) applied a Markov Chain Monte Carlo (MCMC) approach for parameter estimation. The resulting parameter estimates revealed which parameters were well-constrained and where uncertainty remained. Modeled respiration agreed with established empirical relationships and highlighted two key findings: (1) all chronosequence sites favored models that explicitly included microbial carbon as a distinct pool, and (2) parameters related to aboveground litter inputs were better constrained than those for root turnover. These results held regardless of soil depth or the form of the autotrophic respiration moisture response. These findings indicate that direct field measurements of litterfall rates would reduce model uncertainty, and that targeted sampling during seasonal transitions (e. g. , freeze–thaw periods) would provide critical constraints on microbial activity when respiration dynamics are most variable.
Zobitz et al. (Thu,) studied this question.