Large parameter uncertainty persists in soil food web models integrating stoichiometry, respiration and assimilation

Abstract Soil food web models are essential for linking animal diversity to carbon and nutrient cycling. However, it is unclear how soil animals regulate excess carbon when nutrient availability limits their growth and how to best represent animal respiration in these models. To address these issues, we developed an improved soil food web model. We use it to test two alternative mechanisms for handling nutrient limitation: (1) a respiration‐adjustment mechanism, in which excess carbon is released through increased respiration, and (2) an assimilation‐adjustment mechanism, in which carbon assimilation efficiency is reduced. We also compared two approaches for modelling respiration: biomass‐based respiration and consumption‐based respiration. Using data from three tree plantations in Brazil, we parameterized the model and performed sensitivity analyses across parameter sets and the four alternative model structures. Parameter uncertainty was large with fluxes from the best fitting parameter sets ranging from 0.5% to 2000% of the predicted mean. Our results show that parameters governing animal diet and physiological traits are responsible for the largest uncertainty in predicted fluxes. Model structural choices mattered to a lesser extent, producing variation in outputs that were 3%–35% of parameter uncertainty. However, this variation was still large. Changing the nutrient limitation mechanism from the assimilation‐adjusted to respiration‐adjusted model increased the contribution of animals to carbon loss by between 16% and 37%, while different consumption‐based and biomass‐based respiration formulations gave estimates of indirect effects that were two orders of magnitude apart. While microbes dominated total mineralization, soil macrofauna contributions were significant, with termites, earthworms and beetles as major contributors. We propose a 13 C tracer experiment that could distinguish between the model structures. Our findings underscore the need for targeted measurements of animal diet and physiology to identify dominant mechanisms as well as for higher taxonomic resolution in animal feeding preferences. This explicit link between model development and empirical research would enhance the reliability of soil food web model predictions and make them more useful for ecosystem management and biodiversity conservation. Read the free Plain Language Summary for this article on the Journal blog.