Sensitivity analysis of C and N modules in biogeochemical cropand grassland models following manure addition to soil

International audience

anglais. Process-based crop and grassland models estimating carbon (C) and nitrogen(N) dynamics are widely used to investigate best management practices in agricul-ture. They integrate several processes in a complex structure, but studies wheremodules corresponding to specific processes extracted from the whole model struc-ture are assessed independently are uncommon. With the support of documentedaerobic incubation trials in manure-amended soils, a sensitivity analysis was per-formed on the C–N cycling processes of four modules (MOD1–4), correspondingto the models APSIM, EPIC, FASSET and STICS. The results showed that theparameter‘substrate use efficiency’had the most effect on the predicted values ofnet CO2emissions and net N mineralization, together with the C/N ratio of the soilmicrobial biomass. They explained 74–75% on average of both output variances,whereas parameters determining manure C and N partitioning and first-orderdecomposition constants of manure pools explained, on average, an additional17–19%. Efforts should be focused on calibrating these parameters for more accu-rate simulations. The greater sensitivity of both outputs to parameters related tomanure pools in more complex modules (MOD2–4) facilitates their adaptation tospecific contexts, whereas MOD1 probably requires that parameters related to soilpools are also adapted to specific applications. Parameter interactions were limited,becoming noticeable only in situations of N-limited soil organic matter decomposi-tion. Models MOD1 and MOD3 allowed the C/N ratio of the soil microbial bio-mass to vary temporarily; therefore, they were less sensitive to mineral Navailability and more easily adapted to a wide range of situations. This study pro-vides essential information to support the development of state-of-the-art biogeo-chemical models.