Modelling agroecosystem nitrogen functions provided by cover crop species in bispecific mixtures using functional traits and environmental factors

Cover crops are used during fallow periods to produce ecosystem services, especially those related to N management such as (i) capturing mineral-N from soil to reduce nitrate leaching, and (ii) improving N availability for the next main crop (green manuring). Bispecific mixtures consisting of legume and non-legume species could simultaneously produce these two services of nitrate saving and green manuring. The magnitude of these services can be estimated from indicators of agroecosystem functions such as crop growth rate, crop N acquisition rate and the C:N ratio of the cover crop. We developed a conceptual model for each indicator which was described using general linear models. A three-step procedure was used: (1) represent the behavior of each species based on a sub-model and calibrate each species in bispecific mixtures; (2) validate the complete-mixture models, corresponding to the sum of the two species sub-models, and the proportion of each species in the whole cover, and (3) validate the generality of sub-models and complete-mixture models to predict the agroecosystem function indicators of species in mixture not used for calibration. The combined use of (i) potential agroecosystem functions measured in sole crop in non-limiting conditions, (ii) difference in leaf functional traits, as indicators of plant strategies and (iii) environmental factors, was efficient in fitting and predicting the level of agroecosystem functions provided by a cover crop species in mixture in actual conditions. The models fitted for bispecific mixtures were efficient to represent the behavior of each species in mixture and to estimate the legume proportion which expressed the species dominance. The models were evaluated as satisfactory for crop growth rate and C:N ratio for their generality in predicting the agroecosystem functions provided in mixtures by other species not used in the model calibration step, which illustrates the relevance and robustness of the approach.