Multi-metric evaluation of an ensemble of biogeochemical models for the estimation of organic carbon content in long-term bare fallow soils

As part of benchmarking actions at international level (FACCE-JPI project CN-MIP), the C-MIP action was initiated in 2016 to address the question of whether ensemble modelling could bring some improvement to the simulation of soil organic carbon (SOC) dynamics. A multi-model ensemble with 25 process‐based integrated C-N models was implemented to compare simulations (before and after model calibration) to SOC data from a network of six long-term bare fallow experimental sites (one site with two options) in Europe. To evaluate single models and the model ensemble, multiple evaluation metrics were aggregated into a single modular indicator, not only accounting for the agreement between model estimates and actual data but also taking into account their structural complexity. Illustrative results of simulated against observed SOC dynamics while also discussing the potential of the multi-metric aggregated indicator to help identifying areas where structural changes in models may be needed to better represent such dynamics.