Abstract: • First model inter-comparison on crop rotations. • Continuous simulation of multi-year crop rotations yields outperformed single-year simulation. • Low accuracy of yield predictions in less commonly modelled crops such as potato, radish, grass vegetation. • Multi-model mean prediction was found to minimise the likely error arising from single-model predictions. • The representation of intermediate crops and carry-over effects in the models require further research efforts.
Diversification of crop rotations is considered an option to increase the resilience of European crop production under climate change. So far, however, many crop simulation studies have focused on predicting single crops in separate one-year simulations. Here, we compared the capability of fifteen crop growth simulation models to predict yields in crop rotations at five sites across Europe under minimal calibration. Crop rotations encompassed 301 seasons of ten crop types common to European agriculture and a diverse set of treatments (irrigation, fertilisation, CO2 concentration, soil types, tillage, residues, intermediate or catch crops). We found that the continuous simulation of multi-year crop rotations yielded results of slightly higher quality compared to the simulation of single years and single crops. Intermediate crops (oilseed radish and grass vegetation) were simulated less accurately than main crops (cereals). The majority of models performed better for the treatments of increased CO2 and nitrogen fertilisation than for irrigation and soil-related treatments. The yield simulation of the multi-model ensemble reduced the error compared to single-model simulations. The low degree of superiority of continuous simulations over single year simulation was caused by (a) insufficiently parameterised crops, which affect the performance of the following crop, and (b) the lack of growth-limiting water and/or nitrogen in the crop rotations under investigation. In order to achieve a sound representation of crop rotations, further research is required to synthesise existing knowledge of the physiology of intermediate crops and of carry-over effects from the preceding to the following crop, and to implement/improve the modelling of processes that condition these effects.

Abstract: Eddy covariance data from four European grassland sites are used to probabilistically invert the CARAIB (CARbon Assimilation In the Biosphere) dynamic vegetation model (DVM) with 10 unknown parameters, using the DREAM((ZS)) (DiffeRential Evolution Adaptive Metropolis) Markov chain Monte Carlo (MCMC) sampler. We focus on comparing model inversions, considering both homoscedastic and heteroscedastic eddy covariance residual errors, with variances either fixed a priori or jointly inferred together with the model parameters. Agreements between measured and simulated data during calibration are comparable with previous studies, with root mean square errors (RMSEs) of simulated daily gross primary productivity (GPP), ecosystem respiration (RECO) and evapotranspiration (ET) ranging from 1.73 to 2.19, 1.04 to 1.56 g C m(-2) day(-1) and 0.50 to 1.28 mm day(-1), respectively. For the calibration period, using a homoscedastic eddy covariance residual error model resulted in a better agreement between measured and modelled data than using a heteroscedastic residual error model. However, a model validation experiment showed that CARAIB models calibrated considering heteroscedastic residual errors perform better. Posterior parameter distributions derived from using a heteroscedastic model of the residuals thus appear to be more robust. This is the case even though the classical linear heteroscedastic error model assumed herein did not fully remove heteroscedasticity of the GPP residuals. Despite the fact that the calibrated model is generally capable of fitting the data within measurement errors, systematic bias in the model simulations are observed. These are likely due to model inadequacies such as shortcomings in the photosynthesis modelling. Besides the residual error treatment, differences between model parameter posterior distributions among the four grassland sites are also investigated. It is shown that the marginal distributions of the specific leaf area and characteristic mortality time parameters can be explained by site-specific ecophysiological characteristics.