Abstract: Multi-model forecasting has drawn some attention in crop science for evaluating effect of climate change on crop yields. The principle is to run several individual process-based crop models under several climate scenarios in order to generate ensembles of output values. This paper describes a simple Bayesian method – called Bayes linear method- for updating ensemble of crop model outputs using yield observations. The principle is to summarize the ensemble of crop model outputs by its mean and variance, and then to adjust these two quantities to yield observations in order to reduce uncertainty. The adjusted mean and variance combine two sources of information, i.e., the ensemble of crop model outputs and the observations. Interestingly, with this method, observations collected under a given climate scenario can be used to adjust mean and variance of the model ensemble under a different scenario. Another advantage of the proposed method is that it does not rely on a separate calibration of each individual crop model. The uncertainty reduction resulting from the adjustment of an ensemble of crop models to observations was assessed in a numerical application. The implementation of the Bayes linear method systematically reduced uncertainty, but the results showed the effectiveness of this method varied in function of several factors, especially the accuracy of the yield observation, and the covariance between the crop model output and the observation. (C) 2015 Elsevier B.V. All rights reserved.

Abstract: Process-based models (PBM) for simulation of weather dependent grass growth can assist farmers and plant breeders in addressing the challenges of climate change by simulating alternative roads of adaptation. They can also provide management decision support under current conditions. A drawback of existing grass models is that they do not take into account the effect of winter stresses, limiting their use for full-year simulations in areas where winter survival is a key factor for yield security. Here, we present a novel full-year PBM for grassland named BASGRA. It was developed by combining the LINGRA grassland model (Van Oijen et al., 2005a) with models for cold hardening and soil physical winter processes. We present the model and show how it was parameterized for timothy (Phleum pratense L.), the most important forage grass in Scandinavia and parts of North America and Asia. Uniquely, BASGRA simulates the processes taking place in the sward during the transition from summer to winter, including growth cessation and gradual cold hardening, and functions for simulating plant injury due to low temperatures, snow and ice affecting regrowth in spring. For the calibration, we used detailed data from five different locations in Norway, covering a wide range of agroclimatic regions, day lengths (latitudes from 59 degrees to 70 degrees N) and soil conditions. The total dataset included 11 variables, notably above-ground dry matter, leaf area index, tiller density, content of C reserves, and frost tolerance. All data were used in the calibration. When BASGRA was run with the maximum a-posteriori (MAP) parameter vector from the single, Bayesian calibration, nearly all measured variables were simulated to an overall normalized root mean squared error (NRMSE) <0.5. For many site x experiment combinations, NRMSE was <0.3. The temporal dynamics were captured well for most variables, as evaluated by comparing simulated time courses versus data for the individual sites. The results may suggest that BASGRA is a reasonably robust model, allowing for simulation of growth and several important underlying processes with acceptable accuracy for a range of agroclimatic conditions. However, the robustness of the model needs to be tested further using independent data from a wide range of growing conditions. Finally we show an example of application of the model, comparing overwintering risks in two climatically different sites, and discuss future model applications. Further development work should include improved simulation of the dynamics of C reserves, and validation of winter tiller dynamics against independent data. (C) 2016 Elsevier B.V. All rights reserved.

Abstract: Crop models, used to make projections of climate change impacts, differ greatly in structural detail. Complexity of model structure has generic effects on uncertainty and error propagation in climate change impact assessments. We applied Bayesian calibration to three distinctly different empirical and mechanistic wheat models to assess how differences in the extent of process understanding in models affects uncertainties in projected impact. Predictive power of the models was tested via both accuracy (bias) and precision (or tightness of grouping) of yield projections for extrapolated weather conditions. Yields predicted by the mechanistic model were generally more accurate than the empirical models for extrapolated conditions. This trend does not hold for all extrapolations; mechanistic and empirical models responded differently due to their sensitivities to distinct weather features. However, higher accuracy comes at the cost of precision of the mechanistic model to embrace all observations within given boundaries. The approaches showed complementarity in sensitivity to weather variables and in accuracy for different extrapolation domains. Their differences in model precision and accuracy make them suitable for generic model ensembles for near-term agricultural impact assessments of climate change.