Multimodel ensembles of wheat growth: many models are better than one.

Crop models of crop growth are increasingly used to quantify the impact of global changes due to climate or crop management. Therefore, accuracy of simulation results is a major concern. Studies with ensembles of crop models can give valuable information about model accuracy and uncertainty, but such studies are difficult to organize and have only recently begun. We report on the largest ensemble study to date, of 27 wheat models tested in four contrasting locations for their accuracy in simulating multiple crop growth and yield variables. The relative error averaged over models was 24-38% for the different end-of-season variables including grain yield (GY) and grain protein concentration (GPC). There was little relation between error of a model for GY or GPC and error for in-season variables. Thus, most models did not arrive at accurate simulations of GY and GPC by accurately simulating preceding growth dynamics. Ensemble simulations, taking either the mean (e-mean) or median (e-median) of simulated values, gave better estimates than any individual model when all variables were considered. Compared to individual models, e-median ranked first in simulating measured GY and third in GPC. The error of e-mean and e-median declined with an increasing number of ensemble members, with little decrease beyond 10 models. We conclude that multimodel ensembles can be used to create new estimators with improved accuracy and consistency in simulating growth dynamics. We argue that these results are applicable to other crop species, and hypothesize that they apply more generally to ecological system models.

How do various maize crop models vary in their responses to climate change factors?

Potential consequences of climate change on crop production can be studied using mechanistic crop simulation models. While a broad variety of maize simulation models exist, it is not known whether different models diverge on grain yield responses to changes in climatic factors, or whether they agree in their general trends related to phenology, growth, and yield. With the goal of analyzing the sensitivity of simulated yields to changes in temperature and atmospheric carbon dioxide concentrations [CO2 ], we present the largest maize crop model intercomparison to date, including 23 different models. These models were evaluated for four locations representing a wide range of maize production conditions in the world: Lusignan (France), Ames (USA), Rio Verde (Brazil) and Morogoro (Tanzania). While individual models differed considerably in absolute yield simulation at the four sites, an ensemble of a minimum number of models was able to simulate absolute yields accurately at the four sites even with low data for calibration, thus suggesting that using an ensemble of models has merit. Temperature increase had strong negative influence on modeled yield response of roughly -0.5 Mg ha(-1) per °C. Doubling [CO2 ] from 360 to 720 µmol mol(-1) increased grain yield by 7.5% on average across models and the sites. That would therefore make temperature the main factor altering maize yields at the end of this century. Furthermore, there was a large uncertainty in the yield response to [CO2 ] among models. Model responses to temperature and [CO2 ] did not differ whether models were simulated with low calibration information or, simulated with high level of calibration information.

Performance of several models for predicting budburst date of grapevine (Vitis vinifera L.).

The budburst stage is a key phenological stage for grapevine (Vitis vinifera L.), with large site and cultivar variability. The objective of the present work was to provide a reliable agro-meteorological model for simulating grapevine budburst occurrence all over France. The study was conducted using data from ten cultivars of grapevine (Cabernet Sauvignon, Chasselas, Chardonnay, Grenache, Merlot, Pinot Noir, Riesling, Sauvignon, Syrah, Ugni Blanc) and five locations (Bordeaux, Colmar, Angers, Montpellier, Epernay). First, we tested two commonly used models that do not take into account dormancy: growing degree days with a base temperature of 10 degrees C (GDD(10)), and Riou's model (RIOU). The errors of predictions of these models ranged between 9 and 21 days. Second, a new model (BRIN) was studied relying on well-known formalisms for orchard trees and taking into account the dormancy period. The BRIN model showed better performance in predicting budburst date than previous grapevine models. Analysis of the components of BRIN formalisms (calculation of dormancy, use of hourly temperatures, base temperature) explained the better performances obtained with the BRIN model. Base temperature was the main driver, while dormancy period was not significant in simulating budburst date. For each cultivar, we provide the parameter estimates that showed the best performance for both the BRIN model and the GDD model with a base temperature of 5 degrees C.