Search

Crop models are increasingly used to simulate crop yields at the global scale, but so far there is no general framework on how to assess model performance. Here we evaluate the simulation results of 14 global gridded crop modeling groups that have contributed historic crop yield simulations for maize, wheat, rice and soybean to the Global Gridded Crop Model Intercomparison (GGCMI) of the Agricultural Model Intercomparison and Improvement Project (AgMIP). Simulation results are compared to reference data at global, national and grid cell scales and we evaluate model performance with respect to time series correlation, spatial correlation and mean bias. We find that global gridded crop models (GGCMs) show mixed skill in reproducing time series correlations or spatial patterns at the different spatial scales. Generally, maize, wheat and soybean simulations of many GGCMs are capable of reproducing larger parts of observed temporal variability (time series correlation coefficients (r) of up to 0.888 for maize, 0.673 for wheat and 0.643 for soybean at the global scale) but rice yield variability cannot be well reproduced by most models. Yield variability can be well reproduced for most major producing countries by many GGCMs and for all countries by at least some. A comparison with gridded yield data and a statistical analysis of the effects of weather variability on yield variability shows that the ensemble of GGCMs can explain more of the yield variability than an ensemble of regression models for maize and soybean, but not for wheat and rice. We identify future research needs in global gridded crop modeling and for all individual crop modeling groups. In the absence of a purely observation-based benchmark for model evaluation, we propose that the best performing crop model per crop and region establishes the benchmark for all others, and modelers are encouraged to investigate how crop model performance can be increased. We make our evaluation system accessible to all crop modelers so that other modeling groups can also test their model performance against the reference data and the GGCMI benchmark. [ABSTRACT FROM AUTHOR]

Reliable regional-scale representation of crop growth and yields has been increasingly important in earth system modeling for the simulation of atmosphere-vegetation-soil interactions in managed ecosystems. While the parameter values in many crop models are location specific or cultivar specific, the validity of such values for regional simulation is in question. We present the scale dependency of likely parameter values that are related to the responses of growth rate and yield to temperature, using the paddy rice model applied to Japan as an example. For all regions, values of the two parameters that determine the degree of yield response to low temperature (the base temperature for calculating cooling degree days and the curvature factor of spikelet sterility caused by low temperature) appeared to change relative to the grid interval. Two additional parameters (the air temperature at which the developmental rate is half of the maximum rate at the optimum temperature and the value of developmental index at which point the crop becomes sensitive to the photoperiod) showed scale dependency in a limited region, whereas the remaining three parameters that determine the phenological characteristics of a rice cultivar and the technological level show no clear scale dependency. These results indicate the importance of using appropriate parameter values for the spatial scale at which a crop model operates. We recommend avoiding the use of location-specific or cultivar-specific parameter values for regional crop simulation, unless a rationale is presented suggesting these values are insensitive to spatial scale. [ABSTRACT FROM AUTHOR]

This study quantifies the ranges of climate model biases in surface air temperature for July and August (summer temperature) and daily total insolation for May–October (warm-season insolation) that can give simulated regional paddy rice yields with a bias within ±2.5% of the 20-yr mean observed regional yield. The following four sets of three meteorological elements (daily maximum and minimum temperatures and daily total insolation) from daily climate model outputs were used as meteorological inputs for a large-scale crop model for irrigated paddy rice: 1) raw climate model outputs of all meteorological elements, 2) bias-corrected temperatures and raw climate model outputs of insolation, 3) bias-corrected insolation and raw climate model outputs of temperatures, and 4) bias-corrected climate model outputs of all meteorological elements. These meteorological inputs were sourced from seven coupled general circulation models, one regional climate model, and one reanalysis dataset. Crop model simulations with artificially biased meteorological inputs were also used. By using the approximation formula derived from these crop model simulation results and the Monte Carlo simulation technique, it was found that climate model outputs with biases within ±0.6°C and ±3% for summer temperature and warm-season insolation, respectively, could result in a simulated regional paddy rice yield with a bias within ±2.5% of the 20-yr mean observed regional yield. The simulated regional yield was less biased not only when the biases of two meteorological inputs were small but also when the cold or warm bias of summer temperature and the overestimation of warm-season insolation were balanced through the crop model processes. The methodology presented here will lead to a better and more comprehensive understanding of the nature of error propagation from a climate model to an application model and will facilitate the selection of climate models suitable for specific applications. [ABSTRACT FROM AUTHOR]

• The global gridded crop model (GGCM) performance is evaluated at national level. • Years of irrigated and rainfed simulations need be matched with national reports. • Planting window is considered using the planting date ensemble simulations. • The quality of crop statistics, misreports and conversion errors are explored. • This study focuses on rice but has implications for evaluations for other crops. Global gridded crop models (GGCMs) are increasingly used for climate risk assessments and adaptation planning in agriculture. GGCM historical simulation performance is therefore crucial for such applications. However, GGCM performance is lower for rice than for other crops at an aggregated administrative unit level despite the lack of a clear difference in model performance at the site level. Here, we present key factors that need to be considered in the spatial and temporal aggregation of GGCM outputs to improve the evaluation of GGCM historical rice simulations at the country scale. The factors include an adjustment for the harvesting year in GGCM rainfed and irrigated simulations, the removal of misreports from reference data, a consideration of the quality of national crop statistics (census reports), and the explicit incorporation of a planting window. The effect of each individual factor is demonstrated by analyzing a multi-GGCM dataset and performing a planting date ensemble simulation of two GGCMs. We reveal that among others, aligning the harvesting year in the GGCM simulations with that in national reports is pivotal. Although our analysis focuses specifically on rice, the findings of this study are useful for improving the country-level evaluations of GGCM historical simulations for other crops. [ABSTRACT FROM AUTHOR]