AgMIP: Lessons Learned From Initial Crop Model Intercomparisons.

Crop model intercomparison has been one of the first activities of the Agriculture Model Intercomparison Project (AgMIP) with the goals of improved characterization of model uncertainty and climate change assessment. Wheat, maize, and rice model intercomparisons are the most complete, and sugarcane, soybean, groundnut, potato, sorghum, and millet model intercomparisons are at various stages of planning and implementation. Twenty-seven wheat models, about 20 maize models, and 13 rice models are participating, comparing non-calibrated and calibrated results and sensitivities to climate variables at sites that represent major growing regions around the world. Several types of scientific and procedural issues have arisen, including the need for high-quality field trial and climate inputs, definitional differences in simulation of crop processes, varying practices of model calibration, differences in interpretation of inputs, difficulties associated with different formats and definitions of inputs, and lack of understanding of trans-disciplinary linkages. For example, many crop models differ in simulation of evapotranspiration, and therefore ET cannot be generated centrally; rather it must be calculated by each model and considered as part of model uncertainty. Characterization of uncertainty through the use of climate change scenarios has also been a challenge. Key products arising from this activity are protocols for rigorous comparisons among different models and the AgMIP Sentinel Site data quality categories, with silver, gold, and platinum designating increasing level of quality for crop model development and intercomparison.