Testing CERES-wheat with free-air carbon dioxide enrichment (FACE) experiment data:: CO2 and water interactions

Dynamic crop-growth models are used to project the effects of rising atmospheric CO2 concentration and associated climate change on crop yields. Such model predictions are largely untested in the field, for lack of experimental data. We tested the CERES-Wheat model, modified to include leaf-level photosynthesis response to elevated CO2, using field data from 2 yr of Free-Air Carbon Dioxide Enrichment (FACE) experiments with spring wheat (Triticum aestivum L. cv. Yecora Rojo) in Mariclopa, AZ. Two irrigation treatments (well-watered, WW; water-deficit stressed, WS) and two atmospheric CO2 concentrations (ambient, 350 mu mol mol(-1); elevated, 550 mu mol mol(-1)) were simulated. The model was evaluated using measurements of crop phenology, aboveground dry matter (DM) production, grain yield, and evapotranspiration (ET). Model calculations of crop phenology were within 2 to 3 d of observed values under WW, ambient CO2 conditions in both years. The model did not simulate the accelerated crop phenology (5-8 d at physiological maturity) observed in the WW and elevated CO2 treatments, indicating the need to include effects of increased stomatal resistance on canopy temperature. Simulations of DM and grain yield were within 10% of measured values, except for a tendency to overcalculate DM response to CO2 by 10 to 15% in Year 1 for WS treatments. The model undercalculated cumulative ET under WW conditions by 15%; model sensitivity analyses suggest that simulation of potential evapotranspiration (PET) was too low for this arid site. The model reproduced measured dynamics of CO2-water interactions. Simulated reductions in water loss due to elevated CO2 were about 4%, in agreement with measurements. The model simulated larger increases in DM production and yield due to elevated CO2 under WS than under WW conditions. In Year 1, simulated crop response to CO2 was 2% larger (measured: 3%) under WS than under WW conditions; in Year 2, it was 11% larger (measured: 9%). The ability to simulate CO2-water interactions, though it needs to be further evaluated with additional experimental datasets, is an important attribute of models used to project crop yields under elevated CO2 and climate change.