In the ESPACE-Wheat programme, 25 open-top chamber experiments were carried out in 1994, 1995 and 1996, on nine locations, divided over eight European countries. In most experiments, spring wheat cv. Minaret was subjected to two levels of atmospheric CO2 and two levels of ozone. Grain yields in the control treatments (ambient levels of CO2 and O3) varied strongly between sites. Also, yield response to elevated CO2 and O3 showed great variation. The present study was conducted to determine whether climatic differences between sites could account for the observed variation.Two simulation models were used for the analysis: AFRCWHEAT2-O3 and LINTULCC. AFRCWHEAT2-O3 simulates phenology, canopy development and photosynthesis in greater detail than LINTULCC. Both models account for the effects of radiation and temperature on crop growth. New algorithms were developed to simulate the effects of CO2 and O3. Weather data that were measured in the experiments were used as input, and simulated growth responses to CO2 and O3 were compared with measurements. No attempt was made to merge the two models. Thus two independent tools for analysis of data related to climate change were developed and applied.The average measured grain yield in the control treatment, across all 25 experiments, was 5.9 tons per hectare (tha-1), with a standard deviation (SD) of 1.9t ha-1. The models predicted similar average yields (5.5 and 5.8tha-1 for AFRCWHEAT2-O3 and LINTULCC, respectively), but smaller variation (SD for both models: 1.2tha-1). Average measured yield increase due to CO2-doubling was 30% (SD 22%). AFRCWHEAT2-O3 expected a slightly lower value (24%, SD 9%), whereas LINTULCC overestimated the response (42%, SD 11%). The average measured yield decrease due to nearly-doubled O3 levels was 9% (SD 11%). Both models showed similar results, albeit at lower variation (7% yield decrease at SDs of 6 and 4%). Simulations accounted well for the observation that, at elevated CO2, the percentage yield loss due to O3 was lower than at ambient CO2.The models predicted lower variation among sites and years than was measured. Yield response to CO2 and O3 was predicted to depend on the climate, with a predominant effect of temperature on the response to CO2. In the measurements, these climatic effects were indeed observed, but a greater part of the variation was not related to light intensity, temperature, CO2, or O3. This unexplained variability in the measured dataset was probably caused by factors not accounted for in the models, possibly related to soil characteristics.We therefore conclude that even perfect information on the climate variables examined in ESPACE-Wheat, i.e. light intensity and temperature, by itself would be insufficient for accurate prediction of the response of spring wheat to future elevated levels of CO2 and O3.