In weather data sets used by crop modellers, irregularities occur as inaccuracies in data or as missing values. In this investigation, the effect of such irregularities in temperature and global radiation data on simulation results is studied for a spring wheat crop growth simulation model. From the literature, the inaccuracy in temperature and global radiation data was estimated to be 1 °C and 10% respectively. Systematic over- or underestimation of the data using these values resulted in deviations in simulated yield of about 10%. Four methods of estimating missing values were compared: use of average values over 30 yr, over 1 mo and over 10 d, and use of daily data from another meteorological station. When all daily data were replaced by estimates, data from a nearby station gave the best results: only a small deviation in simulated yield was found. The use of averages resulted in overestimations of yield of up to 35% in some years. For global radiation data the effect of estimates based on sunshine duration data was also considered; use of these data gave a better result than data from a nearby station. When only 10% of the daily temperature and radiation data were replaced randomly by estimates, no effects on simulation results were found.