The effect of cold air advection on the estimates of daily thermal time with a base temperature of 8°C (DTT8), calculated using maximum and minimum screen temperatures (max./min. values), was examined. Screen temperature data for at least three maize-growing seasons (October—April) at Cedara, Dundee and Potchefstroom were obtained from thermohygrograph charts. Two-hourly values on the charts were used to calculate unbiased estimates of DTT8. The max./min. and two-hourly estimates of DTT8were closely correlated. Mean bias errors (MBE) associated with the max./min. DTT8, which were defined as being equal to the difference between mean max./min. DTT8and mean two-hourly DTT8, were positive for all seasons at the three sites. MBE in max./min. DTT8was positively related to the fraction of days experiencing cold air advection (FDCAA), withr2= 0.64. Since cold air advection in South Africa is most often due to the movement inland of cold marine air, proximity to the sea may be a strong indicator of MBE in max./min. DTT8. The Cedara and Dundee sites, which are relatively close to the coastline, experienced cold air advection more often than did Potchefstroom (mean FDCAA of 0.229, 0.225and 0.144, respectively) and had greater MBE in max./min. DTT8(mean MBE of 0.62, 0.70 and 0.37°C d, respectively). Using the max./min. and two-hourly DTT8, comparisons were made of predicted silking and maturity dates for dryland maize generated by the CERES-Maize V.2.10 crop model. The largest under-predictions of silking date (–4.9 days on average) and physiological maturity (–13.3 days on average) when using max./min. DTT8were generated for Dundee, followed by Cedara (−3.9 and−7.7) and then Potchefstroom (−1.2 and−2.2). The results indicate that caution must be exercised when using max./min. estimates of daily thermal time for predicting maize growth and development at sites where cold air advection occurs frequently. At such sites, estimates of thermal time should be based on frequently-sampled air temperatures.