AbstractThe sampling frequency of temperature data is examined. The range of sampling from hourly to twice daily are explored to determine the uncertainty that is introduced by sampling less frequently than hourly in the determination of daily temperature and daily temperature extrema. The standards for comparison for the daily average temperature are the average of hourly data and the average of the daily maximum and minimum temperature. Hourly temperature data from 12 Canadian climate stations are examined for several decades leading up to 2017. Daily average temperatures were calculated using data sampled 24 times a day (hourly), 12 times, 8 times, 6 times, 4 times and twice daily. Two triad algorithms from the literature and an experimental one are assessed relative to these sampling frequencies. The sampling frequency analysis was remarkably consistent across all climate stations. The departure from the hourly estimate ranged from 0.1°C for the bi‐hourly sampling to ~1°C for the twice daily sampling. The uncertainty associated with the min/max method consistently fell within that of three and four samples per day. Comparison of triad algorithms, based on a quantitative criterion for determination of best sampling hours, revealed a station specific triad that outperforms algorithms from the literature and thrice daily evenly spaced sampling. Minimum and maximum estimates were compared across the different sampling frequencies for all stations as well. The accuracy of estimating temperature extrema decreases with lower sampling rates with the exception of the 8 hr sampling where hour of sampling influences accuracy. The results demonstrate that the local climate characteristics needs to be considered when choosing the optimal sampling frequency and calculation method for daily means and extrema.