Measurement interval and the accurate assessment of ground temperature trends

High frequency, multi‐year ground temperature measurements (every eight hours) obtained for the Norman Wells to Zama Pipeline Thermal Monitoring Program were used to create synthetic ‘manual’ data subsets with measurement intervals between 30 and 90 days. Mean annual temperatures were estimated from the subsets using cubic spline interpolation. By repeatedly producing collections of subsets with a given long term measurement interval, the uncertainty about the underlying temperature wave when sampled at that interval could be summarized using standard statistical measures (averages and standard deviations). Analysis of data from 0.5–5.3 m deep sensors indicates a predictable pattern of uncertainty for any given measurement interval. The annual temperature regime is estimated with increasing precision at greater depths for a given time interval. The standard deviation of the temperature estimates tends to be greatest where the change in curvature of the true temperature time series is greatest. At all depths, the standard deviation of both the instantaneous and the mean annual temperature estimates is a nearly constant proportion of the annual variation at that depth. The magnitude of error that will result for a given measurement interval can be estimated using the annual temperature range. © 1996 John Wiley & Sons, Ltd.