Abstract
We present a simple, generic method for parameterising the sampling and measurement uncertainty for time series data that is based on the number of observations that constitutes each data point. The method is demonstrated using basinβaveraged monthly time series of (1) the mean temperature above 220 m depth (T~220m~); (2) the mean temperature above the 14 Β°C isotherm (T~14C~) and (3) the depth of the 14 Β°C isotherm (D~14C~). The T~220m~ and T~14C~ analyses are in some sense βequivalentβ, since the spatialβtimeβmean depth of the 14 Β°C isotherm is approximately 220 m. In agreement with oceanographic theory, we find that the T~14C~ time series has a consistently lower associated sampling uncertainty than T~220m~ for all ocean basins considered in this study. We also note that the sampling uncertainty for all quantities for a given number of observations is dependent on ocean basin. Our results suggest that with the current observing array we are able to monitor global monthly values of T~220m~, T~14C~ and D~14C~ with a measurement and sampling uncertainty of approximately Β± 0.07 Β°C, Β± 0.04 Β°C and Β± 2.4 m, respectively. However, these estimated uncertainties will become much larger on moving to regional or smaller scales. Β© Crown Copyright 2010. Published by John Wiley & Sons Ltd.