Previous - 3.4.2 Large-scale Sampling Uncertainty Index Next - 3.5 Reconstruction Techniques and Other Structural Choices
Uncertainties arising from under-sampling at a grid-box level are easy to assess if the observations can be assumed to be randomly distributed within a grid box. However, sampling is not random. The effect of this is reduced in most analyses by the calculation of super-observations that combine nearby measurements; however, optimal methods to minimize uncertainty are not generally applied. Simple estimates of large-scale sampling uncertainty in the global-average SST from subsampling well-sampled periods suggest a value of at most 0.2K even in poorly observed years. However, there are potential limitations of these simple methods and they should be considered together with the range of statistical reconstructions to get a more complete idea of uncertainty in large-scale averages.
Previous - 3.4.2 Large-scale Sampling Uncertainty Index Next - 3.5 Reconstruction Techniques and Other Structural Choices