Previous - 3.4.3 Summary of Sampling Uncertainty Index Next - 3.5.1 Critique of Reconstruction Techniques
Creating global SST analyses is challenging because of the relative sparseness of observations before the satellite era and the non-stationarity of the changing climate. A large number of different SST data sets based on in situ data have been produced employing a variety of statistical methods. The structural uncertainties associated with estimating SSTs in data voids and at data-sparse times are therefore somewhat better explored than structural uncertainties in the pervasive systematic errors. Data sets used in this paper have been summarized in Table 5 and global averages for these data sets are shown in Figure 5.
| Data set | Input data set | Interpolation method | Resolution |
|---|---|---|---|
| ICOADS summaries [Woodruff et al., 2011] | ICOADS 2.5 | None | 2°x2° monthly |
| HadSST2 [Rayner et al., 2006] | ICOADS 2.1 | None | 5°x5° monthly |
| HadSST3 [Kennedy et al., 2011b; Kennedy et al., 2011c] | ICOADS 2.5 | None | 5°x5° monthly |
| HadSST4 [Kennedy et al. 2019] | ICOADS 3.0 | None | 5°x5° monthly |
| TOHOKU [Yasunaka and Hanawa, 2002] | ICOADS 2.1 | None | 5°x5° monthly |
| HadISST1.1 [Rayner et al., 2003] | Met Office Marine Databank and COADS, AVHRR satellite retrievals | Reduced Space Optimal Interpolation | 1°x1° monthly |
| ERSSTv3b [Smith et al., 2008] | ICOADS 2.1 | Separate low and high frequency reconstructions. High frequency component based on EOTs | 2°x2° monthly |
| ERSSTv4 [Huang et al., 2015] | ICOADS 2.5 | Separate low and high frequency reconstructions. High frequency component based on EOTs | 2°x2° monthly |
| ERSSTv5 [Huang et al., 2017] | ICOADS 3.0 | Separate low and high frequency reconstructions. High frequency component based on EOTs | 2°x2° monthly |
| COBE [Ishii et al., 2005] | ICOADS 2.1 and Kobe collection | Optimal interpolation | 1°x1° monthly |
| COBE-2 [Hirahara et al., 2013] | ICOADS 2.5 and Kobe collection, AVHRR satellite retrievals | Multi scale analysis based on EOFs | 1°x1° daily and monthly |
| Kaplan [Kaplan et al., 1998] | Met Office Marine Databank | Reduced Space Optimal Smoothing | 5°x5° monthly |
| NOCS [Berry and Kent, 2011] | ICOADS 2.5 | Optimal Interpolation | 1°x1° daily and monthly |
| VBPCA [Ilin and Kaplan, 2009] | ICOADS 2.5 | Variational Bayesian Principal Component Analysis | 5°x5° monthly |
| GPFA [Luttinen and Ilin, 2009] | ICOADS 2.5 | Gaussian Process Factor Analysis | 5°x5° monthly |
| GP [Luttinen and Ilin,2012] | ICOADS 2.5 | Gaussian Process | 5°x5° monthly |
Table 5: List of datasets used and referred to in the review.
Previous - 3.4.3 Summary of Sampling Uncertainty Index Next - 3.5.1 Critique of Reconstruction Techniques