This page shows real-time reconstructions and forecasts of SST using the ERSSTv5 dataset. The time series used for reconstruction and prediction are the 11 principal components from a time-extended, rotated EOF analysis of seasonal (non-overlapping) SSTs from 1900-present. Leading patterns of variability include four ENSO modes, 2 Pacific decadal patterns, Atlantic multidecadal variability, and the nonstationary trend. This page updates quarterly in March, June, September, and December. All anomalies displayed are relative to the 1981-2010 base period. Note that the ENSO, Decadal, and Trend components will not add up to the Total reconstructed or forecast field. This is for two reasons: 1) There are three other principal components, including NAO-based SST variability, that do not fall into those categories, and 2) the reconstructions are relative to the 1900-present climatology, and then adjusted to be relative to the current WMO climatology, but this adjustment is done after the sum of anomaly contributions in all cases. The figure displaying the autocorrelation curves of the leading prinicpal components is here.
Reconstructions | Forecasts |
---|---|
Total | Total |
ENSO | ENSO |
Decadal | Decadal |
Trend | Trend |
The next set of figures shows the foreast of 2-m temperature (GHCN+CAMS) as standardized anomalies for the next three overlapping seasons using the various PCs are predictors in a linear regression model. The regression coefficients are determinied from 1950-present. All standardized anomalies are relative to the 1981-2010 base period.
Lead-1 | Lead-2 | Lead-3 |
---|---|---|
Total | Total | Total |
ENSO | ENSO | ENSO |
Decadal | Decadal | Decadal |
Trend | Trend | Trend |