While a 5 year average is visually compelling, a more rigorous statistical method to determine any trend is to apply a line of best fit to the data. (skepticalscience.com)
But then one wonders what happens if other non-predicted regions were instead hand corrected — is their Pacific spot special or could one pin down other parts of the surface and generate even better fits to the data? (judithcurry.com)
Both models showed good fit to the data, but the more parsimonious one - factor model was retained. (link.springer.com)