Multi‐Year Predictability of Global Sea Surface Temperature Using Model‐Analogs

Abstract

AbstractThe multi‐year predictability of global sea surface temperature (SST) is examined by applying a model‐analog method to four control simulations to make forecasts at leads of 1–36 months over 1961–2015. The forecasts are found to have skill for annual mean SST at Year 2 (i.e., leads of 13–24 months) or even Year 3 (i.e., leads of 25–36 months) in the tropical Pacific, the North and South Pacific, the southwest Indian Ocean, and the northwestern tropical Atlantic. The seasonality in forecast skill suggests that July is the best time to initialize multi‐year forecasts. The evolution of forecast skill for tropical Pacific SSTs indicates that the predictability of some El Niño Southern Oscillation events is nearly the same at leads of 6 and 18 or 24 months. Further, these El Niño and La Niña events can be predicted at leads of up to 24 and 30 months, respectively.