After a false alert in the northern winter 2014/15, in 2015/16 one of the strongest El Niño events on record occurred in terms of the standard Niño 3.4 (N3.4) sea surface temperature anomaly index. A study based on reanalyses revealed that the 2015/16 El Niño was also associated with unusual energy transfers (see ECMWF Newsletter No. 155). Most notably, the Indonesian Throughflow (ITF) was exceptionally weak, which led to the retention of warm waters in the Pacific Ocean. The weak ITF appeared related to the relatively high sea level in the Indian Ocean with respect to the Western Pacific.
Recent results confirm the earlier hypothesis that the exceptionally warm Indian Ocean state in 2014 was responsible for the unprecedented reduction of observed ITF transports and retention of heat in the Pacific. The Indian Ocean state also decreased the probability of an extreme warm event in the Pacific in 2014/15 and favoured the probability of warm conditions in 2015/16, as later materialised. The results demonstrate the importance of the Indian Ocean low-frequency variability and trends for seasonal forecasts. They also highlight the potential merit of two-year-long ENSO (El Niño – Southern Oscillation) predictions, as forecasters may have interpreted the predictions in 2014 differently with information of lead times beyond year one.
Impact of Indian Ocean on 2-year ENSO predictions
We investigated the impact of the anomalously warm Indian Ocean state in 2014 on the subsequent evolution of ENSO and its energetics with two sets of two-year long twin seasonal forecast experiments. The reference experiments were initialised in February 2014 and 1997, using the SEAS5 forecasting system.
These two initial dates were chosen because the state in the Pacific was similar, but the Indian Ocean was much cooler in 1997. In the perturbed experiments, we swapped the Indian Ocean initial conditions between these two years.
Consideration of predicted ENSO probabilities shows that the two reference experiments are able to discern the contrasting ENSO evolution during 1997–99 and 2014–16 (see the figure). The forecasts starting in Feb 1997 predict a 100% chance of an El Niño event in year 1 (Dec 1997) with a 48% chance of that event being extreme. In year 2 (Dec 1998), cool La Niña conditions are predicted with a 50% probability. The observed outcome was indeed an extreme warming in 1997/98 followed by strong cold conditions in 1998/99. The forecasts starting in Feb 2014 predict a 98% chance of warming in year 1, but only an 18% chance of a strong El Niño, much lower than the 1997 prediction. For year 2 (Dec 2015), they predict high probabilities (58%) for the continuation of warm conditions and only a small chance of a La Niña event (4%).
The perturbed forecasts exhibit a marked shift in ENSO probabilities compared to their respective reference, especially in year 2. The perturbed 2014–16 forecast shows a more than doubled (44%) probability of a strong El Niño event in year 1 compared to the reference forecast (see the figure). Diagnostics and further experiments showed that the main reason for this change is the increased probability of a positive Indian Ocean Dipole event, which promoted development of a strong El Niño event through atmospheric teleconnections. In year 2, the perturbed forecasts exhibit an enhanced probability of a La Niña event (18%). Analysis of the ocean heat budget (not shown) reveals that the perturbed forecast produces a much stronger Pacific heat loss than the reference forecast during 2014–15, mainly due to much stronger ITF transports in the perturbed experiment, thus preparing the ground for the switch to cooler La Niña conditions. The perturbed 1997–98 experiment confirms that the Indian Ocean initial conditions matter for the 24-month ENSO predictions. Compared to the reference forecast, the likelihood of a La Niña event in year 2 is much reduced (20%). Similar to the situation in 2014–16, this is due to much weaker Pacific Ocean heat loss, which is caused by reduced ITF transports.
For further details, see our article in Climate Dynamics published in 2021, https://doi.org/10.1007/s00382-020-05607-6.