Technical ENSO Update

16 September 2010

Current Conditions

Expected Conditions

Presently, the models and observations taken together indicate probabilities of approximately 98% for maintaining La Niña conditions, near 2% for returning to ENSO-neutral conditions, and about 0.1% for developing El Niño conditions during the September-October-November season in progress. Going forward, probabilities for La Niña continue at 95% or greater for the three-month periods extending through December-January-February 2010/11. In northern spring 2011 the probabilities for La Niña begin to weaken at a faster rate, declining to about 62% by March-April-May, and to 37% for April-May-June.

The above assessment was made in part on the basis of an examination of the current predictions of ENSO prediction models as well as the observed conditions. For purposes of this discussion, El Niño SST conditions are defined as SSTs in the NINO3.4 region being in the warmest 25% of their climatological distribution for the 3-month period in question over the 1950-present timeframe. The corresponding cutoff in terms of degrees C of SST anomaly varies seasonally, being close to 0.40 degrees C in boreal late-spring to early-summer season and as high as 0.75 degrees C in late boreal autumn. La Niña conditions are defined as NINO3.4 region SSTs being in the coolest 25% of the climatological distribution. Neutral conditions occupy the remaining 50% of the distribution. These definitions were developed such that the most commonly accepted El Niño and La Niña episodes are reproduced.

The models show nearly unanimous agreement regarding very high probabilities for La Niña conditions over the remainder of 2010 and into the early months of 2011, but vary somewhat in their predictions of the strength of the event. Nearly all statistical and dynamical models call for at least moderate La Niña strength (stronger than -1C); A few models of either type predict a strong event. Collectively the average of the model predictions is for moderate to strong La Niña conditions with peak strength at approximately -1.6 C during the Oct-Nov-Dec and Nov-Dec-Jan seasons. For the current September-November season, 100% of the models are predicting La Niña conditions, while none predict ENSO-neutral conditions. This strong tilt of the odds weakens only very slightly for the Oct-Nov-Dec, Nov-Dec-Jan and Dec-Jan-Feb seasons with 5% of the models (i.e., one model) calling for cool-neutral conditions. Starting in Jan-Feb-Mar 2011, more models begin to predict neutral conditions, but not to the point of being a majority until late northern spring. At lead times of 4 or more months into the future, statistical and dynamical models that incorporate information about the ocean's observed subsurface thermal structure generally exhibit higher predictive skill than those that do not. Among models that do use subsurface temperature information, none indicate El Niño conditions for the Jan-Feb-Mar season, 3 of 15 (20%)  predict ENSO-neutral SSTs, and 12 of 15 (80%) predict La Niña conditions. (Note 1). (Note that La Niña conditions for Jan-Feb-Mar require a NINO3.4 SST anomaly of -0.55 or stronger.) Caution is advised in interpreting the distribution of model predictions as the actual probabilities. At longer leads, the skill of the models degrades, and skill uncertainty must be convolved with the uncertainties from initial conditions and differing model physics, leading to more climatological probabilities in the long-lead ENSO Outlook than might be suggested by the suite of models.  Furthermore, the expected skill of one model versus another has not been established using uniform validation procedures, which may cause a difference in the true probability distribution from that taken verbatim from the raw model predictions.

An alternative way to assess the probabilities of the three possible ENSO conditions is to use the mean of the predictions of all models, and to construct a standard error function centered on that mean. The standard error would be Gaussian in shape, and would have its width determined by an estimate of overall expected model skill for the season of the year and the lead time. Higher skill would result in a relatively narrower error distribution, while low skill would result in an error distribution with width approaching that of the historical observed distribution. This method shows probabilities for La Niña at about 97% for both October-December and November-January, remaining at 90% or more through February-April 2011. (Note that the threshold for La Niña weakens from approximately -0.75C to -0.45C between October-December and February-April, due to the seasonality of the interannual variance). The same cautions mentioned above for the distribution of model predictions apply to this alternative method of inferring probabilities, due to differing model biases and skills. In particular, this approach considers only the mean of the predictions, and not the range across the models, nor the ensemble range within individual models.

The IRI's probabilistic ENSO prediction takes into account the indications of this set of models, the outcome of the standard error approach described above, and additional factors such as the very latest observations that may have developed after the initialization times of some of the models. It indicates a 98% probability for La Niña conditions in the September-November season in progress, and only slightly less (still at least 95%) through December-February 2010/11. Probabilities for neutral ENSO conditions rise from 2% in September-November to 9% in January-March, and to 50% by April-June 2011. The probability for a return to El Niño conditions remain at very low levels from the present through early 2011, not exceeding 10% until April-June and reaching 23% by June-August.

See also: 

• Summary of model predictions