ENSO Forecast Navigation

ENSO Forecasts

ENSO Forecast

December 2021 Quick Look

Published: December 20, 2021

A monthly summary of the status of El Niño, La Niña, and the Southern Oscillation, or ENSO, based on the NINO3.4 index (120-170W, 5S-5N)

In mid-December, Sea Surface Temperatures remain well below normal in the central-eastern equatorial Pacific. The evolution of key oceanic and atmospheric variables is consistent with weak La Niña conditions, and therefore, a La Niña Advisory remained in place for Dec 2021. A large majority of the models predict SSTs to stay below-normal during boreal winter, and then return to ENSO-neutral levels during spring. Similar to the most-recent official CPC/IRI ENSO Outlook issued on December 9, 2021, this objective model-based ENSO outlook also anticipates a continuation of the weak La Niña event with high probability during Dec-Feb, persisting until Feb-Apr, dissipating in Mar-May (36%) and return to ENSO-neutral conditions with high probabilities for rest of the forecast period.

Historically Speaking

    El Niño and La Niña events tend to develop during the period Apr-Jun and they
  • Tend to reach their maximum strength during October - February
  • Typically persist for 9-12 months, though occasionally persisting for up to 2 years
  • Typically recur every 2 to 7 years

IRI ENSO Forecast

IRI Technical ENSO Update

Published: December 20, 2021

Note: The SST anomalies cited below refer to the OISSTv2 SST data set, and not ERSSTv5. OISSTv2 is often used for real-time analysis and model initialization, while ERSSTv5 is used for retrospective official ENSO diagnosis because it is more homogeneous over time, allowing for more accurate comparisons among ENSO events that are years apart. These two products may differ, particularly during ENSO events. The difference between the two datasets may be as much as 0.5 C. Additionally in some years, the ERSSTv5 may tend to be cooler than OISSTv2 in the context of warming trends, because ERSSTv5 is expressed relative to a base period that is updated every 5 years, while the base period of OISSTv2 is updated every 10 years. In February 2021, both datasets were updated to reflect the 1991-2020 climatology period.

Recent and Current Conditions

La Niña conditions are well established, with Sea Surface Temperatures (SST) well below average across most of the equatorial Pacific Ocean. The November SST anomaly for NINO3.4 region (5S-5N; 170W-120W) was -0.89 C, and for Sep-Nov season it was -0.80. The IRI’s definition of a weak La Niña, like NOAA/Climate Prediction Center’s, requires that the SST anomaly in the NINO3.4 region be between  -0.5 C and -1.0 C. The most recent weekly SST anomaly in the NINO3.4 region for the week ending 8 December 2021 was -1.1 C. Subsurface temperatures in the eastern equatorial Pacific remain below-average, the traditional and equatorial Southern Oscillation Indices show sustained positive values, and above-normal Trade Winds are observed near and west of the Date Line. The upper-level westerly wind anomalies that would accompany a large-scale response to La Niña conditions are present, together with reduced cloudiness near the date line and increased rainfall over Indonesia, all of which are consistent with weak La Niña conditions.

In summary, the most recent observations of key oceanic and atmospheric variables indicate well established, though weak La Niña conditions. A La Niña advisory from CPC remained in effect.

Expected Conditions

Note – Only models that produce a new ENSO prediction every month are considered in this statement.

What is the outlook for the ENSO status going forward? The most recent official diagnosis and outlook was issued on 09 December 2021 in the NOAA/Climate Prediction Center ENSO Diagnostic Discussion, produced jointly by CPC and IRI. It states that La Niña conditions continued during the month of November, and highly favored during the Northern Hemisphere winter, gradually decreasing and transitioning to ENSO-Neutral in spring of 2022.

The latest set of model ENSO predictions from mid-December is now available in the the IRI/CPC ENSO prediction plume. These are used to assess the probabilities of the three possible ENSO conditions by using the average value of the NINO3.4 SST anomaly predictions from all models in the plume, equally weighted. Currently, however, the NASA-GMAO model is not factored into the probabilistic update, even though it appears on the ENSO plume-of-models graphic. A standard Gaussian error is imposed over that average forecast, and its width is determined by an estimate of overall expected model skill for the season of the year and the lead time. Higher skill results in a relatively narrower error distribution, while low skill results in an error distribution with width approaching that of the historical observed distribution.

Using this method, La Niña is highly probable (91%) during Dec-Feb, while chances for ENSO-neutral is just 9%. Going forward, probabilities for La Niña decrease to 80% for Jan-Mar, 59% for Feb-Apr, 36% for Mar-May, and less than the La Niña climatological threshold probabilities for the rest of the forecast period. While highly probable until boreal spring, the plume diagram indicates a gradual further weaking of the current La Niña event.  Chances of the ENSO-neutral state rise above 50% beginning in Mar-May, reaching above 71% for Apr-Jun, and decreasing afterwards; thus, the ENSO-neutral state becomes the most likely outcome from Mar-May 2022 onwards. El Niño probabilities start at 1% in Mar-May and reach up to 29% at the end of the forecast period (Aug-Oct). A plot of the probabilities summarizes the forecast evolution. The climatological probabilities for La Niña, neutral, and El Niño conditions vary seasonally, and are shown in a table at the bottom of this page for each 3-month season.

Caution is advised in interpreting the forecast distribution from the Gaussian standard error as the actual probabilities, due to differing biases and performance of the different models. In particular, this approach considers only the mean of the predictions, and not the total range across the models, nor the ensemble range within individual models. At longer leads, the skill of the models degrades, and uncertainty in skill must be convolved with the uncertainties from initial conditions and differing model physics, which leads 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.

In summary, the probabilities derived from the models in the IRI/CPC plume indicate a high preference for a weak La Niña relative to neutral conditions during boreal winter and possibly extending into the early spring of 2022, after which ENSO-neutral conditions becomes the most likely outcome through the remaining forecast periods. The likelihood for El Niño development remains very low during winter and spring time; however, it increases up to 29% at the end of the forecast period.

A caution regarding the model-based ENSO plume predictions released mid-month, is that factors such as known specific model biases and recent changes in the tropical Pacific that the models may have missed, are not considered. This approach is purely objective. Those issues are taken into account in the official outlooks, which are generated and issued early in the month by CPC and IRI, and which will include some human judgment in combination with the model guidance.

IRI ENSO Forecast Histogram Image
Season La Niña Neutral El Niño
DJF 91 9 0
JFM 80 20 0
FMA 59 41 0
MAM 36 63 1
AMJ 25 71 4
MJJ 22 66 12
JJA 22 57 21
JAS 22 52 26
ASO 25 46 29

ENSO Forecast

IRI Model-Based Probabilistic ENSO Forecast

Published: December 20, 2021

A purely objective ENSO probability forecast, based on regression, using as input the model predictions from the plume of dynamical and statistical forecasts shown in the ENSO Predictions Plume. Each of the forecasts is weighted equally. It is updated near or just after the middle of the month, using forecasts from the plume models that are run in the first half of the month. It does not use any human interpretation or judgment. This is updated on the third Thursday of the month.

IRI ENSO Forecast Histogram Image

Season La Niña Neutral El Niño
DJF 91 9 0
JFM 80 20 0
FMA 59 41 0
MAM 36 63 1
AMJ 25 71 4
MJJ 22 66 12
JJA 22 57 21
JAS 22 52 26
ASO 25 46 29

ENSO Forecast

IRI ENSO Predictions Plume

Published: December 20, 2021

Note on interpreting model forecasts

The following graph and table show forecasts made by dynamical and statistical models for SST in the Nino 3.4 region for nine overlapping 3-month periods. Note that the expected skills of the models, based on historical performance, are not equal to one another. The skills also generally decrease as the lead time increases. Thirdly, forecasts made at some times of the year generally have higher skill than forecasts made at other times of the year--namely, they are better when made between June and December than when they are made between February and May. Differences among the forecasts of the models reflect both differences in model design, and actual uncertainty in the forecast of the possible future SST scenario.

Interactive Chart

You can highlight a specific model by hovering over it either on the chart or the legend. Selecting An item on the legend will toggle the visibility of the model on the page. You can also select DYN MODELS or STAT MODELS to toggle them all at once. Clicking on the "burger" menu above the legend will give you options to download the image or expand to full screen. If you have any feedback on this new feature, please let us know at webmaster@iri.columbia.edu.

List of Models Used

Forecast SST Anomalies (deg C) in the Nino 3.4 Region

  Seasons (2021 – 2022)
Model DJF JFM FMA MAM AMJ MJJ JJA JAS ASO
Dynamical Models
AUS-ACCESS -0.90 -0.70 -0.43 -0.20          
BCC_CSM11m
CMC CANSIP -1.11 -1.01 -0.78 -0.56 -0.33 -0.10 0.12 0.27 0.34
COLA CCSM4 -1.12 -1.05 -0.81 -0.49 -0.23 -0.08 -0.03 -0.04 -0.07
CS-IRI-MM -0.77 -0.56 -0.29 -0.06 0.14 0.29      
DWD
ECMWF -0.94 -0.75 -0.49 -0.26 -0.03        
GFDL SPEAR -0.47 -0.27 -0.08 0.09 0.23 0.35 0.42 0.37 0.25
IOCAS ICM -0.90 -0.84 -0.72 -0.63 -0.58 -0.52 -0.46 -0.42 -0.47
JMA -0.92 -0.74 -0.49 -0.25 -0.02        
KMA -1.32 -1.22 -1.00 -0.79 -0.67        
LDEO -0.28 -0.02 0.20 0.34 0.41 0.48 0.57 0.56 0.50
MetFRANCE -1.03 -0.77 -0.51 -0.34 -0.18        
NASA GMAO -2.90 -3.19 -2.80 -2.32 -1.99 -1.71 -1.40    
NCEP CFSv2 -1.16 -1.27 -1.21 -1.03 -0.88 -0.70 -0.46    
SAUDI-KAU -0.91 -0.65 -0.32 -0.10 0.04 0.08 0.11 0.16 0.26
SINTEX-F -0.69 -0.59 -0.41 -0.25 -0.03 0.24 0.47 0.58 0.58
UKMO -1.10 -0.98 -0.73 -0.52          
Average, Dynamical models -1.034 -0.913 -0.679 -0.461 -0.295 -0.167 -0.074 0.211 0.200
Statistical Models
BCC_RZDM
CPC CA -0.94 -0.81 -0.54 -0.38 -0.27 -0.20 -0.11 -0.01 0.11
CPC MRKOV -0.99 -0.80 -0.64 -0.47 -0.30 -0.16 -0.04 0.09 0.26
CSU CLIPR -0.76 -0.61 -0.46 -0.31 -0.27 -0.23 -0.19 -0.11 -0.03
IAP-NN -1.06 -1.05 -0.95 -0.80 -0.66 -0.52 -0.41 -0.33 -0.28
NTU CODA -0.98 -0.84 -0.66 -0.50 -0.26 -0.20 -0.22 -0.27 -0.28
UCLA-TCD -0.87 -0.77 -0.61 -0.44 -0.31 -0.23 -0.23 -0.31 -0.46
Average, Statistical models -0.933 -0.813 -0.643 -0.484 -0.345 -0.256 -0.201 -0.157 -0.113
Average, All models -1.006 -0.886 -0.669 -0.467 -0.310 -0.201 -0.125 0.041 0.056

Discussion of Current Forecasts

A slim majority of the models in the set of dynamical and statistical model predictions issued during mid-December 2021 show moderate La Niña SST anomalies, while various other models suggest a weak La Niña conditions in next two seasons (Dec-Feb, and Jan-Mar), which gradually decreases during spring of 2022. The ENSO-neutral state is likely to be increasing steadily during this time, and dominant in Mar-May and during rest of the forecast period, while the chances of El Niño conditions are very low throughout the forecast period.

Dynamical and statistical models show consistent and high probabilities of La Niña conditions during boreal winter time, and then decreasing in early spring of 2022. ENSO-neutral conditions are again more likely category than La Niña during rest of forecast period. Based on the multi-model mean prediction, and the expected skill of the models by start time and lead time, the probabilities (X100) for La Niña, neutral and El Niño conditions (using -0.5 C and 0.5 C thresholds) over the coming 9 seasons are:

Season La Niña Neutral El Niño
DJF 91 9 0
JFM 80 20 0
FMA 59 41 0
MAM 36 63 1
AMJ 25 71 4
MJJ 22 66 12
JJA 22 57 21
JAS 22 52 26
ASO 25 46 29

Summary of forecasts issued over last 22 months

The following interactive plot shows the model forecasts issued not only from the current month (as in the plot above), but also from the 21 months previous to this month. The observations are shown up to the most recently completed 3-month period. The plots allow comparison of plumes from the previous start times, or examination of the forecast behavior of a given model over time.
Hovering over any single model will highlight that particular model in the chart.
Clicking a particular model will hide/show that model in the chart.
At the bottom of the plot, you can select which models to show in the chart: all the models, the dynamical models only, or the statistical models only.

Notes on the data 

Only models producing forecasts on a monthly basis are included. This means that some models whose forecasts appear in the Experimental Long-Lead Forecast Bulletin (produced by COLA) do not appear in the table.

Once an IRI ENSO probability forecast has been published, the results stand even if a model reports an error and changes their data. When this happens we will update the plume with the model's correct values even though our forecast hasn't changed. What this means is that our forecast is always the same, but the underlying data may be different from what we based our forecast on.

The SST anomaly forecasts are for the 3-month periods shown, and are for the Nino 3.4 region (120-170W, 5N-5S). Often, the anomalies are provided directly in a graph or a table by the respective forecasting centers for the Nino 3.4 region. In some cases, however, they are given for 1-month periods, for 3-month periods that skip some of the periods in the above table, and/or only for a region (or regions) other than Nino 3.4. In these cases, the following means are used to obtain the needed anomalies for the table:

  • Temporal averaging
  • Linear temporal interpolation
  • Visual averaging of values on a contoured map

The anomalies shown are those with respect to the base period used to define the normals, which vary among the groups producing model forecasts. They have not been adjusted to anomalies with respect to a common base period. Discrepancies among the climatological SST resulting from differing base periods may be as high as a quarter of a degree C in the worst cases. Forecasters are encouraged to use the standard 1991-2020 period as the base period, or a period not very different from it.

Historical SST Anomalies Image

ENSO Forecast

Forecast Probability Distribution Based on the IRI ENSO Prediction Plume

Published: December 20, 2021

The plots on this page show predictions of seasonal (3-month average) sea surface temperature (SST) anomaly in the Niño3.4 region in the east-central tropical Pacific (5°N-5°S, 120°-170°W), covering the nine overlapping seasons beginning with the current month. The predictions are based on the large (20+) set of dynamical and statistical models in the plume of model ENSO predictions.

  • Model Based Prediction Percentiles Image

    Figure 5

    Predictions of ENSO are probabilistic. The ensemble mean prediction is only a best single guess. On either side of that prediction, there is a substantial uncertainty distribution, or error tolerance. The second plot (Figure 2) shows the estimated probability distribution of the predictions, showing a set of percentiles within that distribution for each lead time. The distribution is modeled as a normal (Gaussian) distribution, so that the overall mean forecast represents the center, or 50 percentile, in the distribution. The overall mean is formed using equal weighting among all models. On either side, other percentile values are shown symmetrically, ranging from 1 to 99 and including some intermediate percentiles (5 and 95, 15 and 85, and 25 and 75). The plot enables a user to estimate the probability of the Niño3.4 SST anomaly to be greater or less than some critical value, or within some interval. If, for example, the 85 percentile falls at 1.8° C above average, the probability of the SST exceeding 1.8° C can be estimated at 15%. Probabilities for exceeding or not exceeding values not exactly on percentile line can be roughly interpolated by eye. The overall width of the probability distribution is derived from the historical skill of the hindcasts of the models, from 1982 to present, for the specific forecast start time and lead time. This method of defining the probability distribution represents one of two general approaches, the other approach being a direct counting of ensemble members within each of the percentile bands. This second approach assumes that the ensemble spreads of the models are true representations of the uncertainty. Individual model spreads have often been found to be somwehate narrower than they should be, although in multi-model ensembles this tendency has been shown to be milder or even eliminated.

  • Model Based Prediction Distribution Image

    Figure 6

    Figure 6, sometimes called a spaghetti diagram, shows synthetically generated prediction scenarios that are equally likely. Here, 100 scenarios are shown; any number can be generated for such a diagram. Each scenario is produced using a random number generator, combined with knowledge of the mean forecast and its uncertainty, as well as the amount of persistence of anomalies. The degree of persistence of anomalies is based on the correlation of prediction errors from one lead time to another. In other words, the individual lines are designed to show the correct amount of persistence as expected in nature, rather than jumping around more randomly from one lead time to the next. The uncertainty and persistence statistics are based on the set of 7 NMME (North American Multimodel Ensemble) models, as it is assumed that these statistics are approximately applicable to all of the models. Sometimes the “spaghetti density” may appear asymmetric about the mean of all the forecasts or outside of the 85 and 15 percentile lines. This is purely sampling variability, and would not occur if many thousands of such lines were plotted. But with that many lines, most of the plot would be too crowded to get a sense of the behavior of the lines near the center of the distribution. The main purpose of the diagram is to serve users who want to assess realistic individual scenarios of ENSO behavior rather than statistical summaries of the forecast like the percentiles shown in the second plot.

New Article

Real-time ENSO forecast skill evaluated over the last two decades, with focus on the onset of ENSO events. Ehsan, M.A., L’Heureux, M.L., Tippett, M.K., Robertson, A.W, Turmelle, J.P., npj Clim Atmos Sci, 2024.

The IRI ENSO forecast is released on the 19th of each month. If the 19th falls on a weekend or holiday, it is released on the closest business day.

Forecast and model data used in our probabilistic forecast can be accessed by submitting a Request to Access IRI ENSO Data.

All data from this website is covered under the Creative Commons Attribution 4.0 License. When citing IRI ENSO images or data, please use "Images [or Data] provided by The International Research Institute for Climate and Society, Columbia University Climate School", with a link to https://iri.columbia.edu/ENSO.