Official websites use .gov
A .gov website belongs to an official government organization in the United States.

Secure .gov websites use HTTPS
A lock ( ) or https:// means you’ve safely connected to the .gov website. Share sensitive information only on official, secure websites.

Developing an Optimum Multimodel ENSO Prediction

The ENSO prediction plume product, issued monthly by the International Research Institute for Climate and Society (IRI), shows a large set of model forecasts for the ENSO state out to 10 months in advance. Although helpful, the product currently has a number of serious deficiencies. First, it shows predicted SST anomalies with respect to differing climatological base periods, because the modeling centers from which the predictions come do not use identical base periods. Secondly, many modeling centers do not correct systematic errors in their predictions, and these errors are perpetuated in the predictions on the plume. Some of the forecast models, while showing skill in their real-time ENSO forecasts, do not have adequate hindcast histories from which to treat such systematic errors. Finally, only a rudimentary method is used to consolidate the predictions into a logically derived multimodel mean prediction, and no attempt is made to develop a forecast probability distribution about the mean prediction. Thus, users gain some idea of the forecast probability distribution solely from visual inspection of the inter-model ensemble mean forecast disagreement-a by-far suboptimum criterion.

The proposed work will substantially improve the calibration as well as the multimodel ensembling inadequacies, leading to a more accurate multimodel deterministic and probabilistic ENSO prediction. Additionally, the product will have a more user-friendly format such that probabilities of the full range of possible values are provided more explicitly. Several methods will be tested to develop the forecast probability density function, including equally and skill based weighting, and including an ensemble regression method that uses the models’ individual ensemble members for those models producing ensemble predictions. The forecast probability distributions from the selected methodology mayor may not closely coincide with the spread of the individual model ensemble means; when they do not, the resulting forecast distribution would be the proper indicator of uncertainty that is missing from the current plume product.

The ENSO prediction plume product will have several versions, one of which will be a multimodel ensemble using only the NOAA models participating in the national multimodel ensemble (NMME) experiment, using the same strategy as for the larger set of models. In examining this smaller model set, the issue of the number of models in a multimodel ensemble will be addressed: How many acceptably skillful models tend to produce the best possible multimodel prediction skill, given the high inter-relatedness of the model ENSO predictions? Experimentation with weighting schemes and number of models wi1llead to new methodological knowledge, and to an optimum version of the all-model and the NOAA NMME model plumes.

More accurate, usable predictions of the ENSO state are the fundamental aims of improving the ENSO prediction plume product. Because the ENSO state is related to seasonal climate, better and more easily understood ENSO predictions are of benefit to users in the U.S. and worldwide. Better ENSO predictions lead to better predictions of seasonal climate in known seasons/locations in the U.S. and the globe. Better seasonal climate prediction, in turn, is relevant to the mission of the MAPP Program, seeking to advance intra-seasonal to decadal climate prediction. This work is also relevant to the Next Generation Strategic Plan, as enhanced climate predictions leads to more valued, relied-upon climate services for the benefit of any climate-sensitive sector (e.g., water management, coastal sustainability). The combination of the effects of climate change and ENSO has potential for the hazard of record-breaking climate extremes.

Scroll to Top