MRF Ensembles at CDC

Using Ensemble Forecasts to Assess the Impact of El Niño/La Niña on Weather Systems

Motivation: One of the most difficult, and most common, questions we receive is "Was that storm caused by El Niño/La Niña?" Using the observational data record and general circulation model (GCM) results, it is possible to identify certain regions where the probability of weather events is altered by El Niño. However, using these tools it is impossible to say whether a specific weather event was influenced by El Niño or La Nina. In collaboration with scientists at the National Centers for Environmental Prediction (NCEP), we have performed some experiments aimed at identifying how El Niño/La Niña affects forecasts of weather events. This was done by running an ensemble of model forecasts, with and without boundary conditions containing the anomalous sea surface temperatures (SSTs) associated with El Niño/La Niña. Strictly speaking, the difference between the two forecasts tells us how the forecast of the weather event was influenced by the anomalous tropical SSTs, during the course of the forecast. However, to the extent that the forecast of the weather event is accurate, and the impact of the SSTs on the model solution is realistic, the results of these experiments will help us answer that oft asked question.

For more details, see the manuscript that we have submitted to the Bulletin of the Americal Meteorological Society.

Experimental Design: From December 1997 through March 1998 (and again starting in December 1998), an 11 member ensemble was run here at CDC every other day, all members at T62 resolution, starting from the operational NCEP initial perturbations at 00 UTC. This ensemble is essentially identical to the operational 00 UTC ensemble run at NCEP in every respect, except for the sea surface temperature (SST) boundary conditions. These are set to climatological values (using the 1982-1996 climatology from NCEP reanalysis) equatorward of 20 degrees latitude. Below is the difference in the SST boundary conditions imposed in the two ensembles for New Year's Day 1998 (the recent evolution of the tropical SSTs can be seen here.).

By comparing differences between our ensemble mean and NCEP's, the effect of anomalous SSTs (associated with El Niño/La Niña) on the forecasts can be assessed.

Results with initial conditions from (YYMMDD)...

The color-filled regions on the difference plots indicate those areas in which the differences are significant at the 5% level. The significance test used does NOT take advantage of the fact that the samples are paired, i.e. corresponding runs from each ensemble are correlated. If a significance test appropriate for paired samples were used, much more of the maps would be significant. We use a more conservative test to try to identify those regions in which the differences are insensitive to the choice of initial perturbations. In other words, the shading is an attempt to highlight those areas in which the difference field would be similar even if the two sets of ensembles were initialized with different (but still equally plausible) initial perturbations.

[990331]

ENSEMBLE FORECAST PLOTS: CDC NCEP Univ. of Utah

ENSO: Effects on Climate Current Conditions References
Education Forecasts Differences Description Glossary FAQ

NOAA-CIRES Climate Diagnostics Center
Document maintained by Jeff Whitaker (Jeffrey.S.Whitaker@noaa.gov)
Updated: Apr 20, 2004 10:46:44 MDT
http://www.cdc.noaa.gov/people/jeffrey.s.whitaker/mrf/results/index.html