87th AMS Annual Meeting

Wednesday, 17 January 2007: 9:15 AM
Statistical Climate Prediction for the southwestern U.S.: A 7-Year Assessment
206A (Henry B. Gonzalez Convention Center)
Klaus E. Wolter, NOAA/ERL/CDC, Boulder, CO
New and improved "climate divisions" can be used for the statistical prediction of climate anomalies (here: precipitation) in the interior southwestern U.S. After initial success with simple ENSO composites, the lack of clear-cut ENSO phases around 2000-2001 prompted further exploration with other known teleconnection indices, and various statistical techniques. Stepwise linear regression was chosen as the most straightforward statistical technique to test and develop prediction equations.

Seasonal forecasts were trained on station data for the Water Years 1951 through 1999. This includes the implementation of a statistical ensemble approach that uses five different base periods, bias-corrected for its performance within the 49-year record. After some experimentation, publicly issued forecasts started in late 2001, and are continually updated monthly on the internet (http://www.cdc.noaa.gov/people/klaus.wolter/SWcasts/). Since the original data training period ended in September 1999, there will be seven years of verification data available by the end of 2006.

Seasonal forecast skill for the interior southwestern U.S. appears to be linked not only to ENSO (and its various 'flavors'), but also to SST regions further afield (Indian Ocean) as well as closer to the U.S. (eastern subtropical Pacific and Caribbean). Other useful predictors include northern hemispheric teleconnection patterns, and antecedent regional precipitation anomalies. Verification skill for the last seven years exhibits large regional and seasonal variations, but has remained positive for all seasons, in contrast with official Climate Prediction Center forecasts for much of the region. This presentation highlights the regions and seasons with the highest skill, and reports on possible contributors to this skill.

Supplementary URL: