Cluster analysis of multimodel ensemble data over new England

Nusrat Yussouf, David J. Stensrud, S. Lakshmivarahan

Research output: Contribution to journalArticlepeer-review

17 Scopus citations


An ensemble of 48-h forecasts from 23 cases during the months of July and August 2002, which was created as part of a National Oceanic and Atmospheric Administration pilot program on temperature and air quality forecasting, is evaluated using a clustering method. The ensemble forecasting system consists of 23 total forecasts from four different models: the National Centers for Environmental Prediction (NCEP) Eta Model (ETA), the NCEP Regional Spectral Model (RSM), the Rapid-Update Cycle (RUC) model, and the fifth-generation Pennsylvania State University-National Center for Atmospheric Research (PSU-NCAR) Mesoscale Model (MM5). Forecasts of 2-m temperature, 850-hPa u-component wind speed, 500-hPa temperature, and 250-hPa u-component wind speed are bilinearly interpolated to a common grid, and a cluster analysis is conducted at each of the 17 output times for each of the case days using a hierarchical clustering approach. Results from the clustering indicate that the forecasts largely cluster by model, with these intramodel clusters occurring quite often near the surface and less often at higher levels in the atmosphere. Results also indicate that model physics diversity plays a relatively larger role than initial condition diversity in producing distinct groupings of the forecasts. If the goal of ensemble forecasting is to have each model forecast represent an equally likely solution, then this goal remains distant as the model forecasts too often cluster based upon the model that produces the forecasts. Ensembles that contain both initial condition and model dynamics and physics uncertainty are recommended.

Original languageEnglish (US)
Pages (from-to)2452-2462
Number of pages11
JournalMonthly Weather Review
Issue number10
StatePublished - Oct 2004

All Science Journal Classification (ASJC) codes

  • Atmospheric Science


Dive into the research topics of 'Cluster analysis of multimodel ensemble data over new England'. Together they form a unique fingerprint.

Cite this