A multivariate Visualization for Uncertainties in Hurricane Model Ensemble

Thursday, 21 April 2016: 2:15 PM
Ponce de Leon C (The Condado Hilton Plaza)
Keqin Wu, IMSG at NOAA/NWS/NCEP/EMC, College Park, MD; and Z. Zhang, W. Wang, and V. Tallapragada

Given the increasing demand for accurate hurricane predictions, there is an urgent need to understand how the forecast is affected by various uncertainties in hurricane forecast models. The complicated interaction between the uncertainties of different atmospheric variables is one important factor causing inaccurate hurricane track forecasts. One major goal of this study is to depict the uncertainty of the key physical structure of one variable in the context of that of another variable. To convey multiple variables and their uncertainties in a single image, the underlying distribution of the ensemble members is approximated by bootstrapping and several state-of-the-art visualization techniques including topology extraction, colormaps, and glyph are utilized to the render the bootstrapped means and variances. This produces an novel multivariate map with a good preservation of key structures, accurate reading of quantities,and minimal perceptual interference between variables and their uncertainties. We apply the new method to visualize HWRF ensembles for several major hurricanes including Joaquin and Andrew and demonstrate its effectiveness in helping meteorologists to gain insights into the cause of hurricane forecast uncertainties.
- Indicates paper has been withdrawn from meeting
- Submission entered in competition