Ensemble Visualization of Hurricane Models
Open Access
- Author:
- Jin, Qi
- Graduate Program:
- Information Sciences and Technology
- Degree:
- Master of Science
- Document Type:
- Master Thesis
- Date of Defense:
- March 28, 2017
- Committee Members:
- Xiaolong Zhang, Thesis Advisor/Co-Advisor
- Keywords:
- Ensemble Forecasting
Hurricane Models
Ensemble Forecasting
Hurricane Prediction
Information Visualization
Coordinated Views
Interactive Visualization - Abstract:
- Visualization has been widely accepted as one of the most efficient methods in ensemble forecasting, a tool in weather prediction to reveal the uncertainty of forecasting models. An ensemble consists of multiple members of forecasts so as to provide more accurate analysis than traditional deterministic forecasting. However, due to the large-scale and multivariate data and the difficulty in interpreting the results, it becomes important to search for methods and techniques to organize the data and reveal the relations and patterns among different ensemble members and dimensions of outputs. This thesis reports a study to explore visualization design to support the ensemble analysis of hurricane models. We developed a set of visualization tools to present the results of the ensembles on the Hurricane Sandy (2012). We reviewed the methodology and identified the major tasks based on the ensemble analysis of the Sandy. Specifically, our focus is on the composite and sensitivity analyses of the prediction of track, near surface wind speed, and sea level pressure. The proposed visualization tools were developed mainly based on Google Maps API JavaScript and D3.js. We also conducted a preliminary user evaluation of our design. The user evaluation involved two experts in meteorology. Both of them agreed that clustering main groups and changing time slider to see the evolution of the prediction were the most helpful features, while the coordinated views and needed to be improved. Based on their trial of our design and the feedback from them, we recommended a set of design considerations that future design can take, including a clearer view of the clustering groups and the improvement of the diagrams and interactivity.