A 3-Dimensional Watershed Transform Technique for Storm Extraction on Gridded Data

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner
Tuesday, 4 February 2014: 4:30 PM
Room C204 (The Georgia World Congress Center )
Andrew J. MacKenzie, Univ. of Oklahoma, Norman, OK; and A. McGovern, V. Lakshmanan, A. J. Clark, and R. A. Brown

Many storm tracking methods are typically performed on a spatial grid. While visualizations are common in 3D, they are not designed to identify and track individual storms throughout their lifetime. Additionally, they are typically fairly simplistic, contouring along a hard threshold for a single variable to denote a boundary. The proposed method utilizes a 3D adaptation of the enhanced watershed transform. The algorithm grows storms based on several calculated parameters, resulting in a full 3D evaluation of the storm's location. The output provides an objective, adaptive boundary ideal for extraction of object properties for data mining and other analysis. Performance on multiple model types, resolutions, and grid sizes is demonstrated, with additional discussion on parameter selection and algorithm tuning. Comparisons are made with other common storm object identification methods such as the Storm Cell Identification and Tracking algorithm and the Thunderstorm Identification, Tracking, Analysis, and Nowcasting algorithm. General methodology and future avenues of research are also discussed.