1:45 PM - 3:15 PM
[MIS24-P05] Precipitation Cell Detection and Tracking for Phased Array Weather Radar using Density-Based Spatial Clustering of Application with Noise
Keywords:Weather Radar, Density-Based Spatial Clustering of Application with Noise
In recent years, extreme weather phenomena such as torrential rains, and wind gusts have been occurring frequently in Japan. These weather phenomena are brought about by short-term, locally generated storm. In order to observe them in detail, Phased Array Weather Radar (PAWR) has been developed.
We used Density-Based Spatial Clustering of Application with Noise (DBSCAN) to identify cells in areas where the radar reflectivity factor observed by PAWR is more than the threshold of 15 dBZ, and tracks the cells based on their center of gravity. The DBSCAN method is based on the center-of-gravity position of a cell. However, this method has a problem that when a cell fusion occurs, it is misidentified as the development of a cumulonimbus cloud.
In this study, a strong precipitation area (precipitation core) is also defined inside a precipitation cell, and the core is discriminated and tracked. A higher threshold > 45dBZ is used to discriminate cores. The information from the precipitation cores assists in the analysis of precipitation cells and detects and predicts the location of heavy rainfall in more detail. The proposed method is also effective for the future development of heavy rainfall forecasting technology using deep learning.
We used Density-Based Spatial Clustering of Application with Noise (DBSCAN) to identify cells in areas where the radar reflectivity factor observed by PAWR is more than the threshold of 15 dBZ, and tracks the cells based on their center of gravity. The DBSCAN method is based on the center-of-gravity position of a cell. However, this method has a problem that when a cell fusion occurs, it is misidentified as the development of a cumulonimbus cloud.
In this study, a strong precipitation area (precipitation core) is also defined inside a precipitation cell, and the core is discriminated and tracked. A higher threshold > 45dBZ is used to discriminate cores. The information from the precipitation cores assists in the analysis of precipitation cells and detects and predicts the location of heavy rainfall in more detail. The proposed method is also effective for the future development of heavy rainfall forecasting technology using deep learning.