1.3
Recent developments in convective initiation forecasting using GOES and MODIS
John R. Mecikalski, University of Alabama, Huntsville, AL; and K. Bedka and S. J. Paech
This presentation highlights the recent improvements in the 0-1 h forecasting (“nowcasting”) of the initiation of thunderstorms. The goals for accurate high-temporal convective initiation (CI) forecasting has resulted in improvements to the Mecikalski and Bedka (2005) GOES-based CI nowcasting methodology, which has subsequently broadened our understanding of how GOES senses growing cumulus clouds. Convective initiation is defined here as a transition from below to above 35 dBZ echoes as observed by NEXRAD radar.
The goals of this project include to: 1) demonstrate and describe the absolute skill in our GOES-based procedures in terms of Probability Of Detection (POD) and False Alarm Rate (FAR) scores, 2) present a more complete understanding of the relative importance of each possible GOES infrared field to nowcasting CI, and 3) present an overall update to the CI forecasting procedure that maximize POD scores while minimizing FARs. Presently, maximum PODs of 85% are found using the 13.3-10.7 um channel difference, while a minimum FAR of 39% occurs when the time trend of this channel difference is monitored for moving and growing cumulus clouds.
At the present time, research involves processing co-located GOES visible/infrared, and comparing these data to WSR-88D level radar data, for several CI events across North Alabama/South Tennessee. For six events, 1 km GOES-NEXRAD data are processed, which involves parallax correcting GOES observations, and interpolated both into the same coordinate system. High-resolution satellite-derived wind information is used to track moving clouds (Bedka and Mecikalski (2005), with the monitoring of cloud-top infrared temperature trends possible (e.g., Roberts and Rutledge 2003) juxtaposed with WSR-88D data. Within all six events, approximately 54,000 1 km pixels of CI become available for statistical analysis. A combination of linear discriminant analysis and other approaches, we are able to evaluate the relative importance of eight infrared CI indicators, which are combinations of the 6.7, 10.7 and 13.3 um channels from GOES-12. Channel differencing and time trends of channel (differences) are specifically evaluated. Preliminary evidence suggests that the 13.3 um channel is particularly important given its ability to detect particularly large updraft development.
Through research involving the “AutoNowcaster” expert system (Mueller et al. 2003) we are fortunate to be able to test these improvements within a methodology to nowcast CI. Work already completed shows that GOES information from moving convection has value above that previously used within this system. This presentation will report on our progress and most recent findings.
.Session 1, Retrievals and Cloud Products: Part I
Monday, 30 January 2006, 10:45 AM-12:00 PM, A305
Previous paper Next paper