|Item type||Location||Call number||Copy||Status||Date due|
|REPORT||Mesa Lab||102113 (Browse shelf)||1||Available|
Published in English
Doctoral Thesis-- Ludwig-Maximilians-Universitat Munchen.
Newly developing convective clouds can be detected and monitored by satellite, but which of these clouds will grow to mature thunderstorms is difficult to predict from one data source alone. Within this thesis it is shown how the quality of satellite-based convection initiation (CI) detections can be raised substantially by the use of additional data sources which quantify available moisture, airmass instability, and lift for the analyzed clouds, the necessary ingredients for thunderstorms to develop.
Regions of interest for possible CI are detected by the Cb-TRAM1 algorithm using 5 minute rapid scan satellite data. Cb-TRAM combines satellite channel data making it possible to distinguish newly developing (initiating), fast growing, and mature convective storms. Furthermore, these detections are extrapolated into the future, producing nowcasts for up to 60 minutes.
For evaluating the quality of the satellite-based CI detection and to quantify the achievable improvement by the use of additional data, a suitable verification method for these CI detections is very important. An object-based verification
approach for these Cb-TRAM CI objects is introduced, which has been newly develop ed within this study. In order to derive sound statistics, the verification is performed over a whole summer period (May 15 - August 31, 2009) and for the whole Central European area. The CI detections can be categorized as developing (hits) and non-developing (false alarms). The verification results show a large amount of false alarms which has to be reduced in order to get more meaningful CI detections.
The possibility of individual Cb-TRAM CI detections to grow further is analyzed using additional data from surface observations and numerical weather prediction (NWP) model output, in order to gain the information on available moisture, instability, and lift for each CI detection object. This information is combined using fuzzy logic to obtain a so-called "CI forcing" value per object. Finally the CI forcing value is translated into a probability of further development to a thunderstorm for each cell.
Within this thesis the benefit of using multiple data sources to improve CI nowcasting is demonstrated. The additional information provided by the newly incorporated data raises the CI detection and nowcast quality by allowing, depending on the user-selectable amount of omitted hits (in the range of 0 - 25%), a substantial reduction of the amount of false alarms (5 - 65%). The methodology can also easily be adapted or extended for further additional data sources. An early identification of regions where mature storms will evolve allows for more adequate, user-oriented warnings.