Cb-LIKE – Cumulonimbus Likelihood: Thunderstorm forecasting with fuzzy logic

• Published in: Meteorologische Zeitschrift (Berlin, Germany : 1992) Vol. 26; no. 2; pp. 127 - 145
• Main Authors: Köhler, Martin, Tafferner, Arnold, Gerz, Thomas
• Format: Journal Article
• Language: English
• Published: Borntraeger 01.01.2017
• Subjects: best member; ensemble; forecasting; fuzzy logic; Nowcasting; numerical weather prediction; seamless prediction; thunderstorms; verification
• ISSN: 0941-2948; 1610-1227
• DOI: 10.1127/metz/2016/0805

Aviation is heavily affected by thunderstorms. Approaching storm cells with accompanying effects such as heavy rain, hail or downdrafts cause delays and flight cancellations. Consequently, the airlines and airport operators have to bear high additional costs, with reduced flight safety and passenger comfort. A reliable thunderstorm forecast up to several hours ahead saves time for decision makers, as the airport authorities, air traffic control, airline operation centre, and the crew in the cockpit for an appropriate and harmonised reaction, and for mitigating the consequences of a thunderstorm. The algorithm Cb-LIKE (Cumulonimbus-Likelihood) has been developed to provide such forecasts. Cb-LIKE is an automated system which designates areas with possible thunderstorm development by using output of the COSMO-DE numerical weather prediction model operated by the German Meteorological Service (DWD). The algorithm includes a newly developed “Best-Member-Selection” which allows the automatic selection of that member of a COSMO-DE ensemble that matches best with the current weather situation. An innovative fuzzy logic system combines selected model data and calculates a thunderstorm indicator for each grid point of the model domain for the following six hours in one hour intervals. Comparing thunderstorm observations by radar with Cb-LIKE forecasts in the summer period of 2012 verifies the algorithm and demonstrates the system's performance. The verification shows that Cb-LIKE gives better thunderstorm forecasts in comparison with the COSMO-DE radar reflectivity. Moreover, the verification results allow transformation of the Cb-LIKE indicator field into a field of thunderstorm probability. These probability forecasts are valid for thunderstorms within a certain spatial distance from the observed storms.