Developing Parameters to Nowcast Intense Storms within the 0–1 hour Timeframe

Two independent methods for predicting initiation (e.g., Mecikalski and Bedka 2006; Mecikalski et al. 2010) and the near-term intensity of convective storms (Rosenfeld et al. 2008) are combined into one comprehensive method that will be able to nowcast the initiation of locally intense/severe convective storms. Convective storm initiation (CI) has been predicted using a combination of geostationary-based infrared “interest fields” that were statistically found to have the most predictive capability for diagnosing which cumulus clouds later became thunderstorms over the forthcoming 30-60 min. In addition, the severity of storms is determined to a large extent by the atmospheric instability, which in turn affects the potential updraft intensity, the most direct dynamic feature that manifests the storm intensity. Updraft strength determines the maximum size of the hailstones that can be produces by the storm, and also the intensity of the low level convergence and downdrafts that may be responsible to tornadoes and downbursts. Updraft intensity can be inferred indirectly by the fact that cloud drops in stronger updrafts have shorter time to grow and glaciate, thereby having smaller effective radii (Re) for a given cloud top temperature (T), and also colder glaciation temperature (Tg). The Re at Tg is smaller for stronger updrafts (Rosenfeld et al. 2008).

Using a database of several hundred convective initiation events, formed from 5-min frequency imagery over 2 hours on four days in Europe in August 2010 and July 2012, including days in 2013 with 2.5 min MSG rapid-scan data (20 June and 30 July), the CI and T-Re fields are analyzed, together with MSG Global Instability Index (GII) and cloud feature expansion metrics. Both threshold scoring, probabilistic and decision tree/variable importance approaches are considered as the storm intensity estimation nowcast procedure is formed. Satellite attributes used to determine intensity are maximum (over 2 hours) cloud feature expansion rates and overshooting top magnitude (in terms of the depth of the overshoot). Validation of the predictions is made against surface reports of the European Severe Weather Database (ESWD), and the presence of cold-rings and V-shapes (enhanced-V) anvil-level features. The eventual goal is to use the merger CI–TRe algorithm in a real-time nowcasting system for use over Europe and the U.S. (in light of GOES-R).

Results show that trends in cloud growth are linked to Re, with more rapidly growing clouds indeed exhibiting delayed glaciation (i.e. low Tg) and relatively small Re. Relationships between Tg and CI-indicator brightness temperature differences that determine cloud-top glaciation are formed, with stronger storms verifying both a delay in glaciation relative to Tg and defined cloud-top ice signatures as thick anvils form (as compared to weaker storms). Analysis also shows that 5-6 predictors gain increased/decreased importance over 10-min timeframes from 5-15 min to 35-45 min of cumulus cloud growth, arguing for a continuum approach to nowcasting storm intensity. Use of 2.5 min frequency data shows that nowcast lead-time is gained, with meaningful predictions made using only 5 min analysis intervals.