Nowcasting 0–2 hour Storm Intensity within the GOES-R Convective Initiation modeling Framework

Convective initiation (CI) nowcasting uses geostationary-based infrared fields that are statistically related to the occurrence of thunderstorms 30-60 min later, and form the GOES-R CI algorithm as routinely used today. Beyond nowcasting CI (or any first-time occurrence of a ≥35 dBZ radar echo at the surface), the GOES-R CI algorithm has evolved to nowcast the first-time occurrence of lightning from a convective cloud (based on Harris et al. 2010, and another forthcoming manuscript), and also provides now a very early detection of growing cumulus clouds that are destine to become severe storms, which will be the emphasis of this conference presentation. The GOES-R CI algorithm uses a logistic regression modeling framework to identify new CI events, achieving an accuracy of ~80-85% (Mecikalski et al. 2015). Further processing of the CI fields using quadratic-discriminant analysis (QDA; based on Apke et al. 2015) has refined the widespread occurrence of CI objects to fewer objects more likely to undergo the CI process, while new funded work with the GOES-R NearCast algorithm promises to provide further refinements in how the CI algorithm is presented to forecasters as a means of highlighting events of high impact, and/or those CI events likely to grow up-scape into larger convective systems.

A fundamental component is that storm severity is largely determined by atmospheric instability that affects updraft intensity. Updraft intensity can be inferred indirectly by the fact that cloud drops in stronger updrafts have shorter time to grow and glaciate, thereby possess smaller effective radii (Re) for a given cloud-top temperature, and also colder glaciation temperatures (Tg). Re at Tg is smaller for stronger updrafts. Using 340 CI events, formed from 5-min imagery and monitored for 3 hours (4 days in August 2010 and July 2012), and with 2.5 min MSG rapid-scan data in 2013 (20 June and 29 July), the CI and T-Re fields were analyzed with cloud feature metrics. Severe storm predictor fields were analyzed over 10-min timeframes from 5 to 45 min of cloud growth. Validation of the predictions is made against QC2-level severe weather reports of the European Severe Weather Database. Results show (Mecikalski et al. 2016) that trends in cloud growth are linked to Re, with more rapidly growing clouds exhibiting low Tg and relatively small Re when comparisons are made across a mesocale (~125-240 km) region experiencing active convective cloud growth. This study helped confirm earlier field experiment and observational analysis by Rosenfeld et al. (2008). Relationships between Tg and CI-indicators that estimate cloud-top glaciation were also 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).

Recently, the logistic regression model driving the GOES-R CI algorithm was tailored to identify pending severe storm events (so-called “Severe CI” algorithm), relying on 15-min GOES observations over the U.S. Besides including the typical thermodynamic parameters of CAPE and CIN, other variables such as low- and mid-level lapse rates have shown to be very important in whether developing clouds will go on to produce severe weather. Coupling GOES-R cloud property datasets to the Severe CI will subsequently lead to an integration of the T-Re concepts into the nowcasting framework, toward providing forecasters as early a lead-time as possible for pending severe storm development. With the coming of GOES-R in late 2016, integration of 5-min routine and 1-min super rapid scan datasets into the Severe CI algorithm will also significantly increase forecast lead times, with the added infrared channels from the Advanced Baseline Imager increasing skill in the algorithm's overall performance.