A Billion Dollar Flash Flood in Toronto - Challenges for Forecasting and Nowcasting

July 8^th 2013 seemed like a typical summer day in southern Ontario – warm and humid with a chance of thunderstorms in the afternoon. Forecasters expected storms to be marginally severe at best with heavy rain and gusty winds the main threats. However, despite the lack of the usual indicators for a significant flash flood event, copious amounts of rain fell over the Greater Toronto Area (GTA) over a period of several hours with one observation of 126 mm. Flood-related damage resulted in insured losses estimated at $890 million and many millions more in uninsured losses.

The interaction between the Lake Ontario lake-breeze front and an outflow boundary generated by an initial area of storms north of the GTA appeared to play a critical role in the event, causing unexpected storm development along the lake-breeze front toward the Lake and directly over the GTA. Large and rapid increases in in-cloud lightning preceded the flash flooding by more than 20 minutes, and occurred more than an hour before a warning was issued.

Real-time NWP output and post-event, very high-resolution simulations both significantly under-predicted precipitation amounts. Rainfall accumulations from Canadian radars were significantly underestimated as well. However, algorithms using dual-polarization products from the King City radar post event were able to mostly correct for attenuation problems and gave accumulations exceeding 150 mm. Social media provided some of the first and only real-time indications that an historic flash flooding event might be underway.

We will discuss the evolution of what is now “Ontario’s costliest natural disaster”, and how mesoscale boundary information, total lightning trends and social media reports can be combined with the latest radar, satellite and NWP data using an object-based approach to improve detection, nowcasting and alerting for severe and extreme weather events.