"Here is the series of visible and infrared satellite pictures for the storm cell that we issued the tornado warnings. The Convective Initiation product gave a strong signal as this cell began developing, but the Cloud Top Cooling product never identified this storm. The infrared pictures are sampled at the coolest part of the cell at each time step, about 5 to 10 minutes. The storm grew explosively and produced one inch hail along with strong rotation on the radar."
This is a good example of why bringing products into a operational-type setting and demonstrating them in an experimental sense is very beneficial towards improvement and a successful transfer from research to operations. In this case, the developers were able to quickly identify the issue that the forecaster expressed confusion with. Until products are run in real-time over a large domain, some small events may be missed that are still important to an operational user's perspective. It is essential to identify and address issues within any experimental product prior to operational implementation.
Chris thanks for sharing this case with us.
ReplyDeleteAs I communicated to you, I can fairly confidently say this storm was not detected due to the existing anvil immediately to its northeast. In the algorithm we use a series of tests using the small/large boxes to separate vertical cooling between two images from horizontal cloud advection/expansion. In this case the large box likely encountered the anvil to the northeast and hence this storm was omitted. Typically we do get these backbuilding storms, but when occurring very close to the mature anvil it can be omitted.
This goes back to the conservative nature of the algorithm, we will miss things, occasionally a strong storm even, but the false alarm rate is kept quite low as a consequence, so when a forecaster sees the signal they can with high confidence know something will occur.
I think this case illustrates a potential improvement we could make to the algorithm in which we more intelligently use the small/large box design by including model winds to only look 'upstream' from a potential developing storm when performing our cooling rate filtering.
Justin UW/CIMSS