Unfortunately, Mother Nature did not cooperate and the area stayed pretty well capped into the forecast period. In the short time I was away, the decision was made to shift the focus area to Bismarck, ND, where we had a better chance of seeing some storms. Shortly after my return, some storms popped, giving the forecasters something to look at, analyze, and even issue some statements and a warning on.
So how did SATCAST do? Well, we got the convection in North Dakota as the system moved east, approaching out original focus area. It did catch the 3 cells that caused the increased activity, with lead-times of about 30 minutes as well as some other areas of convection as the afternoon progressed.
The area of clouds (marked by the green arrow) consistently flagged CI in subsequent satellite images. These storms started to show up on radar at 1958 UTC and went on to be warned on as early as 2330 UTC.
Unfortunately, SATCAST also indicated CI sporadically across the northern part of Minnesota that never had a chance of getting to verify. The capped environment just couldn't be pushed through by the rapidly growing systems that tended to pop up along the boundary. End result, some False Alarms.
A known problem area of the techniques includes high CAPE areas that are capped. The algorithm exploits the satellite signals to flag rapidly growing clouds. Although there are post-processing techniques in place to help eliminate false alarms (such as masking out areas that have no CAPE) it doesn't pick up as well in areas where significant CIN is also present. We as a research group are still hunting for the "best" way to reduce the false alarms without masking out the good data. Also compounding the problem in this area, was the presence of high level cirrus associated with the low-pressure system. The cloud-typing algorithm does a good job identifying ares that need to be masked out, yet high level cirrus seems to be the bane of both CI algorithms tested during this experiment, as evidenced by the CI flags along the boundaries of the system as the afternoon progressed.
In contrast, the UWCI algorithm did not have as many CI flags in the Grand Forks/Duluth focus areas. The one or two flags they did have, verified shortly thereafter. And yes, the UAH SATCAST algorithm did pick up on those cases as well. To be fair, I did not get a chance to see if UWCI flagged the storms in the Bismarck area later, so I am only commenting on the 2-3 hours over the first focus area. But it does bring up an interesting question, is it better to have a few indicators that verify at the cost of missing many of the other cases, or is it better to have more indicators in/of the environment, that although will catch the definite cases of CI, but will also result in higher false alarms?
I asked each of the forecasters and got a few different answers. For one, watching where SATCAST flagged rapidly growing clouds caused him to focus in on the associated environment to ensure a good understand of what is occurring at that time and check for agreement with previous thinking. If he wasn't looking at it, he was checking to see if the environment had changed since he looked last. In short, a situational awareness tool. For another, watching where the algorithm was flagging CI ahead of the warning area he was responsible for queued him in to watch for radar echoes. There were also some statements indicating that the jury was still out. Overall, a broad spectrum of answers yet not overall negative, especially after playing with both tools for only one day. I imagine that in the next day or two, more definite opinions will be formed and shared...
Then again, isn't that the point?
No comments:
Post a Comment