Reposted from the Cliff Mass Weather Blog
In the New York area, more people died from “remnants” of Hurricane Ida last night than Louisiana and Mississippi, as one of the strongest hurricanes of the century hit a low-lying coastal zone. (Current number: about two dozen in the NY area, about 8 in Louisiana)
Think about it. And as I’m going to suggest on this blog, we can do a lot better, both in forecasting and communicating serious weather threats. Investment and new policies are needed.
Dramatic video of what happened last night
Heavy rain, flooding, and even tornadoes hit a relatively narrow band that stretched southwest-northeast of Pennsylvania and New Jersey across New York, Connecticut and Massachusetts, according to a recent rainfall analysis by the NOAA / NWS Weather Prediction Center for the 48th shows -h ends this morning at 6 p.m. EDT (see below).
Quite a large area of over 6 inches of rainfall with some places reaching 8-10 inches.
Much of this rain fell during brief, intense downpours associated with thunderstorms. Newark, NJ, seasoned 3.24 inches in an hour, with NY hit Central Park by 3.15 inches in an hour. Both are hourly records for these sites.
The intense convection (thunderstorm) of the rain is illustrated by a weather radar image last night at 9:50 p.m. EDT, with the red colors indicating exceptionally high rates of precipitation.
Upon landing, Hurricane Ida turned into a tropical storm and then passed an extra-tropical transition where it assumed the characteristics of a medium-latitude cyclone. The National Weather Service sometimes calls the resulting storm a post-tropical cyclone.But there is a real danger in this transition.
Extra-tropical cyclones have strong upward movement, often associated with frontal zones where temperatures and winds change rapidly. And transitional tropical storms often bring with them large amounts of tropical moisture, which can lead to heavy rainfall, as the humid air is forced to rise by the storm circulation. This moisture can very quickly be converted into rain during strong thunderstorms / convection. That’s exactly what happened last night.
Below is a map of sea level pressure and atmospheric humidity (called rainwater) for last night at 10:00 p.m. PDT; You can see the low pressure center and the plume of moisture (green colors) moving from the southwest.
To the east of the lower center, there was a warm front, as indicated by National Weather Service analysis for 8:00 p.m. EDT (indicated by the black semicircles).
The warm front had warm south winds on the south side and cooler east winds on the north side, with the warm, moist, unstable air in the south being forced to rise by the front, causing heavy convection showers in the north line. That is why the intense rainfall ran parallel to the front.
My colleagues from the National Weather Service almost always issued warnings of heavy rainfall and the potential for flash floods yesterday, with a flash flood advising to watch out more than a day in advance.
Our models were useful, but had some intensity and position issues. High resolution is critical to this type of forecast for many reasons, including the convective nature (thunderstorms) of the heavy rainfall and the sharp frontal boundary that helped create the rain. 3-4 km grid spacing is a practical minimum.
The highest resolution model that is run by the National Weather Service several times a day is the HRRR model … the High-Resolution Rapid Refresh model that runs on a 3km grid. The total precipitation for a run that started at 8 a.m. on Wednesday for precipitation for the next day showed the gang but was shifted slightly to the north and underplayed the precipitation intensity a little.
The National Weather Service also runs a small (7-8 member) ensemble of several high-resolution simulations (approx. 3 km grid spacing) called HREF. Better.
Anyone who deals with numerical weather forecasting knows what this country needs in order to get such forecasts right: a relatively large (30-50 members) high resolution (3 km or better for grid spacing) ensemble of predictions carefully calibrated to provide good probability / uncertainty predictions.
Committee by committee, workshop by workshop have recommended this. This is what the National Weather Service’s own modeling experts say.
But the investment is never made to do so. This means acquiring the necessary computer resources and setting up the modeling / statistics post-processing system. Very, very frustrating that this critical ability is being postponed into the future. (Senator Cantwell, please do this!)
We obviously failed last night. About a dozen people died in flooded cellars. Many people took to the flooded streets. Abandoned cars were parked everywhere.
The time-dependent warning process consists of two phases. Yesterday, hours before heavy rain, it was clear that a serious event was imminent and that people must be warned to stay off the road and prepare.
By late afternoon it was clear that a serious event was imminent and we had to get the message across to the people not only to get out of the street, but also to get out of the low-lying basement apartments. That didn’t happen.
There’s no reason so many people died last night from an event that we knew was coming and that we could observe with weather radar and surface observation.