More people died last night in the New York area from the “remnants” of Hurricane Ida than over Louisiana and Mississippi as one of the most powerful hurricanes of the century made landfall on a low-lying coastal zone. (Current count: about two dozen in the NY area, about 8 in Louisiana)
Think about that. And as I will suggest in this blog, we can do much better, both in terms of forecasting and communicating serious weather threats. There are investments and new policies that are needed.
Heavy rain, flooding, and even tornadoes struck in a relatively narrow band stretching southwest-northeast from Pennsylvania and New Jersey, through New York, Connecticut and Massachusetts, as shown by a recent precipitation analysis by the NOAA/NWS Weather Prediction Center for the 48-h ending 6 AM EDT this morning (see below).
A fairly large area of more than 6 inches of rainfall, with some locations hitting 8-10 inches.
Much of that rain fell during short intense bursts of precipitation associated with thunderstorms. Newark, NJ, experienced 3.24 inches in one hour, with NY Central Park hit by 3.15 inches in one hour. Both are hourly records for those sites.
The intense convective (thunderstorm) nature of the rain is illustrated by a weather radar image at 9:50 PM EDT last night, with the red colors indicating extraordinarily high rainfall rates.
After landfall, Hurricane Ida transitioned into a tropical storm and then underwent extratropical transition in which it took on the characteristics of a midlatitude cyclone. The National Weather Service sometimes calls the resulting storm a Post-Tropical Cyclone.But there was a real danger in this transition.
Extratropical cyclones have strong upward motion, often associated with frontal zones, where temperatures and winds change rapidly. And tropical storms undergoing transition often entrain large amounts of tropical moisture that can result in heavy precipitation as the moist air is forced to rise by the storm circulation. This moisture can be converted to rain very rapidly in strong thunderstorms/convection.This is exactly what happened last night.
Below is a map of sea level pressure and atmospheric moisture (called precipitable water) for 10 PM PDT last night; you can see the low-pressure center and the plume of moisture (green colors) moving in from the southwest.
The warm front had warm, southerly winds on the south side and cooler, easterly winds on the north side, with the warm, moist, unstable air to the south forced to rise by the front, resulting in heavy convective showers to the north of the line. That is why the intense precipitation paralleled the front.
My colleagues at the National Weather Service had warnings out much of yesterday for heavy precipitation and the potential for flash flooding, with a flash flood watch out more than a day ahead of time.
Our models were useful but had some issues in both intensity and position. High resolution is critical for this kind of forecast for many reasons, including the convective (thunderstorm) nature of the heavy rainfall, and the sharp frontal boundary that helped produce the rain. 3-4 km grid spacing is a minimum that is viable.
The highest resolution model run by the National Weather Service several times a day is the HRRR model….the High-Resolution Rapid Refresh Model, run with 3-km grid spacing. Its total precipitation for a run starting at 8 AM on Wednesday for precipitation for the next day showed the band, but it was displaced a bit north and somewhat underplayed the precipitation intensity.
The National Weather Service also runs a small (7-8 member) ensemble of several high-resolution simulations (around 3-km grid spacing), called HREF A statistically enhanced mean of these runs (starting at the same time) is shown below. Better.
Everyone involved in numerical weather prediction knows what this country needs to get such forecasts correct: a relatively large (30-50 member) high-resolution (3-km or better for the grid spacing) ensemble of forecasts that are carefully calibrated to give good probabilistic/uncertainty predictions.
Committee after committee, workshop after workshop, have recommended this. The National Weather Service’s own modeling experts say the same thing.
But the investment is never made to do this. This means acquiring the necessary computer resources and building the modeling/statistical post-processing system. Very, very frustrating that this critical capability is pushed off into the future. (Senator Cantwell please take this on!)
We obviously failed last night. Around a dozen people died in flooded basements. Many people took to the roads that were flooded out. Abandoned cars were everywhere.
There are two stages to the warning process based on time. Yesterday, hours before heavy rain, it was clear that a serious event was in store and people had to be warned in the strongest terms to stay off the road and to prepare.
By late in the afternoon it was clear that a severe event was about to take place, and we needed to get the message to folks not only to stay off the road but to get out of low-lying basement apartments. That did not happen.
There is no reason why so many people had to die last night from an event we knew was coming and which we could watch unfold with weather radar and surface observations.