Unraveling the Sandy Storm-Surge Forecast
The track forecast for Sandy has been widely acclaimed as a success of modern meteorology, and rightly so. The advance notice provided by the European Center’s ECMWF computer forecast model was truly remarkable. And the National Hurricane Center’s forecast showed an alarming threat to the Northeast 4 1/2 days before the worst of Sandy’s surge came ashore. But the little-discussed National Weather Service storm-surge forecast for New York and New Jersey was a jumble of confusing and conflicting numbers – a victim of a broken system of forecasting and communicating the threat from rising water. With critical decisions on evacuations in the balance, city and state officials needed a specific, clear assessment of the threat to coastal sections. But based on the publicly distributed bulletins, they didn't get it.
Those of us who have been critical of the National Weather Service’s handling of the storm have focused on the government’s communications protocols and procedures that resulted in no Hurricane Warning being issued for the coastline north of North Carolina. But the real measure of failure was the stunning misunderstanding of Sandy’s threat by the top people in New York City. In addition, those officials in the Northeast and Mid-Atlantic that understood that a dangerous event was unfolding got started too late to execute a fully successful evacuation-communications plan. Beginning the discussion of evacuations just the afternoon before people over a vast area need to leave home would never have produce the best possible end result.
Early this month, the New York Times reported on the shocking decision by city and state health commissioners to leave nursing-home patients in the evacuation zone, even after the Mayor announced that everyone else should leave. Normally the frail and elderly are the FIRST to be moved. Health officials cited “confusing and changing forecasts”, which sounds like a lot of CYA, but they have at least a partial point.
Friday – The First Decision Day
According to the Times, Friday, October 26, three days before the worst of the storm, was a key decision day for city health commissioners. The report says that a “well-organized evacuation…would have taken at least two days”, meaning it would have had to start the next day, Saturday. The commissioners and the city administration were getting direct internal briefings from the National Weather Service – a meteorologist from the local NWS Upton office was embedded at emergency management. We don’t know exactly what was conveyed privately, of course, but the impression that Sandy would NOT be a historic event and instead be no worse than Irene conforms with the public messages from the Upton office.
On Friday morning, Upton’s Area Forecast Discussion (AFD) – where NWS forecasters share their forecast reasoning – used the word MINOR to describe the worst of the expected storm surge in some unfortunate wording:
A FULL MOON IS EXPECTED FOR MONDAY. A STRENGTHENING EASTERLY FLOW WILL AT THE LEAST RESULT IN A MINOR COASTAL FLOOD EVENT SUN INTO TUE...WITH THE POTENTIAL FOR WIDESPREAD SIGNIFICANT COASTAL FLOODING BASED ON THE TRACK OF A COASTAL STORM ASSOCIATED WITH
TROPICAL CYCLONE SANDY.
The “minor” wording was used in 10 AFD updates through the day on Friday until 3:54 pm when the descriptive word was changed to “moderate”:
WIDESPREAD MODERATE COASTAL FLOODING IS POSSIBLE BY THE MONDAY NIGHT HIGH TIDE CYCLE AS THE STRONGEST WINDS BEGIN AFFECTING THE AREA. THEN THE EXACT TRACK/TIMING AND EVOLUTION OF SANDY WILL DETERMINE THE MAGNITUDE OF COASTAL FLOODING WITH THE TUESDAY MORNING/EARLY AFTERNOON HIGH TIDE...WITH POTENTIAL FOR SIGNIFICANT AND WIDESPREAD COASTAL FLOODING RESULTING IN FLOODING/DAMAGE IN HISTORICALLY FLOOD PRONE SPOTS.
This appears to be the “moderate” language that the health commissioners digested to make their decision. If the nursing homes were not in “historically flood-prone spots”, the impression conveyed by the National Weather Service was, apparently, that they would NOT be flooded.
In addition, we know that the internal briefings from the NWS included language comparing Sandy to Hurricane Irene. During Irene just a year before, the flood threat was also characterized by the NWS as “moderate”, so you see the confusion. In the end, of course, Irene did NOT cause the type of flooding that needed an evacuation. The memory of Irene seems to have been a key player in the decision making.
This whole line of thinking and wording from the NWS Upton forecasters is baffling. First, Sandy was ALWAYS forecast to be stronger than Irene, and second, it was on a track that would produce more storm surge, even if it were the same strength. Everybody that has studied the New York/New Jersey hurricane threat knows that a storm center heading toward the west or northwest into New Jersey is the worst-case track for surge. Irene was weaker and NOT on that worst-case track, so why not make the comparison clear?
The key communications point should have been that Sandy was expected to produce a higher storm surge, perhaps significantly higher, than Irene. Period. End of story. End of confusion.
So the key decision day passed without clear, emphatic public language outlining the extreme threat. You might say that the caveats of a chance of “significant flooding” should have been enough to prompt an evacuation order. But forecasters often throw in worst-case language for public consumption, but say in private that they are really expecting a non-worst-case outcome. We can’t tell which way the Upton forecasters were leaning from the Friday language, but we find out Saturday morning.
At 5:46 AM, the Upton office issued a Coastal Flood Watch for the entire coastline. The “moderate” language continued:
WIDESPREAD MODERATE FLOODING ASSOCIATED WITH THE HIGH TIDES MONDAY COULD LEAD TO WIDESPREAD FLOODING OF VULNERABLE AREAS ALONG ALL THE COAST LINES.
So “historically flood-prone spots” became “vulnerable areas”, but it was still a pretty weak statement given the threat. And, they gave some hard forecast numbers for the first time, which tell you which way they were leaning on the “moderate” scale. They also threw in some more CYA, which was barely accurate:
TIDAL DEPARTURES… 2 TO 3 FT ABOVE ASTRONOMICAL TIDES MONDAY MORNING INTO TUESDAY MORNING...WITH POSSIBLE HIGHER DEPARTURES DEPENDENT UPON THE TRACK OF SANDY.
In actual fact, a significant change in Sandy’s track would have resulted in approximately the SAME storm surge, but more importantly, 2 to 3 feet is significantly LESS than Irene, which came in at more than 4 feet. As we’ve seen, the Irene comparison was critical for evacuation decision making, so this was a good-news forecast. At its peak, Upton thought that Sandy was most likely to produce less surge than Irene.
And there’s more. The bulletin also included another number, the forecast for water rise “above MLLW” for lower Manhattan Monday morning. For comparison, Irene’s storm tide was 9.5 feet above MLLW.
THE BATTERY NYC.....831 AM.....8.7.......MODERATE
MLLW means, essentially, LOW tide. And remember, the forecast was for Sandy to peak around HIGH tide. So, good luck to the poor non-meteorologist decision maker who has to sort out tidal departures from MLLW from high tide. What a nightmare.
And the National Weather Service agrees. The post-storm assessment of Hurricane Isabel in 2003 found that “WFOs should highlight any Storm Tide forecasts in their tropical cyclone products and emphasize the total height of water expected” because emergency managers, media, and storm drainage personnel found it “more effective”. In other words, quit mixing the way the storm-surge forecast is communicated and stick with the worst case water-rise number. Also, quit hiding the information at the bottom of obscure bulletins and advisories. But, somehow, none of that got adopted.
The Mount Holly National Weather Service, serving the Jersey Shore (up to Sandy Hook just south of New York City), had stronger language in its Coastal Flood Watch and a somewhat better forecast early that Saturday morning:
…MODERATE TO MAJOR COASTAL FLOODING IS ANTICIPATED FOR BOTH HIGH TIDE CYCLES ON MONDAY…
And buried in the RAINFALL text:
…RIVERS WILL NOT BE ABLE TO DISCHARGE EXCESS RUNOFF INTO THE 3 TO 5 FOOT TIDAL SURGE AIDED BY AN INCOMING WIND GUSTING TO 60 MPH… SANDY HOOK MAY SEE A COASTAL FLOOD OF RECORD MONDAY EVENING IF THE STORM CROSSES THE COAST SOUTH OF SANDY HOOK NEAR THE TIME OF THE MONDAY EVENING HIGH TIDE…
Bizarrely, the storm-surge forecast ONLY shows up in the RAINFALL section of the Coastal Flood Watch bulletin. Talk about hiding the most important information!
The bottom line: Upton forecast 2 to 3 feet (and was communicating with the New York City officials), which was well less than Irene, and Mt. Holly forecast 3 to 5 feet. Sandy produced a storm surge of 9 to 10 feet. Clearly there was bad computer guidance, bad analysis, or bad something because the forecast track, strength, and size of the storm were essentially correct. And they are the key parameters used in to forecast storm surge.
This was a colossal failure of the U.S. forecasting system and should be first on the list of issues to be examined.
The National Hurricane Center issued its first storm-surge forecast for the Northeast Saturday at 11:00 AM: 4 to 8 feet for the New York City area. The top end of that range ended up being more than a foot too low, but this estimate was drastically better than the forecasts from the local offices issued that morning. And, a potential 8-foot surge forecast should have gotten every emergency planner’s attention… although the forecast for Irene was also 4 to 8 feet, which was a point of confusion. In retrospect, an NHC forecast of 4 to 9 feet would have been a more impactful, just to highlight the difference from Irene.
In any case… voila, the Upton office increased its forecast and intensified its wording almost three hours later. (Why did it take that long?) The headline in the Coastal Flood Watch issued at 1:39 PM was:
...MODERATE TO MAJOR COASTAL FLOODING LIKELY MONDAY MORNING INTO TUESDAY MORNING...
And the body of the text included:
…THE POTENTIAL IS INCREASING FOR A MAJOR COASTAL FLOODING EVENT WITH POSSIBLE DAMAGE IN HISTORICALLY FLOOD PRONE SPOTS.
They should NOT have included the word “moderate” in the headline, and most of the homes that went under water would not have been considered to be “historically flood prone”, but they were moving in the right direction. When the Watch was upgraded to a Warning late in the afternoon, they continue with the same wording.
Unfortunately, when Mt. Holly upgraded to a Warning for the Jersey Shore at 5:08 PM, they quit forecasting storm surge numbers altogether and switched exclusively to the height above MLLW. And, they did NOT adopt the higher NHC forecast! Mt. Holly forecast 9.0 to 9.5 feet above MLLW for Monday morning at Sandy Hook, NJ (which is LOWER than Irene’s peak) and 8.0 feet for Monday evening (much lower than Irene). Strange but true. The final height above MLLW during the worst of Sandy that Monday evening was at least 13.5 feet.
So Mt. Holly had a terrible forecast AND a bad communications policy AND ignored the superior numbers issued by the National Hurricane Center. A triple-colossal systemic failure.
It is true that people along the Jersey Shore who are used to looking at tide gauges understand the “above-MLLW” numbers. But it’s inexplicable and inexcusable that an average person needs a Master’s Degree in Storm Surge to decipher public warnings for a life-threatening event like Sandy. Not to mention the terrible forecasts.
You might think… surely they put information in the bulletin explaining how to convert from the MLLW numbers in the new forecast to the storm surge numbers in their previous forecasts (in this area subtract about 4 to 5 feet). Nope. Surely has left the building.
The NHC Storm Surge Forecast
The National Hurricane Center increased their storm-surge forecast at 11:00 PM Saturday night for Long Island Sound (north of NYC) and Raritan Bay (south of NYC) to 5 to 10 feet. New York Harbor stayed at 4 to 8 feet. At 8:00 AM Sunday they upped the forecast to 6 to 11 for the entire New York City area.
The NHC used this sentence to describe its storm-surge numbers:
THE WATER COULD REACH THE FOLLOWING DEPTHS ABOVE GROUND IF THE PEAK SURGE OCCURS AT THE TIME OF HIGH TIDE...
I’m not crazy about the “above ground” reference, but I get the point. They are trying to relate the surge to something people can understand. The problem is that anybody living near the water immediately asks, “Do you mean above the ground in front of my house?” The answer is almost always “no”, of course, and you eventually come around to explaining that the storm-surge forecast is how much above normal the ocean water is expected to be pushed by to the storm. It would be good if the statement were revised to eliminate the obvious above-which-ground confusion. But, that’s a quibble.
The fact is, the people that know how to forecast storm surge are at the National Hurricane Center in Miami. There is a whole unit there that lives with the challenges of storm surge forecasting and communicating every day. There may be excellent weather forecasters at the local National Weather Service offices, but they can’t be expected to be storm-surge experts and tornado experts and every-other-kind-of-weather experts.
The confirmed-and-then-retracted report from the NOAA Hurricane Meeting this month says that the NHC’s much-delayed Storm Surge Warning program will go into internal-experimental mode this year, with public release in 2015. How the Storm Surge Warning would have been implemented in non-hurricane-warned Hurricane Sandy is an open question. But, we can only hope that this will be a first step toward improving storm surge forecasts and standardizing communications from local National Weather Service offices as well.
The National Hurricane Center exists because hurricanes are a specialized slice of meteorology. The local offices are wasting their time trying to forecast storm surge in parallel with the NHC, and there are other things, like communicating, they should be doing better. There is a systemic change required here that only the National Weather Service Headquarters in Washington can implement, and it needs to be done right away.
In addition to adding a Storm Surge Warning, the NHC needs to change their surge-forecasting timetable to reflect the reality on the ground in a mega event like Sandy. The first surge forecasts are currently released two days before landfall, normally coincident with the Hurricane Watch. That will work in most cases, but not an extreme event. As we have seen, critical decisions were required in New York City on Friday, the day before the NHC issued its first storm-surge numbers for that area. The key message, “the surge threat for New York from Sandy is significantly greater than it was in Irene” was needed THAT day to clear away the comparative confusion that was hindering decision makers. Plainly we need new and clear storm-surge messaging for use three and four days in advance of the storm’s landfall in situations like Sandy.
Surge Forecasts from Local NWS Offices on Sunday
The Upton, NY National Weather Service office, which serves the area in and around New York City, increased their storm surge forecasts in step with the National Hurricane Center, sort of. Here’s their wording at 8:00 AM Sunday morning:
…POTENTIAL FOR 5 TO 10 FT ABOVE MONDAY NIGHT INTO TUESDAY MORNING. THE HIGHER END OF THE RANGE RELEGATED TO THE MOST FLOOD PRONE COASTAL LOCATIONS WHERE TIDAL DEPARTURES WILL REACH 6 TO 11 FT. THIS INCLUDES WESTERN LONG ISLAND SOUND.
What does that mean? How is 6 to 11 at the higher end of 5 to 10? At that time the NHC was forecasting 6 to 11 feet ALL around New York City, including New York Harbor, so why hedge the numbers? You get the feeling that Upton didn't fully buy into the NHC forecast.
In the same bulletin, Upton issued a specific forecast of 8.4 feet above MLLW (remember, subtract 4 to 5 feet to get the surge). First, issuing a storm surge forecast to the tenth of a foot is a terrible idea. Why don’t we issue temperature forecasts as 75.4 degrees? Obviously, because it’s bad science. Storm surge can't be forecast with that kind of precision.
But also, that forecast is for Monday MORNING. It’s a confoundingly confusing combination of numbers and formats and time periods. I know it doesn't seem so confusing to people that use these bulletins every day, but people in the media and decision makers in city and state government are NOT in that category. Numbers stand out… that’s where your eye goes. And confusing numbers make a confusing message.
Mt. Holly increased its forecast numbers as well, but at a slower rate. It wasn't until 4:41 PM on Sunday afternoon, just over a day before landfall, that they finally issued a forecast representative of the actual storm surge that occurred. Though their numbers on Sunday morning would have raised alarm in any decision maker who had the decoder ring for the above-MLLW forecasts.
The Hurricane Isabel report had the answer. Let’s get storm surge forecasts standardized and in a format that everybody can understand.
Don’t Forget the Rivers
That 6 to 11 foot NHC storm-surge forecast for New York Harbor had ramifications beyond the immediate coastline. As the harbor filled, water pushed up the Hudson River raising the water level in Newark Bay and pushing water up the Passaic and Hackensack Rivers… with devastating results. You may have seen the damage in Little Ferry and Moonachie, which are south of Hackensack in Bergen County, NJ. Water poured through homes blocks from the river.
The problem is, nobody told those people the water was coming.
The Upton NWS office, which is responsible for that area, included a broad statement in their Sunday afternoon Coastal Flood Warning. The affected area included:
…THE TIDALLY AFFECTED PORTIONS OF THE HACKENSACK AND PASSAIC RIVERS…
But, they did NOT include Bergen County in the list of counties that were warned. No county listed meant no warning on NOAA weather radio, cable systems, and the like. Yet another systemic failure.
Confusing the Decision Makers?
When a plane crashes these days, it’s usually a chain reaction of all the wrong things – somebody spills his coffee, so somebody doesn't notice a gauge, then this then that. If just one person had done something different, the day could have been saved. The Sandy communications and forecasting fiasco was a similar cascade of missed opportunities.
If the folks at the Upton NWS office really understood the threat, why didn't they speak up when the health commissioners and the Mayor were headed in the wrong direction on Friday and Saturday? Somebody should have said, “Yo, Mayor, you’re going to look like an idiot if you don’t get with the program”.
The thing is… Upton’s bulletins and statements indicate that they didn't understand the threat. But where was the rest of the National Weather Service? And are those New York City people in such a hardened bubble that the extreme alerts that were on The Weather Channel and in the New York Times didn't penetrate at all? Governor Christi and mayors up and down the coast got the message. There is clearly something wrong at New York City emergency management that confusion was allowed to rule the day.
Standard operating procedure in an extreme situation should be to plan on the WORST credible level of threat, not the best-case scenario. New York City didn't follow that rule. Bad on them.
But in the series of blown forecasts and systemic failures, the National Weather Service’s decision to NOT issue a Hurricane Watch and Warning had a nasty side effect that might have played into the problem as well. No Hurricane Local Statements were issued by the local NWS offices because Sandy was forecast to morph into a nor'easter right before landfall (a policy that hopefully will never be used again). The Hurricane Local Statement is an annoyingly long synopsis of the threats to the local area, but without it the local message is fragmented in little known and, as we've seen, arcane bulletins in a variety of technocratic formats. This is bad communications plan at any time, let alone leading up to a historic disaster.
Had the Hurricane Watch been issued on Saturday, Hurricane Local Statements would have come out summarizing the local threats. Even though that would have been late in the city’s decision-making process, would it have made a difference? You never know, but it couldn't have hurt. And, sometimes it only takes one person to notice that something’s not right to stop a cascade of bad decisions.