October 24, 2012

Will Hurricane Sandy Make Landfall on the Northeast U.S.?

We have a fascinating and potentially dangerous situation on hand for the Northeast U.S. and one possessing considerable uncertainty.

Sandy is now a category 1 hurricane with estimated maximum sustained winds of 85 mph and central pressure of 970 hPa (see picture).  My community is all aflutter about this storm, particularly since our forecast models are not in agreement--some suggest the storm will intensify and head straight in to the Middle Atlantic coastline, some take it out to sea, and others move it out to sea before swinging it westward to hit New England or the Canadian maritimes.
Hurricane Sandy is not well formed at this time
Such uncertainty is not unusual under the present circumstances:  we have a storm that will undergo what we call extratropical transition as it moves northward (changing energy sources in the process) and interacts with the stronger flow and temperature gradients of the midlatitudes.  Very complex interactions that greatly reduce predictability.

 Let me illustrate.

First, lets start with the gold standard.  The best hurricane prediction system is embarrassingly not American, but the European Center (ECMWF) high-resolution global model (note to U.S. Congressmen and Senators--you folks need to attend to this issue!).   A model that has the equivalent of 15-km grid spacing over the entire planet.

You will not believe the forecast.  Here is the sea level pressure analysis (solid lines, the shading is relative humidity) as the storm makes landfall along the Delaware coast.  A 940 hPa low! That is very low.


And here is the associated wind field--sustained 50 kts plus over the ocean.


If this were true there would be major coastal damage.  Most models tend to overdo such systems, but still, a huge threat.

But this is 2012 and we can do a lot better than a single "deterministic" forecast....there is plenty of uncertainty in the forecasts and one way we deal with that is with ensemble forecasts..running the model many times, starting a bit differently or with varying physics (e.g., different ways of describing clouds, precipitation, and the like).  Ensembles are run with less resolution than the single high-resolution forecast shown above.

 Here is the sea level pressure output from the ECMWF ensemble mean (the average of all the simulations) on the left side (solid lines) and the single high-res forecast on the right, both for 5 AM PDT on Monday, 29 October.  The ensemble mean is a bit washed out (we are averaging the pressures of many runs and those diffusing things a bit), and the lowest mean pressure somewhat offshore of the high-res run.  The color shading indicates that thelevel of uncertainty, based on the differences between the ensemble runs,  is considerable. 

Here is the same figure for 24 h later (Tuesday morning).  The ensemble mean (left side) low center is making landfall on New England, but there is a lot of uncertainty.  Clearly most of the ensembles are making landfall, but some forecast lows stay out to sea.   The bottom line:  a real threat, but it is not nearly 100%.


What about the American model, the GFS?  Statistics show it is now tied for number two with the UKMET office model.  Our high-res version (resolution about 25 km) does something very different (see graphics showing sea level pressure--black lines, and precipitation--shading).  The storm stays offshore for a long time, bypassing the central Atlantic states and then swings towards the Canadian Maritime Provinces.
Saturday, 11 AM PDT
Monday 11 PM PDT

Wednesday 2 PM PDT
The East Coast is safe?   Don't be so sure.  Let's check the U.S. global ensemble (much coarser than the European Center, unfortunately) at 5 AM on Tuesday..   The solid line is the mean (or average) pressure and the colors indicate uncertainty--orange is the most uncertain--where the simulations disagree the most. Wow...HUGE uncertainty.  Some solutions are taking the storm into the mid-Atlantic states like the European Center model, while others are heading the storm offshore.

Add caption
Here is the ensemble mean and uncertainty the next day. The ensemble mean makes landfall on the U.S.:  many solutions are hitting the NE states, but some move offshore (like our single high-res solution).


 By the way, what other famous storm started as a tropical storm, headed out to sea and then turned westward towards New England at almost exactly the same time of the year?   Answer:  The 1991 Halloween Storm....a.k.a. the Perfect Storm of book and movie fame.

So what do you come away with from all this?  First, you know what it is like to be a meteorologist in a real-world, very serious situation.  There is the potential for substantial loss of life, damage, and disruption.  Large costs to protect assets.  What would you recommend?  This is why meteorologists make the big bucks (right!).

Our models..both high res and ensembles..indicate there is a real threat to the U.S., from roughly Virginia to New England.  Folks in that region should pay close attention to the forecasts--which will get more certain as we get closer to landfall.  Not too soon to think about preparation.  There is still considerable uncertainty--and next week we could see anything from a catastrophic storm to nothing.  There is considerable chance of a serious landfall since our best model is going that way and BOTH U.S. and ECMWF ensemble systems showing a considerable number of forecasts with that solution.

There are many things I have not discussed here...how the tropical storm interacts with the midlatitude winds and why that produces uncertain forecasts...that will have to be for another time.   And yes, it is going to rain in Seattle this weekend.

16 comments:

  1. Christopher McDonald is chomping at the bit right now.

    Someone please get a hold of Bob Gunton. Tell him to turn around, he's not going to make it to Bermuda.

    ReplyDelete
  2. Cliff - for what its worth, you say its embarassing that the gold standard in hurricane prediction is not American.

    I would far rather the scientific community move towards a unified approach to things that looks past nations. No need to be patriotic - the science works regardless of your origin.

    ReplyDelete
  3. Unfortunately this:

    http://www.npr.org/2012/10/23/163499530/italian-seismologists-convicted-of-manslaughter

    should probably hold more sway in a meteorologist's mind right now than the uncertainty in the numerical weather forecasts.

    Times are scary.

    ReplyDelete
  4. Cliff, how does Sandy compare with the 1938 hurricane?

    My father-in-law rode out that one. He and his college buddies wanted to go down to the breakwaters outside of Providence and watch the waves - but were blocked by fallen trees in their driveway. So they missed being swept away by the 18' storm surge!

    KW

    ReplyDelete
  5. As of this morning the Euro had it heading for Delmarva and the GFS had it going to Nova Scotia. I know the Euro is the better model, but I remember the GFS winning the hurricane that hit NO this summer. The Euro had that one going to Brownsville I think and GFS had it going to AL coast.

    ReplyDelete
  6. As a devout Bayesian, I am all cheers about your probabilistic take on forecasting.

    But clearly, it matters how you do the averaging. Minimum (over space) of expected (over model runs) pressure at time t+n may be a very different thing from, say, expected (over model runs) minimum (over time interval) pressure at a particular location.

    In other words, *somebody* is going to get 960hpa with the associated mess, we just don't know who, and everybody's expectation is maybe 980hPa.

    Do you have a more technical discussion somewhere about what the respective permutations of E and min are?

    It's something that has bothered me for a long time about "40% chance of rain". What does that even *mean*?

    ReplyDelete
  7. The 12Z GFS has it making the left hook into Mass, Delaware, etc Monday/Tuesday.

    It is still a little early to really tell with the NAM, but this morning's 84hour forecast had the storm skirting the coast of Virginia.

    Cliff, do you know of any universities that would have high resolution models of the Eastern seaboard similar to the UW's models for the Pac NW?

    ReplyDelete
  8. All NWS Upper Air offices (about 75 in the CONUS) are conducting special upper air releases at 06Z and 18Z (11am and 11pm PDT) to support the models prediction of this storm. In my 17 years at the NWS Spokane office, we've never done this.

    ReplyDelete
  9. Ron,

    That is impressive, I'm aware such soundings are usually done by individual offices for periods of severe weather. So, for all office to do it says something about the necessity to get this right.

    However, your language was unclear on something. Is this just for today, or will the NWS be doing this from now until the storm makes landfall (or doesn't make landfall as the case may be)?

    ReplyDelete
  10. A truly great resource for keeping up on this storm is the Capital Weather Gang. Imagine this blog, except with 5 Dr. Mass'. A lot of great updates, weather/climate articles, and they'll have real-time forecasting for this event. I'm always envious of what those guys have built back there! :)

    http://www.washingtonpost.com/blogs/capital-weather-gang

    ReplyDelete
  11. Western Region NWS offices are doing the special soundings Thursday and Friday. Central Region will do them Thursday through Monday, and Southern/Eastern Region offices will do the special soundings Thursday through Tuesday. The reason is that the upper level short wave that is moving into Washington on Friday is the wave that is forecast to dig into the Ohio Valley by Sunday. The models expect this to cause Sandy to take the left turn and come onshore. So getting this short wave right is pretty key.

    ReplyDelete
  12. Since El Nino is dissipating, does a neutral winter affect weather patterns in the Atlantic? If I recall correctly from a previous blog, there could be a direct relationship between a La Nina and the strength of upper level troughs moving across the US. A product of which is related to some major tornado outbreaks, and what I assume would sheer or move tropical depressions easterly. Can a near-neutral or minor El Nino weather pattern in the Pacific play a significant role in Atlantic hurricane storm paths? As someone who grew up on the Maine coast, and absolutely infatuated with weather, I would always be glued to the TV/radio when the minor threat of a Hurricane was announced. 99% of the time the storm would turn west…except Gloria who hit us with a whopping sustained winds of 38mph (Something I sleep right through, now that I live in Seattle).

    Curious…What were the tropical Pacific temperatures during hurricanes Donna (1960), Gloria (1995), Bob (1991), and Grace (1991) which was instrumental in the “Perfect Storm”?

    ReplyDelete
  13. Cliff,
    I moved from Seattle to Central New Jersey this past summer, so I'm reading your blog with more than just passing interest. Thanks for your analysis, it's much appreciated.

    ReplyDelete
  14. Does anyone know the comparison between this hurricane (Sandy)
    and Hurricane Gloria that hit the East Coast in 1985??

    ReplyDelete
  15. Cliff, looking at the long term UW models for next Friday, should be interesting around here too, not just on the east coast. Too soon to tell?

    ReplyDelete

Please make sure your comments are civil. Name calling and personal attacks are not appropriate.

Should I Let AI Write My Blogs?

I spend a lot of time thinking about and writing my blogs.   Could AI apps do the work for me?   Could any of you tell the difference? Well,...