February 25, 2013

Second-Rate U.S. Numerical Weather Prediction: Why You Should Care


In several of my previous blogs I noted that U.S. numerical weather prediction is lagging behind the European Center and others--a diagnosis pretty much universally accepted in my field.  I listed some of the reasons:  inferior computers, poor management, lack of effective leadership, inability to tap the large U.S. weather research community, and others.

But some highly placed folks in the National Weather Service (NWS) and elsewhere have argued that U.S. inferiority in numerical weather prediction really doesn’t matter, since U.S. government forecasters have access to the superior forecasts of the models of the European Center (EC), the UK Met Office, and others.   As a prime example, they noted that EC model forecasts played a major role in NWS warnings in the week before Hurricane Sandy made landfall.  Some European Center officials have said the same thing--the current situation is fine!  Weather prediction is a global entity, so why be concerned if they are ahead?

In this blog, I will take issue with these arguments and will suggest that first-rate numerical weather prediction by the U.S. National Weather Service is crucial for the nation and of great benefit to the entire world.  That it is one of the cost-effective investments our nation could make.

Let me make a few points.

Point 1:  U.S. has the potential to be far superior to the European Center and others.

Forget the defeatist rhetoric.  The U.S. has by far the largest meteorological research establishment in the world.  We spend more money on weather research than any other nation.   Why should anyone believe that the best we can do is to follow or equal the EC?  As a numerical modeler myself I firmly believe that our global model could far exceed the performance of the rather conservative EC effort.   To go technical on you for a second, I believe that using ensemble-based data assimilation (in space and time), far better model physics, higher resolution, and better use of observational assets, we could produce vastly improved forecasts, far superior to the EC, with huge economic and safety benefits to the nation.

To put it another way.  Think of the U.S. and the European Center weather prediction efforts as two cars.  One (the U.S.)  has a much bigger engine (knowledge and research base) than the other.   But the big-engine car has a very inefficient transmission (and other deficits) and ends up going slower than the small-engined competition (EC).  Yes, we can hitch a ride with Europeans, but our car could leave them in the dust, if only we had the will to do so.
The U.S. governmental weather prediction effort looks like this...a very big research engine, without the ability to use U.S. massive research power to move our weather prediction forward rapidly.  The guy in the red pants is a NOAA bureaucrat.

Point 2:  U.S. national and regional prediction is a shadow of where it should be.

A lot of the discussion in this blog and elsewhere has been about the inferiority of the U.S. global model (the GFS) to global models of others (EC and UKMET).   But as important as the global models are, they are only part of the story.  The lack of computer power and poor coordination between research and operational weather communities in the U.S. has crippled our ability to move forward towards the high-resolution weather prediction capability that we know represents the future:  probabilistic prediction.  And remember the EC only does global modeling--they are not interested in high-resolution prediction over the U.S. 

It is clear that the future of weather prediction will be to forecast probabilistically for all parameters, with the essential infrastructure to do this being high-resolution ensembles (meaning we will run our forecast models MANY times --say 25-100 times--using different model starting points and model physics).  A number of National Academy study groups have recommended this approach (I have been a member of several of them!) and noted that such ensembles must be high enough resolution (2-4 km grid spacing) to  resolve convection (thunderstorms).   You want to predict major convective outbreaks, like the UNFORECAST derecho (strong convective system with powerful winds) that hit the northeast U.S. last June?  You need this capability.  But the NWS Environmental Modeling Center (EMC) does not have this critical capability because they don't have the computer resources, among other reasons.

In fact, they don't have the computer power to run even current generation weather technology.  For example, the NOAA Earth System Research Lab (ESRL) has developed a new high-resolution prediction system called HRRR (High Resolution Rapid Refresh) that was able to predict the powerful derecho hours before (see graphic).


In contrast, the NWS EMC model (NAM) failed (see graphic below).  The NWS EMC does not have the computer power today to run HRRR operationally.
The NWS NAM model missed this important event 12h before

Point 3:  U.S. modeling inferiority is costing the U.S. private sector big bucks and denying real-time access to the U.S. research community.

U.S. business and governmental interests, such as weather prediction firms and utilities, need the best forecasts, with even modest differences in skill having big financial implications.  Thus, a number of companies and U.S. entities are paying hundreds of the thousands of dollars EACH to get European Center model output.  Yes, we are talking about millions of dollars that is being used to support the EC modeling effort.  The Europeans have a different financial model than the U.S. National Weather Service, one patterned after ancient empires:  they require financial tribute from those wishing the best meteorological "protection The U.S., to its credit, provides model output for free, which not only fosters commerce, but assists nations and researchers all over the world.  But what we give away is clearly not as good as EC's global model, thus allowing the rest of the world to help support their modeling efforts.

Even the U.S. research community is required to pay for real-time access to EC grids.  A long time ago I was able to get EC grids because I was working with an EC researcher.  The EC bureaucrats caught this infraction, cut off the grids, and offered me a "deep discount" rate of $50, 000 a year!  My research using the superior EC grids (which required real-time access) was over.  And my situation is repeated many times over with the rest of the U.S. weather research community.

Epilogue

This weekend I went to a talk by a Stanford political scientist who noted that great nations generally don't fail from external threats--rather they weaken from within.  They get lazy, inefficient, lose their edge, and start making bad decisions.   U.S. operational numerical weather prediction is a prime example of a nation resting on it laurels and falling behind.  We invented numerical weather prediction.  Most scientific and technical advances  in weather prediction have occurred here and STILL DO,   Our research community is still largest and dominant. And with all of that we have lost leadership and have fallen well back into the pack.  And the cost is not just in prestige, but also a weakening of  our nation's economic prowess and needlessly jeopardizing life and limb.

As explained in my previous blogs the route to fixing U.S. numerical weather prediction is clear, including:

(1) Secure sufficient computer power, either by redirecting some of the huge computer resources acquired for climate prediction or using new funds (like money already appropriated in the Hurricane Sandy relief bill)

(2)  Integrate numerical weather prediction research and operations, including replacing the ineffective division between the NWS EMC and the NOAA labs.  NOAA must dedicate enough extramural research funds to entrain the U.S. research community.

(3) The need for effective leadership in NOAA, including a clear vision where weather prediction is going in the U.S.  

There is a real cost to inferior U.S. numerical weather prediction and I am confident the numbers are easily in the billions of dollars.  We have let a crucial piece of the technological infrastructure of our nation to weaken and atrophy.  We must take a new course.

5 comments:

  1. Cliff: "[Great nations] get lazy, inefficient, lose their edge, and start making bad decisions."

    Or no decisions, which after all is still a form of decision-making, akin to throwing darts backwards over one's shoulder while blindfolded and wearing heavy gloves. The budget fiasco now being enacted in the form of Kabuki theater in DC makes grand thinking and objectives such as Cliff's quite impossible; who can build excellence in such an atmosphere?

    Civilization costs money and government can't be operated from a imaginary, bottomless well of hypothetical increased efficiency. Historically low taxation is a wonderful thing, -if- it still allows us have a properly functioning civil society including such features as a reliable Earth observing satellite constellation, predictable and safe employment for civil servants dedicated to maintaining our safety, the ability to exploit research gains in the form of deployed computational facilities of the type Cliff speaks of, etc. So far the empirical evidence suggests that our fascination with free lunch is unjustified.

    ReplyDelete
  2. As someone who worked at NOAA (in research) for many years, I'm fairly certain that none of this will happen unless there are significant changes to NOAA management. The culture there is broken.

    As an example of how bad it is, for many years NOAA had a program (the HPCC (High Performance Computing Consortium) that funded research on IT projects. It was managed by NOAA's CIO office. Proposals were solicited, and a committee would rank and, theoretically, select the best of the proposals. The catch was that committee members were allowed to submit their own proposals! As a result, the vast majority of the money went to committee members who selected their own proposals.

    Unfortunately, there are many more stories like this.

    Given this small example of how corrupt NOAA's culture is, it's difficult to see how the changes Cliff proposes will happen unless someone at the undersecretary level is willing to address these kinds of management issues.

    ReplyDelete
  3. Cliff just an FYI. Using IE9, I cannot view the most recent posting. I can see the heading but no content. Under Firefox, I have no problem. I normally surf using IE9 and this is the first time there has been a compatibility issue with your page.

    ReplyDelete
  4. Ahh. (?) ... "U.S. Numerical Wx-Prediction, .. Aide." (?)

    Just "Joshing" here of course.

    My main question with having followed this theme and having appreciated its importance, would be: So beyond "Write your Congress persons", is there .. something other, that "Joe Wx-Enthusiast", even Professional, can do toward helping to get this, big steel ball, rolling down the road a little quicker.

    — I've written my own two Senators (I'm from CA.). I don't put much stock in my main "Representative", more local.

    ReplyDelete
  5. It is not clear to me why you complain about access to data in real-time. Of course the value of a research institute is its forecasts. Getting access in real-time to data without paying is like software piracy.
    On the other hand, correct me if I am wrong, we can get access to previous forecasts of past events by ECMWF (and so without commercial value) for doing all kind of scientific study that we want.
    The system seems fairly reasonable to me.

    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,...