April 21, 2015

The U.S. Air Force Turns to a Foreign Weather Forecasting System

The United States Air Force has decided to drop its American weather prediction system (the Weather Research and Forecasting model, WRF) for the forecasting system developed by the United Kingdom (UKMET Office Unified Model).

As described below, this decision is a terrible mistake and will ensure substantial damage to U.S. weather prediction capability, waste precious financial resources, and  undermine the U.S. Air Force's capacity to provide the best possible forecasts for U.S. pilots and Air Force operations.


This blog will tell you about this unfortunate situation, document a flawed decision-making process, describe the downside of this decision, and call for better-informed public officials and legislators to intervene.

The current situation

Today, the US Air Force makes regional forecasts around the world using the WRF model.  WRF is an extraordinary success story;  developed at the National Center For Atmospheric Research  (NCAR) in Boulder, Colorado, WRF is used by thousands of users in the U.S., and is the predominant model used in the research community and the private sector.  WRF is also heavily used by the National Weather Service and by many thousands of individuals, groups, and weather forecast entities around the world.    WRF is probably the best example of a community model: highly flexible, state of the art, adaptive, with advances from the research community flowing into the effort, resulting in constant improvement.


During the mid-2000s, the AF took on WRF as their main regional modeling tool, using the U.S. global GFS (Global Forecasting System) model for their global predictions.  The global model is used to provide boundary conditions for the regional model (WRF).

The AF adoption of WRF was a win-win for the nation.   AF funding contributed to maintaining and improving WRF; in fact, the AF was the largest financial supporter of WRF.   The AF in turn had the best possible regional model, one that was easy to use and highly capable, and a model that took advantage of the efforts of the vast U.S. weather research community.   An improved WRF helped drive U.S. weather modeling research and was taken on by many private sector firms.  The U.S. was clearly the world leader in this domain.

The Air Force fumbles

Late last year it became known the Air Force Weather Agency, which runs AF numerical forecast models, had decided to drop WRF and NOAA's GFS model, and turn to a foreign modeling system:  the UKMET office model.   A recent story in the Washington Post discussed this decision.   This decision was made without talking to U.S. national weather modeling partners (the National Weather Service and the U.S. Navy) and appears to be the decision of one individual, Ralph Stoffler, acting head of the Air Force Weather Agency.  Mr. Stoffler was an AF weather officer and has a BS in meteorology.

Ralph Stoffler

Checking with my contacts in the Air Force, I have learned that there were no long-term verification/comparison runs to demonstrate that the the UKMET office model would be superior to WRF.  It is stunning that such a major decision would be taken without strong evidence of improvement.

Mr. Stoffler's plan greatly expands AF modeling into the global arena, moving to DUPLICATE the U.S. global prediction efforts completed by the National Weather Service and the U.S. Navy's Fleet Numerical Meteorology and Oceanography Center.   There has always been an unfortunate relationship between Navy and Air Force weather operations, with substantial duplication of efforts.   But the new AF plan goes beyond this and is highly wasteful of U.S. weather prediction resources.

To run a state-of-the-art global model requires large resources, including the acquisition and quality control of vast amounts of data from many different satellites.  A high-resolution global model also demands huge computer resources. Clearly, Mr. Stoffler has not considered these issues in depth before proposing his new approach.

Let me underline the fact that there is no evidence that the UKMET office model is a superior regional model.  WRF has far more physics options and is much more widely tested at high resolution around the world.  UKMET Office global forecasts have slightly better verification scores in the Northern Hemisphere that the NOAA GFS, but these differences are small.   Furthermore, the NOAA GFS model is now undergoing rapid improvements (made possible by the new supercomputers NOAA is getting this year) and I suspect that by the end of 2016, the GFS will be as good, if not better, than UKMET.  Thus, the AF could well end up with an inferior global forecast.

But it is worse than that.   The UKMET office model is known to be difficult and unwieldy to use, and there will be a hugely expensive spin up at the AF to run this model and connect it to their production suite of products.  Resource demands in running a state of the science global model are huge. And as I have described in previous blogs, the U.S. has TOO MANY models running, resulting in division of effort and waste.  The AF is taking the wrong road.

But let's be honest here.  This situation is a warning to the National Weather Service and the U.S. weather modeling efforts---if the U.S. Air Force is making plans to use overseas modeling systems, this is not a good sign.

Major Impacts on WRF

Air Force funding has been critical for the viability of the national regional weather forecast system (WRF), the one used here at the University of Washington, by the way.   The AF has been the main Federal financial supporter of WRF.  The loss of AF funding will greatly undermine WRF and its future development (including the revolutionary global MPAS model that would be its successor).   WRF is the model used in many key forecasting systems in the U.S., such as the National Weather Service High Resolution Rapid Refresh system. The economic and scientific impacts of the AF action would be large and damaging to the U.S. weather prediction enterprise.


What needs to be done

The U.S. meteorological community and others need to speak loudly to Air Force management, the current administration, Congress, and others to stop this ill-advised AF action.  The damage to the U.S. weather prediction capacity and AF weather prediction will be substantial if the proposed plan is followed.  There is time to turn this around and restore a rational approach to weather prediction modeling the in the Air Force.   Here in Washington State, I hope our Senators, Patty Murray and Maria Cantwell, will intervene.


8 comments:

  1. "Furthermore, the NOAA GFS model is now undergoing rapid improvements (made possible by the new supercomputers NOAA is getting this year) and I suspect that by the end of 2016, the GFS will be as good, if not better, than UKMET."

    If I understand your comment - the UKMET is not getting any new computing power and/or improvements to their model over the next two years? Any idea why they're deciding to keep the model static?

    ReplyDelete
  2. Expecting US Navy and USMC missions to execute with the "same forecast" now seems even less likely. "Joint Command's" weather coordination challenges were (80s - 90s) handled knowledgeably and successfully. Now with USN and USMC not even on the team, I shudder to think of breaking through security issues to integrate a joint forecast for a TS mission. Could we have convinced Saudi Arabia to turn over a daily direct feed of weather observations to the UK versus the USAF for integration into a global model forecast? Political sensitivities will vary regionally. I question this solution to a computer modeling problem. I also wonder about other US plans involving cooperation of the three services and the NWS, a la NEXRAD.

    ReplyDelete
  3. One more effort to 'disappear' the US as a sovereign nation, I fear. Probably another "penny wise, pound ($0) foolish" move Whether through incompetence or plan, it's a terrible thing to watch. Time once again to contact our fed legislators. Thanks for posting this, Cliff. Stuff like this must drive you batty.

    ReplyDelete
  4. The military should not be counting on services for foreign nations.

    ReplyDelete
  5. As an Air Force Weather veteran, it confuses me to see such a stark move from 2nd Weather Group at Offutt AFB; the WRF has been heavily invested in and helps forecasters shape their mesoscale and sub-synoptic pattern very well. The claim that they require a different model to produce dust and aerosol forecasts is disingenuous at best, as I completed an applied meteorology study while deployed in the Middle East that demonstrated the viability of using hydrometeor-based models for lithometeor forecasting....

    As a proud U.S. Navy Sailor these days, serving as a Meteorology & Oceanography Forecaster in the Reserves and a practicing meteorologist in state government who specializes in lithometeors (smoke), this gives me some excellent questions to ask when I visit the heart of U.S. Navy Weather in a few weeks...

    ReplyDelete
  6. It is a bit unfair to say the UM has not been compared with WRF for hi-res prediction, as it has been demonstrated as part of NOAA's HWT Spring Forecast Experiment for several years in a row, and it consistently out-performed the WRF based models. That said, the forecasts were run by the UKMET office - where they undoubtedly have exceptional expertise to achieve good forecasts and extensive computing capacity that the AFWA likely won't have. They get a software license, the rest is on AFWA to make it perform well.

    ReplyDelete
  7. Gman,
    I disagree. The HWT Spring Forecast Experiment was not a valid comparison. Different large scale forcing was used (GFS versus UKMET met model) and model set up (e.g., vertical resolution) differed. Not a fair comparision of the mesoscale forecasting ability of both systems...cliff

    ReplyDelete
  8. "The military should not be counting on services for foreign nations." - There are actually many nations relying on the US Army too much, worldwide. Despite the fact, that e.g. in many European countries US Army activities in the world are highly unpopular, many nations not only in Europe just rely on our army to cover their backs in case anything happens. E.g. with Russia, they wouldn´t have a chance! / Scott @ usmilitarylendingcorp.com

    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,...