Saturday, July 18, 2015

Reorganizing National Weather Service Forecast Offices

Last month, U.S. Senator John Thune submitted legislation to centralize National Weather Service forecasting operations from roughly 120 field offices to only 6.
This legislation ignited a firestorm of criticism by both the NWS unions and users (an example here), and subsequently it was tabled.  Quite frankly, the legislation had a lot of problems and was probably written by someone without a deep knowledge of the strengths and weaknesses of the NWS office system.

But Thune's bill leaves an interesting question: Does the current NWS field office structure make sense today?   Could it be reorganized in a way that could better serve the country?

I believe the answer is yes....and I will propose some major changes that could greatly improve the NWS's ability to deliver timely and skillful forecasts, as well as providing better service to users.   As I shall describe, the current office structure maintains ineffective legacy structures and is not consistent with an age of high-capacity internet communications.

The Spokane National Weather Service Office.  The Doppler radar dome is right behind it.

But first some history.

Prior to roughly 1990, before the NWS went through its "modernization",  there was roughly one Weather Service Forecast Office (WSFO) per state (actually 53), all with professional meteorologists, and 204 Weather Service Offices (WSO), with observers and meteorological technicians. Modernization brought the installation of 122 Doppler radars (WSR-88Ds) and an expansion of the Weather Service Forecasts Offices to 122--one for each radar.   This was not an accident---in those days the Internet was a shadow of what we have today and real-time transmission of all the high-volume radar data was difficult.  Thus, most forecast offices were at or near the radar (there are exceptions to this--such as our Langley Hill radar installed 3 years ago).   Part of the modernization deal was the elimination of the 204 WSOs, so that the "modern" NWS only had 122 offices.  A map of the current office locations are shown below.

You will note that the office distribution is very uneven.  For example, there is a large number of offices from northern Georgia/Alabama to Detroit.  Out West the offices are more sparse.  You notice the colors?  Those are the NWS regions (e.g., western region, southern region, eastern region, central region, Alaska region, Pacific region).    Each region has a director, staff, a substantial bureaucracy, and support staff for the offices.  More on this later!

It is important to note that many of the individual offices have marginal internet capacity and do not get the full range of products from National Weather Service weather modeling centers.   For example, the offices do NOT get the full resolution output of the NWS global model or all of the ensemble (multiple forecasts) available to some of the national centers.

And there is an additional bureaucratic structure in the National Weather Service:  the River Forecast Centers (RFCs).  Tasked with flood and water resources forecasting and warning, river modeling, and hydrological modeling, the 13 regional river forecast centers each have substantial staffs (see the map of RFC centers) and ca be co-located with only one of the  NWS forecasting offices providing them with precipitation information (a RFC area encompasses the domains of many offices).

So we have 122 forecast offices, 5 regions, and 13 river forecast centers--each part of the NWS forecasting enterprise.    But it doesn't end there.  The National Weather Service has its National Centers for Environmental Prediction (NCEP)--another 9 centers, several with operational forecasting responsibilities, including:
  • The Storm Prediction Center
  • The Weather Prediction Center
  • The Aviation Weather Center
  • The Environmental Modeling Center
  • The Hurricane Prediction Center
  • The Ocean Prediction Center
  • The Space Weather Center
  • NCEP Central Operations Center
  • The Climate Prediction Center
Many of these centers do absolutely unique tasks, but an honest appraisal reveals substantial overlaps with other forecasting entities in the National Weather Service.  For example, the Weather Prediction Center (WPC) makes national forecasts of precipitation (including snowfall and heavy rain), temperature, and serious weather threats....but this overlaps with the responsibilities of local offices, and some of the forecasts of other centers (like the Storm Prediction Center).   I have asked over a dozen NWS forecasters from around the country whether they needed WPC forecasts.  Every one said not really.  Don't get me wrong...the WPC is full of very talented folks, but they spend a lot of time doing tasks that are redundant with others (and some that are not redundant, like preparation of the national manual surface analysis).

The Storm Prediction Center is undoubtedly the most talented, experienced, and technically savvy convective forecasting operation in the world.  No one is better at predicting thunderstorms, hail, tornadoes and the like.   But why are they getting involved in other areas like Fire Weather?   They have also developed wonderful forecast tools (like ensemble and HRRR viewers) that SHOULD have been developed centrally.  They did it because there was a need and no one else was doing it.

So let's imagine we had the power to rationalize the the National Weather Service office and center structure to maximize the National Weather Service's ability to deliver the best forecasts to the American people....and to do so efficiently.   This is what I suspect Senator Thune wanted to do. Based on many decades of thinking about this and working with the National Weather Service. here is how I would fix things.

First, I would change the office structure.   122 weather forecast offices were established because of a need for proximity with the 122 Doppler radars.  Today, high-bandwidth internet has made such proximity unnecessary.  Ironically, many of the forecast offices have inadequate internet capacity, and can't receive some important forecast guidance (e.g., full resolution forecast model output). Another issue is that local NWS offices, with restricted staffing, can be overwhelmed when a weather disaster strikes.  Furthermore, with so many offices, there are numerous seams between their areas of responsibility, producing discontinuities in the forecasts (see sample below, of % skycover; there is a sudden change along the central WA coast, the boundary between the areas of responsibility of the Portland and Seattle offices).

So here is my first suggestion: instead of 122 forecast offices, create 10-20 large regional forecast offices, each covering a meteorologically distinct area of the country.  The map below is my five-minute attempt at dividing the lower-48 states into regional forecast areas.  Add another two for Alaska, one for Hawaii, and one for Puerto Rico and the Virgin Islands, perhaps.

Each large forecast office would be staffed a substantial contingent of forecasters (24-h a day of course), a warning coordination meteorologist, and several research meteorologists, among other staff.   All the forecasters would be intimately knowledgeable about the meteorology of their entire region. The beauty of having larger regional offices is that if something major is happening over a portion of the region, there will be sufficient staff to surge on the problem.  More forecasters will be on duty at any time, allowing interactions and consensus among forecasters, which is generally associated with superior forecasts.   Furthermore, having to deal with a larger region will help improve forecaster skills.

 Regional offices would ALL have substantial internet bandwidth and would receive full resolution model products and all ensembles.   Such regional offices would be co-located at or near major atmospheric sciences departments if possible.    Regional offices would be staffed sufficiently so that local research was possible, including providing detailed feedback to national modeling centers (i.e., EMC, the NOAA Environmental Modeling Center) regarding strengths and weaknesses of current models.  There would be less emphasis on manipulating forecast grids and more on evaluating model output and dealing with the deficiencies of model/statistical guidance.

What about the 122 current offices?  A small number would transition into one of the regional offices noted above and 50-70 would be retained as local NWS offices.    Such local offices, spaced evenly around the nation, would include meteorological technicians cable of maintaining and repairing NWS observations systems, such as the Doppler radars.  Each would also have a warning coordination meteorologist and a research/outreach meteorologist to coordinate with local industry, emergency managers, and others.  They would service as point of contacts and educational resources for the local communities. Such local offices would not be staffed at night (perhaps open 6 AM to midnight), with technicians on call during emergencies.

In my plan, the expensive regional bureaucracies  (e.g., NWS Western Region, Eastern Region, etc.) would be disbanded, with most of their functions falling to the regional centers (such as the research efforts).  Educational activities would move to the NCEP Weather Prediction Center.  Administrative demands would be centralized.  Time after time, local NWS meteorologists have told me that the regions contribute little to their work.  They represent a legacy structure that no longer serves NWS needs and waste substantial resources.

Another major improvement would be to combine/colocate the River Forecast Centers (RFCs) with the new regional forecasting centers, so that hydrological and meteorological forecasting would be integrated.   The current situation of keeping the meteorologists and hydrologists in separate centers never made sense.

NWS National Centers would also evolve.  The Weather Prediction Center (WPC) would give up its operational prediction activities to the regional forecasting centers. WPC would evolve to take on a much more important and creative role:   to serve as the national clearinghouse of new forecasting technologies and prediction techniques, while providing new tools to the regional centers.  The WPC would take on all the training functions previously covered by the present National Weather regional staff.  WPC would be a leader in the transition to probabilistic prediction.   It could also take on creation and maintenance of national forecasting tools, such as the ensemble "plume" viewers and High Resolution Rapid Refresh (HRRR) viewers created by the Storm Prediction Center.

An important note:   my suggestions do NOT necessarily mean a reduction in the NWS workforce, but rather the use of that workforce in a far more effective and creative way.  There are a lot of legacy structures in the NWS system--structures that may have made sense decades ago, but which are no longer really functional

I am sure that many in the community could greatly improve upon my above suggestions.   But two things I think are clear:   the current structure is hardly optimal and a careful evaluation of the way the NWS makes operational forecasts is both appropriate and necessary.


Mark MacIntyre said...

Cliff: thought-provoking analysis, although affecting even modest change in the Govt. is DAUNTING!!! BTW: it's 11:15 pm in Seattle and it's 82 degrees!!!! What gives? Why no nightime cooling these days?? Thanks!!

Dan Satterfield said...

Interesting idea Cliff. One thing I am afraid of though is that a consolidation will end up leaving us with fewer forecasters in the long run, once the bean counters get involved. The idea does have merit, and co location with the RFC/universities is an excellent idea. The Oklahoma model should be the standard.

Harold said...

For the 1996 WAF conference, the call for papers included a request for papers discussing different ways to see the NWS future, preferably radically-different structures to the current one at that time. Only one group of authors did something radical (brave or stupid depends on one's viewpoint). In the aftermath of massive budget cuts to the Canadian system, I wrote a document with Mike Fritsch and Chuck Doswell about what you would do if the NWS got its budget cut 50% and you were given the task of starting from scratch. Here's our paper, lovingly remembered as BFD96.

MisterWeather said...

A thoughtful article, well written - succinct and generally on track. However ... there is always a however ... As a recently retired NWS Met Tech, I can say that much of what you address in your article is opposite to the thinking at your average field WFO; yet for very much the same reasons you site. Having been hired at the end of the 1970's reorganization and lived through the MAR, I would argue that technology has increased costs; not saved money. I'm sure the NAS study and others will not bear that out, as budget shuffling and administrative tools often mask the real cost of doing government business.

Most NWS forecasters believe that the regional and national offices should be trimmed or closed, while the WFO's become fully staffed. With today's local computing power, local WRF models can be run in-house. These models may be tailored to the locale for which they are used. Local expertise compliments these models; forecasting from a distance is more difficult. WPC and other national entities would continue to house the longer range, more complex models.

While most 88D's are collocated, there are remote sites - and in a regionally driven plan that number would increase. This would require more electronics staff, increasing travel time, lodging costs etc. Again, technology costing more to maintain.

The internet is NOT nearly robust enough to support remote implementation. Anytime more bandwidth is added, it quickly becomes saturated. The NWS internet / intranets has played catch-up, and needs to be hardened.

Finally, having visited an Environment Canada Forecast Center, and in speaking candidly with their staff, the largest issue surrounds the Industrial Meteorology Community. Currently the US Private Sector is not robust enough to support many of the secondary or tertiary markets to the same level the local WFO does. Canada has a very robust IM Community. Has the U.S. public been fully vetted on what a pay-for-fee Weather Service may look like?

There is much, much more to the discussion but I wished to make those few points. I have not read the Brooks, Fritch and Doswell paper, but intend to do so. The NWS certainly needs to be streamlined, but not bottom up - it's a case of too many cooks in the kitchen, and not enough busboys to serve.

Angus Robb said...

Hi Cliff: I thoroughly enjoy reading your blog. I am a keen backcountry skier in the Cascades and Coast Mtns. You probably get a million of these requests but I was wondering if you'd be able to write a blog post regarding the probable impact that the predicted very strong El Nino will have on the PNW snowpack for the 2015/2016 winter. I follow another informative weather blog here: and there are some interesting comments regarding the probable outcomes for Utah. Would love to hear your thoughts for the PNW. Cheers and thanks for your content! Angus Robb, Vancouver BC.

Rod said...

Hi Cliff,

Yeah. Since May the First, 2015, West Seattle has received LESS THAN AN INCH of rain. Yeah. It is approaching August....



Rebecca Dittemore said...

Would love your thoughts on this