Is the U.S. Government Shutdown Hurting Weather Model Accuracy?

My phone has been ringing from media asking whether the partial government shutdown is degrading the skill of the National Weather Service model, the GFS. 

What is driving this interest?    First, a number of media outlets, including the well-known Washington Post Capital Weather Gang, have made the claim of worsening U.S. forecasts (see examples below).

These stories describe a situation in which important observations, the input data streams for numerical weather prediction, are not being used or are degraded because of changes in coding of data formats.  As a result, the initialization (starting point) of U.S. global forecasts are degraded, lessening the skill of the predictions.

Then there is the U.S. model forecast skill itself, which appears to have worsened (compared to the European Center or the UK MET office model) just after the shutdown (on December 22nd).

The graph below from the National Weather Service illustrates this. Skill is plotted against time for the 5-day, upper level forecast for day 5 over the northern hemisphere (1 is a perfect forecast).  The black line shows the skill of the U.S. GFS model and the red dotted line presents the skill of the world-leading ECMWF model.  U.S. skill seems decline, both in absolute sense and relative to ECMWF right after December 22nd.

So are these claims of shutdown forecast problems well founded?   Is U.S. forecast skill drifting because of a lack of skilled NWS personnel minding the shop?

I think these claims are baseless.  

One of the first things I did was to check with some very well connected colleagues in NOAA and they confirmed these stories are nonsense.  Yes, many NOAA/NWS employees are not working, but since the models are considering essential for national security some staff are working--monitoring the global forecast system doing whatever it takes to keep it working smoothly.  Those NOAA/NWS staff are a dedicated lot!

But being a fact-driven type of guy, I decided to see if I could PROVE that the degradation claims are false with real data.

There are other explanations for the forecasts getting worse that have nothing to do with poor initial data.  For example, certain weather regimes may be more difficult for the U.S. model and perhaps that is all that we are seeing.

If the initial data was bad the short term forecasts would be less skillful.  So let's look at the skill of the 1-day forecast in the middle troposphere (500 hPa) for the U.S. GFS model (black line) and several others (the European Center is red).   No hint of any shutdown changes.

What about quality of the wind initialization at 500 hPa when compared to radiosondes (balloon-launched weather instruments)?  Little evidence of a shut-down effect (see below).

I have looked at many other fields and the answer is the same:  there is NO EVIDENCE that the initialization of the U.S. global model has been degraded as a result of the partial government shutdown.

In contrast, the predictions of the University of Washington regional prediction systems HAVE been hurt, because one of the important data streams we get from NOAA--the NOAA/NWS RAP model grids--has been cut  off during the government closure.  The shutdown is a disaster for weather prediction and weather research, but degradation of the U.S. global forecasts is really not an issue.