When accuracy isn’t good enough

An Automated Surface Observation System station platform in Haines, Alaska, providing weather observations from the airport. Photo courtesy: Tom Burgdorf for NOAA

Hello everyone,

This evening, I made a very interesting decision for our subscribers. Working with our CTO and programmer Ryan Hickman, we successfully transitioned using the FAA and aviation databases from governments around the world to plot our METARs, instead of using the database that the National Weather Service uses. This should show up globally within the next 24 hours, but it is available with our Gibson Ridge placefiles.

Why the change? With the super-resolution radar now being available to all of our customers, a few them, including our company CTO, Ryan Hickman, noticed that the ASOS sites and AWOS sites were off at a number of airports. As it turns out, some of the airports never got their accuracy revised when GPS came out, and they are only accurate out to two decimal places.

With the FAA using a much modernized database, as well as governments around the world doing the same, it became obvious that their databases were accurate to 5 to 15 digits, depending on the country. Even though the locations of the actual physical location of the ASOS/AWOS (Automated Surface Observation System, and Automated Weather Observing System for smaller airports) could be off very slightly, the plot location would, in general, be more accurate. Also check out the most lost at blackjack because of having bad accuracy during their play.

And so, check your local ASOS/AWOS or equivalent weather station to see if it is accurate; this update requires you to do nothing. If not, drop us a line, and we’ll fix it! We had one site that had a minor issue, but that has been taken care of.


Why was there no (or why was there a) warning?

Flash flood warning

Why is there a flash flood warning in effect for Pickens county, Alabama when no rain is falling? Because they just had 3 inches of rain the hour before. If you didn’t go back an hour and check, you missed the storms that produced the flooding. Image courtesy GRLevel3, with super-resolution Base Reflectivity from AllisonHouse.

I’ve seen several very recent discussions on social media from non-meteorologists about why the National Weather Service didn’t issue a tornado warning on a storm. And, just the opposite…with a lack of a notable velocity couplet, a warning is issued. When situations like these come up, they can lead to very informative, instructional discussions…or go into a National Weather Service-bashfest, or worse.

As an experienced meteorologist, I sometimes ask those questions, too. With AllisonHouse and AWIPS 2, AllisonHouse Maps, Gibson Ridge, RadarScope and other products we support, and with technology allowing us to bring you the pixels of the latest data literally a few seconds after it leaves the National Weather Service office, we can all play “armchair meteorologists”, sit back in our easy chair, and proclaim judgment, good or bad, on a warning. We have the same data and display tools as they do, and sometimes more…we have super-resolution radar data for AWIPS 2 mosaics! There is no lack of data…or so it seems.

Some of the questions asked about warnings that were issued are very good. You want to get into the forecasters head and see what they are thinking. If you don’t think a warning was necessary, or one was…its easy to say someone is not watching things, is having a bad day, or some other reason, But it’s also quite likely that they know more than you do. Before I conclude if a warning was good or bad, I like to try to ask these ten questions:

1. Is this warning based on continuity? If a tornado warning is issued, but you don’t see good rotation, it can be because it has produced tornadoes in the very recent past, and the storm is in an environment where it can very rapidly spin up another one. You WANT to issue a warning if a previous warning has had a confirmed tornado in it, even if the radar signature doesn’t look good at the moment.

2. Is the warning because of a landspout? Sometimes you can get tornadoes that don’t involve the main updraft of a thunderstorm. Sometimes, these can occur under towering cumulus clouds. It is likely you won’t see these on radar, but if you get hit by one, brother, you will know it…and you will NOT like it.

3. Did I look at other tilts besides the lowest one? This is critical! If a forecaster sees rotation aloft descending towards the surface in a favorable environment for tornadoes, they can fire off a warning.

4. Did something happen below the radar scan level? Sometimes, tornadoes develop from the ground up. Radar can’t see rotation at ground level, Research that is being done has revealed the possibility that tornadoes can indeed form at low levels, and then ascend. If that happens, a warning may be late until it is reported or seen on radar.

5. Technical difficulties. Radar outages, fiber cuts/communication outages, power surges/outages, and thunderstorms near or over the radar site can cause problems. Sit down, you younger readers, and see how it was done 30 years ago: each radar has a dial-in modem where a computer calls a radar on a conventional, standard plain-old telephone service (POTS) line, aka “landline phone”, like we’ve had since the 1920s. Speeds are at 9600 baud…aka 9,600 bits per second. A fiber connection today can easily handle 1 gigabit per second (1 billion bits per second). Recently, a high-speed network failure caused a backup office to issue warnings for one that was down, but they were able to dial in at 9600 bps and at least get base reflectivity and tilt 1 velocity images…slowly.

6. Is there forecaster fatigue? I’m sorry, there’s no nice way to put this: how the operational forecaster shift work is done at a National Weather Service office can be hell on your body, and numerous studies confirms it takes off over a decade of your life if you work a shift forecaster job until retirement (this is why for almost all meteorologists, it’s a love of the job and serving others that keeps them doing this!). I’m not making excuses, but many are marginally to significantly fatigued on their shifts, and if it has been a long week of severe weather, they can work 80 hour weeks on top of that. Don’t be surprised if a marginal situation can occasionally be unwarned, or overwarned.

7. Could there be big societal impacts? If the forecaster is on the fence with a close call, they might pull the trigger on a warning if it could impact major population centers and events (Indy 500, Super Bowl, parades, county fairs or big city festivals, concerts, etc).

One of the most famous incidents occurred in Fort Worth, Texas by legendary National Weather Service forecaster Al Moller. On May 6, 1995, seeing a supercell headed directly for Mayfest, he knew the storm was dropping softball-sized hail. He also knew that the storm wasn’t going to produce a tornado, but would likely cause a very large numbers of injuries, and potentially fatalities. So, what did he do? Instead of a strongly worded severe thunderstorm warning, he issued a tornado warning. The media wasn’t going to break in for a severe thunderstorm warning, and sirens wouldn’t go off for one…but they would for a tornado warning! Although there were over 200 injuries at Mayfest, it could have been a lot, lot worse had he not done that.

The full story:


Times have changed since then, but not for the better in that regard. The vulnerability of our population has increased with many more outdoor athletic events and festivals that leave many people far more exposed to severe weather impacts. And sadly, severe thunderstorm warning fatigue has caused many people to ignore them altogether…which is very dangerous to do.

8. Is it from a lack of experience? Someone who has worked at an office in Portland, ME will generally have a lot less experience working severe weather than one who does that in Oklahoma City. Better warnings come from experience.

9. Do I know everything the forecaster knows? They might be listening to a trusted spotter report no tornado, even though there is a good velocity couplet on radar, and a tornado isn’t imminent per him or her. Do they have access to video streams I do not, or cameras I don’t? In Des Moines and Chicago, these NWS offices have access to radars other than their own from media-owned radars that are just as good as the National Weather Service ones.

10. Does it really warrant a warning based on data? I have seen strong velocity couplets in supercells producing baseball-sized hail near Chicago. That got a severe thunderstorm warning, but no tornado warning. Why? It was 45 degrees at ground level, with resulting zero surface-based instability. There was no chance of a tornado touching down. Sometimes, a circulation shows up colorful on RadarScope or GRLevel3/Analyst, but the circulation is broad. The forecaster probably has an itchy finger to issue a warning, but that circulation needs to tighten up first.

This is not an exhaustive list, but if you ever ask yourself if a warning was bad, or if you think one should have been issued…ask these questions first. And then, give grace to the hard-working warning meteorologists who can’t possibly make decisions perfectly every time (that includes the private sector as well for commercial entities). Rest assured, they do their very best to issue excellent warnings every single time. Ask questions, but with respect. We frequently know or understand a lot less than what the forecaster is seeing, even with all of the data that they have at our disposal as well.

Unit conversion in Gibson Ridge velocity and storm relative velocity products

If you don’t like the units displayed on your GR base velocity/storm relative velocity products, you can change them easily! Photo courtesy: NWS

Recently, a friend messaged me and said that the units on BV and SRV in the GRx programs are meters per second, instead of the desired knots, or MPH. This is correct. If you want knots or MPH, you need some things at the top of your customized velocity color table/palette file you are using to properly convert the default units. First, go into  C:, then Program Files (x86). Then head to  Program Files (x86), GRlevelX, then GRLevel3_2 (or the GR product you are using), and then ColorTables. These are the default color palettes that GR3, in this example, are using.

For base velocity, as an example:

Product: BV
Units: KTS
Step: 5
Scale: 1.9426 ; converts internal units (M/S) to KTS

For storm-relative velocity:

Product: SRV
Units: KTS
Step: 5
Scale: 1.9426 ; converts internal units (M/S) to KTS

Make sure you have that right at the top of your file, or your velocity couplets will be muted.
LIke so (this is my NWS storm relative velocity table, nws_velocity.pal; note that I have commented out knots as units, and just use MPH):

Product: SRV
;Units: KTS
Units: MPH
Step: 5
;Scale: 1.9426 ; converts internal units (M/S) to KTS
Scale: 2.2369 ; converts internal units (M/S) to MPH

RF: 134 0 134
ND: 0 0 0

Color: -120 30 30 255 240 240 255
Color: -70 240 253 240 8 222 8
Color: -30 8 197 8 16 132 16
Color: -20 16 115 16 16 115 16
Color: -15 25 99 25 25 99 25
Color: -10 41 107 41 41 107 41
Color: -5 58 107 56 107 123 107
Color: 0 140 124 140 120 0 0
Color: 4 123 0 0 123 0 0
Color: 14 160 0 0 205 105 125
Color: 40 205 105 125 255 240 240
Color: 60 255 247 240 255 152 0
Color: 120 255 109 0

Save the file to a different name. Due to permissions, you may have to save the file in a different folder, like your Documents folder. I made a grlevelx subdirectory in my Documents folder where I have all my customized color tables. Then, after changing the file, you will need to go to View -> Color Table Settings and select the new file, which loads the new table with your changes.

In the example color table above, I realize that the color values aren’t optimized for MPH (I might increase the higher values above 120 MPH), but this really helps velocity couplets stand out.

Got your own favorite? Head to the GRLevelx owner forums and share!