Saturday, April 16, 2011

How to recognize rotation on Doppler radar

One of the most frequent questions I am asked by people is how we can see tornadoes on our Doppler radar.  Given that the shortwave that has trekked across the southern US the last few days has produced (and is still producing...) dozens of tornadoes, I collected a few images from radars to try and show the different signatures associated with tornadoes.  I believe that the more people know about looking for threatening weather, even in this sophisticated way, the safer we can be.  Nevertheless, personal diagnosis of radar images is no substitute for warnings from the experts at the National Weather Service.  Always pay attention to the warnings and advisories issued by the Weather Service--they're there to protect you.

Let's start with some basics.  The basic structure of a thunderstorm consists of two parts--warm, moist air rising in an updraft, and cool, rain-laden air sinking in a downdraft.

The kinds of thunderstorms that most often produce long-lived or strong tornadoes are called "supercells".  At the heart of a supercell storm is a rotating column of air called the mesocyclone.  Just because this rotating column of air is present doesn't mean that the storm is producing or will produce a tornado--all supercells have a rotating mesocyclone, but only some produce a tornado. The presence of rotation in the middle of a storm will somewhat twist and turn how the updraft and downdraft actually look.  Let's switch to an idealized block diagram looking down from the top of a storm.  The warmer, moister air is usually to the south or southeast of the storm, so that's where the storm starts drawing in its moisture.  Because there is a rotating core in supercells, it can draw in warm moist air somewhat laterally instead of just vertically.  This air that is drawn in is called the "inflow".
The inflow rises up as it enters the storm, forming the updraft.  As the warm, moist air rises, it cools off, and as the air cools off, water vapor condenses back out into liquid water and rain is formed.  Eventually, at the top of the cloud, there is rather cold, rain-filled (or hail and ice filled too) air that is no longer buoyant because it's so cold.  So, it starts sinking.  Now, for the storm to keep growing and maintain itself, the downdraft can't sink into the updraft--this would choke off the updraft and kill the storm.  But in an organized storm like a supercell, this doesn't happen.  The downdraft air takes one of two paths (usually).  A lot of the air gets pushed out in front of the updraft (to the east, typically) by the strong upper-level jets that we find at the top of the storm.  The same jet streaks that are providing the divergence aloft to support the updraft can also carry the cold, rain-filled air off ahead of the storm and away from the updraft.  Thus, we'd expect to see a lot of radar returns out ahead of the mesocyclone area, since radar beams reflect off of raindrops.  This large area of precipitation and sinking air out in front of the mesocyclone is called the "front flank downdraft".

However, because the center of the storm is rotating, some of the downdraft air can get wrapped around the back side of the storm as well.  Thus, in most supercells, we see an intense, though narrow band of rain wrapped around on the back side of the mesocyclone.  This part of the downdraft is usually called the "rear flank downdraft".  Here's a block diagram of the two paths I just metioned:

So what would our resulting radar reflectivity image look like for this kind of situation.  We expect to see lots of returns out ahead of the storm because that's where most of the rain-filled downdraft air is going.  There would also probably be a weaker area of returns (or no returns) where the inflow region happened to be, since that warm, buoyant and rising and rain is probably not falling through that area.  Then we also expect a narrow band of radar reflectivity returns behind the mesocyclone corresponding to that portion of the downdraft that gets wrapped around by the rotating air.  The light green outline here shows generally what we'd expect to see on radar:

Basically we expect to see radar reflectivity returns everywhere that the downdraft is dominant.  Does this structure start to look familiar?  This pattern of reflectivity return is the iconic signature of a supercell thunderstorm.  Notice how since the rear flank downdraft--the downdraft air getting wrapped around the mesocyclone--is forced out into the inflow air, and there is a protruding area of reflectivity return on the southern side of the storm.  This defines what is commonly called the "hook echo" that you so often hear about on TV. Hook echoes are a sign of a strong rear flank downdraft coupled with a strong inflow region.  This is only really feasible if you have very strong rotation in the storm.  Notice how there is no real reflectivity return in the inflow region.  Because air is rising in this region, no rain is falling there and we get no radar return.  However, this seems to "notch out" the otherwise round reflectivity signature of the storm.  This area of no return cutting into the storm is referred to as the "inflow notch".  It helps define the hook echo and shows you the area where the storm is bringing in its warm, moist air.  A deep inflow notch often means strong inflow and updrafts and correspondingly a strong storm.

So what does this look like on an actual radar reflectivity image?  Here's a picture of a radar reflectivity image from a supercell thunderstorm from this afternoon in North Carolina that was producing a tornado at the time:
Radar 0.5 degree Base Reflectivity image from KRAX, 2040Z, April 16, 2011. 
I've overlaid labels of several of the features we talked about.  You can see all that enhanced return to the north and east of the center of the storm--that's all rain and hail falling in the front-flank downdraft region.  That sharp cut into the reflectivity on the eastern side is the inflow notch--this is the area where the storm is drawing in its warm, moist air.  Rising motion dominates here, and as such no rain is falling (and we see no reflectivity returns).  However, there is a clear, hook-shaped appendage hanging down from the storm--a classic hook echo.  This is that other part of the downdraft air (the rear flank downdraft) that is wrapped around the mesocyclone as it descends, bringing rain (and, consequently, radar returns) with it.  Some of that enhanced return in the big ball at the end of the hook might actually be the radar reflecting off of debris being kicked up by a tornado on the ground.  The hook echo and the inflow notch wrap around the area of strongest rotation--the mesocyclone.  If the mesocyclone becomes very tightly wrapped at low-levels, it can form a tornado.

So that's all and good for reflectivity images--we can recognize these kinds of structures that show us what rotation is doing to the environment.  But what about seeing rotation itself--looking at the wind velocities?  Our Doppler radars can do this, and that's a big reason the Doppler capability was added to the radars.  They do have some limitations, though.  But first...let's get back to concepts.

We have an area of strong rotation in the mesocyclone. Winds there are probably much faster than they are elsewhere, and they also change speed and direction over a very small spatial area.  Our Doppler radars can measure wind speeds, but only in two directions--air moving toward the radar and air moving away from the radar.  The radar measures wind speed by looking at phase differences caused by moving targets along the radar beam.  However, things moving across the radar beam perpendicular to it do not cause phase shifts along the beam.  So, we can actually only measure wind in two directions--towards and away.  Turns out that's all we really need.

Let's go back to the mesocyclone diagram and insert a theoretical radar location.  Since the radar measures winds moving toward and away from it, the radar's location is going to dictate what the return looks like.  In the image below I've randomly added a radar located to the southeast of the storm.

As the radar looks toward the storm, its beam follows a path similar to the dotted line.  You can see that to the right of the dotted line, winds in the mesocyclone (which is rotating counter-clockwise here as most mesocyclones do...but that's a topic for another blog...) are generally moving away from the radar.  But, to the left of the dotted line, winds are generally moving toward the radar.  We can assume that if there's a compact space where winds suddenly shift from moving away from the radar to moving toward the radar, there's probably something rotating there.  Unfortunately, we just can't tell what's going on when winds are blowing perpendicular to the radar beam (unless you use another radar located at another location, for instance).  But, in general, anything rotating should have one side strongly blowing away from the radar and the other side strongly blowing toward the radar.

What will this look like on a radar velocity image?  Usually, warm colors like reds and oranges are used for winds blowing away from the radar.  Cool colors like greens and blues are used for winds blowing toward the radar.  If we assume that the rotation is very concentrated in the mesocyclone, we'd then expect the radar velocity return to look something like this:
We'd expect a small area of bright reds (or oranges, or other warm colors) to be located right next to a small area of bright greens (or some other cool color) with a line parallel to the radar beam running in between them.  This is important--the red and green enhancements must be on either side of a line drawn radially outward from the center of the radar.  If they are not, it implies other things (like convergence or divergence--which are also important, but not rotation...).  The reds once again indicate air moving away from the radar and the greens indicate air moving toward the area.  If they are right next to each other, that implies that the winds are changing very rapidly across a small-distance--and that there's probably strong rotation.

What does this look like in real life?  Here's the base velocity image corresponding to the radar reflectivity image of the storm in North Carolina that I showed above.
Radar 0.5 degree Base Velocity image from KRAX, 2040Z, April 16, 2011.
Note that in the radar image above, the radar location is actually off to the upper left side of the image.  I've drawn a black dotted line along what is approximately the radar beam path.  Notice how there's a very localized area of bright green colors right next to an area of bright red colors exactly where we guessed the mesocyclone to be from the reflectivity image.  If you follow the arrows I drew showing the approximate rotation of the mesocyclone, you'll see that the bright green side corresponds to air moving toward the radar and the bright red side corresponds to air moving away from the radar. we have a rotation signature right where we expected it to be--in the mesocyclone region.  This close pairing of bright red and green colors on the radar is known as a "velocity couplet". It's a hallmark of rotation.

So those are the basics.  Now let's look at a couple more examples, just for practice.  Here's a reflectivity image of the powerful supercell and tornado that did lots of damage right as it moved over downtown Raleigh, North Carolina.
Radar 0.5 degree Base Reflectivity image from KRAX, 1940Z, April 16, 2011.
Once again, off the north and the east of the storm is the front flank downdraft--where most of the downdraft air, hail and rain is falling.  We see an inflow notch cutting in and a strong hook echo wrapping around the bottom side of the storm.  Right in between we would expect to see the mesocyclone and tornado.  You can really visualize in these reflectivity images how the mesocyclone is wrapping things up around it, forming that hook echo. How does this look on the velocity image?
Radar 0.5 degree Base Velocity image from KRAX, 1940Z, April 16, 2011.
The couplet is not as strong, but there's lots of air inbound to the radar on the southern side of the mesocyclone.  You can actually see where the radar is located in this image--in the middle of that black circle on the lower right side.  Once again, if you match up the arrows I've drawn showing the rotation to the colors, you can see that there's green on the side where the air is blowing toward the radar and red where the air is blowing away from the radar.  It matches up very well.

Of course, you'll notice on these velocity images that there are a lot of places where green and red velocity colors are right next to each other--and obviously they don't all mean rotation.  This is why I also discussed the reflectivity images--it's best to use the reflectivity images to know where to look for rotation before looking at the velocity images to verify your suspicions. If you just use the velocity images, you can sometimes get rather lost.

But, then again, sometimes the reflectivity images don't tell you everything and don't show the clear structures we saw in today's North Carolina storms.  For example, here's another tornadic supercell in Oklahoma two days ago in reflectivity:
Radar 0.5 degree Base Reflectivity image from KTLX, 202Z, April 15, 2011.
Not the clearest-cut structure here.  You might guess that that sharp cut of lower reflectivity values into the southern side of the storm could be an inflow notch.  But, then the hook to the west of that really doesn't look like a hook.  This is what's called an "HP supercell" for 'high precipitation".  When there's a whole lot of rain or hail, sometimes there is so much reflectivity return in different places that it can mask the signatures of rotation that would otherwise have been seen in the reflectivity.  But it's a strong enough storm that a look at velocity is warranted.  Here it is:
Radar 0.5 degree Base Velocity image from KTLX, 202Z, April 15, 2011..
Wow--that's some intense rotation right there.  I drew in the arrows again to show the area we're looking at.  The radar is located off to the northwest of this storm.  That's a very strong velocity couplet indeed.  But, the evidence of this rotation was somewhat hidden in the reflectivity image.  Also, the further you get from the radar, not only does your resolution decrease, but your radar beam is getting higher and higher above the ground.  This storm is rather far from the radar, and we're probably not looking at the "low-level" features of this storm.  Nevertheless, this storm is clearly rotating and with a strong velocity couplet like that, it's no wonder it was producing a tornado.

As I said before, often it can be somewhat difficult to really tell what is rotation and what's not by just looking at the velocity image alone  Often, as we saw above, the reflectivity image may not be very much help either.  In these cases, you just have to investigate and be persistent--check higher tilts of the radar to see if there's clearer signs of rotation aloft, look at a loop of images to see any suspicious signs in the recent past, and so on. However, the velocity couplets for strong storms will usually stand out.  Just as an exercise, here's a wider view of the reflectivity images of the storms in eastern Oklahoma on Thursday evening:
Radar 0.5 degree Base Reflectivity image from KSRX, 231Z, April 15, 2011.
A mess of several nasty storms.  Could that be a very fat hook echo on the southwest side of the furthest northeast cell?  Or what about the smaller appendage sticking out of that same cell just to the east of the fat hook?  And what about these other cells further to the southwest--they're definitely intense, but do they have rotation in them?

Here's the raw velocity image for that same time:
Radar 0.5 degree Base Velocity image from KSRX, 231Z, April 15, 2011.
There's a lot going on in that velocity image.  Can you spot areas that may be rotation?  On the image below I've circled some of the areas that I would want to look at more closely.  Some are pretty obvious...others are just suspicious enough to warrant a look:
Well, that's all I have for now.  I hope this has given you enough background to start spotting potentially rotating storms on radar.  Once again--just because a storm is rotating doesn't mean it's producing a tornado. And, even more importantly, just because you don't think you see any rotation on the radar but the Weather Service has a warning out does not mean you should ignore the warning. There are professional meteorologists whose job it is to analyze radar images and lots of other data to determine what is going on and who to warm.  Trust their judgment and take cover when there's a warning.


  1. Thanks for the great information!
    Is there any publicly available software for tracking the weather, something like what you are using for the radar images? I can't find any live Doppler radar images on the web.

  2. Hi Rob,

    It's very difficult to find high-resolution radar images on the internet. There are two types of radar data--level 3 data is extensively post-processed data data that's widely used for images on the web, but it's at a lower resolution. The images I've been showing use level 2 data, which is closer to the raw data and is at a higher resolution before it has been post-processed. These level 2 data files are very large--each radar volume scan can be several megabytes. This is why you usually don't see these online.

    To make the images on this site, I used a program called Gibson Ridge Level 2 Radar viewer. The company offers a few different products and their website is They offer a 21 day trial of their programs before you have to purchase them. If you use one of their level 2 radar viewers, you'll also have to input an online database of level-2 radar data. Searching online can find a few free sources of the level 2 data.

    As far as just websites, the National Weather Service website radars or the NEXRAD images at the College of Dupage website ( are what I use when I don't have Gibson Ridge. They're level 3 images, but I just make do.

    Hope that helps.

  3. Great explanation I'm student who had never taken a class on weather, I had an assignment where the book couldn't explain this subject very good . But once I read it my mind just cleared.

    Thank you


  4. nice explanation , well done!

  5. This article was awesome. We just got a tornado in Jersey yesterday and most people I know would love to have a better understanding of how they work. I've linked this article to my site As you probably are aware we don't really get tornadoes or even funnel clouds so the fact that we had multiple funnel clouds yesterday and one touch down is unheard of. Thanks for posting this!

  6. Excellent explanations! I am ready for the spring severe weather season now. Thanks again!

  7. This is by far the best explanation of the use of radar data for identifying storm rotation that I have seen. Thanks for the posting.

  8. Best explanation and best use of actual radar images that I've seen as examples. A lot of storms don't exhibit that classic hook The examples of how to compare the two sets of images is terrific.

  9. For live radar, get RadarScope ( It's available for iPhone/iPad, Mac and Android.