Thursday, December 22, 2011

Using dual-pol radar to look at temperature and moisture structure

This evening, the Chicago area looked to be just on the northern edge of a shield of precipitation over much of the Ohio River Valley.  In fact, the base reflectivity looked like this around 5:30 PM CST:

You'll notice a wide area of rather light reflectivity, including a large area of light returns to the east of the radar. The radar was scanning in a pattern called VCP 31, which is one of the most sensitive scanning patterns the radar can use. This scanning pattern is most often used in clear air situations (when there's not much to look at) or in the case of snow events, since snow tends to have a lower reflectivity than liquid rain.  Anyhow, one might ask what kind of precipitation we are looking at here.  If we assume that the radar beam is propagating normally, then we can guess that the area of light returns to the east of the radar runs from about 10000-15000 feet above ground.  Let's take a look at the RUC forecast sounding from that time:
I've circled that approximate elevation range on the sounding profile.  Remember here that red is the temperature line and green is the dewpoint line.  You can see in the forecast sounding that the temperature and dewpoint are the same in this 3-hour forecast sounding through that layer.  The model thought that the air was saturated there, and sure enough the radar seems to confirm that there is saturation and, consequently, ice crystal formation through that layer.

So now that we have a little bit of confidence in that model sounding, our overall confidence in the model increases.  Let's take a look at what the model forecasts the sounding to be around an hour later.
We see a slight difference in the sounding--the air is no longer saturated in that layer.  So, the model forecasts the air to be drying out aloft.  This would probably lead to water evaporating off any ice crystals that form (or those ice crystals just falling out).  Regardless, if this were to happen, we'd expect to see a reduction in the reflectivity in that area.  Here's the reflectivity image from about 50 minutes after that first image:

And then again an hour later:

We can see that just as the model was predicting, that layer seemed to dry out quite considerably.  By two hours later, there is very little precipitation left in that area to the east of the radar.  If you were to watch the full loop, you'd see it actually eroded from the south to the north, so the ice crystals didn't simply advect away.  They either sublimated or fell out.  Regardless, this seems to confirm the model's forecast of drying in that layer.

But we can do better than that with our radars now.  The Chicago radar is one of the few in the nation that has been upgraded to have dual-polarization.  Now, I haven't talked a whole lot about what dual-polarization means on this blog (though I plan to in the near future), but there are some simple things that I can go over right now to illustrate the new capabilities this adds to the radar.

With dual-polarization, the radar scans with two different beams of two different orientations--one is sensitive to returns in the horizontal direction and another is sensitive to returns in the vertical direction.  This allows us to get a good idea of the relative shape of what the radar is looking at.  We can most directly see this in a product known as differential reflectivity or ZDR.  This simply takes the magnitude of the vertical return and subtracts it from the magnitude of the horizontal return.

Let's imagine we're looking at small raindrops, like in drizzle.  Small raindrops tend to be nearly spherical.
Since a small raindrop is nearly spherical, its horizontal and vertical dimensions are about equal.  Therefore, when the dual-pol radar samples it with its vertical and horizontal pulses, it gets about the same return from each pulse.  When we take the difference between them, we get a value about equal to zero for the ZDR.

However, let's think about big, dendritic snowflakes.  You know, the six-ponted, fancy kind you imagine snowflakes to be.  It turns out when those flakes fall, they tend to fall with their big, flat side facing the ground.  Now let's think about with this means in terms of differential reflectivity (ZDR).
As I've crudely attempted to illustrate here, in the case of these crystals, the horizontal dimension tends to be much greater than the vertical dimension.  This means that when the radar beams hit the snowflakes, the horizontal return will be much greater than the vertical return.  When we take the difference to compute ZDR, it's the horizontal return minus the vertical return.  Since the horizontal return is greater than the vertical return, we would get a positive number for ZDR.  Positive ZDR returns indicate something that has a larger horizontal dimension than vertical dimension.

Now, let's pretend we knew nothing about the temperature structure of the atmosphere at the time of the first radar image I showed above. We saw that area of very light returns to the east of the radar and knew it could either be one of two things--either light rain/drizzle or snow.  Reflectivity doesn't tell us, but ZDR can give us a clue.  Here's the ZDR image for that time:

Notice in that area to the east of the radar, there's a lot of green and yellow.  Going over to our color scale on the left side of the image, we see that that points to ZDR values of 1-2 dB--positive values!  We don't see any blue, which we would expect to see at least some of if the ZDR values were near zero.  Because we're seeing fairly robust positive values for ZDR and knowing what we know about how ZDR relates to hydrometeor shape, we could assume that what we're seeing here is all snow.  It turns out that dendritic crystals like to form in temperature between -12 to -18 degrees Celsius.  Returning to our model sounding, we can see that the temperatures in the saturated layer do indeed overlap with that dendritic crystal temperature range (that range is highlighted where the red temperature trace in the image below becomes yellow).  So, the radar's indication of hydrometeor type matches the model sounding's guess.
However, actually one of the first things we should have done to use the radar to look at the type of hydrometeors was to look for a melting layer.  Looking at the model sounding above, we would guess that we wouldn't expect to see a melting layer, as pretty much the entire sounding is below freezing (except for maybe right near the surface).  We have another dual-pol product that can help us see melting layers, and that is called correlation coefficient.  Without going into the fine details, basically correlation coefficient tells us whether the hydrometeors that the radar is seeing in a given area of space are all the same type (like, all rain or all snow) or if there is a mixture (like melting snow and ice).  If the radar is seeing consistently the same hydrometeor type, then the correlation coefficient is very high--between 0.98 and 1.00.  If the correlation coefficient drops below that, we're probably seeing a mixture of liquid and ice.

So, let's look at the correlation coefficient map for this time.
In this image, anywhere there is white indicates where the correlation coefficient is greater than 0.99.  You can see that pretty much that entire easterly area of radar return is all highly correlated, meaning we're looking at all the same type-in this case, all snow.  But, remember that as the radar beam travels away from the radar it increases in elevation.  We know that--we know those returns to the east are over 10,000 feet above the ground.  What's interesting in the image above, however, is that when you get close to the radar (which also means where the radar beam is still close to the ground), we see many areas where the correlation coefficient drops below 0.98 (areas that are purple and red).  This slight drop in correlation is enough to start implying that what we're seeing isn't all the same in those areas.  In fact, based on that temperature sounding above, I'd guess that the radar is seeing these snow crystals starting to melt right before they reach the ground.  Looking at surface temperatures from the observations plotted underneath the radar images, the temperatures at the surface are in the 32-36 degree Fahrenheit range--just above freezing. So, it looks like the radar is telling us that much of this snow is starting to melt right before it hits the ground.  However, since this melting region goes all the way into the radar (and therefore all the way down to the ground), it doesn't look like the snow is completely melting before it hits the ground.  Therefore, I'd expect to be seeing big, wet snowflakes making a light, but slushy mess on the south side of Chicago.

So there we see a simple case where the addition of dual-pol products helps us to glean a lot more from a radar image than just what reflectivity alone may tell us.  The radar can be used to qualitatively verify model forecast soundings to be sure that what the model is telling us is what's actually happening.  We can also use the model soundings to help inform us as to what we're seeing on radar, letting us make better short-term forecasts of precipitation type.

No comments:

Post a Comment