Correlation Coefficient (CC or RHO)


Correlation Coefficient is a product that measures how similarly the two radar pulses (horizontal and vertical) are behaving. What does that mean? Basically, it looks for how homogenous the returns are in each pulse. If returns “look” the same in each pulse, the CC is high. If they are vastly different, the CC is low. 

How does this apply meteorologically? If you have a very high CC, confidence is pretty high that whatever is occurring in that area, it’s probably uniform. In other words, if its rain, it’s probably all rain. If it’s snow, it’s probably all snow.  As CC decreases, that means there is likely a mixture of returns. It could be a mixture of melting snow, rain, ice or even hail. If CC gets very low, its means the returns are so diverse, it likely not even weather-related, it must be something biological.

What values should you look for?

CC can range from 0 to 1.05…the higher the number…the more uniformity. In terms of meteorological significance, the range of values you want is much narrower though.

For uniform precipitation, look for values of .95 or greater. This applies for all precipitation types.  As CC decreases to .80 to .95…there is likely some mixture going on. It could be melting snow, it could be rain-coated hail, or it could be “big drops” of rain. Big drops usually occur on the outer edge of convection and are low-rate rainfall. In hail, values closer to .80 may be more suggestive of very large hail (because it typically becomes more jagged in shape). In snow, values closer to .80 are usually more suggestive of very large, wet flakes. Dry snow will always be closer to 1.

If CC is less than .80, it’s very rarely something weather-related. It could be clutter, biological targets, chaff, or debris.

Applications (Applications will be covered in more detail later):

•   Differentiation of weather vs. non-weather returns
•   Melting layer identification
•   Precip transitions
•   Large hail detection
•   Tornadic debris detection

Specific Differential Phase (KDP)


First, don’t confuse this product with PHI, or Standard Differential Phase, which may be viewable in some radar software (mainly GR2Analyst). Without going into detail, this is a useless product for meteorological interpretation and doesn’t need to be covered.

KDP is a very hard product to explain in basic terms, so I’m going straight to the values/applications section.

What values should you look for?

KDP can range from -2 to 10. KDP is most useful in rain/heavy rain events and will be the main reason precipitation estimations will vastly improve in dual-pol. As KDP increases, rain rate increases. KDP over 1 is very heavy rain though like in ZDR extreme values may be indicative of water-coated, small hail.  Pure hail and snow have near 0 KDP. KDP will not compute for non-weather returns as it will be far too noisy, but this is a drawback as it may mistakenly not compute for legitimate weather echoes and that can have an impact on precip estimations.

Applications (Applications will be covered in more detail later):

•   Heavy rain detection
•   Will better differentiate standard rain processes from others (like tropical rainfall)

Applications

Like with the standard radar products, the best applications of dual-pol data will be when used in combination with all the various products (including the standard ones). With very few exceptions, you won’t determine something out of the data from looking solely at one specific product. Also with very few to no exceptions, dual-pol data won’t replace looking at the standard products. They will remain as essential as before.

1) Winter Weather

By far the most important application in winter weather will be identification of the melting layer. For years, standard Doppler has been able to detect the melting layer in some situations through “bright banding” or an area (usually circular) of enhanced reflectivity which is detecting the melting snow as it falls into the above freezing air. Dual polarization will greatly enhance the ability to detect this, especially through Correlation Coefficient. As explained, CC lowers to between .8 and .95 when returns are mixed. Of course, this is what is happening in the melting layer. So, CC is an excellent means to find the melting layer. An algorithm has also been developed which plots the CC-based melting layer on the map, and this will be available eventually in some radar software (most notably GRLevel3, it’s not available in Radarscope as of yet). Because the CC data is so good, this will actually be a fairly reliable algorithm though not 100% foolproof. The top of the melting layer (the point furthest out from the radar site) will be the wet bulb zero height and getting a 4-6 minute update of this information will be incredibly helpful in the future.


How can the melting layer be helpful in precipitation-typing? Obviously, if a melting layer is present, once precipitation falls below it, snow is no longer an option. It will either be rain, freezing rain, or sleet. Unfortunately, there’s no real way to differentiate between those three on dual-pol, so you’ll have to refer to environmental conditions.  In VERY rare cases, and only very near the radar site, you may see CC lower once again below the melting layer. If this is happening, this may be an indication of refreezing which is indicative of sleet. Again, this will be very rare – don’t expect to see this happening. Also keep in mind CC will lower near the radar site anyway sometimes due to clutter which makes it even harder to point out.


Remember for snow, CC is usually high and ZDR low, closer to zero. Snow near the melting layer will obviously be wetter, with lower CC and higher ZDR. In transitions from rain to snow, you will see ZDR lower quickly and CC increase, even in a wet snow transition (though the change may not be as extreme). Mixed rain/snow will keep an increased ZDR and decreased CC.  Graupel/Ice crystals will usually only be easily detected within or above the melting layer


Limitations: Again, this can’t be emphasized enough, winter precip typing by dual-pol will, at best, be only applicable at the radar beam height. While this will help tremendously, it won’t replace surface obs/reports. Additionally, with the melting layer detection, keep in mind there may sometimes be more than one melting layer. And, the melting layer may not always be perfectly circular and thus can “blend” in with the other data. This is especially true in transition situations. As always, assessing and knowing environmental conditions first will be most important.

3) Tornadic Debris Signature (TDS)

This may be the biggest operational advantage in all of dual polarization. Dual-pol products greatly enhance the ability to detect debris lofted by tornadoes, and several cases in OK have shown detection in even weaker tornadoes (as low as EF1).

Differentiating between weather and non-weather related echoes is a definite selling point of dual-pol. The signatures between such are greatly different, and since debris is non-weather related, if those signatures begin showing up when surrounded by legitimate weather, and coincident with a strong velocity signature suggestive of a tornado, there’s no doubt whatsoever that debris is present, and thus, a tornado occurring.

What to look for? High reflectivity and the strong velocity signature are how the TDS has been identified previously, and that won’t change. But, reflectivity won’t need to be nearly as high especially in weaker tornadoes. If the dual-pol signature is there w/velocity, that’s all you need. CC will go into the non-weather range, less than .80. ZDR will also lower, generally to values near or just below 0. If you have those signatures but there is not a velocity couplet, it’s NOT debris. Simple as that.


A properly identified TDS is 100% confirmation of a damaging tornado. NWS is already suggesting that WFOs relay TDS information in warning products, and they are also advising that they should be LSR’d and then surveyed because there will be no question about it.


The TDS was seen in several of the tornadoes associated with Hurricane Irene via the Morehead City, NC that was beta testing. The TDS was noted in weaker cases and in situations where reflectivity was not really enhanced at all. The fact it picked them up in low-topped, shallow, tropical supercells is very encouraging overall.


Limitations: The TDS obviously has no impact on lead-time. Once you see it, the tornado is there. These will NOT improve initial tornado warnings, but can greatly help in communicating a serious and confirmed threat to the public. TDS also has no direct correlation to tornado strength, as mentioned; it’s been detected in weak tornado cases.  It will be important not to assume too much now from debris detection now that they will be more easily seen. All that said, it will NOT be detected in every tornado. Obviously, something has to be there, and weak tornadoes will be hit and miss for sure. Finally, ability for detection will significantly decrease as you go further out from the radar site.

Nice work, Kevin!  I need to take some time to read through this more thoroughly.

But, some of the details sound like a real breakthrough in weather radar... probably the most significant since the advent of Doppler.

Hey Kevin, do you have any objections to me printing this series of posts for my own future reference?

Not at all! Feel free! Glad to know its of use to you!

Hi Kevin, I living in Turkey and ask some questions about dual pole radar.
1.We have 4 dual pole radar but 3 of them built on the mountains. Their heights approximately 1000 m. So we see the precipitiations on high levels. In my opinion we dont need dual pole radars becouse of their height but our met office want to take more. I wonder what is your oponion about this subject?

2.Why dual pole radars need zero degree height. Why dont they classify precipitation without 0 degre height. Everybody who know 0 degree height can classify precipitation with single radars.

3.Some of the case we can see melting hail high Dbz and high Zdr. High zdr also show heavy precipitation. How can we understand  it is shower or hail?