The NWS is installing a dual-polarization upgrade at its WSR-88D Doppler radar sites. What new capabilities will this provide? I've heard that it will help to distinguish among different precipitation types. Will it also help in tornado detection?
The dual-polarization upgrade to the 88D radars is underway and will continue until May 2013. Before describing the new capabilities, it is useful to list those things that will not change. The 88D radars will still acquire data on reflectivity, radial velocity, and spectrum width.
Reflectivity is a measure of the energy scattered back toward the antenna from targets. When a pulse of energy emitted by the antenna hits a target (raindrops, snow, hail, drizzle, birds, insects, buildings, or even airborne debris), a small part of that energy “bounces” back toward the antenna. Radial velocity is the component of the wind measured along the direction in which the antenna is pointing; it is the component of wind toward or away from the radar. Spectrum width is a measure of scatter in the values of radial velocity obtained within a volume of atmosphere sampled by the radar.
The 10-cm wavelength of the radar, the 1.2° effective beam width, the scanning strategies (for example, the rotation rate of the antenna on its pedestal, and the selection of antenna tilt angles), and the algorithms for diagnosing features in the radar displays (for example, rotation signatures) will also remain unchanged.
Whereas the standard 88D radar emits pulses of energy that are horizontally polarized, the dual polarization upgrade will make it possible to emit a second kind of pulse—a vertically polarized pulse. Figure 1 illustrates dual polarization. A very simple example will illustrate just one of many new capabilities the upgrade will provide.
Caption: Figure 1. Horizontally (blue) and vertically (red) polarized pulses emitted by a dual-polarization radar, lower left. The pulses hit targets (raindrops, snow crystals, and hail in this example) within the atmospheric volume illuminated by the pulse.
Imagine a cloud of raindrops of uniform size. The reflectivity of a raindrop increases with the sixth power of its diameter. Thus a raindrop 1.0 mm in diameter will scatter 64 times more energy back toward the radar than a raindrop only 0.5 mm in diameter. A cloud of 1.0-mm drops will therefore have a much higher reflectivity than a cloud with the same number of 0.5-mm drops. With only horizontal polarization, the radar responds to the horizontal dimension of the drop, that is, to the width of the drop but not its height. For small droplets, which are spherical, this is fine because the cross-section of the drop is the same regardless of the direction of view. However, when raindrops fall, larger drops assume the shape of an oblate spheroid (a flattened sphere), which is wider than it is high. With dual polarization radars, the returned power from a large drop is greater from the horizontally polarized pulse than from the vertically polarized pulse because the former is detecting the width while the latter is detecting the height of the drop. In fact, using the difference in reflectivity factors, ZH − ZV, expressed in decibels (dB_a logarithmic scale), it is possible to infer the size of the drop because larger drops are flatter, and ZH − ZV increases with “flatness.”
I'll mention other capabilities occasioned by the dual-polarization upgrade, but I'll have to omit details because of space limitations.
Three new base products will appear with the dual-polarization upgrade: Differential Reflectivity (ZDR), Correlation Coefficient (CC), and Specific Differential Phase (KDP).
Differential reflectivity is just the difference between the reflectivity factor from horizontally polarized pulses and that from vertically polarized pulses: ZDR = ZH − ZV. The range of values is from −7.9 to +7.9 dB.
Figure 2 shows typical ZDR values for various targets. As hinted above, ZDR for nearly spherical targets is close to zero, but values increase with drop size as the drops become more oblate. Very large hail sometimes produces negative ZDR values. Small hail, not water-coated, tends to tumble randomly and thus, to the radar, appears to be spherical (ZDR near zero). If hail becomes coated with liquid, it appears to the radar as a giant raindrop, and ZDR increases. Similarly, if rain and hail are in the same sample volume, the ZDR increases. Graupel, often the embryo for hail, is a low-density white pellet that is filled with air bubbles. Dry graupel often has a conical shape and may be taller than it is wide, and so yields negative ZDR values. If graupel becomes water-coated, its ZDR value increases. Dry snow generally has ZDR values close to zero, but if it begins to melt, the ZDR increases markedly. Considering ice crystals, if many of them are interlocked in a flake, the ZDR is close to zero. Individual crystals assume many shapes, none even remotely spherical. Needles, columns, and plates tend to fall with their long axes horizontal, and so their ZDR values can be large. Ground clutter refers to objects on the ground seen by the radar, either because they are tall (buildings) or because the temperature stratification in the lower atmosphere bends the radar beam (anomalous propagation–AP) so that the radar sees more of the ground than usual. Biological targets include birds, insects, and bats. Chaff refers to thin strips or wires of aluminum sometimes dropped near or inside of clouds to track air flows and sometimes for military purposes. Debris refers to objects lofted from the ground by high winds, including tornadoes.
Caption: Figure 2. Typical values of differential reflectivity for various targets as seen by WSR-88D radar.
The Correlation Coefficient expresses the similarity (or dissimilarity) in the behavior of the horizontally and vertically polarized pulses within the same pulse volume. CC is a dimensionless number. The range in displayed values is 0.2 to 1.05. Mathematical correlation can never exceed 1.0. CC values greater than 1.0 arise when the signal-to-noise ratio is low and cannot be trusted.
Figure 3 shows typical CC values for various targets. Stratiform rain or snow generally yield CC values from 0.97 to 1.00. Hydrometeors with complex shapes or a mixture of hydrometeors will have CC values from 0.85 to 0.95. Any form of water-coated ice has rather low CC values. For small, dry hail and dry graupel, CC is close to 1.0. The melting layer, where falling snow begins to melt and become water-coated, produces CC values between 0.65 and 0.95. Ground clutter, biological targets, chaff, and wind-blown debris have CC values at the lower end of the scale, typically less than 0.7.
Caption: Figure 3. Typical values of the correlation coefficient for various targets as measured by a WSR-88D radar with the dual-polarization upgrade.
The definition of Specific Differential Phase is more complicated than that for ZDR and CC. The radar pulses are emitted as oscillating electromagnetic waves. Associated with these waves are amplitude and phase. The latter is a measure of position within a single wave, with 360° marking the distance between successive peaks in the train of waves. When the polarized waves pass through precipitation, they slow down very slightly but by a measurable amount. The slowing is measured by a shift in the phase of the waves. The peak in the waves arrives slightly later after passing through precipitation than it would if the air had been cloud-free. Horizontally polarized waves tend to slow down slightly more than the vertically polarized waves. The differential phase, ØDP, is defined by ØDP = ØH – ØV. Because the phase shift of the horizontally polarized wave ØH is usually greater than that for the vertically polarized wave ØV after passing through precipitation, ØDP is usually positive. Note that ØDP does not change as the pulse passes through clear air, nor does it reset after passing through precipitation; it is a nondecreasing function of range.
The Specific Differential Phase is defined as the change in differential phase measured along a single radial:
where r1 and r2 are ranges along the same radial, with r2 > r1. The “two” in the denominator allows for the round trip of the pulse, out to the target and back to the antenna. With this definition, it is easy to see that KDP is zero in clear air and nonzero in precipitation because that is where phase shifting occurs. KDP is reported in units of degrees per kilometer. The range of values is −2 to +10 deg/km.
Two other important products are derived from six data types: ZH, ZDV, CC, KDP, and the texture (ranging from smooth to very noisy) of the ZH and ØDPfields. These are the Hydrometeor Classification (HC) product and the Melting Layer (ML) product.
The Hydrometeor Classification product uses the six data types to determine the target most likely responsible for reflectivity in a given area. The target types are described and color-coded in Figure 5. Most are familiar and have already been discussed. Large drops may arise from melted hail, have diameters measured in millimeters, and have large ZDR values. UK means an unknown type, which the algorithm failed to classify. RF means range folding. It is not a valid classification. Dual-polarization radar cannot distinguish between rain and freezing rain. It is important to remember that the classification scheme applies at the location of the radar target, which is defined by the range, azimuth, and elevation angle. If the algorithm diagnoses snow at 8,000 feet, the snow may melt long before it reaches the surface and arrives in your backyard as rain.
Caption: Figure 4. KDP values typical for various targets detected by WSR-88D radars equipped with the dual polarization upgrade.
Caption: Figure 5. The 12 classifications produced by the Hydrometeor Classification algorithm that runs on dual-polarization data collected by 88D radars.
Figure 6 is an example of the Hydrometeor Classification product. On the left is a standard reflectivity image acquired at 0.5° elevation angle. The radar is located just off the lower right-hand corner of the image. Note the areas of light rain (green) and several strong convective storms with reflectivity values exceeding 60 dBZ. On the right is the corresponding hydrometeor classification image. The light blue-green color at lower right indicates light rain, but there are small patches of big drops mixed in. Farther from the radar, even at a half-degree tilt angle, the beam penetrates the freezing level. The large area of blue indicates dry snow, but this area is punctuated with smaller areas of graupel and hail, precisely where the reflectivity is greatest.
Caption: Figure 6. Left: A reflectivity image with the antenna tilted 0.5° above the horizontal. Color-coded values of reflectivity (dBZ) are at upper left. The radar is located just off the lower right-hand corner of the image. The greater the distance from the radar, the higher the radar beam in the atmosphere. Right: The corresponding Hydrometeor Classification product. The most likely target type is color-coded at upper left. A more complete color bar is shown in Figure 5. Note that hail and graupel signatures correspond to areas of high reflectivity.
The Melting Layer product is generated in areas that have stratiform (gentle) precipitation. As snow falls below the 0°C altitude, it begins to melt. After further descent, measured in hundreds of meters, the snow has completely melted into raindrops. In between these altitudes, the snow is partially melted and has a unique signature, easily detected by dual polarization radars. Wet snow is indicated when the reflectivity Z is between 30 and 47 dB, ZDR is between 0.8 and 2.2 dB, and CC is greater than 0.85. From the range and elevation angle, the altitude of radar volumes containing melting (wet) snow is easily computed. Because the reflectivity of melting snow is greater than that of dry snow or raindrops from melted snow, the melting layer stands out as a ring, concentric about the radar, of higher reflectivity (the bright band) and lower correlation. The melting layer indicates the snow level in mountainous areas, and the lower limit indicates icing potential in clouds.
Radar estimates of precipitation are traditionally based on a relationship between reflectivity and rainfall rate. The instantaneous rainfall rate is integrated over time to give hourly or storm-total estimates of precipitation. With the additional information on precipitation type from dual-polarization radar and independent temperature profile data from models (which helps a forecaster decide what type of precipitation is reaching the ground), it is now possible to refine estimates of precipitation significantly. For example, if an area of 60 dBZ reflectivity is known to be mostly rain instead of all hail, the estimated precipitation rate will be higher, possibly prompting a flash flood warning. As another example, the bright band of higher reflectivity will not result in a higher estimate of rainfall rate at the ground.
Here are a few more benefits of dual-polarization radar not yet mentioned:
Improved detection of non-meteorological targets (birds, bats, bugs, wind farms, ground clutter)
Improved hail detection for severe thunderstorm warnings
Positive identification of debris lofted by tornadic circulations. This works best for tornadoes close to the radar, when the beam height is close to the ground. This detection will not necessarily improve the lead time for tornado warnings, but it does confirm a tornado on the ground independent of an eyewitness.
KDP significantly improves the detection of heavy rain
More details about dual polarization radar and excellent training materials for meteorologists and non-meteorologists alike are available at http://www.wdtb.noaa.gov/courses/dualpol/outreach/#mets. I thank Paul Schlatter and Clark Payne for help with the figures. Clark also checked the text for accuracy.
Weatherwise Contributing Editor THOMAS SCHLATTER is a retired meteorologist and volunteer at NOAA's Earth System Research Laboratory in Boulder, Colorado. Submit queries to the author at firstname.lastname@example.org, or by mail in care of Weatherwise, Taylor & Francis, 325 Chestnut St., Suite 800, Philadelphia, PA 19106