One of the most frequent questions I'm asked these days relates to how Doppler radar operates and whether Doppler radar can automatically detect or predict the formation of tornadoes. What I have done to try to answer that question is to write both a brief summary and a more complete explanation. In each section of the brief summary below, you will find a link to a more thorough explanation. We start out with a very basic outline of how Doppler weather radar operates, then talk about how computers are used to analyze and display data coming from the doppler radar, and finally we'll look at the question of how computers (more precisely, computer software) are used to detect severe storms and tornadoes.

Basics of Weather Radar: The type of radar used in detecting weather such as precipitation and storms has been around for more than fifty years. A radar is really a very precise transmitter and receiver capable of sending and receiving radio waves at a highly precise frequency, mounted on a pedestal so that the sending and receiving antenna can be rotated through 360 degrees of azimuth (i.e. through all points of the compass). <more>

Detecting and Locating Storms: When a radio wave emitted by the radar strikes a "scatterer" (a raindrop or anything else capable of   reflecting a portion of the radar wave) some of the energy is reflected directly back toward the radar antenna. Weather radars operate on frequencies that optimize the reflection of radio waves from precipitation (what meteorologists call "hydrometeors"). The radio waves travel at the speed of light, so even at a range of several hundred miles from the radar, the roundtrip time from transmission to reception of the return signal takes only milliseconds. Since we know the speed at which the radar wave was traveling, simple mathematics can tell us how far away from the radar the scatterer was when the radar wave bounced off of it, and because the radar "knows" exactly where it was pointed when it transmitted and received the signal, we can locate that point exactly (i.e. we know its direction and distance from the radar unit). <more>

Detecting Wind Data with Radar: Remember that the radar unit transmits the radar beam at a very precise frequency. The emitted radar wave retains that frequency as it travels through space away from the radar. If the wave strikes a scatterer that is neither moving toward nor away from the radar (i.e. it has zero velocity relative to the radar), then the portion of the wave reflected to the radar unit will retain the precise frequency at which it was originally transmitted. But (drum roll!) ... if the scatterer has any velocity that is either toward or away from the radar unit, the frequency of the reflected energy will be slightly shifted (higher frequency if the scatterer has velocity toward, lower frequency if the scatterer has velocity away). This is what is called the "Doppler shift" and is the basis for measuring winds within weather systems remotely using radar. You can experience this phenomenon while waiting for a railroad engine to pass a crossing. Notice that the pitch of the engine's whistle changes as the engine approaches and passes your location; the pitch is higher as the train approaches and lower after it has passed your location. <more>

Other Important Issues: (1) As the radar beam travels away from the  radar site, it becomes less compact (i.e. it "spreads"). The result of this beam-spreading is decreasing resolution per unit of space as the distance from the radar site increases. (2) A complex scientific  principle effectively limits the range at which velocity data can be measured accurately as the radar transmitter cannot "fire" again until the return pulse is received or a specified waiting period has expired. <more>

Radar Detection of Severe Storms: Studies conducted by researchers have demonstrated time and again that severe thunderstorms have certain reflectivity features that can be used to help identify them as being severe. For this type of data, you don't need the Doppler aspect of weather radar. <more>

Non-automated Radar Products: Any Doppler radar produces several "raw" products that can be used separately or together to help identify storms that have become severe, or may be about to. In addition to base reflectivity, the radars produce base velocity and spectrum width products. <more>

Automated Radar Products:  The National Weather Service WSR-88D (often called "nexrad") radar units automatically generate many products designed to aid the meteorologist in a preliminary identification of severe thunderstorms. These automated WSR-88D products are constantly being refined by the researchers at the Nexrad Support Facility, where efforts continue to improve the accuracy of these products, and to develop new ones. (There is also software marketed by private companies that ingests data from the WSR-88D units and produces output that may aid in identifying features, including circulations, within storms.) With respect to these automated products, the Nexrad Support Facility makes the following recommendation: Warning forecasters should integrate information from a variety of sources, including Doppler radar data from multiple radars (when available), algorithm (automated) guidance, reliable spotter reports, storm history, remote sensing instruments (surface, satellite and lightning observations), and a good understanding of the Near-Storm Environment (NSE). <more>

Monitoring the Near-Storm Environment:     Research in the past decade has demonstrated the importance of local variations in atmospheric conditions in determining which storms may become severe, and which of those may produce tornadoes. Among other things, we have learned that relatively small-scale boundaries (such as may occur after earlier storms have passed through an area) can aid in the development of local conditions that enhance the tornado threat if a severe storm interacts with the boundary. In other cases, local conditions may preclude the development of tornadoes despite very favorable conditions aloft. <more>

Detecting and Locating Storms:
    When the radio wave emitted by the radar (remember, in a relatively narrow beam) strikes a "scatterer" some of the energy in the wave is reflected directly back toward the radar antenna. That energy also travels at the speed of light, so even at a range of hundreds of miles from the radar unit, the roundtrip time from transmission to reception of the return signal takes only milliseconds. Since we know the speed at which the radar wave was traveling, simple mathematics can tell us how far away from the radar the scatterer was when the radar wave bounced off of it, and because the radar "knows" exactly where it was pointed when it transmitted and received the signal, we can locate that point exactly (i.e. we know its direction and distance from the radar unit). Of course, we would also like to know how high above the ground that scatterer was when the radar wave bounced off of it, and we can calculate that by determining the extent to which the radar antenna was pointed above the horizon (i.e. its elevation or tilt). Although there are various scatterers that may cause radar waves to bounce back to the radar unit, the kind that weather radar is designed to detect are called "hydrometeors" (a broad class that includes rain drops. snow flakes, hail stones, and other forms of precipitation).

    This latter calculation, however, involves slightly more complex mathematics, because the earth is round, not flat. So, the radar beam travels away from (and back to) the radar unit in a straight line, but as distance from the radar unit increases, the earth "falls away" from the beam because of the earth's curvature. Thus, the radar beam, traveling in a straight line at 0.1 degree elevation when transmitted, will be at an elevation of 1,700 feet above ground when it is 50 miles from the radar. If the elevation of the antenna is 0.5 degrees, then the same radar beam will be at 3,100 feet when it is 50 miles from the radar. Virtually all weather radars are designed to operate in various tilts from near 0.0 degrees elevation to 20 degrees elevation or more.

    Thus, using the known properties of the radar unit and the beam it projects, along with mathematics, it is possible to locate the scatterer from which the radar wave bounced in three dimensions (direction from radar, distance from radar, and elevation above the nominal elevation of the radar unit). This process happens repeatedly, with almost incredible frequency, as the radar antenna rotates through the 360 degrees of the compass, over and over. To handle all of the data, modern weather radar units feed the data to a computer that accomplishes all of the computations and supplies a variety of displays for the radar operator to view. A radar need not be operated in "Doppler" mode to produce this kind of data.

Detecting Wind Data with Radar:
   Remember that the radar unit transmits the radar beam at a very precise frequency. The emitted radar wave retains that frequency as it travels through space away from the radar. If the wave strikes a scatterer that is neither moving toward nor away from the radar (i.e. it has zero velocity relative to the radar), then the portion of the wave reflected to the radar unit will also retain the precise frequency at which it was originally emitted. But (drum roll!) ... if the scatterer has any velocity that is either toward or away from the radar unit, the frequency of the reflected energy will be slightly shifted (higher frequency if the scatterer has velocity toward, lower frequency if the scatterer has velocity away). This is what is called the "Doppler shift" and is the basis for measuring winds within weather systems remotely using radar.

    Of course, the computer that is assembling the data from the radar unit has no way of knowing whether the scatterer was traveling directly toward the radar (or directly away from the radar) but it can detect (through the frequency shift) whatever component of motion was toward or away from the radar. To complicate matters, a scatterer that is truly stationary would have a Doppler velocity of zero, but so would a scatterer that was moving at 50mph but perpendicular to the radar beam (more precisely, tangentially to the radar). This seems to present a huge problem but in reality, since the radar (and its computer) can present a display of not just one scatterer but of all* of those detected within range of the radar, it is possible to infer storm-scale (and sometimes, smaller) wind flow and circulations.

    [*Actually, because the radar is constantly sending and receiving hundreds of pulses each second, it is impossible to display individual scatterers or their motion. Instead, the computer drops the digital data into a bin that contains the data for many, many individual scatterers and the bin average is displayed.]

    All Doppler radars actually produce just three products: base reflectivity, base velocity, and spectrum width. Let's examine each of these briefly. Base reflectivity is the basic radar output of the location of scatterers in space, the kind of radar output that can be produced without operating in Doppler mode. It is the kind of output with which you are probably most familiar from seeing it on television or on the Internet. Base velocity is the basic output of the velocity with respect to the radar unit of  hydrometeors (and other scatterers, such as insects, dust, and other atmospheric impurities). Both base reflectivity and base velocity output has been "binned" and the displayed data is a bin-average for each location on the radar display. Spectrum width, the third "raw" output from Doppler radar units, is a measure of the variation in data values within each bin into which base velocity data was dumped (i.e. it is a numeric expression of the amount of variation in the individual values that were averaged to produce the bin-average base velocities). Spectrum width has some very specialized uses but is certainly the least used of the three basic outputs from Doppler radars.

Other Important Issues:
    I've already mentioned that the radar beam lies at increasing height above the ground as distance from the radar increases. The significance of this physical principle cannot be overstated. From the standpoint of detecting severe storm features, the meteorologist frequently wants to look as close to the ground as possible in the vicinity of the storm. Because the radar beam lies higher and higher above the ground as distance from the radar increases, the ability to detect low-level features in storms is reduced and eventually eliminated with distance from the radar site.

    Two other scientific issues also impact the effectiveness radar and must be mentioned:

    (1) As the radar beam travels away from the  radar site, it becomes less compact (i.e. it "spreads"). The result of this beam-spreading results in decreasing resolution per unit of space with distance from the radar site. This impacts both reflectivity and velocity products, but more acutely the latter. Detection of storm-scale and lesser circulations within storms depends on high resolution of velocity data. This is best accomplished by narrow beam widths, so that hydrometeor motion, especially changes in speed and direction over short distances, is highly resolved. At greater range from the radar site, the circulations may fall entirely within the beam width, not across it, and in such cases the circulations may go undetected. (Spectrum width can provide a backup in such situations in some cases.)

    (2) The other issue involves a limitation inherent in the Doppler radar process. After the radar transmitter has pulsed and sent out a radar beam, it must wait for the return of any reflected energy so that the calculation of distance to the reflecting source can be calculated. Of course, if the pulse of energy doesn't strike a reflector, it won't bounce back and after waiting a specified interval, the radar fires again. (Remember, all of this happens in very rapid sequence, hundreds of times per second.) This physical principle effectively limits the range at which velocity data can be measured accurately. For the WSR-88D (NWS NEXRAD) radars, the limit occurs at around 125 miles from the radar. For other radars, it may be as close as 75 miles.

    To summarize, there are some important operational issues for all weather radars and some additional ones for those that detect wind velocity data using the Doppler principle.

Radar Detection of Severe Storms:
    An experienced radar operator can frequently deduce a lot of information about a thunderstorm without ever looking at velocity products (i.e. those derived using the Doppler principle). Before the advent of Doppler radar, meteorologists depended on what they could see of storm structure in the reflectivity data that is still the data most often displayed by television stations on-air and on the Internet. Over several decades, studies conducted by researchers demonstrated time and again that severe thunderstorms had certain reflectivity features that could be used to help identify them. Included among those features is the "hook" echo, which was associated early-on with a thunderstorm capable of producing a tornado. Other important features may include an echo overhang with increasing height, the "flying-eagle" signature, and the BWER (bounded weak echo region). Examples of some of these are shown below.

   I previously mentioned that one of the base products produced by the NWS's NEXRAD radars was base velocity. This product displays the velocity relative to the radar of any hydrometeors (and other scatterers). The product is generated for several elevations (more often called "tilts") beginning at 0.5 degrees above the local horizon. The display generated uses colors to depict the magnitude of motion toward (cool colors) and away (warm colors) relative to the radar. If a storm is not moving at all relative to the radar (a rarity) or if the storm motion is perpendicular to the radar beam (i.e. across the beam), then the base velocity product can be useful in examining storm-scale circulations. Otherwise, storm motion "contaminates" the base velocity product and meteorologists must resort to the storm-relative motion velocity product, which is generated by applying an average storm motion to the base velocity data before creating the display seen by the meteorologist. The result is to "remove" that part of the velocity data that may be attributed to the overall motion of the storm, thereby revealing the circulations within the storm.

    You will notice that the elapsed time between the two products displayed above is 55 minutes, and the elapsed time between the "hook echo" in the reflectivity product and the operator-defined TVS is over an hour. This storm (which, by the way, produced a tornado that produced F2 damage on the outskirts of Thornton in Limestone County) gave repeated indications that it was capable of producing a tornado. Between noon and 1:16pm, the storm was repeatedly interrogated utilizing various non-automated and automated products from the WSR-88D NWS radars, as well as with Doppler10 Live radar.

    When the WSR-88D (the NWS NEXRAD radar) was being developed, studies suggested that identifying cyclonic circulations (what we now call mesocyclones) in thunderstorms would yield an important tool for anticipating which thunderstorms were or might become severe. Since the deployment of the WSR-88D network, what has surprised meteorologists is how frequently thunderstorms develop Doppler indications of a mesocyclone for at least a brief part of the storms' lifetimes. Thus, meteorologists now look for more qualitative information about the indicated mesocyclone, and one approach to this problem is to look at how strong the circulation is, how deep it is, whether it is broad or compact, etc. Many of these questions can best be answered by a trained meteorologist who examines the non-automated products from a Doppler radar, whether that be a WSR-88D or some other Doppler weather radar, such as Doppler10Live at KWTX-TV.

Automated Radar Products:
    The WSR-88D radar units automatically generate many products designed to aid the meteorologist in a preliminary identification of severe thunderstorms. These products are generated every five to ten minutes (depending upon the type of scanning pattern the radar is in) and are available not only to the National Weather Service but to the public sector, including the forecasters in the Doppler10 Forecast Center. We receive these products automatically via a satellite feed as soon as they are produced by the WSR-88D units that cover our area. And we do make use of these products as a part of the continuous process of analysis and nowcasting when severe storms threaten our viewing area. The automated products are truly too numerous (and technical) to mention, but to give you an idea of them, the products help identify storms that have mesocyclones, storms (if any) that have indications of a tornado vortex signature (a smaller cyclonic circulation that may indicate the presence of a tornado),  storms that contain hail and/or large hail, the height of the tops of the storms, the rainfall produced over the are in the preceding one hour, three hours or since precipitation began, etc.

    These automated WSR-88D products are constantly being refined by the researchers at the Nexrad Support Facility, where efforts continue to improve the accuracy of these products, and to develop new ones. Scientists from both the U.S. and the international meteorological community are invited and encouraged to review the manner in which the software is designed and to suggest revisions that would lead to improved products. There is also software marketed by private companies that will ingest data from the WSR-88D units and produce output that may aid in identifying circulations within storms. The privately marketed software has not been subjected to professional scientific review in the same manner that the WSR-88D software has been critiqued and reviewed.

    Without getting into an argument about the relative merits of the WSR-88D automated products versus those of the private companies, let's take a look at what the National Weather Service recommends with respect to the use of automated products for identifying severe storms and tornadoes. The recommendations are a product of published papers on identification of severe storms during storm intercept projects (such as VORTEX), applied research utilizing case studies from prior severe storm events, and observations from NWS forecast offices during actual warning operations. Chief among these recommendations is the admonition that "whenever possible, one should not rely solely on radar data for making warning decisions. *** Warning forecasters should integrate information from a variety of sources, including Doppler radar data from multiple radars (when available), algorithm (automated) guidance, reliable spotter reports, storm history, remote sensing instruments (surface, satellite and lightning observations), and a good understanding of the Near-Storm Environment (NSE). This requires acute situational awareness ...." I think the most important thread inherent in this admonition is that the warning-decision process must involve humans making use of the tools, including both raw Doppler output and the automated products.

    Remember in an earlier section I mentioned that one of the storm features that can be detected by Doppler radar is the mesocyclone, sometimes referred to as the rotating updraft that supplies the storm with buoyant air that is truly the fuel for the storm. I also mentioned that experience with the WSR-88D network has yielded a surprise for meteorologists because we now know that many thunderstorms exhibit Doppler-indications of a mesocyclone for at least a brief part of the storms' lifecycles. That means that the mere occurrence of brief periods of rotation are less helpful in identifying storms that are or may become severe. The branch of the weather service charged with researching the utility of NEXRAD data and training NWS meteorologists in the use of that data says that "it is extremely important to understand that the presence of a tornado vortex signature (TVS) or a mesocyclone does not imply the presence of a tornado (they are tornadic only 20-40% of the time, according to the latest statistics)." What is more important is the persistence of a mesocyclone for tens of minutes, the strength and dimensions of the mesocyclone and the trends of those parameters, and the Near-Storm Environment (NSE), which can in many cases be the single most important factor in whether a storm goes on the produce a tornado.

Monitoring the Near-Storm Environment:
    Research in the past decade has demonstrated the importance of local variations in atmospheric conditions in determining which storms may become severe, and which of those may produce tornadoes. Among other things, we have learned that relatively small-scale boundaries (such as may occur after earlier storms have passed through an area) can aid in the development of local conditions that enhance the tornado threat if a severe storm interacts with the boundary. In other cases, local conditions may preclude the development of tornadoes despite very favorable conditions aloft.

    If monitoring the NSE is so important, perhaps we ought to briefly discuss how it is accomplished. We are fortunate in Central Texas to have a reasonable number of automated weather observation stations that report to conditions at least once an hour via the national weather report system. These stations (called ASOS and AWOS) are located at airports and report (at a minimum) cloud information, temperature and dewpoint, wind direction and velocity, and air pressure. The AWOS stations send a report every 20 minutes on the national circuit; the ASOS stations report hourly and may generate special reports when significant changes occur between regular reports. In this area, stations are located at the following airports: Waco, McGregor, Temple, Killeen, Brownwood, Burnet, Georgetown, Corsicana, College Station. There are also several AWOS stations that exist but are not yet connected to the national circuit: TSTC Airport, San Saba, Brady.

    You might think that with these stations, monitoring the weather conditions near storms would be easy. Unfortunately, these stations are too sparsely located to afford an accurate picture of the NSE in many instances. That's because severe thunderstorms can respond to variations in surface weather conditions where the variation is entirely between the airport stations. To address this problem, the State of Oklahoma has placed at least one weather station in every county! This is called to Oklahoma Mesonet and it has proven instrumental in monitoring the NSE in many storm cases in that state. Of course, in Texas we have a much larger geographic area than Oklahoma. Texas Tech University has a pilot mesoscale network project in the counties around Lubbock, which hopefully will someday be expanded to cover more of the state. In the meantime, we are fortunate at KWTX to have the AWS Schoolwatch stations to help us monitor conditions with greater resolution. In the last few years, we have seen the Schoolwatch network expand, and we hope it will continue to do so.