1022JOURNAL OF APPLIED METEOROLOGYVOLUME 16Real-Time Wind Measurement in Extratropical Cyclones by Means of Doppler RadarH. .W. BAYNTON, R. J. SERAFIN, C. L. FRUSH AND G. R. GRAY National Center for Atmospheric Research,` Boulder, Colo. 80307P. V. HOBBS, R. A. HOUZE, JR., AND J. D. LOCATELLIDepartntent of Atmospheric Sciences: University of Washington, Seattle 98195(Manuscript received 4 February 1977, in revised form 6 August 1977)ABSTRACTColor displays of the velocities of precipitation particles detected with a C-band Doppler radar in wide-spread cyclonic storms provide a variety of real-time information on the atmospheric wind field.Vertical profiles of wind speed and direction indicated by the real-time color displays agree well withrawinsonde measurements. Veering winds (or warm advection) produce a striking S-shaped pattern onthe color display and backing winds (or cold advection) produce a backward S. A maximum in the verticalprofile of wind speed is indicated by a pair of concentric colored rings, one upwind and one downwind ofthe radar. Vertically sloping velocity maxima are indicated by asymmetries in the color displays, as areconfluent and difluent winds. Divergence and convergence computed from the real-time color displays areof reasonable magnitude.1. IntroductionA C-band Doppler radar developed by the Field Ob-serving Facility of the National Center for AtmosphericResearch (NCAR) has been used to obtain measure-ments in extratropical cyclonic storms in the PacificNorthwest as part of the University of Washington'sCYCLES (Cyclonic Extratropical Storms) Project. Theradar is equipped with a real-time Doppler processor(Lhermitte, 1972) and a color display (Gray et al., 1975)for showing velocities or range-normalized reflectivities.As the antenna was rotated about a vertical axis at aseries of constant elevation angles, detailed displayswere obtained of the radial component of the targetvelocity in coordinates of slant range and azimuth.Each display was protrayed in 15 colors with a velocityresolution of 2 m s-l.Lhermitte and Atlas (1961) first described how theVelocity Azimuth Display (VAD) could be used tomeasure the horizontal wind field in stratiform pre-cipitation. Later work as exemplified by Browning andWexler (1968) extended the measurement of divergenceand deformation as well. The real-time color presenta-tion described here provides essentially a multiplicityof VAD's which may be viewed simultaneously thuspermitting real-time estimates of the complete hori-zontal wind profile within the precipitating region.Techniques of pattern recognition are described forThe National Center for Atmospheric Research is sponsoredDepartment of Atmospheric Sciences Contribution No. 433.by the National Science Foundation.detecting veering and backing winds and wind velocitymaxima. Confluence and difluence are also detectableand the presence of convergence is sometimes evidentfrom inspection of the display. Procedures for comput-ing wind profiles and convergence are given and theresults of such computations are presented. Being pre-sented with this operationally useful information, theinvestigator is able to opitmize his deployment of air-craft, radars, rawinsondes, etc., in real time and con-centrate the rigorous analyses, after the fact, on themore interesting situations.2. Elements of the colored Doppler velocitypatternsPulsed Doppler radars are characterized by a maxi-mum unambiguous velocity (V,,,) given byVmsx= (PRF) (X/4), (1)where PRF is the pulse repetition frequency (Hz), andX the radar wavelength. For the NCAR C-band radar,X=0.0545 m, and for the CYCLES Project the PRFwas set at 1071 Hz. Hence Vmax= 14.60 m s-'.Velocities greater than the maximum unambiguousvelocity are "folded over" and then assigned values lessthan the maximum unambiguous velocity. In practice,with the color display, there is no difficulty in resolvingthese ambiguities provided that the targets are spatiallycontinuous. Velocities ranging between f 1 Vmax I areassigned, unambiguously, one of 15 colors. In theCYCLES Project, the width of each color band was0~~0~~~1977 BAYNTON ET AL. 1023thus simply 2 (14.60)/ 15= 1.95 m s-l. Grey is used toindicate targets with radial velcoities of Of0.97 m SKI.Ground clutter, as well as all targets viewed from adirection normal to their velocity, have zero radialvelocity and therefore appear grey, no matter how fastthey may be moving.The TV monitor or color display can be thought ofas a fine-mesh grid. Each square of the grid is assignedone of the 15 colors corresponding to the observed radialvelocity of targets, or black if there are no targets. The15 colors are also numbered sequentially, +7 through0 to - 7, where + 7 and - 7 correspond to the maximumunambiguous velocity away from and toward the radar.If the velocity toward the radar rises above 14.60 m(in magnitude), the display will shift from color -7(purple) to +7 (red). If further increases occur, thecolors will sequence through +6, +5, etc. The fifteenthcolor in this kind of sequence is grey, which wouldrepresent 29.2f0.97 m s-l in magnitude, the sign beingdetermined by the context. Clearly, if the target is aspatially continuous one, extending all the way outfrom the radar, there is no ambiguity. However, withisolated targets, there is ambiguity, since the samecolor might represent several different velocities.Also contributing to the patterns on the TV monitoris the increase in height of the radar beam with in-creasing slant range. Using the $-earth-radius conven-tion (Battan, 1973) to allow for average refraction andearth curvature, we haveR217 000z=--+R sina, (2)where z is the height (km) of the beam above the sur-face at slant range R (km) and a is the elevation angle.Finally, since strong thermal gradients characterize ex-tratropical cyclones, winds change in both direction andspeed with changing height.To illustrate how these elements are combined bythe color display into a colored picture, we consider atypical winter wind profile for Seattle. Fig. 1 showsthe distribution of radial components, with slant rangeand azimuth, for this profile. Concentric arcs on thefigure are drawn at increasing slant ranges to correspondto beam heights of 1, 2, 3, 4, 5, 6 and 7 km. Radialcomponents are then plotted along each height contour.The dashed lines show where the boundaries of the darkgreen color (number, -1) would lie. When Fig. 1 isexpanded to the full circle and radial velocity contoursare drawn at the color boundaries, Fig. 2 is obtained.Observing that winds veer with height for this profile,we note that the zero velocity band (grey on the colordisplay) bisects the picture in the shape of an S. Whenwinds back with height, the grey traces a backward S.3. Pattern recognitionQualitative information about the wind field is avail-able in real time based on pattern recognition. Someexamples are given below.a. Patterns typical of warm advection and warm frontsWhen there is warm air advection, winds veer withheight and, as we have already seen from our considera-tion of the typical winter wind profile for Seattle, the290, 119 I.270--179- -I6 2--I44TYPICAL WINTER WINDSIN SEATTLEFIG. 1. Radial components of typical winter winds at Seattle. Coordinates are azimuth(deg) and slant range corresponding to beam heights of 1-7 km. Dashed lines show boundariesof the dark green region on the radar color display.1024JOURNAL OF APPLIED METEOROLOGY VOLUME 16N -0.97 0.97SFIG. 2. Contours of radial components constructed from thetypical winter wind profile for Seattle for an elevation angle of11". Contours are drawn at intervals of 1.95 m s-l correspondingto the width of each color band in the color display.band of zero velocities then bisects the display in theshape of an S. Color displays with this feature werecommon in the CYCLES Project during the wintersof 1974-75 and 1975-76. Fig. 3, observed at 2205 PST13 January 1976, is typical. Fig. 3 also illustrates afeature of the color display whereby any of the fifteencolors-here grey-can be replaced by white.When the radar beam penetrates a warm front, anS-shaped warm advection pattern is seen on the colordisplay in the baroclinic zone below the frontal surface,with little evidence of wind shear appearing in therelatively uniform air mass above the warm frontalsurface. An example of a warm frontal pattern was ob-served at 2036 PST 26 January 1976 (Fig. 4). With theecho extending to a height of 5 km, virtually all of thewind shear occurred below 1.2 km, the apparent heightof the warm frontal surface.6. Pattern typical of occlusionsComplex thermal gradients are typically associatedwith occlusions (e.g., see Kreitzberg, 1964; Elliott andHovind, 1965 ; Kreitzberg and Brown, 1970; Browninget al., 1973; Houze et al., 1976). Cold advection in thelowest layers might be accompanied by warm advectionaloft or vice versa. During January 1976 mixed patternsof both types were observed. A striking example ofwarm advection below and cold advection aloft wasobserved on 22 January. Fig. 5 is the color displayobtained at 0705 PST. The picture was bisected by aperfect S between the surface and 1.9 km, while above1.9 km the pattern was reversed by backing winds. Fig.6 is a black and white drawing based on Fig. 5 showinga wind profile derived from the colored velocity pattern.The procedure for obtaining the wind profile is explainedin the caption of Fig. 6. Hodographs for the layers0.4-1.9 km and 1.9-4.2 km, included as inserts in Fig. 6,verify the presence of warm advection in the lower layerand cold advection in the upper layer.c. Wind maximaA maximum in the vertical profile of horizontal windspeed above the radar is indicated by a pair of con-centric colored rings, one upwind and one downwind,around a maximum radial velocity core. Such a windmaximum may be seen inside the innermost slant range(or height) circle in Fig. 3. The storm of 14 January1976, producing the display shown in Fig. 7, featuredtwo wind maxima, one at a height of about 1.8 km andanother at about 4 km. Although the color displayprovides no information as to the lateral extent of awind maximum, estimates can be made of the heightand magnitude of the maximum wind, both upwindand downwind. Height is determined from the slantrange and elevation using Eq. (2).From upwind and downwind estimates of height,wind maxima often appear to be sloping in the vertical.On 14 January 1976 the higher altitude wind maximumhas a slope of 1 : 30 or 1.8". The lower wind maximumzone is nearly horizontal. Several wind maxima, withslopes from 0.6" to 1.8" have been observed in theCYCLES Project, all sloping upward in the downwinddirection. These slopes must be accepted as real sincethe radar antenna is aligned to an accuracy of 0.1" usinga plumb line, a clinometer and the sun's C-band radia-tion as a target with known coordinates.It is tempting to assume that the observed windmaximum zones were due to continuous jets of air whichfollowed a sloping path over the radar. However, suchan interpretation applied to the upper wind maximumzone of Fig. 7 would imply an upward velocity of 1 ms-1, a value typical of cumulus-scale convection, butnot of broader mesoscale precipitation areas such as theone containing the wind maximum seen in Fig. 7. Con-sequently, we reject the hypothesis of a continuousupward-sloping jet and suggest instead that air at alllevels moved horizontally through the upward-slopingzone of maximum wind speed.d. Asymmetric patternsIf the wind field within the echo region is uniform, themeasured Doppler velocity will exhibit symmetry, ex-cept for the contribution due to the vertical motion ofthe particles. In the cases described here, since theelevation angle was 7" and the combination of fallspeedand vertical air motion was generally less than 3 m s-l,this contribution would be at most 0.4 m s-l (Le., 3sin7"). But an asymmetry of at least 1.95 m s-l is0~~0~~~1977 BAYNTON ET AL. 1025needed to be detected as a color change. Accordingly,any observed asymmetric patterns in the display canbe related to some nonuniformity of the wind field.The example of an upward sloping wind maximum hasalready been given. The evidence was that the slantrange of the wind maximum was greater downwindthan upwind. Another kind of asymmetry is evident inFig. 4, which shows a display typical of a warm front.It is also a good example of difluence.A quick test for difluence or confluence is to lay astraight edge across the display from the middle of thezero-velocity band on one side and through the centerof the display. If the straight edge does not interceptthe middle of the band at the same range on the oppositeside of the radar, the flow is either difluent or confluent.Applying this test to Fig. 4, at a slant range of 40 km(z= 5 km), shows that the wind direction is 247" northof the radar and 260" south of the radar.FIG. 3. Color display typical of warm air advection. Picture isfor 2205 PST 13 January 1976, Tacoma, elevation 7.1". Rangemarks at 20 and 40 km, north at top of picture, velocity color codein the right column. The zero-velocity band bisecting the picturein the shape of an S indicates winds veering from south near theground to west-southwest aloft. Continuing color changes to thewest-southwest and east-northeast out to full range also indicatethat the veering winds extend to the echo top.FIG. 5. Color display indicating an occlusion with warm airadvection near the surface and cold air advection aloft. Pictureis for 0705 PST 22 January 1976, Tacoma, elevation 7.1") rangemarks at 20 and 40 km. The S within 15 km slant range indicatesveering winds and warm air advection to a height of 1.9 km.Beyond 15 km the zero-velocity band traces a counterclockwisespiral, indicating backing winds and cold air advection.FIG. 4. Color display typical of a warm front. Picture is for2036 PST 26 January 1976, Tacoma, elevation 7.1", range marksat 20 and 40 km. Confinement of the S-shape of the zero-velocityband to the lowest 1.2 km (to 10 km slant range), and the twopie-shaped grey wedges extending in from 40 to 10 km slant range,indicate that the veering wind is confined to the lowest 1.2 kmand place the warm frontal surface at 1.2 km.FIG. 7. Color display of two velocity maxima at 0006 PST 14January 1976 (date incorrect in picture). Elevation 11" rangemarks at 20 and 40 km. Each velocity maximum produces a pairof oval-shaped rings, one upwind and one downwind. The lowervelocity maximum core is horizontal, appearing at the same slantrange (10 km) both upwind and downwind. The upper velocitymaximum core slopes upward, appearing at a greater slant rangedownwind than upwind.1026JOURNAL OF APPLIED METEOROLOGYVOLUME 16N22 JAN 19760705 PST .E O 7"EN1 WARM AIR ADVECTION IN LAYER 0.4-1.9 km COLD AIR ADVECTIONI IN LAYER 1.9-4.2 kmA4.2 km /y,-THERMALWINDSFIG. 6. Elements derived from the color display shown in Fig. 5. Wind directions are obtained from the zero-velocity band bisectingthe figure by subtracting 90" from the azimuth on the low-pressure northwest side or adding 90" to the azimuth on the high-pressuresoutheast side. The two direction scales make this adjustment. Upwind and downwind speeds read from the color display are shownon the height contours. The hodographs confirm warm-air advection in the layer of veering winds and cold-air advection in the layerof backing winds.Occasionally convergence may he suspected by in-spection of the color display. Fig. 3 is an example. Ata slant range of 20 km (z= 2.5 km) the maximum radialspeed downwind (northeast sector) nearly coincideswith the "folded over" contour (red to purple) of14.60 m sV1, whereas upwind of the radar (southwestsector) the display shows three colors, or nearly 6 mgreater than the "folded over" contour. The discussionin section 5 verifies the presence of convergence, atthe same height, 30 min later than the time of Fig. 3.4. Comparison with rawinsonde dataWind profiles were obtained from the radar colordisplay using the procedure described in the captionof Fig. 6. Three comparisons between radar wind pro-files and simultaneous rawinsonde data, obtained at theUniversity of Washington in Seattle, 45 km north ofthe radar site, are illustrated in Fig. 8. Since the ob-servations were taken during periods of frontal passages,some differences in the soundings are to be expecteddue to atmospheric spatial variations. In spite of this,the results from the two independent methods are verysimilar.During CYCLES the antenna scan sequence includeda series of conical scans, vertically pointing observationsand RHI scans. A single 360" conical scan, with eleva-tion fixed, required 36 s and the sequence of 4", 7", ll",15" and 19" conical scans was repeated as often as every7 min. These data provided a valuable supplement tothe conventional rawinsondes which were obtained nomore frequently than every 45 min.5. Divergence computationsProcedures of computing divergence from Dopplerradar data are described by Caton (1963) and byBrowning and Wexler (1968). While their techniqueswere applied to digitally recorded data, the underlyingprinciples are readily adapted to photographs of thecolor display.Divergence may be evaluated from the expression(3)where is the average radial velocity at an elevationangle LY and a slant range of R during a 360" scan, andOCTOBER 1977 BAYNTON ET AL.1027Radarsonde0736 0740439700 g I Wav)WaCI800 2a9001000PSTFIG. 8. Comparison of radar and rawinsonde wind data for 10 January 1976. Half wind barb is for 2.5 m s-l, full barb is for 5 m s-l and flag is for 25 m s7.v, is the mean vertical velocity of the precipitationparticles. R is expressed in meters and velocities are inmeters per second, positive away from the radar.Generally, one is interested in a vertical profile ofdivergence, and in order that the profile may apply overthe same horizontal area one must use a series of scansat different elevation angles. Our procedure is to com-pute v at a slant range of 20 km, the inner rangemarker on the color display, at elevation angles of 4",7", ll", 15" and 19" corresponding to heights of 1.55,2.59, 3.97, 5.33 and 6.66 km above mean sea level.This series of scans is completed in about 7 min. Thechoice of 20 km complies with Browning and Wexler'srecommendation that the slant range should be lessthan 23 km in order to reduce height errors.Color transparencies for each elevation angle are pro-jected onto polar coordinate graph paper. The 20 kmrange mark and the location of all color transitions at20 km are traced in pencil onto the graph paper. Thisproduces a diagram similar to the inner portion of Fig.2. Then, proceeding in 6" steps around the 20 km circle,radial velocities are determined for each step and vis computed from60v= c li/60,i=l(4)TABLE 1. Mean vertical velocity t, of precipitation particles,ground clutter effects removed, positive upward, between 2208and 2258 PST, and divergence at 2237 PST 13 January 1976.Height (km MSL) 1.55 2.59 3.97 5.33 6.66Divergence s-l) -1.62 -1.07 -0.44 f1.05 +1.37P, (m s-1) -1.9 -1.2 -1.0 -0.8 -0.7where I; is the mean radial velocity of the color bandintercepted at the ith step.Vertical velocities of precipitation particles [Vz inEq. (3)] can be estimated by pointing the antennavertically. During the CYCLES Project verticallypointing scans are interspersed regularly with theconical scans. Generally small fallspeeds, characteristicof snow, were observed as one would expect with thelow freezing levels. Fallspeed spectra were recordedduring the CYCLES Project and these spectra havebeen used to remove the influence of ground clutter inestimating average vertical velocities of the precipita-tion particle^.^ During 23 min of vertical scanning, be-tween 2208 and 2258 PST on 13 January 1976, thevertical velocities given in Table 1 were observed. Thesevalues of vz are believed to be representative of PacificNorthwest storms and were used to obtain the diver-gence profile given in Table 1. The magnitudes of theconvergence and divergence in this case are consistentwith the results of Matsumoto et al. (1967) who found,from measurements with closely spaced rawinsondes,that divergence in the mesoscale precipitation areasof winter storms over Japan was -lop4 s-'.Errors in the estimates of divergence that are peculiarto this procedure can be evaluated. Since only the term3 When vertically pointing, a second data system was activatedwhich digitized the raw Doppler signals and recorded them ondigital magnetic tape. Doppler spectra were computed later fromthe complex time series using a fast-Fourier transform algorithmand general purpose computer. Bias of the mean velocity estimatesdue to ground clutter was eliminated by removing the zero velocityspectral line from the spectrum. The mean vertical Dopplervelocities were then obtained from the spectra using the objectivethresholding method of Hildebrand and Sekhon (1974) to removebias effects due to the background noise spectral density.1028 JOURNAL OF APPLIED METEOROLOGY VOLUME 1628/R cosa in Eq. (3) is derived by this procedure, wecan obtain the desired error estimates by determiningav, the standard deviation of v, and evaluating(2/R cosa)uv.From the expression for v in Eq. (4) we have forthe variance of v60Var(8)=C (1/60)2 Var(Bi), (5)i=land since 6i is obtained by rounding off within an inter-val of approximately 2 m s~~ Var(gi)=22/12=+ m2 s-~.Thus from Eq. (5) we obtain av=0.07 m s-l and thedivergence errors peculiar to this procedure are foundto be 0.07X10-4 at a height of 1.55 km, increasingto 0.08X10-4 sP1 at 6.66 km. Since these errors are anorder of magnitude smaller than the divergence esti-mates of Table 1, the procedure is evidently quiteaccurate.6. ConclusionsExperience using the NCAR CP-3 radar in the Uni-versity of Washington's CYCLES Project has shownthat colored Doppler velocity patterns provide a varietyof useful information in real time on the wind field insituations of widespread precipitation. These displaysdo not substitute for the more precise and accuratemotion field analyses that may be performed after thefact. Rather, they augment such analyses by permittingmeteorologists to identify significant features that arelikely candidates for further analysis. The displays per-mit rapid measurements of wind speed and directionas a function of height, and divergence which can becomputed from pictures of the radar color display withsuitable speed and accuracy for mesoscale analysis.Readily recognizable patterns in the color displayshave shown warm and cold advection layers, windmaxima and zones of difluence or confluence. Warmadvection is indicated by an S-shaped zero velocityband, while cold advection is indicated by a backwardS. Maximum wind layers appear as concentric colorrings around a maximum velocity core. Difluence orconfluence is indicated by asymmetry in the zero-velocity band. The displays also permit the meteo-rologist to optimize his data collection methodologythrough real-time decision making, thus concentratinghis instrument resources (radars, aircraft, rawinsonde,etc.) on the most interesting situations.Acknowledgments. The CYCLES Project is supportedby Grant ATM 74-14726-A02 to the University ofWashington from the Meteorology Program of theAtmospheric Research Section of the National ScienceFoundation. The successful operation of the radar wasdue to the skill and dedication of NCAR techniciansRobert Bowie, Alan Sorenson and Dale Zalewski.Thomas J. Matejka of the University of Washingtonassisted in the collection of data. REFERENCESBattan, L. J., 1973: Radar Observation of the Atmosphere. The University of Chicago Press, 324 pp.Browning, K. A,, and R. Wexler, 1968: The determination ofkinematic properties of a wind field using Doppler radar.J. Appl. Meteor., 7, 105-113.-, M. E. Hardman, T. W. Harrold and C. W. Pardoe, 1973:The structure of rainbands within a mid-latitude depression.Quart. J. Roy. Meteor. SOC., 99, 215-231.Caton, P. A. F., 1963: Wind measurement by Doppler radar.Meteor. Mag., 92, 213-222.Elliott, R. D., and E. L. Hovind, 1965: Heat, water, and vorticitybalance in frontal zones. J. Appl. Meteor., 4, 196-211.Gray, G. R., R. J. Serafin, D. Atlas, R. E. Rinehart and J. J.Boyajian: 1975 : Real-time color Doppler radar display.Bull. A7ner. Meteor. Soc., 56, 580-588.Hildebrand, Peter H., and R. S. Sekhon, 1974: Objective deter-mination of the noise level in Doppler spectra. J. Appl.Meteor., 13, 808-81 1.Houze, R. A,, Jr., J. D. Locatelli and P. V. Hobbs, 1976: Dynamicsand cloud microphysics of the rainbands in an occludedfront. J. Atmos. Sci., 33, 1921-1936.Kreitzberg, Carl W., 1964: The structure of occlusions as deter-mined from serial ascents and vertically-directed radar. Res.Rep. AFCRL-64-26, Air Force Cambridge Research Labo- ratories, 121 pp.-, and H. A. Brown, 1970: Mesoscale weather systems withinan occlusion. J. Appl. Meteor., 9, 417432.Lhermitte, R. H., 1972 : Real time processing of meteorologicalDoppler radar signals. Preprints 15th Radar Meteorology Conf.,Champaign-Urbana, Amer. Meteor., Sac. 364-367.-, and D. Atlas, 1961 : Precipitation motion by pulse Doppler.Preprints Ninth Radar Meteorology Conf., Kansas City,Amer. Meteor. Sac., 218-223.Matsumoto, S., K. Nimomiya and T. Akiyama, 1967: Cumulus activities in relation to the mesoscale convergence field.J. Meteor. SOC. Japan, 45, 292-304.