1184 JOURNAL OF ATMOSPHERIC AND OCEANIC TECHNOLOGY VOLUME 11Boundary Layer Clear-Air Radar Echoes: Origin of Echoes and Accuracy of Defive~ W~nds JAMES W. WILSON National Center for Atmospheric Research,* Boulder, Colorado TAMMY M. WECKWERTHNational Center for Atmospheric Research, Boulder, Colorado and University of California at Los Angeles, Los Angeles, CaliforniaJ. V1VEKANANDAN National Center for Atmospheric Research,- Boulder, Colorado ROGER M. WAKIMOTO University of California at Los Angeles, Los Angeles, California ROBERT W. RUSSELL University of California at Irvine, lrvine, California (Manuscript received 2 August 1993, in final form 7 March 1994) AI~TRACT Boundary layer clear-air echoes are routinely observed with sensitive, microwave, Doppler radars similar tothe WSR-88D. Operational and research meteorologists are using these Doppler velocities to derive winds. Theaccuracy of the winds derived from clear-air Doppler velocities depends on the nature of the scatterers. Thispaper uses dual-wavelength and dual-polarization radars to examine the cause of these echoes and the use ofDoppler velocities from the clear-air return to estimate winds. The origin of these echoes has been an ongoingcontroversy in radar meteoroloID'. These echoes have been attributed to refractive-index gradient (Bragg scattering)and insects and birds (paniculate scattering). These echoes are most commonly observed over land from springthrough autumn. Seldom do they occur over large bodies of water. Widespread clear-air echoes have also beenobserved in winter when temperatures are above 10-C. Radar refiectivity comparisons of clear-air echoes in Florida and Colorado were made at radar wavelengthsof 3, 5, and 10 cm. These comparisons, when analyzed along with a theoretical backscattering model, indicatethat the echoes result from both paniculate and Bragg scattering with paniculate scattering dominating in thewell-mixed boundary layer. The return signal in this layer is highly horizontally polarized with differentialreflectivity Zog values of 5-10 riB. This asymmetry causes the backscattering cross section to be considerablylarger than one for a spherical water droplet of equal mass. At X band and possibly even at C and S band thescattering enters the Mie region. It is concluded that insects are primarily responsible for the clear-air echo inthe mixed boundary layer. At and above the top of the well-mixed boundary layer, Bragg scattering dominatesand is frequently observed at S band. When insects and birds are not migrating, the Doppler velocities can be used to estimate horizontal windsin the boundary layer. Viewing angle comparisons of Zvg values were made to determine if migrations wereoccurring. Migrations were not observed in Horida and Colorado during summer daylight hours. Limitedcomparison of winds derived from Doppler radar with balloon-sounding winds showed good agreement. However,a more extensive study is recommended to determine the generality of this conclusion. Dual-Doppler analyses show that thin-line echoes are updraft regions. Comparison of these radar-derivedvertical velocities with aircraft-measured vertical velocities showed a correlation coefficient of 0.79. In addition,the position of small-scale updraft maxima (1-2 km in diameter) along the sea-breeze front correspond toindividual cumulus clouds. The good agreement between dual-Doppler-derived vertical motion fields and theseother independent vertical velocity measurements provides evidence that the dual-Doppler-derived wind fieldsin the dear-air boundary layer are accurate and capable of providing details of the wind circulations associatedwith horizontal convective rolls and the sea breeze. * The Nat/onal Center for Atmospheric Research is partially sponsored by the National Science Foundation. Corresponding author address: James W. Wilson, NCAR, RemoteSensing Facility, P.O. Box 3000, Boulder, CO 80307.1. Intr~duc~on The purpose of this paper is twofold: first, throughthe use of mulfiparameter radar data to examine theorigin and characteristics of boundary layer clear-airc 1994 American Meteorological SocietyOCTOBER 1994 WILSON ET AL. 1185radar echoes and, second, to examine the accuracy ofclear-air wind fields derived from Doppler radar data.Results of these inquiries are pertinent to both researchand operational meteorologists using Doppler radar tomeasure winds. Radar echoes from the clear-air atmosphere--thatis, free of cloud and precipitationmhave been observedsince the introduction of radars. There has been considerable debate in the literature and at scientific meetings since the early 1940s on the cause of these echoes.This history has been well documented by Fletcher(1990), Katz and Harney (1990), Hardy and Gage(1990), and Gossard (1990). Some researchers concluded that birds and insects were the cause (Crawford1949), while others concluded the source was refractive-index gradients from moisture and temperaturegradients on a scale of one-half the radar wavelength(Atlas 1960). Battan (1973) reviewed an extensivenumber of studies on clear-air echoes and stated thatthe echoes "are caused primarily by insects, birds orregions of the atmosphere where there are strong refractive-index gradients." Hardy and Katz (1969), using the Wallops Island three-wavelength radar, concluded that the echoes were primarily from insects andbirds. Other selected discussions on this subject maybe found in Doviak and Zrni6 (1984), Vaughn (1985),Mueller and Larkin (1985), Wilson and Schreiber(1986), and Achtemeier ( 1991 ). Research Doppler radars and now the operationalWSR-88D and Terminal Doppler Weather Radar(TDWR) routinely observe, from spring to autumn,clear-air echo in the convective boundary layer toranges of at least 50-100 km. Clear-air echo is alsoobserved during the winter but not as frequently norto as great a range. Boundary layer wind-shift lines, 1generated by synoptic-scale and mesoscale weatherevents, are usually easily identified as lines of enhancedradar reflectivity factor (i.e., thin-line echoes) and / orregions of radial velocity convergence. This capabilitymakes it possible to routinely monitor important windfeatures in the boundary layer in a manner similar tothat from satellite imagery (Purdom 1982) but withgreater temporal and spatial resolution. It has become common to assume that the scatterersresponsible for the clear-air echoes are passively cardedby the wind. For example, the WSR-88D has an operational algorithm that routinely provides operationalforecasters with wind profiles (Klazura 1993) that arebased on the velocity-azimuth display (VAD) technique (Browning and Wexler 1968). Researchers areusing the clear-air echoes to describe kinematic detailsof clear-air wind features (Doviak and Berger 1980;Rabin and Doviak 1982; Mueller and Carbone 1987;Mahoney 1988; Carbone et al. 1990; Wilczak et al. ~ The term "lines" is used to emphasize the small width ( 1-3 km)of the convergence zones.1992; Wakimoto and Martner 1992; Wilson et al.1992). The reliability of using echoes from birds andinsects as a means for estimating winds depends onwhether their direction of flight within the radar sampling volume is random or they are being carded passively by the wind. In the case of insects and birds, theaccuracy of this assumption is not clear; for migratingbirds or roosting birds dispersing to their daily feedingareas, this assumption can be incorrect (Larkin 1991;Vaughn 1985). Even for insects this assumption canbe wrong. Mueller and Larkin ( 1985 ) and Achtemeier(1991 ) have shown that insects may fly in mass witha common heading that does not coincide with thewind velocity. Thus, it is important to determine theprimary scattering mechanism. If it is birds and insects,considerable error could occur when wind fields arededuced from Doppler radar measurements. This paper examines data from four different climatic regions to determine the primary characteristicsand causes of the clear-air echoes and addresses theiraccuracy for obtaining wind velocities. The primary emphasis is with data from Florida andColorado where a dual-polarization radar was available.The Florida data were collected during the Convectionand Precipitation/Electrification (CAPE) Project(Wakimoto and Lew 1993), which was conductedduring the summer of 1991. Considerable dual-Dopplerdata were collected during that experiment with theintent of studying clear-air kinematic features of thesea-breeze front, horizontal convective rolls, and associated initiation of convective storms. There is a specific interest in whether vertical motions derived fromdual-Doppler analyses are sufficiently accurate for theabove purposes. Observed dual-wavelength and dual-polarizationcharacteristics of the clear-air echo are used in a theoretical backscattering model to estimate the type, size,and shape of the targets. Based on these results andinsect and bird observations in the literature, inferencesare made about the origin of the targets. Section 2 will describe the locations and characteristics of the radars and analysis methods used for thestudies. Section 3 will provide examples of clear-airechoes. Section 4 will provide theoretical and observational evidence concerning the origin and characteristics of the scatterers. Section 5 will provide datato address the accuracy of Doppler-derived horizontaland vertical winds.2. Radar locations and characteristics and dual Doppler analysis procedures Measurements are presented from four different climatological regions: Florida, Colorado, Kansas, andthe tropical Pacific Ocean. The Florida data were collected in 1991 during CAPE. The radars during CaPEwere the National Center for Atmospheric Research(NCAR) multiparameter radar, CP-2, and two C-band1186 JOURNAL OF ATMOSPHERIC AND OCEANIC TECHNOLOGY VOLUMEI1 FIG. 1. Display of radar reflectivity factor showing clear-air echofrom the east coast of Florida near Cape Canaveral. The echo abruptlyends at the shoreline accurately outlining Cape Canaveral northeastof the radar. The north-south thin-line echo is associated with theeast coast sea-breeze front. The data were collected with CP-4 at 1913UTC 4 August 1991 during the CaPE project. Locations of CaPEradars are indicated. Range marks are at 20-km intervals. Values ofdBZe are given by the color bar at the bottom of the image.Doppler radars, CP-3 and CP-4. The dual-Doppleranalyses that are presented in section 5 were obtainedfrom CP-3 and CP-4, which were located 23 km apartand about 40 km southwest of the Kennedy SpaceCenter. The locations of all three radars in CaPE areshown in Fig. 1. The data from Colorado were collectedwith CP-2 in 1992 during the Real-time Analysis andPrediction of Storms (RAPS) Project (Neilley et al.1993) from a location 10 km south of Boulder, Colorado. The data from Kansas were collected with CP4 in February and March 1991 during STORM-FEST(STORM-Fronts Experiment System Test). The datafrom the tropical Pacific were collected by the Massachusetts Institute of Technology (MIT) C-bandDoppler radar, which was located on the ship Vickers,during the Tropical Ocean Global AtmosphereCoupled Ocean-Atmosphere Response Experiment(TOGA COARE) (Webster and Lukas 1992) in December 1992. Table 1 provides some of the more pertinent characteristics of the NCAR radars, CP-2, CP-3, and C.P4, and the MIT radar. The data in Table 1 for the CPradars are for the CaPE project. Footnotes to Table 1give minimum sensitivity values for the STORM-FESTexperiment in 1992 and the RAPS experiment in 1992.. Particular care was taken in section 4 to verify therelative accuracy of the measured radar reflectivity factors of each CP radar. This was done by intercomparingradar reflectivity factors from different radars in care~fully selected precipitation echoes where measuredvalues should be equal--that is, in Rayleigh scatteringregions where the individual radars are sampling thesame volume and where contamination from groundtargets and/or attenuation of the radar signal by intervening precipitation was felt to be at a minimum.These comparisons showed the need to reduce theCP-2 S-band reflectivities between 1.5 and 3 dB. Basedon these findings the CP-2 reflectivities were reduced2 dB for the comparisons made in section 4. Synthesized horizontal and vertical velocity windfields from dual-Doppler data are presented in section5. Established analysis programs were used to errorcheck, manually edit, and analyze the original radarobservations. The data utilized in the dual-Dopplersyntheses were edited to remove ground clutter biasing,aliased Doppler velocities, and second-trip echoes.These functions were accomplished using the NCARResearch Data Support System interactive computersoftware (Oye and Carbone 1981). Specific editingtechniques were described by Wilson et al. (1984).TABLE 1. Selected radar characteristics. The values for the CP radars are for CAPE. Footnotes provide sensitivity values during RAPS and STORM-FEST. The letters H and V for polarization indicate horizontal and vertical polarization, respectively.CP-2 CP-3 CP-4 MITS band X band C band C band C bandWavelength (cm) 10.7Polarization H, VAverage transmitted power (dBm) 59.1Beamwidth (deg) 0.91Radar system gain (dB) 42.3Radar constant 73.3Minimum detectable retlectivity factorat 50 km (dBZe)a -6b 3.2 5.5 5.5 5.4HH H H43 57.1 56.6 51.90.95 0.9 0.9 1.645.1 43.5 41.5 40.574.2 67.3 70.4 72.0-5~ -18a -15e -9For a signal to noise ratio of -6 dB.-11 dBZ~ for RAPS in 1992.-7 dBZ~ for RAPS in 1992.-22 dBZ~ for STORM-FEST in 1992.-21 dBZ~ for STORM-FEST in 1992.OCTOBER 1994 WILSON ET AL. 1187 An NCAR software package called REORDER wassubsequently used to interpolate the edited data ontoa Cartesian grid using the Cressman scheme (Cressman1959). The radius of influence of the Cressmanweighting function was set equal to 300 m or less depending on data spacing. The grid spacing was 300 m.The CEDRIC (custom editing and display of reducedinformation in Cartesian space) software package(Mohr et al. 1986 ) was used to obtain syntheses of thehorizontal and vertical winds. The gridded horizontalvelocities were smoothed with a two-step Leise (1982)filter. This essentially eliminates wavelengths less thanfour times the grid interval (4 x 300 = 1200 m). Datawere collected with azimuth and elevation steps ofabout 0.4-. Carbone et al. ( 1985 ) indicated that aboutsix to eight independent measurements are required torecover roughly 80% of the magnitude of a wave. Inthis case the data spacing is 300 m suggesting 80% resolution of 2-km wavelengths. However, the data onthe 300-m grid are not completely independent becausethe radar half-power beamwidth at the range of interestis roughly 500 m; thus recovery will be less than 80%for a 2-km wave. Since the full wave would be madeup of the updraft and downdraft, the convective updrafts on a scale of 1 km should be detected, althoughwith reduced magnitude. Vertical velocities were obtained by a variationaladjustment of the anelastic mass continuity equationsuch that the integration was forced to meet both upperand lower boundary conditions (O'Brien 1970). Thelower boundary condition was set to zero at the earthsurface and the upper boundary condition was set tozero one-half grid point above the top of the clear-airor cloud echo. In addition, vertical velocities were alsoobtained by upward only integration. Because of theshallow depth over which the integration is performedit might be expected that upward integration alonewould be superior to the variational technique. However, comparison with independently deduced verticalmotions discussed in section 5 showed that the variational technique provided better results. The poorerresults from upward integration likely can be attributedto errors in divergence estimates near the earth surfacecaused by clutter contamination or antenna beamovershooting. The effect of such errors is mitigated inthe variational scheme by inclusion of the additionalupper boundary condition. Rationale for assuming thevertical velocity is zero at the top of the clear-air echofollows. As will be shown and discussed later, the clearair echo is primarily particulates borne aloft by updrafts. These echoes fill the well-mixed boundary layer.The top of the particulate echo tends to mark the topof the well-mixed boundary layer. Caughey and Palmer(1979) have shown the vertical velocity variance approaches zero at this level and at least in a statisticalsense it is expected that the vertical velocity is relativelysmall. In addition to errors in the resulting synthesizedwinds caused by possible insect or bird flight, there arenumerous other possible error sources that have received extensive discussion in the literature. These arestatistical uncertainties in the radial velocity winds(Doviak et al. 1976), geometrical considerations(Miller and Strauch 1974; Ray et al. 1979), nonsimultaneity of the observations (Miller and Kropfli1980), contamination of the radial velocity estimatesby ground clutter, inadequate sampling of the lowestlevel divergence field, and inappropriate upper andlower boundary conditions on the vertical velocities.The larger errors are probably associated with cluttercontamination, inadequate sampling of the very lowestlevel divergence, and errors in boundary conditions.Using the theoretical techniques described by Ray andWagner (1976), Ray et al. (1978), and Wilson et al.(1984), the error variances of derived vertical velocitiesvaried between 0.5 and 3.0 m2 s-2. These values arefor a height of 700 m located 10 km west and 10 kmsouth of CP-4, which is near the center of the analysisregion discussed in section 5. The higher error varianceassumes a standard error of I m2 s-2 for the interpolated radial velocities and the lower vertical velocityerror assumes a value of 0.1 m2 s-2. There is no methodfor estimating the actual errors in the radial velocityestimates since they are certain to vary in space andtime. Thus, theoretical means are not very useful forassessing the accuracy of dual-Doppler-derived verticalvelocities. Rather, accuracies are based on comparisonsin section 5 of dual-Doppler-derived vertical velocitieswith those measured by aircraft and inferred from cumulus cloud locations.3. Examples of clear-air echo Figures 1-4 show examples of clear-air return fromC-band radars at several locations and during variousmeteorological situations. Figure 1 is an example ofclear-air return from a summer afternoon in Florida.With the exception of echo from ground targets withinapproximately 5 km of the radar (yellow echoes), theechoes are from the clear air. The shoreline and CapeCanaveral to the northeast of the radar are vividly outlined by the absence of return over the ocean. Thegeneral background clear-air radar reflectivity factorover land is between -5 and 5 dBZe (blue and greencolors). The yellow north-south thin-line echo, withmaximum radar reflectivity factors of 20-30 dBZe, isthe leading edge of the east coast sea-breeze front thathas penetrated 10-20 km inland. There are a numberof thin lines of 10-15 dBZe (dark green) that are produced by horizontal convective rolls that are not welldefined in this case. A more illustrative example of horizontal convectiverolls is shown in Fig. 2 for a warm winter day in Kansas.Satellite imagery shows that there are no clouds, whichis to be expected from the very dry boundary layer1188 JOURNAL OF ATMOSPHERIC AND OCEANIC TECHNOLOGY VOLUME 11156-E during the TOGA COARE in December 1992.The thin-line echo north and northeast of the radar inadvance of the more intense precipitation echoes isapparently the leading edge of the outflow from thesestorms. This thin line has a radar reflectivity factor of-10 to 0 dBZe. In marked contrast to the previousfigures there are no other clear-air echoes. The circtdarecho surrounding the radar is from precipitation. Perusal of many hours of data from this ship-based radarshows no other clear-air echoes except for a rare thinline similar to the one in Fig. 3. Figure 4 shows an east-west-oriented thin-line echowith reflectivity factors of about -15 to -5 dBZe. llt isassociated with a synoptic-scale cold front movingthrough Kansas during February. The -25 to -20dBZe north-south thin lines north of the front appearto be caused by horizontal convective rolls, whose orientation is consistent with the wind direction indicatedby the Doppler velocities (not shown). Widespread clear-air echo, as in Fig. 1, is commonover land areas during the warm season. Radar reflectivity factors outside convergence lines are typically 010 dBZe. Clear-air echo over land during the wintermonths is less prevalent, usually being confined to closeranges and displaying weaker intensity. Extensive wintertime clear-air return, as in Fig. 2, occurs only on 'thewarmer days. In Hawaii, Florida, and the northerncoast of Australia (Keenan et al. 1991 ), clear-air echoover the ocean has [arely been observed except for shortdistances offshore in strong offshore wind conditions.The origin of the clear-air echoes shown in Figs. 1-4is discussed in section 4b. FIG. 2. (a) Display of radar reflectivity factor showing thin-lineechoes associated with horizontal convective rolls. Data were collectedwith CP-4 on 1 March 1992 in Kansas during STORMoFEST. Rangemarks are at 20-kin intervals. (b) Doppler velocity display corresponding to (a). Negative (positive) values depict motion toward(away from) the radar. The velocity scale is given by the color barat the bottom of the image.shown by a nearby sounding. When sufficient moistureis present, cloud streets often exist that coincide withthe thin-line echoes (Christian and Wakimoto 1989).The reflectivity factors in Fig. 2a are between -20 and-5 dBZ~; 10-15 dB less than those in Fig. 1. The radialvelocities in Fig. 2b show the winds are about 240-,which is similar to the orientation of the thin lines inFig. 2a. Perturbations in the Doppler velocity field associated with the horizontal convective rolls are alsoevident in Fig. 2b. Figure 3 shows a gust front observed by a ship-basedDoppler radar in the tropical Pacific Ocean near 2-S,Front FIG. 3. Display of radar reflectivity factor showing a thin-line echoassociated with a gust front. Data were collected with the MIT radaron the Vickers ship on 9 December 1992 north of the Solomon Islandsduring TOGA COARE. Range marks are at 20-km intervals.OCTOBER 1994 WILSON ET AL. 1189 FIG. 4. Display of radar reflectivity factor showing an east-westthin-line echo associated with a cold front. Data were collected withCP-4 on 10 February 1992 in Kansas during STORM-FEST. Rangemarks are at 20-km intervals.4. Origin and characteristics of clear-air echoa. Theory1) PARTICULATE SCATTERING Radar echoes from the atmosphere are either fromparticulates, such as precipitation particles and insects,or from small-scale spatial variations in refractive index(henceforth, called refractive-index turbulence). Forparticulate scattering from spherical particles that areconsiderably smaller than the radar wavelength (about0.1 the wavelength), the Rayleigh scattering approximation can be used. This is typically the case for meteorological radars looking at rain. Mie scattering occurs from larger and/or nonspherical particles of evensmaller sizes. For Rayleigh scattering, the average power receivedPr by the radar due to rain can be written as (Battan1973)Pr = C[KI2ZD6r-2,(1)where C is a constant depending on the characteristicsof the radar set, I KI is the polarizability factor and itdepends on the refractive-index factor for water, r isthe range of the resolution volume, and D is the raindrop diameter and is summed over all particles in aunit volume. Here, ZD6 is defined as the familiar radarreflectivity factor Z. Since it is often not known whattype of scattering is occurring, the quantity "effectiveradar reflectivity factor" Ze is used in place of Z orZD6 in ( 1 ). Typically, Ze values are expressed in theirlogarithmic form: dBZe = 10 IogZe (mm6 m-3). (2)For Mie scattering from large or nonspherical particles,the received power is not simply dependent on the sixthmoment of the size distribution as in ( 1 ). Instead, themore exact formulation of the backscattering crosssection of the particles is required (see Battan 1973). Data that follow indicates that the clear-air scatterersin the well-mixed boundary layer are not spherical.This is largely based on differential radar reflectivityZr~R data to be shown in the following section, as wellas literature on insects and birds. Here, Zr>R is definedas Zr>R = 10 log ~ (dB), (3)where Zh and Zv refer to the horizontal and verticalcopolarized reflectivity factors, respectively. Vaughn (1985) has written a review article of thepublished literature concerning birds and insects as radar targets. He presents data that indicate insects areof sufficient size and concentration that they could account for the clear-air signals observed. He further reports that width-to-length ratios for birds vary between1:2 and 1:3 and insects vary between 1:3 and 1:10. Asa result he states that radar cross sections are moreaccurately modeled by prolate spheroids than they areby a spherical water droplet of equivalent mass, as doneby Riley (1985). When horizontal linearly polarizedradar is used, as is the case here, Vaughn (1985) reportsthat cross sections for some insects may be several orders of magnitude greater than that of an equivalentmass water sphere. Since it is likely that the scatterers are nonsphericaland to assist in interpretation of multifrequency radarobservations of clear-air signals in section 4b, theoretical estimates of radar reflectivity factor are made assuming that insects can be characterized as water droplets with shapes of prolate spheroids. Calculations aremade for S-, C-, and X-band radar with a variety ofwidth to length ratios. Results are provided in Fig. 5. The calculations that led to Fig. 5 were made withthe assumption that there is a monodispersed size distribution of water droplets (insects) with a concentration of 1 per 1000 m3 at a temperature of 15-C. In thecase of spherically shaped scatterers, Mie theory (Mie1908) was used to compute the horizontal reflectivityfactor. The nonspherical targets (width to length rationot 1 ) were assumed to be prolate spheroids with themajor axis on the X-Yplane and a uniform probabilitydistribution of orientation in the azimuthal (rI,) direction. The insert in Fig. 5a illustrates characteristics andorientation of the prolate spheroids. A brief descriptionof the technique that is used for calculating theoreticalreflectivity factors for prolate spheroids is given in theappendix. The effect of width-to-length ratio on the horizontalreflectivity factor is shown in Fig. 5a where the S-band1190 JOURNAL OF ATMOSPHERIC AND OCEANIC TECHNOLOGY VOLUME 11 a) S-band~ .--"''" 1:3 ................................. -~ .~'~ ..................... 1:1 20 / ...... ,- y.- //,....[i'~ '~ -20 -/.// ///"// I I I~2 4 6 8 10 Equivalent Spherical Diameter (ram)1510-10 b) S-band minus X-band~ 1 '3 ?~/? 1:1~ ~I I I5 lO 15 2025 Equivalent Spherical Diameter (mm) FIG. 5. Polarireetric radar reodel coreputations (sec appendix) for S, C, and X bands as a functionof equivalent spherical diameter. Both spherical and prolate spheroidal water drops arc considered.Vertical-to-horizontal aspect ratios are labeled on the curves. The orientation of the prolate spheroidsis shown in (a) and discussed in the text. (a) The S-band horizontal reflectivity, (b) refiectivitydifference between S and X bands, (c) reflectivity differences between S and C bands, and (d)S-band differential reflectivity.horizontal reflectivity factor is plotted as a function ofequivalent spherical diameter Deq (the equivalent diameter of a spherical water drop that has the samemass as a prolate spheroid). For example, in Fig. 5a aprolate spheroid with a width-to-length ratio of 1:5 anda Deq of 8 mm will have a Zh value 1 8 dB greater thana spherical droplet ( 1:1 ) of equivalent mass. Provided that Rayleigh scattering is occurring at twodifferent wavelengths, Ze will be equal for each wavelength. However, if Mie scattering is occurring at oneor both of the wavelengths, the Ze values will diffi~r.This is illustrated in Figs. 5b and 5c, which show thedifference in dBZe values between S band and X bandand between S band and C band as a function of widthOCTOBER 1994 WILSON ET AL. 119110-5-1020 Ic) S-band minus C-bandI5 10 15 20 25Equivalent Spherical Diameter (ram)I I I I I I I ~ /d) S-band ...t" j// // /// 1:5 .... --''''"2 1:2I I I I I I t 4 6 8Equivalent Spherical Diameter (mm) FIG. 5. (Continued)to-length ratio and De~. Non-Rayleigh scattering beginsto be a factor for Deq > 3 mm for S band and X bandcomparisons (Fig. 5b) and greater than 6 mm for Sand C-band comparisons (Fig. 5c). In addition sincethe CP-2 S-band radar measures differential radar reflectivity, Fig. 5d was constructed to show the effect ofwidth-to-length ratio of the scatterers on ZDR as a function of De,. It is apparent that insects and birds, whichare highly nonspherical, would exhibit large ZDR signals. Figure 5 will be used in section 4b to obtain estimates of the size and shape of the clear-air scatterers.2) BRAGG SCATTERING Scattering from refractive-index turbulence (frequently referred to as Bragg scattering) in the atmosphere typically arises from variations in moisture, andto a lesser extent, temperature on a scale of one-half1192 JOURNAL OF ATMOSPHERIC AND OCEANIC TECHNOLOGY VOLUME 11 the radar wavelength. Gossard and Strauch (1983) and Doviak and Zrni6 (1984) discuss Bragg scattering in some detail. For Bragg scattering the strength of the returned sig nal is related to the refractive-index structure constant C2~ and is given by Ottersten (1969) as ~/= 0.38Cn2X-1/3, (4) where ,/is the radar reflectivity and X the radar wave length. Here, C2n depends on the spatial variance of moisture, and to lesser extent, temperature. Following the discussion of Knight and Miller (1993), C~ can be equated to Ze as C2, - IIS[rl2Ze 0.38Xll/3 . (5) They went on to show that for Bragg scattering, the ratio of Ze values for two radar wavelengths, designated by subscripts 1 and 2, is Zel ~/~.1 ~11/3 ~e2 = IX2/ , (6) or by expressing Ze in the conventional dBZe form given in (2) the difference in dBZe between two wave lengths is [~Xl\11/3 dBZ~l - dBZe2 = l0 logic] dB. (7) For Bragg scattering (7) gives the difference between S band and X band as 19 dB and between S band and C band as 11 dB with the S band being larger. These values are approximate values assuming that refractive-index turbulence is described by the in ertial subrange of turbulence where the kinetic energy spectrum obeys the k-5/3 law where k is wavenumber. See Knight and Miller (1993) for a more in-depth discussion. Gossard (1990) and Pratte and Keeler (1986) have summarized data from several sources to compare the radar return from particulates and refractive-index turbulence. They reported that Rayleigh particulate scattering from insects would be expected to range from roughly - 10 dBZe for small insects to 20 dBZe for very large insects with higher values possible in convergence lines. Scattering from refractive-index turbulence is more often observed at longer wavelengths. Reflectivity at 10-cm wavelengths from refractive-index turbulence, except for localized regions, should generally be less than roughly -25 dBZe (Pratte and Keeler 1986). However, Gossard (1990) felt that C2~ values greater than 10-12 m-2/3 (greater than -8 dBZe at 10-cm wavelength) were not unreasonable in limited regions associated with strong density currents. Knight and- Miller ( 1993 ) showed that cumulus clouds, when they first become visible at 5-cm wavelengths, have reflec tivities of about -10 dBZe in the Tropics and -20dBZ~ in the High Plains. Later with additional datafrom dual-wavelength radar, they showed this scatteringto be from refractive-index turbulence.b. Observational evidence of origin A comparison of dBZe values for S, C, and X bandwas made for a portion of the north-south thin-lineecho shown in Fig. 6. This feature delineates the Floridaeast coast sea-breeze front. The comparison was conducted over the boxed area in Fig. 6, which is locatedapproximately midway between CP-2 and CP-3. Theheight of the comparison ranged between 320 and590 m with the CP-3 heights averaging about 100 mlower than CP-2 heights. Figure 7 shows the area covered by each dBZe value for all three wavelengths. Itis evident that the reflectivity of the S and C bands areabout equivalent, whereas the X-band averages areabout 7 dB less. A histogram of gate by gate differencingof the S- and X-band reflectivities (not shown) alsoshowed that the distribution of differences peaked al: 7dB. In addition the S-band Zr>R values within the delineated comparison region in Fig. 6 averaged 7 dB;that is, the horizontally polarized reflectivity factorswere 7 dB greater than the vertical. These observations indicate that the predominantscattering mechanism is particulates. As discussedabove, if the scattering were predominantly Bragg scattering, the difference between S and X band and between S and C band should be about 19 and 11 dB,respectively, rather than the observed 7 and 0 dB. Al;soMueller and Larkin ( 1985 ) argue that, if scattering isfrom refractive-index fluctuations on a scale of one FIG. 6. Radar reflectivity factor display showing the region (whitepolygon) of reflectivity comparison between CP-2 and CP-3. Thenorth-south thin line (yellow) illustrates the Florida east coast seabreeze front. Data are shown from CP-3 at 1913 UTC 25 July 1991.Range marks are at 20-km intervals.OCTOBER 1994 WILSON ET AL. 11935 -10 0 10 20 30 RADAR REFLECTIVITY (dBZe ) FIG. 7. Comparison of radar reflectivity values between the CP-2S-band and X-band radars and the CP-3 C-band radar. Comparisonsof area (kin2) versus radar reflectivity factor (dBZe) are for the outlinedregion across the sea-breeze front shown in Fig. 6 at a height rangingbetween 320 and $90 m.half the wavelength due to isotropic turbulence, thevalues of Zr~R will be zero. A nonzero Zr~R value wouldthen indicate that the return is likely due to particulatesrather than Bragg scattering. Utilizing Fig. 5, an estimate of the particle size andshape can be made. First from Fig. 5d a Zr~R value of7 dB indicates that the width-to-length ratio is probablybetween 1:2 and 1:3. For this ratio and a 7-dB differencebetween the S and X band Zh, Fig. 5b would indicatethe equivalent spherical diameter should be between 8and 10 mm. Since there was essentially no differencebetween the S- and C-band Zh values, Fig. 5c showsthat the only size that can satisfy all three observedmeasurements is for a width-to-length ratio of about1:3 and a Deq of about 10 mm. Vaughn's (1985) data indicate that this particlesize is more consistent with insects being the predominant scatterer than birds. In addition he reportsthat, except for migrating birds or a few varieties ofsoaring birds, they usually fly at altitudes below100 m. It is quite common to observe in the radardata widely scattered point targets that have reflectivities and radial velocities inconsistent with othersurrounding clear-air echoes. It is likely that manyof these echoes are soaring birds, though some areknown to be airplanes. The nearly uniform, widelyspread nature of the majority of the clear-air signalthroughout the mixed boundary layer is more consistent with the targets being insects that are passivelycarried by the wind than being strong flyers like birdsthat, except when migrating, are commonly distributed in a patchier nature. Thus the clear-air returnof Fig. 6 is likely caused primarily by insects. Figure 3 shows an apparent clear-air thin-line echo(- 10 to 0 dBZe) over the ocean far from any land withno other clear-air return evident. The thin-line originated near the precipitation echo, expanded in length,and moved in advance of that echo, similar to gustfront observations over land. It is possible that the thinline echo was caused by very high values of refractiveindex turbulence (Bragg scattering) associated with agust front. If this echo was caused by Bragg scattering,it would require C2~ >10-tl m-2/3, which would be atthe high end of expected C2n values possible with theleading edge of density currents (Gossard 1990). Otherpossibilities are very light rain echo from clouds alongthe gust front or from rough water caused by the windshift and increased wind speed. Available satellite cloudimagery was not of sufficient resolution to determineif a line of cumulus clouds was associated with the thinline. The origin of the wintertime boundary layer clearair echo observed by CP-4 in Figs. 2 and 4 from Kansascannot be determined unequivocally. This is becauseonly one radar wavelength and one polarization wasavailable. In the case of the horizontal convective rollsin Fig. 2, where echoes of -15 to -5 dBZe were observed almost continuously out to a range of 40-60km, a C2n value of 10-12 m-2/3 would be required forthe echo to be from Bragg scattering. This is far aboveC,2 values that have been observed over an extendedarea (Gossard 1990). Thus it is more likely that thescattering in Fig. 2 is from particulates. The origin ofthe thin-line echo in Fig. 4 is not evident. For the echoto be the result of Bragg scattering at C band, C2, valueswould again need to be greater than 10-it m-2/3. Asstated above, Gossard (1990) has indicated this maybe possible in frontal zones and this thin line is associated with a cold front. Further evidence regarding the origin of the wintertime clear-air echoes comes from a comparisonof the intensity and range of the clear-air echo withboundary layer thermodynamic data. Surface temperature and lapse rate information from 3 Februarythrough 12 March 1992 were obtained from asounding site 50 km west of CP-4. Figure 8 showsthe intensity and maximum range of the clear-airecho as a function of the surface temperature. Toqualify as a case in Fig. 8 a 360- PPI (plan positionindicator) scan, not dominated by precipitation, atan elevation angle of 1.4- was required within 15min of the sounding release time. It is evident fromFig. 8 that the intensity and range of the clear-airecho increases with temperature. With the exceptionof one case, a temperature of 10-C was the dividingline between clear-air echo and no clear-air echo.There was also a tendency for the range and intensityof the clear-air echo to increase when there was a1194 JOURNAL OF ATMOSPHERIC AND OCEANIC TECHNOLOGY VOLUME 1125201510-5 17 2o---- 10-- 18 20-40' -- 10- -- 10- 20-30.-- 40 -15-2050-60 km40-50.30-40'NONE -30 -25 -20 -15 -10 -5 0 dBZe FIG. 8. Surface temperature versus intensity (dBZe) of clear-air echo for 25 cases between 3February and 18 March 1992 as observed by the CP-4 radar in Kansas during the STORM-FESTexperiment. The range to which the clear-air echo was observed at an antenna elevation of 1.4-is given by the numbers along each line. The short lines above "none" indicate there was noclear-air echo.deep dry-adiabatic lapse rate. The one case with clearair echo at temperatures below 10-C occurred on 11March and had a dry-adiabatic layer extending 575m above the ground. The coldest case with the -3-Ctemperature occurred 6 h prior to this case. Figure2 depicts the warmest case (25-C). The winter of1992 was abnormally warm in Kansas and the diurnal cycle of clear-air echo suggested there were anumber of local insects. It is speculated that if therehad been more normal periods of cold temperatures,fewer local insects would have been present and theclear-air echo would have been less prevalent. Data on the effect of temperature on insect flighthas been reported by Cockbain ( 1961 ), Taylor (1963),and Walters and Dixon (1984). Lower temperaturethresholds for flight vary both within and between insect species. However, the studies of the above authorsshow that flight is seldom observed below 10-C andfrequency of flight increases rapidly between 10- and20-C. This is consistent with the strong dependence ofclear-air return on temperature shown in Fig. 8. Thisfurther supports the notion that insects are the sourceof most of the clear-air return observed during thistime period in Kansas.c. Vertical characteristics Examples of the vertical distributions of clear-airecho are shown from the X-band (Fig. 9a) and S-ba~ad(Fig. 9b) wavelengths of CP-2 in Florida. These datawere obtained from RHI (range-height indicator) scansthrough a particularly intense thin-line echo producedby a gust front. The gust front thin line is at a radarrange between 8 and 12 km. It can be seen that thereflectivity factor in the precipitation echo at the farright of each RHI is the same at both wavelengths.However, the maximum reflectivity factor in the thinline is 30-35 dBZe at S band compared to only 15-120dBZe at X band. The reflectivity in the clear-air of Fig.9 decreases rapidly at 1-km height, near the top of theadiabatically mixed boundary layer. Furthermore, theclear-air echo at S band extends to a height of about 4km compared to only 1 km for the X band. The averagedifference between the S- and X-band reflectivity factoralong this gust front was 13 dB (determined from PPIscans). This is 6 dB greater than the case in Figs. 5and 7. The ZDR values within the thin line were primarily between 7 and 10 dB, indicating particul~ttescattering. Above 1 km the ZDR values at S band wereOCTOBER 1994 WILSON ET AL. 1195 FlG. 9. Range-height (RHI) radar reflectivity factor display through a gust front approaching CP-2 on 4 August 1991 during CAPE. TheRHI is directed toward 332-. The leading edge of the gust front is near 8 km. The distance scales are in kilometers; (a) X-band radar and(b) S-band radar.near 0 dB, suggesting Bragg scattering in this region.This case was unusual in that the thin-line echo reflectivity factors were larger than typical and the difference between S and X band was greater, suggestingmore elongated insects. Figures 10a and 10b show vertical profiles of clearair radar reflectivity factor within thin lines for the CP2 S- and X-band radars and ZDR for the S-band radarfor representative cases from Florida and Colorado,respectively. The Florida case is for the 4 August seabreeze front in Fig. 1, except for a time 1.5 h earlier.Median reflectivity factors shown were determinedwithin rectangular areas of about 50 km2 centered ata range of 15 km. The right half of each figure showsa nearby sounding. In the Florida case the atmosphereis well mixed (adiabatic lapse rate) to 1 km comparedwith 3 km for the Colorado case. The depth of theclear-air signal is considerably deeper in Colorado thanFlorida corresponding to the much deeper mixedboundary layer. In both cases the X-band reflectivityis less than the S band and the reflectivity factor decreases with height. There is no detectable X-band return above the adiabatically mixed layer at the range( --~ 15 km) of these examples. In both cases the S-bandreturn extends above the mixed layer. The signal abovethe mixed layer is variable, ranging primarily between-10 and 0 dBZe in Florida and between -15 and -5dBZe in Colorado. The ZvR at S band above the mixedlayer is near 0 dB (shown by the numbers next to theS-band curves), suggesting Bragg scattering. This becomes more evident at very close ranges where thereis also return at X band above the mixed layer. In thesecases (not shown), as would be expected for Braggscattering, the X band is roughly 20 dB less than theS band and Zr~R is near zero. In contrast to Florida, the difference between the Sand X-band reflectivities for Colorado decreases withheight and becomes almost equal at the top of themixed layer. This suggests that in Colorado the sizeand/or number of the insects is decreasing with height.The ZvR signal also decreases with height and becomeszero above the mixed layer. It is hypothesized that thelarger insects are more successful at resisting the updrafts and that only the smaller less elongated insectsare present at the higher, colder levels in Colorado.This is partially supported by Achtemeier ( 1991 ), whoshowed that grasshoppers were not carried to higherlevels in updraft regions. The vertical profiles shown in Fig. 10 were takenwithin thin-line echoes. The Florida case was for a seabreeze front and the Colorado case was for a stationaryconvergence line of unknown origin; it was not a gustfront. Vertical profiles similar to those in Fig. 10 butfor areas close to but not in the thin line show reflectivity factors about 5 dB less. The S-band reflectivityfactors, as before, extend above the mixed layer. Forthe Colorado case the X-band return ends roughly500 m below the top of the mixed layer and in theFlorida case it extends to the top of the mixed layer.The ZVR values decrease with height, similar to thosewithin the thin lines, but average about 1-2 dB less atlower levels. An examination of numerous cases from Florida andColorado showed that the examples in Fig. 10 are representative of most situations. That is, the particulatescattering tends to extend throughout the adiabaticallymixed layer, the reflectivity decreases with height, particularly near the top of the adiabatic layer, and theZVR signals are highly positive within the mixed layer.Typically the depth of the particulate scattering layer1196 JOURNAL OF ATMOSPHERIC AND OCEANIC TECHNOLOGY VOLUME 114a) 4 AUG 91 Florida ,~! !.04' .../ ~'"~03K ..~ ,2~k,-, ....'" 7 .....'"'"' ~.. --"X~t_ ...'"'"" -10 0 10 dBZe.~ 3v~ 2 b) 6 JUL 92 Colorado 5 j ~0 K 3;o ~.~ ~-' ..."' ~ ,,, '... / /-4 ............ .... ? ?? S-ba ....?m .....'"' / .......... ..... ....... 4 // ' ~'"~'" ~0~' X-band ~ ~ ~/ ..'/o"N /"' .."" 5 / .."' ..'" .'" , ./.X .? , , ./ ?/ %/' -10 0 10 dBZe BG. 10. Ve~ical profiles of radar reflectivity hctor (dBZ,) from the CP-2 S- and X-band radarsare ~ven on the left side of the figures. The profiles am average values for an approximate50-km2 area across convergence lines. The numbers listed along the S-band curves are mediandifferential radar reflectivity values (dB) for the areas of compahson. Proximity soundings oftemperature and dewpoint are ~ven on the h~t side of the figure. A near-surface lifted parcel isshown. (a) Data are for a re, on along the ~ofida east coast sea breeze on 4 August 1991. (b)Data are for a re, on along a convergence line in Colorado on 6 July 1992.is a good indication of the depth of the adiabatic layer.Dramatic exceptions to this were noted on two daysin Colorado. For these cases the particulate scatteringextended only through the lower third of a 3-km adiabatic layer. For both cases there were high clouds obscuring the sun and horizontal convective rolls werenot apparent. This suggests that thermals were weakand not able to provide sufficient lift to carry the insectsto the top of the adiabatic layer. The above observationsof the vertical decrease in clear-air reflectivity factorand suggestion that the insects are passively carriedvertically by thermals is consistent with actual insectbehavior reported by Johnson ( 1957a and 1957b) andIsard et al. (1990), who report that the density of insectsduring the day decreases continuously with height,which can be described by an inverse power law, andthat densities are correlated with temperature lapse rate. Since Bragg scattering was not observed beyondabout 10 km in range for the X band it was often quiteuseful for identifying the top of the adiabatically mixedlayer. This was not true for the C band and S bandsince Bragg scattering was often observed above themixed layer to 30-40 km at C band and further at Sband.5. Accuracy of kinematic features from Doppler velocities Results from the previous sections have strongly suggested that the boundary layer clear-air return is generallyfrom insects. Since they have velocities of their own, theOCTOBER1994 WILSON ET AL. 1197question now arises as to what extent Doppler velocitiesfrom these echoes can be used to determine the wind 4.5field. For the bird and insect velocities not to bias theDoppler velocities they must either be carried passivelyby the wind or they must be flying randomly, so that the ~.integrated velocity of all insects and birds in the beam is cazero relative to the wind. ~Mueller and Larkin (1985) and Achtemeier (1991) Ehave shown examples where the observed clear-air x,~ 3.0Doppler horizontal velocities are not representative of ~ EBthe actual wind when migrating insects are present. Mi- .$grations of both birds (Kerlinger and Moore 1989) and -rlarge insects (Drake and Farrow 1989) occur mostly atnight. In contrast, migrations of small insects are primarilydiurnal and downwind (Drake and Farrow 1989). The 1.5remainder of this section examines the accuracy of horizontal and vertical winds obtained from the Dopplerradar measurements in optically clear air during summerdaylight hours in Colorado and Florida.a. Horizontal winds The utility of using Doppler velocities to determinehorizontal winds depends on whether the biologicaltargets are flying in a preferred horizontal direction orin a random manner. This was examined with boththe Florida and Colorado data during summer daylighthours; night observations were not made. Horizontal winds were determined from both theVAD technique (Browning and Wexler 1968) anddual-Doppler synthesis. Figure 11 is from Coloradoand compares horizontal winds from the VAD technique with those taken by a Cross-chain Loran Atmospheric Sounding System (CLASS) released fromthe radar site at the same time. The VAD uses the 7-elevation angle from a radar range of 1-24 km. Thetwo techniques show close agreement in both wind direction and wind speed. Figures 12 and 13 are fromFlorida and also include dual-Doppler winds derivedfrom CP-3 and CP-4. The dual-Doppler winds weredetermined for a 100-km2 region centered over theballoon-sounding release location. The maximum,minimum, and median Doppler velocity winds for thesquare are given in Figs. 12 and 13. While there isgeneral agreement between the radar and soundingwinds there are regions, particularly for wind speed,where the sounding winds are considerably different.In particular, note between about 0.9 and 1.3 km inFig. 13b where differences approach 4 m s-~. A VADfrom CP-4, however, showed very close agreement withthe dual-Doppler winds in this region. It is apparentfrom the differences in the VAD winds from radar toradar (not shown) and between the maximum andminimum winds obtained by the dual-Doppler synthesis over a 10-km square that there is considerablenatural variability in wind velocity over small distances.This is also apparent in wind velocity data that wereobtained from two instrumented research aircraft dura) Wind Direction90 180 270Wind Direction (deg)3604.5b) Wind Speed-- VAD...... SOUNDING0 8 . - I , I i I , , I I2 4 6Wind Speed (m/s) FIG. I 1. Comparison of horizontal (a) wind direction and (b)wind speed profiles obtained from single-Doppler radar using theVAD technique and from a balloon sounding taken at the same timereleased from the radar site. The data are from Colorado at 2000UTC 19 June 1992.ing CAPE. Thus direct comparison between techniquesis difficult and an extensive statistical comparisonwould be required to precisely determine how well theballoon and Doppler radar winds compare. However,based on this limited analysis, the Doppler and balloonwinds appear to agree within limits expected from naturally occurring spatial variations in the wind. The question of bias in horizontal winds determinedby Doppler velocities from insect or bird migration isaddressed through the use of ZDR data from CP-2. Ifthe insects and birds were flying in a particular directionthey would have a preferred alignment that would resultin a change in Zr~R with viewing angle (Achtemeier1991 ). Many scans of ZDR were examined from numerous days from Colorado and Florida and no ob1198 JOURNAL OF ATMOSPHERIC AND OCEANIC TECHNOLOGY VOLUME I11.8"-" 1.2ET 0.6a) Wind Direction-- VAD---- DDOP AVERAGES--- ODOP MAX& MIN.... SOUNDING0.0 i 0 90 360 ' '" I,~' '\ '~ t / \ ,/ ' ~ / t ', 180 270Wind Direction (deg) 'b) Wind Speed 1.15 ' ' ' ' ....... ' ' ' ' ' i \,\ ./---'"' i._~z 0.6 . ' ' <~ ,/-- VAD '"" '~ i---- DDOP AVERAGES j ', ~ :--- DDOP MAX & MINJ i './I i---SOUND..G 0.0 .... "/ ....... , , , , ~ , , , - 0 2 4 6 8 10 Wind Speed (m/s) FIG. 12. Comparison of horizontal (a) wind direction and (b)wind speed profiles obtained from the VAD technique, dual-Doppler(DDOP) synthesis, and balloon sounding. The dual-Doppler windsare for a 100-km2 region centered over the balloon release location.The average, maximum, and minimum dual-Doppler winds withinthe square are given for each height. The VAD is for CP-4. The datawere collected in Florida at 1700 UTC 4 August 1991.vious systematic variation in ZDR with viewing anglewas observed. The field itself is highly variable fromradar gate to radar gate, but the medium does not varysignificantly with viewing angle. This is illustrated inFig. 14 for typical cases from Florida and Colorado.Figure 14 shows representative curves of median Zmvalues as a function of azimuth for a Florida and Colorado case. The Florida data are median values for 10-azimuth sectors at a range between 10 and 20 km foran elevation angle of 2.2-. The Colorado data are for10- azimuth sectors at a range of 5-10 km at an elevation angle of 4.2-. There are azimuthal interruptionsin the data because of ground clutter from nearbymountains west of the radar in Colorado and becauseof the absence of echo northeast of the radar in Floridadue to the close proximity of the ocean. The medianvalue does not vary more than 1-2 dB and there is riosystematic variation with azimuth. If the insects or binlshad a common heading, a maximum and minimumin ZDR values would occur at 90- intervals. The only systematic difference observed was thatZD~ values were about 1-2 dB greater in many th!inlines and the values looking down thin lines were aboutI dB greater than looking across thin lines. Theobservations, at least during summer daylight hours,suggest that the insects dominating the radar returnFlorida and Colorado are generally in random flightand thus the Doppler velocities should closely representthe wind. However, the likelihood of bird and insect.migrations increases significantly at night, particularlyduring the spring and fall.b. Vertical velocity 1 ) COMPARISON WITH THIN LINES The observations of thin lines of enhanced echo associated with boundary layer convergence lines suggest.E ~v~.~t C~.~Q~T2.01.51.00.50.0 0a) Wind Direction -- VAD ," ---- DDOP AVERAGES I -- - DDOP MAX & MIN I , ---- SOUNDING i t I I,I !:[ \ ~ , I ~ I ~ 90 180 270 360 Wind Direction (deg) b) Wind Speed 2.0 ......... VA~DDOP ^VERAGrS'.''''' '.~_~:'' ' ' .... ODOP ~ & MIN',. ~ SOUNDING \ ' \ .~ , 1.5 ' ". x \ \, ~-- . .....'\ /,/' :'~'"' .':i:' 1.0 j ~(,.....?')"':. .... 0.5 ,/ :', / ",,,'\ 0.0 , I/~/'--"."--x ...... 0 2 4 6 8 Wind Speed (m/s)FIG. 13. Same as Fig. 12 except for 1700 UTC 6 August 1991 and the VAD is from CP-3.OCTOBER 1994 WILSON ET AL. 119910..... Florida0 90 180 270 360 Azimuth (deg)FIG. 14. Typical example of the variation of median ZDR valuesas a function of radar azimuth for Florida and Colorado.that the insects do resist being carried upward by thewinds. If they were carried passively by the wind, theywould be carried upward in the convergence regionsat the rate at which they were being brought in horizontally and their density would not increase as required by mass continuity considerations, similar tothose for air. Achtemeier (1991) has presented dataindicating that grasshoppers resist being carried tocolder heights by flying against the updraft or by foldingtheir wings and dropping, thus increasing the low-leveldensity of grasshoppers within updraft regions. The density of insects should increase in an updraft region if the insects actively fly downwardagainst the updraft or if they become too cold tomaintain wing beating and settle at their terminalfall speed. The increase in insect density will be dependent upon the relative magnitudes of insect vertical velocities to those of the boundary layer verticalair motions. Wellington ( 1945 ) has reported on thefall speeds of numerous insects varying in mass from1 to 24 rag. The range of fall speeds varied between0.4 and 2.3 m s-l, with the larger insects falling thefastest. Thomas et al. (1977) have extensively studied aphids with an average mass of approximately0.5 mg and show corroboratory data of fall speedsof 1.8 m s-1 with wings folded and 0.8 m s-~ withwings open. He also reported that they can activelyfly downward at speeds of 0.7 m s-~. It is not knownwhat type of insects are the primary radar targetsbut aphids are likely on the small side of the distribution. Nevertheless it would appear that most insects, whether actively flying downward or fallingbecause of immobilization from cold, will have vertical velocities similar to updraft velocities in theboundary layer. It is likely that smaller insects willbe found at the top of the stronger updrafts. The vertical velocity of the insects has little impacton the dual-Doppler-synthesized velocity fields sincethe vertical velocities are based on the horizontal divergence fields from the dual-Doppler horizontal winds.At the low-elevation angles used (<9-) the contributionof the insects' vertical velocity to the Doppler radialvelocities measured by the radar will be less than 10%.This is based on assuming a very large insect verticalvelocity of 3 m s-~ and an average horizontal wind of5ms-~' The above arguments indicate that thin linesshould be regions of updrafts. Figure 15 shows vertical velocities obtained from a dual-Doppler synthesis overlaid upon radar reflectivity factor. Thisexample is from Florida showing thin lines associated with the sea-breeze front and between pairsof oppositely rotating horizontal convective rolls.The vertical velocity field shown in Fig. 15 is basedupon a dual-Doppler analysis of the. CP-3 and CP4 data as described in section 3. There is a markedtendency for the updraft regions (white contours)to be located over the enhanced reflectivity regions(green colors) indicating that these two independentfields are well correlated. A gridpoint-by-gridpointcomparison of the retlectivity and vertical velocityvalues of the data west of the sea-breeze front andnot including the far south west corner in Fig. 15gives a correlation coefficient of 0.67. Similar comparisons for the example shown in Fig. 18 (to bediscussed later) is 0.66. Thus, enhanced reflectivityregions are generally associated with regions of updraft but the magnitude of the reflectivity is onlymoderately correlated with the magnitude of thevertical motion. At radar ranges of about 40 km,the agreement between updrafts and thin lines decreases rapidly. For example, note the poor agreement between the thin lines and vertical velocity inthe southwest corner of Fig. 15. This is primarilybecause the radar beam height increases with range,and near 40 km it no longer adequately samples thelow-level divergence-convergence fields. Thus thedual-Doppler vertical motion fields, which are determined from integrating the divergence field, arelikely to be in error. 2) COMPARISON WITH CLOUD LOCATIONS While Fig. 15 supplies considerable evidence thatthe general pattern of updrafts and downdrafts is correct, the question arises about the accuracy of the detailsin the vertical velocity field. For example, along thesea-breeze front, which runs from roughly (x, y)= (-17, 2) to (x, y) = (-2, -22), there are a numberof updraft centers of 0.8 m s-~ or greater. Figure 16 isan enlargement of this region with the addition of cloudlocations. The cloud locations are based on radarechoes above the lifting condensation level with reflectivities between - 10 and 0 dBZe. These echoes were1200 JOURNAL OF ATMOSPHERIC AND OCEANIC TECHNOLOGY VOLUME 11 WEST OF CP-4 (krn) FIG. 15. Horizontal cross section of the CP-3 radar reflectivity field (color in dBZe is given by scale along the right side) at 0.7 km AGLat 1715 UTC 6 August 1991 during CAPE. Positive vertical velocities derived from dual-Doppler analyses are contoured every 0.4 m s-~starting at zero. CP-4 is at the grid origin. Note that the updrafts correlate well with the enhanced reflectivity thin lines delineating horizontalconvective rolls and the east coast sea-breeze front.also confirmed by photogrammetry to be cumulusclouds. The clouds, as seen from two photo sites, areshown in Fig. 17. "Photo 4" was taken from CP-4 at(x, y) = (0, 0) and "Photo South" was taken from (x,y) = ( 15.8, -9.7). For cross reference between Figs.16 and 17, the clouds are numbered from north tosouth along the sea-breeze front. Cloud 0 in Fig. 16was only visible from the photo site at CP-3 (notshown); this cloud was not associated with a definitiveecho. As can be seen in Fig. 16 the clouds are closelyassociated with updraft maxima. The fact that themaximum updrafts are associated with cloud locationsprovides evidence that the dual-Doppler analysis is ableto identify updrafts with diameters as small as about1 km (one-half wavelength). 3) COMPARISON WITH AIRCRAFT MEASUREMENTS The accuracy of the vertical velocities derived fromthe dual-Doppler synthesis was examined by comparing them with the vertical velocities measured by ~heWyoming King Air research aircraft. Figure 18 showsthe location of a flight leg (labeled AA ') superimpo~;edupon a radar reflectivity image. Upward vertical velocities from the dual-Doppler synthesis are also shownOCTOBER 1994 WILSON ET AL. 1201 5 '~ i i O-5 m-10 --15-20-25 I I -20 -15II I 6 August 19911715 UTCW @ 0.7 km AGLCloud Locations-5 X (km east of CP4) DO. 16. Expanded horizontal cross section of the sea-breeze verticalvelocity field shown in Fig. 15. Contours are every 0.2 m s-~ startingat 0.8 m s-'. Local maxima tend to correlate with the shaded regionsthat represent CP-4 radar refiectivity factor greater than -10 dBZeabove 1.2 km AGL. These echoes correspond with cloud locationsas determined from photogrammet~y. The clouds are numbered fromnorth to south for cross referencing with the clouds in Fig. 17. Cloud0 was not associated with an obvious echo but was identified byphotogrammetry from the photo site at CP-3.in Fig. 18 by the white contours. Similar to Fig. 15,there is excellent agreement between thin-line echoesand dual-Doppler-derived updrafts. It can be seen thatthe flight leg crosses five thin-line echoes. The dualDoppler analysis is for a height of 700 m and the.aircraftflight altitude was 547 m. The time of this comparisonis 40 min after Fig. 15; there was no aircraft data available at the time of Fig. 15. Figure 19a compares the aircraft-measured verticalvelocities with the dual-Doppler vertical velocities alongthe flight leg marked AA' in Fig. 18. For the aircraft,both the original 1-s vertical velocity data (short dashes)and filtered vertical velocity data (long dashes) areshown. A fast Fourier transform (FFT) filter was usedto remove wavelengths less than 1910 m. A 1910-mfilter width for the aircraft-observed vertical velocitieswas chosen because it provided the best agreement withthe dual-Doppler vertical velocities (thick solid line). The rationale for why this scale is optimum is explored below. Examination of the unfiltered aircraftdata shows two dominant scales: the first less thanabout 1 km and the second is 3.7 km associated withthe horizontal convective rolls. The radar data are toocoarse to resolve the small-scale features in the measured aircraft vertical velocities; in fact a filter was applied to the radar data to remove wavelengths less than1.2 km. As a first approximation it would be expectedthat the aircraft data processed with the 1910-m filterwould resolve approximately the same scales as theradar data. Theoretically, as discussed in section 2, theradar data on a 300-m grid, could resolve about 80%of the magnitude of a 2000-m wave (Carbone et al.1985). However, there are secondary considerationsthat make it impossible to perfectly match scales between the two sets of observations. Complicating factorsinclude: (a) the radar half-power beamwidth is about500 m at the average 17-km range along the flight pathso that data on the 300-m grid are not completely independent as is assumed for the argument that 300-mgridpoint data can resolve 80% ofa 2000-m wave, (b)the vertical velocities are obtained from finite-differencederivatives of the horizontal winds on the 300-m gridthat lead to further smoothing, (c) due to the shallowdepth of the boundary layer ( ~ 1 km) only a very fewindependent observations are available in the verticaldirection, and (d) the radar data are smoothed in threedimensions, while the filtered aircraft data aresmoothed in only one dimension. Figure 19a showsthat, in general, there is remarkably good agreementbetween the filtered aircraft vertical velocity data andthe dual-Doppler-estimated values (coefficient of correlation equals 0.79). The radar data appear to adequately detect vertical velocity features present withthe horizontal rolls and the sea-breeze front but underestimate the maximum vertical velocities. Figure 19b shows the radar reflectivity factor fromCP-4 (thin solid line) and the radar Vertical velocities(thick solid line). The correlation coefficient is 0.71,which is similar to the 0.67 value in Fig. 15. The above data show that dual-Doppler synthesistechniques can be used within about 30-40 km of theradar to detect small-scale motions on the scale of thehorizontal convective rolls and sea breeze when radarsare placed 23 km apart.6. Conclusions Boundary layer clear-air radar echoes are a commonoccurrence over land areas from spring through falland were even observed in Kansas during winter whenthe temperature was above 10-C. Clear-air echo overwater is rare. These observations of clear-air echo areconsistent with the activities of birds and insects assummarized by Vaughn (1985), who stated that,"From spring through fall, birds and/or insects aregenerally common to abundant in the atmosphere to'1202 OCEANIC TECHNOL-OGY VOLUME 11'.Photo South . 17'15:00.'UTC'JOURNAL OF ATMOSPHERIC AND 6 August 1991 ~ii~l~ ~ ~ ~-,~ ~:,~r~ ~ ~ ~ g~;~'.t-::.~: ~ ~::~,~g~ ~g!~g g:~tg~ g g~ ~ ~=~ . g ~g~,:~::~,~g'~: ::~: :~ :,g ~ ~g~ g~;~ ~ ~:: ,~: ~~;~ ~~~~~ ; ~~ [5 ~ ~ [~[[[ [ = ~ ~=~ ~. '~ ~ :~ ~~5-~:~dg.-g g;g)~ m *~&~ g!~< ~,~;~g~:~' ~ ~,~,~.250- Photo 4 ~ ~ ~'*~:~ ~ ~ ~ .... ~ ,~i~, i ~ ~;~~ ~ ~.~,~'~ ~?~,~ ~ ~ '~ ~ ~ ~ .......~ ~'~'~~~ ~ ~.~~~~~~~~ ~. ~.~ ~ -~ ~ ~ ~ ~ ~.~ ~ ~,::?:,~t,~g:~, ~?~,~ ~ ~g%~ ~gS~ ~ -.~ '~t;~: ~?~*~ ~t:;~-'~:~g '~*~'~ ~ t~ ~?~;~,: ~ .:~,t~ ,~::~'~?~ ~,~%~ ~,~-~?~t :~:? :?~%~%~:~g ;~,~ g t%'~?~*~:~*t ~t~::,~r~$~-~:~,~ g ~ t.-t~:t~ ~ '~ ~?~:~; ~i~t~?~ 5 ~- ~*~ ~ :, ~ C~g~:~g ~:~ ~t~,~%,~:~5~/~,~g~,~:[~gt:~g:~: :~:~:': g~*:~:~: ~ ~:~: ~ g~-t- _ ~ .,~ ~, ..~:~;~,, ~ ....... ~-~ ~.~ .....~ ~.~, ~~~~~~~~~~~~~~~ , ,,~ ~:, ~:~%~.;,~;~,, : ~, 170- 180- 190- 200- 210e ~20- 230* ~40o 250- FIG. 17. Cloud photographs from two photo sites: "Photo South" located at (x, y) = (15.8, -9.7) km and '~Photo 4" located at CF-4.The pictures were taken looking west toward the sea-breeze front. The an~es represent directions from the photo sites. The numberedclouds co~espond to the echoes shown in Fig, 16,'240- 250- 260- 270- 280- 6 August 1991 !7!5':051,UTc~'an altitude of 1-2 km over most land areas of theworld?' Available data on insect densities and radar~ross sections generally agree.with observed Clear-airradar' reflectivity factors that typically range between-5 and 10 dBZ~ with maxima within convergence linesas high as 20-35 dBZ~. Dual-wavelength and dual-polarization data obtained from Colorado and Floridashowed that the boundary layer e~hoes are dominatedby particulate scatterers. Theoretical calculations thatassume insects can be modeled as prolate spheroidstogether with Observed differential radar reflectivityvalues and horizontal reflectivity differences measuredbetween S, C, and X band suggest the insects havevertical to horizontal ratios of 1:3 and equivalentspherical diameters of 10 mm. This study further support~ the conclusions of Hardy and Katz (1969) thatinsects and birds are largely responsible for the boundary layer Clear-air radar return, it is suspected that insects are the primary contributor to the widespreadnature of the echo while birds are more responsible forwidely scattered poifit targets of intense echo. Literatureon the behavior of insects as a function of absolutetemperature and vertical temperature profiles is alsoconsistent with the observed occurrence and verticalprofiles of clear-air radar echo, At least during summer daylight hours it is concluded that the Doppler velocities from the clear-airecho can generally be used to estimate boundary layerhorizontal wind velocities. During these time. periodsthere was no evidence in Colorado or Florida of preferred flight that would bias the Doppler wind velocities. This conclusion is based on the observation thatthe differential radar reflectivity did not change in aSystematic manner with 'viewing angle. If the insectsor birds were flying in mass in a particular direction,they would have a preferred orientation. This wouldcause changes in the differential radar reflectivity valuesas the radar viewing angle changed, but no such changeswere observed. Other evidence of the absence of nilgration comes from the generally good agreement observed between balloon-sounding winds and those fromthe VAD technique and dual-Doppler synthesis. Inview of the likelihood that considerable operationalUse may be made with winds from the VAD techniqaeit is recommended that an extensive evaluation 'bemade for numerous localities and times. The common observation of thin-line echoes in thedear-air boundary layer associated with convergence linesis hypothesized to result from insects actively flyingdownward to resist being carded to colder heights or infree fall from immobilization caused by:cold. This resultsi.n an increase in their density since their, flux into ~heupdraft region is greater than the flux upward. Typically,the reflectivity factors within the thin lines are 5-15 dBgreater than the background value. Monitoring the locationof these thin lines is particularly useful for forecasting convective weather events (Wilson and Mueller 1993). Most importantly, it was found that the clear-air echocould be used to identify updrafts as Small as approximately 1 km in diameter. These fields of vertical motion were obtained from integration of the anelasticmass continuity equation using dual-Doppler-synthesized horizontal wind fields. This finding was based onthree independent observations. First, there was a closeOCTOBER 1994 6 AUG i991 6.~13.5-11.5WILSON ET AL.17:57 GMT0.7 km AGL 1203-lO.O -14.tl -41].1~ -37.5 -35.0 -32.5 -30.0 -;27.5 -25.0 -22.5 -2~.~ X (km East of CP4) FIG. 18. Horizontal cross section of the CP-3 radar reflectivity field (gray shade in dBZe is given by the scale along the right side) at 0.7km AGL at 1757 UTC 6 August 1991 during CAPE. Positive vertical velocities derived from dual-Doppler analyses are contoured every 0.4m s-~ starting at zero. The thick white line labeled AA' is the path of the University of Wyoming King Air aircraft that is used in Fig. 19to compare vertical velocities. The grid is in kilometers relative to CP-4.agreement between the radar-derived updraft regionsand reflectivity thin lines. Second, localized radar updraft maxima along the Florida sea-breeze front corresponded to the location of individual cumulus clouds.Third, the radar vertical velocities agreed closely withresearch aircraft-measured vertical velocities (coefficient of correlation equal 0.79). Comparison of theradar-derived vertical velocities with the aircraft-measured vertical velocities showed that the radar was ableto detect the vertical motion pattern within the horizontal convective rolls although the magnitude of theupdrafts/downdrafts in some cases were damped byroughly a factor of two. In summary it is concluded that insects are the primary cause of boundary layer clear-air echoes. However, a confirmatory study that includes direct measurement of insects and refractive-index turbulence isencouraged. Second, provided care is taken, as wasdone in this analysis, to ensure that the insects andbirds are not migrating and that the radar data arecarefully collected and edited, kinematic details can beobtained of boundary layer wind features such as horizontal convective rolls, sea breezes, and gust fronts. Acknowledgments. The authors are most appreciative of the excellent efforts of Dan Megenhardt of1204 JOURNAL OF ATMOSPHERIC AND OCEANIC TECHNOLOGY VOLUME 11,.o a) 6 AUG 91 17:55:20 - 17:59:00 UTC 0 S 110 115 A A' Distance (km )4.0b) 6 AUG 912.Oo.o17:55:20 - 17:59:00 UTC !~-..~ .~..~:~.? .'~ .'~ .....A A'Distance ( km )-10 FI(3. 19. (a) Comparison of vertical velocities (W) measured bythe University of Wyoming King Air aircraft and synthesized dualDoppler radar data (solid thick line) along the path marked AA' inFig. 18. An FFT filter was used to remove wavelengths less than 1910m for the filtered aircraft curve (long dashes). The other aircraftcurve is the original 1 -s observations (short dashes). (b) Same as (a)except the synthesized dual-Doppler vertical velocities (thick solidline) with scale on the left are compared to radar reflectivity factor(dBZe) from CP-4 (see scale on the fight).responsible for collecting the aircraft data. CindyMueller and Jay Miller of NCAR and Nolan Atkinsof the University of California, Los Angeles, providedin-depth reviews of the paper. This work was partiallyfunded by NSF and FAA through interagency agreemerit DTFA01-90-Z-02049, as well as NSF GrantsATM 9008683 and ATM 9221951. APPENDIX Theoretical Technique for Calculating Ze for Nonspherieal Targets The transition matrix or the T matrix is used forcomputing backscattering characteristics of nonspherical targets such as spheroids and cones (Waterman1969; Barber and Yeh 1975 ). The T matrix essentiallyrelates the unknown scattered field expansion coefficients to the known incident field coefficients (or fieldtransmitted by the radar). Scattering characteristics ofraindrops, graupel, and hail particles in microwave fi:equencies are computed using the T-matrix method(Aydin and Seliga 1984; Vivekanandan et al. 1993).In this method the incident, scattered, and internalelectric fields are expanded in terms of vector sphericalharmonics functions. Vector spherical harmonic functions are composed of associated Legendre functions,sinusoidal functions, and Bessel functions (Morse andFishback 1953). The unknown scattered wave is a,btained by a surface integral equation method. The extended boundary condition method and the analyticcontinuity are used to formulate the surface integralequations. The T matrix for a particular scatterer of given shape,size, and composition are computed only once with itssymmetry axis along Z axis.~ Then to obtain averagedscattering properties of an arbitrarily oriented scatterer,the technique proposed by Wang (1979) is used. Depending on the orientation of scatterer, rotation oftlheincident and scattered wave directions, as well astation of the unit vectors describing the polarizationstates, are performed to compute the orientation-averaged scattered fields.TABLE A 1. Dielectric constants (of water at 15 -C) usedin the model for the indicated frequencies.NCAR, who prepared most of the figures and compiledthe STORM-FEST data in a manner that was easy toanalyze. Thanks to Bob Bowie of NCAR, who informed us of the thin-line echo observed during TOGACOARE and to Steve Rutledge of Colorado State University for providing permission to present these data.We are also grateful to the NCAR technicians and engineers responsible for operating the radars and theaircraft crews at NCAR and the University of WyomingFrequency Real Imaginary S band 78.93 15.51 C band 0.66 27.38 X band 57.0 35.98 2 Here backscatter characteristics ofprolate spheroidal water dropsat 15-C at S, C, and X bands were modeled. The correspondingdielectric constant at these frequencies are listed in Table A1 (Ray1972).OCTOBER 1994 WILSON ET AL. 1205REFERENCESAchtemeier, G. L., 1991: The use of insects as tracers for "clear-air" boundary-layer studies by Doppler radar. J. Atmos. Oceanic Technol., 8, 746-765.Atlas, D., 1960: Radar detection of the sea breeze. J. Meteor., 17, 244-258.Aydin, K., and T. A. Seliga, 1984: Radar polarimetric backscattering properties of conical graupel. J. Atmos. Sci., 41, 1887-1892.Barber, P., and C. Yeh, 1975: Scattering of electromagnetic waves by arbitrarily shaped dielectric bodies. Appl. Opt., 14, 2864 2872.Battan, L. J., 1973: Radar Observation of the Atmosphere. University of Chicago Press, 324 pp.Browning, K. A., and R. Wexler, 1968: A determination of kinematic properties of a wind field using Doppler radar. J. Appl. Meteor., 7, 105-113.Carbone, R. E., M. J. Carpenter, and C. D. Burghart, 1985: Doppler radar sampling limitations in convective storms. J. Atmos. Oceanic Technol., 2, 357-361. , J. W. Conway, N. A. Crook, and M. W. Moncrieff, 1990: The generation and propagation of a nocturnal squall line. Part I: Observations and implications for mesoscale predictability. Mon. Wea. Rev., 118, 26-49.Caughey, S. J., and S. G. Palmer, 1979: Some of the aspects of tur bulence structure through the depth of convective boundary layer. Quart. J. Roy. Meteor. Soc., 105, 811-827.Christian, T. W., and R. M. Wakimoto, 1989: The relationship between radar reflectivities and clouds associated with horizontal roll con vection on 8 August 1982. Mort. Wea. Rev., 117, 1530-1544.Cockbain, A. J., 1961: Low temperature thresholds for flight in aphis fabae Scop. Entomol. Exp. Appl., 4, 211-219.Crawford, A. B., 1949: Radar reflections in the lower atmosphere.Proc. IRE, 37, 404-405. Cressman, G. P., 1959: An operational objective analysis system. Mort. Wea. Rev., 87, 367-374.Doviak, R. J., and M. Berger, 1980: Turbulence and waves in the optically clear planetary boundary layer resolved by dual-Dopp ler radars. Radio Sci., 15, 297-317. , and D. S. Zrnir, 1984: Doppler Radar and Weather Obser vations. Academic Press, 458 pp. -, P. S. Ray, R. G. Stmuch, and L. J. Miller, 1976: Error estimation in wind fields derived from dual-Doppler radar measurement. J. Appl. Meteor., 15, 868-878.Drake, V. A., and R. A. Farrow, 1989: The influence of atmospheric structure and motions on insect migration. Ann. Rev. Entomol., 33, 183-210.Fletcher, S. O., 1990: Early development of weather radar during World War I1. Radar in Meteorology, D. Atlas, Ed., Amer. Me teor. Soc., 3-6.Gossard, E. E., 1990: Radar research on the atmospheric boundary layer. Radar in Meteorology, D. Atlas, Ed., Amer. Meteor. Soc., 477-527.--, and R. G. Strauch, 1983: Radar Observations of ClearAir and Cloud. Elsevier, 280 pp.Hardy, K. R., and I. Katz, 1969: Probing the atmosphere with high power, high resolution radars. Atmospheric Exploration by Re mote Probes, Vol. 2, National Academy of Science, 217-243.--, and K. S. Gage, 1990: The history of radar studies of the clear atmosphere. Radar in Meteorology, D. Atlas, Ed., Amer. Meteor. Soc., 130-142.Isard, S. C., N. E. Irwin, and S. E. Hollinger, 1990: Vertical distribution of aphids (Homoptera, Aphididae) in the planetary boundary. Environ. Entomol., 19, 1473-1484.Johnson, C. C., 1957a: The distribution of insects in the air and theempirical relation of density to height. J. Anim. Ecol., 26, 479494.--, 1957b: The vertical distribution of aphids in the air and the tem perature lapse rate. Quart. J. Roy. Meteor. Soc., 83, 194-201.Katz, I., and P. J. Harney, 1990: Radar meteorology at Radiation Laboratory, MIT, 1941-1947. Radar in Meteorology, D. Atlas, Ed., Amer. Meteor. Soc., 16-21.Keenan, T., R. J. Potts, and J. Wilson, 1991: The Darwin area forecast experiment: Description and preliminary results. A ust. Meteor. Mag., 39, 211-222.Kerlinger, P., and F. R. Moore, 1989: Atmospheric structure and avian migration. Current Ornithology, Vol. 6, D. M. Power, Ed., Plenum Press, 106-142.Klazura, G. E., 1993: Description of the initial set of analysis products available from the NEXRAD WSR-88D system. Bull. Amer. Meteor. $oc., 74, 1293-1321.Knight, C. A., and L. J. Miller, 1993: First radar echoes from cumulus clouds. Bull. Amer. Meteor. $oc., 74, 179-188.Larkin, R. P., 1991: Sensitivity of NEXRAD algorithms to echoes from birds and insects. Preprints, 25th Int. Conf on Radar Me teorology, Paris, France, Amer. Meteor. Soc., 203-205.Leise, J. A., 1982: A multidimensional scale-telescoped filter and data extension package. NOAA Tech. Memo. ERL WPL-82, 19 pp. [Available from NOAA, 325 Broadway, Boulder, CO 80303.]Mahoney, W. P., III, 1988: Gust front characteristics and the kine matics associated with thunderstorm outflows. Mon. Wea. Rev., 116, 1474-1491.Mie, G., 1908: Beitr/ige zur Optik tr/iber Medien, speziell kolloidaler Metalltsungen [Contribution to the optics of suspended media, specifically colloidal metal suspensions]. Ann. Phys., 25, 377 445.Miller, L. J., and R. G. Strauch, 1974: A dual Doppler radar methodfor the determination of wind velocities within precipitatingweather systems. Remote Sens. Environ., 3, 219-235.--, and R. A. Kropfli, 1980: The multiple Doppler radar workshop, November 1979. Part II: Experimental design and procedures. Bull. Amer. Meteor. Soc., 61, 1173-1177.Mohr, C. G., L. J. Miller, R. L. Vaughn, and H. W. Frank, 1986: The merger of mesoscale data sets into a common Cartesian format for efficient and systematic analyses. J. Atmos. Oceanic Technol., 3, 143-161.Morse, P. M., and H. Fishback, 1953: Methods of Theoretical Physics. McGraw-Hill, 1978 pp.Mueller, C. K., and R. E. Carbone, 1987: Dynamics of a thunderstorm outflow. J. Atmos. Sci., 44, 1879-1898.Mueller, E. A., and R. P. Larkin, 1985: Insects observed using dual polarization radar. J. Atmos. Oceanic Technol., 2, 49-54.Neilley, P. P., N. A. Crook, E. A. Brandes, M. Dixon, C. Kessinger, C. Mueller, R. Roberts, and J. Tuttle, 1993: RAPS-92: The real time analysis and prediction of storms--1992 program. Pre prints, 26th Int. Conf on Radar Meteorology, Norman, OK, Amer. Meteor. Soc., 135-137.O'Brien, J. J., 1970: Alternative solutions to the classical vertical velocity problem. J. Appl. Meteor., 9, 197-203.Ottersten, H., 1969: Radar backscattering from the turbulent clear atmosphere. Radio Sci., 4, 1251-1255.Oye, R., and R. E. Carbone, 1981: Interactive Doppler editing soft ware. Preprints, 20th Conf on Radar Meteorology, Boston, Amer. Meteor. Soc., 683-689.Pratte, J. F., and R. J. Keeler, 1986: Sensitivities of operational weatherradars. Preprints, 23d Radar Meteorology and Cloud PhysicsConf, Snowmass, CO, Amer. Meteor. Soc., JP333-JP336.Purdom, J. F. W., 1982: Subjective interpretation of geostationary satellite data for nowcasting. Nowcasting, K. Browning, Ed., Academic Press, 149-166.Rabin, R. M., and R. J. Doviak, 1982: Doppler radar observations of momentum flux in a cloudless convective layer boundary layer with rolls. J. Atmos. Sci., 39, 851-863.Ray, P. S., 1972: Broad-band complex refractive indices of ice and water. Appl. Opt., 11, 1836-1844.--, and K. Wagner, 1976: Multiple Doppler radar observation ofstorms. Geophys. Res. Lett., 3, 189-191.1206 JOURNAL OF ATMOSPHERIC AND OCEANIC TECHNOLOGY VOLUME 11 --, K. W. Johnson, J. J. Stephens, W. C. Bumgarner, and ~. A. Mueller, 1978: Triple Doppler observations of a convective storm. J. Appl. Meteor., 17, 1201-1212. , J. J. Stephens, and K. W. Johnson, 1979: The multiple Doppler radar workshop, November 1979. Part IV: Motion field synthesis and radar placement. Bull. Amer. Meteor. Soc., 61, 1184 1189.Riley, J. R., 1985: Radar cross sections of insects. Proc. IEEE, 73, 228-232.Taylor, L. R., 1963: Analysis of the effect of temperature on insects in flight. J. Anim. Ecol., 32, 99-117.Thomas, A. A. G., A. R. Ludlow, and J. S. Kennedy, 1977: Sinking speeds of falling and flying Aphisfabae Scopoli. Ecol. Entomol., 2, 315-326.Vaughn, C. R., 1985: Birds and insects as radar targets: A review. Proc. IEEE, 73, 205-227.Vivekanandan, J., 1986: Multiparameter radar measurement in Col orado convective storms. Ph.D. thesis, Colorado State University, 175 pp. -, R. Raghavan, and V. N. Bringi, 1993: Polarimetric radar mod eling of mixtures of precipitation particles. IEEE Trans. Geosci. Remote Sens., 31, 1017-1030.Wakimoto, R. M., and B. E. Martner, 1992: Observations of a Col orado tornado. Part II: Combined photogrammetric and Doppler analysis. Mon. Wea. Rev., 120, 522-543.--, and J. K. Lew, 1993: Observations of a Florida waterspout during CAPE. Wea. Forecasting, 8, 412-423.Waiters, K. F. A., and A. F. G. Dixon, 1984: The effect of temperature and wind on the flight activity of cereal aphids. Ann. Appl. Biol., 104, 17-26.Wang, D. Y., 1979: Light scattering by nonspherical multilayered particles. Ph.D. thesis, University of Utah, 167 pp.Waterman, P. C., 1969: Scattering by dielectric obstacles. Alta Freq., 38, 348-352.Webster, P. J., and R. Lukas, 1992: TOGA COARE: The Coupled Ocean-Atmosphere Response Experiment. Bull. Amer. Meteor. Soc., 73, 1377-1416.Wellington, W. G., 1945: Conditions governing the distribution of inse:ts in the free atmosphere. Can. Entomol., 77, 21-28.Wilczak, J. W., T. W. Christian, D. E. Wolfe, R. J. Zamora, and B. Stankov, 1992: Observations of a Colorado tornado. Part I: Me soscale environment and tornadogenesis. Mort. Wea. Rev., 120, 497-521.Wilson, J. W., and W. E. Schreiber, 1986: Initiation of convective storms at radar-observed boundary-layer convergence lines. Mort. Wea. Rev., 114, 2516-2536. , and C. K. Mueller, 1993: Nowcasts of thunderstorm initiation and evolution. Wea. Forecasting, 8, 113-131. , R. D. Roberts, C. Kessinger, and J. McCarthy, 1984: Microbu:rst wind structure and evaluation of Doppler radar for airport wind shear detection. J. Appl. Meteor., 23, 898-915. , G. B. Foote, N. A. Crook, J. C. Fankhauser, C. G. Wade, J. D. Tuttle, and C. K. Mueller, 1992: The role of bounding layer convergence zones and horizontal rolls in the initiation of thunderstorms: A case study. Mon. Wea. Rev., 120, 1785-1815.