the edop radar system on the high-altitude nasa er-2 aircraft · electra] rely on primarily...

/3u07 0460 Mp 795 Friday Jul 05 09:47 AM AMS: J Tech (August 96) 0460v5291

795AUGUST 1996 H E Y M S F I E L D E T A L .

The EDOP Radar System on the High-Altitude NASA ER-2 Aircraft

GERALD M. HEYMSFIELD,* STEVEN W. BIDWELL,* I. JEFF CAYLOR,† SYED AMEEN,† SHAUN NICHOLSON,#

WAYNE BONCYK,@ LEE MILLER,& DOUG VANDEMARK,** PAUL E. RACETTE,* AND LOUIS R. DOD††

*Goddard Space Flight Center, Greenbelt, Maryland†Science Systems and Applications, Inc., Lanham, Maryland

#TRW, Redondo Beach, California@USGS, Eros Data Center, Sioux Falls, South Dakota

&Clemson University, Clemson, South Carolina**Wallops Flight Facility, Wallops Island, Virginia

††Swales, Inc., Beltsville, Maryland

(Manuscript received 1 November 1995, in final form 16 February 1996)

ABSTRACT

The NASA ER-2 high-altitude (20 km) aircraft that emulates a satellite view of precipitation systems carriesa variety of passive and active (lidar) remote sensing instruments. A new Doppler weather radar system at Xband (9.6 GHz) called the ER-2 Doppler radar (EDOP) has been developed and flown on the ER-2 aircraft.EDOP is a fully coherent Doppler weather radar with fixed nadir and forward pointing (337 off nadir) beamsthat map out Doppler winds and reflectivities in the vertical plane along the aircraft motion vector. Dopplerwinds from the two beams can be used to derive vertical and along-track air motions. In addition, the forwardbeam provides linear depolarization measurements that are useful in discriminating microphysical characteristicsof the precipitation. This paper deals with a general description of the EDOP instrument including the measure-ment concept, the system configuration and hardware, and recently obtained data examples from the instrument.The combined remote sensing package on the ER-2, along with EDOP, provides a unique platform for simulatingspaceborne remote sensing of precipitation.

1. Introduction

Airborne weather radar systems have played an im-portant role in studying mesoscale convective systems(MCS) and other mesoscale and cloud-scale phenom-ena in recent years. These radars have provided an im-portant tool to help understand kinematic and dynam-ical aspects of MCSs, such as the importance of therear inflow jet, mesoscale up- and downdrafts, the sus-tinence of anvil precipitation, etc. (e.g., Smull andHouze 1987; Jorgensen et al. 1991). MCSs often havelong lifetimes (12–24 h), cover large areas (severalhundred kilometers) , and advect considerable dis-tances over their lifetime. As a result, ground-basedradars may not be suitably located for high-resolutionmeasurements of the vertical and horizontal structureof MCSs due to large radar slant ranges or from atten-uation affects; or the MCSs are located over openocean, which precludes ground-based radars. All air-borne radar systems have, however, limited flight en-durance that is sometimes short relative to MCS life-times.

Corresponding author address: Dr. Gerald M. Heymsfield, NASA/GSFC, Code 912, Greenbelt, MD 20771.

Airborne Doppler radars have been previously op-erated on low-altitude (Ç6 km maximum) aircraft andmedium altitude (Ç12 km maximum) aircraft. Themost recent low-altitude (Ç6 km) airborne Dopplersystems at X band (9–9.5 GHz) perform scanningabout the aircraft longitudinal axis with looks 30–407forward and aft of the fuselage-normal plane to providea quasi-dual-Doppler measurement of the winds (e.g.,Jorgensen et al. 1994; Hildebrand et al. 1994). Thelow-altitude turboprop aircraft [National Oceanic andAtmospheric Administration (NOAA) WP-3D, andNational Center for Atmospheric Research (NCAR)Electra] rely on primarily side-looking views by theradar and do not penetrate large updrafts and/or highreflectivity regions because of safety. The ER-2 Dopp-ler radar is intended to fill some of this gap by over-flying intense convection.

One of the most difficult measurements with Dopplerradars has been the vertical air velocity w . For ground-based and low-altitude airborne radars, w is calculatedindirectly from upward or downward integration of themass continuity equation using horizontal winds de-rived from horizontal divergence estimates (e.g., Ray1990; Carbone et al. 1985). The problem is there arefour unknowns (Cartesian wind components u , £, w ,and hydrometeor fallspeed £t) and three knowns (tworadial velocity vectors taken from different viewing lo-


796 VOLUME 13J O U R N A L O F A T M O S P H E R I C A N D O C E A N I C T E C H N O L O G Y

cations and a fallspeed estimate typically based on theradar reflectivity) . The horizontal winds are calculatedby geometrically combining the two Doppler windswith £t removed, from multiple ground-based radars ordifferent airborne views (forward and aft beams or or-thogonal flight legs) at a common point in the three-dimensional region being analyzed. Estimation of w re-quires integration of the anelastic mass continuity equa-tion with an upper or lower w boundary condition.Errors in the derived horizontal and vertical winds canresult from poor spatial resolution of the raw Dopplermeasurements used to compute mass divergence, in-adequate estimates for the top and bottom boundaryvalues used for w and mass divergence, inaccuracy ofthe fallspeed estimates, etc. The low-altitude airbornesystems provide somewhat higher resolution measure-ments than the ground-based systems, but the airbornemeasurements are subject to additional uncertainty dueto the aircraft platform motions in a turbulent atmo-sphere. Vertical motions have also been calculated ‘‘di-rectly’’ with NOAA WP-3D tail radar measurementsusing zenith and nadir azimuths to produce time–height images of reflectivity and vertical motions(Marks and Houze 1987). These time–height imagesare readily interpreted as distance–height cross sec-tions using the aircraft ground speed to convert the timeaxis to spatial distance.

The NCAR ELDORA system with two beams about18.57 forward and aft of the plane normal to the fu-selage and with a relatively high scan rate, can mapout horizontal winds in one linear flight track (Hil-debrand et al. 1994) . Flight plans are thus simplified,and spatial resolution is improved over the previoussingle-look P3 measurements that required orthogonalflight tracks; recently, the P3 has implemented dualantennas but with a single transmitter, which resultsin reduced along-track resolution (Jorgensen et al.1994) . Vertical velocities with ELDORA and the newP3 configuration are calculated in a manner similar toprevious airborne and ground-based systems, exceptthat the higher density of measurements and better co-incidence in time provide less uncertainty in the de-rived winds. Using two aircraft simultaneously, called‘‘quad Doppler,’’ can reduce errors in w further sincethe system of equations is overdetermined and it is notnecessary to provide fallspeed estimates (Jorgensen etal. 1994) .

Spaceborne precipitation radar systems in devel-opment have the capability to provide more directmeasurements of precipitation structure than the cur-rent passive microwave radiometric instruments.Spaceborne radars have monumental problems interms of spatial and/or Doppler velocity resolutiondue to the attendent high ground-track velocities. Thefirst spaceborne radar for precipitation measurementswill be the Ku-band (14 GHz) precipitation radar(PR) on the Tropical Rainfall Measuring Mission(TRMM) (Simpson et al. 1988) . Two airborne down-

looking radar systems that simulate spaceborne rainmeasurements, the Communication Research Labo-ratory (CRL) radar and Airborne Rain Mapping Radar(ARMAR), have been flown on the NASA DC-8 me-dium altitude aircraft. The CRL dual-frequency (10.5and 35 GHz) incoherent radar-radiometer was confi-gured for nadir-pointing DC-8 measurements in sup-port of spaceborne radar algorithms (Kumagai et al.1993) . Rain estimates from the CRL radar have beenevaluated along with microwave radiometric mea-surements (Wang et al. 1994) . The ARMAR thattransmits at 13.8 GHz to simulate TRMM, is cross-track scanning and has Doppler and polarizationmodes (Durden et al. 1994) . The ARMAR can pro-vide direct measures of vertical hydrometeor motions,but along- and cross-track winds cannot be obtainedas for the low-altitude aircraft since multiple looks arenot obtained from the same region. While the mediumaltitude of the DC-8 is adequate to overfly many pre-cipitation systems, there are often deep Midwest andtropical MCSs that cannot be safely penetrated by thisaircraft during a single flight track. Convectiveregions of deep MCSs often have cloud tops that ex-tend above 16 km, convective cores with reflectivitieslarger than 50 dBZ , and peak updraft vertical veloc-ities in excess of 15 m s01 in the Tropics and 30 m s01

in midlatitudes (e.g., Cotton and Anthes 1989) .A new dual-beam (nonscanning) Doppler weather

radar system at X band (9.6 GHz) called the ER-2Doppler radar (EDOP) has been installed in the noseof the NASA ER-2. This aircraft flies at a nominal highaltitude of 20 km with a ground speed of about 200m s01 , has a 7-h flight duration, and a payload capacityof about 1350 kg. A downlooking radar on this high-altitude platform is desirable for several reasons. First,the ER-2 has a satellite-like perspective of MCSs andthunderstorms since it flies at an altitude virtually aboveall cloud tops, including high-level tropical thunder-storm anvils and intense Midwest thunderstorms withconvective tops overshooting into the stratosphere.Even under these conditions, the ER-2 is a very stableplatform. A downlooking radar from the ER-2 altitudedirectly measures the vertical reflectivity and windstructure from deep convective systems; this is gener-ally not possible with other low- and medium-altitudeaircraft. Because the ER-2 is relatively fast (200 m s01)and has few aircraft traffic control restrictions at itscruising altitude, the ER-2 can readily overfly largeMCSs in an uninterrupted fashion. A second importantreason for a radar on the ER-2 is the valuable intercom-parisons that can be made with the existing suite ofremote sensing instruments for precipitation, clouds,and cloud radiation. The ER-2 can be configured withscanning radiometers at visible, near-infrared, and in-frared wavelengths (0.5–13 mm), microwave frequen-cies ranging from 10 to 325 GHz, and a nadir pointingbackscatter lidar system (e.g., Kakar 1993; Griffin etal. 1994). This instrument suite has been valuable for



FIG. 1. EDOP measurement concept.

understanding various algorithms for space-based ob-servations (Adler et al. 1990; Smith et al. 1994;Heymsfield et al. 1995, etc.) . Several studies have com-pared the ER-2 passive microwave measurements withthe underlying hydrometeor structure using ground-based radars to improve the understanding of the mi-crowave measurements (e.g., Fulton and Heymsfield1991; Turk et al. 1994). EDOP-measured radar reflec-tivities coupled with the ER-2 passive microwave ra-diometeric temperatures will provide considerably bet-ter precipitation validation than is possible withground-based radars.

The current paper provides a general description ofthe EDOP measurement concept, system configuration,and hardware, as well as some examples of recentlyobtained data. Along with other ER-2 remote sensinginstruments, EDOP has several important precipitation-oriented objectives: 1) to intercompare visible throughhigh-frequency microwave passive radiometric tem-peratures with coincident radar measurements for as-sessing the limitations of various precipitation estima-tion algorithms, such as will be used in TRMM andother future spaceborne systems; 2) to better under-stand the microphysics and dynamics of deep precipi-tating systems such as differences in vertical structurebetween convective and stratiform regions; 3) to pro-vide full coverage of MCSs, and 4) to evaluate newradar measurement techniques such as, for example, theproposed dual-beam rain retrieval approach for space-borne rain measurements (Testud and Amayenc 1989).

Section 2 focuses on the EDOP measurement conceptand will provide a discussion of the calculation of airmotions from the Doppler measurements. Section 3gives an overview of the system specifications resultingfrom the science requirements and an overview of theEDOP microwave and data system hardware. Section4 presents examples of recent reflectivity and Dopplermeasurements collected from flights along the UnitedStates gulf coast.

2. Motivation for EDOP configurationa. EDOP measurement concept

The EDOP measurement concept (Fig. 1 ) makesuse of two fixed radar beams: one pointed at nadirand the other pointed at approximately 33.57 forwardof nadir. A transmit pulse with vertical linear polar-ization is sent to both antennas. Reflectivity andDoppler information are received from both the nadirand forward antennas. The nadir antenna measuresthe copolarized reflectivity Zn££ , Doppler velocity £n ,and Doppler spectral width sn , while the forwardbeam measures copolarized reflectivity Zf ££ , Dopplervelocity £f , spectral width sf , and the cross polarizedreturn Zf h£ .1 Here subscripts £, h , n , and f denote

1 Here, Z is taken here to be the equivalent radar reflectivity factornormally denoted by Ze, which is effectively the water equivalentvalue of the reflectivity factor for Rayleigh scattering since the com-plex index of refraction ÉK 2

É Å 0.93 is assumed.



vertical polarization, horizontal polarization, nadirbeam, and forward beam, respectively; ££ indicatesvertical transmit and vertical receive, and h£ indi-cates vertical transmit and horizontally polarized re-ceive. Data are collected simultaneously from bothbeams as the ER-2 flies typical linear flight segments,thus providing independent time–height sections ofthe precipitation region with one pass of the aircraft.The high-altitude viewing perspective of EDOP andits close proximity to deep thunderstorm updrafts anddowndrafts allows for relatively high-resolution ver-tical structure measurements when compared withtypical ground-based radars.

The motivation for dual antennas on EDOP is three-fold. First, the two beams can be used together to pro-vide time–height sections of vertical and along-trackair motions as will be described in more detail in thenext section. Using aircraft ground-speed, these time–height sections are readily interpreted as distance–height cross sections, which are more meaningful me-teorologically; time–height measurements are referredto as cross sections hereafter. The nadir beam £n is usedto calculate cross sections of vertical air motions in theprecipitation regions by removing aircraft motions andhydrometeor fallspeeds (similar to Marks and Houze1987). The forward beam £f can be used along with £n

to estimate the along-track wind component at eachlevel. The high-resolution two-dimensional wind andreflectivity time–height profiles provide useful infor-mation on vertical circulations in various phenomenasuch as squall lines. The EDOP system is nonscanning,and thus the wind component normal to the aircrafttrack cannot be obtained. Also, EDOP cross sectionsare along the aircraft flight track, and thus it is desirableto have coordinated flights with other airborne andground-based radars that provide a more completethree-dimensional view of the precipitation under in-terrogation. With the above approach, the absolute ac-curacy of the vertical motion is ultimately limited byaccuracies of the fallspeed estimates and aircraft mo-tions.2

Second, the dual beams on EDOP provide the op-portunity to test the ‘‘stereographic’’ approach pro-posed by Testud and Amayenc (1989) for estimationof the specific attenuation K . In their approach, a pre-cipitation cell is viewed from two angles of incidenceby a dual-beam airborne radar. Their variational ap-proach for estimating K requires reflectivity observa-tions from fore and aft views of an airborne radar. Thestereographic technique has been validated with side-

2 The surface, ‘‘direct,’’ and ‘‘mirror image’’ returns are understudy to determine their use in compensating for pitch and roll errorsand estimation of aircraft vertical velocities. Testud et al. (1995) haveused the velocity of the surface measured with airborne radar to de-termine systematic biases in the Doppler velocities due to navigationand other errors.

looking observations from the NOAA-WP3D airborneradar (Kabeche and Testud 1995), and it has been pro-posed for use in rain estimation from nadir-directedspaceborne radar (Testud and Amayenc 1989). Otherapproaches exist for estimating K , but the main advan-tage of the stereographic approach is that it provides Kwithout knowledge of drop size distributions and hy-drometeor species. The dual-beam EDOP measure-ments on the relatively stable ER-2 aircraft provide anopportune dataset to test this approach. The nadir andforward antennas on EDOP also provide informationon the path-integrated attenuation and other propertiesof the rain layer. These attenuation measurements areused in combination with the Doppler measurementsfor obtaining a more complete understanding of theinteraction between the storm microphysics and dy-namics.

Third, the forward beam that has a fixed, high-quality polarization antenna, can be used to explorepolarization techniques from the airborne platform.Polarization measurements have been made bydownlooking airborne measurements and have pro-vided useful microphysical information (Kumagai etal. 1993) . On EDOP, the linear depolarization ratio(LDR) is obtained from the ratio of the receivedpower at two orthogonal polarizations when a lin-early polarized pulse is transmitted: LDR Å 10log(Zf h£ /Zf ££ ) . LDR can provide information on par-ticle nonsphericity and orientation, even at high in-cidence angles such as with the EDOP configuration.This measurement is related to particle axial ratio andrefractive index, although for larger particles such ashail it is related to tumbling, which causes a largeparticle canting angle (e.g., Jameson 1987; Kumagaiet al. 1993) . From LDR, it is possible to infer particlephase and ice particle habit (e.g., graupel, snow,hail ) . The LDR values are very low in the rain region(õ 030 dB) , large in the wet hail regions (Ç 015dB) , and intermediate in the dry snow region(Ç 025 dB) . This microphysical information is im-portant for understanding MCSs and the interpreta-tion of the other ER-2 remote sensing measurements.

b. Relation of measured Doppler windsto air motions

Calculation of Cartesian air motion componentsfrom the measured EDOP Doppler velocities requiresremoval of the aircraft motions and the hydrometeorfallspeeds. The subject has been dealt with for nadirviewing radars (Marks and Houze 1987; Heymsfield1989) and in a more general fashion for scanning air-borne radars (Lee et al. 1994) where equations are de-rived for relating the measured Doppler velocities tovarious aircraft platform-produced and air motionterms. The general equation for the EDOP forward andnadir Doppler velocity can be obtained from simplifi-cation of Lee et al.’s Eqs. (15) and (25) as follows:



0GS(cosD cost cosR sinP 0 sinD sinR cost / cosD cosP sint )£ Å F Gr / w (sinP sint 0 cosP cost cosR)p

V plane

u(0cosH sinR cost / sinH cosR cost sinP / sinH cosP sint )/ / £(sinH sinR cost / cosH cosR cost sinP / cosH cosP sint )GF

/ (w 0 £ ) sinP sint 0 cosP cost cosR)t

Vhydrometeor

Å V / V ,plane hydrometeor (1)

where GS, D , H , P , R , and t are the aircraft groundspeed, drift angle, heading, pitch, roll, and tilt angle ofthe antennas off nadir, respectively. Also, the three Car-tesian wind components are denoted by u , £, and w ,the aircraft vertical motion wp , and the radar pulse-volume averaged hydrometeor fall speed £t with theconvention that £t ú 0 for falling hydrometeors; Vplane

and Vhydrometeor represent aircraft motion produced termsand hydrometeor motion terms, respectively. Note thatfor downlooking measurements along the aircraft trackas in Fig. 1, the azimuth angle in Lee et al.’s Eq. (15)has been fixed at 0p. For the EDOP nadir antenna tÇ 07, and for the forward antenna, t Ç 33.57.

The ER-2 can generally fly stable, straight flightlines, which is advantageous for constructing two-di-mensional (height–time) images of radar data. Roll an-gle typically varies less than 0.257 on straight flightlines, but on occasion it may have variations {17–27with periods of about 10 s near intense thunderstorms.The ER-2 angle of attack varies by about 27 during a6–7-h flight as a result of fuel burnoff, and althoughthe pitch angle is relatively constant, it can be perturbedby as much as {27 near thunderstorms. At the 20-kmER-2 altitude, the ambient winds are typically veryweak (õ5 m s01) so that drift angles are relativelysmall (õ1.57) . This relative stability of the aircraft re-duces the magnitude of the aircraft component of thecorrection to the data. However, it is still large enoughto have a profound effect on air motion calculations.The parameters GS, D , H , P , and R are provided bythe navigation system on the ER-2, and wp is calculatedin postflight analysis from vertical acceleration data us-ing a pressure altitude feedback loop similar to thatused by Scott et al. (1990).

Calculation of air motions from the measured Dopp-ler velocities first requires removal of the aircraft mo-tions (Vplane ) from (1). Equation (1) can be simplifiedfor the EDOP measurements by setting H Å p /2, sothat the u* horizontal component is defined along theaircraft heading, and D Å 0 and R Å 0 since they aretypically near zero and they produce only a small cor-

rection to Vplane in (1) . The forward and nadir Dopplerwinds can then be given by

£ Å u *(cost sinP / cosP sint )n n n n n

/ (w 0 £ )(sinP sint 0 cosP cost ) (2)t n n n n

£ Å u *(cost sinP / cosP sint )f f f f f

/ (w 0 £ )(sinP sint 0 cosP cost ) . (3)t f f f f

Here, the nadir and forward tilt angles are given by tn

Ç 07 and tf Ç 33.57. Assuming the Doppler velocitiesfrom the two beams are mapped to a common time–height grid and the aircraft three-dimensional motionshave been removed from the measured nadir £n andforward £f Doppler velocities, the approximate equa-tions for the along-track u * component and the verticalvelocity w are given by

c £ 0 c £1n rf 1f rnw Å £ / (4)t c c 0 c c2n 1f 2 f 1n

c £ 0 c £2n rf 2 f rnu * Å , (5)c c 0 c c2n 1f 2 f 1n

where c1i Å (sinPi costi / cosPi sinti ) , c2i Å (sinPi

1 sinti 0 cosPi costi ) , and i Å n and f for nadir andforward beams, respectively. Note that calculation ofthe u * component requires only £n and £f , and does notrequire £t . Vertical air motions (w) are obtained byremoving hydrometeor motions using a £t estimate froman empirical Z– £t relations for rain and snow similarto Marks and Houze (1987), where a snow, rain, andintermediate transition region is defined. While the cal-culation of w in (4) uses £n and £f , the u * term of £n in(5) is small and w can be estimated using only £n . An-other special case exists when the antenna is stabilizedat nadir (tn Ç 0P), which implies simplification to wÇ £t 0 £n . The largest errors in estimating the along-track wind component are due to 1) storm structureevolution in the time between the forward and nadirlooks, and 2) displacements between the forward andnadir beam locations due to a nonzero ER-2 drift angleor heading changes.



TABLE 1. EDOP system specifications.

TransmitterFrequency (wavelength) 9.6 GHz (3.123 cm)Peak power (nominal) 25 kWPulse width 0.25, 0.5, 1.0 ms (0.5 ms typ.)Pulse repetition frequency

(PRF) 2200, 4400 Hz (4400 Hz typ.)Traveling wave tube

amplifier (TWTA)

ReceiverIF bandwidth 60 MHzDynamic range with gain

control 110 dBMinimum detectable signal

at 10 km 05 dBZ (for 0.5-ms pulse width)Linear Doppler channels 2Log. reflectivity channels 3IF filter bandwidth 2, 8 MHz

AntennasType: offset fed parabolic

reflector using‘‘matched feed’’concept

Antenna diameter 0.76 mAntenna beamwidth 2.97Gain 36 dBFirst sidelobe level õ 026 dBCross-polarization level õ 030 dBNadir transmit/receive

polarization V/VForward transmit/receive

polarization V/V and H

Data processingA/D converters 7 channels 1 12 bits, 2 MHzSignal processors 24 1 AT&T DSP32CDoppler processing type pulse pairGate spacing 37.5, 75, 150 m (75 m typ.)Gates 436 (reflect.), 360 (Doppler)Integration cycle 0.25–1.0 s (0.5 s typ.)Products

Nadir Zn££ , £n , sn , SNRn

Forward Zf££ , Zfh£ , £f , sf , SNRf

Total system weight 180 kgTotal system input power 1500 W

3. EDOP instrument overview

a. Specifications

The science objectives for EDOP provided the ba-sis for the hardware design. Since no single radarcould satisfy all the scientific requirements, giventhe size and weight constraints of the ER-2 aircraft,a tradeoff analysis was performed prior to devel-opment of the system. Calculations were made todetermine the optimum operating frequency tocover a wide range of precipitation situations. Animportant requirement for the system was to providesurface reference measurements for calibration pur-poses. The surface also provides a means to cali-brate the antenna pointing angles that, if in error,will bias the Doppler velocities according to (1 ) .The X-band frequency ( 9.6 GHz ) and 25-kW peakpower (12.5 kW per antenna ) were found to satisfymost of the requirements, given that higher frequen-cies can have severe attenuation in high-reflectivityregions, and lower, less attenuating frequencies re-quire too large an antenna for the ER-2. The generalspecifications for EDOP are given in Table 1 andare briefly described in the following.

Pulse repetition frequency (PRF) selection is dic-tated by the maximum unambiguous velocity to be ac-commodated, the desired velocity precision (which re-quires that PRF be maximized), and the maximum de-sired slant range from the radar. A PRF of 4400 Hzresults in a Nyquist velocity (ÅPRF l/4) of 34.4 m s01

and a maximum unambiguous range (Åc /2 PRF) of34.1 km. This PRF provides the best compromise forcases where the updrafts are large such as in the mid-western United States thunderstorms, for the large £f

resulting from the 200 m s01 ER-2 horizontal transla-tion, and for sufficient range to reach the surface fromthe 20-km ER-2 altitude. The lower 2200-Hz PRF (Ny-quist velocity 17 m s01 and maximum range of 68 km)is available for cases when EDOP is used for ground-based measurements when longer slant ranges are de-sired.

The transmit pulse widths are selectable from0.25 to 1.0 ms, which results in range sampling in-tervals from 37.5 to 150 m. For most of the precip-itation measurements, vertical resolution of 75 m orbetter is desired. The received signal sensitivity isproportional to the pulse width: doubling the pulsewidth improves sensitivity by 6 dB but cuts therange resolution in half.3 The present configurationof EDOP is for a 1-ms maximum pulse width, butthe system was designed for a maximum pulse widthof 2 ms as determined by the 0.1% duty cycle of the

3 A 6-dB improvement results because there is a 3-dB increase insignal power and the hardware employs a bandpass filter matched tothe transmit pulse, which improves the noise floor by an additional3 dB.

traveling wave tube (TWT) and the high-voltagepower supply design.

A narrow antenna beamwidth is desirable for im-proved resolution of small-scale features in convectiveand other regions and for better accuracy of the Dopp-ler velocity estimates. The EDOP offset-fed antennaswith pencil beams are the largest apertures (0.76 m)and smallest beamwidths (2.97) that could be installedin the ER-2. For the nadir beam, the effective beamshape is degraded slightly from circular to oblongalong-track, due to averaging multiple radar pulses inthe process of obtaining reflectivity or velocity esti-mates. The forward beam pulse volumes are tilted by33.57, and thus the vertical structure is slightly smearedand some loss of data occurs near the surface due topartial beam filling over several gates by surface re-



turn.4 The uncertainty in the mean Doppler velocityestimates increase for larger beamwidths due to aircraftmotion, although this is not of significant magnitude tobe of concern for EDOP (see below).

Radar observations of intense thunderstorms requirea large receiver dynamic range since the entire stormcan encompass a dynamic range of greater than 80 dB.Maximum reflectivities can approach 65 dBZ , the up-per ice regions of the storm may be characterized byvalues less than 0 dBZ , and reflectivity gradients canapproach 60 dB km01 . As a result, this wide dynamicrange places stringent requirements on the receiver.The logarithmic receiver used in EDOP is capable of80-dB dynamic range as presently implemented. Hard-ware exists for a computer-controlled automatic gaincontrol (AGC) to improve the receiver dynamic rangeand prevent saturation for the strongest signals.

Estimation of the uncertainty of the EDOP reflectivityand Doppler measurements requires calculation of thenumber of independent samples integrated during a radardwell. The coherence interval, or the time to independ-ence or decorrelation time Ti , is determined either fromthe random motion of the scatterers within the radar pulsevolume, the beam broadening, and the time required forthe aircraft to move a distance equal to the spatial foot-print of the antenna. The latter aircraft translational effecthas a fairly long decorrelation time on the order of a fewseconds. The decorrelation due to reshuffling of meteo-rological targets and to beam broadening may be derived,based on the Doppler spectral width s

£given according

to Ti Å 2l/s£, where Ti is in units of milliseconds, l is

in centimeters, and s£

in meters per second (e.g., Sau-vageot 1992). For the EDOP parameters (Table 1), thiscorresponds to Ti Ç 2 ms assuming a s

£of 3.4 m s01 (s

£

is calculated below). Based on these estimates, there areapproximately 250 independent samples for a typical 0.5-s dwell.

The accuracy of the reflectivity measurements fromEDOP can be estimated as follows. For video matched-filtering, the Rayleigh fading signal statistics evolveinto a chi-squared distribution with mean PV and stan-dard deviation s(PV ) Å PV (2N)01/2 (Davenport andRoot 1987, Eq. 12-23), where N Å 250 is the numberof statistically independent samples for a 0.5-s dwell.Thus, s(PV )É 4.5% of PV and the one-sigma uncertaintyin the reflectivity data, averaged over a typical 0.5-sdwell is 10 log[1{ s(PV )]Å 0.2 dB. Note that an exactcalculation would be based on an assumed form for the

4 The vertical beam dimension V of the forward beam can be givenby VÅ R sinBW sint / L cost, where R is the range from the aircraft,t Å 33.57, L Å 75 m is the pulse length, and BW Å 2.97 is the 3-dBbeamwidth of the antenna. For a 20-km aircraft altitude, R Å 24 kmfor the forward beam resulting in V Å 620 m. The number N offorward beam range gates affected by the surface is N Å R BW tant/L Å 11 gates; thus, about 6 gates above the surface are affected bythe surface return. These calculations neglect sidelobes, which wouldincrease V and N.

signal correlation properties. This measurement repre-sents a relative uncertainty and the absolute accuracyof the reflectivity measurements are determined by theprecision of the calibration. Recent flights have indi-cated absolute calibration accuracy to better than 2 dBfrom comparison to ground-based radar (Caylor et al.1994).

The uncertainty in the Doppler velocity estimatescan be given as follows. For autocovariance processing,which is used in EDOP for Doppler velocity and spec-tral width estimation, the mean Doppler velocity mea-surement uncertainty for large signal-to-noise ratioss

V£

and narrow spectral widths is given by

s PRFl£2 __√s Å , (6)

V£

8M p

where M is the number of pulse-pairs [Eq. (6.23) inDoviak and Zrnic 1993]. The uncertainty of the meanDoppler velocity is increased for broader Doppler spec-tra under the assumption of Gaussian and non-Gaussianspectra. The total Doppler spectral width of precipita-tion return is primarily a function of beam broadeningdue to aircraft motion, wind shear, turbulence, and fallvelocity distribution. Doppler spectral broadening re-sults from the aircraft horizontal motion, which pro-duces a cross-beam wind component. The standard de-viation of this beam broadening contribution for thenadir beam is given by s

£Å 0.3BTAS, where B is the

antenna beamwidth between the 3-dB points, and TASis the true airspeed (e.g., Atlas 1964; Sauvageot 1992).

2 2s Å [s (beam broad) / s / (shear)£ £ £

2/ s ( turbulence)£

2 0.5/ s (fallspeed distribution)]£

01 2 01 2É [(3.0 m s ) / (0.5 m s )01 2 01 2 0.5/ (1.0 m s ) / (1.0 m s ) ]

01É 3.4 m s , (7)

where the beam broadening term is estimated aboveand reasonable estimates are made for the other terms.Using (6) with EDOP parameters and M Å 2200 pairsfor 0.5-s sampling dwells, the standard deviation of themean Doppler velocity is roughly 0.1 m s01 , whichs

V£

exceeds the observational requirements of EDOP. Thisfigure represents measurement precision; absolute ac-curacy, or systematic error characteristics arising fromfactors such as aircraft attitude errors are not includedin this number. Note that for well-behaved wind dis-tributions such as uniform linear wind shears across theradar beam, the spectral width is broadened and theaccuracy of the mean Doppler velocity estimate de-grades. The above estimate assumes relatively well-be-haved Doppler spectra. This assumption may be invalidfor highly non-Gaussian spectra such as when severalspectral modes occur with small-scale subbeam scale



FIG. 2. EDOP minimum detectable reflectivity for 0.25-, 0.5-, and1.0-ms pulse widths. Theoretical curves are given along with a curvebased on actual measurements.

FIG. 3. Simplified block diagram of EDOP in ER-2 nose cone.

air motions or widely disperse particle distributionssuch as associated with hail.

Finally, an important aspect of the radar system isits sensitivity to weak cloud regions, such as in iceregions near cloud top and weak rain near the surface.While there are a number of different methods to cal-culate the sensitivity of radar systems, the approachused here calculates the minimum detectable signal PV min

based on a noise power level for the system and thenumber of independent samples. After substituting theEDOP parameters (Table 1) into the radar equation asdescribed by Caylor et al. (1994), an equation is ob-tained to convert PV min to an equivalent radar reflectivityfactor. EDOP employs an IF filter bandwidth matchedto the transmit pulsewidth and therefore losses associ-ated with a matched filter (e.g., Doviak and Zrnic1993) must be accounted for in the radar equation. ThePV min is obtained from the mean noise power for EDOP,which can be given by PV n Å kT0Bn(Fn 0 1) Å kTsBn ,where k is Boltzman’s constant, T0 Å 290 K is thephysical antenna temperature, Ts is the system noisetemperature, Fn is the system noise figure, and Bn is thenoise bandwidth (Skolnik 1990). For EDOP, Fn

Å 01.8 dB, Bn Å 2 MHz, which results in Ts Å 437.6K and PV n Ç 0109 dBm. According to Atlas (1964),for greater than 10 independent samples and a 97.3%confidence limit, the minimum detectable signal can bedistinguished from the average noise at a level PV min

Å 2PV n(N 1/20 1)01Å0117 dBm. Using this PV min valuefor the 0.5-ms pulse width case in the radar equationprovides estimates of the EDOP minimum detectablereflectivity shown in Fig. 2; curves for two other pulsewidth settings (0.25 and 1.0 ms) are also shown. Thesecurves indicate reasonably good sensitivity near cloudtop and for weak rain near the surface. Actual PV min mea-

surements are about 2 dB lower than the above esti-mate.

b. RF hardware description

The EDOP system is configured for operation in arefurbished military radar nose for the ER-2. Amongthe system operational concerns are 1) severe environ-mental conditions, 2) limited pilot interaction, and 3)size and weight restrictions, impacting the electronics.Conditions within the nose of the ER-2 present a dif-ficult environment for electronic instrumentation. Inparticular, air temperatures of 0207C and partial pres-surization of 0.3 atm (9.1-km altitude) are typical ofambient conditions in the ER-2 nose. Single pilot op-eration complicates instrument design since only twoswitches and two indicator lights are available duringflight for instrument power-up, monitor, and control.The entire radar system must therefore have the capa-bility to function in a ‘‘turn key’’ mode, with internalmonitoring of any faults that may occur in the trans-mitter or data system, and the capability to reset thesystem automatically in the event that a recoverablesystem failure is detected.



FIG. 4. Layout of EDOP system in ER-2 nose cone.

A simplified block diagram of EDOP is shown inFig. 3, and the layout of the entire system in the ER-2nose is shown in Fig. 4. The RF system consists oftransmitter and receiver subsystems that are housed intwo separate enclosures. The principal system compo-nents that govern spectral purity of the Doppler signalare the reference oscillator (or system clock), thephase-locked oscillators, and the TWTA (travelingwave tube amplifier) and related power supply.

The EDOP transmitter consists of 1) two phase-locked oscillators, one at 9.66 GHz (local oscillator,LO) and the other at 60 MHz (intermediate frequency,IF), both locked to a 10-MHz reference oscillator, 2)an exciter circuit generating the 0.25–1.0 ms pulseddrive signal at the coherent transmit frequency (9.6GHz) to the power amplifier, and 3) a high gain 25-kW Litton air-cooled TWTA with output coupledthrough a power divider and cascaded Ferrite circula-tors to the antennas. The peak transmit signal is split(reducing power per channel by 3 dB) for simultaneoustransmission through the nadir and forward antennas.Both the oscillator and TWTA phase-noise levels arenegligible in comparison to the Doppler uncertainty fortypical SNR values. Forced air is used to cool theTWTA. In order to reduce weight, sulfur hexaflourideis used in lieu of transformer oil to prevent high voltagebreakdown in the transmitter high-voltage power sup-ply compartment. The remainder of the transmitter en-closure and waveguides are filled with nitrogen to pre-vent condensation during aircraft descent and to pre-vent high-power microwave breakdown within thewaveguides.

Following the received signal path, the receiver con-sists of 1) low-noise GaAs preamplifiers for the threereceived channels (nadir copolarized, forward co- andcross-polarized); 2) computer controlled RF variableattenuators providing automatic gain control (AGC);3) mixers/IF preampliers to up-convert the receivedsignal; 4) bandpass filters; 5) a splitter to divide thecopolarized returns into logarithmic and to IF linearchannels; 6) logarithmic detection and amplification on

the reflectivity channels; 7) a computer controlledAGC on the linear (Doppler) channels; and 8) in-phaseand quadrature detection on the linear channels. Sep-arate logarithmic IF channels are used for acquisitionof reflectivity data because of their inherently wide dy-namic range (80 dB) of the logarithmic IF amplifiersand the convenience of logarithmic data acquisition.Because of their inherent nonlinear characteristics, log-arithmic receivers are not suitable for Doppler pro-cessing and thus the linear channels for the nadir andforward antennas are processed separately. Thoughsoftware has not yet been implemented, the option ex-ists with the current hardware to implement a pro-grammed AGC range profile that uses a priori knowl-edge of the reflectivity profile. AGC action is based onthe average of several preceding range bins, so that the‘‘local mean’’ of the Rayleigh fading signals can beestimated.

The EDOP antenna design consists of two separateoffset-fed parabolic antennas mounted in the nose ofthe ER-2 to generate the downlooking and forward-beams (Fig. 4) . The antenna requirements for EDOPare for low sidelobes, high gain (ú35 dB), and highpolarization purity (õ030 dB isolation). The conven-tionally fed single offset reflector exhibits all of thesefeatures except for the cross-polarization performance.The short focal length required to fit the antennas inthe EDOP nose further degrades the cross-polarizationmeasurements. To overcome this disadvantage with theconventionally fed antenna, ‘‘trimode’’ feeds wereused that employ focal plane field matching based onthe ‘‘matched feed’’ concept (Rudge and Adatia1975). With this technique, excellent polarization char-acteristics are achieved. The radome which covers thelower half of the ER-2 nose is designed for 9.6-GHztransmission and is an ‘‘A’’ sandwich design consistingof a low dielectric constant center core with thin rein-forcing face sheets. Radio frequency testing of thecurved radome with the offset parabolas has demon-strated low loss (õ0.25 dB) at the EDOP transmit fre-quency.

/3u07 0460 Mp 804 Friday Jul 05 09:47 AM AMS: J Tech (August 96) 0460


The requirement to calculate air motions from theDoppler measurements by (1) has provided the moti-vation to stabilize the nadir antenna. The major correc-tion to (1) before calculating w is the terms involvingaircraft motions (Vplane ) . Heymsfield (1989) has sug-gested that horizontal winds (u , £) in (1) which are notknown a priori, can also introduce errors in w . A simpleerror analysis showed that without stabilizing the an-tennas for aircraft pitch variations, the worst case errorsin w excluding the error in the fallspeed would be about0.5 m s01 . The nadir pointing beam has thus been de-signed for pitch stabilization (i.e., P Å 0tn Ç 0),which simplifies (4) to wÅ £t0 £n . Hardware currentlyexists for stabilization of the nadir antenna, but soft-ware has not yet implemented until further experienceis obtained with the EDOP system and with a new high-speed navigation interface on the ER-2.

c. Data system hardware and processing

The data system addresses the analog-to-digital con-version of radar output, signal processing of data, ac-quisition of aircraft navigation data, storage of pro-cessed and housekeeping data, and overall control andinitialization of the data system. Details of this datasystem and the hardware implementation of the pro-cessing algorithms are described in Nicholson (1994).The on-board real-time processing provides estimatesof Doppler velocity, reflectivity, and spectral width.For Doppler processing, the dual-lag pulse-pair algo-rithm is used on the in-phase (I ) and quadrature (Q)time series data (Srivastava et al. 1979). The proces-sors perform reflectivity averaging and calculate thezeroth, first, and second lag complex autocovariancesat each gate. The dual-lag pulse-pair estimates for thevelocity and spectral width estimates are performed inpostflight using the processed autocovariances. Thereal-time processor for EDOP is designed to accom-modate the extremely high data and processing ratesrequired by the system’s 4400-Hz PRF, the four linear( i.e., two Doppler I and Q pairs) and three logarithmicreceiver channels, 37.5-m range resolution (0.25 ms) ,0.5-s dwell times. The current implementation providesprocessing all channels simultaneously for a PRF of4400 Hz, 436 gates, and 0.5-s dwells. Note that theEDOP processor requirements are somewhat higherthan those for typical ground-based systems since theEDOP PRF is about four times that of these systemsand both beams are being processed simultaneously.The processing of the mean received power for thethree reflectivity estimates is performed with samplesfrom the logarithmic receiver channels. The bias intro-duced by logarithmic integration is removed in post-processing (Zrnic 1975).

The data system is based on a VME chassis with a68030 host computer that provides basic control andinitialization of the system. The host computer queuesprocessed radar returns and aircraft navigation data to

a 2-Gbyte disk drive, which is housed in a pressurizedenclosure. The radar interface and control function ishandled by a custom-designed radar interface board(RIB). The analog-to-digital conversion of the linear(I , Q) and log (reflectivity) channels is achieved withtwo custom designed four-channel, 2 MHz per channel,12-bit acquisition boards. The signal processing is per-formed with three C programmable processing boardsdeveloped by Martin Marietta, which use eight AT&TDSP-32C chips to provide 200 Mflops of computa-tional capability per board. The digitized data from theseven channels is distributed over two high-speed buses(each 40 Mbyte s01) by the RIB to the 24 digital signalprocessing (DSP) zones on the three processor boards.The processing boards then calculate the reflectivitymeans and the first three lags of the complex autoco-variance required for the pulse-pair velocity and spec-tral width calculation. These processed values are readalong with status and navigation information by thehost computer over the VME bus and merged into asingle datastream prior to data storage.

As discussed earlier, accurate aircraft attitude andposition information is required for removal of thethree-dimensional aircraft motion vector from theDoppler measurements in postflight analysis. The ER-2 is equipped with a Litton LTN-92 inertial navigationsystem (INS) with Global Positioning System (GPS)update. The high-speed navigation data from the INSis distributed to the ER-2 instruments on the aircraftover two ARINC-429 buses. EDOP has an additionalVME board to capture high-speed (up to 64 Hz) nav-igation data for each dwell of radar data along with theprocessed radar measurements and status information.Currently, 123 parameters of navigation data are storedat the 0.5-s dwell intervals. The data system also hascustom interfaces for automatic gain control to achievelarger dynamic range in the logarithmic and linear re-ceiver channels and pitch control of the nadir antenna.These items, although existing in hardware, have notyet been implemented in software.

d. Calibration procedures

The calibration procedure for EDOP reflectivitymeasurements is described in detail in Caylor et al.(1994). Calibration of EDOP receiver involves con-verting the received power from engineering units indigitized counts to physical units (dBm). A radar con-stant incorporating fixed parameters in the radar equa-tion was calculated for each receiver channel based onEDOP parameters and measured losses in the receiverchain. Reflectivities (dBZ ) are calculated by combin-ing the radar constant, the measured received power,and the range-squared correction. EDOP calibration istypically performed in the laboratory by injectingknown RF power levels into the antenna ports prior toand after a deployment. During flight, the calibrationstability of the three RF receivers is checked at regular



intervals ( typically 30 min) using an internal calibra-tion routine that injects a continuous wave signal intothe receiver front-ends. Long-term stability of thetransmitter power also affects the overall calibration.This measurement is accomplished indirectly since thehigh power transmitted pulse, although significantly at-tenuated, leaks through the circulator and transmit–re-ceive switch into the receiver. Since the data systembegins sampling just prior to the leading edge of thetransmit pulse, the transmit pulse is sampled qualita-tively throughout the flight. Calibration data from boththe receiver and transmitter have shown stability towithin about 0.25 dB during flight after an initial 30–45-min warmup period following takeoff. Caylor et al.(1994) have shown that the EDOP calibrated reflectiv-ity agrees to within a few decibels with data from theMelbourne, Florida, WSR-88D radar. The ocean sur-face scattering cross section so has also been used tovalidate the calibration. The nadir and forward anten-nas that are beamwidth and pulse length limited, re-spectively, require different forms of the radar equationfor so calculations (Eqs. 2-39 and 2-40 in Nathanson1969). Recent flights of EDOP have provided addi-tional confidence in the receiver stability and the ab-solute reflectivity calibration.

Calibration of the antenna pointing angles is impor-tant for using both the reflectivity and Doppler mea-surements from EDOP. The EDOP antennas do notscan, and thus to estimate the mounting angles of theantennas, the ER-2 occasionally performs calibrationmanuevers during level, cloud-free flight over theocean in which the aircraft pitches up and down byapproximately 57 and subsequently rolls left and rightby about 107. These manuevers provide a means to es-timate antenna tilt (pitch) and azimuth (roll) mountingangles from variations in the surface echo Doppler ve-locity (Testud et al. 1995). In addition to these man-uevers, flight tracks during ascent and descent, wherepitch angles were greater than 57, were also examined.For the observations presented later, the forward andnadir tilt angles were estimated to be within 0.17 of33.57 and 1.47, respectively. The azimuth angle for bothnadir and forward antennas were determined to be lessthan 0.57; roll angle biases cannot be estimated moreaccurately since they produce only small variations inthe measured Doppler velocities [see (1)] . No attemptswere made to separate the antenna mounting anglesfrom absolute deviations in the INS. This will not affectcalculations using (1) since the Doppler velocity is de-pendent on the sum of the pitch and tilt angles.

4. EDOP observations

EDOP collected its first data from the Convectionand Atmospheric Moisture Experiment (CAMEX)based at Wallops Island, Virginia, during Septemberand October 1993. The ER-2 carried a full suite of ra-diometers ranging from visible to microwave wave-

lengths during this deployment. EDOP was still in adevelopment mode for CAMEX and was set up to re-cord primarily raw reflectivity measurements that wereaveraged in postflight analysis. On one of the flightsduring CAMEX, data were collected from a group ofthunderstorms on 5 October 1993 in the southern Flor-ida region. Results intercomparing the EDOP reflectiv-ity measurements along with microwave and infraredradiometric measurements for this case have been re-ported in Heymsfield et al. (1995).

A series of flights based out of Houston called Hous-ton Precipitation Experiment (HOPEX) were con-ducted during January 1995. EDOP operated with fullreflectivity and Doppler processing for the first time.During a flight on 13 January 1995, the ER-2 overflewan intense squall line with an extensive stratiform re-gion along the Gulf Coast. The squall line developedin the western part of the gulf during the early morninghours and advanced eastward to beyond Mobile, Ala-bama at the time of the ER-2 overpass (Ç1800 UTC).Some hail was reported with this squall line and therewere a number of strong surface wind gusts reportedprior to the squall-line passage, although no extensivedamage was reported with this line. Figure 5 providesreflectivity images from the nadir beam during the out-going flight leg. The data have been calibrated using aradar constant derived as in Caylor et al. (1994) butmodified for several configuration changes duringHOPEX. The ER-2 flew at 19.4-km altitude duringwhich the surface return remained roughly in the samerange gate. The aircraft was quite stable during thisflight line with pitch, roll, and drift values of 00.627{ 0.447, 00.137 { 0.047, and 00.127 { 0.917, respec-tively. Ranges greater than 19.4 km from the ER-2 thatinclude features such as ‘‘mirror image’’ returns aretruncated in Fig. 5 for presentation purposes. Also, thedata display has been thresholded to a minimum re-ceived power of 0110 dBm, which corresponds to aminimum detectable reflectivity of 1 dBZ at 10-km al-titude. Note the bottom panel (240–340-km distance)has a different reflectivity scale because the reflectivi-ties are much lower than the other panels. The figureshows a leading convective region at about 25-km dis-tance with tops extending to about 12 km, a westwardtilt, and peak reflectivities of about 55 dBZ . An exten-sive trailing stratiform region is evident between 100and 200 km with a well-defined bright band and peakreflectivities near 50 dBZ . The bright band is centeredat about 3 km close-in to the convective region, butabruptly undergoes lowering by about 500 m at 160 kmdistance. A trailing anvil ice layer (240- and 340-kmdistance) is apparent between 4- and 9-km altitude withmaximum reflectivities of about 25 dBZ near 5-km al-titude. Considerable small-scale structure is present inthe observations such as precipitation streamers 2–3km across below the melting level.

Figure 6 presents ‘‘gridded’’ nadir and forward re-flectivity and nadir Doppler velocity with aircraft mo-



FIG. 5. Nadir reflectivity Zn cross section covering full 360-km width (Ç30 min) of 13 January 1995 squallline. Convective region is located at 25-km distance (top panel), stratiform region between 100 and 200 km(middle panel), and thin trailing anvil cirrus is at 0–100-km distance (bottom panel). Distance increasestoward the east. Bottom panel has different color scale.

tions removed [i.e., Vp term in (1)] for the same squallline convective region in Fig. 5. The nadir and forwardbeam radar data have been mapped to common gridsto facilitate later wind and reflectivity analyses usingthe dual-beam information. This gridding is accom-plished in several steps as follows. First, the gate con-taining the surface peak echo is determined for everynadir and forward dwell. For each range gate, an x(along-track distance) –z (height) pair relative to the

aircraft reference frame is computed using the pitch,roll, drift, antenna tilt, and slant range information inthe track-relative coordinate transformation derived byLee et al. (1994). The z coordinate for each gate isoffset by the z value at the surface, thus giving heightabout the surface. A flat earth is assumed, which isvalid for the EDOP geometry, where the forward beamat the surface is only 13 km ahead of the subnadir point.The x coordinates are incremented by the distance trav-



FIG. 6. Convective region cross sections from the squall line shown in Fig. 5 for nadir hydrometeor motionsVhydrometeor (top panel), nadir reflectivity Zn (middle panel), and forward reflectivity Zf (bottom panel). TheDoppler velocities have aircraft motions removed and represent hydrometeor motions with positive velocitiesdownward. The forward reflectivity has been remapped into its true projection using aircraft attitude infor-mation.

eled from some reference time such as the beginningof the flight line (i.e., the product of the elapsed timeand the ER-2 ground speed). Finally, the forward andnadir data are interpolated onto a regularly spaced rect-angular grid using linear interpolation.

The nadir and forward reflectivity panels in Fig. 6are similar qualitatively. Prominant features such as the

bright band and high reflectivity cores in the convectiveregion are reasonably similar between the two beams.Major differences between the two fields are primarilyattributed to 1) changes and/or advection in the stormstructure in the interval between forward and nadir ob-servation, and 2) attenuation along the propagationpath. In the convective regions of the squall line (20-



km distance) , there is up to 17 dB of attenuation onthe nadir channel and 12 dB of attenuation on the for-ward (Caylor et al. 1995). The attenuation on the twopaths accounts for the differences in structure in theregions of high precipitation. A few additional differ-ences between the forward and nadir reflectivity imagesare also apparent. The forward beam is more smearedout in the vertical due to the 33.57 tilt of the radar pulsevolumes (see footnote 4). The ocean surface is evidentin images, but for the forward beam, the high reflectiv-ity surface return covers roughly 15 gates total in low-reflectivity regions, as a result of the surface return en-tering into the antenna main beam and sidelobes (seefootnote 4). Another difference between the nadir andforward reflectivities is in the height of the echo tops.This can be totally attributed to the forward beam hav-ing about 2 dB less sensitivity than the nadir beam,which results in lower detected cloud tops.

The Doppler velocity panel in Fig. 6, which provideshydrometeor motions [Vhydrometeor in (1)] , shows an ob-vious increase in velocities from above (1–2 m s01) tobelow the bright band (ú6 m s01) due to the increasein fallspeeds from snow to rain. Considering that theupdraft region has reflectivities greater than 45 dBZ ,the particle fallspeeds are likely to be at least 5 m s01

using a Z–£t relation (e.g., Doviak and Zrnic 1993).With this in mind, a strong westward tilted updraft witha core maximum at least 15 m s01 at 4–5-km altitudeis present, with downdrafts along the periphery of theupdraft core. The squall-line updraft region is com-prised of a number of discrete reflectivity and hydro-meteor motion pulses rather than being continuous.This may be partially explained by the way in whichthe cross section intersects the updraft region, particu-larly if the updrafts are three-dimensional rather thantwo-dimensional. Marks and Houze (1987), using na-dir and zenith beams from the NOAA WP3 tail radardata, have provided vertical velocity estimates by re-moving hydrometeor motions £t using separate Z– £t

relations for the snow, rain, and intermediate transitionregions. The £t estimate is the most critical assumptionin obtaining w , since £t depends on many factors suchas particle phase, size distributions, etc. In rain regions,£t can be estimated within 1 m s01 using Z– £t relations.But difficulty occurs in mixed-phase regions and whenlarge ice particles (hail) are present and errors aresomewhat larger. Another factor of relevance is thatreflectivities are attenuated in the convective region andthey must be corrected before using Z–£t relations. Ini-tial attempts for correcting this attenuation have beenreasonably successful (Caylor et al. 1995).

5. Conclusions

EDOP is a new Doppler radar system with a uniquedownlooking, high-altitude ER-2 viewing perspective.The radar operates at 9.6-GHz wavelength and hasfixed nadir and forward pointing beams that map out

Doppler winds and reflectivities in the vertical planealong the aircraft motion vector. Doppler winds fromthe two beams can be used to derive vertical and along-track air motions. The forward beam also provides po-larization information. The dual-beam geometry andthe Doppler and polarization capabilities provide theopportunity for interesting remote sensing studies rel-evant to spaceborne applications and to basic under-standing of precipitation system structure. Futureflights using EDOP combined with other remote sens-ing instruments will maximize the use of the EDOPobservations. Preliminary data from the instrument isencouraging, with indications that most of the originalscience objectives are achievable. The reflectivity andDoppler observations have provided high-resolutionvertical cross sections from squall lines and convectiveregions. Calculation of derived products such as along-track and vertical air velocities, attenuation retrievalfrom the dual-beam approach (Testud and Amayenc1989), and linear depolarization measurements, will bereported in future papers. Methods to improve the ac-curacy of the along-track and vertical air motion esti-mates are currently being examined. Aircraft motionshave been removed from the Doppler measurementsusing low data rate GPS climb rate estimates; high-speed navigation data from these flights will be used inthe future, but the uncertainty in estimation of £t stillremains the largest source of error in estimating w . Bet-ter methods to estimate £t are currently being exploredusing the information from the dual beams and the lin-ear depolarization (LDR) measurements provided byEDOP.

Acknowledgments. The development of this instru-ment was a joint project between the Goddard Meso-scale Atmospheric Processes Branch and the Micro-wave Sensors Branch. It has involved a variety of ef-forts that deserve special recognition. Mr. R. Aldridgehas provided invaluable technician services for the ra-dar. Mr. M. Triesky is appreciated for his assistance ontechnical issues such as antenna testing. Efforts on theEDOP data system by the Goddard MicroelectronicsBranch are acknowledged. Pulse Technology of Mari-etta, Georgia is appreciated for supporting technical is-sues related to the radar microwave hardware. TheNASA Ames High-Altitude Branch was responsiblefor the integration of the instrument on the ER-2 alongwith Lockheed engineers. Many other individuals havehelped in one way or another. This work was supportedunder NASA’s Mission To Planet Earth by Dr. RameshKakar at NASA Headquarters.

REFERENCES

Adler, R. F., R. A. Mack, N. Prasad, H.-Y. Yeh, and I. M. Hakkari-nen, 1990: Aircraft microwave observations and simulations ofdeep convection from 18 to 183 GHz. Part I: Observations. J.Atmos. Oceanic Technol., 7, 377–391.

Atlas, D., 1964: Advances in radar meteorology. Advances in Geo-physics, Vol. 10, Academic Press, 317–478.



Carbone, R. E., M. J. Carpenter, and C. D. Burghart, 1985: Dopplerradar sampling limitations in convective storms. J. Atmos. Oce-anic Technol., 2, 357–361.

Caylor, I. J., G. M. Heymsfield, S. Bidwell, and S. Ameen, 1994:NASA ER-2 Doppler radar reflectivity calibration for the CA-MEX project. NASA Tech. Memo. 104611, 15 pp. [Availablefrom NASA Center for Aerospace Information, 800 ElkridgeLanding Rd., Linthicum Heights, MD 21090-2934]., , , and , 1995: Estimating the rain rate profile inthe presence of attenuation using EDOP airborne radar obser-vations. Preprints, 27th Radar Meteorology Conf., Vail, CO,Amer. Meteor. Soc., 783–785.

Cotton, W. R., and R. A. Anthes, 1989: Storm and Cloud Dynamics.Academic Press, 880 pp.

Davenport, W. B., Jr., and W. L. Root, 1987: An Introduction to theTheory of Random Signals and Noise. IEEE Press, 393 pp.

Doviak, R. J., and D. S. Zrnic, 1993: Doppler Radar and WeatherObservations. Academic Press, 562 pp.

Durden, S. L., E. Im, F. K. Li, W. Ricketts, A. Tanner, and W. Wil-son, 1994: ARMAR: An airborne rain-mapping radar. J. Atmos.Oceanic Technol., 11, 727–737.

Fulton, R., and G. M. Heymsfield, 1991: Microphysical and radiativecharacteristics of convective clouds during COHMEX. J. Appl.Meteor., 30, 98–116.

Griffin, V. L., A. R. Guillory, M. Susko, and J. E. Arnold, 1994:Operations summary for the Convection and Moisture Experi-ment (CAMEX). NASA Tech. Memo. TM-108445, 180 pp.

Heymsfield, G. M., 1989: Accuracy of vertical air motions from nadirviewing Doppler airborne radars. J. Atmos. Oceanic Technol.,6, 1079–1082., I. J. Caylor, J. M. Shepherd, W. S. Olson, S. W. Bidwell, S.Ameen, and W. Boncyk, 1996: Structure of Florida thunder-storms using high-altitude aircraft radiometer and radar obser-vations. J. Appl. Meteor., in press.

Hildebrand, P. H., C. A. Walther, C. Frush, J. Testud, and G. Baudin,1994: The ELDORA/ASTRAIA airborne Doppler weather ra-dar. Proc. IEEE, 82, 1873–1890.

Jameson, A. R., 1987: Relations among linear and circular polariza-tion parameters measured in canted hydrometeors. J. Atmos.Oceanic Technol., 4, 634–645.

Jorgensen, D. P., M. A. LeMone, and B. J.-D. Jou, 1991: Precipitationand kinematic structure of an oceanic mesoscale convective sys-tem. Part I: Convective line structure. Mon. Wea. Rev., 119,2608–2637., T. J. Matejka, and J. D. DuGranrut, 1996: Multi-beam tech-niques for deriving wind fields from airborne Doppler radars. J.Meteor. Atmos. Phys., 59, 83–104.

Kabeche, A., and J. Testud, 1995: Stereoradar meteorology: A newunified approach to process data from airborne or ground-basedmeteorological radars. J. Atmos. Oceanic Technol., 12, 783–799.

Kakar, R., Ed., 1993: Mission summary reports. NASA COARE, 236pp. [Available from NASA Headquarters, Code YS, Washing-ton, DC 20546-0001.]

Kumagai, H., R. Meneghini, and T. Kozu, 1993: Preliminary resultsfrom multiparameter airborne rain radar measurement in thewestern Pacific. J. Appl. Meteor., 32, 433–440.

Lee, W.-C., P. Dodge, F. D. Marks Jr., and P. H. Hildebrand, 1994:Mapping of airborne Doppler radar data. J. Atmos. OceanicTechnol., 11, 572–578.

Marks, F. D., Jr., and R. A. Houze Jr., 1987: Inner core structure ofHurricane Alicia from airborne Doppler radar observations. J.Atmos. Sci., 44, 1296–1317.

Nathanson, F. E., 1991: Radar Design Principles: Signal Processingand the Environment. McGraw-Hill, 720 pp.

Nicholson, S. R., 1994: The design and development of signal-pro-cessing algorithms for an airborne X-band Doppler weather ra-dar. Remote Sensing Lab. Tech. Report 10540-1, University ofKansas, 124 pp. [Available from University of Kansas, 2291Irving Hill Road, Lawrence, KS 66045-2969.]

Ray, P., 1990: Convective dynamics. Radar in Meteorology, D. At-las, Ed., Amer. Meteor. Soc., 383–387.

Rudge, A. W., and N. A. Adatia, 1975: New class of primary-feedantennas for use with offset parabolic-reflector antennas. Inst.Elec. Eng. Electron. Lith., 11, 597–599.

Sauvageot, H., 1992: Radar Meteorology. Artech, 366 pp.Scott, S. G., P. Buji, and K. R. Chan, 1990: The meteorological mea-

surement system on the NASA ER-2 aircraft. J. Atmos. OceanicTechnol., 7, 525–540.

Simpson, J., R. F. Adler, and G. R. North, 1988: A proposed TropicalRainfall Measuring Mission (TRMM) satellite. Bull. Amer. Me-teor. Soc., 69, 278–295.

Skolnick, M., 1990: Radar Handbook. 2d ed. McGraw Hill, 2.1–2.68.

Smith, E. A., X. Xiang, A. Mugnai, R. E. Hood, and R. W. Spencer,1994: Behavior of an inversion-based precipitation retrieval al-gorithm with high-resolution AMPR measurements including alow-frequency 10.7 GHz channel. J. Atmos. Oceanic Technol.,11, 858–873.

Smull, B. F., and R. A. Houze, 1987: Dual-Doppler radar analysis ofa midlatitude squall line with a trailing region of stratiform rain.J. Atmos. Sci., 44, 2128–2148.

Srivastava, R. C., A. R. Jameson, and P. H. Hildebrand, 1979: Time-domain computation of mean and variance of Doppler spectra.J. Appl. Meteor., 18, 189–194.

Testud, J., and P. Amayenc, 1989: Stereoradar meteorology: A prom-ising technique for observation of precipitation from a mobileplatform. J. Atmos. Oceanic Technol., 6, 89–108., P. H. Hildebrand, and W.-C. Lee, 1995: A procedure to correctfor airborne Doppler radar data for navigation errors using theecho returned from the earth’s surface. J. Atmos. Oceanic Tech-nol., 12, 800–820.

Turk, J., J. Vivekanandan, F. S. Marzano, R. E. Hood, R. W. Spencer,and F. J. LaFontaine, 1994: Active and passive microwave remotesensing of precipitating storms during CaPE. Part I: AdvancedMicrowave Precipitation Radiometer and polarimetric radar mea-surements and models. Meteor. Atmos. Phys., 54, 3–27.

Wang, J. R., and Coauthors, 1994: Airborne active and passive mi-crowave observations of Typhoon Flo. IEEE Trans. Geosci. Re-mote Sens., 32, 231–242.

Zrnic, D. S., 1975: Moments of estimated input power for finite sam-ple averages of radar receiver outputs. IEEE Trans. AerospaceElectron. Syst., AES-11, 109–113.

the edop radar system on the high-altitude nasa er-2 aircraft · electra] rely on primarily...

Documents