JOURNAL OF GEOPHYSICAL
Coherent
radar
RESEARCH, VOL. 106, NO. D24, PAGES 33,761-33,772, DECEMBER
ice thickness
measurements
27, 2001
over the Greenland
ice sheet
S. Gogineni,• D. Tammana,• D. Braaten,• C. Leuschen,• T. Akins,g J. Legarsky,3
P. Kanagaratnam,• J. Stiles,• C. Allen,• and K. Jezek4
Abstract. We developedtwo 150-MHz coherent radar depth soundersfor ice thickness
measurementsover the Greenland ice sheet.We developedone of these using
connectorizedcomponentsand the other usingradio frequencyintegrated circuits
(RFICs). Both systemsare designedto use pulsecompression
techniquesand coherent
integrationto obtain the high sensitivityrequired to measurethe thicknessof more than 4
km of cold ice. We used these systemsto collect radar data over the interior and margins
of the ice sheet and severaloutlet glaciers.We operated both radar systemson the NASA
P-3B aircraft equippedwith GPS receivers.Radar data are taggedwith GPS-derived
location information and are collectedin conjunctionwith laser altimeter measurements.
We have reduced all data collectedsince 1993 and derived ice thicknessalong all flight
linesflown in supportof Programfor RegionalClimate Assessment
(PARCA)
investigationsand the North Greenland Ice Core Project. Radar echogramsand derived
ice thicknessdata are placedon a serverat the Universityof Kansas(http://tornado.
rsl.ukans.edu/Greenlanddata.htm)
for easyaccessby the scientificcommunity.We
obtainedgood ice thicknessinformation with an accuracyof _+10 m over 90% of the flight
lines flown as a part of the PARCA initiative. In this paper we provide a brief description
of the systemalongwith samplesof data over the interior, along the 2000-m contour line
in the south and from a few selectedoutlet glaciers.
1.
Introduction
In 1991, NASA started a polar researchinitiative aimed at
determining the mass balance of the Greenland ice sheet.
This programconsistedof coordinatedsurface,airborne, and
spaceborne
measurements
for determiningthe massbalanceof
the ice sheet.The initial airborneprogramconsistedof a laser
altimeterand a Ku-bandradar altimeterfor measuringsurface
elevationof the ice sheetalongselectedflight lines.In 1993the
airborneinstrumentationsuitewas expandedto include a radar depthsounderto collectice thicknessdata alongthe same
flight lines. Ice thickness is a key variable in the timedependentequationof continuityand is essentialto any study
of ice sheet dynamics.
Rajuetal. [1990]developed
a coherentradarsounderfor measurementsin the Antarctic.We usedthis systemto collectice
thickness
data duringthe 1993field season.Althoughthe system
collectedgood quality data in certain areas in the north and
centralpartsof the ice sheet,its performance
waslessthan optimum for obtainingice thicknessdata over a few partsof the ice
sheetin southernGreenland.Theseare in temperateareasof the
ice sheet with thick, warm ice. To overcome its limitations and
improveits performance,we developedtwo new systems:one
•RadarSystems
andRemoteSensing
Laboratory,
The Universityof
Kansas, Lawrence, Kansas.
2jet Propulsion
Laboratory,
Pasadena,
California.
3Departmentof ElectricalEngineering,Universityof MissouriColumbia, Columbia, Missouri.
4ByrdPolarResearchCenter,The Ohio StateUniversity,Columbus, Ohio.
usingconnectorizedcomponentsand the other usingradio frequencyintegratedcircuits(RFICs). The transmitterand receiver
prototypesfor the systemusingRFICs were developedby senior
undergraduate
studentsaspart of a capstonedesignproject.We
usedtheseprototypesto developan operationalsystemthat was
usedto collectsomeof the data reportedin thispaper.
Both radar systemsoperate at the center frequencyof 150
MHz with a chirpedpulseof 1.6 p•sdurationand peak transmit
power of about 200 W. We also appliedthe frequency-wavenumber(f-k) migrationalgorithm[Gazdagand Sguazzero,
1984;
Stolt, 1978] to processa few selecteddata sets to improve
signal-to-noise
(S/N) ratio and resolution.We haveplacedall
of the ice thicknessdata, includingradio echogramsand ice
thicknesses
derivedfrom theseechograms,on a serverfor the
sciencecommunity to download and use in their research.
Thesedata from two nearlyidenticalradarscoverover 90% of
the flight lines flown on an aircraft equippedwith kinematic
GPS receiversfor navigationas part of the PARCA initiative
and provide a consistentdata set of accuratelygeolocated
measurementsof ice thicknessthroughoutGreenland.
The radar data reportedhere both complementand supplement previousdata setscollectedoverthe Greenlandice sheet
[Robin et al., 1969; Ekholm, 1996]. Also, Barnbetet al. [this
issue]developeda new ice thicknessmap by combininghistorical datawith this new data setcollectedunder the Programfor
RegionalClimate Assesment(PARCA) initiative.
In this paper we provide a brief descriptionof the system
with emphasison the new digital signal-processing
section,
antennabeam steering,and the f-k migration algorithm.We
also provide sampleresultsover the interior of the ice sheet
Copyright2001 by the AmericanGeophysicalUnion.
where ice is about 3 km thick, from the 2000-m contour line in
Paper number2001JD900183.
the south, and from two outlet glaciers: Storstr0mmen
0148-0227/01/2001JD900183 $09.00
Gletscher and Zachariae Isstr0m.
33,761
33,762
GOGINENI
ET AL.' COHERENT
RADAR ICE THICKNESS MEASUREMENTS
Table 1. Important Radar SystemParameters
Description
Characteristics
Center frequency
150
RF bandwidth
17
Waveform
Transmitterpulsewidth
Compressedpulsewidth
Range sidelobes
Peak transmitterpower
Coherent integrations
Incoherentintegrations
chirped pulse
1.6
60
<26
200
32 to 4096
0 to 64000
MHz
Furthermore,to reducethe radiation sidelobes,the signal
amplitudesof each elementwere modified accordingto a binomial weighting.This reducedthe sidelobelevel relative to
the peakfrom about-13 dB for the uniformlyweightedarray
MHz
to about
Unit
mounted
Us
ns
dB
W
-30
dB. The
radar
transmitter
in a rack inside the aircraft
and receiver
and connected
are
to their
respectiveantennaswith 20-m-longRF cablesand a feed network with a combined loss of about 3 dB. The effective trans-
mit power at the antennais thus about 100 W.
A low-noisereceiverwith an overall gain of about 100 dB
filters, amplifies,compresses,
and coherentlydetectsthe reI and Q detector bandwidth
8.5
MHz
ceivedsignal.The digital signalprocessor,consistingof two
A/D dynamicrange
72
dB
Antennas
four-elementdipole array
12-bitA/D converters
andadders,digitizesthe in-phase(I) and
Samplingfrequency
18.75
MHz
quadrature(Q) outputsignalsfrom the coherentdetectorand
Samplingdelay
0.080-300
Us
integratesthem. The radar controller consistsof circuitsfor
Range resolutionin ice
4.494
m
Tx. feed network and cable losses
3
dB
generatingdelayedpulserepetitionfrequencyand for setting
Rx. cable and feed network losses
3
dB
the samplingwindowand the clocksignalfor the A/D converters.Table 1 providesimportantspecifications
of the system.
The radarsystemdescribed
by Gogineniet al. [1998]usedthe
digitalsignalprocessor(DSP) developedin 1988 [Xin, 1989].
2. System Description
This DSP systemused 8-bit A/D converterswith a dynamic
range
of about 48 dB. It was housedin an external casethat
A detaileddescriptionof the Universityof Kansascoherent
radar systemalongwith a few sampleresultsis givenby Gog- was connectedto the host computerthrougha ribbon cable.
ineni et al. [1998].Recently,we developeda new digital signal This interface limited the data transfer rate between the DSP
processorand incorporatedantennabeam-steeringcircuitsto systemand the host computer.A minimum of 150 coherent
improve the radar performancefurther. In this sectionwe integrationsat a pulserepetitionfrequency(PRF) of 4 kHz or
provide a brief descriptionof the radar and the new digital lesshad to be performedby the DSP systembeforedata could
be transferredto the main computer,whichlimited the typesof
system.
that couldbe done.
The coherent radars we used for collectingdata over the postprocessing
To improve the systemsensitivity,increasethe dynamic
Greenlandice sheettransmita chirpedpulse,whichis linearly
range, and reducethe minimumnumber of onboardcoherent
frequencymodulatedover a bandwidthof 17 MHz, of 1.6
duration,and hasa peak transmitpowerof about200 W. They integrations,we developeda new DSP systemusing12-bitA/D
use complementarysurfaceacousticwave (SAW) dispersive converters and high-speed field-programable gate arrays
delay lines to generatethe chirp signaland to compressthe (FPGAs). This allowedfor fast integrationsand reductionof
receivedsignal.The compressed
pulsewidth is about 60 ns. the system'sphysicalsize.We developedthe data acquisition
The antennasare two four-element dipole arrays mounted card to interfacewith the PCI bus. Figure 1 showsthe block
under eachwing of the aircraft: one for transmissionand the diagramof the new DSP system.It consistsof two cardsthat
other for reception.To compensatefor the aircraft wing tilt, are pluggedinto a rack-mountablecomputer.One of these
the lengthsof the cablesdrivingeachelementwere adjustedto cards,designedto interfacewith PCI bus,digitizesthe dataand
causethe peak of the far-field radiation pattern from these performsa numberof coherentandincoherentintegrations
(if
arraysto occurdirectlybeneaththe aircraft.
desired),and the other card generatesPRF, the clocksignal
37.5 MHz
;P
• vAidGePo
• Display
PRF
Radar
Control
ISA card
DIGITAL
Pentiron
SYSTEM
Motherboard
A/D conversion
and Integrator
PCI card
IChannel
QChannel
12 bit, 30 MHz
2-Channel
II
PCI
•l•
Bus
I•
I.I vwscsI
I
VWSCSI•_• Hard I
Controller
i i DriveI
Figure 1. Block diagramof the radar digital signalprocessor.
GOGINENI
ET AL.:
,."
COHERENT
RADAR
ICE
THICKNESS
MEASUREMENTS
33 763
:
......... ¾- •
?.'
'.
. •,
:t\.
-:
•.
.,
':
,.-.
-,--' .5.
- -, ...._:-:_•--.. -
,.ß
•
..,
:
,..
1:5.
.: -:-lb.
'"
:
' ßI : •
•
-
.,':'"
'
-..
\ \...- :¾'
:
; .--'"".
-
""
.:•-
;
:,
..:..:
•
\
'*
\
,
:'
:.-::.::
'
"
-..
".
..
4 .!• "¾'" '"
'"'
ß
..,,.......
:
t:
,
:
:
;
'
-/
•
........
".
ß
:
-:
•
.
:
•
-lr
.........
I
-----['.......
•
•.......
ß
;
:
:
I
;
;
;
:
;..,- ........
I
ß-
I
---
ß
:
..
:
': '
"-
-'"
'-
• t5250--•"-"
'
ß
,---T-
.
.•......
Red
solid
line=Flight
lines
which
have
data
i
ß
......... •..............
-•o.
--:-. -
'
.i ........
:.
Plate 1. Locationsof ice thicknessdata collectedfor Greenland with the University of Kansasradar depth
sounderas part of the PARCA initiative.All measurementsare accuratelygeolocatedusingkinematicGPS.
for the A/D converters,and the delayed PRF signal to start
digitization.The host computer providesoperator interface,
storesdigitized data on removablehard disks, and displays
data in real time [/tkins,1999].We usedthe new digitalsystem
to collect data during 1998 and 1999.
The new digital systemis capableof collectingdata with only
32 presummedcoherentreturns at a PRF of 9.2 kHz. It can
also collect data over 2048 range bins and operate with a
maximumPRF of 18.4 kHz, asopposedto 1024 rangebinsand
a maximumPRF of 9.2 kHz for the old system.To operatewith
the higherPRF of 18.4 kHz, the number of rangebins mustbe
decreasedto 800 (whichcorresponds
to a maximumice thick-
nessof about3 km for a radar operatingat an altitude of 500 m
over the ice surface),and the coherentpresumsmust be increasedto 64 or more. The new digital systemalso provides
better gray-scaleresolution (64 versus4) of real-time data
display for the operator. This allows the operator to make
appropriatereal-time receiver-gainadjustmentsto collectdata
in differentpartsof the ice sheet.We developedsoftwarethat
enablesthe operatorto changeradar parametersin flight and
record
onboard
GPS
location
information
and master
clock
data into every radar data record. The aircraft is normally
flown in terrain-followingmode at an altitude of about 450 m
above the ice surface.
33,764
GOGINENI
ET AL.' COHERENT RADAR ICE THICKNESS MEASUREMENTS
10G
ZOO
4OO
7OO
1000
Figure 2a. Radar echogramfrom datacollectedoverthe interiorof the ice sheet.(The radardatahavebeen
averagedover a distancecoveredby aircraftin 1 s (-130 m).)
3.
thicknessby tracking the peaks of surface and ice-bedrock
Results
Since1993,we havecollecteda largevolumeof ice thickness interfacereturnsand by computingthe numberof rangecells
data asa part of the PARCA initiative.Reportswith first-order betweenthe surfaceand the ice-bedrockinterfacepeaks.We
resultsshowingradar echogramsand correspondingderived multipliedthis numberby 4.494,whichis obtainedby dividing
ice thicknesses
for each flight line are providedboth in elec- free-spacerange-celldimensionby the refractiveindex of ice,
to estimate ice thickness. Errors in the derived ice thickness
tronic form (http://tørnadø'rsl'ukans'edu/Greenlanddata'htm)
and in hard copyformat [Chuahet al., 1996a,1996b;Legarsky have been estimated for several Greenland ice sheet locations.
et al., 1997;Wonget al., 1998;Teeet al., 1999].We estimatedice For high elevation,interior locations,we comparedthe radar-
'.:,,." ...'
..' ',....'.
•00
......
.,
.:'
:•
: '-
.,:
. ..'
.... ..':
.....!
i .i
;
!:.....
,:
.'
i
'.
",
; i
'
':
: ..... :...... : ..... 4-..... : '
,"
:
ß .•i::.'.....':..' :,,
,."
...
.....,,".•z'.•:
:ß
' t
'.'.,
: : ",',,:•.,.
i
i
:
'
',
I
;
'r/. -.....
•
!
i
i
i
½'.
....
.,,.'
.:/ i.......
i,.......
i.......
.......
i......
ß..:
•ooo....
,
,
I
;
:: "":.":..,.:•1 /
,
•'" ':/.•..:: , ,
,. . :.::.....•
/.: ,:',',/'
,: ';:':i,...i ri........
'"•.•
i •:."'
.....
¾'"".•
:'" i "'?........
•-'-<
....:i........
", ,
::"
-
.• •00- ß
•
•
:........
• / ." ',"• i "': ';, '::
?"
,:
:
'"'.' ..
i
;
!
,,'
:: !i
!
:
:
:•000 .....
•................
i' .......
:•500 .....
i.....
i',. .
ß
:
...
3000
ß
Geolocation
Distance:
0.00 km
•8.04 km
542.19km
Longitude: 42.8G7W
Latitude:
75.SG9N
41.715W
75.432N
40.7GSW
7S.•gH
78.33 km
39.8•2W
75.142•
104.38 km
38.891W
74.98g•
130.53 km
37.9GSW
74.83•
Figure 2b. Ice thicknessobtainedfrom radar data over the interior of the ice sheet.(The radar data have
been averagedover a distancecoveredby aircraftin 1 s (-130 m).)
GOGINENI
ET AL.: COHERENT
RADAR
ICE THICKNESS
MEASUREMENTS
?
33,765
I
100
2O0
300
6O0
700-.
Ice -. bedrock Interface
800-
900-
1000
•
100
Distance:0.00
200
300
3.59
Longitude: 41.753W
Lattf'ude: 68.634N
41.692W
6.8.613N
400
500
600
7.23Geolocation
10.84
41.629W
68.592N
41.56TW
68.57N
700
800
900
1000
14.44
41.504W
68.548N
Figure 3a. Radar echogramfrom data collectedover the centralpart of the ice sheet.(The radar data have
been averagedover a distancecoveredby aircraft in ---0.139s (---18 m).)
derived ice thickness information
with GISP
These results showed that radar-derived
and GRIP
ice thicknesses
cores.
were
within _+10 m of that for these cores.We did not make any
altitude of about 450 m abovethe ice surfacewith an airspeed
of about 130 m/s. The distance traveled in the time to collect
256 samplesat 9.2 kHz PRF is about 3.6 m. This distanceis
much less than an unfocusedSAR aperture, which is about
For ice thicknessmeasurementsin the vicinity of outlet gla- 21 m for a target locatedat a rangeof 450 m. We processedthe
ciers, Luthi [2000] measured ice thicknessat two boreholes data further by coherentlyintegratingtwo to four samplesand
(828 and 831 m) on Jacobshaven
Isbrae and found thesemea- incoherentlyintegrating10-20 samplesto reduce fading. Figsurements to be within a few meters of the radar-determined
ures 2a and 2b show an example echogram and derived ice
ice thicknessof 830 m. Cross-pointcomparisonsat sevenlo- thicknessfor data collectedand processedover the interior of
cationsin the vicinity of outlet glaciers(geolocationof ice the ice sheet. Ice thickness varies between 3.1 and 2.9 km over
thicknesspairsseparatedby <75 m) gave an averageabsolute a distanceof 100 km from the start of this flight segment.The
difference of 23.6 m.
bottom topographyis relativelyflat over the first 120 km, with
We obtainedgood ice thicknessdata over 90% of the flight a smallpeak of 400 m at the end. The S/N ratio for the bottom
linesflown during PARCA, as shownin Plate 1. Our definition echo varies between 30 and 40 dB for the data shown in the
of goodice thicknessdata is that the signal-to-noiseratio for figure. This confirmsthat the radar can provide consistentice
the bedrock echo is greater than 6 dB and is continuousover thickness information for more than 3 km of cold ice. Also,
a distanceof 20 km alongthe flight path. This is an extensive echoesassociatedwith internal layersare visibleto a depth of
new data set in which all ice thicknessdata are accurately about2.5 km. The white bandbetween100 and 200 pixelsis the
geolocatedusingkinematic GPS. We generatedthis plate by result of usinga sensitivitytime control (STC) circuit to predividingthe ice sheetinto a grid with a resolutionof 0.02ø and vent receiversaturationfrom the strongprimary and multiple
0.06ø latitude and longitude, respectively.The grid extends echoes from the air-ice interface.
from about 59.5 ø to 84 ø N in latitude and 75 ø to 10 ø W in
For about 80% of the flight lines,we had to perform only a
longitude.We identifiedgrid cellswith valid thicknessdata or minimal amount of postprocessing
(additional coherent inteno data along each flight line. In this sectionwe show a few grations(two to four) and incoherentintegrations(10-20) to
examplesof data collectedover the interior of the ice sheet, reducefading) to obtain goodice thicknessdata. Resultsover
around the margin, and on two outlet glaciers.We will show these lines are similar to those discussed above with the S/N
examplesof data where we obtained good ice thicknessdata ratio for bottom echoesvarying between 10 and 40 dB. Bewithoutany processing
andwherewe appliedthef-k migration cause of increased absorption loss for warm ice over some
algorithmto improvethe signal-to-noiseratio.
areasin the south and central parts of the ice sheet,we had to
Normally, we collecteddatawith a radar PRF of 9.2 kHz and perform additionalpostprocessing
on thesedata usingthef-k
the integratorsetto sum256 samplescoherentlyto reducedata migration algorithm to extract ice thickness. Gazdag and
volume.The aircraftwasflown in terrain-followingmode at an Sguazzero
[1984] and Stolt [1978]providea detaileddiscussion
correction for firn effect, which could reduce this error further.
33,766
GOGINENI
ET AL.: COHERENT
RADAR ICE THICKNESS MEASUREMENTS
Icesurface
100
1•
Dts•n•:
0.•
2•
3•
3•
4•
•
•
7•
7•3••tion •10.
•
•N
•:613N
68•N
8•
•
14.•
Figure 3b. Rada• echog•amf•om thesedata afte• p•ocessi•gwith the•-k mig•atio• algorithm.(•
data haw bee• aw•agcd ove• a distancecove•edby aircraftJ• •0.]39 s (•]8 m).)
500
1000
rada•
........................................................................................................................................................................................
..........................................................................................................................................................................................
.• 1500
..................................................................................................................................................
2000
........................................................................................................................................................................................
I
I
I
Geolocation
Dis•nce:
0.00 km
3.-59 km
7.23 km
10.84 km
14.44 km
Longitude:
41.753W
41•92W
41•29W
41.567W
41.504W
Latitude:
68.634N
68•13N
68.592N
68.57N
68.548N
Figure 3c. Ice thickness
obtainedfrom the radar data collectedoverthe centralpart of the ice sheet.(The
radar data havebeen averagedover a distancecoveredby aircraftin •-0.139 s (---18m).)
GOGINENI
i
ET AL.: COHERENT
I
I
RADAR
I
ICE THICKNESS
I
MEASUREMENTS
I
I
I
33,767
..........
1
.
300
........
...............................
.,
400
5OO
I
600
I
I
5oo
1000
•soo
5000
5500
6000
zooo
z500
30%
3500
4ooo
6500
7000
7500
8000
8500
o
lOO
;zoo
300
400
500
600
4500
Geolocation
Longitude: 47.067W 46.137W 44.741W 43.841W 43.497W 43.161W 42.629W 41.g72W 41.613W
Latitude:
61.g24N
61.804N
61.75N
62.04N
62•B81N
63.283N
63.864N
64.339N
64.875N
Figure 4. (a, b) Radar echogramsof data collectedalongthe 2000-mcontourline in the south.(The radar
data havebeen averagedover a distancecoveredby aircraftin 1 s (---130m).)
of the f-k migration technique;therefore only a summaryis
provided here. The f-k migration algorithm is essentiallya
correlationprocessor(i.e., a matchedfilter) implementedwith
Fourier processing.The mathematicsof this correlationprocessorcan be interpreted as reversemigrating (i.e., reverse
propagating)the receivedsignalsback to their sources(i.e.,
scatterers)within the ice. Thus the processingconsistsof a
term that describesthe migrationof the receivedsignalbackto
the air-ice interface, as well as a secondterm describingthe
propagationthroughthe ice layer.
The coherent depth-soundingdata received at the moving
aircraft are denotedby Srx(t,X), which is a functionboth in
time t and along-trackdistancex. Accordingly,thesedata can
be representedin the frequencydomain as
-ax5
(3)
wherec is the speedof electromagneticpropagationin air. The
functionSsmface
(tO,kx) canbe interpretedasthe Fourier transform of Ssmface(X,
t), the scattered(scalar)wave on the ice
surface.In other words, Hai•(tO, kx) migratesthe received
signalfrom the radar back to the ice surface.
The next term of the f-k migration processfurther backpropagatesthiswave,downthroughthe ice layer to a depthd.
Site(tO,kx) = Hice(tO,kx)Ssurfac•(tO,
kx),
(4)
where
H•c•(tO,kx) = exp (j2kzid)
Srx(to,kx) = F{srx(t, x)},
where F is the two-dimensionalFourier transform operator
and to and kx are, respectively,the temporal and spatial Fourier frequencyvariables.
We then multiplythis resultwith Hair(to, kx), the first term
of the f-k migration algorithm,
Ssurface(to,
kx) ---Hair(to,kx)Srx(to,
kx)
=
(2)
In the ice the spatialfrequencyvariable kzi is expressedas
kzi
= t,,-•7•ek•
2,
(5)
where •ce is the velocityof electromagneticpropagationwithin
the ice, in this case a value-assumed constant.
The inverseFourier transformof Sice(to,kx) again can be
interpretedas the signalS•ce(t,x) at a depth d under the ice
surface.Recall that f-k migration is effectivelya correlation
Hair(tO,kx) = exp (j2kz.h).
processor,and a correlationoperationcanbe implementedby
The valueh is the heightof the radar abovethe ice surface,and evaluatingthe outputof a matchedfilter at a specifictime. Simkza is spatialfrequencyin the verticalz dimension.This value ilarly,a two-dimensional
subsurface
image,u(x,z), isgeneratedin
kza is a functionof both tOand kx,
f-k migrationby evaluating
&ce(t,X) at a specifictime ?.
where
33,768
GOGINENI ET AL.: COHERENT RADAR ICE THICKNESS MEASUREMENTS
2000
1500
I
-
I• 500
I
I
I
I
I
I
I
GOGINENI
ET AL.: COHERENT
RADAR
ICE THICKNESS
MEASUREMENTS
33,769
300
4OO
• 00
60O
700
•
800
90O
1000
z
•
•
Geoiocation
Distance:
0.00 km
11.63 km
232.7 km
34.91 km
46.54 km
58.18 km
Longitude: 23.903W
23.54W
23.2W
22.862W
22.555W
22.24W
Latitude:
77.405N
77.362N
77.327N
77.277N
77.228N
77.448N
Figure 5b. Ice thicknessalong a flight acrossStorstrCmmenGletscheron May 22, 1996. (The radar data
have been averagedover a distancecoveredby aircraftin 1 s (-130 m).)
tt(X, g = -d) = Sice(t: T, x).
(6)
these radars if the measurements
are made in coherent
mode
before a significantsurfacemelt startsand near the ice sheet
The value t = •- can be interpretedas the time when Sice(t, marginswith the aircraft flown at different altitudes for the
x) representsthe wave energydue to scatteringspecifically remaining 10% of the flight lines.
from depthd, as opposedto later times(t > •-) whereSice(t,
Figures 4a and 4b show radar echogramsand Figures 4c
x) will be due to the scatteredenergyfrom lower layersthat and 4d show derived ice thicknessesalong the 2000-m conhave propagatedup to depth d.
tour line in the south. We collected these data with a radar
A complete subsurfaceimage u(x, z) can thus be conPRF of 15 kHz and the integrator set to presum 128 returns
structed using f-k migration by calculating and evaluating
coherently. We postprocessedrecorded data by coherently
Sice(t,X) at all depthsz = -d. Figures3a and 3b showradar
integrating four presummed returns and incoherently sumechogramsgeneratedwith normal processingand the f-k miming 30 coherentlyintegrated returns. With an airspeed of
gration algorithm.The inset in thesefiguresshowsthe ampliabout 130 m/s, spacingbetween points along the x axis is
tude of echoesasa functionof depth.Correspondingice thick-130 m. We low-passfiltered the derived ice thicknessdata
nessesfor this echogramare plottedin Figure3c.The S/N ratio
and
plotted the average ice thickness obtained after lowfor the bottom echo using conventionaldata processingis
pass
filtering, shown as a dashedline in Figures 4c and 4d.
about4 dB, whereasthe S/N ratio for data processedwith the
Ice
thickness
gradually decreasedfrom about 1700 m at the
f-k migration algorithm is 10 dB. Over the entire image we
obtained a 4-6 dB improvementin the S/N ratio for data start of the flight line (inset in Figure 4c) to about 1500 m
processed
with the f-k algorithmover that obtainedwith con- over the first 100 km. It then decreasedrapidly from about
1500 m to about 600 m over the next 350 km with significant
ventionalprocessing.
surface
relief. The maximum peak-to-peak ice thicknessdeWe only processeddata with weak bottom echoeswith f-k
migrationbecauseit is computationallyintensive.We obtained viation is about 800 m in this region. The small gap in ice
good thicknessinformation for about an additional 10% for thickness between 3500 and 3700 units is caused by weak
the flight lines beyond the 80% obtainedwith conventional bottom echoessubmergedin clutter from off-angle surface
processing.
The areaswhere we couldnot obtain ice thickness returns and ice-surfacemultiples. The averageice thickness
consistedprimarily of the deep, narrow channelbeneath the increased from about 500 to 1200 m over the next 10 km,
Jacobshavn
outlet glacierand near the marginsof the ice sheet and then it decreasedgraduallyfrom 1200 to 800 m over the
where bottom echoeswere mergedwith multiples of the ice next 200 km. Bottom topography is very rough in this area
surface. Much of the radar data over the Jacobshavn area were
with peak-to-peak deviationsof about 700 m over a distance
collectedin incoherentmode during 1997, and we could not of about 10 km. We believe that these are the first highpostprocess
these data to improvethe S/N ratio. We believe quality ice thicknessdata collectedover this region of the ice
that we can obtain ice thickness within the narrow channel with
sheet.
33,770
GOGINENI
ET AL.: COHERENT
RADAR ICE THICKNESS MEASUREMENTS
800
600
40O
200
0
-200
-600
Di•ance:
-800
0.00 km
13.02 km
26.05 km
Geolocation
39.17 lun
52.19 km
65.21 km
78.34 km
91.36 km
Longitude: 19.07W
19.664W
20.26W
20.852W
21A26W
21.967W
22.535W
23.164W
Latitude:
78.92N
78.915N
78.925N
78.942N
78.973N
79.005N
79.056N
78.927N
Figure 6a. Ice surfaceelevationand ice bottomelevation(WGS84 ellipsoid)alonga flight up Zachariae
IsstrOmon May 19, 1999. (The radar data have been averagedover a distancecoveredby aircraft in 1 s
(-130 m).)
3.1.
Outlet
Glaciers
Along the Greenlandice sheetmargin,outlet glaciersplay a
major role in ice dischargeand accountfor about half of the
total masslossfrom the ice sheet[Reehet al., 1999].Ice thicknessmeasurementsobtained from the coherentradar depth
sounder in conjunctionwith high-resolutionsurface images
(e.g., Landsatand satellitesyntheticapertureradar (SAR)),
permit three-dimensionalcharacterizationsof the outlet glaciers. Surface imagesprovide important informationon the
lateral surfacedimensionsof outlet glaciers,the locationof the
calvingfront, the surroundingtopography,and ice surfaceflow
lines.Surfaceimagescan alsobe usedto determinethe ice flow
velocity, and Rignot et al. [1997] have used interferometric
SAR to determine the surfacevelocityof outlet glacierswith
ness and ATM
ice surface
elevation
measurements
to be
placed in proper contextwith the local complexterrain. We
present detailed characterizationsof two exampleoutlet glaciers located in northeast Greenland: StorstrOmmen Gletscher
(77.0øN, 22.6øW) and ZachariaeIsstrOm(79.0ø N, 20.0øW).
For each of theseoutlet glaciers,ice thicknessprofilesalong
the glacier axis and acrossthe glacier are presented.Also
includedare crossoverpoints to quantifythe repeatabilityof
these measurements.
StorstrOmmenGletscher is roughly oriented in the northsouth direction and is -41 km wide in the entranceregion,
narrowingto -24 km in the main channel.Weidick[1995]gives
the speedof thisglacierasrangingfrom200to 1800m yr-•.
Figure5a showsice surfaceelevationand ice bottomelevation
an accuracy
of about1 m yr-•. Radaricethickness
measure- (WGS84 ellipsoid)from a flighton May 24, 1997,whichbegan
mentsprovide the critical third dimensionthat allowsassessment of volume flux, bed topography,and the locationsof
groundinglines. The groundingline locationis an especially
importantparameterin assessing
massbalancesincesubglacial
meltingof floatingice canbe large [Reehet al., 1997;Mayeret
al., 2000].
Using GPS-derivedpositionsassociatedwith the radar ice
thicknessmeasurements
and the corresponding
ice surfaceelevationmeasurements
from the airbornetopographicmapper
(ATM), flight pathsover outlet glacierswere plottedon highresolutionSAR imagesusingthe National Snowand Ice Data
Center (NSIDC) LIMB (large imagemap browser)software.
NSIDC distributesthissoftwareto viewthe digitalSAR mosaic
of the Greenlandice sheetthat wascompiledfrom ERS-1 data
obtainedduringAugust 1992. This allowsthe radar ice thick-
inland of the calvingfront and over the center of the glacier
axis.The flighttrajectorywasup the glacierat a slightlyoff-axis
angle that took it within about 8 km of its easternboundary.
Initially, the ice thickness(obtained by subtractingthe ice
bottom elevationfrom the ice surfaceelevation)increasedby
85 m to -440 m at a distanceof about9.5 km alongthe flight
path, due to a decreasein the ice base elevation. The ice
thicknessdecreasedslightlyto 415 m between9.5 and 19 km
due mainly to an increasein ice base elevation.Ice thickness
then increased at a rate of about 3.76 m km-• to -550 m over
the next 40 km as the ice surface elevation increased and the bed
elevationremainednearlyconstantat approximately-250 m.
Beyondthis point, the bed elevationincreasedas the flight trajectoryleft the main glacialchannel.
As seenin Figure 5a, the bed topographyof StorstrOmmen
GOGINENI
ET AL.: COHERENT
RADAR ICE THICKNESS MEASUREMENTS
33,771
loo
200
300
40O
500
6O0
700
800
900
lO0O
Distance:0.00
kin
23.50
km
Longitude: 21.375W
21.557W
21.75W
21.949W
Latitude:
78.859N
78.978N
79.097N
78.745N
Geøløcatiø•1.80
km
65.30
km
Figure 6b. Ice thickness
alonga flightacrossZachariaeIsstrOmon May 24, 1997.(The radar datahavebeen
averagedover a distancecoveredby aircraft in 1 s (--•130m).)
Gletscherwas quite smooth. Figure 5b showsice thickness
from a flight on May 22, 1996, havinga trajectoryacrossthe
entrance region of StorstrOmmenGletscher from the west to
the east side.The westernportion of the glacierbasinwas the
deepest,with an ice thicknessof more than 800 m extending
over --•19 km. The easternhalf of the glacier entrance region
had a thicknessaveraging632 m. A crossoverpoint in the two
flights described above was compared, showing very close
agreementin the ice thicknessmeasurements.The 1997 upglacier flight measuredan ice thicknessof 530.0 m, and the
1996 cross-glacierflight measuredan ice thicknessof 529.7 m
at roughlythe same GPS location.
ZachariaeIsstrOmis roughlyorientedin the east-westdirection. This glacier is --•60 km wide in the entrance region,
narrowingto --•26km over a distancedown the glacierof --•35
km. Weidick[1995]statesthat the speedof this glacieris 500 m
yr-•. Figure6ashows
theicesurface
elevation
andtheicebase
ZachariaeIsstr0mis very roughin comparisonwith Storstr0mmen Gletscher.Figure 6b showsice thicknessfrom a flight on
May 24, 1997,havinga trajectoryacrossthe entranceregionof
Zachariae Isstr0mfrom the southto the north. The averageice
thicknessacrossthe main glacier bed was 770 m, with the
deepestportion of the glacier basin havingan ice thicknessof
more than 900 m. A crossoverpoint for thesetwo flightsalso
showsgoodagreement.The 1999 up-glacierflight measuredan
ice thicknessof 712.6 m, and the 1997 cross-glacier
flight measured an ice thicknessof 727.1 m at roughly the same GPS
location.This is consistentwith _ 10 m accuracyinferred from
the comparison of radar-derived thicknessesto GISP and
GRIP
4.
ice cores.
Conclusions
We have developedtwo coherent radars for measuringice
thicknessand have conductedextensiveaircraft surveyscompiling a large, consistent,geolocateddata set of ice thickness
on the Greenland ice sheet as part of the NASA PARCA
initiative. These radars obtain the high sensitivityrequired to
soundglacialice more than 3 km thick usingpulsecompression
and coherentsignalprocessing
while transmittingat only about
200 W peak power. We obtained good ice thicknessinformaincrease in ice surface elevation between 25 km and 40 km.
tion for over 80% of the flight lines flown as a part of the
plus an addiAlong this 15-km segmentof the flight path, ice thickness PARCA initiative with normal postprocessing
increasedat a rate of about39.1m km-• to --•810m, a thick- tional 10% of the flight linesusingthef-k migrationalgorithm.
eningrate that is more than 10 timesof what was observedfor We have presentedexamplesof data from the interior, over
StorstrOmmen
Gletscher.The 40 km point alongthe flight path the marginsand on the outlet glaciersof the Greenland ice
correspondsto the smallestlateral extent of the glacier.
sheet,which demonstratethe widely varyingbedrock topograIt is alsoapparentfrom Figure 6a that the bed topographyof phy for different regionsof the ice sheet.We distributethese
elevation from a flight on May 19, 1999, which began over
floatingice and continuedup the centerof the glacieraxis.The
ice thickness
increasedfrom zero (ice surfaceelevationis equal
to the ice bed elevation)to --•250m in the baybeyondthe exit
region of the glacier at a point about 25 km along the flight
path. Beyondthis,thickeningrapidlyoccurredprimarily due to
a large decrease in the ice bed elevation and a moderate
33,772
GOGINENI ET AL.: COHERENT RADAR ICE THICKNESS MEASUREMENTS
data to the scientificcommunityboth in hard copyform and in
electronic form from our server. The developmentof these
radars involved significantparticipation by Electrical Engineering and Computer Scienceundergraduatestudentsas a
part of their education.
Radar Syst. and Remote Sens.Lab., Univ. of Kansas,Lawrence,
Kansas, 1996a.
Chuah,T. S., S. Gogineni,C. Allen, and B. Wohletz, Radar thickness
measurementsover the northern part of the Greenland ice sheet
[1995data],RSL Tech.Rep.10470-3,Radar Syst.and RemoteSens.
Lab., Univ. of Kansas, Lawrence, Kansas, 1996b.
Ekholm, S., A full coverage,high-resolutiontopographicmodel of
Greenland composedfrom a variety of digital elevationdata, J.
Geophys.Res., 101, 21,961-21,972, 1996.
Gazdag,J., and P. Sguazzero,Migration of seismicdata,Proc.IEEE,
Notation
s = receivedradar signalas a function of time
S = receivedradar signalas a functionof
frequencyand distance.
1998.
Ssurface(6O,
kx): Fourier transformof Ssurface(X,
t), the
scattered(scalar)waveon ice surface.
Hair = matchedfilter responsefor a target on
the surface.
Hice = matchedfilter responsefor a target on
the ice.
t = time.
Fourier
Remote Sens. Lab., Univ. of Kansas, Lawrence, Kansas, 1997.
Luthi, M.P., Rheologyof cold frin and dynamicsof a polythermalice
stream,Ph.D. dissertation,Eidg. Tech. Hochsch.,Zurich, Switzerland, 2000.
2289-2292, 2000.
transform
Raju, G., W. Xin, and R. K. Moore, Design,development,field operationsandpreliminaryresultsof the coherentAntarcticradar depth
sounder(CARDS) of the Universityof Kansas,USA, J. Glaciol.,36,
operator.
to = temporal frequencyvariable.
kx = spatialfrequencywave numberin x
247-258, 1990.
Reeh, N., H. Thomsen, O. Olesen, and W. Starzer, Mass balance of
North Greenland, Science,278, 205-206, 1997.
direction.
h - height of radar aboveice surface.
kza - spatialfrequencyin verticaldimension,z,
in air.
Reeh, N., C. Mayer, H. Miller, H. Thomsen,and A. Weidick,Present
andpastclimatecontrolon fjord glaciationsin Greenland:Implicationsfor IRD-depositionin the sea,Geophys.
Res.Lett.,26(8), 10391042, 1999.
kzi = spatialfrequencyvariablein vertical
dimension, z, in ice.
c = speedof electromagneticpropagationin
air.
d = depth.
l,,ic
e : velocityof electromagneticpropagation
within
Legarsky,J., S. P. Gogineni,C. Allen, T. S. Chuah,and Y. C. Wong,
Radar thicknessmeasurements
overthe northernpart of Greenland
ice sheet--1996 results,RSL Tech.Rep. 10470-6, Radar Syst.and
Mayer, C., N. Reeh, F. Jung-Rothenh/iusler,
P. Huybrechts,and H.
Oerter, The subglacialcavityand implied dynamicsunder NioghalvfjerdsfjordenGlacier, NE-Greenland,Geophys.
Res.Lett., 27(15),
x = distance.
F = two-dimensional
72, 1302-1315, 1984.
Gogineni, S., T. Chuah, C. Allen, K. Jezek, and R. K. Moore, An
improvedcoherentradar depth sounder,J. Glaciol.,44, 659-669,
and distance.
ice.
Rignot,E. J., S. P. Gogineni,W. B. Krabill, and S. Ekholm,North and
northeastGreenlandice dischargefrom satelliteradar interferometry, Science,276, 934-937, 1997.
Robin, G. de O., S. Evans,and J. T. Bailey, Interpretationof radio
echosoundingin polar ice sheets,Philos.Trans.R. Soc.London,Ser.
.4,265, 437-505, 1969.
Stolt,R. H., Migrationby Fouriertransform,Geophysics,
43, 23-48, 1978.
Tee, K. L., W. K. Chong, H. Coulter, T. Akins, S. P. Gogineni, C.
Allen, and J. Stiles, Radar thicknessmeasurement over the southern
u (x, z) - two-dimensional
subsurface
image.
r - specifictime.
part of the Greenlandice sheet--1998 results,RSL TR 13720-10,
Radar Syst.and Remote Sens.Lab., Univ. of Kansas,Lawrence,
Kansas, 1999.
Acknowledgments.The work is supportedby NASA grantNAG58758. We would like to thank William
Krabill from the NASA
God-
dard SpaceFlight Center for providingsurfaceelevationdata and
Doug Young, alsofrom Goddard,who designedthe antennamount.
We thank the aircraft crew for the excellentsupportprovidedwhile
conductingtheseexperiments.We alsoacknowledge
Greg Rosenand
Meng Wu for their help in plotting flight trajectoriesand Donnis
Graham for editing this paper. The analysesof outlet glacierspresentedin this paperbenefitedgreatlyfrom the digitalSAR mosaicof
the Greenlandice sheetproducedby Mark Fahnestock(Department
of Meteorology,Universityof Maryland,CollegePark) andRon Kwok
(Jet PropulsionLaboratory,CaliforniaInstituteof Technology,Pasadena).The helpfulsuggestions
andcommentsof tworeviewersare also
greatlyappreciated.
References
Weidick, A., Greenland[with a sectionof Landsatimagesof Greenlandby R. S. WilliamsJr. andJ. G. Ferrigno],in SatelliteImageAtlas
of Glaciersof the World, edited by R. S. Williams Jr. and J. G.
Ferrigno,U.S. Geol. Surv.Prof. Pap. 1386-C, 141 pp., Washington,
D.C.,
1995.
Wong, Y. C., S. Gogineni,C. Allen, J. Stiles,J. Legarsky,T. Akins, and
K. L. Tee, Radar thicknessmeasurements over the Greenland ice
sheet--1997 results,RSL Tech.Rep. 10470-7, Radar Syst.and Remote Sens. Lab., Univ. of Kansas, Lawrence, Kansas, 1998.
Xin, W., Data processing
systemof the coherentAntarcticradardepth
sounder,RSL Tech. Rep. 8070-3, Radar Syst. and Remote Sens.
Lab., Univ. of Kansas, Lawrence, Kansas, 1989.
C. Allen, D. Braaten, P. Gogineni,P. Kanagaratnam,C. Leuschen,
J. Stiles,and D. Tammana,Remote SensingLaboratory,Universityof
Kansas, 2335 Irving Hill Road, Lawrence, KS 66045-7612.
([email protected])
T. Akins, Jet PropulsionLaboratory,4800 Oak Grove Dr., Pasa-
Akins, T., Design and developmentof an improveddata acquisition dena, CA 91109.
K. Jezek,Byrd Polar ResearchCenter, The Ohio State University,
systemfor the coherentradar depth sounder,M.S. thesis,Electr.
108 Scott Hall, 1090 Carmack Road, Columbus OH 43210-1002.
Eng. and Comput. Sci., Univ. of Kansas,Lawrence,Kansas,1999.
J. Legarsky,Department of Electrical Engineering,University of
Bamber,J. L., R. L. Layberry,and S. P. Gogineni,A newice thickness
and bed data setfor the Greenlandice sheet,1,J. Geophys.
Res.,this Missouri-Columbia,223 EngineeringBuilding West, Columbia,MO
issue.
Chuah,T. S., S. Gogineni,C. Allen, B. Wohletz, Y. C. Wong, P. Y. Ng,
and E. Ajayi, Radar thicknessmeasurements
overthe southernpart
of the Greenland ice sheet [1993 data], RSL Tech.Rep. 10470-2,
65211.
(ReceivedJuly31, 2000;revisedFebruary5, 2001;
acceptedFebruary15, 2001.)