Open Research OnlineThe Open Universityrsquos repository of research publicationsand other research outputs

Autonomous underwater vehicles (AUVs) andinvestigations of the ice-ocean interface deploying theAutosub AUV in Antarctic and Arctic watersJournal ItemHow to cite

Dowdeswell J A Evans J Mugford R Griffiths G McPhail S Millard N Stevenson P Brandon MA Banks C Heywood K J Price M R Dodd P A Jenkins A Nicholls K W Hayes D Abrahamsen EP Tyler P Bett B Jones D Wadhams P Wilkinson J P Stansfield K and Ackley S (2008) Autonomousunderwater vehicles (AUVs) and investigations of the ice-ocean interface deploying the Autosub AUV in Antarcticand Arctic waters Journal of Glaciology 54(187) pp 661ndash672

For guidance on citations see FAQs

ccopy [not recorded]

Version Version of Record

Link(s) to article on publisherrsquos websitehttpdxdoiorgdoi103189002214308786570773httpwwwigsocorgjournal54187j07j141pdf

Copyright and Moral Rights for the articles on this site are retained by the individual authors andor other copyrightowners For more information on Open Research Onlinersquos data policy on reuse of materials please consult the policiespage

oroopenacuk

Instruments and Methods

Autonomous underwater vehicles (AUVs) and investigations of theicendashocean interface in Antarctic and Arctic waters

JA DOWDESWELL1 J EVANS1 R MUGFORD1 G GRIFFITHS2 S McPHAIL2

N MILLARD2 P STEVENSON2 MA BRANDON3 C BANKS3 KJ HEYWOOD4

MR PRICE4 PA DODD4 A JENKINS5 KW NICHOLLS5 D HAYES5

EP ABRAHAMSEN5 P TYLER6 B BETT6 D JONES6 P WADHAMS78

JP WILKINSON9 K STANSFIELD10 S ACKLEY11

1Scott Polar Research Institute University of Cambridge Lensfield Road Cambridge CB2 1ER UKE-mail jd16camacuk

2National Marine Facilities National Oceanography Centre Universtiy of Southampton Southampton SO14 3ZH UK3Department of Earth and Environmental Sciences Open University Walton Hall Milton Keynes MK7 6AA UK

4School of Environmental Sciences University of East Anglia Norwich NR4 7TJ UK5British Antarctic Survey Natural Environmental Research Council Madingley Road Cambridge CB3 0ET UK

6Deep-Sea Biology Group National Oceanography Centre University of Southampton Southampton SO14 3ZH UK7Department of Applied Mathematics and Theoretical Physics University of Cambridge Cambridge CB3 0WA UK

8Laboratoire drsquoOceanographie de Villefranche Universite Pierre et Marie Curie UMR 7093 BP 2806234 Villefranche-sur-Mer Cedex France

9Scottish Association for Marine Science Dunstaffnage Marine Laboratory Dunbeg Oban Argyll PA37 1QA UK10Ocean Observing and Climate Group National Oceanography Centre University of Southampton

Southampton SO14 3ZH UK11Department of Earth and Environmental Science University of Texas at San Antonio San Antonio Texas 78249 USA

ABSTRACT Limitations of access have long restricted exploration and investigation of the cavitiesbeneath ice shelves to a small number of drillholes Studies of sea-ice underwater morphology arelimited largely to scientific utilization of submarines Remotely operated vehicles tethered to a mothership by umbilical cable have been deployed to investigate tidewater-glacier and ice-shelf margins buttheir range is often restricted The development of free-flying autonomous underwater vehicles (AUVs)with ranges of tens to hundreds of kilometres enables extensive missions to take place beneath sea iceand floating ice shelves Autosub2 is a 3600 kg 67m long AUV with a 1600m operating depth andrange of 400 km based on the earlier Autosub1 which had a 500m depth limit A single direct-drive dcmotor and five-bladed propeller produce speeds of 1ndash2m sndash1 Rear-mounted rudder and stern-planecontrol yaw pitch and depth The vehicle has three sections The front and rear sections are free-flooding built around aluminium extrusion space-frames covered with glass-fibre reinforced plasticpanels The central section has a set of carbon-fibre reinforced plastic pressure vessels Four tubescontain batteries powering the vehicle The other three house vehicle-control systems and sensors Therear section houses subsystems for navigation control actuation and propulsion and scientific sensors(eg digital camera upward-looking 300 kHz acoustic Doppler current profiler 200 kHz multibeamreceiver) The front section contains forward-looking collision sensor emergency abort the homingsystems Argos satellite data and location transmitters and flashing lights for relocation as well asscience sensors (eg twin conductivityndashtemperaturendashdepth instruments multibeam transmitter sub-bottom profiler AquaLab water sampler) Payload restrictions mean that a subset of scientificinstruments is actually in place on any given dive The scientific instruments carried on Autosub aredescribed and examples of observational data collected from each sensor in Arctic or Antarctic watersare given (eg of roughness at the underside of floating ice shelves and sea ice)

INTRODUCTION

The undersides of floating ice shelves and sea ice in theAntarctic and Arctic are among the least accessibleenvironments on Earth Ice shelves several hundred metresthick fed from parent ice sheets float above submarinecavities that are up to 1 km deep and cover areas as large asabout 500 000 km2 (eg Jenkins and Doake 1991 Mayer

and others 2000) The irregular calving of icebergs frommarginal ice cliffs makes the close approach to both floatingice shelves and grounded tidewater-glacier margins hazar-dous Sea ice with ridged submarine keels that reach tens ofmetres deep covers about 15106 km2 of the Arctic Oceanand surrounds the Antarctic continent during winter(Cavalieri and others 2003) The interactions between iceshelves sea ice and the ocean are of considerable scientific

Journal of Glaciology Vol 54 No 187 2008 661

interest not least because the nature and rate of freezing andmelting processes that take place are of wider significanceto the global environmental system through their influenceon for example water masses that flow equatorward as adriver of the thermohaline circulation of the oceans (egBroecker 1991)

Limitations of access have long precluded the thoroughexploration and investigation of the cavities beneath iceshelves where only a small number of drillholes have givenaccess to the waters beneath (eg Nicholls 1996 Nichollsand others 2001) Studies of the underwater morphology ofsea ice have been restricted largely to the scientificutilization of submarines (eg Wadhams 1978 1988) Theuse of emerging technology in the form of unmannedunderwater vehicles provides a method by which these veryinaccessible and inhospitable parts of the global oceanand cryosphere can be investigated safely (Francois 1977Wadhams and others 2004) Remotely operated vehicles(ROVs) tethered to a mother ship by an umbilical power andcontrol cable have been deployed to investigate tidewater-glacier and ice-shelf margins but their range is restrictedto hundreds or at most a few thousands of metres (egDowdeswell and Powell 1996 Powell and others 1996)However the development of free-flying autonomousunderwater vehicles (AUVs) provides a means to operate

over ranges of tens to hundreds of kilometres and to depthsbelow even the thickest floating ice shelves

Here we describe briefly the specification of theAutosub2 AUV (Fig 1) the scientific instrument packagesit has deployed and examples of applications to ice-shelfglaciology sea-ice studies oceanography and glacial geol-ogy in Antarctic and Arctic waters Full details of the at-seaoperations together with an inventory of the scientific datafor the four cruises comprising the Autosub Under Iceprogramme are available on the website of the BritishOceanographic Data Centre (BODC) at wwwbodcacukprojectsukauicruise_programme

THE AUTOSUB AUTONOMOUS UNDERWATERVEHICLE (AUV)

Design and powerAutosub2 is a 3600 kg 67m long AUV with a 1600moperating depth and range of 400 km at a forward speed of17m sndash1 This vehicle is based on the earlier Autosub1which had a 500m depth limit (Millard and others 1998)Autosub2 is referred to as Autosub through the remainder ofthis paper For the polar science campaigns considered hereit was instrumented as illustrated in Figure 1b (Stevenson

Fig 1 (a) The Autosub AUV being deployed in Courtauld Fjord East Greenland from RRS James Clark Ross Autosub is 67m longPhotograph by JA Dowdeswell (b) The major systems of Autosub2 and the science sensors that were installed for the Autosub polarmissions of 2003ndash05 The vehicle displaced 36 t and had a range of 400 km at 17m sndash1 with the payload of sensors shown

Dowdeswell and others Instruments and methods662

interest not least because the nature and rate of freezing andmelting processes that take place are of wider significanceto the global environmental system through their influenceon for example water masses that flow equatorward as adriver of the thermohaline circulation of the oceans (egBroecker 1991)

Limitations of access have long precluded the thoroughexploration and investigation of the cavities beneath iceshelves where only a small number of drillholes have givenaccess to the waters beneath (eg Nicholls 1996 Nichollsand others 2001) Studies of the underwater morphology ofsea ice have been restricted largely to the scientificutilization of submarines (eg Wadhams 1978 1988) Theuse of emerging technology in the form of unmannedunderwater vehicles provides a method by which these veryinaccessible and inhospitable parts of the global oceanand cryosphere can be investigated safely (Francois 1977Wadhams and others 2004) Remotely operated vehicles(ROVs) tethered to a mother ship by an umbilical power andcontrol cable have been deployed to investigate tidewater-glacier and ice-shelf margins but their range is restrictedto hundreds or at most a few thousands of metres (egDowdeswell and Powell 1996 Powell and others 1996)However the development of free-flying autonomousunderwater vehicles (AUVs) provides a means to operate

over ranges of tens to hundreds of kilometres and to depthsbelow even the thickest floating ice shelves

Here we describe briefly the specification of theAutosub2 AUV (Fig 1) the scientific instrument packagesit has deployed and examples of applications to ice-shelfglaciology sea-ice studies oceanography and glacial geol-ogy in Antarctic and Arctic waters Full details of the at-seaoperations together with an inventory of the scientific datafor the four cruises comprising the Autosub Under Iceprogramme are available on the website of the BritishOceanographic Data Centre (BODC) at wwwbodcacukprojectsukauicruise_programme

THE AUTOSUB AUTONOMOUS UNDERWATERVEHICLE (AUV)

Design and powerAutosub2 is a 3600 kg 67m long AUV with a 1600moperating depth and range of 400 km at a forward speed of17m sndash1 This vehicle is based on the earlier Autosub1which had a 500m depth limit (Millard and others 1998)Autosub2 is referred to as Autosub through the remainder ofthis paper For the polar science campaigns considered hereit was instrumented as illustrated in Figure 1b (Stevenson

Fig 1 (a) The Autosub AUV being deployed in Courtauld Fjord East Greenland from RRS James Clark Ross Autosub is 67m longPhotograph by JA Dowdeswell (b) The major systems of Autosub2 and the science sensors that were installed for the Autosub polarmissions of 2003ndash05 The vehicle displaced 36 t and had a range of 400 km at 17m sndash1 with the payload of sensors shown

Dowdeswell and others Instruments and methods662

and others 2003) Mechanically the vehicle consisted ofthree sections The front and rear sections were free-flooding built around aluminium extrusion space-framesand covered with (replaceable) glass-fibre reinforced plastic(GFRP) panels The central section comprised seven 3mlong carbon-fibre reinforced plastic (CFRP) pressure vesselswithin a cylindrical matrix of syntactic foam ndash one centralpressure vessel and the surrounding six at 608 intervalsThese pressure vessels limited the vehicle operating depth to1600m at a safety factor of two Four of the tubes housed thebattery system of up to 5184 lsquoDrsquo size primary manganesealkaline cells With a total weight of 720 kg these providedup to 60 kWh (220MJ) of energy (depending upon usage rateand ambient temperature) Thermal insulation between thecells and the CFRP tubes enabled an internal temperature ofgt158C to be maintained using the waste heat of the cellsdespite external temperatures as low as ndash28C (Stevenson andothers 2002) The three other tubes housed electronicschassis for the control systems and sensors

The rear section of Autosub housed essential subsystems(navigation control actuation and propulsion) and scientificsensors (eg digital camera upward-looking 300 kHz Tele-dyne RDI acoustic Doppler current profiler (ADCP) 200 kHzmultibeam receiver) A single brushless direct-drive (nogearbox) direct-current motor and five-bladed propeller gavethe vehicle a speed of 1ndash2m sndash1 A rear-mounted rudder andstern-plane controlled vehicle yaw pitch and depth

The free-flooding front section of Autosub housed otheressential vehicle subsystems (forward-looking collisionsensor emergency abort the homing system Argos trans-mitters and flashing lights for relocation) as well as sciencesensors (eg twin Seabird 911 conductivityndashtemperaturendashdepth (CTD) instruments multibeam transmitter EdgeTechsub-bottom profiler Envirotech AquaLab water-samplingsystem) Payload restrictions meant that on any given dive itwas possible to deploy in the front and rear sections ofAutosub only a subset of the instruments actually available

The control and data system for Autosub was based upona distributed and networked control architecture (McPhailand Pebody 1998) With such architecture it is relativelystraightforward to add new sensors onto the vehicle withoutaffecting the safe operation of the control system

Navigation and controlThe rationale for a highly accurate navigation system wasthat when executing under-ice missions the vehicle wouldbe required to travel 100 km or more without the possibilityof global positioning system (GPS) position fixes or trackingfrom the mother ship Furthermore it would need tonavigate its way back to a relatively small hole in the iceor to a polynya The two primary sensors for navigation werea 150 kHz Teledyne RDI ADCP and Ixsea-Oceano PHINSfibre-optic gyro-based inertial navigation system (INS) Toobtain the best possible navigational accuracy (errors oflt02 of distance travelled were typically achieved even at808N) the downward-looking ADCP must be able to trackthe seabed The missions under the ice shelf needed as greata bottom-tracking range as possible This meant the use of arelatively low-frequency (150 kHz) ADCP (with a bottomtrack range of 500m) rather than higher-frequency versionstypically fitted to AUVs Both the INS and the 150 kHzADCP were housed within a single pressure case so that thevital mechanical alignment between the ADCP and INScould be maintained accurately between missions The

navigation system was also able to utilize the velocity-tracking data from the upward-looking 300 kHz ADCP Thiswould be used when the Autosub was flying within a cavityunder an ice shelf close enough to obtain useful velocity-tracking data from the underside of the essentially stationaryice shelf (at lt150m range) but too far from the seabed touse the (preferred) seabed-tracking mode

Collision-avoidance and emergency beacon systemsIn polar waters uncharted bathymetry icebergs sea-icepressure ridges and the undersides of ice shelves with theirunknown topography are all possible collision risks This ledto the development of a strategy system and algorithms forcollision avoidance specifically designed for polar oper-ations The system relied upon the use of sensors and dataalready available on the vehicle these being

Paroscientific Digiquartz pressure sensor

Four upward-looking ADCP beam ranges

Four downward-looking ADCP beam ranges

Forward-looking echo sounder (Simrad Mesotech120 kHz)

The approach was to keep the hardware and software assimple as possible triggering one straightforward yeteffective behaviour upon detecting that a collision wasimminent

Collision-avoidance mode was entered if

the forward-looking echo sounder detected an objectcontinuously closing on the vehicle and at a rangelt100m or

there was lt50m depth of water in which the vehiclecould operate

Once collision-avoidance mode was triggered the vehiclewas programmed to backtrack for 1 km along its previouscourse It then returned along its original route but with anoffset of up to 500m either side of its pre-planned trackSimultaneously the AUV adjusted its depth safety limitsincreasing the margin of safety in the vertical plane If animminent collision was detected again the vehicle repeatedthe collision-avoidance manoeuvre but with a newrandomly chosen track offset Once clear of the obstaclethe original course and safety limits were restored Anexample of the operation of the collision-avoidance algo-rithm allowing Autosub to circumvent an obstructingobject is shown in Figure 2

Fig 2 Plan view of collision-avoidance behaviour triggered bydetection of a 30m deep iceberg keel ahead on mission 365 offnortheast Greenland (Wadhams and others 2006) It took threeattempts for Autosub to avoid the hazard and continue eastwards onits programmed course Axes are in decimal degrees north and west

Dowdeswell and others Instruments and methods 663

If the Autosub systems detected a critical failure or ifthere was a catastrophic power loss during the mission anemergency acoustic beacon would be dropped on a 15mcable transmitting a 45 kHz chirp once per minute Thebeacon would be heard on the ship using a vertical-arrayreceiver deployed to a depth of up to 100m By timing thearrival of the received signal at each of three or more shiprsquospositions it would be possible to triangulate Autosubrsquosposition up to a range of 30 km

Deployment and recovery in iceWhereas many handling problems of AUVs in the openocean are a result of high sea states in polar waters the maindifficulties arise from fast-changing ice conditions Duringthe first Autosub campaign in the Antarctic the vehicle waslaunched and recovered in areas that were virtually clear ofice (Brierley and others 2002) However sea-ice coverremained a hazard if the wind was blowing the ice out tosea In such circumstances the launch and recovery positioncould be tens of kilometres away from the area of interestand a large proportion of battery energy was expendedsimply getting Autosub to the ice edge

For subsequent campaigns a sink-weight release systemwas developed allowing the vehicle to be launched in anyice-free patch of water This avoided the need for a largeopen-water area to allow Autosub to dive from the surface ata shallow inclination At a predetermined depth usually15ndash20m a 20 kg weight slung beneath the nose wouldrelease and the propulsion motor would start The wholeautonomous mission from that point would be carried outsubmerged A disposable passive hydrostatic safety releasewas fitted between the weight and the programmablerelease to ensure that the weight fell away in the event ofthe programmable release failing The sink-weight systemgreatly enhanced the effectiveness of under-ice work byfacilitating deployment close to the ice edge

Water-density gradientsAutosub is ballasted to be 8ndash12 kg positively buoyant inwater and final ballast adjustments were usually made onthe working site after taking water-density measurementsWhen working close to an East Greenland tidewater glaciera density difference of 4 kgmndash3 between the surface and 6mwas present This density difference produced a 10 kgincrease in buoyancy for the vehicle at its working depthand was on the verge of making the vehicle unable tocontrol its depth The solution adopted to cope with these

density variations was to fit lsquowingletsrsquo (160mm half-span by254mm chord) slightly aft of the centre of gravity of thevehicle to produce additional downward force whilemoving These proved effective and had the added benefitof reducing body pitch angle

Surfacing in ice and homing systemAutosub had two Argos satellite transmitter antennae andone WiFi Ethernet radio antenna mounted externally Thesewere vulnerable and could be damaged by the ice before thevehicle had been located visually potentially resulting invehicle loss simply because it could not be located Evenwhen ice cover was very light it was clear that antennaecould be broken easily and the vehicle could become ice-covered (Fig 3) Autosub is particularly dependent on theArgos transmissions for location at the end of missions and aGFRP tube was made to fit over the forward antenna to givesome protection against ice damage The experience high-lighted the need to control the final stages of the end of themission remotely rather than trying to second-guess iceconditions at the mission programming stage

On occasions due to drifting sea ice it was not possibleat the time of launch to be certain that the recovery positionwould be clear of ice Further there was a need to be able tocope with situations in which the mother ship could notreach the intended recovery position or when unexpectednavigation system drift or some other failure could leavethe AUV still operating but a long way off the intendedrecovery position

To deal with these eventualities a homing system wasdeveloped able to guide Autosub back towards the mothership at ranges of up to 15 km The shipboard homing beacontransmits regularly spaced swept frequencies (chirps)centred at 4504 kHz On the vehicle were three sphericalhydrophones which together with the three-channel correl-ation receiver allowed the direction of arrival of the homingsignal to be determined Once the system had detected fourconsecutive transmissions with the correct temporal spa-cing the AUV entered homing mode and headed towardsthe source of the signal This system was used successfullyduring the northeast Greenland campaign of 2004 where onmission 365 the intended recovery position had beencovered by sea ice as a result of changing wind conditions

INSTRUMENTATION FOR THE AUTOSUB AUVGeophysical instrumentsA swath-bathymetry system or multibeam echo sounder wasincluded in the Autosub instrument package to measure thegeometry of both the underside of ice shelves and sea icetogether with sea-floor morphology The system could beconfigured to operate in upward- or downward-lookingmode for glaciological and glacialndashgeological applica-tions respectively A Kongsberg Simrad EM-2000 swath-bathymetry system was used It operates at a frequency of200 kHz has 111 individual beams an angular coverage ofover 1208 and a swath width of up to 300m (depending ondistance from target) Quantitative data on elevation can begridded at a horizontal resolution of 1m Root-mean square(rms) errors of better than 10 cm can be achieved in thevertical A swath-bathymetry system was preferred to aconventional marine side-scan sonar instrument because itprovides quantitative three-dimensional (3-D) morphologic-al data of high absolute accuracy rather than imaging

Fig 3 Autosub surfacing in light sea-ice cover Note the possibilityof damage to antennae protruding from the vehicle

Dowdeswell and others Instruments and methods664

changes in backscatter that relate to both geometry andsurface properties

An EdgeTech chirp penetration echo sounder was alsomounted on Autosub to investigate the acoustic stratigraphyof the sea floor The profiler operates at 2ndash16 kHz andpenetrates through up to several tens of metres of sedimentdepending on grain-size density and pore-water character-istics with a vertical resolution of 6ndash10 cm Both geophysicalinstruments log data to internal hard drives for the durationof each Autosub mission

Oceanographic instrumentsAutosub was equipped with a Seabird 911 plus CTD systemwhich included two pairs of temperature and conductivitysensors A dissolved oxygen sensor was also attachedalthough for accurate oxygen measurements this needed tobe calibrated against laboratory measurements on concur-rent water samples The Seabird conductivity and tempera-ture sensors were in general remarkably stable so that withtwo pairs and regular calibration high accuracy measure-ments were possible The specified drift rate for thetemperature and conductivity sensors was 00028Candash1 and00024 siemensmndash1 andash1 (00023mndash1 andash1) respectively Thetwo essentially independent sensor pairs provided a checkon the data quality and we have typically found the pairsdiffered by no more than 00018C in temperature and nomore than 0002 in derived salinity The deployment of ashipboard CTD before or after an Autosub mission provideda cross-check that the Autosub CTD sensors were makingreasonable measurements The Autosub CTD data wereprocessed using the software provided by the manufacturerfollowing the standard processing pathway set out in theuser manual for the software This included calculatingsalinity and other derived variables

Upward- and downward-looking Teledyne RDI ADCPswere fitted which were used both for navigation andscientific measurements The downward-looking 150 kHzinstrument typically returned current measurements up to150ndash200m below Autosub The upward-looking 300 kHzinstrument typically provided more limited range up toabout 100m above Autosub The Autosub ADCP data wereprocessed using a system of dedicated MATLABTM scripts

Photographic instrumentsAutosub was equipped with a Starlight SXV-H9 which is ablack-and-white charge-coupled device (CCD) imager toobtain images of the sea floor and the marine benthos closeto and beneath floating ice The imager was selected for itshigh sensitivity (particularly in the important blue part of thespectrum) low readout noise (about 20 photons rms equiva-lent) and high dynamic range An integral data loggerrecords the images on hard disk which can be accessed viathe Autosub radio network The camera is installed in the tailsection of Autosub and a Minolta zoom flash is installedwithin a pressure case at the nose of Autosub The imagesensor has an array of 1040 1392 of 6 mm square pixelsmaking an imager size of 624 835mm With an airndashwater magnification factor of 14 this equates to an imagediagonal of 22m at a flying altitude of 10m The cameradata are stored in a raw 16-bit binary format

Water-sampling instrumentsAutosub carried a compact water sampler to allow themeasurement of a wide range of water properties The

sampler was an Envirotech AquaLab which consists of amechanical syringe that draws water into one of 49 EVAplastic sample bags by means of a rotary valve (Fig 4)Samples are suitable for most tracer and nutrient studies butnot for the measurement of trace gases due to the gas-permeable nature of the EVA bags used

Before deployment sample bags must be filled with asmall volume of lsquoprimersquo fluid so that hydrostatic forces donot crush connecting parts Ordinarily this fluid is flushedout of each bag in situ before a sample is collectedHowever this time-consuming procedure can be avoided ifbags are filled with a prime fluid in which the properties ofinterest are known and later accounted for (Dodd andothers 2006) This approach allows a 200mL sample to becollected in 8ndash10min during which Autosub would traveljust under 1 km at a cruising speed of 16m sndash1 Samples ofany size can be collected but multiple syringe strokes arerequired to collect samples larger than 200mL

The water sampler is capable of operating in an autono-mous mode in which samples are collected at predeter-mined times or it can be triggered by Autosub to collectsamples at specific locations It is also possible for Autosubto circle while a sample is collected and continue onlywhen the sampler reports that sampling is complete Todate the sampler has provided samples for oxygen-isotoperatio and barium concentration measurements (Dodd andothers 2006)

AUV OBSERVATIONS OF ICE OCEAN AND SEAFLOOR SOME EXAMPLESIce shelves the underside of an Antarctic ice shelfTraditional methods for determining the topography of anice-shelf base have used either downward-looking radarfrom above the upper surface of the ice shelf generally froman aircraft platform or the inversion of elevation data ofthe upper surface under the assumption that the ice columnfloats in hydrostatic equilibrium (Bamber and Bentley 1994Dowdeswell and Bamber 2007) Both techniques havetheir advantages inversion of (typically) satellite-derivedsurface elevations gives very good spatial coverage athorizontal scales somewhat longer than the ice is thick

Fig 4 The AquaLab water sampler located in the nose of theAutosub vehicle Individual water-sample bags are visible Photo-graph by P Dodd

Dowdeswell and others Instruments and methods 665

downward-looking radar gives a detailed view at horizontalscales determined by the ice thickness and the wavelengthof the radar waves in ice Reliance on the results from thesetechniques has reinforced the notion that the base of iceshelves is generally rather smooth and can be regarded ashaving a drag coefficient at the icendashwater interface similar tothat of the sea floor Neither technique however is able toreveal basal topography at the scales important to thefriction exerted on water flow beneath the ice shelf aparameter important to the realistic modelling of flowbeneath the ice shelf

Autosubrsquos mission 382 beneath Fimbul Ice Shelf EastAntarctica yielded the first acoustic image of the base of anice shelf (Fig 5) The upward-looking multibeam echosounder gave a 150m wide image of the base of this iceshelf along 18 km of the mission track For most of the swaththe vehicle was 90m below the ice base The draft of the iceshelf as seen by the multibeam system is shown in Figure 5aThe breadth of the swath is an indication of the specularity ofthe ice base with a broader swath indicative of a rougherbase Much of the swath suggests an almost specularreflection consistent with the generally held view that thebase of an ice shelf is rather smooth (Holland and Feltham2006) A substantial fraction however is very roughFigure 5b shows a 3-D visualization of a rough portion ofthe swath from 189 to 215 km along track illustrating thatthe basal ice-shelf topography is quite chaotic at horizontallength scales of 10m or less with vertical scales similarly ofthe order of 10m In reality the image in Figure 5 is asubstantially smoothed visualization as the extreme natureof the terrain caused excessive shadowing which has beenfilled-in in a smooth manner

The rough portions of the swath data correspond on theice-shelf surface to flow traces These flow traces are linear

features visible from aerial or satellite imagery (Fahnestockand others 2000) that are often initiated at glacial featuressuch as shear margins or associated with regions of fastflow Flow traces are ubiquitous on ice shelves and if theyare generally underlain by an ice base with such dramatictopography it is clear that the frictional drag imposed on theocean circulation beneath the ice shelf needs to bereassessed (Nicholls and others 2006)

Sea ice a three-dimensional viewIn August 2004 the Autosub AUV in operations off northeastGreenland obtained the first successful multibeam sonarmeasurements under sea ice giving a quantitative map ofthe 3-D nature of the under-ice surface (Wadhams andothers 2006) The vehicle operating from RRS James ClarkRoss obtained more than 450 track-km of under-ice multi-beam sonar data using the Kongsberg EM-2000 systemFigure 6 shows examples of imagery from first- and multi-year ice including young ridges old hummocks andundeformed melting ice The imagery was obtained frommission 365 on 21ndash22 August 2004 which headed westacross the shallow Belgica Bank at 79830rsquoN under partiallygrounded multi-year ice then penetrated further over the500m deep Norske Trough occupied mainly by unde-formed first-year fast ice Each of the displayed images is aperspective view of the underside of the ice obtained withthe AUV at 40m depth with scenes shown as if illuminatedby a sun of elevation 208

Two swath-bathymetric images of the underside ofArctic sea ice are shown Figure 6a illustrates the deepestridge encountered during mission 365 which has a 33mdraft This ridge is embedded in a larger multi-year floe(from 3200 to 3800m) that probably drifted out from theArctic Ocean the previous summer The undeformed ice

Fig 5 (a) Multibeam data from mission 382 beneath Fimbul Ice Shelf East Antarctica showing ice-shelf draft (m) the track starting beneaththe ice shelf (0 km) and ending at the ice front (26 km) (b) 3-D rendering of swath-bathymetric data showing a portion of the ice base belowa flow trace at 20 km including the smooth base either side of the feature (from Nicholls and others 2006)

Dowdeswell and others Instruments and methods666

Fig 6 Examples of EM-2000 swath-bathymetric images of the under surface of sea ice offshore of northeast Greenland The perspectiveviews are illuminated by a sun elevation of 208 (a) An embedded multi-year floe with a 33m deep sea-ice ridge The floe is surrounded byundeformed shorefast sea ice (b) A multi-year ridged floe of draft 3ndash5m embedded in undeformed shorefast ice of draft 18m Fast iceshows a pattern of depressions due to mirroring of surface melt pools The floe contains a pressure ridge of maximum draft 11m which haspartly disintegrated into individual ice blocks of diameter 5ndash20m (from Wadhams and others 2006)

Dowdeswell and others Instruments and methods 667

surrounding this floe is 175m in draft and is almostcertainly first-year ice Since the individual ice blocks thatmake up ridges are quite small the ridge is a relativelyuniform triangle in cross-section representing the angle ofrepose of a pile of buoyant ice A number of thinner floes10ndash15m in draft are also visible in the image

Figure 6b shows an old multi-year ridged floe of thickness3ndash5m which is embedded in younger fast ice of draft 18mThe edges of the floe are sharp and linear as would occurwith a fracture that occurred just before embedding Theridge which occupies half of the floe has maximum draft of11m and contains separate ice blocks of typical diameters5ndash20m In the ice surrounding the embedded floe a numberof small floes with drafts of about 10m are present Thefaint pattern of depressions in the underside of the thinnerice occurs because of the presence of meltwater pools on theupper surface These pools preferentially absorb incoming

radiation giving a heat flux that enhances bottom melt andgenerates a bottom depression which mirrors the position ofeach pool on the top side (Wadhams and Martin 1990Wadhams 2000)

Oceanography the nature of a water-filled cavitybeneath an ice shelfThe majority of Antarctic Bottom Water (AABW) is thoughtto have its origins in processes that take place over theAntarctic continental shelf These processes therefore reflectthe importance of AABW as a key component in the globalthermohaline circulation As a consequence interactionsbetween the Southern Ocean and Antarctic ice shelveswhich cover 40 of the Antarctic continental shelf are alsoimportant Historically exploration of the processes beneathice shelves has been restricted to what can be achieved bydrilling access holes and deploying oceanographic instru-mentation into the water column beneath The process ofmaking access holes is demanding logistically and a rathersmall number of holes can be made in any given Antarcticfield season In fact fewer than 30 access points have beenmade across all ice shelves in Antarctica Clearly AUVs offeran opportunity to improve substantially our ability to obtaindata from this unique environment

During mission 382 to the cavity beneath Fimbul Ice ShelfAutosub executed a simple inndashout track with a total tracklength of 60 km 53 km of which was beneath the ice shelfThe in-going track was at an elevation above the seabed of150m The vehicle then turned on a reciprocal trackascending to an elevation of 400m The fact that the seabedshallows towards the ice front combined with an overridinginstruction to maintain a minimum headroom from the icebase of 90m meant that Autosub was terrain-following offthe base of the ice shelf for much of the return track

The temperature salinity and current-speed data obtainedfrom the primary oceanographic instruments during themission are shown in Figure 7 These data exhibit a wealth ofdetail and are discussed by Nicholls and others (2006) inthe context of data obtained from the front of the ice shelfusing the ship The principal conclusion of Nicholls andothers (2006) was that as the properties of some of thewaters observed within the cavity did not relate to the watersobserved along the front of Fimbul Ice Shelf at the time ofthe mission the cavity must be flushed episodically byrelatively warm water that crosses the continental-shelfbreak from the north possibly during the winter

An intriguing dataset acquired by the Autosub ADCPs isshown in Figure 7a The effective range of an ADCP in largepart depends on the number and type of scatterers in thewater column and their size with respect to the wavelengthsin the acoustic pulse With a wavelength of 10mm the150 kHz downward-looking instrument generally has agreater range than its 300 kHz upward-looking counterpartThis can be seen outside the cavity on the left side ofFigure 7a Once Autosub has passed beyond one or two tidalexcursions into the cavity (a distance of about 3 km) therange of both instruments decreases markedly and the perfor-mance of the 300 kHz ADCP overtakes that of the 150 kHzinstrument The performance of the ADCPs indicates adifferent biological assemblage beneath the ice shelfimplying a reduction in the volume density of biologicalmaterial and a shift towards smaller-sized scatterers Therapid fall-off with distance into the cavity of scatterer volumedensity also suggests that this is an area of outflow

Fig 7 Oceanographic data from mission 382 obtained beneathFimbul Ice Shelf Antarctica (a) Mission trajectory (red and bluelines indicate the outward and return Autosub legs respectively)The vertical dashed line at 265 km gives the position of the icefront referenced to the turning point in the mission the horizontaldashed line at 570m depth shows the depth of a nearby sill at thecontinental-shelf break Also shown are the ADCP data illustratingthe dramatically reduced range beneath the ice shelf that implies adearth of appropriately sized scatterers in the water column Theupward-looking instrument operated at 300 kHz and the down-ward-looking instrument at 150 kHz The data are for the northndashsouth velocity component (positive northward approximatelyperpendicular to the ice front) which have been averaged using ahorizontal window 100m wide The inset shows the ADCP data inthe vicinity of the ice front for the outward leg (b) Verticallyaveraged ADCP currents after subtraction of the modelled tide(c) Salinity (bold) and potential temperature () The thin green near-horizontal dashed line is the freezing point of the water at surfacepressure for salinities measured on the outward journey (fromNicholls and others 2006)

Dowdeswell and others Instruments and methods668

Although investigation of the ocean processes withincavities beneath ice shelves will always require mooredinstruments capable of collecting data over periods ofmonths or years Autosubrsquos ability as a platform that canuse sophisticated oceanographic instrumentation to gener-ate spatially extensive datasets has given us a unique view ofone of the least accessible parts of the worldrsquos oceans

Autosub was lost under the Fimbul Ice Shelf on mission383 the one following that described above Its low-frequency acoustic beacon signalled that an abort had beentriggered and that the vehicle was stuck at a positionapproximately 17 km from the ice front A full investigation(Strutt 2006) concluded that either an open-circuit ornetwork failure was the most likely cause of an abort orloss of power This was the only time the vehicle deployed itslong-range acoustic beacon

Oceanography fjord circulation and meltwater fluxThe circulation and mixing processes of water masses withinfjords can be complex so one advantage of a rapid andcontinuous surveying device such as Autosub is to enablea more synoptic survey than is achievable with a shipTypically saline ocean water enters a fjord at mid-depthabove the sill and fresh meltwater from the surroundingglaciers or rivers exits the fjord as a surface layer (Syvitski andothers 1987) The deep waters within the fjord are renewedonly sporadically However this steady-state simple picturecan be complicated by the presence of tides cross-fjordflows sea ice entering andor leaving the fjord and the flowsinduced by inertial oscillations following storms The netexchange of fresh water between a fjord and the continental-shelf environment is of importance in determining forexample the influence of meltwater from the Greenlandice sheet on the formation of dense water masses in the seassurrounding Greenland

The detailed current-velocity structure revealed by theAutosub ADCPs at the mouth of Kangerdlussuaq Fjord onthe east coast of Greenland is shown in Figure 8 In the

6 hours of this survey three passes across the fjord weremade at depths of 70 190 and 400m The upward- anddownward-looking ADCPs are very consistent betweenadjacent passes some 4hours apart implying that tidal (orother temporally varying) flows are not dominant hereHowever the velocity structure is very different from thesimple three-layer flow suggested above The primaryinflows are on the southwest side of the mouth at 300ndash400m and in the upper 100m The primary outflow is atabout 200ndash350m on the northwest side of the fjord mouthimplying a clockwise circulation of open ocean water in thebay at the mouth of the fjord There is a suggestion that thewater in the top 10m may be a thin layer of ice melt leavingthe fjord Thus Autosub has revealed in unprecedenteddetail a snapshot of the complex exchanges between anArctic fjord environment and the adjacent continental shelf

Oceanography attenuation of waves by sea iceA serendipitous result for the behaviour of waves propa-gating in sea ice was obtained from the upward-lookingADCP surface track velocity recorded on Autosub Becausethe surface track ping has longer range than the profile pingthe velocity of sea ice relative to Autosub could be measuredduring runs as deep as 200m This was the first use of anAUV to measure directional and scalar wave propertiesduring surface wave propagation through sea ice (Hayes andothers 2007) Since ice-edge detection was also possiblefrom the surface track ping (verified by ship observations)dependence of the above wave properties on distance fromthe edge of the marginal ice zone could be examined

As an example during mission 324 on 25 March 2003 inthe marginal ice zone of the Bellingshausen Sea Antarcticathe horizontal velocity of the ice was oscillating Themagnitude of this oscillation also decayed with distancefrom the ice edge both on the inward and outward segments(Fig 9a) In the observed regime of small ice floes (lt20m)and long wavelength (100ndash350m) the floes nearly follow thecircular path of a point on the water surface Therefore the

Fig 8 Cross-section of the current velocity (colour scale in m sndash1)into and out of the mouth of Kangerdlussuaq Fjord East Greenlandacquired from the upward- and downward-looking ADCPs mountedon Autosub The Autosub navigated horizontal paths at 70 190 and400m (marked as black lines) descending or rising in betweentaking 6 hours to complete the survey Positive values denote waterflowing into the fjord negative values indicate water flowing out ofthe fjord Southwest is to the left and northeast to the right

Fig 9 Sea-ice velocity from Autosub mission 324 The upward-looking ADCP measured the surface track velocity upon (a) enteringthe ice pack at 90m and (b) exiting the ice pack at 90m Themagnified inset shows a typical segment analyzed here Note thestrong periodicity in both components as well as mean currenttowards the southeast (modified from Hayes and others 2007)

Dowdeswell and others Instruments and methods 669

surface track velocity is regarded as a measurement of sur-face wave orbital velocity superimposed on mean icevelocity (southeastward in the case of mission 324) Theseries is divided into a number of blocks (Fig 9b) to analyzethe surface velocity The directional and scalar wave spectraare calculated for each segment so any trend in significantwave height mean and peak wave periods as well as anychange in the energy wave direction or spread of variousfrequency components can be detected (Fig 10) The char-acter of waves propagating through sea ice that was observedusing Autosub agrees with most of the previous observational(Wadhams and others 1986 1988 Liu and others 1991) andnumerical (Meylan and others 1997) experiments

Glacial geology submarine glacial landforms andacoustic stratigraphyThe morphology and stratigraphy of the sea floor provideimportant evidence for the reconstruction of the dimensionsand flow of former ice sheets (eg Anderson 1999) Whereice flows across a sedimentary bed landforms diagnostic ofice-flow direction and dynamics are produced These land-forms which are often streamlined are preserved underwater as ice retreats across continental shelves and fjordsduring interglacial and interstadial periods (eg Andersonand others 2002 Ottesen and others 2005 Evans andothers 2006) Characteristic assemblages of these submarinelandforms are indicators of for example ice-stream flowpast glacier-surge activity and former grounding lines (egPowell and others 1996 Canals and others 2000 O Cofaighand others 2002 Ottesen and Dowdeswell 2006)

The swath-bathymetry system on Autosub when mountedin downward-looking mode produces data that yield high-resolution digital-terrain models and 3-D images of the seafloor Figure 11 shows the floor of an East Greenland fjordwhere the fast-flowing Kangerdlussuaq Glacier one of themajor outlet glaciers of the Greenland ice sheet (Rignot andKanagaratnam 2006) has produced streamlined sediment-ary bedforms which are preserved in several hundred metres

of water after ice retreat from its position at the Last GlacialMaximum (Syvitski and others 1996) Shallow acousticstratigraphy provides further information on the structure ofthe upper few metres to tens of metres of sediment In theexample shown in Figure 11b the acoustic profiler onAutosub penetrates the fine-grained and acoustically lamin-ated sediments in the deepest part of Kangerdlussuaq Fjordwith less transparent and probably coarser-grained sedi-ments characteristic of glacial diamicts or tills to either sideAutosub which has been deployed close to the calvingtidewater margins of Courtauld Glacier East Greenland(Fig 1a) can be used to image areas of the sea floor inpreviously inaccessible locations near calving ice cliffs andbeneath ice shelves

In addition to geophysical instruments the digital cameraequipment on Autosub provides detailed information on theform and composition of the sea floor and the marine biotathat inhabit it Figure 12 shows an example of a sea-floorphotograph acquired by Autosub in Kangerdlussuaq FjordBoth individual dropstones released by iceberg melting andbottom-dwelling marine organisms are shown The presenceof deposit-feeding species is indicated by faunal traces on thesediment surface Evidence of disturbance to the seabed andfauna from iceberg-keel ploughing was also observed inphotographs of the sea floor at water depths less than about500m reducing faunal density and diversity as well asproducing a sedimentologically heterogeneous environment

CONCLUSIONSThe Autosub AUV provided a platform for the deployment ofa number of geophysical and oceanographic instruments inhazardous polar environments that ships and other mannedvehicles cannot access

Fig 10 (a) Mean wave period and (b) significant wave height forAutosub missions 322ndash324 The label lsquoinrsquo refers to the seriescollected upon entering the ice pack while lsquobackrsquo refers to thereturn series Period and wave height are derived from the one-dimensional wave spectrum of 512 s blocks (with the exception ofthe return trip in mission 323 in which 256 s blocks were analyzed)

Fig 11Multibeam echo-sounder image of the glacially streamlinedsea floor of Kangerdlussuaq Fjord acquired from a 200 kHz swath-bathymetry system mounted on Autosub The swath width isapproximately 200m Water depth is 710ndash840m The swath-bathymetry data are gridded at a resolution of 1m in the horizontalThe lower panels show acoustically stratified sediments on the fjordfloor acquired from the chirp 2ndash16 kHz sub-bottom profiler onAutosub The acoustic profile is located in the multibeam image

Dowdeswell and others Instruments and methods670

Ice-covered environments investigated using Autosubinclude a cavity beneath the Fimbul Ice Shelf and therelatively shallow and poorly charted waters beneath sea iceon the East Greenland continental shelf

The multibeam echo sounder of Autosub has imaged theunderside of an ice shelf for the first time showing that someareas are very rough with implications for the modelling ofwater flow and melt rates The underside of sea ice has alsobeen imaged in detail and quantitative shape parametersextracted Swath images and bottom photographs of theglacial geology and marine biota close to the margins ofArctic tidewater glaciers have also been obtained

Oceanographic data such as salinity temperature andwater velocity have been derived continuously during Auto-sub missions beneath floating Arctic and Antarctic ice pro-viding observations with a very dense spatial coverage inenvironments where previously few or no data have beenavailable

The Autosub3 vehicle successor to the lost Autosub2and AUV technology in general is likely to be used in-creasingly in hazardous polar marine environments for thecollection of detailed geophysical and oceanographic dataclose to and beneath floating ice These data in turn areimportant in the calibration and testing of numerical modelsrelating to ice-sheet interactions with the polar waters

Not all of the scenarios for AUV operations in polar seashave yet been achieved in practice although many havebeen described and discussed by scientists and engineers(Griffiths and Collins 2007 Collins and Griffiths 2008) InAugust 2007 the first AUV campaign took place to searchfor and then examine hydrothermal sites at the slow-spreading Gakkel Ridge in the Arctic Ocean an area ofextensive multi-year pack ice (Reves-Sohn and others 2007)Other plans include multidisciplinary studies beneath theRoss Ice Shelf Antarctica and surveys of Southern Oceankrill populations in winter

ACKNOWLEDGEMENTSThis work was supported by the Autosub Under Ice The-matic Programme of the UK Natural Environment ResearchCouncil (Programme Chair S Ackley Programme ManagerK Collins) We are grateful to the Autosub Technical Teamand the officers and crew of RRS James Clark Ross for theirinvaluable contributions on four Autosub cruises

REFERENCESAnderson JB 1999 Antarctic marine geology Cambridge etc

Cambridge University PressAnderson JB SS Shipp AL Lowe JS Wellner and AB Mosola

2002 The Antarctic ice sheet during the last glacial maximumand its subsequent retreat history a review Quat Sci Rev21(1ndash3) 49ndash70

Bamber JL and CR Bentley 1994 A comparison of satellite-altimetry and ice-thickness measurements of the Ross Ice ShelfAntarctica Ann Glaciol 20 357ndash364

Brierley AS and 11 others 2002 Antarctic krill under sea iceelevated abundance in a narrow band just south of ice edgeScience 295(5561) 1890ndash1892

Broecker WS 1991 The great ocean conveyor Oceanography4(2) 79ndash89

Canals M R Urgeles and AM Calafat 2000 Deep sea-floorevidence of past ice streams off the Antarctic PeninsulaGeology 28(1) 31ndash34

Cavalieri DJ CL Parkinson and KY Vinnikov 2003 30-Yearsatellite record reveals contrasting Arctic and Antarctic decadalsea ice variability Geophys Res Lett 30(18) 1970 (1010292003GL018031)

Collins K and G Griffiths eds 2008 Workshop on AUV sciencein extreme environments collaborative Autosub science inextreme environments Proceedings of the International ScienceWorkship 11ndash13 April 2007 Scott Polar Research InstituteUniversity of Cambridge UK London Society for UnderwaterTechnology

Dodd PA MR Price KJ Heywood and M Pebody 2006Collection of water samples from an autonomous underwatervehicle for tracer analysis J Atmos Oceanic Technol 23(12)1759ndash1767

Dowdeswell JA and JL Bamber 2007 Keel depths of modernAntarctic icebergs and implications for sea-floor scouring in thegeological record Mar Geol 243(1ndash4) 120ndash131

Dowdeswell JA and RD Powell 1996 Submersible remotelyoperated vehicles (ROVs) for investigations of the glacierndashoceanndashsediment interface J Glaciol 42(140) 176ndash183

Evans J JA Dowdeswell C O Cofaigh TJ Benham and JB And-erson 2006 Extent and dynamics of the West Antarctic IceSheet on the outer continental shelf of Pine Island Bay during thelast glaciation Mar Geol 250(1ndash2) 53ndash72

Fahnestock MA TA Scambos RA Bindschadler and G Kvaran2000 A millennium of variable ice flow recorded by the RossIce Shelf Antarctica J Glaciol 46(155) 652ndash664

Francois RE 1977 High resolution observations of under-icemorphology Seattle WA University of Washington AppliedPhysics Laboratory Tech Rep APL-UW-7112

Griffiths G and K Collins eds 2007 Masterclass in AUVtechnology for polar science collaborative autosub science inextreme environments Proceedings of the International Master-class 28ndash30 March 2006 National Oceanography CentreSouthampton UK London Society for Underwater Technology

Hayes DR and A Jenkins 2007 Autonomous underwater vehiclemeasurements of surface wave decay and directional spectra inthe marginal sea ice zone J Phys Oceanogr 37(1) 71ndash83

Holland PR and DL Feltham 2006 The effects of rotation andice shelf topography on frazil-laden ice shelf water plumesJ Phys Oceanogr 36(12) 2312ndash2327

Fig 12 Example photograph from Autosub mission 377 showingthe floor of outer Kangerdlussuaq Fjord (imaged from an altitude of9m at a depth of 564m) One cobble-sized iceberg-rafted drop-stone three large burrows and numerous tubeworms are visibleThe photograph is about 1m across

Dowdeswell and others Instruments and methods 671

If the Autosub systems detected a critical failure or ifthere was a catastrophic power loss during the mission anemergency acoustic beacon would be dropped on a 15mcable transmitting a 45 kHz chirp once per minute Thebeacon would be heard on the ship using a vertical-arrayreceiver deployed to a depth of up to 100m By timing thearrival of the received signal at each of three or more shiprsquospositions it would be possible to triangulate Autosubrsquosposition up to a range of 30 km

Deployment and recovery in iceWhereas many handling problems of AUVs in the openocean are a result of high sea states in polar waters the maindifficulties arise from fast-changing ice conditions Duringthe first Autosub campaign in the Antarctic the vehicle waslaunched and recovered in areas that were virtually clear ofice (Brierley and others 2002) However sea-ice coverremained a hazard if the wind was blowing the ice out tosea In such circumstances the launch and recovery positioncould be tens of kilometres away from the area of interestand a large proportion of battery energy was expendedsimply getting Autosub to the ice edge

For subsequent campaigns a sink-weight release systemwas developed allowing the vehicle to be launched in anyice-free patch of water This avoided the need for a largeopen-water area to allow Autosub to dive from the surface ata shallow inclination At a predetermined depth usually15ndash20m a 20 kg weight slung beneath the nose wouldrelease and the propulsion motor would start The wholeautonomous mission from that point would be carried outsubmerged A disposable passive hydrostatic safety releasewas fitted between the weight and the programmablerelease to ensure that the weight fell away in the event ofthe programmable release failing The sink-weight systemgreatly enhanced the effectiveness of under-ice work byfacilitating deployment close to the ice edge

Water-density gradientsAutosub is ballasted to be 8ndash12 kg positively buoyant inwater and final ballast adjustments were usually made onthe working site after taking water-density measurementsWhen working close to an East Greenland tidewater glaciera density difference of 4 kgmndash3 between the surface and 6mwas present This density difference produced a 10 kgincrease in buoyancy for the vehicle at its working depthand was on the verge of making the vehicle unable tocontrol its depth The solution adopted to cope with these

density variations was to fit lsquowingletsrsquo (160mm half-span by254mm chord) slightly aft of the centre of gravity of thevehicle to produce additional downward force whilemoving These proved effective and had the added benefitof reducing body pitch angle

Surfacing in ice and homing systemAutosub had two Argos satellite transmitter antennae andone WiFi Ethernet radio antenna mounted externally Thesewere vulnerable and could be damaged by the ice before thevehicle had been located visually potentially resulting invehicle loss simply because it could not be located Evenwhen ice cover was very light it was clear that antennaecould be broken easily and the vehicle could become ice-covered (Fig 3) Autosub is particularly dependent on theArgos transmissions for location at the end of missions and aGFRP tube was made to fit over the forward antenna to givesome protection against ice damage The experience high-lighted the need to control the final stages of the end of themission remotely rather than trying to second-guess iceconditions at the mission programming stage

On occasions due to drifting sea ice it was not possibleat the time of launch to be certain that the recovery positionwould be clear of ice Further there was a need to be able tocope with situations in which the mother ship could notreach the intended recovery position or when unexpectednavigation system drift or some other failure could leavethe AUV still operating but a long way off the intendedrecovery position

To deal with these eventualities a homing system wasdeveloped able to guide Autosub back towards the mothership at ranges of up to 15 km The shipboard homing beacontransmits regularly spaced swept frequencies (chirps)centred at 4504 kHz On the vehicle were three sphericalhydrophones which together with the three-channel correl-ation receiver allowed the direction of arrival of the homingsignal to be determined Once the system had detected fourconsecutive transmissions with the correct temporal spa-cing the AUV entered homing mode and headed towardsthe source of the signal This system was used successfullyduring the northeast Greenland campaign of 2004 where onmission 365 the intended recovery position had beencovered by sea ice as a result of changing wind conditions

INSTRUMENTATION FOR THE AUTOSUB AUVGeophysical instrumentsA swath-bathymetry system or multibeam echo sounder wasincluded in the Autosub instrument package to measure thegeometry of both the underside of ice shelves and sea icetogether with sea-floor morphology The system could beconfigured to operate in upward- or downward-lookingmode for glaciological and glacialndashgeological applica-tions respectively A Kongsberg Simrad EM-2000 swath-bathymetry system was used It operates at a frequency of200 kHz has 111 individual beams an angular coverage ofover 1208 and a swath width of up to 300m (depending ondistance from target) Quantitative data on elevation can begridded at a horizontal resolution of 1m Root-mean square(rms) errors of better than 10 cm can be achieved in thevertical A swath-bathymetry system was preferred to aconventional marine side-scan sonar instrument because itprovides quantitative three-dimensional (3-D) morphologic-al data of high absolute accuracy rather than imaging

Fig 3 Autosub surfacing in light sea-ice cover Note the possibilityof damage to antennae protruding from the vehicle

Dowdeswell and others Instruments and methods664

changes in backscatter that relate to both geometry andsurface properties

An EdgeTech chirp penetration echo sounder was alsomounted on Autosub to investigate the acoustic stratigraphyof the sea floor The profiler operates at 2ndash16 kHz andpenetrates through up to several tens of metres of sedimentdepending on grain-size density and pore-water character-istics with a vertical resolution of 6ndash10 cm Both geophysicalinstruments log data to internal hard drives for the durationof each Autosub mission

Oceanographic instrumentsAutosub was equipped with a Seabird 911 plus CTD systemwhich included two pairs of temperature and conductivitysensors A dissolved oxygen sensor was also attachedalthough for accurate oxygen measurements this needed tobe calibrated against laboratory measurements on concur-rent water samples The Seabird conductivity and tempera-ture sensors were in general remarkably stable so that withtwo pairs and regular calibration high accuracy measure-ments were possible The specified drift rate for thetemperature and conductivity sensors was 00028Candash1 and00024 siemensmndash1 andash1 (00023mndash1 andash1) respectively Thetwo essentially independent sensor pairs provided a checkon the data quality and we have typically found the pairsdiffered by no more than 00018C in temperature and nomore than 0002 in derived salinity The deployment of ashipboard CTD before or after an Autosub mission provideda cross-check that the Autosub CTD sensors were makingreasonable measurements The Autosub CTD data wereprocessed using the software provided by the manufacturerfollowing the standard processing pathway set out in theuser manual for the software This included calculatingsalinity and other derived variables

Upward- and downward-looking Teledyne RDI ADCPswere fitted which were used both for navigation andscientific measurements The downward-looking 150 kHzinstrument typically returned current measurements up to150ndash200m below Autosub The upward-looking 300 kHzinstrument typically provided more limited range up toabout 100m above Autosub The Autosub ADCP data wereprocessed using a system of dedicated MATLABTM scripts

Photographic instrumentsAutosub was equipped with a Starlight SXV-H9 which is ablack-and-white charge-coupled device (CCD) imager toobtain images of the sea floor and the marine benthos closeto and beneath floating ice The imager was selected for itshigh sensitivity (particularly in the important blue part of thespectrum) low readout noise (about 20 photons rms equiva-lent) and high dynamic range An integral data loggerrecords the images on hard disk which can be accessed viathe Autosub radio network The camera is installed in the tailsection of Autosub and a Minolta zoom flash is installedwithin a pressure case at the nose of Autosub The imagesensor has an array of 1040 1392 of 6 mm square pixelsmaking an imager size of 624 835mm With an airndashwater magnification factor of 14 this equates to an imagediagonal of 22m at a flying altitude of 10m The cameradata are stored in a raw 16-bit binary format

Water-sampling instrumentsAutosub carried a compact water sampler to allow themeasurement of a wide range of water properties The

sampler was an Envirotech AquaLab which consists of amechanical syringe that draws water into one of 49 EVAplastic sample bags by means of a rotary valve (Fig 4)Samples are suitable for most tracer and nutrient studies butnot for the measurement of trace gases due to the gas-permeable nature of the EVA bags used

Before deployment sample bags must be filled with asmall volume of lsquoprimersquo fluid so that hydrostatic forces donot crush connecting parts Ordinarily this fluid is flushedout of each bag in situ before a sample is collectedHowever this time-consuming procedure can be avoided ifbags are filled with a prime fluid in which the properties ofinterest are known and later accounted for (Dodd andothers 2006) This approach allows a 200mL sample to becollected in 8ndash10min during which Autosub would traveljust under 1 km at a cruising speed of 16m sndash1 Samples ofany size can be collected but multiple syringe strokes arerequired to collect samples larger than 200mL

The water sampler is capable of operating in an autono-mous mode in which samples are collected at predeter-mined times or it can be triggered by Autosub to collectsamples at specific locations It is also possible for Autosubto circle while a sample is collected and continue onlywhen the sampler reports that sampling is complete Todate the sampler has provided samples for oxygen-isotoperatio and barium concentration measurements (Dodd andothers 2006)

AUV OBSERVATIONS OF ICE OCEAN AND SEAFLOOR SOME EXAMPLESIce shelves the underside of an Antarctic ice shelfTraditional methods for determining the topography of anice-shelf base have used either downward-looking radarfrom above the upper surface of the ice shelf generally froman aircraft platform or the inversion of elevation data ofthe upper surface under the assumption that the ice columnfloats in hydrostatic equilibrium (Bamber and Bentley 1994Dowdeswell and Bamber 2007) Both techniques havetheir advantages inversion of (typically) satellite-derivedsurface elevations gives very good spatial coverage athorizontal scales somewhat longer than the ice is thick

Fig 4 The AquaLab water sampler located in the nose of theAutosub vehicle Individual water-sample bags are visible Photo-graph by P Dodd

Dowdeswell and others Instruments and methods 665

downward-looking radar gives a detailed view at horizontalscales determined by the ice thickness and the wavelengthof the radar waves in ice Reliance on the results from thesetechniques has reinforced the notion that the base of iceshelves is generally rather smooth and can be regarded ashaving a drag coefficient at the icendashwater interface similar tothat of the sea floor Neither technique however is able toreveal basal topography at the scales important to thefriction exerted on water flow beneath the ice shelf aparameter important to the realistic modelling of flowbeneath the ice shelf

Autosubrsquos mission 382 beneath Fimbul Ice Shelf EastAntarctica yielded the first acoustic image of the base of anice shelf (Fig 5) The upward-looking multibeam echosounder gave a 150m wide image of the base of this iceshelf along 18 km of the mission track For most of the swaththe vehicle was 90m below the ice base The draft of the iceshelf as seen by the multibeam system is shown in Figure 5aThe breadth of the swath is an indication of the specularity ofthe ice base with a broader swath indicative of a rougherbase Much of the swath suggests an almost specularreflection consistent with the generally held view that thebase of an ice shelf is rather smooth (Holland and Feltham2006) A substantial fraction however is very roughFigure 5b shows a 3-D visualization of a rough portion ofthe swath from 189 to 215 km along track illustrating thatthe basal ice-shelf topography is quite chaotic at horizontallength scales of 10m or less with vertical scales similarly ofthe order of 10m In reality the image in Figure 5 is asubstantially smoothed visualization as the extreme natureof the terrain caused excessive shadowing which has beenfilled-in in a smooth manner

The rough portions of the swath data correspond on theice-shelf surface to flow traces These flow traces are linear

features visible from aerial or satellite imagery (Fahnestockand others 2000) that are often initiated at glacial featuressuch as shear margins or associated with regions of fastflow Flow traces are ubiquitous on ice shelves and if theyare generally underlain by an ice base with such dramatictopography it is clear that the frictional drag imposed on theocean circulation beneath the ice shelf needs to bereassessed (Nicholls and others 2006)

Sea ice a three-dimensional viewIn August 2004 the Autosub AUV in operations off northeastGreenland obtained the first successful multibeam sonarmeasurements under sea ice giving a quantitative map ofthe 3-D nature of the under-ice surface (Wadhams andothers 2006) The vehicle operating from RRS James ClarkRoss obtained more than 450 track-km of under-ice multi-beam sonar data using the Kongsberg EM-2000 systemFigure 6 shows examples of imagery from first- and multi-year ice including young ridges old hummocks andundeformed melting ice The imagery was obtained frommission 365 on 21ndash22 August 2004 which headed westacross the shallow Belgica Bank at 79830rsquoN under partiallygrounded multi-year ice then penetrated further over the500m deep Norske Trough occupied mainly by unde-formed first-year fast ice Each of the displayed images is aperspective view of the underside of the ice obtained withthe AUV at 40m depth with scenes shown as if illuminatedby a sun of elevation 208

Two swath-bathymetric images of the underside ofArctic sea ice are shown Figure 6a illustrates the deepestridge encountered during mission 365 which has a 33mdraft This ridge is embedded in a larger multi-year floe(from 3200 to 3800m) that probably drifted out from theArctic Ocean the previous summer The undeformed ice

Fig 5 (a) Multibeam data from mission 382 beneath Fimbul Ice Shelf East Antarctica showing ice-shelf draft (m) the track starting beneaththe ice shelf (0 km) and ending at the ice front (26 km) (b) 3-D rendering of swath-bathymetric data showing a portion of the ice base belowa flow trace at 20 km including the smooth base either side of the feature (from Nicholls and others 2006)

Dowdeswell and others Instruments and methods666

Fig 6 Examples of EM-2000 swath-bathymetric images of the under surface of sea ice offshore of northeast Greenland The perspectiveviews are illuminated by a sun elevation of 208 (a) An embedded multi-year floe with a 33m deep sea-ice ridge The floe is surrounded byundeformed shorefast sea ice (b) A multi-year ridged floe of draft 3ndash5m embedded in undeformed shorefast ice of draft 18m Fast iceshows a pattern of depressions due to mirroring of surface melt pools The floe contains a pressure ridge of maximum draft 11m which haspartly disintegrated into individual ice blocks of diameter 5ndash20m (from Wadhams and others 2006)

Dowdeswell and others Instruments and methods 667

surrounding this floe is 175m in draft and is almostcertainly first-year ice Since the individual ice blocks thatmake up ridges are quite small the ridge is a relativelyuniform triangle in cross-section representing the angle ofrepose of a pile of buoyant ice A number of thinner floes10ndash15m in draft are also visible in the image

Figure 6b shows an old multi-year ridged floe of thickness3ndash5m which is embedded in younger fast ice of draft 18mThe edges of the floe are sharp and linear as would occurwith a fracture that occurred just before embedding Theridge which occupies half of the floe has maximum draft of11m and contains separate ice blocks of typical diameters5ndash20m In the ice surrounding the embedded floe a numberof small floes with drafts of about 10m are present Thefaint pattern of depressions in the underside of the thinnerice occurs because of the presence of meltwater pools on theupper surface These pools preferentially absorb incoming

radiation giving a heat flux that enhances bottom melt andgenerates a bottom depression which mirrors the position ofeach pool on the top side (Wadhams and Martin 1990Wadhams 2000)

Oceanography the nature of a water-filled cavitybeneath an ice shelfThe majority of Antarctic Bottom Water (AABW) is thoughtto have its origins in processes that take place over theAntarctic continental shelf These processes therefore reflectthe importance of AABW as a key component in the globalthermohaline circulation As a consequence interactionsbetween the Southern Ocean and Antarctic ice shelveswhich cover 40 of the Antarctic continental shelf are alsoimportant Historically exploration of the processes beneathice shelves has been restricted to what can be achieved bydrilling access holes and deploying oceanographic instru-mentation into the water column beneath The process ofmaking access holes is demanding logistically and a rathersmall number of holes can be made in any given Antarcticfield season In fact fewer than 30 access points have beenmade across all ice shelves in Antarctica Clearly AUVs offeran opportunity to improve substantially our ability to obtaindata from this unique environment

During mission 382 to the cavity beneath Fimbul Ice ShelfAutosub executed a simple inndashout track with a total tracklength of 60 km 53 km of which was beneath the ice shelfThe in-going track was at an elevation above the seabed of150m The vehicle then turned on a reciprocal trackascending to an elevation of 400m The fact that the seabedshallows towards the ice front combined with an overridinginstruction to maintain a minimum headroom from the icebase of 90m meant that Autosub was terrain-following offthe base of the ice shelf for much of the return track

The temperature salinity and current-speed data obtainedfrom the primary oceanographic instruments during themission are shown in Figure 7 These data exhibit a wealth ofdetail and are discussed by Nicholls and others (2006) inthe context of data obtained from the front of the ice shelfusing the ship The principal conclusion of Nicholls andothers (2006) was that as the properties of some of thewaters observed within the cavity did not relate to the watersobserved along the front of Fimbul Ice Shelf at the time ofthe mission the cavity must be flushed episodically byrelatively warm water that crosses the continental-shelfbreak from the north possibly during the winter

An intriguing dataset acquired by the Autosub ADCPs isshown in Figure 7a The effective range of an ADCP in largepart depends on the number and type of scatterers in thewater column and their size with respect to the wavelengthsin the acoustic pulse With a wavelength of 10mm the150 kHz downward-looking instrument generally has agreater range than its 300 kHz upward-looking counterpartThis can be seen outside the cavity on the left side ofFigure 7a Once Autosub has passed beyond one or two tidalexcursions into the cavity (a distance of about 3 km) therange of both instruments decreases markedly and the perfor-mance of the 300 kHz ADCP overtakes that of the 150 kHzinstrument The performance of the ADCPs indicates adifferent biological assemblage beneath the ice shelfimplying a reduction in the volume density of biologicalmaterial and a shift towards smaller-sized scatterers Therapid fall-off with distance into the cavity of scatterer volumedensity also suggests that this is an area of outflow

Fig 7 Oceanographic data from mission 382 obtained beneathFimbul Ice Shelf Antarctica (a) Mission trajectory (red and bluelines indicate the outward and return Autosub legs respectively)The vertical dashed line at 265 km gives the position of the icefront referenced to the turning point in the mission the horizontaldashed line at 570m depth shows the depth of a nearby sill at thecontinental-shelf break Also shown are the ADCP data illustratingthe dramatically reduced range beneath the ice shelf that implies adearth of appropriately sized scatterers in the water column Theupward-looking instrument operated at 300 kHz and the down-ward-looking instrument at 150 kHz The data are for the northndashsouth velocity component (positive northward approximatelyperpendicular to the ice front) which have been averaged using ahorizontal window 100m wide The inset shows the ADCP data inthe vicinity of the ice front for the outward leg (b) Verticallyaveraged ADCP currents after subtraction of the modelled tide(c) Salinity (bold) and potential temperature () The thin green near-horizontal dashed line is the freezing point of the water at surfacepressure for salinities measured on the outward journey (fromNicholls and others 2006)

Dowdeswell and others Instruments and methods668

Although investigation of the ocean processes withincavities beneath ice shelves will always require mooredinstruments capable of collecting data over periods ofmonths or years Autosubrsquos ability as a platform that canuse sophisticated oceanographic instrumentation to gener-ate spatially extensive datasets has given us a unique view ofone of the least accessible parts of the worldrsquos oceans

Autosub was lost under the Fimbul Ice Shelf on mission383 the one following that described above Its low-frequency acoustic beacon signalled that an abort had beentriggered and that the vehicle was stuck at a positionapproximately 17 km from the ice front A full investigation(Strutt 2006) concluded that either an open-circuit ornetwork failure was the most likely cause of an abort orloss of power This was the only time the vehicle deployed itslong-range acoustic beacon

Oceanography fjord circulation and meltwater fluxThe circulation and mixing processes of water masses withinfjords can be complex so one advantage of a rapid andcontinuous surveying device such as Autosub is to enablea more synoptic survey than is achievable with a shipTypically saline ocean water enters a fjord at mid-depthabove the sill and fresh meltwater from the surroundingglaciers or rivers exits the fjord as a surface layer (Syvitski andothers 1987) The deep waters within the fjord are renewedonly sporadically However this steady-state simple picturecan be complicated by the presence of tides cross-fjordflows sea ice entering andor leaving the fjord and the flowsinduced by inertial oscillations following storms The netexchange of fresh water between a fjord and the continental-shelf environment is of importance in determining forexample the influence of meltwater from the Greenlandice sheet on the formation of dense water masses in the seassurrounding Greenland

The detailed current-velocity structure revealed by theAutosub ADCPs at the mouth of Kangerdlussuaq Fjord onthe east coast of Greenland is shown in Figure 8 In the

6 hours of this survey three passes across the fjord weremade at depths of 70 190 and 400m The upward- anddownward-looking ADCPs are very consistent betweenadjacent passes some 4hours apart implying that tidal (orother temporally varying) flows are not dominant hereHowever the velocity structure is very different from thesimple three-layer flow suggested above The primaryinflows are on the southwest side of the mouth at 300ndash400m and in the upper 100m The primary outflow is atabout 200ndash350m on the northwest side of the fjord mouthimplying a clockwise circulation of open ocean water in thebay at the mouth of the fjord There is a suggestion that thewater in the top 10m may be a thin layer of ice melt leavingthe fjord Thus Autosub has revealed in unprecedenteddetail a snapshot of the complex exchanges between anArctic fjord environment and the adjacent continental shelf

Oceanography attenuation of waves by sea iceA serendipitous result for the behaviour of waves propa-gating in sea ice was obtained from the upward-lookingADCP surface track velocity recorded on Autosub Becausethe surface track ping has longer range than the profile pingthe velocity of sea ice relative to Autosub could be measuredduring runs as deep as 200m This was the first use of anAUV to measure directional and scalar wave propertiesduring surface wave propagation through sea ice (Hayes andothers 2007) Since ice-edge detection was also possiblefrom the surface track ping (verified by ship observations)dependence of the above wave properties on distance fromthe edge of the marginal ice zone could be examined

As an example during mission 324 on 25 March 2003 inthe marginal ice zone of the Bellingshausen Sea Antarcticathe horizontal velocity of the ice was oscillating Themagnitude of this oscillation also decayed with distancefrom the ice edge both on the inward and outward segments(Fig 9a) In the observed regime of small ice floes (lt20m)and long wavelength (100ndash350m) the floes nearly follow thecircular path of a point on the water surface Therefore the

Fig 8 Cross-section of the current velocity (colour scale in m sndash1)into and out of the mouth of Kangerdlussuaq Fjord East Greenlandacquired from the upward- and downward-looking ADCPs mountedon Autosub The Autosub navigated horizontal paths at 70 190 and400m (marked as black lines) descending or rising in betweentaking 6 hours to complete the survey Positive values denote waterflowing into the fjord negative values indicate water flowing out ofthe fjord Southwest is to the left and northeast to the right

Fig 9 Sea-ice velocity from Autosub mission 324 The upward-looking ADCP measured the surface track velocity upon (a) enteringthe ice pack at 90m and (b) exiting the ice pack at 90m Themagnified inset shows a typical segment analyzed here Note thestrong periodicity in both components as well as mean currenttowards the southeast (modified from Hayes and others 2007)

Dowdeswell and others Instruments and methods 669

surface track velocity is regarded as a measurement of sur-face wave orbital velocity superimposed on mean icevelocity (southeastward in the case of mission 324) Theseries is divided into a number of blocks (Fig 9b) to analyzethe surface velocity The directional and scalar wave spectraare calculated for each segment so any trend in significantwave height mean and peak wave periods as well as anychange in the energy wave direction or spread of variousfrequency components can be detected (Fig 10) The char-acter of waves propagating through sea ice that was observedusing Autosub agrees with most of the previous observational(Wadhams and others 1986 1988 Liu and others 1991) andnumerical (Meylan and others 1997) experiments

Glacial geology submarine glacial landforms andacoustic stratigraphyThe morphology and stratigraphy of the sea floor provideimportant evidence for the reconstruction of the dimensionsand flow of former ice sheets (eg Anderson 1999) Whereice flows across a sedimentary bed landforms diagnostic ofice-flow direction and dynamics are produced These land-forms which are often streamlined are preserved underwater as ice retreats across continental shelves and fjordsduring interglacial and interstadial periods (eg Andersonand others 2002 Ottesen and others 2005 Evans andothers 2006) Characteristic assemblages of these submarinelandforms are indicators of for example ice-stream flowpast glacier-surge activity and former grounding lines (egPowell and others 1996 Canals and others 2000 O Cofaighand others 2002 Ottesen and Dowdeswell 2006)

The swath-bathymetry system on Autosub when mountedin downward-looking mode produces data that yield high-resolution digital-terrain models and 3-D images of the seafloor Figure 11 shows the floor of an East Greenland fjordwhere the fast-flowing Kangerdlussuaq Glacier one of themajor outlet glaciers of the Greenland ice sheet (Rignot andKanagaratnam 2006) has produced streamlined sediment-ary bedforms which are preserved in several hundred metres

of water after ice retreat from its position at the Last GlacialMaximum (Syvitski and others 1996) Shallow acousticstratigraphy provides further information on the structure ofthe upper few metres to tens of metres of sediment In theexample shown in Figure 11b the acoustic profiler onAutosub penetrates the fine-grained and acoustically lamin-ated sediments in the deepest part of Kangerdlussuaq Fjordwith less transparent and probably coarser-grained sedi-ments characteristic of glacial diamicts or tills to either sideAutosub which has been deployed close to the calvingtidewater margins of Courtauld Glacier East Greenland(Fig 1a) can be used to image areas of the sea floor inpreviously inaccessible locations near calving ice cliffs andbeneath ice shelves

In addition to geophysical instruments the digital cameraequipment on Autosub provides detailed information on theform and composition of the sea floor and the marine biotathat inhabit it Figure 12 shows an example of a sea-floorphotograph acquired by Autosub in Kangerdlussuaq FjordBoth individual dropstones released by iceberg melting andbottom-dwelling marine organisms are shown The presenceof deposit-feeding species is indicated by faunal traces on thesediment surface Evidence of disturbance to the seabed andfauna from iceberg-keel ploughing was also observed inphotographs of the sea floor at water depths less than about500m reducing faunal density and diversity as well asproducing a sedimentologically heterogeneous environment

CONCLUSIONSThe Autosub AUV provided a platform for the deployment ofa number of geophysical and oceanographic instruments inhazardous polar environments that ships and other mannedvehicles cannot access

Fig 10 (a) Mean wave period and (b) significant wave height forAutosub missions 322ndash324 The label lsquoinrsquo refers to the seriescollected upon entering the ice pack while lsquobackrsquo refers to thereturn series Period and wave height are derived from the one-dimensional wave spectrum of 512 s blocks (with the exception ofthe return trip in mission 323 in which 256 s blocks were analyzed)

Fig 11Multibeam echo-sounder image of the glacially streamlinedsea floor of Kangerdlussuaq Fjord acquired from a 200 kHz swath-bathymetry system mounted on Autosub The swath width isapproximately 200m Water depth is 710ndash840m The swath-bathymetry data are gridded at a resolution of 1m in the horizontalThe lower panels show acoustically stratified sediments on the fjordfloor acquired from the chirp 2ndash16 kHz sub-bottom profiler onAutosub The acoustic profile is located in the multibeam image

Dowdeswell and others Instruments and methods670

Ice-covered environments investigated using Autosubinclude a cavity beneath the Fimbul Ice Shelf and therelatively shallow and poorly charted waters beneath sea iceon the East Greenland continental shelf

The multibeam echo sounder of Autosub has imaged theunderside of an ice shelf for the first time showing that someareas are very rough with implications for the modelling ofwater flow and melt rates The underside of sea ice has alsobeen imaged in detail and quantitative shape parametersextracted Swath images and bottom photographs of theglacial geology and marine biota close to the margins ofArctic tidewater glaciers have also been obtained

Oceanographic data such as salinity temperature andwater velocity have been derived continuously during Auto-sub missions beneath floating Arctic and Antarctic ice pro-viding observations with a very dense spatial coverage inenvironments where previously few or no data have beenavailable

The Autosub3 vehicle successor to the lost Autosub2and AUV technology in general is likely to be used in-creasingly in hazardous polar marine environments for thecollection of detailed geophysical and oceanographic dataclose to and beneath floating ice These data in turn areimportant in the calibration and testing of numerical modelsrelating to ice-sheet interactions with the polar waters

Not all of the scenarios for AUV operations in polar seashave yet been achieved in practice although many havebeen described and discussed by scientists and engineers(Griffiths and Collins 2007 Collins and Griffiths 2008) InAugust 2007 the first AUV campaign took place to searchfor and then examine hydrothermal sites at the slow-spreading Gakkel Ridge in the Arctic Ocean an area ofextensive multi-year pack ice (Reves-Sohn and others 2007)Other plans include multidisciplinary studies beneath theRoss Ice Shelf Antarctica and surveys of Southern Oceankrill populations in winter

ACKNOWLEDGEMENTSThis work was supported by the Autosub Under Ice The-matic Programme of the UK Natural Environment ResearchCouncil (Programme Chair S Ackley Programme ManagerK Collins) We are grateful to the Autosub Technical Teamand the officers and crew of RRS James Clark Ross for theirinvaluable contributions on four Autosub cruises

REFERENCESAnderson JB 1999 Antarctic marine geology Cambridge etc

Cambridge University PressAnderson JB SS Shipp AL Lowe JS Wellner and AB Mosola

2002 The Antarctic ice sheet during the last glacial maximumand its subsequent retreat history a review Quat Sci Rev21(1ndash3) 49ndash70

Bamber JL and CR Bentley 1994 A comparison of satellite-altimetry and ice-thickness measurements of the Ross Ice ShelfAntarctica Ann Glaciol 20 357ndash364

Brierley AS and 11 others 2002 Antarctic krill under sea iceelevated abundance in a narrow band just south of ice edgeScience 295(5561) 1890ndash1892

Broecker WS 1991 The great ocean conveyor Oceanography4(2) 79ndash89

Canals M R Urgeles and AM Calafat 2000 Deep sea-floorevidence of past ice streams off the Antarctic PeninsulaGeology 28(1) 31ndash34

Cavalieri DJ CL Parkinson and KY Vinnikov 2003 30-Yearsatellite record reveals contrasting Arctic and Antarctic decadalsea ice variability Geophys Res Lett 30(18) 1970 (1010292003GL018031)

Collins K and G Griffiths eds 2008 Workshop on AUV sciencein extreme environments collaborative Autosub science inextreme environments Proceedings of the International ScienceWorkship 11ndash13 April 2007 Scott Polar Research InstituteUniversity of Cambridge UK London Society for UnderwaterTechnology

Dodd PA MR Price KJ Heywood and M Pebody 2006Collection of water samples from an autonomous underwatervehicle for tracer analysis J Atmos Oceanic Technol 23(12)1759ndash1767

Dowdeswell JA and JL Bamber 2007 Keel depths of modernAntarctic icebergs and implications for sea-floor scouring in thegeological record Mar Geol 243(1ndash4) 120ndash131

Dowdeswell JA and RD Powell 1996 Submersible remotelyoperated vehicles (ROVs) for investigations of the glacierndashoceanndashsediment interface J Glaciol 42(140) 176ndash183

Evans J JA Dowdeswell C O Cofaigh TJ Benham and JB And-erson 2006 Extent and dynamics of the West Antarctic IceSheet on the outer continental shelf of Pine Island Bay during thelast glaciation Mar Geol 250(1ndash2) 53ndash72

Fahnestock MA TA Scambos RA Bindschadler and G Kvaran2000 A millennium of variable ice flow recorded by the RossIce Shelf Antarctica J Glaciol 46(155) 652ndash664

Francois RE 1977 High resolution observations of under-icemorphology Seattle WA University of Washington AppliedPhysics Laboratory Tech Rep APL-UW-7112

Griffiths G and K Collins eds 2007 Masterclass in AUVtechnology for polar science collaborative autosub science inextreme environments Proceedings of the International Master-class 28ndash30 March 2006 National Oceanography CentreSouthampton UK London Society for Underwater Technology

Hayes DR and A Jenkins 2007 Autonomous underwater vehiclemeasurements of surface wave decay and directional spectra inthe marginal sea ice zone J Phys Oceanogr 37(1) 71ndash83

Holland PR and DL Feltham 2006 The effects of rotation andice shelf topography on frazil-laden ice shelf water plumesJ Phys Oceanogr 36(12) 2312ndash2327

Fig 12 Example photograph from Autosub mission 377 showingthe floor of outer Kangerdlussuaq Fjord (imaged from an altitude of9m at a depth of 564m) One cobble-sized iceberg-rafted drop-stone three large burrows and numerous tubeworms are visibleThe photograph is about 1m across

Dowdeswell and others Instruments and methods 671

changes in backscatter that relate to both geometry andsurface properties

An EdgeTech chirp penetration echo sounder was alsomounted on Autosub to investigate the acoustic stratigraphyof the sea floor The profiler operates at 2ndash16 kHz andpenetrates through up to several tens of metres of sedimentdepending on grain-size density and pore-water character-istics with a vertical resolution of 6ndash10 cm Both geophysicalinstruments log data to internal hard drives for the durationof each Autosub mission

Oceanographic instrumentsAutosub was equipped with a Seabird 911 plus CTD systemwhich included two pairs of temperature and conductivitysensors A dissolved oxygen sensor was also attachedalthough for accurate oxygen measurements this needed tobe calibrated against laboratory measurements on concur-rent water samples The Seabird conductivity and tempera-ture sensors were in general remarkably stable so that withtwo pairs and regular calibration high accuracy measure-ments were possible The specified drift rate for thetemperature and conductivity sensors was 00028Candash1 and00024 siemensmndash1 andash1 (00023mndash1 andash1) respectively Thetwo essentially independent sensor pairs provided a checkon the data quality and we have typically found the pairsdiffered by no more than 00018C in temperature and nomore than 0002 in derived salinity The deployment of ashipboard CTD before or after an Autosub mission provideda cross-check that the Autosub CTD sensors were makingreasonable measurements The Autosub CTD data wereprocessed using the software provided by the manufacturerfollowing the standard processing pathway set out in theuser manual for the software This included calculatingsalinity and other derived variables

Upward- and downward-looking Teledyne RDI ADCPswere fitted which were used both for navigation andscientific measurements The downward-looking 150 kHzinstrument typically returned current measurements up to150ndash200m below Autosub The upward-looking 300 kHzinstrument typically provided more limited range up toabout 100m above Autosub The Autosub ADCP data wereprocessed using a system of dedicated MATLABTM scripts

Photographic instrumentsAutosub was equipped with a Starlight SXV-H9 which is ablack-and-white charge-coupled device (CCD) imager toobtain images of the sea floor and the marine benthos closeto and beneath floating ice The imager was selected for itshigh sensitivity (particularly in the important blue part of thespectrum) low readout noise (about 20 photons rms equiva-lent) and high dynamic range An integral data loggerrecords the images on hard disk which can be accessed viathe Autosub radio network The camera is installed in the tailsection of Autosub and a Minolta zoom flash is installedwithin a pressure case at the nose of Autosub The imagesensor has an array of 1040 1392 of 6 mm square pixelsmaking an imager size of 624 835mm With an airndashwater magnification factor of 14 this equates to an imagediagonal of 22m at a flying altitude of 10m The cameradata are stored in a raw 16-bit binary format

Water-sampling instrumentsAutosub carried a compact water sampler to allow themeasurement of a wide range of water properties The

sampler was an Envirotech AquaLab which consists of amechanical syringe that draws water into one of 49 EVAplastic sample bags by means of a rotary valve (Fig 4)Samples are suitable for most tracer and nutrient studies butnot for the measurement of trace gases due to the gas-permeable nature of the EVA bags used

Before deployment sample bags must be filled with asmall volume of lsquoprimersquo fluid so that hydrostatic forces donot crush connecting parts Ordinarily this fluid is flushedout of each bag in situ before a sample is collectedHowever this time-consuming procedure can be avoided ifbags are filled with a prime fluid in which the properties ofinterest are known and later accounted for (Dodd andothers 2006) This approach allows a 200mL sample to becollected in 8ndash10min during which Autosub would traveljust under 1 km at a cruising speed of 16m sndash1 Samples ofany size can be collected but multiple syringe strokes arerequired to collect samples larger than 200mL

The water sampler is capable of operating in an autono-mous mode in which samples are collected at predeter-mined times or it can be triggered by Autosub to collectsamples at specific locations It is also possible for Autosubto circle while a sample is collected and continue onlywhen the sampler reports that sampling is complete Todate the sampler has provided samples for oxygen-isotoperatio and barium concentration measurements (Dodd andothers 2006)

AUV OBSERVATIONS OF ICE OCEAN AND SEAFLOOR SOME EXAMPLESIce shelves the underside of an Antarctic ice shelfTraditional methods for determining the topography of anice-shelf base have used either downward-looking radarfrom above the upper surface of the ice shelf generally froman aircraft platform or the inversion of elevation data ofthe upper surface under the assumption that the ice columnfloats in hydrostatic equilibrium (Bamber and Bentley 1994Dowdeswell and Bamber 2007) Both techniques havetheir advantages inversion of (typically) satellite-derivedsurface elevations gives very good spatial coverage athorizontal scales somewhat longer than the ice is thick

Fig 4 The AquaLab water sampler located in the nose of theAutosub vehicle Individual water-sample bags are visible Photo-graph by P Dodd

Dowdeswell and others Instruments and methods 665

downward-looking radar gives a detailed view at horizontalscales determined by the ice thickness and the wavelengthof the radar waves in ice Reliance on the results from thesetechniques has reinforced the notion that the base of iceshelves is generally rather smooth and can be regarded ashaving a drag coefficient at the icendashwater interface similar tothat of the sea floor Neither technique however is able toreveal basal topography at the scales important to thefriction exerted on water flow beneath the ice shelf aparameter important to the realistic modelling of flowbeneath the ice shelf

Autosubrsquos mission 382 beneath Fimbul Ice Shelf EastAntarctica yielded the first acoustic image of the base of anice shelf (Fig 5) The upward-looking multibeam echosounder gave a 150m wide image of the base of this iceshelf along 18 km of the mission track For most of the swaththe vehicle was 90m below the ice base The draft of the iceshelf as seen by the multibeam system is shown in Figure 5aThe breadth of the swath is an indication of the specularity ofthe ice base with a broader swath indicative of a rougherbase Much of the swath suggests an almost specularreflection consistent with the generally held view that thebase of an ice shelf is rather smooth (Holland and Feltham2006) A substantial fraction however is very roughFigure 5b shows a 3-D visualization of a rough portion ofthe swath from 189 to 215 km along track illustrating thatthe basal ice-shelf topography is quite chaotic at horizontallength scales of 10m or less with vertical scales similarly ofthe order of 10m In reality the image in Figure 5 is asubstantially smoothed visualization as the extreme natureof the terrain caused excessive shadowing which has beenfilled-in in a smooth manner

The rough portions of the swath data correspond on theice-shelf surface to flow traces These flow traces are linear

features visible from aerial or satellite imagery (Fahnestockand others 2000) that are often initiated at glacial featuressuch as shear margins or associated with regions of fastflow Flow traces are ubiquitous on ice shelves and if theyare generally underlain by an ice base with such dramatictopography it is clear that the frictional drag imposed on theocean circulation beneath the ice shelf needs to bereassessed (Nicholls and others 2006)

Sea ice a three-dimensional viewIn August 2004 the Autosub AUV in operations off northeastGreenland obtained the first successful multibeam sonarmeasurements under sea ice giving a quantitative map ofthe 3-D nature of the under-ice surface (Wadhams andothers 2006) The vehicle operating from RRS James ClarkRoss obtained more than 450 track-km of under-ice multi-beam sonar data using the Kongsberg EM-2000 systemFigure 6 shows examples of imagery from first- and multi-year ice including young ridges old hummocks andundeformed melting ice The imagery was obtained frommission 365 on 21ndash22 August 2004 which headed westacross the shallow Belgica Bank at 79830rsquoN under partiallygrounded multi-year ice then penetrated further over the500m deep Norske Trough occupied mainly by unde-formed first-year fast ice Each of the displayed images is aperspective view of the underside of the ice obtained withthe AUV at 40m depth with scenes shown as if illuminatedby a sun of elevation 208

Two swath-bathymetric images of the underside ofArctic sea ice are shown Figure 6a illustrates the deepestridge encountered during mission 365 which has a 33mdraft This ridge is embedded in a larger multi-year floe(from 3200 to 3800m) that probably drifted out from theArctic Ocean the previous summer The undeformed ice

Fig 5 (a) Multibeam data from mission 382 beneath Fimbul Ice Shelf East Antarctica showing ice-shelf draft (m) the track starting beneaththe ice shelf (0 km) and ending at the ice front (26 km) (b) 3-D rendering of swath-bathymetric data showing a portion of the ice base belowa flow trace at 20 km including the smooth base either side of the feature (from Nicholls and others 2006)

Dowdeswell and others Instruments and methods666

Fig 6 Examples of EM-2000 swath-bathymetric images of the under surface of sea ice offshore of northeast Greenland The perspectiveviews are illuminated by a sun elevation of 208 (a) An embedded multi-year floe with a 33m deep sea-ice ridge The floe is surrounded byundeformed shorefast sea ice (b) A multi-year ridged floe of draft 3ndash5m embedded in undeformed shorefast ice of draft 18m Fast iceshows a pattern of depressions due to mirroring of surface melt pools The floe contains a pressure ridge of maximum draft 11m which haspartly disintegrated into individual ice blocks of diameter 5ndash20m (from Wadhams and others 2006)

Dowdeswell and others Instruments and methods 667

surrounding this floe is 175m in draft and is almostcertainly first-year ice Since the individual ice blocks thatmake up ridges are quite small the ridge is a relativelyuniform triangle in cross-section representing the angle ofrepose of a pile of buoyant ice A number of thinner floes10ndash15m in draft are also visible in the image

Figure 6b shows an old multi-year ridged floe of thickness3ndash5m which is embedded in younger fast ice of draft 18mThe edges of the floe are sharp and linear as would occurwith a fracture that occurred just before embedding Theridge which occupies half of the floe has maximum draft of11m and contains separate ice blocks of typical diameters5ndash20m In the ice surrounding the embedded floe a numberof small floes with drafts of about 10m are present Thefaint pattern of depressions in the underside of the thinnerice occurs because of the presence of meltwater pools on theupper surface These pools preferentially absorb incoming

radiation giving a heat flux that enhances bottom melt andgenerates a bottom depression which mirrors the position ofeach pool on the top side (Wadhams and Martin 1990Wadhams 2000)

Oceanography the nature of a water-filled cavitybeneath an ice shelfThe majority of Antarctic Bottom Water (AABW) is thoughtto have its origins in processes that take place over theAntarctic continental shelf These processes therefore reflectthe importance of AABW as a key component in the globalthermohaline circulation As a consequence interactionsbetween the Southern Ocean and Antarctic ice shelveswhich cover 40 of the Antarctic continental shelf are alsoimportant Historically exploration of the processes beneathice shelves has been restricted to what can be achieved bydrilling access holes and deploying oceanographic instru-mentation into the water column beneath The process ofmaking access holes is demanding logistically and a rathersmall number of holes can be made in any given Antarcticfield season In fact fewer than 30 access points have beenmade across all ice shelves in Antarctica Clearly AUVs offeran opportunity to improve substantially our ability to obtaindata from this unique environment

During mission 382 to the cavity beneath Fimbul Ice ShelfAutosub executed a simple inndashout track with a total tracklength of 60 km 53 km of which was beneath the ice shelfThe in-going track was at an elevation above the seabed of150m The vehicle then turned on a reciprocal trackascending to an elevation of 400m The fact that the seabedshallows towards the ice front combined with an overridinginstruction to maintain a minimum headroom from the icebase of 90m meant that Autosub was terrain-following offthe base of the ice shelf for much of the return track

The temperature salinity and current-speed data obtainedfrom the primary oceanographic instruments during themission are shown in Figure 7 These data exhibit a wealth ofdetail and are discussed by Nicholls and others (2006) inthe context of data obtained from the front of the ice shelfusing the ship The principal conclusion of Nicholls andothers (2006) was that as the properties of some of thewaters observed within the cavity did not relate to the watersobserved along the front of Fimbul Ice Shelf at the time ofthe mission the cavity must be flushed episodically byrelatively warm water that crosses the continental-shelfbreak from the north possibly during the winter

An intriguing dataset acquired by the Autosub ADCPs isshown in Figure 7a The effective range of an ADCP in largepart depends on the number and type of scatterers in thewater column and their size with respect to the wavelengthsin the acoustic pulse With a wavelength of 10mm the150 kHz downward-looking instrument generally has agreater range than its 300 kHz upward-looking counterpartThis can be seen outside the cavity on the left side ofFigure 7a Once Autosub has passed beyond one or two tidalexcursions into the cavity (a distance of about 3 km) therange of both instruments decreases markedly and the perfor-mance of the 300 kHz ADCP overtakes that of the 150 kHzinstrument The performance of the ADCPs indicates adifferent biological assemblage beneath the ice shelfimplying a reduction in the volume density of biologicalmaterial and a shift towards smaller-sized scatterers Therapid fall-off with distance into the cavity of scatterer volumedensity also suggests that this is an area of outflow

Fig 7 Oceanographic data from mission 382 obtained beneathFimbul Ice Shelf Antarctica (a) Mission trajectory (red and bluelines indicate the outward and return Autosub legs respectively)The vertical dashed line at 265 km gives the position of the icefront referenced to the turning point in the mission the horizontaldashed line at 570m depth shows the depth of a nearby sill at thecontinental-shelf break Also shown are the ADCP data illustratingthe dramatically reduced range beneath the ice shelf that implies adearth of appropriately sized scatterers in the water column Theupward-looking instrument operated at 300 kHz and the down-ward-looking instrument at 150 kHz The data are for the northndashsouth velocity component (positive northward approximatelyperpendicular to the ice front) which have been averaged using ahorizontal window 100m wide The inset shows the ADCP data inthe vicinity of the ice front for the outward leg (b) Verticallyaveraged ADCP currents after subtraction of the modelled tide(c) Salinity (bold) and potential temperature () The thin green near-horizontal dashed line is the freezing point of the water at surfacepressure for salinities measured on the outward journey (fromNicholls and others 2006)

Dowdeswell and others Instruments and methods668

Although investigation of the ocean processes withincavities beneath ice shelves will always require mooredinstruments capable of collecting data over periods ofmonths or years Autosubrsquos ability as a platform that canuse sophisticated oceanographic instrumentation to gener-ate spatially extensive datasets has given us a unique view ofone of the least accessible parts of the worldrsquos oceans

Autosub was lost under the Fimbul Ice Shelf on mission383 the one following that described above Its low-frequency acoustic beacon signalled that an abort had beentriggered and that the vehicle was stuck at a positionapproximately 17 km from the ice front A full investigation(Strutt 2006) concluded that either an open-circuit ornetwork failure was the most likely cause of an abort orloss of power This was the only time the vehicle deployed itslong-range acoustic beacon

Oceanography fjord circulation and meltwater fluxThe circulation and mixing processes of water masses withinfjords can be complex so one advantage of a rapid andcontinuous surveying device such as Autosub is to enablea more synoptic survey than is achievable with a shipTypically saline ocean water enters a fjord at mid-depthabove the sill and fresh meltwater from the surroundingglaciers or rivers exits the fjord as a surface layer (Syvitski andothers 1987) The deep waters within the fjord are renewedonly sporadically However this steady-state simple picturecan be complicated by the presence of tides cross-fjordflows sea ice entering andor leaving the fjord and the flowsinduced by inertial oscillations following storms The netexchange of fresh water between a fjord and the continental-shelf environment is of importance in determining forexample the influence of meltwater from the Greenlandice sheet on the formation of dense water masses in the seassurrounding Greenland

The detailed current-velocity structure revealed by theAutosub ADCPs at the mouth of Kangerdlussuaq Fjord onthe east coast of Greenland is shown in Figure 8 In the

6 hours of this survey three passes across the fjord weremade at depths of 70 190 and 400m The upward- anddownward-looking ADCPs are very consistent betweenadjacent passes some 4hours apart implying that tidal (orother temporally varying) flows are not dominant hereHowever the velocity structure is very different from thesimple three-layer flow suggested above The primaryinflows are on the southwest side of the mouth at 300ndash400m and in the upper 100m The primary outflow is atabout 200ndash350m on the northwest side of the fjord mouthimplying a clockwise circulation of open ocean water in thebay at the mouth of the fjord There is a suggestion that thewater in the top 10m may be a thin layer of ice melt leavingthe fjord Thus Autosub has revealed in unprecedenteddetail a snapshot of the complex exchanges between anArctic fjord environment and the adjacent continental shelf

Oceanography attenuation of waves by sea iceA serendipitous result for the behaviour of waves propa-gating in sea ice was obtained from the upward-lookingADCP surface track velocity recorded on Autosub Becausethe surface track ping has longer range than the profile pingthe velocity of sea ice relative to Autosub could be measuredduring runs as deep as 200m This was the first use of anAUV to measure directional and scalar wave propertiesduring surface wave propagation through sea ice (Hayes andothers 2007) Since ice-edge detection was also possiblefrom the surface track ping (verified by ship observations)dependence of the above wave properties on distance fromthe edge of the marginal ice zone could be examined

As an example during mission 324 on 25 March 2003 inthe marginal ice zone of the Bellingshausen Sea Antarcticathe horizontal velocity of the ice was oscillating Themagnitude of this oscillation also decayed with distancefrom the ice edge both on the inward and outward segments(Fig 9a) In the observed regime of small ice floes (lt20m)and long wavelength (100ndash350m) the floes nearly follow thecircular path of a point on the water surface Therefore the

Fig 8 Cross-section of the current velocity (colour scale in m sndash1)into and out of the mouth of Kangerdlussuaq Fjord East Greenlandacquired from the upward- and downward-looking ADCPs mountedon Autosub The Autosub navigated horizontal paths at 70 190 and400m (marked as black lines) descending or rising in betweentaking 6 hours to complete the survey Positive values denote waterflowing into the fjord negative values indicate water flowing out ofthe fjord Southwest is to the left and northeast to the right

Fig 9 Sea-ice velocity from Autosub mission 324 The upward-looking ADCP measured the surface track velocity upon (a) enteringthe ice pack at 90m and (b) exiting the ice pack at 90m Themagnified inset shows a typical segment analyzed here Note thestrong periodicity in both components as well as mean currenttowards the southeast (modified from Hayes and others 2007)

Dowdeswell and others Instruments and methods 669

surface track velocity is regarded as a measurement of sur-face wave orbital velocity superimposed on mean icevelocity (southeastward in the case of mission 324) Theseries is divided into a number of blocks (Fig 9b) to analyzethe surface velocity The directional and scalar wave spectraare calculated for each segment so any trend in significantwave height mean and peak wave periods as well as anychange in the energy wave direction or spread of variousfrequency components can be detected (Fig 10) The char-acter of waves propagating through sea ice that was observedusing Autosub agrees with most of the previous observational(Wadhams and others 1986 1988 Liu and others 1991) andnumerical (Meylan and others 1997) experiments

Glacial geology submarine glacial landforms andacoustic stratigraphyThe morphology and stratigraphy of the sea floor provideimportant evidence for the reconstruction of the dimensionsand flow of former ice sheets (eg Anderson 1999) Whereice flows across a sedimentary bed landforms diagnostic ofice-flow direction and dynamics are produced These land-forms which are often streamlined are preserved underwater as ice retreats across continental shelves and fjordsduring interglacial and interstadial periods (eg Andersonand others 2002 Ottesen and others 2005 Evans andothers 2006) Characteristic assemblages of these submarinelandforms are indicators of for example ice-stream flowpast glacier-surge activity and former grounding lines (egPowell and others 1996 Canals and others 2000 O Cofaighand others 2002 Ottesen and Dowdeswell 2006)

The swath-bathymetry system on Autosub when mountedin downward-looking mode produces data that yield high-resolution digital-terrain models and 3-D images of the seafloor Figure 11 shows the floor of an East Greenland fjordwhere the fast-flowing Kangerdlussuaq Glacier one of themajor outlet glaciers of the Greenland ice sheet (Rignot andKanagaratnam 2006) has produced streamlined sediment-ary bedforms which are preserved in several hundred metres

of water after ice retreat from its position at the Last GlacialMaximum (Syvitski and others 1996) Shallow acousticstratigraphy provides further information on the structure ofthe upper few metres to tens of metres of sediment In theexample shown in Figure 11b the acoustic profiler onAutosub penetrates the fine-grained and acoustically lamin-ated sediments in the deepest part of Kangerdlussuaq Fjordwith less transparent and probably coarser-grained sedi-ments characteristic of glacial diamicts or tills to either sideAutosub which has been deployed close to the calvingtidewater margins of Courtauld Glacier East Greenland(Fig 1a) can be used to image areas of the sea floor inpreviously inaccessible locations near calving ice cliffs andbeneath ice shelves

In addition to geophysical instruments the digital cameraequipment on Autosub provides detailed information on theform and composition of the sea floor and the marine biotathat inhabit it Figure 12 shows an example of a sea-floorphotograph acquired by Autosub in Kangerdlussuaq FjordBoth individual dropstones released by iceberg melting andbottom-dwelling marine organisms are shown The presenceof deposit-feeding species is indicated by faunal traces on thesediment surface Evidence of disturbance to the seabed andfauna from iceberg-keel ploughing was also observed inphotographs of the sea floor at water depths less than about500m reducing faunal density and diversity as well asproducing a sedimentologically heterogeneous environment

CONCLUSIONSThe Autosub AUV provided a platform for the deployment ofa number of geophysical and oceanographic instruments inhazardous polar environments that ships and other mannedvehicles cannot access

Fig 10 (a) Mean wave period and (b) significant wave height forAutosub missions 322ndash324 The label lsquoinrsquo refers to the seriescollected upon entering the ice pack while lsquobackrsquo refers to thereturn series Period and wave height are derived from the one-dimensional wave spectrum of 512 s blocks (with the exception ofthe return trip in mission 323 in which 256 s blocks were analyzed)

Fig 11Multibeam echo-sounder image of the glacially streamlinedsea floor of Kangerdlussuaq Fjord acquired from a 200 kHz swath-bathymetry system mounted on Autosub The swath width isapproximately 200m Water depth is 710ndash840m The swath-bathymetry data are gridded at a resolution of 1m in the horizontalThe lower panels show acoustically stratified sediments on the fjordfloor acquired from the chirp 2ndash16 kHz sub-bottom profiler onAutosub The acoustic profile is located in the multibeam image

Dowdeswell and others Instruments and methods670

Ice-covered environments investigated using Autosubinclude a cavity beneath the Fimbul Ice Shelf and therelatively shallow and poorly charted waters beneath sea iceon the East Greenland continental shelf

The multibeam echo sounder of Autosub has imaged theunderside of an ice shelf for the first time showing that someareas are very rough with implications for the modelling ofwater flow and melt rates The underside of sea ice has alsobeen imaged in detail and quantitative shape parametersextracted Swath images and bottom photographs of theglacial geology and marine biota close to the margins ofArctic tidewater glaciers have also been obtained

Oceanographic data such as salinity temperature andwater velocity have been derived continuously during Auto-sub missions beneath floating Arctic and Antarctic ice pro-viding observations with a very dense spatial coverage inenvironments where previously few or no data have beenavailable

The Autosub3 vehicle successor to the lost Autosub2and AUV technology in general is likely to be used in-creasingly in hazardous polar marine environments for thecollection of detailed geophysical and oceanographic dataclose to and beneath floating ice These data in turn areimportant in the calibration and testing of numerical modelsrelating to ice-sheet interactions with the polar waters

Not all of the scenarios for AUV operations in polar seashave yet been achieved in practice although many havebeen described and discussed by scientists and engineers(Griffiths and Collins 2007 Collins and Griffiths 2008) InAugust 2007 the first AUV campaign took place to searchfor and then examine hydrothermal sites at the slow-spreading Gakkel Ridge in the Arctic Ocean an area ofextensive multi-year pack ice (Reves-Sohn and others 2007)Other plans include multidisciplinary studies beneath theRoss Ice Shelf Antarctica and surveys of Southern Oceankrill populations in winter

ACKNOWLEDGEMENTSThis work was supported by the Autosub Under Ice The-matic Programme of the UK Natural Environment ResearchCouncil (Programme Chair S Ackley Programme ManagerK Collins) We are grateful to the Autosub Technical Teamand the officers and crew of RRS James Clark Ross for theirinvaluable contributions on four Autosub cruises

REFERENCESAnderson JB 1999 Antarctic marine geology Cambridge etc

Cambridge University PressAnderson JB SS Shipp AL Lowe JS Wellner and AB Mosola

2002 The Antarctic ice sheet during the last glacial maximumand its subsequent retreat history a review Quat Sci Rev21(1ndash3) 49ndash70

Bamber JL and CR Bentley 1994 A comparison of satellite-altimetry and ice-thickness measurements of the Ross Ice ShelfAntarctica Ann Glaciol 20 357ndash364

Brierley AS and 11 others 2002 Antarctic krill under sea iceelevated abundance in a narrow band just south of ice edgeScience 295(5561) 1890ndash1892

Broecker WS 1991 The great ocean conveyor Oceanography4(2) 79ndash89

Canals M R Urgeles and AM Calafat 2000 Deep sea-floorevidence of past ice streams off the Antarctic PeninsulaGeology 28(1) 31ndash34

Cavalieri DJ CL Parkinson and KY Vinnikov 2003 30-Yearsatellite record reveals contrasting Arctic and Antarctic decadalsea ice variability Geophys Res Lett 30(18) 1970 (1010292003GL018031)

Collins K and G Griffiths eds 2008 Workshop on AUV sciencein extreme environments collaborative Autosub science inextreme environments Proceedings of the International ScienceWorkship 11ndash13 April 2007 Scott Polar Research InstituteUniversity of Cambridge UK London Society for UnderwaterTechnology

Dodd PA MR Price KJ Heywood and M Pebody 2006Collection of water samples from an autonomous underwatervehicle for tracer analysis J Atmos Oceanic Technol 23(12)1759ndash1767

Dowdeswell JA and JL Bamber 2007 Keel depths of modernAntarctic icebergs and implications for sea-floor scouring in thegeological record Mar Geol 243(1ndash4) 120ndash131

Dowdeswell JA and RD Powell 1996 Submersible remotelyoperated vehicles (ROVs) for investigations of the glacierndashoceanndashsediment interface J Glaciol 42(140) 176ndash183

Evans J JA Dowdeswell C O Cofaigh TJ Benham and JB And-erson 2006 Extent and dynamics of the West Antarctic IceSheet on the outer continental shelf of Pine Island Bay during thelast glaciation Mar Geol 250(1ndash2) 53ndash72

Fahnestock MA TA Scambos RA Bindschadler and G Kvaran2000 A millennium of variable ice flow recorded by the RossIce Shelf Antarctica J Glaciol 46(155) 652ndash664

Francois RE 1977 High resolution observations of under-icemorphology Seattle WA University of Washington AppliedPhysics Laboratory Tech Rep APL-UW-7112

Griffiths G and K Collins eds 2007 Masterclass in AUVtechnology for polar science collaborative autosub science inextreme environments Proceedings of the International Master-class 28ndash30 March 2006 National Oceanography CentreSouthampton UK London Society for Underwater Technology

Hayes DR and A Jenkins 2007 Autonomous underwater vehiclemeasurements of surface wave decay and directional spectra inthe marginal sea ice zone J Phys Oceanogr 37(1) 71ndash83

Holland PR and DL Feltham 2006 The effects of rotation andice shelf topography on frazil-laden ice shelf water plumesJ Phys Oceanogr 36(12) 2312ndash2327

Fig 12 Example photograph from Autosub mission 377 showingthe floor of outer Kangerdlussuaq Fjord (imaged from an altitude of9m at a depth of 564m) One cobble-sized iceberg-rafted drop-stone three large burrows and numerous tubeworms are visibleThe photograph is about 1m across

Dowdeswell and others Instruments and methods 671

downward-looking radar gives a detailed view at horizontalscales determined by the ice thickness and the wavelengthof the radar waves in ice Reliance on the results from thesetechniques has reinforced the notion that the base of iceshelves is generally rather smooth and can be regarded ashaving a drag coefficient at the icendashwater interface similar tothat of the sea floor Neither technique however is able toreveal basal topography at the scales important to thefriction exerted on water flow beneath the ice shelf aparameter important to the realistic modelling of flowbeneath the ice shelf

Autosubrsquos mission 382 beneath Fimbul Ice Shelf EastAntarctica yielded the first acoustic image of the base of anice shelf (Fig 5) The upward-looking multibeam echosounder gave a 150m wide image of the base of this iceshelf along 18 km of the mission track For most of the swaththe vehicle was 90m below the ice base The draft of the iceshelf as seen by the multibeam system is shown in Figure 5aThe breadth of the swath is an indication of the specularity ofthe ice base with a broader swath indicative of a rougherbase Much of the swath suggests an almost specularreflection consistent with the generally held view that thebase of an ice shelf is rather smooth (Holland and Feltham2006) A substantial fraction however is very roughFigure 5b shows a 3-D visualization of a rough portion ofthe swath from 189 to 215 km along track illustrating thatthe basal ice-shelf topography is quite chaotic at horizontallength scales of 10m or less with vertical scales similarly ofthe order of 10m In reality the image in Figure 5 is asubstantially smoothed visualization as the extreme natureof the terrain caused excessive shadowing which has beenfilled-in in a smooth manner

The rough portions of the swath data correspond on theice-shelf surface to flow traces These flow traces are linear

features visible from aerial or satellite imagery (Fahnestockand others 2000) that are often initiated at glacial featuressuch as shear margins or associated with regions of fastflow Flow traces are ubiquitous on ice shelves and if theyare generally underlain by an ice base with such dramatictopography it is clear that the frictional drag imposed on theocean circulation beneath the ice shelf needs to bereassessed (Nicholls and others 2006)

Sea ice a three-dimensional viewIn August 2004 the Autosub AUV in operations off northeastGreenland obtained the first successful multibeam sonarmeasurements under sea ice giving a quantitative map ofthe 3-D nature of the under-ice surface (Wadhams andothers 2006) The vehicle operating from RRS James ClarkRoss obtained more than 450 track-km of under-ice multi-beam sonar data using the Kongsberg EM-2000 systemFigure 6 shows examples of imagery from first- and multi-year ice including young ridges old hummocks andundeformed melting ice The imagery was obtained frommission 365 on 21ndash22 August 2004 which headed westacross the shallow Belgica Bank at 79830rsquoN under partiallygrounded multi-year ice then penetrated further over the500m deep Norske Trough occupied mainly by unde-formed first-year fast ice Each of the displayed images is aperspective view of the underside of the ice obtained withthe AUV at 40m depth with scenes shown as if illuminatedby a sun of elevation 208

Two swath-bathymetric images of the underside ofArctic sea ice are shown Figure 6a illustrates the deepestridge encountered during mission 365 which has a 33mdraft This ridge is embedded in a larger multi-year floe(from 3200 to 3800m) that probably drifted out from theArctic Ocean the previous summer The undeformed ice

Fig 5 (a) Multibeam data from mission 382 beneath Fimbul Ice Shelf East Antarctica showing ice-shelf draft (m) the track starting beneaththe ice shelf (0 km) and ending at the ice front (26 km) (b) 3-D rendering of swath-bathymetric data showing a portion of the ice base belowa flow trace at 20 km including the smooth base either side of the feature (from Nicholls and others 2006)

Dowdeswell and others Instruments and methods666

Fig 6 Examples of EM-2000 swath-bathymetric images of the under surface of sea ice offshore of northeast Greenland The perspectiveviews are illuminated by a sun elevation of 208 (a) An embedded multi-year floe with a 33m deep sea-ice ridge The floe is surrounded byundeformed shorefast sea ice (b) A multi-year ridged floe of draft 3ndash5m embedded in undeformed shorefast ice of draft 18m Fast iceshows a pattern of depressions due to mirroring of surface melt pools The floe contains a pressure ridge of maximum draft 11m which haspartly disintegrated into individual ice blocks of diameter 5ndash20m (from Wadhams and others 2006)

Dowdeswell and others Instruments and methods 667

surrounding this floe is 175m in draft and is almostcertainly first-year ice Since the individual ice blocks thatmake up ridges are quite small the ridge is a relativelyuniform triangle in cross-section representing the angle ofrepose of a pile of buoyant ice A number of thinner floes10ndash15m in draft are also visible in the image

Figure 6b shows an old multi-year ridged floe of thickness3ndash5m which is embedded in younger fast ice of draft 18mThe edges of the floe are sharp and linear as would occurwith a fracture that occurred just before embedding Theridge which occupies half of the floe has maximum draft of11m and contains separate ice blocks of typical diameters5ndash20m In the ice surrounding the embedded floe a numberof small floes with drafts of about 10m are present Thefaint pattern of depressions in the underside of the thinnerice occurs because of the presence of meltwater pools on theupper surface These pools preferentially absorb incoming

radiation giving a heat flux that enhances bottom melt andgenerates a bottom depression which mirrors the position ofeach pool on the top side (Wadhams and Martin 1990Wadhams 2000)

Oceanography the nature of a water-filled cavitybeneath an ice shelfThe majority of Antarctic Bottom Water (AABW) is thoughtto have its origins in processes that take place over theAntarctic continental shelf These processes therefore reflectthe importance of AABW as a key component in the globalthermohaline circulation As a consequence interactionsbetween the Southern Ocean and Antarctic ice shelveswhich cover 40 of the Antarctic continental shelf are alsoimportant Historically exploration of the processes beneathice shelves has been restricted to what can be achieved bydrilling access holes and deploying oceanographic instru-mentation into the water column beneath The process ofmaking access holes is demanding logistically and a rathersmall number of holes can be made in any given Antarcticfield season In fact fewer than 30 access points have beenmade across all ice shelves in Antarctica Clearly AUVs offeran opportunity to improve substantially our ability to obtaindata from this unique environment

During mission 382 to the cavity beneath Fimbul Ice ShelfAutosub executed a simple inndashout track with a total tracklength of 60 km 53 km of which was beneath the ice shelfThe in-going track was at an elevation above the seabed of150m The vehicle then turned on a reciprocal trackascending to an elevation of 400m The fact that the seabedshallows towards the ice front combined with an overridinginstruction to maintain a minimum headroom from the icebase of 90m meant that Autosub was terrain-following offthe base of the ice shelf for much of the return track

The temperature salinity and current-speed data obtainedfrom the primary oceanographic instruments during themission are shown in Figure 7 These data exhibit a wealth ofdetail and are discussed by Nicholls and others (2006) inthe context of data obtained from the front of the ice shelfusing the ship The principal conclusion of Nicholls andothers (2006) was that as the properties of some of thewaters observed within the cavity did not relate to the watersobserved along the front of Fimbul Ice Shelf at the time ofthe mission the cavity must be flushed episodically byrelatively warm water that crosses the continental-shelfbreak from the north possibly during the winter

An intriguing dataset acquired by the Autosub ADCPs isshown in Figure 7a The effective range of an ADCP in largepart depends on the number and type of scatterers in thewater column and their size with respect to the wavelengthsin the acoustic pulse With a wavelength of 10mm the150 kHz downward-looking instrument generally has agreater range than its 300 kHz upward-looking counterpartThis can be seen outside the cavity on the left side ofFigure 7a Once Autosub has passed beyond one or two tidalexcursions into the cavity (a distance of about 3 km) therange of both instruments decreases markedly and the perfor-mance of the 300 kHz ADCP overtakes that of the 150 kHzinstrument The performance of the ADCPs indicates adifferent biological assemblage beneath the ice shelfimplying a reduction in the volume density of biologicalmaterial and a shift towards smaller-sized scatterers Therapid fall-off with distance into the cavity of scatterer volumedensity also suggests that this is an area of outflow

Fig 7 Oceanographic data from mission 382 obtained beneathFimbul Ice Shelf Antarctica (a) Mission trajectory (red and bluelines indicate the outward and return Autosub legs respectively)The vertical dashed line at 265 km gives the position of the icefront referenced to the turning point in the mission the horizontaldashed line at 570m depth shows the depth of a nearby sill at thecontinental-shelf break Also shown are the ADCP data illustratingthe dramatically reduced range beneath the ice shelf that implies adearth of appropriately sized scatterers in the water column Theupward-looking instrument operated at 300 kHz and the down-ward-looking instrument at 150 kHz The data are for the northndashsouth velocity component (positive northward approximatelyperpendicular to the ice front) which have been averaged using ahorizontal window 100m wide The inset shows the ADCP data inthe vicinity of the ice front for the outward leg (b) Verticallyaveraged ADCP currents after subtraction of the modelled tide(c) Salinity (bold) and potential temperature () The thin green near-horizontal dashed line is the freezing point of the water at surfacepressure for salinities measured on the outward journey (fromNicholls and others 2006)

Dowdeswell and others Instruments and methods668

Although investigation of the ocean processes withincavities beneath ice shelves will always require mooredinstruments capable of collecting data over periods ofmonths or years Autosubrsquos ability as a platform that canuse sophisticated oceanographic instrumentation to gener-ate spatially extensive datasets has given us a unique view ofone of the least accessible parts of the worldrsquos oceans

Autosub was lost under the Fimbul Ice Shelf on mission383 the one following that described above Its low-frequency acoustic beacon signalled that an abort had beentriggered and that the vehicle was stuck at a positionapproximately 17 km from the ice front A full investigation(Strutt 2006) concluded that either an open-circuit ornetwork failure was the most likely cause of an abort orloss of power This was the only time the vehicle deployed itslong-range acoustic beacon

Oceanography fjord circulation and meltwater fluxThe circulation and mixing processes of water masses withinfjords can be complex so one advantage of a rapid andcontinuous surveying device such as Autosub is to enablea more synoptic survey than is achievable with a shipTypically saline ocean water enters a fjord at mid-depthabove the sill and fresh meltwater from the surroundingglaciers or rivers exits the fjord as a surface layer (Syvitski andothers 1987) The deep waters within the fjord are renewedonly sporadically However this steady-state simple picturecan be complicated by the presence of tides cross-fjordflows sea ice entering andor leaving the fjord and the flowsinduced by inertial oscillations following storms The netexchange of fresh water between a fjord and the continental-shelf environment is of importance in determining forexample the influence of meltwater from the Greenlandice sheet on the formation of dense water masses in the seassurrounding Greenland

The detailed current-velocity structure revealed by theAutosub ADCPs at the mouth of Kangerdlussuaq Fjord onthe east coast of Greenland is shown in Figure 8 In the

6 hours of this survey three passes across the fjord weremade at depths of 70 190 and 400m The upward- anddownward-looking ADCPs are very consistent betweenadjacent passes some 4hours apart implying that tidal (orother temporally varying) flows are not dominant hereHowever the velocity structure is very different from thesimple three-layer flow suggested above The primaryinflows are on the southwest side of the mouth at 300ndash400m and in the upper 100m The primary outflow is atabout 200ndash350m on the northwest side of the fjord mouthimplying a clockwise circulation of open ocean water in thebay at the mouth of the fjord There is a suggestion that thewater in the top 10m may be a thin layer of ice melt leavingthe fjord Thus Autosub has revealed in unprecedenteddetail a snapshot of the complex exchanges between anArctic fjord environment and the adjacent continental shelf

Oceanography attenuation of waves by sea iceA serendipitous result for the behaviour of waves propa-gating in sea ice was obtained from the upward-lookingADCP surface track velocity recorded on Autosub Becausethe surface track ping has longer range than the profile pingthe velocity of sea ice relative to Autosub could be measuredduring runs as deep as 200m This was the first use of anAUV to measure directional and scalar wave propertiesduring surface wave propagation through sea ice (Hayes andothers 2007) Since ice-edge detection was also possiblefrom the surface track ping (verified by ship observations)dependence of the above wave properties on distance fromthe edge of the marginal ice zone could be examined

As an example during mission 324 on 25 March 2003 inthe marginal ice zone of the Bellingshausen Sea Antarcticathe horizontal velocity of the ice was oscillating Themagnitude of this oscillation also decayed with distancefrom the ice edge both on the inward and outward segments(Fig 9a) In the observed regime of small ice floes (lt20m)and long wavelength (100ndash350m) the floes nearly follow thecircular path of a point on the water surface Therefore the

Fig 8 Cross-section of the current velocity (colour scale in m sndash1)into and out of the mouth of Kangerdlussuaq Fjord East Greenlandacquired from the upward- and downward-looking ADCPs mountedon Autosub The Autosub navigated horizontal paths at 70 190 and400m (marked as black lines) descending or rising in betweentaking 6 hours to complete the survey Positive values denote waterflowing into the fjord negative values indicate water flowing out ofthe fjord Southwest is to the left and northeast to the right

Fig 9 Sea-ice velocity from Autosub mission 324 The upward-looking ADCP measured the surface track velocity upon (a) enteringthe ice pack at 90m and (b) exiting the ice pack at 90m Themagnified inset shows a typical segment analyzed here Note thestrong periodicity in both components as well as mean currenttowards the southeast (modified from Hayes and others 2007)

Dowdeswell and others Instruments and methods 669

surface track velocity is regarded as a measurement of sur-face wave orbital velocity superimposed on mean icevelocity (southeastward in the case of mission 324) Theseries is divided into a number of blocks (Fig 9b) to analyzethe surface velocity The directional and scalar wave spectraare calculated for each segment so any trend in significantwave height mean and peak wave periods as well as anychange in the energy wave direction or spread of variousfrequency components can be detected (Fig 10) The char-acter of waves propagating through sea ice that was observedusing Autosub agrees with most of the previous observational(Wadhams and others 1986 1988 Liu and others 1991) andnumerical (Meylan and others 1997) experiments

Glacial geology submarine glacial landforms andacoustic stratigraphyThe morphology and stratigraphy of the sea floor provideimportant evidence for the reconstruction of the dimensionsand flow of former ice sheets (eg Anderson 1999) Whereice flows across a sedimentary bed landforms diagnostic ofice-flow direction and dynamics are produced These land-forms which are often streamlined are preserved underwater as ice retreats across continental shelves and fjordsduring interglacial and interstadial periods (eg Andersonand others 2002 Ottesen and others 2005 Evans andothers 2006) Characteristic assemblages of these submarinelandforms are indicators of for example ice-stream flowpast glacier-surge activity and former grounding lines (egPowell and others 1996 Canals and others 2000 O Cofaighand others 2002 Ottesen and Dowdeswell 2006)

The swath-bathymetry system on Autosub when mountedin downward-looking mode produces data that yield high-resolution digital-terrain models and 3-D images of the seafloor Figure 11 shows the floor of an East Greenland fjordwhere the fast-flowing Kangerdlussuaq Glacier one of themajor outlet glaciers of the Greenland ice sheet (Rignot andKanagaratnam 2006) has produced streamlined sediment-ary bedforms which are preserved in several hundred metres

of water after ice retreat from its position at the Last GlacialMaximum (Syvitski and others 1996) Shallow acousticstratigraphy provides further information on the structure ofthe upper few metres to tens of metres of sediment In theexample shown in Figure 11b the acoustic profiler onAutosub penetrates the fine-grained and acoustically lamin-ated sediments in the deepest part of Kangerdlussuaq Fjordwith less transparent and probably coarser-grained sedi-ments characteristic of glacial diamicts or tills to either sideAutosub which has been deployed close to the calvingtidewater margins of Courtauld Glacier East Greenland(Fig 1a) can be used to image areas of the sea floor inpreviously inaccessible locations near calving ice cliffs andbeneath ice shelves

In addition to geophysical instruments the digital cameraequipment on Autosub provides detailed information on theform and composition of the sea floor and the marine biotathat inhabit it Figure 12 shows an example of a sea-floorphotograph acquired by Autosub in Kangerdlussuaq FjordBoth individual dropstones released by iceberg melting andbottom-dwelling marine organisms are shown The presenceof deposit-feeding species is indicated by faunal traces on thesediment surface Evidence of disturbance to the seabed andfauna from iceberg-keel ploughing was also observed inphotographs of the sea floor at water depths less than about500m reducing faunal density and diversity as well asproducing a sedimentologically heterogeneous environment

CONCLUSIONSThe Autosub AUV provided a platform for the deployment ofa number of geophysical and oceanographic instruments inhazardous polar environments that ships and other mannedvehicles cannot access

Fig 10 (a) Mean wave period and (b) significant wave height forAutosub missions 322ndash324 The label lsquoinrsquo refers to the seriescollected upon entering the ice pack while lsquobackrsquo refers to thereturn series Period and wave height are derived from the one-dimensional wave spectrum of 512 s blocks (with the exception ofthe return trip in mission 323 in which 256 s blocks were analyzed)

Fig 11Multibeam echo-sounder image of the glacially streamlinedsea floor of Kangerdlussuaq Fjord acquired from a 200 kHz swath-bathymetry system mounted on Autosub The swath width isapproximately 200m Water depth is 710ndash840m The swath-bathymetry data are gridded at a resolution of 1m in the horizontalThe lower panels show acoustically stratified sediments on the fjordfloor acquired from the chirp 2ndash16 kHz sub-bottom profiler onAutosub The acoustic profile is located in the multibeam image

Dowdeswell and others Instruments and methods670

Ice-covered environments investigated using Autosubinclude a cavity beneath the Fimbul Ice Shelf and therelatively shallow and poorly charted waters beneath sea iceon the East Greenland continental shelf

The multibeam echo sounder of Autosub has imaged theunderside of an ice shelf for the first time showing that someareas are very rough with implications for the modelling ofwater flow and melt rates The underside of sea ice has alsobeen imaged in detail and quantitative shape parametersextracted Swath images and bottom photographs of theglacial geology and marine biota close to the margins ofArctic tidewater glaciers have also been obtained

Oceanographic data such as salinity temperature andwater velocity have been derived continuously during Auto-sub missions beneath floating Arctic and Antarctic ice pro-viding observations with a very dense spatial coverage inenvironments where previously few or no data have beenavailable

The Autosub3 vehicle successor to the lost Autosub2and AUV technology in general is likely to be used in-creasingly in hazardous polar marine environments for thecollection of detailed geophysical and oceanographic dataclose to and beneath floating ice These data in turn areimportant in the calibration and testing of numerical modelsrelating to ice-sheet interactions with the polar waters

Not all of the scenarios for AUV operations in polar seashave yet been achieved in practice although many havebeen described and discussed by scientists and engineers(Griffiths and Collins 2007 Collins and Griffiths 2008) InAugust 2007 the first AUV campaign took place to searchfor and then examine hydrothermal sites at the slow-spreading Gakkel Ridge in the Arctic Ocean an area ofextensive multi-year pack ice (Reves-Sohn and others 2007)Other plans include multidisciplinary studies beneath theRoss Ice Shelf Antarctica and surveys of Southern Oceankrill populations in winter

ACKNOWLEDGEMENTSThis work was supported by the Autosub Under Ice The-matic Programme of the UK Natural Environment ResearchCouncil (Programme Chair S Ackley Programme ManagerK Collins) We are grateful to the Autosub Technical Teamand the officers and crew of RRS James Clark Ross for theirinvaluable contributions on four Autosub cruises

REFERENCESAnderson JB 1999 Antarctic marine geology Cambridge etc

Cambridge University PressAnderson JB SS Shipp AL Lowe JS Wellner and AB Mosola

2002 The Antarctic ice sheet during the last glacial maximumand its subsequent retreat history a review Quat Sci Rev21(1ndash3) 49ndash70

Bamber JL and CR Bentley 1994 A comparison of satellite-altimetry and ice-thickness measurements of the Ross Ice ShelfAntarctica Ann Glaciol 20 357ndash364

Brierley AS and 11 others 2002 Antarctic krill under sea iceelevated abundance in a narrow band just south of ice edgeScience 295(5561) 1890ndash1892

Broecker WS 1991 The great ocean conveyor Oceanography4(2) 79ndash89

Canals M R Urgeles and AM Calafat 2000 Deep sea-floorevidence of past ice streams off the Antarctic PeninsulaGeology 28(1) 31ndash34

Cavalieri DJ CL Parkinson and KY Vinnikov 2003 30-Yearsatellite record reveals contrasting Arctic and Antarctic decadalsea ice variability Geophys Res Lett 30(18) 1970 (1010292003GL018031)

Collins K and G Griffiths eds 2008 Workshop on AUV sciencein extreme environments collaborative Autosub science inextreme environments Proceedings of the International ScienceWorkship 11ndash13 April 2007 Scott Polar Research InstituteUniversity of Cambridge UK London Society for UnderwaterTechnology

Dodd PA MR Price KJ Heywood and M Pebody 2006Collection of water samples from an autonomous underwatervehicle for tracer analysis J Atmos Oceanic Technol 23(12)1759ndash1767

Dowdeswell JA and JL Bamber 2007 Keel depths of modernAntarctic icebergs and implications for sea-floor scouring in thegeological record Mar Geol 243(1ndash4) 120ndash131

Dowdeswell JA and RD Powell 1996 Submersible remotelyoperated vehicles (ROVs) for investigations of the glacierndashoceanndashsediment interface J Glaciol 42(140) 176ndash183

Evans J JA Dowdeswell C O Cofaigh TJ Benham and JB And-erson 2006 Extent and dynamics of the West Antarctic IceSheet on the outer continental shelf of Pine Island Bay during thelast glaciation Mar Geol 250(1ndash2) 53ndash72

Fahnestock MA TA Scambos RA Bindschadler and G Kvaran2000 A millennium of variable ice flow recorded by the RossIce Shelf Antarctica J Glaciol 46(155) 652ndash664

Francois RE 1977 High resolution observations of under-icemorphology Seattle WA University of Washington AppliedPhysics Laboratory Tech Rep APL-UW-7112

Griffiths G and K Collins eds 2007 Masterclass in AUVtechnology for polar science collaborative autosub science inextreme environments Proceedings of the International Master-class 28ndash30 March 2006 National Oceanography CentreSouthampton UK London Society for Underwater Technology

Hayes DR and A Jenkins 2007 Autonomous underwater vehiclemeasurements of surface wave decay and directional spectra inthe marginal sea ice zone J Phys Oceanogr 37(1) 71ndash83

Holland PR and DL Feltham 2006 The effects of rotation andice shelf topography on frazil-laden ice shelf water plumesJ Phys Oceanogr 36(12) 2312ndash2327

Fig 12 Example photograph from Autosub mission 377 showingthe floor of outer Kangerdlussuaq Fjord (imaged from an altitude of9m at a depth of 564m) One cobble-sized iceberg-rafted drop-stone three large burrows and numerous tubeworms are visibleThe photograph is about 1m across

Dowdeswell and others Instruments and methods 671

Fig 6 Examples of EM-2000 swath-bathymetric images of the under surface of sea ice offshore of northeast Greenland The perspectiveviews are illuminated by a sun elevation of 208 (a) An embedded multi-year floe with a 33m deep sea-ice ridge The floe is surrounded byundeformed shorefast sea ice (b) A multi-year ridged floe of draft 3ndash5m embedded in undeformed shorefast ice of draft 18m Fast iceshows a pattern of depressions due to mirroring of surface melt pools The floe contains a pressure ridge of maximum draft 11m which haspartly disintegrated into individual ice blocks of diameter 5ndash20m (from Wadhams and others 2006)

Dowdeswell and others Instruments and methods 667

surrounding this floe is 175m in draft and is almostcertainly first-year ice Since the individual ice blocks thatmake up ridges are quite small the ridge is a relativelyuniform triangle in cross-section representing the angle ofrepose of a pile of buoyant ice A number of thinner floes10ndash15m in draft are also visible in the image

Figure 6b shows an old multi-year ridged floe of thickness3ndash5m which is embedded in younger fast ice of draft 18mThe edges of the floe are sharp and linear as would occurwith a fracture that occurred just before embedding Theridge which occupies half of the floe has maximum draft of11m and contains separate ice blocks of typical diameters5ndash20m In the ice surrounding the embedded floe a numberof small floes with drafts of about 10m are present Thefaint pattern of depressions in the underside of the thinnerice occurs because of the presence of meltwater pools on theupper surface These pools preferentially absorb incoming

radiation giving a heat flux that enhances bottom melt andgenerates a bottom depression which mirrors the position ofeach pool on the top side (Wadhams and Martin 1990Wadhams 2000)

Oceanography the nature of a water-filled cavitybeneath an ice shelfThe majority of Antarctic Bottom Water (AABW) is thoughtto have its origins in processes that take place over theAntarctic continental shelf These processes therefore reflectthe importance of AABW as a key component in the globalthermohaline circulation As a consequence interactionsbetween the Southern Ocean and Antarctic ice shelveswhich cover 40 of the Antarctic continental shelf are alsoimportant Historically exploration of the processes beneathice shelves has been restricted to what can be achieved bydrilling access holes and deploying oceanographic instru-mentation into the water column beneath The process ofmaking access holes is demanding logistically and a rathersmall number of holes can be made in any given Antarcticfield season In fact fewer than 30 access points have beenmade across all ice shelves in Antarctica Clearly AUVs offeran opportunity to improve substantially our ability to obtaindata from this unique environment

During mission 382 to the cavity beneath Fimbul Ice ShelfAutosub executed a simple inndashout track with a total tracklength of 60 km 53 km of which was beneath the ice shelfThe in-going track was at an elevation above the seabed of150m The vehicle then turned on a reciprocal trackascending to an elevation of 400m The fact that the seabedshallows towards the ice front combined with an overridinginstruction to maintain a minimum headroom from the icebase of 90m meant that Autosub was terrain-following offthe base of the ice shelf for much of the return track

The temperature salinity and current-speed data obtainedfrom the primary oceanographic instruments during themission are shown in Figure 7 These data exhibit a wealth ofdetail and are discussed by Nicholls and others (2006) inthe context of data obtained from the front of the ice shelfusing the ship The principal conclusion of Nicholls andothers (2006) was that as the properties of some of thewaters observed within the cavity did not relate to the watersobserved along the front of Fimbul Ice Shelf at the time ofthe mission the cavity must be flushed episodically byrelatively warm water that crosses the continental-shelfbreak from the north possibly during the winter

An intriguing dataset acquired by the Autosub ADCPs isshown in Figure 7a The effective range of an ADCP in largepart depends on the number and type of scatterers in thewater column and their size with respect to the wavelengthsin the acoustic pulse With a wavelength of 10mm the150 kHz downward-looking instrument generally has agreater range than its 300 kHz upward-looking counterpartThis can be seen outside the cavity on the left side ofFigure 7a Once Autosub has passed beyond one or two tidalexcursions into the cavity (a distance of about 3 km) therange of both instruments decreases markedly and the perfor-mance of the 300 kHz ADCP overtakes that of the 150 kHzinstrument The performance of the ADCPs indicates adifferent biological assemblage beneath the ice shelfimplying a reduction in the volume density of biologicalmaterial and a shift towards smaller-sized scatterers Therapid fall-off with distance into the cavity of scatterer volumedensity also suggests that this is an area of outflow

Fig 7 Oceanographic data from mission 382 obtained beneathFimbul Ice Shelf Antarctica (a) Mission trajectory (red and bluelines indicate the outward and return Autosub legs respectively)The vertical dashed line at 265 km gives the position of the icefront referenced to the turning point in the mission the horizontaldashed line at 570m depth shows the depth of a nearby sill at thecontinental-shelf break Also shown are the ADCP data illustratingthe dramatically reduced range beneath the ice shelf that implies adearth of appropriately sized scatterers in the water column Theupward-looking instrument operated at 300 kHz and the down-ward-looking instrument at 150 kHz The data are for the northndashsouth velocity component (positive northward approximatelyperpendicular to the ice front) which have been averaged using ahorizontal window 100m wide The inset shows the ADCP data inthe vicinity of the ice front for the outward leg (b) Verticallyaveraged ADCP currents after subtraction of the modelled tide(c) Salinity (bold) and potential temperature () The thin green near-horizontal dashed line is the freezing point of the water at surfacepressure for salinities measured on the outward journey (fromNicholls and others 2006)

Dowdeswell and others Instruments and methods668

Although investigation of the ocean processes withincavities beneath ice shelves will always require mooredinstruments capable of collecting data over periods ofmonths or years Autosubrsquos ability as a platform that canuse sophisticated oceanographic instrumentation to gener-ate spatially extensive datasets has given us a unique view ofone of the least accessible parts of the worldrsquos oceans

Autosub was lost under the Fimbul Ice Shelf on mission383 the one following that described above Its low-frequency acoustic beacon signalled that an abort had beentriggered and that the vehicle was stuck at a positionapproximately 17 km from the ice front A full investigation(Strutt 2006) concluded that either an open-circuit ornetwork failure was the most likely cause of an abort orloss of power This was the only time the vehicle deployed itslong-range acoustic beacon

Oceanography fjord circulation and meltwater fluxThe circulation and mixing processes of water masses withinfjords can be complex so one advantage of a rapid andcontinuous surveying device such as Autosub is to enablea more synoptic survey than is achievable with a shipTypically saline ocean water enters a fjord at mid-depthabove the sill and fresh meltwater from the surroundingglaciers or rivers exits the fjord as a surface layer (Syvitski andothers 1987) The deep waters within the fjord are renewedonly sporadically However this steady-state simple picturecan be complicated by the presence of tides cross-fjordflows sea ice entering andor leaving the fjord and the flowsinduced by inertial oscillations following storms The netexchange of fresh water between a fjord and the continental-shelf environment is of importance in determining forexample the influence of meltwater from the Greenlandice sheet on the formation of dense water masses in the seassurrounding Greenland

The detailed current-velocity structure revealed by theAutosub ADCPs at the mouth of Kangerdlussuaq Fjord onthe east coast of Greenland is shown in Figure 8 In the

6 hours of this survey three passes across the fjord weremade at depths of 70 190 and 400m The upward- anddownward-looking ADCPs are very consistent betweenadjacent passes some 4hours apart implying that tidal (orother temporally varying) flows are not dominant hereHowever the velocity structure is very different from thesimple three-layer flow suggested above The primaryinflows are on the southwest side of the mouth at 300ndash400m and in the upper 100m The primary outflow is atabout 200ndash350m on the northwest side of the fjord mouthimplying a clockwise circulation of open ocean water in thebay at the mouth of the fjord There is a suggestion that thewater in the top 10m may be a thin layer of ice melt leavingthe fjord Thus Autosub has revealed in unprecedenteddetail a snapshot of the complex exchanges between anArctic fjord environment and the adjacent continental shelf

Oceanography attenuation of waves by sea iceA serendipitous result for the behaviour of waves propa-gating in sea ice was obtained from the upward-lookingADCP surface track velocity recorded on Autosub Becausethe surface track ping has longer range than the profile pingthe velocity of sea ice relative to Autosub could be measuredduring runs as deep as 200m This was the first use of anAUV to measure directional and scalar wave propertiesduring surface wave propagation through sea ice (Hayes andothers 2007) Since ice-edge detection was also possiblefrom the surface track ping (verified by ship observations)dependence of the above wave properties on distance fromthe edge of the marginal ice zone could be examined

As an example during mission 324 on 25 March 2003 inthe marginal ice zone of the Bellingshausen Sea Antarcticathe horizontal velocity of the ice was oscillating Themagnitude of this oscillation also decayed with distancefrom the ice edge both on the inward and outward segments(Fig 9a) In the observed regime of small ice floes (lt20m)and long wavelength (100ndash350m) the floes nearly follow thecircular path of a point on the water surface Therefore the

Fig 8 Cross-section of the current velocity (colour scale in m sndash1)into and out of the mouth of Kangerdlussuaq Fjord East Greenlandacquired from the upward- and downward-looking ADCPs mountedon Autosub The Autosub navigated horizontal paths at 70 190 and400m (marked as black lines) descending or rising in betweentaking 6 hours to complete the survey Positive values denote waterflowing into the fjord negative values indicate water flowing out ofthe fjord Southwest is to the left and northeast to the right

Fig 9 Sea-ice velocity from Autosub mission 324 The upward-looking ADCP measured the surface track velocity upon (a) enteringthe ice pack at 90m and (b) exiting the ice pack at 90m Themagnified inset shows a typical segment analyzed here Note thestrong periodicity in both components as well as mean currenttowards the southeast (modified from Hayes and others 2007)

Dowdeswell and others Instruments and methods 669

surface track velocity is regarded as a measurement of sur-face wave orbital velocity superimposed on mean icevelocity (southeastward in the case of mission 324) Theseries is divided into a number of blocks (Fig 9b) to analyzethe surface velocity The directional and scalar wave spectraare calculated for each segment so any trend in significantwave height mean and peak wave periods as well as anychange in the energy wave direction or spread of variousfrequency components can be detected (Fig 10) The char-acter of waves propagating through sea ice that was observedusing Autosub agrees with most of the previous observational(Wadhams and others 1986 1988 Liu and others 1991) andnumerical (Meylan and others 1997) experiments

Glacial geology submarine glacial landforms andacoustic stratigraphyThe morphology and stratigraphy of the sea floor provideimportant evidence for the reconstruction of the dimensionsand flow of former ice sheets (eg Anderson 1999) Whereice flows across a sedimentary bed landforms diagnostic ofice-flow direction and dynamics are produced These land-forms which are often streamlined are preserved underwater as ice retreats across continental shelves and fjordsduring interglacial and interstadial periods (eg Andersonand others 2002 Ottesen and others 2005 Evans andothers 2006) Characteristic assemblages of these submarinelandforms are indicators of for example ice-stream flowpast glacier-surge activity and former grounding lines (egPowell and others 1996 Canals and others 2000 O Cofaighand others 2002 Ottesen and Dowdeswell 2006)

The swath-bathymetry system on Autosub when mountedin downward-looking mode produces data that yield high-resolution digital-terrain models and 3-D images of the seafloor Figure 11 shows the floor of an East Greenland fjordwhere the fast-flowing Kangerdlussuaq Glacier one of themajor outlet glaciers of the Greenland ice sheet (Rignot andKanagaratnam 2006) has produced streamlined sediment-ary bedforms which are preserved in several hundred metres

of water after ice retreat from its position at the Last GlacialMaximum (Syvitski and others 1996) Shallow acousticstratigraphy provides further information on the structure ofthe upper few metres to tens of metres of sediment In theexample shown in Figure 11b the acoustic profiler onAutosub penetrates the fine-grained and acoustically lamin-ated sediments in the deepest part of Kangerdlussuaq Fjordwith less transparent and probably coarser-grained sedi-ments characteristic of glacial diamicts or tills to either sideAutosub which has been deployed close to the calvingtidewater margins of Courtauld Glacier East Greenland(Fig 1a) can be used to image areas of the sea floor inpreviously inaccessible locations near calving ice cliffs andbeneath ice shelves

In addition to geophysical instruments the digital cameraequipment on Autosub provides detailed information on theform and composition of the sea floor and the marine biotathat inhabit it Figure 12 shows an example of a sea-floorphotograph acquired by Autosub in Kangerdlussuaq FjordBoth individual dropstones released by iceberg melting andbottom-dwelling marine organisms are shown The presenceof deposit-feeding species is indicated by faunal traces on thesediment surface Evidence of disturbance to the seabed andfauna from iceberg-keel ploughing was also observed inphotographs of the sea floor at water depths less than about500m reducing faunal density and diversity as well asproducing a sedimentologically heterogeneous environment

CONCLUSIONSThe Autosub AUV provided a platform for the deployment ofa number of geophysical and oceanographic instruments inhazardous polar environments that ships and other mannedvehicles cannot access

Fig 10 (a) Mean wave period and (b) significant wave height forAutosub missions 322ndash324 The label lsquoinrsquo refers to the seriescollected upon entering the ice pack while lsquobackrsquo refers to thereturn series Period and wave height are derived from the one-dimensional wave spectrum of 512 s blocks (with the exception ofthe return trip in mission 323 in which 256 s blocks were analyzed)

Fig 11Multibeam echo-sounder image of the glacially streamlinedsea floor of Kangerdlussuaq Fjord acquired from a 200 kHz swath-bathymetry system mounted on Autosub The swath width isapproximately 200m Water depth is 710ndash840m The swath-bathymetry data are gridded at a resolution of 1m in the horizontalThe lower panels show acoustically stratified sediments on the fjordfloor acquired from the chirp 2ndash16 kHz sub-bottom profiler onAutosub The acoustic profile is located in the multibeam image

Dowdeswell and others Instruments and methods670

Ice-covered environments investigated using Autosubinclude a cavity beneath the Fimbul Ice Shelf and therelatively shallow and poorly charted waters beneath sea iceon the East Greenland continental shelf

The multibeam echo sounder of Autosub has imaged theunderside of an ice shelf for the first time showing that someareas are very rough with implications for the modelling ofwater flow and melt rates The underside of sea ice has alsobeen imaged in detail and quantitative shape parametersextracted Swath images and bottom photographs of theglacial geology and marine biota close to the margins ofArctic tidewater glaciers have also been obtained

Oceanographic data such as salinity temperature andwater velocity have been derived continuously during Auto-sub missions beneath floating Arctic and Antarctic ice pro-viding observations with a very dense spatial coverage inenvironments where previously few or no data have beenavailable

The Autosub3 vehicle successor to the lost Autosub2and AUV technology in general is likely to be used in-creasingly in hazardous polar marine environments for thecollection of detailed geophysical and oceanographic dataclose to and beneath floating ice These data in turn areimportant in the calibration and testing of numerical modelsrelating to ice-sheet interactions with the polar waters

Not all of the scenarios for AUV operations in polar seashave yet been achieved in practice although many havebeen described and discussed by scientists and engineers(Griffiths and Collins 2007 Collins and Griffiths 2008) InAugust 2007 the first AUV campaign took place to searchfor and then examine hydrothermal sites at the slow-spreading Gakkel Ridge in the Arctic Ocean an area ofextensive multi-year pack ice (Reves-Sohn and others 2007)Other plans include multidisciplinary studies beneath theRoss Ice Shelf Antarctica and surveys of Southern Oceankrill populations in winter

ACKNOWLEDGEMENTSThis work was supported by the Autosub Under Ice The-matic Programme of the UK Natural Environment ResearchCouncil (Programme Chair S Ackley Programme ManagerK Collins) We are grateful to the Autosub Technical Teamand the officers and crew of RRS James Clark Ross for theirinvaluable contributions on four Autosub cruises

REFERENCESAnderson JB 1999 Antarctic marine geology Cambridge etc

Cambridge University PressAnderson JB SS Shipp AL Lowe JS Wellner and AB Mosola

2002 The Antarctic ice sheet during the last glacial maximumand its subsequent retreat history a review Quat Sci Rev21(1ndash3) 49ndash70

Bamber JL and CR Bentley 1994 A comparison of satellite-altimetry and ice-thickness measurements of the Ross Ice ShelfAntarctica Ann Glaciol 20 357ndash364

Brierley AS and 11 others 2002 Antarctic krill under sea iceelevated abundance in a narrow band just south of ice edgeScience 295(5561) 1890ndash1892

Broecker WS 1991 The great ocean conveyor Oceanography4(2) 79ndash89

Canals M R Urgeles and AM Calafat 2000 Deep sea-floorevidence of past ice streams off the Antarctic PeninsulaGeology 28(1) 31ndash34

Cavalieri DJ CL Parkinson and KY Vinnikov 2003 30-Yearsatellite record reveals contrasting Arctic and Antarctic decadalsea ice variability Geophys Res Lett 30(18) 1970 (1010292003GL018031)

Collins K and G Griffiths eds 2008 Workshop on AUV sciencein extreme environments collaborative Autosub science inextreme environments Proceedings of the International ScienceWorkship 11ndash13 April 2007 Scott Polar Research InstituteUniversity of Cambridge UK London Society for UnderwaterTechnology

Dodd PA MR Price KJ Heywood and M Pebody 2006Collection of water samples from an autonomous underwatervehicle for tracer analysis J Atmos Oceanic Technol 23(12)1759ndash1767

Dowdeswell JA and JL Bamber 2007 Keel depths of modernAntarctic icebergs and implications for sea-floor scouring in thegeological record Mar Geol 243(1ndash4) 120ndash131

Dowdeswell JA and RD Powell 1996 Submersible remotelyoperated vehicles (ROVs) for investigations of the glacierndashoceanndashsediment interface J Glaciol 42(140) 176ndash183

Evans J JA Dowdeswell C O Cofaigh TJ Benham and JB And-erson 2006 Extent and dynamics of the West Antarctic IceSheet on the outer continental shelf of Pine Island Bay during thelast glaciation Mar Geol 250(1ndash2) 53ndash72

Fahnestock MA TA Scambos RA Bindschadler and G Kvaran2000 A millennium of variable ice flow recorded by the RossIce Shelf Antarctica J Glaciol 46(155) 652ndash664

Francois RE 1977 High resolution observations of under-icemorphology Seattle WA University of Washington AppliedPhysics Laboratory Tech Rep APL-UW-7112

Griffiths G and K Collins eds 2007 Masterclass in AUVtechnology for polar science collaborative autosub science inextreme environments Proceedings of the International Master-class 28ndash30 March 2006 National Oceanography CentreSouthampton UK London Society for Underwater Technology

Hayes DR and A Jenkins 2007 Autonomous underwater vehiclemeasurements of surface wave decay and directional spectra inthe marginal sea ice zone J Phys Oceanogr 37(1) 71ndash83

Holland PR and DL Feltham 2006 The effects of rotation andice shelf topography on frazil-laden ice shelf water plumesJ Phys Oceanogr 36(12) 2312ndash2327

Fig 12 Example photograph from Autosub mission 377 showingthe floor of outer Kangerdlussuaq Fjord (imaged from an altitude of9m at a depth of 564m) One cobble-sized iceberg-rafted drop-stone three large burrows and numerous tubeworms are visibleThe photograph is about 1m across

Dowdeswell and others Instruments and methods 671

surrounding this floe is 175m in draft and is almostcertainly first-year ice Since the individual ice blocks thatmake up ridges are quite small the ridge is a relativelyuniform triangle in cross-section representing the angle ofrepose of a pile of buoyant ice A number of thinner floes10ndash15m in draft are also visible in the image

Figure 6b shows an old multi-year ridged floe of thickness3ndash5m which is embedded in younger fast ice of draft 18mThe edges of the floe are sharp and linear as would occurwith a fracture that occurred just before embedding Theridge which occupies half of the floe has maximum draft of11m and contains separate ice blocks of typical diameters5ndash20m In the ice surrounding the embedded floe a numberof small floes with drafts of about 10m are present Thefaint pattern of depressions in the underside of the thinnerice occurs because of the presence of meltwater pools on theupper surface These pools preferentially absorb incoming

radiation giving a heat flux that enhances bottom melt andgenerates a bottom depression which mirrors the position ofeach pool on the top side (Wadhams and Martin 1990Wadhams 2000)

Oceanography the nature of a water-filled cavitybeneath an ice shelfThe majority of Antarctic Bottom Water (AABW) is thoughtto have its origins in processes that take place over theAntarctic continental shelf These processes therefore reflectthe importance of AABW as a key component in the globalthermohaline circulation As a consequence interactionsbetween the Southern Ocean and Antarctic ice shelveswhich cover 40 of the Antarctic continental shelf are alsoimportant Historically exploration of the processes beneathice shelves has been restricted to what can be achieved bydrilling access holes and deploying oceanographic instru-mentation into the water column beneath The process ofmaking access holes is demanding logistically and a rathersmall number of holes can be made in any given Antarcticfield season In fact fewer than 30 access points have beenmade across all ice shelves in Antarctica Clearly AUVs offeran opportunity to improve substantially our ability to obtaindata from this unique environment

During mission 382 to the cavity beneath Fimbul Ice ShelfAutosub executed a simple inndashout track with a total tracklength of 60 km 53 km of which was beneath the ice shelfThe in-going track was at an elevation above the seabed of150m The vehicle then turned on a reciprocal trackascending to an elevation of 400m The fact that the seabedshallows towards the ice front combined with an overridinginstruction to maintain a minimum headroom from the icebase of 90m meant that Autosub was terrain-following offthe base of the ice shelf for much of the return track

The temperature salinity and current-speed data obtainedfrom the primary oceanographic instruments during themission are shown in Figure 7 These data exhibit a wealth ofdetail and are discussed by Nicholls and others (2006) inthe context of data obtained from the front of the ice shelfusing the ship The principal conclusion of Nicholls andothers (2006) was that as the properties of some of thewaters observed within the cavity did not relate to the watersobserved along the front of Fimbul Ice Shelf at the time ofthe mission the cavity must be flushed episodically byrelatively warm water that crosses the continental-shelfbreak from the north possibly during the winter

An intriguing dataset acquired by the Autosub ADCPs isshown in Figure 7a The effective range of an ADCP in largepart depends on the number and type of scatterers in thewater column and their size with respect to the wavelengthsin the acoustic pulse With a wavelength of 10mm the150 kHz downward-looking instrument generally has agreater range than its 300 kHz upward-looking counterpartThis can be seen outside the cavity on the left side ofFigure 7a Once Autosub has passed beyond one or two tidalexcursions into the cavity (a distance of about 3 km) therange of both instruments decreases markedly and the perfor-mance of the 300 kHz ADCP overtakes that of the 150 kHzinstrument The performance of the ADCPs indicates adifferent biological assemblage beneath the ice shelfimplying a reduction in the volume density of biologicalmaterial and a shift towards smaller-sized scatterers Therapid fall-off with distance into the cavity of scatterer volumedensity also suggests that this is an area of outflow

Fig 7 Oceanographic data from mission 382 obtained beneathFimbul Ice Shelf Antarctica (a) Mission trajectory (red and bluelines indicate the outward and return Autosub legs respectively)The vertical dashed line at 265 km gives the position of the icefront referenced to the turning point in the mission the horizontaldashed line at 570m depth shows the depth of a nearby sill at thecontinental-shelf break Also shown are the ADCP data illustratingthe dramatically reduced range beneath the ice shelf that implies adearth of appropriately sized scatterers in the water column Theupward-looking instrument operated at 300 kHz and the down-ward-looking instrument at 150 kHz The data are for the northndashsouth velocity component (positive northward approximatelyperpendicular to the ice front) which have been averaged using ahorizontal window 100m wide The inset shows the ADCP data inthe vicinity of the ice front for the outward leg (b) Verticallyaveraged ADCP currents after subtraction of the modelled tide(c) Salinity (bold) and potential temperature () The thin green near-horizontal dashed line is the freezing point of the water at surfacepressure for salinities measured on the outward journey (fromNicholls and others 2006)

Dowdeswell and others Instruments and methods668

Although investigation of the ocean processes withincavities beneath ice shelves will always require mooredinstruments capable of collecting data over periods ofmonths or years Autosubrsquos ability as a platform that canuse sophisticated oceanographic instrumentation to gener-ate spatially extensive datasets has given us a unique view ofone of the least accessible parts of the worldrsquos oceans

Autosub was lost under the Fimbul Ice Shelf on mission383 the one following that described above Its low-frequency acoustic beacon signalled that an abort had beentriggered and that the vehicle was stuck at a positionapproximately 17 km from the ice front A full investigation(Strutt 2006) concluded that either an open-circuit ornetwork failure was the most likely cause of an abort orloss of power This was the only time the vehicle deployed itslong-range acoustic beacon

Oceanography fjord circulation and meltwater fluxThe circulation and mixing processes of water masses withinfjords can be complex so one advantage of a rapid andcontinuous surveying device such as Autosub is to enablea more synoptic survey than is achievable with a shipTypically saline ocean water enters a fjord at mid-depthabove the sill and fresh meltwater from the surroundingglaciers or rivers exits the fjord as a surface layer (Syvitski andothers 1987) The deep waters within the fjord are renewedonly sporadically However this steady-state simple picturecan be complicated by the presence of tides cross-fjordflows sea ice entering andor leaving the fjord and the flowsinduced by inertial oscillations following storms The netexchange of fresh water between a fjord and the continental-shelf environment is of importance in determining forexample the influence of meltwater from the Greenlandice sheet on the formation of dense water masses in the seassurrounding Greenland

The detailed current-velocity structure revealed by theAutosub ADCPs at the mouth of Kangerdlussuaq Fjord onthe east coast of Greenland is shown in Figure 8 In the

6 hours of this survey three passes across the fjord weremade at depths of 70 190 and 400m The upward- anddownward-looking ADCPs are very consistent betweenadjacent passes some 4hours apart implying that tidal (orother temporally varying) flows are not dominant hereHowever the velocity structure is very different from thesimple three-layer flow suggested above The primaryinflows are on the southwest side of the mouth at 300ndash400m and in the upper 100m The primary outflow is atabout 200ndash350m on the northwest side of the fjord mouthimplying a clockwise circulation of open ocean water in thebay at the mouth of the fjord There is a suggestion that thewater in the top 10m may be a thin layer of ice melt leavingthe fjord Thus Autosub has revealed in unprecedenteddetail a snapshot of the complex exchanges between anArctic fjord environment and the adjacent continental shelf

Oceanography attenuation of waves by sea iceA serendipitous result for the behaviour of waves propa-gating in sea ice was obtained from the upward-lookingADCP surface track velocity recorded on Autosub Becausethe surface track ping has longer range than the profile pingthe velocity of sea ice relative to Autosub could be measuredduring runs as deep as 200m This was the first use of anAUV to measure directional and scalar wave propertiesduring surface wave propagation through sea ice (Hayes andothers 2007) Since ice-edge detection was also possiblefrom the surface track ping (verified by ship observations)dependence of the above wave properties on distance fromthe edge of the marginal ice zone could be examined

As an example during mission 324 on 25 March 2003 inthe marginal ice zone of the Bellingshausen Sea Antarcticathe horizontal velocity of the ice was oscillating Themagnitude of this oscillation also decayed with distancefrom the ice edge both on the inward and outward segments(Fig 9a) In the observed regime of small ice floes (lt20m)and long wavelength (100ndash350m) the floes nearly follow thecircular path of a point on the water surface Therefore the

Fig 8 Cross-section of the current velocity (colour scale in m sndash1)into and out of the mouth of Kangerdlussuaq Fjord East Greenlandacquired from the upward- and downward-looking ADCPs mountedon Autosub The Autosub navigated horizontal paths at 70 190 and400m (marked as black lines) descending or rising in betweentaking 6 hours to complete the survey Positive values denote waterflowing into the fjord negative values indicate water flowing out ofthe fjord Southwest is to the left and northeast to the right

Fig 9 Sea-ice velocity from Autosub mission 324 The upward-looking ADCP measured the surface track velocity upon (a) enteringthe ice pack at 90m and (b) exiting the ice pack at 90m Themagnified inset shows a typical segment analyzed here Note thestrong periodicity in both components as well as mean currenttowards the southeast (modified from Hayes and others 2007)

Dowdeswell and others Instruments and methods 669

surface track velocity is regarded as a measurement of sur-face wave orbital velocity superimposed on mean icevelocity (southeastward in the case of mission 324) Theseries is divided into a number of blocks (Fig 9b) to analyzethe surface velocity The directional and scalar wave spectraare calculated for each segment so any trend in significantwave height mean and peak wave periods as well as anychange in the energy wave direction or spread of variousfrequency components can be detected (Fig 10) The char-acter of waves propagating through sea ice that was observedusing Autosub agrees with most of the previous observational(Wadhams and others 1986 1988 Liu and others 1991) andnumerical (Meylan and others 1997) experiments

Glacial geology submarine glacial landforms andacoustic stratigraphyThe morphology and stratigraphy of the sea floor provideimportant evidence for the reconstruction of the dimensionsand flow of former ice sheets (eg Anderson 1999) Whereice flows across a sedimentary bed landforms diagnostic ofice-flow direction and dynamics are produced These land-forms which are often streamlined are preserved underwater as ice retreats across continental shelves and fjordsduring interglacial and interstadial periods (eg Andersonand others 2002 Ottesen and others 2005 Evans andothers 2006) Characteristic assemblages of these submarinelandforms are indicators of for example ice-stream flowpast glacier-surge activity and former grounding lines (egPowell and others 1996 Canals and others 2000 O Cofaighand others 2002 Ottesen and Dowdeswell 2006)

The swath-bathymetry system on Autosub when mountedin downward-looking mode produces data that yield high-resolution digital-terrain models and 3-D images of the seafloor Figure 11 shows the floor of an East Greenland fjordwhere the fast-flowing Kangerdlussuaq Glacier one of themajor outlet glaciers of the Greenland ice sheet (Rignot andKanagaratnam 2006) has produced streamlined sediment-ary bedforms which are preserved in several hundred metres

of water after ice retreat from its position at the Last GlacialMaximum (Syvitski and others 1996) Shallow acousticstratigraphy provides further information on the structure ofthe upper few metres to tens of metres of sediment In theexample shown in Figure 11b the acoustic profiler onAutosub penetrates the fine-grained and acoustically lamin-ated sediments in the deepest part of Kangerdlussuaq Fjordwith less transparent and probably coarser-grained sedi-ments characteristic of glacial diamicts or tills to either sideAutosub which has been deployed close to the calvingtidewater margins of Courtauld Glacier East Greenland(Fig 1a) can be used to image areas of the sea floor inpreviously inaccessible locations near calving ice cliffs andbeneath ice shelves

In addition to geophysical instruments the digital cameraequipment on Autosub provides detailed information on theform and composition of the sea floor and the marine biotathat inhabit it Figure 12 shows an example of a sea-floorphotograph acquired by Autosub in Kangerdlussuaq FjordBoth individual dropstones released by iceberg melting andbottom-dwelling marine organisms are shown The presenceof deposit-feeding species is indicated by faunal traces on thesediment surface Evidence of disturbance to the seabed andfauna from iceberg-keel ploughing was also observed inphotographs of the sea floor at water depths less than about500m reducing faunal density and diversity as well asproducing a sedimentologically heterogeneous environment

CONCLUSIONSThe Autosub AUV provided a platform for the deployment ofa number of geophysical and oceanographic instruments inhazardous polar environments that ships and other mannedvehicles cannot access

Fig 10 (a) Mean wave period and (b) significant wave height forAutosub missions 322ndash324 The label lsquoinrsquo refers to the seriescollected upon entering the ice pack while lsquobackrsquo refers to thereturn series Period and wave height are derived from the one-dimensional wave spectrum of 512 s blocks (with the exception ofthe return trip in mission 323 in which 256 s blocks were analyzed)

Fig 11Multibeam echo-sounder image of the glacially streamlinedsea floor of Kangerdlussuaq Fjord acquired from a 200 kHz swath-bathymetry system mounted on Autosub The swath width isapproximately 200m Water depth is 710ndash840m The swath-bathymetry data are gridded at a resolution of 1m in the horizontalThe lower panels show acoustically stratified sediments on the fjordfloor acquired from the chirp 2ndash16 kHz sub-bottom profiler onAutosub The acoustic profile is located in the multibeam image

Dowdeswell and others Instruments and methods670

Ice-covered environments investigated using Autosubinclude a cavity beneath the Fimbul Ice Shelf and therelatively shallow and poorly charted waters beneath sea iceon the East Greenland continental shelf

The multibeam echo sounder of Autosub has imaged theunderside of an ice shelf for the first time showing that someareas are very rough with implications for the modelling ofwater flow and melt rates The underside of sea ice has alsobeen imaged in detail and quantitative shape parametersextracted Swath images and bottom photographs of theglacial geology and marine biota close to the margins ofArctic tidewater glaciers have also been obtained

Oceanographic data such as salinity temperature andwater velocity have been derived continuously during Auto-sub missions beneath floating Arctic and Antarctic ice pro-viding observations with a very dense spatial coverage inenvironments where previously few or no data have beenavailable

The Autosub3 vehicle successor to the lost Autosub2and AUV technology in general is likely to be used in-creasingly in hazardous polar marine environments for thecollection of detailed geophysical and oceanographic dataclose to and beneath floating ice These data in turn areimportant in the calibration and testing of numerical modelsrelating to ice-sheet interactions with the polar waters

Not all of the scenarios for AUV operations in polar seashave yet been achieved in practice although many havebeen described and discussed by scientists and engineers(Griffiths and Collins 2007 Collins and Griffiths 2008) InAugust 2007 the first AUV campaign took place to searchfor and then examine hydrothermal sites at the slow-spreading Gakkel Ridge in the Arctic Ocean an area ofextensive multi-year pack ice (Reves-Sohn and others 2007)Other plans include multidisciplinary studies beneath theRoss Ice Shelf Antarctica and surveys of Southern Oceankrill populations in winter

ACKNOWLEDGEMENTSThis work was supported by the Autosub Under Ice The-matic Programme of the UK Natural Environment ResearchCouncil (Programme Chair S Ackley Programme ManagerK Collins) We are grateful to the Autosub Technical Teamand the officers and crew of RRS James Clark Ross for theirinvaluable contributions on four Autosub cruises

REFERENCESAnderson JB 1999 Antarctic marine geology Cambridge etc

Cambridge University PressAnderson JB SS Shipp AL Lowe JS Wellner and AB Mosola

2002 The Antarctic ice sheet during the last glacial maximumand its subsequent retreat history a review Quat Sci Rev21(1ndash3) 49ndash70

Bamber JL and CR Bentley 1994 A comparison of satellite-altimetry and ice-thickness measurements of the Ross Ice ShelfAntarctica Ann Glaciol 20 357ndash364

Brierley AS and 11 others 2002 Antarctic krill under sea iceelevated abundance in a narrow band just south of ice edgeScience 295(5561) 1890ndash1892

Broecker WS 1991 The great ocean conveyor Oceanography4(2) 79ndash89

Canals M R Urgeles and AM Calafat 2000 Deep sea-floorevidence of past ice streams off the Antarctic PeninsulaGeology 28(1) 31ndash34

Cavalieri DJ CL Parkinson and KY Vinnikov 2003 30-Yearsatellite record reveals contrasting Arctic and Antarctic decadalsea ice variability Geophys Res Lett 30(18) 1970 (1010292003GL018031)

Collins K and G Griffiths eds 2008 Workshop on AUV sciencein extreme environments collaborative Autosub science inextreme environments Proceedings of the International ScienceWorkship 11ndash13 April 2007 Scott Polar Research InstituteUniversity of Cambridge UK London Society for UnderwaterTechnology

Dodd PA MR Price KJ Heywood and M Pebody 2006Collection of water samples from an autonomous underwatervehicle for tracer analysis J Atmos Oceanic Technol 23(12)1759ndash1767

Dowdeswell JA and JL Bamber 2007 Keel depths of modernAntarctic icebergs and implications for sea-floor scouring in thegeological record Mar Geol 243(1ndash4) 120ndash131

Dowdeswell JA and RD Powell 1996 Submersible remotelyoperated vehicles (ROVs) for investigations of the glacierndashoceanndashsediment interface J Glaciol 42(140) 176ndash183

Evans J JA Dowdeswell C O Cofaigh TJ Benham and JB And-erson 2006 Extent and dynamics of the West Antarctic IceSheet on the outer continental shelf of Pine Island Bay during thelast glaciation Mar Geol 250(1ndash2) 53ndash72

Fahnestock MA TA Scambos RA Bindschadler and G Kvaran2000 A millennium of variable ice flow recorded by the RossIce Shelf Antarctica J Glaciol 46(155) 652ndash664

Francois RE 1977 High resolution observations of under-icemorphology Seattle WA University of Washington AppliedPhysics Laboratory Tech Rep APL-UW-7112

Griffiths G and K Collins eds 2007 Masterclass in AUVtechnology for polar science collaborative autosub science inextreme environments Proceedings of the International Master-class 28ndash30 March 2006 National Oceanography CentreSouthampton UK London Society for Underwater Technology

Hayes DR and A Jenkins 2007 Autonomous underwater vehiclemeasurements of surface wave decay and directional spectra inthe marginal sea ice zone J Phys Oceanogr 37(1) 71ndash83

Holland PR and DL Feltham 2006 The effects of rotation andice shelf topography on frazil-laden ice shelf water plumesJ Phys Oceanogr 36(12) 2312ndash2327

Fig 12 Example photograph from Autosub mission 377 showingthe floor of outer Kangerdlussuaq Fjord (imaged from an altitude of9m at a depth of 564m) One cobble-sized iceberg-rafted drop-stone three large burrows and numerous tubeworms are visibleThe photograph is about 1m across

Dowdeswell and others Instruments and methods 671

Although investigation of the ocean processes withincavities beneath ice shelves will always require mooredinstruments capable of collecting data over periods ofmonths or years Autosubrsquos ability as a platform that canuse sophisticated oceanographic instrumentation to gener-ate spatially extensive datasets has given us a unique view ofone of the least accessible parts of the worldrsquos oceans

Autosub was lost under the Fimbul Ice Shelf on mission383 the one following that described above Its low-frequency acoustic beacon signalled that an abort had beentriggered and that the vehicle was stuck at a positionapproximately 17 km from the ice front A full investigation(Strutt 2006) concluded that either an open-circuit ornetwork failure was the most likely cause of an abort orloss of power This was the only time the vehicle deployed itslong-range acoustic beacon

Oceanography fjord circulation and meltwater fluxThe circulation and mixing processes of water masses withinfjords can be complex so one advantage of a rapid andcontinuous surveying device such as Autosub is to enablea more synoptic survey than is achievable with a shipTypically saline ocean water enters a fjord at mid-depthabove the sill and fresh meltwater from the surroundingglaciers or rivers exits the fjord as a surface layer (Syvitski andothers 1987) The deep waters within the fjord are renewedonly sporadically However this steady-state simple picturecan be complicated by the presence of tides cross-fjordflows sea ice entering andor leaving the fjord and the flowsinduced by inertial oscillations following storms The netexchange of fresh water between a fjord and the continental-shelf environment is of importance in determining forexample the influence of meltwater from the Greenlandice sheet on the formation of dense water masses in the seassurrounding Greenland

The detailed current-velocity structure revealed by theAutosub ADCPs at the mouth of Kangerdlussuaq Fjord onthe east coast of Greenland is shown in Figure 8 In the

6 hours of this survey three passes across the fjord weremade at depths of 70 190 and 400m The upward- anddownward-looking ADCPs are very consistent betweenadjacent passes some 4hours apart implying that tidal (orother temporally varying) flows are not dominant hereHowever the velocity structure is very different from thesimple three-layer flow suggested above The primaryinflows are on the southwest side of the mouth at 300ndash400m and in the upper 100m The primary outflow is atabout 200ndash350m on the northwest side of the fjord mouthimplying a clockwise circulation of open ocean water in thebay at the mouth of the fjord There is a suggestion that thewater in the top 10m may be a thin layer of ice melt leavingthe fjord Thus Autosub has revealed in unprecedenteddetail a snapshot of the complex exchanges between anArctic fjord environment and the adjacent continental shelf

Oceanography attenuation of waves by sea iceA serendipitous result for the behaviour of waves propa-gating in sea ice was obtained from the upward-lookingADCP surface track velocity recorded on Autosub Becausethe surface track ping has longer range than the profile pingthe velocity of sea ice relative to Autosub could be measuredduring runs as deep as 200m This was the first use of anAUV to measure directional and scalar wave propertiesduring surface wave propagation through sea ice (Hayes andothers 2007) Since ice-edge detection was also possiblefrom the surface track ping (verified by ship observations)dependence of the above wave properties on distance fromthe edge of the marginal ice zone could be examined

As an example during mission 324 on 25 March 2003 inthe marginal ice zone of the Bellingshausen Sea Antarcticathe horizontal velocity of the ice was oscillating Themagnitude of this oscillation also decayed with distancefrom the ice edge both on the inward and outward segments(Fig 9a) In the observed regime of small ice floes (lt20m)and long wavelength (100ndash350m) the floes nearly follow thecircular path of a point on the water surface Therefore the

Fig 8 Cross-section of the current velocity (colour scale in m sndash1)into and out of the mouth of Kangerdlussuaq Fjord East Greenlandacquired from the upward- and downward-looking ADCPs mountedon Autosub The Autosub navigated horizontal paths at 70 190 and400m (marked as black lines) descending or rising in betweentaking 6 hours to complete the survey Positive values denote waterflowing into the fjord negative values indicate water flowing out ofthe fjord Southwest is to the left and northeast to the right

Fig 9 Sea-ice velocity from Autosub mission 324 The upward-looking ADCP measured the surface track velocity upon (a) enteringthe ice pack at 90m and (b) exiting the ice pack at 90m Themagnified inset shows a typical segment analyzed here Note thestrong periodicity in both components as well as mean currenttowards the southeast (modified from Hayes and others 2007)

Dowdeswell and others Instruments and methods 669

surface track velocity is regarded as a measurement of sur-face wave orbital velocity superimposed on mean icevelocity (southeastward in the case of mission 324) Theseries is divided into a number of blocks (Fig 9b) to analyzethe surface velocity The directional and scalar wave spectraare calculated for each segment so any trend in significantwave height mean and peak wave periods as well as anychange in the energy wave direction or spread of variousfrequency components can be detected (Fig 10) The char-acter of waves propagating through sea ice that was observedusing Autosub agrees with most of the previous observational(Wadhams and others 1986 1988 Liu and others 1991) andnumerical (Meylan and others 1997) experiments

Glacial geology submarine glacial landforms andacoustic stratigraphyThe morphology and stratigraphy of the sea floor provideimportant evidence for the reconstruction of the dimensionsand flow of former ice sheets (eg Anderson 1999) Whereice flows across a sedimentary bed landforms diagnostic ofice-flow direction and dynamics are produced These land-forms which are often streamlined are preserved underwater as ice retreats across continental shelves and fjordsduring interglacial and interstadial periods (eg Andersonand others 2002 Ottesen and others 2005 Evans andothers 2006) Characteristic assemblages of these submarinelandforms are indicators of for example ice-stream flowpast glacier-surge activity and former grounding lines (egPowell and others 1996 Canals and others 2000 O Cofaighand others 2002 Ottesen and Dowdeswell 2006)

The swath-bathymetry system on Autosub when mountedin downward-looking mode produces data that yield high-resolution digital-terrain models and 3-D images of the seafloor Figure 11 shows the floor of an East Greenland fjordwhere the fast-flowing Kangerdlussuaq Glacier one of themajor outlet glaciers of the Greenland ice sheet (Rignot andKanagaratnam 2006) has produced streamlined sediment-ary bedforms which are preserved in several hundred metres

of water after ice retreat from its position at the Last GlacialMaximum (Syvitski and others 1996) Shallow acousticstratigraphy provides further information on the structure ofthe upper few metres to tens of metres of sediment In theexample shown in Figure 11b the acoustic profiler onAutosub penetrates the fine-grained and acoustically lamin-ated sediments in the deepest part of Kangerdlussuaq Fjordwith less transparent and probably coarser-grained sedi-ments characteristic of glacial diamicts or tills to either sideAutosub which has been deployed close to the calvingtidewater margins of Courtauld Glacier East Greenland(Fig 1a) can be used to image areas of the sea floor inpreviously inaccessible locations near calving ice cliffs andbeneath ice shelves

In addition to geophysical instruments the digital cameraequipment on Autosub provides detailed information on theform and composition of the sea floor and the marine biotathat inhabit it Figure 12 shows an example of a sea-floorphotograph acquired by Autosub in Kangerdlussuaq FjordBoth individual dropstones released by iceberg melting andbottom-dwelling marine organisms are shown The presenceof deposit-feeding species is indicated by faunal traces on thesediment surface Evidence of disturbance to the seabed andfauna from iceberg-keel ploughing was also observed inphotographs of the sea floor at water depths less than about500m reducing faunal density and diversity as well asproducing a sedimentologically heterogeneous environment

CONCLUSIONSThe Autosub AUV provided a platform for the deployment ofa number of geophysical and oceanographic instruments inhazardous polar environments that ships and other mannedvehicles cannot access

Fig 10 (a) Mean wave period and (b) significant wave height forAutosub missions 322ndash324 The label lsquoinrsquo refers to the seriescollected upon entering the ice pack while lsquobackrsquo refers to thereturn series Period and wave height are derived from the one-dimensional wave spectrum of 512 s blocks (with the exception ofthe return trip in mission 323 in which 256 s blocks were analyzed)

Fig 11Multibeam echo-sounder image of the glacially streamlinedsea floor of Kangerdlussuaq Fjord acquired from a 200 kHz swath-bathymetry system mounted on Autosub The swath width isapproximately 200m Water depth is 710ndash840m The swath-bathymetry data are gridded at a resolution of 1m in the horizontalThe lower panels show acoustically stratified sediments on the fjordfloor acquired from the chirp 2ndash16 kHz sub-bottom profiler onAutosub The acoustic profile is located in the multibeam image

Dowdeswell and others Instruments and methods670

Ice-covered environments investigated using Autosubinclude a cavity beneath the Fimbul Ice Shelf and therelatively shallow and poorly charted waters beneath sea iceon the East Greenland continental shelf

The multibeam echo sounder of Autosub has imaged theunderside of an ice shelf for the first time showing that someareas are very rough with implications for the modelling ofwater flow and melt rates The underside of sea ice has alsobeen imaged in detail and quantitative shape parametersextracted Swath images and bottom photographs of theglacial geology and marine biota close to the margins ofArctic tidewater glaciers have also been obtained

Oceanographic data such as salinity temperature andwater velocity have been derived continuously during Auto-sub missions beneath floating Arctic and Antarctic ice pro-viding observations with a very dense spatial coverage inenvironments where previously few or no data have beenavailable

The Autosub3 vehicle successor to the lost Autosub2and AUV technology in general is likely to be used in-creasingly in hazardous polar marine environments for thecollection of detailed geophysical and oceanographic dataclose to and beneath floating ice These data in turn areimportant in the calibration and testing of numerical modelsrelating to ice-sheet interactions with the polar waters

Not all of the scenarios for AUV operations in polar seashave yet been achieved in practice although many havebeen described and discussed by scientists and engineers(Griffiths and Collins 2007 Collins and Griffiths 2008) InAugust 2007 the first AUV campaign took place to searchfor and then examine hydrothermal sites at the slow-spreading Gakkel Ridge in the Arctic Ocean an area ofextensive multi-year pack ice (Reves-Sohn and others 2007)Other plans include multidisciplinary studies beneath theRoss Ice Shelf Antarctica and surveys of Southern Oceankrill populations in winter

ACKNOWLEDGEMENTSThis work was supported by the Autosub Under Ice The-matic Programme of the UK Natural Environment ResearchCouncil (Programme Chair S Ackley Programme ManagerK Collins) We are grateful to the Autosub Technical Teamand the officers and crew of RRS James Clark Ross for theirinvaluable contributions on four Autosub cruises

REFERENCESAnderson JB 1999 Antarctic marine geology Cambridge etc

Cambridge University PressAnderson JB SS Shipp AL Lowe JS Wellner and AB Mosola

2002 The Antarctic ice sheet during the last glacial maximumand its subsequent retreat history a review Quat Sci Rev21(1ndash3) 49ndash70

Bamber JL and CR Bentley 1994 A comparison of satellite-altimetry and ice-thickness measurements of the Ross Ice ShelfAntarctica Ann Glaciol 20 357ndash364

Brierley AS and 11 others 2002 Antarctic krill under sea iceelevated abundance in a narrow band just south of ice edgeScience 295(5561) 1890ndash1892

Broecker WS 1991 The great ocean conveyor Oceanography4(2) 79ndash89

Canals M R Urgeles and AM Calafat 2000 Deep sea-floorevidence of past ice streams off the Antarctic PeninsulaGeology 28(1) 31ndash34

Cavalieri DJ CL Parkinson and KY Vinnikov 2003 30-Yearsatellite record reveals contrasting Arctic and Antarctic decadalsea ice variability Geophys Res Lett 30(18) 1970 (1010292003GL018031)

Collins K and G Griffiths eds 2008 Workshop on AUV sciencein extreme environments collaborative Autosub science inextreme environments Proceedings of the International ScienceWorkship 11ndash13 April 2007 Scott Polar Research InstituteUniversity of Cambridge UK London Society for UnderwaterTechnology

Dodd PA MR Price KJ Heywood and M Pebody 2006Collection of water samples from an autonomous underwatervehicle for tracer analysis J Atmos Oceanic Technol 23(12)1759ndash1767

Dowdeswell JA and JL Bamber 2007 Keel depths of modernAntarctic icebergs and implications for sea-floor scouring in thegeological record Mar Geol 243(1ndash4) 120ndash131

Dowdeswell JA and RD Powell 1996 Submersible remotelyoperated vehicles (ROVs) for investigations of the glacierndashoceanndashsediment interface J Glaciol 42(140) 176ndash183

Evans J JA Dowdeswell C O Cofaigh TJ Benham and JB And-erson 2006 Extent and dynamics of the West Antarctic IceSheet on the outer continental shelf of Pine Island Bay during thelast glaciation Mar Geol 250(1ndash2) 53ndash72

Fahnestock MA TA Scambos RA Bindschadler and G Kvaran2000 A millennium of variable ice flow recorded by the RossIce Shelf Antarctica J Glaciol 46(155) 652ndash664

Francois RE 1977 High resolution observations of under-icemorphology Seattle WA University of Washington AppliedPhysics Laboratory Tech Rep APL-UW-7112

Griffiths G and K Collins eds 2007 Masterclass in AUVtechnology for polar science collaborative autosub science inextreme environments Proceedings of the International Master-class 28ndash30 March 2006 National Oceanography CentreSouthampton UK London Society for Underwater Technology

Hayes DR and A Jenkins 2007 Autonomous underwater vehiclemeasurements of surface wave decay and directional spectra inthe marginal sea ice zone J Phys Oceanogr 37(1) 71ndash83

Holland PR and DL Feltham 2006 The effects of rotation andice shelf topography on frazil-laden ice shelf water plumesJ Phys Oceanogr 36(12) 2312ndash2327

Fig 12 Example photograph from Autosub mission 377 showingthe floor of outer Kangerdlussuaq Fjord (imaged from an altitude of9m at a depth of 564m) One cobble-sized iceberg-rafted drop-stone three large burrows and numerous tubeworms are visibleThe photograph is about 1m across

Dowdeswell and others Instruments and methods 671

surface track velocity is regarded as a measurement of sur-face wave orbital velocity superimposed on mean icevelocity (southeastward in the case of mission 324) Theseries is divided into a number of blocks (Fig 9b) to analyzethe surface velocity The directional and scalar wave spectraare calculated for each segment so any trend in significantwave height mean and peak wave periods as well as anychange in the energy wave direction or spread of variousfrequency components can be detected (Fig 10) The char-acter of waves propagating through sea ice that was observedusing Autosub agrees with most of the previous observational(Wadhams and others 1986 1988 Liu and others 1991) andnumerical (Meylan and others 1997) experiments

Glacial geology submarine glacial landforms andacoustic stratigraphyThe morphology and stratigraphy of the sea floor provideimportant evidence for the reconstruction of the dimensionsand flow of former ice sheets (eg Anderson 1999) Whereice flows across a sedimentary bed landforms diagnostic ofice-flow direction and dynamics are produced These land-forms which are often streamlined are preserved underwater as ice retreats across continental shelves and fjordsduring interglacial and interstadial periods (eg Andersonand others 2002 Ottesen and others 2005 Evans andothers 2006) Characteristic assemblages of these submarinelandforms are indicators of for example ice-stream flowpast glacier-surge activity and former grounding lines (egPowell and others 1996 Canals and others 2000 O Cofaighand others 2002 Ottesen and Dowdeswell 2006)

The swath-bathymetry system on Autosub when mountedin downward-looking mode produces data that yield high-resolution digital-terrain models and 3-D images of the seafloor Figure 11 shows the floor of an East Greenland fjordwhere the fast-flowing Kangerdlussuaq Glacier one of themajor outlet glaciers of the Greenland ice sheet (Rignot andKanagaratnam 2006) has produced streamlined sediment-ary bedforms which are preserved in several hundred metres

of water after ice retreat from its position at the Last GlacialMaximum (Syvitski and others 1996) Shallow acousticstratigraphy provides further information on the structure ofthe upper few metres to tens of metres of sediment In theexample shown in Figure 11b the acoustic profiler onAutosub penetrates the fine-grained and acoustically lamin-ated sediments in the deepest part of Kangerdlussuaq Fjordwith less transparent and probably coarser-grained sedi-ments characteristic of glacial diamicts or tills to either sideAutosub which has been deployed close to the calvingtidewater margins of Courtauld Glacier East Greenland(Fig 1a) can be used to image areas of the sea floor inpreviously inaccessible locations near calving ice cliffs andbeneath ice shelves

In addition to geophysical instruments the digital cameraequipment on Autosub provides detailed information on theform and composition of the sea floor and the marine biotathat inhabit it Figure 12 shows an example of a sea-floorphotograph acquired by Autosub in Kangerdlussuaq FjordBoth individual dropstones released by iceberg melting andbottom-dwelling marine organisms are shown The presenceof deposit-feeding species is indicated by faunal traces on thesediment surface Evidence of disturbance to the seabed andfauna from iceberg-keel ploughing was also observed inphotographs of the sea floor at water depths less than about500m reducing faunal density and diversity as well asproducing a sedimentologically heterogeneous environment

CONCLUSIONSThe Autosub AUV provided a platform for the deployment ofa number of geophysical and oceanographic instruments inhazardous polar environments that ships and other mannedvehicles cannot access

Fig 10 (a) Mean wave period and (b) significant wave height forAutosub missions 322ndash324 The label lsquoinrsquo refers to the seriescollected upon entering the ice pack while lsquobackrsquo refers to thereturn series Period and wave height are derived from the one-dimensional wave spectrum of 512 s blocks (with the exception ofthe return trip in mission 323 in which 256 s blocks were analyzed)

Fig 11Multibeam echo-sounder image of the glacially streamlinedsea floor of Kangerdlussuaq Fjord acquired from a 200 kHz swath-bathymetry system mounted on Autosub The swath width isapproximately 200m Water depth is 710ndash840m The swath-bathymetry data are gridded at a resolution of 1m in the horizontalThe lower panels show acoustically stratified sediments on the fjordfloor acquired from the chirp 2ndash16 kHz sub-bottom profiler onAutosub The acoustic profile is located in the multibeam image

Dowdeswell and others Instruments and methods670

Ice-covered environments investigated using Autosubinclude a cavity beneath the Fimbul Ice Shelf and therelatively shallow and poorly charted waters beneath sea iceon the East Greenland continental shelf

The multibeam echo sounder of Autosub has imaged theunderside of an ice shelf for the first time showing that someareas are very rough with implications for the modelling ofwater flow and melt rates The underside of sea ice has alsobeen imaged in detail and quantitative shape parametersextracted Swath images and bottom photographs of theglacial geology and marine biota close to the margins ofArctic tidewater glaciers have also been obtained

Oceanographic data such as salinity temperature andwater velocity have been derived continuously during Auto-sub missions beneath floating Arctic and Antarctic ice pro-viding observations with a very dense spatial coverage inenvironments where previously few or no data have beenavailable

The Autosub3 vehicle successor to the lost Autosub2and AUV technology in general is likely to be used in-creasingly in hazardous polar marine environments for thecollection of detailed geophysical and oceanographic dataclose to and beneath floating ice These data in turn areimportant in the calibration and testing of numerical modelsrelating to ice-sheet interactions with the polar waters

Not all of the scenarios for AUV operations in polar seashave yet been achieved in practice although many havebeen described and discussed by scientists and engineers(Griffiths and Collins 2007 Collins and Griffiths 2008) InAugust 2007 the first AUV campaign took place to searchfor and then examine hydrothermal sites at the slow-spreading Gakkel Ridge in the Arctic Ocean an area ofextensive multi-year pack ice (Reves-Sohn and others 2007)Other plans include multidisciplinary studies beneath theRoss Ice Shelf Antarctica and surveys of Southern Oceankrill populations in winter

ACKNOWLEDGEMENTSThis work was supported by the Autosub Under Ice The-matic Programme of the UK Natural Environment ResearchCouncil (Programme Chair S Ackley Programme ManagerK Collins) We are grateful to the Autosub Technical Teamand the officers and crew of RRS James Clark Ross for theirinvaluable contributions on four Autosub cruises

REFERENCESAnderson JB 1999 Antarctic marine geology Cambridge etc

Cambridge University PressAnderson JB SS Shipp AL Lowe JS Wellner and AB Mosola

2002 The Antarctic ice sheet during the last glacial maximumand its subsequent retreat history a review Quat Sci Rev21(1ndash3) 49ndash70

Bamber JL and CR Bentley 1994 A comparison of satellite-altimetry and ice-thickness measurements of the Ross Ice ShelfAntarctica Ann Glaciol 20 357ndash364

Brierley AS and 11 others 2002 Antarctic krill under sea iceelevated abundance in a narrow band just south of ice edgeScience 295(5561) 1890ndash1892

Broecker WS 1991 The great ocean conveyor Oceanography4(2) 79ndash89

Canals M R Urgeles and AM Calafat 2000 Deep sea-floorevidence of past ice streams off the Antarctic PeninsulaGeology 28(1) 31ndash34

Cavalieri DJ CL Parkinson and KY Vinnikov 2003 30-Yearsatellite record reveals contrasting Arctic and Antarctic decadalsea ice variability Geophys Res Lett 30(18) 1970 (1010292003GL018031)

Collins K and G Griffiths eds 2008 Workshop on AUV sciencein extreme environments collaborative Autosub science inextreme environments Proceedings of the International ScienceWorkship 11ndash13 April 2007 Scott Polar Research InstituteUniversity of Cambridge UK London Society for UnderwaterTechnology

Dodd PA MR Price KJ Heywood and M Pebody 2006Collection of water samples from an autonomous underwatervehicle for tracer analysis J Atmos Oceanic Technol 23(12)1759ndash1767

Dowdeswell JA and JL Bamber 2007 Keel depths of modernAntarctic icebergs and implications for sea-floor scouring in thegeological record Mar Geol 243(1ndash4) 120ndash131

Dowdeswell JA and RD Powell 1996 Submersible remotelyoperated vehicles (ROVs) for investigations of the glacierndashoceanndashsediment interface J Glaciol 42(140) 176ndash183

Evans J JA Dowdeswell C O Cofaigh TJ Benham and JB And-erson 2006 Extent and dynamics of the West Antarctic IceSheet on the outer continental shelf of Pine Island Bay during thelast glaciation Mar Geol 250(1ndash2) 53ndash72

Fahnestock MA TA Scambos RA Bindschadler and G Kvaran2000 A millennium of variable ice flow recorded by the RossIce Shelf Antarctica J Glaciol 46(155) 652ndash664

Francois RE 1977 High resolution observations of under-icemorphology Seattle WA University of Washington AppliedPhysics Laboratory Tech Rep APL-UW-7112

Griffiths G and K Collins eds 2007 Masterclass in AUVtechnology for polar science collaborative autosub science inextreme environments Proceedings of the International Master-class 28ndash30 March 2006 National Oceanography CentreSouthampton UK London Society for Underwater Technology

Hayes DR and A Jenkins 2007 Autonomous underwater vehiclemeasurements of surface wave decay and directional spectra inthe marginal sea ice zone J Phys Oceanogr 37(1) 71ndash83

Holland PR and DL Feltham 2006 The effects of rotation andice shelf topography on frazil-laden ice shelf water plumesJ Phys Oceanogr 36(12) 2312ndash2327

Fig 12 Example photograph from Autosub mission 377 showingthe floor of outer Kangerdlussuaq Fjord (imaged from an altitude of9m at a depth of 564m) One cobble-sized iceberg-rafted drop-stone three large burrows and numerous tubeworms are visibleThe photograph is about 1m across

Dowdeswell and others Instruments and methods 671

Ice-covered environments investigated using Autosubinclude a cavity beneath the Fimbul Ice Shelf and therelatively shallow and poorly charted waters beneath sea iceon the East Greenland continental shelf

The multibeam echo sounder of Autosub has imaged theunderside of an ice shelf for the first time showing that someareas are very rough with implications for the modelling ofwater flow and melt rates The underside of sea ice has alsobeen imaged in detail and quantitative shape parametersextracted Swath images and bottom photographs of theglacial geology and marine biota close to the margins ofArctic tidewater glaciers have also been obtained

Oceanographic data such as salinity temperature andwater velocity have been derived continuously during Auto-sub missions beneath floating Arctic and Antarctic ice pro-viding observations with a very dense spatial coverage inenvironments where previously few or no data have beenavailable

The Autosub3 vehicle successor to the lost Autosub2and AUV technology in general is likely to be used in-creasingly in hazardous polar marine environments for thecollection of detailed geophysical and oceanographic dataclose to and beneath floating ice These data in turn areimportant in the calibration and testing of numerical modelsrelating to ice-sheet interactions with the polar waters

Not all of the scenarios for AUV operations in polar seashave yet been achieved in practice although many havebeen described and discussed by scientists and engineers(Griffiths and Collins 2007 Collins and Griffiths 2008) InAugust 2007 the first AUV campaign took place to searchfor and then examine hydrothermal sites at the slow-spreading Gakkel Ridge in the Arctic Ocean an area ofextensive multi-year pack ice (Reves-Sohn and others 2007)Other plans include multidisciplinary studies beneath theRoss Ice Shelf Antarctica and surveys of Southern Oceankrill populations in winter

ACKNOWLEDGEMENTSThis work was supported by the Autosub Under Ice The-matic Programme of the UK Natural Environment ResearchCouncil (Programme Chair S Ackley Programme ManagerK Collins) We are grateful to the Autosub Technical Teamand the officers and crew of RRS James Clark Ross for theirinvaluable contributions on four Autosub cruises

REFERENCESAnderson JB 1999 Antarctic marine geology Cambridge etc

Cambridge University PressAnderson JB SS Shipp AL Lowe JS Wellner and AB Mosola

2002 The Antarctic ice sheet during the last glacial maximumand its subsequent retreat history a review Quat Sci Rev21(1ndash3) 49ndash70

Bamber JL and CR Bentley 1994 A comparison of satellite-altimetry and ice-thickness measurements of the Ross Ice ShelfAntarctica Ann Glaciol 20 357ndash364

Brierley AS and 11 others 2002 Antarctic krill under sea iceelevated abundance in a narrow band just south of ice edgeScience 295(5561) 1890ndash1892

Broecker WS 1991 The great ocean conveyor Oceanography4(2) 79ndash89

Canals M R Urgeles and AM Calafat 2000 Deep sea-floorevidence of past ice streams off the Antarctic PeninsulaGeology 28(1) 31ndash34

Cavalieri DJ CL Parkinson and KY Vinnikov 2003 30-Yearsatellite record reveals contrasting Arctic and Antarctic decadalsea ice variability Geophys Res Lett 30(18) 1970 (1010292003GL018031)

Collins K and G Griffiths eds 2008 Workshop on AUV sciencein extreme environments collaborative Autosub science inextreme environments Proceedings of the International ScienceWorkship 11ndash13 April 2007 Scott Polar Research InstituteUniversity of Cambridge UK London Society for UnderwaterTechnology

Dodd PA MR Price KJ Heywood and M Pebody 2006Collection of water samples from an autonomous underwatervehicle for tracer analysis J Atmos Oceanic Technol 23(12)1759ndash1767

Dowdeswell JA and JL Bamber 2007 Keel depths of modernAntarctic icebergs and implications for sea-floor scouring in thegeological record Mar Geol 243(1ndash4) 120ndash131

Dowdeswell JA and RD Powell 1996 Submersible remotelyoperated vehicles (ROVs) for investigations of the glacierndashoceanndashsediment interface J Glaciol 42(140) 176ndash183

Evans J JA Dowdeswell C O Cofaigh TJ Benham and JB And-erson 2006 Extent and dynamics of the West Antarctic IceSheet on the outer continental shelf of Pine Island Bay during thelast glaciation Mar Geol 250(1ndash2) 53ndash72

Fahnestock MA TA Scambos RA Bindschadler and G Kvaran2000 A millennium of variable ice flow recorded by the RossIce Shelf Antarctica J Glaciol 46(155) 652ndash664

Francois RE 1977 High resolution observations of under-icemorphology Seattle WA University of Washington AppliedPhysics Laboratory Tech Rep APL-UW-7112

Griffiths G and K Collins eds 2007 Masterclass in AUVtechnology for polar science collaborative autosub science inextreme environments Proceedings of the International Master-class 28ndash30 March 2006 National Oceanography CentreSouthampton UK London Society for Underwater Technology

Hayes DR and A Jenkins 2007 Autonomous underwater vehiclemeasurements of surface wave decay and directional spectra inthe marginal sea ice zone J Phys Oceanogr 37(1) 71ndash83

Holland PR and DL Feltham 2006 The effects of rotation andice shelf topography on frazil-laden ice shelf water plumesJ Phys Oceanogr 36(12) 2312ndash2327

Fig 12 Example photograph from Autosub mission 377 showingthe floor of outer Kangerdlussuaq Fjord (imaged from an altitude of9m at a depth of 564m) One cobble-sized iceberg-rafted drop-stone three large burrows and numerous tubeworms are visibleThe photograph is about 1m across

Dowdeswell and others Instruments and methods 671

Jenkins A and CSM Doake 1991 Icendashocean interaction onRonne Ice Shelf Antarctica J Geophys Res 96(C1) 791ndash813

Liu AK B Holt and PW Vachon 1991 Wave propagation in themarginal ice zone model predictions and comparisons withbuoy and synthetic aperture radar data J Geophys Res 96(C3)4605ndash4621

Mayer C N Reeh F Jung-Rothenhausler P Huybrechts andH Oerter 2000 The subglacial cavity and implied dynamicsunder Nioghalvfjerdsfjorden glacier NE Greenland GeophysRes Lett 27(15) 2289ndash2292

McPhail SD and M Pebody 1998 Navigation and control of anautonomous underwater vehicle using a distributed networkedcontrol architecture Underwater Technol 23(1) 19ndash30

Meylan M VA Squire and C Fox 1997 Towards realism inmodelling ocean wave behavior in marginal ice zones J Geo-phys Res 102(C10) 22981ndash22991

Millard NW and 8 others 1998 Versatile autonomous sub-mersibles ndash the realising and testing of a practical vehicleUnderwater Technol 23(1) 7ndash17

Nicholls KW 1996 Temperature variability beneath Ronne IceShelf Antarctica from thermistor cables J Geophys Res101(C1) 1199ndash1210

Nicholls KW S Osterhus K Makinson and MR Johnson 2001Oceanographic conditions south of Berkner Island beneathFilchnerndashRonne Ice Shelf Antarctica J Geophys Res 106(C6)11481ndash11492

Nicholls KW and 21 others 2006 Measurements beneath anAntarctic ice shelf using an autonomous underwater vehicleGeophys Res Lett 33(8) L08162 (1010292006GL025998)

O Cofaigh C CJ Pudsey JA Dowdeswell and P Morris 2002Evolution of subglacial bedforms along a paleo-ice streamAntarctic Peninsula continental shelf Geophys Res Lett 29(8)1199 (1010292001GL014488)

Ottesen D and JA Dowdeswell 2006 Assemblages of submarinelandforms produced by tidewater glaciers in SvalbardJ Geophys Res 111(F1) F01016 (1010292005JF000330)

Ottesen D JA Dowdeswell and L Rise 2005 Submarinelandforms and the reconstruction of fast-flowing ice streamswithin a large Quaternary ice sheet the 2500-km-long Nor-wegian-Svalbard margin (578ndash808N) Geol Soc Am Bull117(7) 1033ndash1050

Powell RD M Dawber JN McInnes and AR Pyne 1996Observations of the grounding-line area at a floating glacierterminus Ann Glaciol 22 217ndash223

Reves-Sohn RA and 22 others 2007 Scientific scope andsummary of the Arctic Gakkel vents (AGAVE) expedition[Abstract OS41C-07] Eos 88(52) Fall Meet Suppl

Rignot E and P Kanagaratnam 2006 Changes in the velocitystructure of the Greenland Ice Sheet Science 311(5673)986ndash990

Stevenson P G Griffiths and AT Webb 2002 The experienceand limitations of using manganese alkaline primary cells in alarge operational AUV In Proceedings of the 2002 Workshop onAutonomous Underwater Vehicles 20ndash21 June San AntonioTexas Piscatawey NJ Institute of Electrical and ElectronicsEngineers 27ndash34

Stevenson P and 7 others 2003 Engineering an autonomousunderwater vehicle for under ice operations In Proceedings ofthe 22nd International Conference on Offshore Mechanics andArctic Engineering 8-13 June 2003 Cancun Mexico New YorkAmerican Society of Mechanical Engineers CD-ROM

Strutt JE 2006 Report of the inquiry into the loss of Autosub2under the Fimbulisen Southampton National OceanographyCentre (Research and Consultancy Report 12)

Syvitski JPM DC Burrell and JM Skei 1987 Fjords processesand products New York Springer-Verlag

Syvitski JPM JT Andrews and JA Dowdeswell 1996 Sedimentdeposition in an iceberg-dominated glacimarine environmentEast Greenland basin fill implications Global Planet Change12(1ndash4) 251ndash270

Wadhams P 1978 Sidescan sonar imagery of sea ice in the ArcticOcean Can J Remote Sens 4(2) 161ndash173

Wadhams P 1988 The underside of Arctic sea ice imaged bysidescan sonar Nature 333(6169) 161ndash164

Wadhams P 2000 Ice in the ocean Amsterdam etc Gordon andBreach Science Publishers

Wadhams P and S Martin 1990 Processes determining thebottom topography of multiyear arctic sea ice In Ackley SFand WF Weeks eds Sea ice properties and processesProceedings of the WF Weeks Sea Ice Symposium HanoverNH US Army Cold Regions Research and Engineering Labora-tory 136ndash141 (CRREL Monogr 90-1)

Wadhams P VA Squire JA Ewing and RW Pascal 1986 Theeffect of the marginal ice zone on the directional wave spectrumof the ocean J Phys Oceanogr 16(2) 358ndash376

Wadhams P VA Squire DJ Goodman AM Cowan andSC Moore 1988 The attenuation rates of ocean waves in themarginal ice zone J Geophys Res 93(C6) 6799ndash6818

Wadhams P JP Wilkinson and A Kaletzky 2004 Sidescan sonarimagery of the winter marginal ice zone obtained from an AUVJ Atmos Oceanic Technol 21(9) 1462ndash1470

Wadhams P JP Wilkinson and SD McPhail 2006 A new viewof the underside of Arctic sea ice Geophys Res Lett 33(4)L04501 (1010292005GL025131)

MS received 11 December 2007 and accepted in revised form 22 May 2008

Dowdeswell and others Instruments and methods672