Sea Surface Topography by Airborne LaserAltimetry and Offshore GPS Buoys in the EasternMediterranean: Comparison with JASON-1 RadarAltimeter Data and GRACE Gravity Field

Limpach, P., Geiger, A., Kahle, H.-G.Geodesy and Geodynamics Laboratory, Institute of Geodesy and Photogrammetry, ETH Zurich, Switzerland

Abstract. Satellite radar altimetry is the basicmeans for global-scale sea surface height (SSH)monitoring, constituting a major source for gravityfield improvements. In order to contribute to the im-provement of sea level monitoring and to providelocal-scale information on the short-wave structureof the marine gravity field, enhanced ground-basedmethods for precise SSH measurements have beendeveloped, consisting in airborne laser altimetry,shipborne multi-antenna GPS measurements andGPS-equipped buoys. Two local survey areas werechosen in the vicinity of JASON-1 ground-tracks inthe Eastern Mediterranean. The gathered SSH datacould ultimately allow to contribute to the valida-tion and calibration of radar altimeter missions.

Preliminary SSH results from airborne laser al-timetry and offshore GPS surveys are presented.Furthermore comparisons with JASON-1 radar al-timeter data and geoid heights from both EGM96and GRACE-based GGM02 models are made.

Keywords. Marine geodesy, airborne laser altime-try, GPS buoys, shipborne GPS, sea surface height

_________________________________________

1 Introduction

Enhanced ground-based methods have been devel-oped for precise ground-truth determination of seasurface heights (SSHs), consisting in airborne laseraltimetry, shipborne multi-antenna GPS measure-ments and GPS buoys. The SSH data provide local-scale information on the short-wave structure of thegravity field and can be used to improve local ma-rine geoid solutions. They also contain informationon local dynamic ocean topography, tides andwaves, and can be used for the validation and cali-bration of radar altimeter missions. In addition, theyprovide a link between offshore radar altimeter dataand tide-gauge stations.

The key area for our SSH investigations is theEastern Mediterranean Sea, where two survey areas

were chosen around Crete and in the North AegeanSea, respectively (Fig. 1). Differences between pre-liminary SSH solutions and both EGM96 (to de-gree/order 360) and GRACE-based GGM02 geoidmodels have been computed for these areas. TheGGM02 model used in the comparisons isGGM02C extended to degree/order 360 usingEGM96 coefficients above degree 200 (Tapleyet al. 2005). The most striking feature revealed inthis area by the geoid model are the extreme gradi-ents between the central Aegean Sea and the Hel-lenic Trench area, which amount to 40 m along adistance of only 400 km (Fig. 1). In the EasternMediterranean, the differences between GGM02and EGM96 geoid heights reveal no systematic off-set, although they locally exceed 0.5 m (Fig. 2). Thedifferences in our survey areas are around 0.3 mnear Crete and 0.6 m in the North Aegean Sea.

Geoid height [m]

Aegean Sea

Hellenic Trench

Crete

Fig. 1. Geoid heights in the Eastern Mediterranean Seafrom GRACE-based GGM02 model extended to de-gree/order 360 using EGM96 coefficients. White rect-angles: survey areas around Crete (airborne laser alti-metry) and in the North Aegean Sea (shipborne/buoyGPS).

GGM02 – EGM96 [m]

Aegean Sea

Hellenic Trench

Crete

Fig. 2. Difference between GGM02 and EGM96 geoidheights in the Eastern Mediterranean Sea.

2 Airborne Laser Altimetry

2.1 Technique

Airborne laser altimetry is based on georeferencinga laser beam carried by an aircraft, yielding a 3Dvector between the aircraft and the ground surface.The key elements are highly precise position and at-titude of the aircraft. For this purpose, the latter wasequipped with an array of four GPS antennas. Oneantenna is used for trajectory recovery and as a ref-erence for moving baseline processing, where theother three antennas are the remote receivers, yield-ing three moving vectors used in attitude determina-tion. As the attitude estimation is based on GPS vec-tors, its accuracy is dependent on the geometry ofthe antenna configuration. With a baseline accuracyon the order of 0.01 m and baseline lengths of about10 m, the expected angular accuracy is 0.05°, whichis sufficient for most applications.

2.2 Preliminary Field Measurement Results

A detailed airborne laser altimetry campaign wascarried out around the island of Crete in the frame-work of the EU project GAVDOS in 2003 (Fig. 3).The aim of the latter was the establishment of aEuropean sea-level monitoring and radar altimetercalibration site for JASON-1, ENVISAT andEURO-GLOSS (Pavlis et al. 2004). The calibrationsite is located on the isle of Gavdos, at a crossoverof two JASON-1 ground-tracks. During the airbornecampaign an area of 200x200 km adjacent to the

Hellenic Trench was covered by 24 flight lines per-formed at an altitude of 700 ft (210 m) with a laserprofiler operated at an observation rate of 1 kHz.

In order to derive a time-independent sea surfacetopography from instantaneous SSHs obtained byaltimetry, several corrections have to be applied, es-pecially for tides and atmospheric effects (inversebarometer effect). In a first computation, tide cor-rections based on the GOT00.2 tide model havebeen applied. Inverse barometer corrections havebeen computed over the entire Mediterranean Seausing ECMWF atmospheric pressure data.

Crete

Gavdos

Fig. 3. Flight-tracks with color-coded SSH profiles fromairborne laser altimetry around Crete.

Crete

Gavdos

Fig. 4. Sea surface topography obtained from airbornelaser altimetry SSH profiles of Fig. 3 (white lines).

Sea surface height – EGM96 [m]

Crete

Gavdos

Height difference [m]

Gavdos

Mean diff. : - 0.41m

Std dev. : 0.39m

Sea surface height – EGM96 [m]

Crete

Gavdos

Fig. 5. Top: Height differences between SSH profilesfrom airborne laser altimetry and EGM96 geoid heightsaround Crete. Center: Distribution of the height differ-ences. Bottom: Height differences shown as surfaces.

Sea surface height – GGM02 [m]

Crete

Gavdos

Gavdos

Height difference [m]

Mean diff. : - 0.17m

Std dev. : 0.38m

Sea surface height – GGM02 [m]

Crete

Gavdos

Fig. 6. Top: Height differences between SSH profilesfrom airborne laser altimetry and GRACE-basedGGM02 geoid heights around Crete. Center:Distribution of the height differences. Bottom: Heightdifferences shown as surfaces.

The SSH results shown in Fig. 3 and Fig. 4 re-veal very strong gradients, with SSH decreasingfrom nearly 30 m in the North-East to 5 m towardsthe Hellenic Trench in the southern part, along adistance of 200 km only.

2.3 Comparisons Between Local SSH andGlobal Geoid Models

Height differences between the SSHs obtained byairborne laser altimetry and both the EGM96(Fig. 5) and the GRACE-based GGM02 (Fig. 6)geoid heights have been computed for the surveyarea around Crete. The distributions of the heightresiduals have similar standard deviation of about0.40 m for both geoid models. In this survey area,EGM96 is systematically higher than the sea surface(mean difference -0.41 m). The offset between theSSH and GGM02 is less pronounced (meandifference -0.17 m), meaning that the long- and mid-wave structure of the GGM02 geoid model fits theSSH better than does the EGM96. A strikinganomaly is the large positive difference around theisle of Gavdos, where the sea surface is about0.75 m above the EGM96 and 1 m above theGGM02 geoid. The highs and lows of the heightresiduals are too pronounced to be explained onlyby dynamic ocean topography effects and, therefore,seem to be related to local short-wave gravityanomalies that are not seen in both global geoidmodels.

3 Sea Surface Heights by GPS Buoysand Shipborne Multi-Antenna GPS

3.1 Technique

The GPS-equipped buoys deployed for ground-truthmeasurements of the SSH are lightweight buoys,carrying high-frequency L1/L2 GPS receivers(Fig. 7). The shell of the buoys is fabricated frommicrowave-transparent polycarbonate, so they canbe waterproof-sealed containing the receiver, anten-na and power supply.

40 cm

Fig. 7. Left: GPS-equipped buoy containing receiver,antenna and battery. Right: sailing boat equipped withan array of four GPS antennas (arrows) for precise po-sition and attitude determination.

Highly-precise GPS positioning of the buoys andthe boat is achieved by simultaneously operating thebuoy receivers, the receivers aboard the boat(Fig. 7) and several permanent terrestrial GPS refer-ence stations with coordinates known in the ITRFreference frame. All receivers are measuring at asampling rate of 1 Hz. The kinematic positions ofthe buoys and the boat are computed through differ-ential carrier phase processing with respect to thereference stations. In addition the multi-antennaconfiguration abord the boat allows for precise atti-tude determination, which is of great importance forprecise SSH retrieval.

3.2 Preliminary Field Measurement Results

Two campaigns for shipborne/buoy GPS SSH sur-veys have been carried out in the North Aegean Seain 2004/2005, totaling more than 1000 nm of shiptracks (Fig. 8). The survey area was chosen in thevicinity of the North Aegean Trough (NAT), whichis a tectonic graben-like feature characterized by azone of deep water reaching 1500 m and trendingfrom north-east to south-west in the North AegeanSea. The NAT is considered to form the westerncontinuation of the seismically active North Anato-lian Fault Zone (McNeill et al. 2004).

GPS reference station

Tide-gauge

North

Aegea

n Tro

ugh

Sea surface height [m]

Fig. 8. GPS surveys in the North Aegean Sea. Boat-tracks with color-coded SSH profiles from combinedshipborne/buoy GPS observations. Black lines: JA-SON-1 ground-tracks. Background: bathymetry withdeep-water zone of the NAT.

In order to derive the sea surface topographyfrom the instantaneous SSHs, the same procedure asdescribed in Chap. 2.2 has been applied. In addi-tion, the local tidal effects have been determined byusing own tide gauges installed in the survey area.

The SSH results (Fig. 8) reveal that the bathy-metric low of the NAT is associated with a distinctdepression of the SSH which reaches a minimum of37.5 m above the WGS84 ellipsoid, while the SSHin the surrounding area is more than 39 m andreaches even more than 40.5 m towards the north ofthe survey area.

3.3 Comparisons Between Local SSH andGlobal Geoid Models

Height differences between the SSHs obtained byshipborne/buoy GPS measurements and both the

Sea surface height – EGM96 [m]

North

Aeg

ean T

roug

h

Height difference [m]

Mean diff. : - 0.75m

Std dev. : 0.44mNorth Aegean Trough

Fig. 9. Top: Height differences between SSH profilesfrom shipborne/buoy GPS data and EGM96 geoidheights. Bottom: Distribution of the height residuals.

EGM96 (Fig. 9) and the GRACE-based GGM02(Fig. 10) geoid heights have been computed for thesurvey area in the North Aegean Sea.

While EGM96 is systematically higher than thesea surface (mean difference -0.75 m), the offset be-tween the SSH and GGM02 is less pronounced(mean difference -0.19 m), which means that thelong-wave structure of GGM02 better fits the seasurface. A striking anomaly is the large negative dif-ference over the North Aegean Trough, where thesea surface is more than 1.5 m below EGM96 and1 m below GGM02. This distinct low of the heightresiduals can be considered as being too pro-nounced to be explicable only by dynamic ocean to-pography effects and seems, therefore, to be relatedto a local short-wave gravity anomaly that is not de-tected by both global geoid models.

Sea surface height – GGM02 [m]

North

Aegea

n Tro

ugh

Height difference [m]

Mean diff. : - 0.19m

Std dev. : 0.42m

North Aegean Trough

Fig. 10. Top: Height differences between SSH profilesfrom shipborne/buoy GPS data and GGM02 geoidheights. Bottom: Distribution of the height residuals.

3.4 Comparison Between JASON-1 andGround-Truth GPS Data

For calibration and validation purposes of radar al-timeter missions, the survey area has been chosen inthe vicinity of JASON-1 ground-tracks. Dedicatedbuoy measurements were performed along thesetracks, including deployments with direct JASON-1cross-overs, which provide precise ground-truthSSH information during the overflight.

First comparisons between JASON-1 radar al-timeter SSH data and the preliminary results ofcombined in situ shipborne and buoy GPS data havebeen successfully performed. Similar tide correc-tions have been previously applied to both datasets,as well as a cross-track correction to account for thehorizontal offset between the radar altimeterground-points and the GPS profiles. The observedheight differences between the GPS buoy data andJASON-1 data varied between 0.10 and 0.12 m onsix overflights. These values are in accordance withNASA/CNES results obtained at calibration sites.

An example of an encounter situated directlyabove the NAT is shown in Fig. 11. The meanheight difference along this profile is0.142 ± 0.042 m standard deviation, whereas themaximum and minimum height difference reached0.224 m and 0.085 m, respectively. This significantvariation of the height differences along the profileseems to be geographically-correlated and is mostlikely due to the effect of the different spatial reso-lutions of the two methods amplified in regions withstrong sea surface height gradients and strong gradi-ent variations.

Difference

JASON-1

GPS

Fig. 11. Comparison between a JASON-1 SSH profile(green line) and the results of combined in situ ship-borne/buoy GPS data (red line). The encounter (closestapproach) is marked by the dashed blue line on thegraph and the blue square on the map.

4 Conclusions

The airborne laser altimetry and shipborne/buoyGPS methods for precise SSH determination in lo-cal areas have been successfully developed. FirstSSH results are extremely promising in terms of ac-curacy and repeatability, leading to a local high res-olution sea surface topography solution at cm level.

Comparisons with JASON-1 radar altimeter dataalong dedicated profiles showed encouraging resultsbut revealed geographically correlated variations ofthe height difference between 0.1 and 0.2 m in areaswith strong sea surface height gradients.

Comparisons between EGM96 and GGM02global geoid models showed considerable local dif-ferences of up to 0.7 m in the Eastern Mediter-ranean. When comparing the SSHs to the EGM96and GGM02 geoid heights, a significant improve-ment of the GRACE-based GGM02 model in thelong- and mid-wave structure was observed, leadingto a reduction of the mean difference. Comparisonwith the SSH's also revealed local anomalies in thevicinity of the Hellenic Trench and the NorthAegean Trough, reaching more than 1 m in these re-gions of high geodynamic activity. These featurescan be considered as being too pronounced to be ex-plicable only by dynamic ocean topography effects.Therefore they can be considered as indications fordistinct mass anomalies causing local gravityanomalies in the high frequency domain that are notdetected by the global geoid models.

References

McNeill LC, Mille A, Minshull TA, Bull JM, Kenyon NH,Ivanov M (2004) Extension of the North Anatolian Faultinto the North Aegean Trough: Evidence for transtension,strain partitioning and analogues for Sea of Marmarabasin models. Tectonics, Vol. 23, TC2016, 2004

Pavlis EC, Mertikas SP and the GAVDOS-Team (2004) TheGAVDOS Mean Sea Level and Altimeter Calibration Fa-cility: Results for Jason1", 3rd Jason special issue, Ma-rine Geodesy, Vol. 27, No. 3-4, pp. 631-655

Rio MH (2004) A Mean Dynamic Topography of theMediterranean Sea Estimated from the Combined use ofAltimetry, In-Situ Measurements and a General Circula-tion Model. Geophysical Research Letter Vol. 6, 03626.

Tapley B, Ries J, Bettadpur S, Chambers D, Cheng M, CondiF, Gunter B, Kang Z, Nagel P, Pastor R, Pekker T, PooleS, Wang F (2005) GGM02 – An Improved Earth GravityField Model from GRACE. J Geod DOI 10.1007/s00190-005-0480-z

Tziavos IN, Andritsanos VD (1999). Recent Geoid Computa-tions for the Hellenic Area. Phys. Chem. Earth (A), Vol-24, No. 1, pp. 91-96.

Tziavos IN, Vergos GS, Kotzev V, Pashova L (2005) Meansea level and sea surface topography studies in the BlackSea and the Aegean. International Association of GeodesySymposia, Vol. 129, Jekeli C, Bastos L, Fernandes J(eds.), Gravity Geoid and Space Missions 2004,Springer–Verlag Berlin Heidelberg, pp. 254-259