Not all bathymetry grids are created equal

Transcription

1 Bathymetry grids Not all bathymetry grids are created equal K. M. Marks NOAA Laboratory for Satellite Altimetry, Silver Spring, MD 20910, USA W. H. F. Smith NOAA Laboratory for Satellite Altimetry, Silver Spring, MD 20910, USA Corresponding author: Karen M. Marks NOAA Laboratory for Satellite Altimetry E/RA 31 SSMC-1, Sta Silver Spring, MD USA Phone: ext. 124 Fax: Karen.Marks@noaa.gov Submitted to: Marine Geophysical Researches, GEBCO Special Issue August 25, 2004

2 Bathymetry grids Abstract We compare and evaluate six bathymetry grids: DBDB2 (Digital Bathymetric Data Base; an ongoing project of the U. S. Naval Research Laboratory), ETOPO2 (Earth Topography; U. S. National Geophysical Data Center 2001), GEBCO (General Bathymetric Charts of the Oceans; British Oceanographic Data Centre 2003), GINA (Geographic Information Network of Alaska; Lindquist et al. 2004), S&S (Smith and Sandwell 1997), and S2004 (Smith, manuscript in preparation). DBDB2 and ETOPO2 are both based on the S&S grid, but we find that their short wavelengths are attenuated by a shift in registration. Further, ETOPO2 is misregistered in both latitude and longitude, and DBDB2 displays data errors that result from ingesting regional datasets that already contain errors. Short wavelengths in the GINA grid are also attenuated, but the cause is not known. The GEBCO grid is interpolated from 500 m contours that were digitized from paper charts at 1:10 M scale, so it is artificially smooth. Contour maps plotted from the GEBCO grid are attractive. The S&S grid, derived from satellite altimetry and ship data combined, provides high resolution mapping of the seafloor, even in remote regions; and in the S2004 grid, GEBCO is incorporated over shallow depths and polar regions. Even higher resolution is available from swath bathymetry surveys, but these are very local and cover only a small fraction of the seafloor. Key words: Bathymetric grids, bathymetry, DBDB2, ETOPO2, GEBCO, GINA, satellite bathymetry, Woodlark Basin Introduction What s a scientist who needs a bathymetry grid to do? There are at least six global digital bathymetry grids available to choose from today (DBDB2, ETOPO2, GEBCO, GINA, S&S, and S2004). These grids are diverse in their sources of data- ranging from ship soundings to satellite altimetry- and some are augmented with regional or local surveys as well. The grid spacing and registration of each grid is different, and one may be more suitable than another for any given purpose. We strive to make the scientist s choice of which bathymetry grid to use a more informed one. In this study, we list the various attributes of each global grid in a succinct table (see Table 1). We compare the grids to determine their strengths and weaknesses, and we point out any problems associated with the grids, or with their construction. We use local grids in the Woodlark Basin (Figure 1) for additional comparison. Bathymetry Grids and Associated Problems GEBCO The General Bathymetric Charts of the Oceans (GEBCO) grid (British Oceanographic Data Center 2003, see Table 1) is a 1-minute grid prepared from the bathymetric contours of the world s oceans that were originally available as a series of paper maps at 1:10M scale, and later as digital contours in the GEBCO Digital Atlas. These maps were

3 Bathymetry grids contoured, by hand, from both digital and analog ship soundings. The ship track coverage, however, is sparse, irregular, and of uneven quality and navigational control. The red lines in Figure 1 (and also plotted for comparison in Figure 2a) locate ship tracks that GEBCO bathymetrists used to produce the contour map of the Woodlark Basin study area. We do not know the locations or density of the points used to form the contours, only that those digital (or analog) data points lie along the track lines. National Geophysical Data Center (NGDC) (Natl. Geophys. Data Center 2003) ship track coverage (Figure 2b) differs from GEBCO coverage because NGDC s database is digital, and it also contains data from some more recent surveys. We show an image of the GEBCO grid in Figure 3. Even though the grid spacing is 1- minute, it has a generally smooth appearance. This is because the 1:10 M scale paper maps used in the digitization were only contoured at 500 m intervals. At this scale, contours can only be drawn about 3 mm apart. Thus the map scale alone limits the horizontal resolution to about 30 km, and the slope of the ocean floor to about 1 degree. Abyssal hills will not be detected because their maximum relief is only 300 m (Goff 1991). Interpolation and gridding of these 500 m contours at a 1-minute spacing cannot recover seafloor details that were never present in the paper charts. The contours appear as terraces in the grid- and are particularly visible around the Louisiade Plateau. The contour values also appear as spikes in the histogram of GEBCO depths (Figure 4a). S&S In 1997, Smith and Sandwell published a 2-minute Mercator-projected grid (S&S) based on bathymetry derived from satellite gravity data, combined with ship measurements (Table 1). It is possible to map the seafloor from space because marine gravity anomalies reflect the underlying topography. The dense satellite track coverage (Figure 2c) and ship data (Figure 2d) permitted even the most remote and inaccessible regions of seafloor to be mapped in detail for the first time. We note this ship track coverage is more complete than is currently available at NGDC (Figure 2b). Smith and Sandwell were able to obtain ship surveys from investigators to use in their bathymetry solution, but these data are not yet available to the public via the NGDC repository. The S&S grid is shown in Figure 5. Short-wavelength tectonic details of the seafloor in the Woodlark Basin are resolved (e.g., north-south trending lineations along about 154.1º E, 155.1º E, and 156.4º E are fracture zones, and small circular bathymetric highs are seamounts). The pervasive bumpy background fabric, while due in part to noise, may also be related to small-scale morphology of the ocean floor (Goff et al. 2004). This bumpy fabric is not present in the GEBCO bathymetric anomalies (Figure 3) because the ship tracks (Figure 2a) are too widely-spaced to resolve short-wavelength tectonic features, and because of the 500 m contour interval. The correlation of satellite-derived gravity with bathymetry should be poor at wavelengths longer than about 160 km, and so Smith and Sandwell rely on interpolation

4 Bathymetry grids of conventional soundings to constrain the longer wavelengths in their grid. Since their track control (Figure 2d) is very different from that used in GEBCO (Figure 2a) one might expect differences in the two grids at large scales. Figure 6 shows low-passfiltered versions of the S&S grid and the GEBCO grid, showing good agreement of the two at wavelengths longer than 160 km. We conclude that the significant differences in the two products lie chiefly at shorter scales, where the S&S product captures information from satellite gravity. We also examined the distribution of depths in the S&S grid. Figure 4b displays a reasonable distribution of depths except for spikes at 4500 m and 0 m. A region of flat sediments in the Coral Sea (southwest portion of the study area) account for the large number of 4500 m depths. The spike at 0 m occurs because Smith and Sandwell constrained the land-sea boundary in their grid with the high-resolution World Vector Shoreline (Wessel and Smith 1996). ETOPO2 The ETOPO2 2-minute bathymetry grid (Table 1) is a product of NGDC. It was assembled from the S&S grid between 64º N and 72º S, from the U.S. Naval Oceanographic Office s (NAVOCEANO) Digital Bathymetric Data Base Variable Resolution (DBDBV) data south of 72º S, and from the International Bathymetric Chart of the Artic Ocean (IBCAO) data north of 64º N. Land topography is from the GLOBE database. Figure 7 is an image of ETOPO2 data in the Woodlark Basin study area. This image looks very similar to Figure 5, because ETOPO2 is based on the S&S grid at these latitudes. Closer inspection, however, reveals a slightly smoother appearance to the ETOPO2 grid- for example, note how the fracture zones and seamounts are not as crisp. This smoothing results from NGDC s interpolation of the pixel-registered S&S grid onto a grid-registered ETOPO2 grid. Figure 8a shows a waveform with amplitudes of 1 and -1 at odd arc-minutes. This is the Nyquist-wavelength signal in a pixel-registered grid at 2-arc-minute sampling, such as the S&S grid. If this waveform is sampled at even arc minutes (as in a grid-registered 2- arc-minute grid such as ETOPO2) the waveform vanishes. Amplitudes at longer wavelengths are proportionally attenuated as well, as shown by the theoretical transfer function (Figure 8b) for moving from a pixel to a grid registration. The spectral density plot (Figure 8c) of ETOPO2 exhibits this attenuation as a reduction of power at wavelengths shorter than about 20 km, as compared to S&S. The transfer function models the ratio of ETOPO2/S&S data well (Figure 8d) at wavelengths >100 km. There is another problem with the ETOPO2 grid: misregistration in latitude and longitude. This occurred while the global grid was being assembled from its major components at NGDC. We demonstrate the systematic offset in Figure 9a. The black lines contour the S&S grid over an enlarged area in the Coral Sea (dash-dot outlined in

5 Bathymetry grids Figure 1). The red lines contour ETOPO2 bathymetry anomalies, which are systematically offset to the northeast. DBDB2 The Naval Research Laboratory (NRL) at Stennis Space Center in Mississippi produced the DBDB2 (Version 2.2) 2-minute bathymetry grid (Table 1). DBDB2 is based on the S&S grid, and is augmented with selected regions of higher resolution bathymetric data from Geoscience Australia, IBCAO, Sung Kyun Kwan University, and others. DBDB2 bathymetry is shown in Figure 10. Like ETOPO2, DBDB2 was interpolated from S&S bathymetry onto a grid-node registration, so it demonstrates the same smoothing as a result. The reduced power in DBDB2 gridded data caused by moving to a grid registration is demonstrated in Fig. 8c. In Figure 8d, the DBDB2/S&S amplitudes are higher than those of the transfer function at shorter wavelengths. This is due to the higher-resolution Australian bathymetry and topography data incorporated into the DBDB2 grid in our study area. Unlike ETOPO2, the DBDB2 grid is correctly registered in latitude and longitude. Blue contour lines from DBDB2 are properly located on top of black contour lines from the S&S grid in Figure 9b. We have identified a different problem with the DBDB2 (Version 2.2) data, however. Red arrows in Figure 10 point to ship tracks that are evident in the DBDB2 bathymetry anomalies. These tracks result from ship track soundings being incorrectly merged into the bathymetry grid. Regional data from the Australian bathymetry and topography grid (Petkovic and Buchanan 2002) (Figure 11) that were incorporated into the DBDB2 solution show the same ship tracks, so these imperfect data were simply ingested into DBDB2, errors and all. We obtained detailed swath bathymetry data (A. Goodliffe, personal communication) covering the Woodlark Basin (Goodliffe et al. 1999) (Figure 12). It appears that the western portion of the detailed swath bathymetry survey was incorporated into the Australian bathymetry and topography grid (compare Figures 11 and 12). Yet the detailed swath bathymetry data do not contain the ship tracks that are evident in DBDB2 and the Australian bathymetry and topography grids. The higher resolution swath bathymetry data do not improve the DBDB2 solution much because they are decimated when gridded at a 2-minute spacing. We note the fracture zones and seamounts that are resolved in the S&S grid are very clear in the swath bathymetry. Recently, NRL produced DBDB2 Version 3.0. In this version, an attempt was made to remove the bad ship tracks in the western Woodlark Basin from the Australian bathymetry and topography data prior to incorporation into the DBDB2 solution. The transition between S&S and Australian data occurs at a greater depth in Version 3.0, causing shallower topography to be smoother than in Version 2.2, in regions not covered by the detailed swath bathymetry. Version 3.0 also has a loss of spectral power (similar to DBDB2 Version 2.2 shown in Fig. 8c) due to the shift in registration. DBDB2 Version 3.0 is correctly registered in latitude and longitude.

6 Bathymetry grids GINA A new attempt to produce a global topography dataset by combining existing grids without a loss in resolution has been undertaken by Lindquist et al. (2004). The resulting Geographic Information Network of Alaska (GINA) topography grid (Table 1) was formed by resampling and merging GTOPO30, S&S, and IBCAO data onto a pixelregistered 30 arc-second grid. Even though the GINA grid is correctly registered in longitude and latitude, and is sampled on a pixel-registered grid that should include the original S&S points, it still exhibits a reduction in power at shorter wavelengths (Figure 8c). Unfortunately, the published documentation (Lindquist et al. 2004) does not reveal details of how the GINA grid was constructed, so we are unable to determine the cause of the smoothing. It is possible that the amplitudes of the GINA/S&S curve (Figure 8d) are higher than those of the transfer function because the 30 arc-second GTOPO30 land data may have been incorporated in the GINA grid covering our Woodlark Basin study area. S2004 Walter Smith developed a new 1-minute global topography grid (S2004, manuscript in preparation) that combines the Smith and Sandwell (1997) and GEBCO grids (Table 1). Smith interpolated the S&S data from its original pixel-registered and Mercator-projected 2-arc-minute grid onto GEBCO's grid-registered, geographical, one-arc-minute sampled grid; attempting to preserve all the short-wavelength power in the S&S grid. Interpolation was by Fourier transform in the east-west direction, resulting in a sinc function interpolant, with an Akima spline used to interpolate from Mercator to geographical spacing in the north-south direction. This resulted in S&S values at fullresolution but on the GEBCO coordinates between ±72º latitude. This was then blended with the GEBCO grid, using GEBCO poleward of 72º and shallower than 200 m depth (and on land), and S&S equatorward of 70º and deeper than 1000 m depth. Areas in between were smoothly blended with a cosine taper. The result achieves global coverage and one-arc-minute geographical coverage, while capturing the seafloor texture of satellite altimetry in deep water areas equatorward of 70º. The spectral density of the S2004 grid (Figure 8c) is slightly lower at shorter wavelengths than for S&S. This is because GEBCO data were added to the S2004 solution. The amplitudes of S2004/S&S in Figure 8d are slightly less than one at shorter wavelengths. This demonstrates how the short wavelength power of S&S is retained in the S2004 grid, but is slightly lowered by the incorporation of GEBCO data. An advantage of the S2004 grid is that it is on a geographic grid, which users may find convenient. Satellite altimeter track coverage can be spotty near land (see Figure 2c) and altimetry is poorly correlated with bathymetry in shallow coastal areas. In some parts of the world the GEBCO grid production team put new effort into capturing information from shallow water contours on navigational charts. For these reasons, the GEBCO grid may be superior to S&S in some shallow water areas. S2004 attempts to capture the best information from both products.

7 Bathymetry grids Other Grid Problems Here we note other problems observed in the grids. In places, the S&S grid displays bad ship tracks (inside red circle in Figure 13). These occur where ship bathymetry data are too poor to be successfully incorporated into the gridded bathymetry solution. Future versions of the S&S bathymetry solution will omit these bad tracks. If these features lie in water deeper than 1000 m and equatorward of 70º, the same errors should appear in the S2004 grid. In the GEBCO grid, regions were interpolated separately and then combined to form the global grid. Seams may show up where some of these regions where patched together. Two such seams are evident southeast of our study area (red arrows point to seams in Figure 13). The tectonic features of the seafloor mapped by the GEBCO grid may appear different in extent and shape than they actually are. An example of this is shown in Figure 14, where the S&S topography grid maps the northwest-southeast trending Foundation Seamount Chain (Mammerickx 1992) (top panel), while the GEBCO grid displays an east-west trending seamount chain (middle panel). This occurs because only seamounts and troughs traversed by ship tracks are detected, so only these detected features are mapped by the grid. However, most seamounts are missed because they lie between ship tracks (bottom panel). A trough is likewise not detected in the GEBCO grid, yet is clearly mapped as a northwest-southeast trending trough in the S&S grid. Concluding Remarks Our comparison of six bathymetric grids reveals that for many purposes, the original S&S grid may be the best choice. Both the DBDB2 and ETOPO2 grids which are based on the S&S grid in fact degrade the solution by interpolating it onto a grid registration. ETOPO2 is also problematic because it is misregistered in latitude and longitude. Some of the higher resolution regional surveys ingested into the DBDB2 solution contain errors. We have also determined that the GINA grid has been smoothed. For researchers who prefer a 1-minute geographic grid with global coverage, S2004 is a good choice. An advantage of the original S&S grid, however, is that it encodes the sounding control information: pixels constrained by echo soundings have depths rounded to the nearest odd meter, while satellite-interpolated pixels have depths rounded to the nearest even meter. This allows the user to map the control locations, or to extract only the measured values and throw away the interpolated ones, if desired. This encoding is lost in all the grids interpolated from S&S, including ETOPO2, DBDB2, and S2004. Future versions of DBDB2 and ETOPO2 can be improved by honoring the pixel registration of the underlying S&S grid. The S&S grid can be improved by making a new solution that incorporates more recent ship data and high resolution regional and local surveys. It would be more user-friendly too if it were made available in a geographic grid. A new satellite mission capable of collecting higher resolution altimetry

8 Bathymetry grids data would yield a greatly improved bathymetry solution containing significant new details. Extending the bathymetry solution to latitudes higher than 72º, as in the S2004 solution, would be desirable. For purposes of displaying smooth 500 m bathymetric contours, the GEBCO grid is a good choice. Contours of the S&S grid are jagged because of the noise contained in the solution. Acknowledgements We thank Eric Bayler for suggesting we look at DBDB2. The following researchers kindly provided us with data: Dong-Shan Ko of NRL provided DBDB2, and Andrew Goodliffe, Fernando Martinez, and Brian Taylor of the School of Ocean and Earth Science, University of Hawaii, provided swath bathymetry in the Woodlark Basin. Graphics were created using the Generic Mapping Tools (GMT) software package (Wessel and Smith 1996). The views, opinions, and findings contained in this report are those of the authors and should not be construed as an official National Oceanic and Atmospheric Administration or U.S. Government position, policy, or decision. References British Oceanographic Data Centre (2003) Centenary Edition of the GEBCO Digital Atlas (CD-ROM). Published on behalf of the Intergovernmental Oceanographic Commission and the International Hydrographic Organization, Liverpool, UK. Goff JA (1991) A global and regional stochastic analysis of near-ridge abyssal hill morphology. J. Geophys. Res ,713-21,737. Goff JA, Smith WHF, and Marks KM (2004) The contributions of abyssal hill morphology and noise to altimetric gravity fabric. Oceanography, 17 (1), Goodliffe A, Taylor B, and Martinez F (1999) Data Report: Marine geophysical surveys of the Woodlark Basin region. In: Taylor B, Huchon P, Klaus A, et al., Proc. ODP, Init. Repts., 180, Ocean Drilling Program, Texas A&M Univ., College Station, TX, 1-20 [CD- ROM]. Jakobsson M, Cherkis NZ, Woodward J, Macnab R, and Coakley B (2000) New grid of Arctic bathymetry aids scientists and mapmakers. EOS Trans. Amer. Geophys. U., 81 (9), Lindquist KG, Engle K, Stahlke D, and Price E (2004) Global topography and bathymetry grid improves research efforts. EOS Trans. Amer. Geophys. U., 85 (19), 186. Mammerickx J. (1993) The Foundation Seamounts: tectonic setting of a newly discovered seamount chain in the South Pacific. Earth Planet. Sci. Lett., 113,

9 Bathymetry grids National Geophysical Data Center (2001) ETOPO2 Global 2 Elevation (CD-ROM). U.S. Dept. of Commerce, Natl. Oceanic and Atmos. Admin., Boulder, CO, USA. National Geophysical Data Center (2003) Worldwide Marine Geophysical Data, version (Geodas CD-ROM). Data Announce MGG-02, U.S. Dep. Of Commer., Natl. Oceanic and Atmos. Admin., Boulder, CO, USA. Petkovic P and Buchanan C (2002) Australian Bathymetry and topography grid (digital dataset). Geoscience Australia, Canberra. Smith WHF and Sandwell DT (1997) Global sea floor topography from satellite altimetry and ship depth soundings. Science, 277, 1,956-1,962. U. S. Geological Survey (1996) Global 30 Arc-Second Elevation Data Set (GTOPO30). EROS Data Center, Land Processes Distributed Active Archive Center, Sioux Falls, SD, USA. Wessel P and Smith WHF (1996) A global self-consistent hierarchical high-resolution shoreline database. J. Geophys. Res.. 101, Wessel P and Smith WHF (1998) New, improved version of Generic Mapping Tools released. EOS Trans. Amer. Geophys. U. 79, 579.

10 Bathymetry grids Table 1. Grid attributes. Grid Spacing Node Projection Coverage Based on DBDB2 1 2 grid geographic global S&S 5 below -1000m depth and between 72º N-66º S ETOPO2 2 2 grid geographic global S&S 5 between 64º N-72º S GEBCO 3 1 grid geographic global hand-drawn contours GINA 4 30 pixel geographic global S&S 5, IBCAO 7, GTOPO30 8 S&S 5 2 longitude pixel Mercator ± 72º latitude satellite gravity S grid geographic global S&S 5 below -1000m depth and equatorward of 72º, GEBCO 3 in shallow water and polar regions 1 NRL (2003); we examined both Versions 2.2 and NGDC (2001); 3 GEBCO (2003); 4 GINA (2004); 5 Smith and Sandwell (1997); 6 Smith; manuscript in preparation 7 Jakobsson et al. (2000); 8 USGS (1996);

11 Bathymetry grids Woodlark Basin Louisiade Plateau AUS Coral Sea meters Fig. 1 GEBCO contour map of Woodlark Basin study area. Contour interval is 500 m. Dashed region is covered by swath bathymetry in Figure 12. Dash-dot region is enlarged in Figure 9. Red lines are locations of ship tracks used in the GEBCO compilation.

12 Bathymetry grids a) Ship tracks used in the GEBCO compilation, b) ship tracks in the NGDC GEODAS database (National Geophysical Data Center 2003), c) satellite and d) ship tracks used in the Smith and Sandwell (1997) bathymetry estimation

13 Bathymetry grids GEBCO meters Fig. 3 Color shaded-relief image of GEBCO bathymetry, illuminated from the east. Note terraces evident in the image. These result from interpolation between contour lines.

14 Bathymetry grids counts x GEBCO a) b) 5 1 S&S meters meters Fig. 4 Histograms of a) GEBCO depths and b) S&S depths, at a bin width of 50 m. The GEBCO depths (a) spike at the 500m contour values. The spike at 4500 m in the S&S grid reflects flat sediments in the southwestern portion of the study area, while the spike at 0 m reflects the use of a high-resolution land-sea boundary.

15 Bathymetry grids S&S meters Fig. 5 Color shaded-relief image of S&S satellite bathymetry, illuminated from the east.

16 Bathymetry grids GEBCO S&S meters Fig. 6 Low-pass (wavelengths > 160 km) filtered bathymetry from a) GEBCO and b) S&S shows good agreement, despite the very different track control used in these two solutions (Figure 2).

17 Bathymetry grids ETOPO meters Fig. 7 Color shaded-relief image of ETOPO2 bathymetry, illuminated from the east.

18 Bathymetry grids a) b) Amplitude longitude, arc-min Amplitude 0.5 c) d) Spectral Density, m Wavelength, arc-min S&S DBDB2 ETOPO2 GINA S Wavelength, km Amplitude TF DBDB2/S&S ETOPO2/S&S GINA/S&S S2004/S&S Wavelength, km 10 Fig. 8

19 Bathymetry grids Figure 8. a) The pixel-registered S&S grid is sampled at odd meridians (black circles). If the S&S grid is interpolated to lie on even meridians (open squares) as was done in DBDB2 and ETOPO2, the amplitudes are reduced to zero at the Nyquist wavelength (4 arc-minutes), and amplitudes at other wavelengths are attenuated as well. For example, b) plots the transfer function for moving between odd (pixel) and even (grid) 2-arc-minute registrations. At a wavelength of 20 arc-minutes (~36 km at 11.5ºN), the amplitude is reduced by half. c) Attenuation of power in DBDB2 and ETOPO2 due to resampling the S&S grid is seen at shorter wavelengths in the spectral density plot. The GINA data are also attenuated at shorter wavelengths, but documentation does not suffice to explain why. The S2004 data are also slightly attenuated, because GEBCO data are incorporated. d) Amplitudes of ETOPO2/S&S and DBDB2/S&S match the transfer function (b) at wavelengths longer than 20 km; DBDB2/S&S and GINA/S&S have higher amplitudes at shorter wavelengths because the higher-resolution Australian bathymetry and topography data were incorporated into DBDB2, and 30 arc-second GTOPO30 data were incorporated into GINA. The S2004/S&S amplitudes remain close to one because no information was lost due to resampling.

20 Bathymetry grids a) b) C.I. = 100 m S&S Bathymetry, meters S&S ETOPO2 DBDB Fig. 9 S&S satellite bathymetry overlain by a) ETOPO2 contours and b) DBDB2 contours. The registration offset to the northeast is seen in the ETOPO contours. DBDB2 contours are correctly registered.

21 Bathymetry grids DBDB meters Fig. 10 Color shaded-relief image of the DBDB2 (Version 2.2) grid, illuminated from the east. Ship tracks visible in the Australian bathymetry and topography grid (Figure 11) are also present in DBDB2 (red arrows point to ship tracks), because these data were incorporated. Elsewhere, the DBDB2 grid is slightly smoother than the S&S grid because of amplitude attenuation caused by moving DBDB2 to a grid registration.

22 Bathymetry grids Australian Bathymetry & Topography meters Fig. 11 Color shaded-relief image of the Australian bathymetry and topography grid (Petkovic and Buchanan 2002), illuminated from the east. Red arrows point to ship tracks visible in grid.

23 Bathymetry grids Color shaded-relief image of swath bathymetry in the Woodlark Basin (Goodliffe et al. 1999). The image is illuminated from the north.

24 Bathymetry grids S&S GEBCO meters Fig. 13 Some bad ship tracks (circled) are present in the S&S grid, that are not in the GEBCO grid. Seams, due to patching together individually interpolated regions, are present in the GEBCO grid (red arrows point to seams) that are not in the S&S grid.

25 Bathymetry grids S&S GEBCO -40 GEBCO meters Fig. 14 The northwest-southeast trending Foundation Seamount chain and an unnamed trough are well resolved in the S&S grid. Only seamounts and troughs traversed by ship tracks (plotted in the bottom panel) show up in the GEBCO grid. Most seamounts in the GEBCO grid are missed because they lie between ship tracks, and seamounts that are traversed by ship tracks appear aligned along the tracks, rather than in their true orientation.