The Center for Coastal &OceanMapping-Joint Hydrographic Center (CCOM-JHC), University of New Hampshire is in charge of acquiring multi-beam bathymetry of US margins for any potential claims of extended continental shelves under Article 76 of the UN Convention on the Law of the Sea. The data have been collected using Simrad EM120 and EM121A, Reson 8150 and SeaBeam 2112 multi-beam echosounders (MBES). More than a million square kilometres have been mapped in the Atlantic, Gulf of Mexico, Arctic, Bering Sea, Gulf of Alaska and in the western Pacific. Although bathymetry is the primary objective of each cruise, acoustic backscatter is readily available from each of the MBES systems, so these data are also processed. Processing this immense amount of data has required careful thought, resulting in standardised processing streams. The discussion below is a synopsis of those processing streams.

Bathymetry Processing
Quality control is maintained by running a calibration patch test prior to each cruise. Static offsets in sensor locations and timing are checked against those already in the MBES system. A cross-line is first run perpendicular to the planned survey lines so that each survey line crosses it and a cross-line analysis is made at each crossing. Each cross-line analysis is a comparison of interpolated depths in the digital terrain models (DTMs) of the cross-line with corresponding measured soundings from the raw MBES file of the survey line. If the comparison of depths shows a difference greater than 1%, the survey is halted and a determination is made on why this precision has not been achieved. Invariably, the problem is minor errors in the static offsets and/or the sound-speed profile.

A suitable mixture of software is important for mapping missions. In particular, the availability of multiple tools for processing deep-water data provides the means to compare processing for flaws, to check on operator actions and to utilise the unique aspects of the various components of the processing to improve the quality control of the data. For example, for Simrad EM120 and EM121A data we employ SwathEd and SAIC/SABER, Reson 8150 and SeaBeam 2112 data are processed using CARIS/HIPS and all initial derivative products are processed using IVS3D software. A primary concern is to preserve effort at all stages, maintain single reference copies for the data and provide the required files for subsequent backscatter processing.

The lack of reliable uncertainty estimates for deep-water mapping systems has resulted in a fairly conventional processing approach based on inspection of point data, rather than the use of a more automated method. However, because the data densities are low, the processing cost can be readily absorbed into collection time and, particularly in the Arctic, the interpretation of features observed at the limits of the data can be a significant part of extracting the best from the available information, necessitating a manual approach.

All of the MBES lines are processed at sea in near real-time. At the end of each survey line, the raw datagram file is copied from a server to a processing computer. The cleaned soundings are used to populate submap DTMs at the highest resolution allowed by the sounding density. Initially, the submaps are assembled in Mercator (or Polar Stereographic for the Arctic) projection; however, ASCII XYZ (longitude, latitude, depth) values for each sounding from each line are extracted and gridded into non-projected georeferenced maps. Later, these maps can be projected as needed.

Acoustic Backscatter
Acoustic backscatter is acquired as an integral part of the MBES surveys. Backscatter contains important information about the sea-floor morphology and geology. If the two-way travel time of the transmitted acoustic pulse is the primary observation from which bathymetry is derived, the intensities of the received time series are the starting point for the backscatter processing. Simrad EM120 and EM121A and Reson 8150 MBES record one complete time series of received intensities per acoustic beam, normally referred to as beam-time-series or snippets. The SeaBeam 2112 system only records one average value of intensity per beam, which is referred to as beam-average backscatter.
Backscatter processing uses Geocoder software, developed at CCOM-JHC, designed specifically for backscatter analysis. The main objective of the backscatter processing is to convert the raw intensity observations into estimates of sea-floor backscatter strength per unit area in decibels. Initially, all the gains that were applied during acquisition are removed from the raw intensity observations. The observations are corrected based on the terms of the sonar equation, which include transmission loss, area of insonification, transducer source level and transmit and receive beam patterns. The acquisition geometry is taken into account by correcting the position of all beams based on the navigation, transducer attitude and sound-speed profiles and by correcting the backscatter values for changes in sea-floor slope, calculated from the bathymetric model generated during the bathymetry processing.

Once the backscatter strength has been calculated, the next step in the processing involves the removal of the backscatter angular response (the variation in backscatter strength with the angle of incidence). The angular response is an intrinsic property of the sea floor, so different sea-floor types have different angular responses. However, the mosaic should be uniform across the swath if the sea floor is uniform. This angular variation is compensated by the use of algorithms for angle-varying gain corrections. Geocoder implements several approaches for this processing and suggests a default algorithm, which reduces the normal-incidence artefact. Finally,the artefacts due to the overlap among adjacent lines are minimised by the use of a feathering algorithm. This algorithm blends the overlapping lines by choosing the highest priority sample, giving lower priority to samples that are close to the normal incidence or are on very shallow grazing angles.
?Geocoder automatically applies the proper adjustments to the backscatter observations, so only minimum user interaction is required. The processing starts with the raw acquisition files and the result is a comprehensive analysis of the acoustic returns.

Full-resolution DTMs and Mosaics
The processed bathymetry and backscatter data are assembled at the completion of each cruise into full-resolution DTMs and co-registered backscatter mosaics. The DTMs and mosaics can be projected and re-projected by standard GIS packages (ESRI, Geomedia, PCI, etc.) as well as specialised software (IVS3D Fledermaus, Global Mapper, etc.). The DTMs and mosaics allow easy visualisation and interpretation of each data set.
A primary issue is to determine which grid-cell resolution provides the highest resolution, justified by the sounding density. Invariably, a completed survey spans a large range of water depths, often from less than a few hundred metres to 4,000m or deeper. For instance, the Arctic surveys range in water depths of 38–3,970m. Consequently, the highest grid-cell resolution for an overview DTM ranges from 6m (the 38m depths) to 100m (the 3,970m depths), depending on what part of the area is being gridded into a DTM. For a regularly spaced grid DTM, an overview of the entire area must use the coarsest grid-cell resolution for the entire DTM, thus creating a lower resolution in the shallower areas. Although not used in the US UNCLOS surveys, one could use an irregular array of points (i.e. TINs) for gridding the data. The pros and cons of regularly spaced grids versus TINs is beyond the scope of this paper, but should be considered before a decision on the type of grid is made.

An example of a regularly spaced grid of the Gulf of Alaska bathymetry is shown in Figures 1 and 2 (a, top). The measured water depths in the data set ranges from 237m to 4238m indicating that the data resolution allows grid resolutions of 12m (shallow) to 100m (deepest). Figure 3 is an example from the Barrow margin, Arctic Ocean. In practice, the mapping project subdivides each region to be mapped into smaller submaps of various pixel resolution based on the expected water depths. Each submap is gridded at its optimum grid-cell resolution determined from the sounding densities. These provide archives of submap DTMs with the highest resolution allowed for each submap.
Once the optimum grid-cell resolution is determined for the overview map (100m in the Gulf of Alaska example), the acoustic-backscatter mosaic is draped over the bathymetry as a georeferenced texture map (Figures 2 b, bottom and 4) using IVS3D software. The optimal resolution for the backscatter mosaic is normally higher than the grid-cell resolution for the bathymetry. The optimum backscatter resolution is calculated based on the bathymetric model, transmit and receive beam widths, transmit pulse length, and on the choice of beam average or beam time series backscatter. The advantage of this approach is that each pixel on the computer screen has a longitude, latitude, depth and backscatter value and the data can be visualised in 3D at any desired vertical exaggeration and at any viewing resolution or view angle. An interpreter can therefore simultaneously investigate the bathymetry and backscatter to better understand the nature of the sea floor.

Metadata
Metadata are critically important for this project because it is likely that there will be years between the time of data collection and processing and the time the data will be analysed to develop an extended continental shelf submission. Consequently, each survey line of raw multi-beam data has a complimentary metadata file that follows the Federal Geographic Data Committee (FGDC) metadata standard ( 8 1). The raw multi-beam data and metadata are archived at the National Geophysical Data Center of NOAA as well as at the University of New Hampshire CCOM-JHC. The archives provide secure storage for the raw data, and are available to the public.

Summary
Vast amounts of new multi-beam bathymetry and associated acoustic backscatter are being collected to support potential US extended continental shelf claims under Article 76 of the UN Convention on the Law of the Sea. The new data have been and are continuing to be collected from US continental and insular margins. The processed bathymetry and backscatter data, together with associated metadata, are posted on the web ( 8 2) within a few weeks of the completion of each cruise in a variety of data formats so that the marine science community, and any other interested party, has access to them in a timely fashion. The raw multi-beam data and associated metadata are archived at the NOAA National Geophysical Data Center where they are also publicly available.