1. Introduction
Antarctica covers an area of some 13 million sq. km, an area rather larger than that of the USA (BAS, 1993). In only around 0.4% of this area is the rock exposed (Fox and Cooper, 1994), the rest being covered by a permanent ice sheet that is in places more that 4500 m thick. The Antarctic Ice Sheet is a major control on world sea level and contains enough water to raise sea level by perhaps 65 m.

In the last ten years mapping the topography of the surface of the Antarctic Ice Sheet has been advanced considerably through field surveying, satellite imagery and satellite altimetry. While mapping is, in places, still limited to scales smaller than 1: 10 000 000 digital maps are now available over the entire continent. Mapping of the true surface of the continent beneath the ice sheet has not, however, kept pace. Whereas present and planned satellites can give an almost complete description of the surface of the ice sheet, neither present nor planned satellites will be able to look through the ice sheet. For the foreseeable future, we are restricted to terrestrial methods of measurement, primarily radar and seismic sounding. The high logistic costs of working in Antarctica have prevented any one nation from attempting to sound the entire continent. Rather, we have seen a largely uncoordinated and patchy coverage develop as many nations produced regional surveys to support their other scientific activities. We are now arriving in a position where no one individual or institute has a complete knowledge of where data have been collected and which areas remain to be visited. Furthermore, bed elevation collected in aid of other scientific disciplines are beginning to be lost as researchers leave the community, and in a short time this will become too difficult to recover.

The BEDMAP project was conceived as an attempt to rationalise the coverage of ice thickness measurements over Antarctica, collect the data together and produce a new topographic model of the bed of the Antarctic Ice Sheet, to provide a new basis for all aspects of Antarctic geoscience.

2. Previous bed elevation mapping
Since 1983, one map of bed elevation in Antarctica has been widely adopted. This is Sheet 3 in the SPRI Glaciological and Geophysical Folio Series by Drewry and Jordan (1983). The contours were drawn from ice thickness data collected on sparse surface traverses and by airborne surveys, including a major airborne campaign involving Scott Polar Research Institute, National Science Foundation and Technical University of Denmark, which covered about one third of the continent at 100 km line-spacing. Contours of bed elevation were drawn over the entire continent although in many areas there were more than 500 km between measurements. Figure 1 shows the coverage of data that were used in this compilation in terms of the number of measurements available in each 100 km grid cell.

Despite its limitations and well-known shortcomings, the SPRI Folio Sheet 3 is still a commonly cited publication in papers from a wide variety of scientific disciplines. This is a clear indication that there is a major requirement for bed elevation data within the scientific community. While the SPRI compilation could be considered as the state-of-the-art in 1983, it is no longer adequate and is no longer the best we can achieve. Considerable advances in data and technology will now allow us to make big improvements in the coverage and accuracy of the product.

! new ice-thickness data have now been collected by many nations, covering many of the sectors not visited before 1983.
! Satellite altimetry has almost entirely removed the need to rely on inaccurate barometric altimetry
! Mapping has been refined so that areas that are classified as ice-free can now be easily incorporated into a bed elevation compilation
! Coastal bathymetry has been much improved and can be easily merged with continental bed elevation
! Advanced GIS systems now allow incorporation of many types of subsidiary data into a single determination of bed elevation

3. Scientific benefits
Perhaps the most immediate benefit of BEDMAP will be in the field of ice sheet modelling. Numerical modelling of the ice sheet covering Antarctica is the goal of many researchers and our politicians and taxpayers are expecting these modellers to provide concrete answers to difficult questions about the future ice sheet and sea level. The models are beginning to include far more realistic treatments of the physics involved, but mapping is not keeping pace. And while field workers are collecting new data each year, the modellers often find these too difficult to obtain and too time consuming to handle, with the result that the models are based on old and substandard base data. Indeed, a recent report produced by EISMINT identified lack of adequate bed elevation data as a major inhibitor of progress in ice sheet modelling. BEDMAP will fill this gap and present the modelling community with a variety of easily used maps and gridded data, reducing a major limitation on the models, namely the lack of a realistic map of bed elevation on which the models rest.

The benefits of an improved topography for the continent will not end with ice sheet modelling: as there are a wide range of other scientific disciplines for which it will be a major benefit.

Geology improved tectonic and structural interpretation
Gravity improved interpretation of gravity/geoid anomalies.
Magnetics improved interpretation of magnetic anomalies especially the continent-wide magnetic anomaly map, ADMAP.
Geophysics improved models of isostasy and sea level
Ice cores improved interpretation of the flow characteristics upstream of ice core drilling sites leading to improved dating of ice cores.

4. Requirements for ice sheet modelling
The Antarctic Ice Sheet is a complex non-linear system, and it is well known that its overall development can be strongly controlled by local basal conditions. For example, around 90% of ice discharge from the continent passes through a series of ice streams. Some of these lie in geological depressions while others do not appear to be structurally controlled. Ice streams are a major control on how the Antarctic Ice Sheet evolves in time and yet there is evidence (Retzlaff and Bentley, 1993) these ice streams can be Aswitched on@ or Aswitched off@ in only a few tens of years by subtle changes in sub-ice drainage patterns. Similarly, the evolution of ice shelves is strongly dependent on the distribution of seabed shoals that might act as nucleation sites for ice rises. To mimic reality ice-sheet models should use bed elevation data that takes account of such features wherever possible. Sadly, many of the models presently running are based on bed elevation maps derived from datasets which contained data-holes covering hundreds of thousands of square km. It is likely that entire mountains and basins were overlooked in these early maps.

Contemporary ice sheet modelling can be broadly classified into two types, fine resolution models dealing with local areas and concentrating on process studies, and whole continent models that attempt to reproduce past behaviour of the ice sheet and give indications about what might happen under future climates. Over most of the ice sheet, the stress driving ice movement in both types of models is approximately

where rho is the density of ice, g is acceleration due to gravity, h is the ice thickness and alpha is the surface slope of the ice sheet averaged over a length of about 10 km (Paterson, 1994). Clearly the ice sheet thickness has a linear effect on the driving stress, but the response of the ice sheet to this driving stress is non-linear, so that in some areas the local ice flux is proportional to the fifth power of ice sheet thickness. Under these conditions even a small error in bed elevation in the models would result in major errors in ice flux and lead to spurious predictions or ice sheet evolution.

The present generation of whole continent models requires a variety of gridded datasets of ice sheet bed / sea bed elevation that cover both the continent and the continental shelf of Antarctica. It seems likely that for most of these models a grid resolution of 5 km will be adequate for the near future.

Fine resolution models require more detailed bed elevation data. It seems unlikely that any single product that could be designed by BEDMAP would be adequate for all these models. What would be of greater value would be for modellers to eventually have access to the original data, in the form of point measurements and digitized profiles. This would allow the generation of bespoke datasets in collaboration with the BEDMAP Consortium

For all models the most crucial areas are close to the grounding line, at the margins where slopes are greatest and ice flow most rapid, in mountain ranges which dam the ice flow from the interior of the continent and on the major outlet glaciers and ice streams through which most of the ice will eventually pass. It is also clear that to prevent logical inconsistencies bed elevation data should satisfy the following requirements that until now have not been achieved:

! Include both ice thickness and sea bed elevation in areas covered by ice shelves
! Extend to the edge of the continent shelf at constant resolution
! Be self consistent regarding the three parameters, surface elevation, bed elevation and ice thickness
! Correctly predict the ice to be at the flotation limit near the mapped grounding line.

5. Status of Ice thickness measurement
Many countries and institutes now have the capability to measure ice sheet thickness, both by radar and seismic sounding and many are actively engaged in collecting new data on a yearly basis. BEDMAP should serve two specific functions, a. to prevent duplication of effort by directing fieldwork towards areas not previously covered, and b. provide a framework for the cross matching of overlapping datasets.

A poll conducted at this workshop showed that as of 1996 only a small part of Antarctica is lacking basic, reconnaissance level ice thickness data. Appendix 10 contains a catalogue of known datasets collected during the workshop.
It was noted that, while there are an increasing number of seismic and radar sounding measurements, these do not represent the full state of knowledge and at the compilation stage BEDMAP should consider all sources of useful data

! Ice sheet thickness
- Airborne radar sounding
- Oversnow radar sounding
- Oversnow seismic sounding
- Rock outcrop data
- Borehole data
- Gravity measurements

! Ice sheet surface elevation
- ERS-1 altimetry
- Airborne and oversnow altimetry
- Map/survey data for the mountainous regions

! Sub-ice shelf seabed
- Oversnow seismic data

! Bathymetry over continental shelf
- New compilations of bathymetry

6. Geographic framework
There now exists a good geographic framework on which to base the BEDMAP project, this is the Antarctic Digital Database (BAS, SPRI & WCMC, 1993), which is available to the entire Antarctic community on CDROM through the SCAR Secretariat, Cambridge.

One of the major problems during the compilation of ice-thickness data will result from navigational inconsistencies. Although most data collected since the mid-1980s have used GPS and so are relatively well fixed with respect to an ellipsoidal reference frame (usually WGS84), data collected before this was not so well fixed. Many methods of navigation were employed, including astronomical fixes, dead reckoning, inertial avionics and Doppler avionics, where each technique has its own sources and propagation routes for error. In addition, many of the ice thickness data were fixed to maps that are now known to have inaccurate control. During the compilation phase a major effort will be needed to evaluate and where possible correct these data.

In addition, to providing a framework for the compilation the ADD map data contains feature types that are more directly relevant to BEDMAP. Areas of rock outcrop define areas of zero ice sheet thickness and so can be included in the compilation of ice thickness. Where map contour lines cross exposed rock, the elevation of these contours should be taken directly as the ice sheet bed elevation.

7. Digital Elevation Models
Before the introduction of satellite positioning systems and satellite surface altimetry the surface of the Antarctic Ice Sheet was mapped using barometric techniques. These techniques are notoriously prone to error and during the compilation of the SPRI Glaciological and Geophysical Folio it was recognised that the surface elevation data contained significant errors. Now that field observations are fixed using satellite positioning systems and satellite altimeters can provide orthometric heights over most of the continent where surface slopes are low, it is possible for us to discard those unreliable surface elevation measurements in favour of a revised digital elevation model.

Several DEMs derived from satellite altimetry are presently available, but it seems likely that once in the compilation phase BEDMAP will need to use the best available DEM derived from several sources

! ERS-1 Satellite altimetry in areas of low slope
! Airborne measurements south of the orbital limit of ERS-1
! Map data in rugged areas
There will be a discussion at the second BEDMAP workshop as to whether any existing DEM is adequate for the compilation or whether a new DEM will have to be commissioned.

8. Project schedule
Discussion during the workshop led to the BEDMAP project being divided into 3 phases:

8.1 Phase - 1 Data collection and storage (10/96-10/97)

8.2 Phase - 2 Data compilation (10/97 - 10/98)

8.3 Phase - 3 Product generation (10/98 onwards)

9. Project management
To implement the plan described in this document the group decided that a BEDMAP Steering Group should be formed comprising, David Vaughan, Janet Thomson, Ian Goodwin and Sergei Popov. The steering group will report regularly to the rest of the BEDMAP Consortium on progress in the project. An offer by the CRC to establish a BEDMAP homepage and contact point on the World Wide Web under the GLOCHANT entry, was warmly accepted by the workshop.
To be a truly valuable exercise, BEDMAP should aim to include as much data as possible. The project will stand or fall depending on the breadth of participants that it attracts and to encourage cooperation, preferential access to the final products will be given to the participating workers. Although local surveys are of little value in terms of broad-scale mapping, taken with other data they can make a significant contribution to the larger picture. BEDMAP will thus encourage participation from any researcher or institute with appropriate data to contribute.

10. Recognised obstacles
It was anticipated that problems may arise for BEDMAP, and some of these were discussed at the workshop.
It was recognised that since new data are being collected on a yearly basis BEDMAP can never be entirely complete. BEDMAP should simply collect all the data on offer and at an appropriate time generate its products. Updated versions of the products could then be generated as new data became available.
It was recognised that some institutes would need to publish their data under their own name before they could be made available to be included in BEDMAP. In order to speed this process and thus make data available as early as possible assistance may be sought from other BEDMAP partners to aid in the publication process.

11. Funding
Since BEDMAP proposes to collect no new data the only requirement for funding would be towards the direct costs of collaboration, data manipulation, publication and dissemination. It is anticipated that no dedicated funding would be required to complete Phase 1 (databasing), but it is likely that some funding will be neccessary to support Phases 2 and 3. At present it is not clear what level of funding would be required to support these activities, or even at which Institute the work will be done. During Phase 1 the steering group will investigate suitable sources. In the meantime an application would be made to EISMINT to fund a second BEDMAP workshop in late-1997. The purpose of this workshop would be to assess the progress made on Phase 1 of BEDMAP and establish the way forward for Phase 2.
A number of funding bodies have, however, already be identified.
! EISMINT
! European Union - Framework V
! US NSF
! European Centre for Earth Observation
! SCAR
! European human mobility fund
In an ideal world, adequate international funding would be available to support BEDMAP, but this is not the case and it is likely that the project will have to proceed by the enthusiasm of the participating researchers. However, there was no objection to individual workers using participation in BEDMAP as the basis for funding proposals. Indeed, BEDMAP will provide letters of support where these proposals are scientifically sound.

12. Final products
There are a variety of possible forms for the final products:
! Hardcopy map to be published by BEDMAP consortium and/or SCAR
! Gridded dataset at various resolutions for modelling
! Internet accessible flight line database
! Publication in quality journal or cornerstone paper by BEDMAP Consortium
! CD-ROM
The exact choice of products has not yet been made but will be reconsidered at the second workshop.

13. Conclusion
The primary outcome of the workshop was consensus that BEDMAP is a worthwhile and feasible goal that should be pursued. All present at the workshop expressed a wish to be involved in BEDMAP at some level, contributing data in one form or another and then participating in the compilation and product generation. A program which includes several concrete goals has been set out for the next year and agreed. A commitment to meet again has been made.

14. Acknowledgements
We would like to thank the SCAR Working Group on Glaciology for their spiritual support in making recommendation Glac-XXIV-2, and the European Ice Sheet Modelling Initiative for their concrete support in funding the first BEDMAP workshop. Phillipa Pirra of the European Science Foundation and Kathy Salisbury of British Antarctic Survey played a large part in organising this workshop.

APPENDICES
APPENDIX 1. The SCAR recommendation
At the XXIV Meeting of the Scientific Committee on Antarctic Research, the Working Group on Glaciology made the following recommendation (Glac-XXIV-2)
Noting:
1. that numerical ice sheet models with realistically proscribed bedrock elevation and other boundary conditions are required to interpret and predict ice sheet changes; and
2. That many National Committees have operated regional, and wider-scale, programmes of measurement of ice sheet thickness and bedrock elevation,
the Working Group recommends that National Committees
recover and collate historic measurements of ice sheet thickness, and contribute these to the coordinated data compilation and bedrock mapping project, BEDMAP, which will update the existing SPRI bedrock data set.

APPENDIX 2. Note of support from Dr David Drewry, co-author of the SPRI Glaciological and Geophysical Folio compilation of Antarctic Ice Sheet Thickness.
Dear David
Thank you for your kind invitation to the BEDMAP meeting. I am sorry to have taken a while to respond - both out of the country and also waiting to see if my diary would allow me to take up your invitation. Alas, this is not to be. I am out of the country on 21st and on 22nd I have an important engagement which has now been confirmed here in Swindon.
I do very much wish you well in the enterprise and still retain an interest in promoting such activity. I would, therefore, be keen to keep in touch with developments in your project as may be possible.
Pass on my encouragement to your group!
Best regards
David

APPENDIX 3. Program of 1st BEDMAP Workshop 21-22/10/96
Carlyle Room, Arundel House Hotel, Cambridge, UK
Monday, October 21^st, 1996
Dr David Vaughan British Antarctic Survey 09:15
Subject: Welcome and overview of the proposal BEDMAP
Dr Ian Goodwin Antarctic CRC 10:00
Subject: The role of BEDMAP within SCAR GLOCHANT
Modelling requirements
Dr Richard Hindmarsh British Antarctic Survey 11:00
Subject: Incorporating real bed elevation data into ice sheet models
Dr Philippe Huybrechts Alfred-Wegener-Institut für 11:25
Subject: A modeller=s specification for an Antarctic bed elevation map
Experiences
Mr Paul Cooper British Antarctic Survey 12:00
Subject: Experiences with compilation of the SPRI folio series
Mrs Janet Thomson British Antarctic Survey 12:25
Subject: Present state of feature mapping in Antarctica & lessons from Antarctic Digital Database
Thickness data holdings -1
Dr Hans Oerter &Astrid Lambrecht Alfred-Wegener-Institut 14:00
Subject Overview of the AWI RES data holdings
Dr Christopher Doake British Antarctic Survey 14:25
Subject: Overview of archival BAS RES data
Hugh Corr British Antarctic Survey 14:50
Subject: Overview of recent BAS RES data holdings
Dr Ian Allison Antarctic CRC 15:15
Subject: Overview of the Australian RES data holdings
Thickness data holdings - 2
Dr. Charlie Bentley University of Wisconsin-Madison 16:00
Subject: Overview of the US RES/Seismic data holdings
Prof. Manfred Lange Westf. Wilhelms-Universität Münster 16:25
Subject: Overview of University of Münster RES data holdings
Sergei Popov Polar Marine Geological Research Expedition (PMGRE) 16:50
Subject: Overview of Russian RES data holdings
Tuesday, October, 22^nd, 1996
Thickness data holdings - 3
Dr Fumihiko Nishio Hokkaido University of Education, Japan 09:00
Subject: Overview of Japanese RES/Seismic data holdings
Prof. I Tabacco University of Milano, Italy 09:25
Subject: Overview of Italian RES/Seismic data holdings
Dr Per Holmlund Stockholm University, Sweden 09:50
Subject: Overview of Swedish RES data holdings
Surface elevation data holdings
Dr Frederique Remy CNRS France 10:35
Subject: Overview of satellite derived Antarctic digital elevation model
Dr Jonathon Bamber University of Bristol, UK 11:00
Subject: Overview of satellite derived Antarctic digital elevation model
Mr Ashley Johnson British Antarctic Survey 11:25
Subject: Recent experiences with Antarctic Digital Magnetic Anomaly Map
Summary & Open discussion
Dr David Vaughan British Antarctic Survey 11:50
Subject: Summary of resources required and a possible route forward
Open discussion

APPENDIX 4. Protocol concerning the use of data within BEDMAP
To foster a sense of cooperation it is important that BEDMAP has a well defined and inclusive data and publication policy that encourages scientific use of the products, but safeguards and offers significant reward to participants.
The following protocols were generally agreed by the participants as providing adequate safeguards for data-collectors, whilst allowing other BEDMAP partners and, eventually, the wider community the opportunity to use BEDMAP data for research purposes. The guidelines adopted would be broadly similar to those adopted by the ADMAP consortium for magnetic anomaly data (ADMAP, 1995).
! At least until the next BEDMAP workshop any data deposited with the BEDMAP data managers would be considered to be for the sole use of BEDMAP and would not be released to any third party for other purposes.
! All gridded datasets and contour maps produced by BEDMAP will be freely available to all bone fide researchers
! Certain restrictions will apply for access to the individual measurements of ice thickness deposited with BEDMAP. For 3 years after collection data are considered to be the sole property of the originators. For the subsequent 3 years the data will be made available for use by other BEDMAP researchers, provided that this transfer is considered as a collaboration between the originator and recipient, and authorship of resulting publications should acknowledge that collaboration.
! For existing datasets the same restrictions will apply but with the date of publication of this report being considered as the beginning of the process.
! A companion paper will be prepared as an accompaniment to the final bed elevation map. This will be published with an authorship reflecting the entire BEDMAP consortium.
! The copyright of the final product of BEDMAP will be vested with SCAR.