The Third EISMINT Workshop on Model Intercomparison was held in the SunStarHotel in Grindelwald from Thursday September 27 th to Saturday September 29 th 1997.The principal objective of this meeting was to discuss the results of the final phase ofthe model intercomparison venture, where existing ice-flow models were comparedunder real-world situations and under much more challenging conditions than hadbeen the case during previous workshops held in Brussels in 1993 and Bremerhavenin 1994.A first planning meeting to prepare the Grindelwald workshop was organized i nBrussels on 29-30 March 1996. This planning meeting was attended by Drs. RichardHindmarsh, Doug MacAyeal, Catherine Ritz, Vincent Rommelaere, Tony Payne,Mikhail Verbitsky, and Philippe Huybrechts. We discussed which tests we consideredmost useful for the third phase of the venture, what suitable data would be available,and how the work could be practically organised. It was decided to define 5intercomparison topics (Greenland models; Antarctic models; Ice-shelf models;Thermomechanical effects; and Grounding-line treatments), each of which would betaken care of by a coordinator. The respective coordinators (C. Ritz, T. Payne, and P.Huybrechts) first performed the experiments they designed themselves, and thenproduced a written description. All was put on the web in September 1996(http://homepages.vub.ac.be/~phuybrec/eismint.html), and widely advertised, bothdirectly to previous participants, as on the E-mail forum of the InternationalGlaciological Society. These documents form part of this report.Participants could directly load the experimental descriptions and the required datasetsfrom the webpages set up in Brussels, with further links to sites in Grenoble andSouthampton. Results of the tests were sollicited before May 1 st 1997. During thesummer of 1997, the respective topic coordinators collated and plotted with somestudent help all results for presentation in Grindelwald.In total, 13 scientists participated in the various experiments, most of them in morethan one topic. Almost all major ice-sheet and ice-shelf modeling groups wererepresented at the meeting, with a dominance of European groups, but also includingseveral groups and individuals from Japan, the USA and Canada. The workshop wasattended by 21 persons from 8 different countries, including 3 non-European.The workshop itself took three days. The first day was entirely devoted to a plenarypresentation and discussion of the model intercomparison results by topiccoordinators. This was followed during the second morning by group discussions ofmodel results per intercomparison topic, and a plenary presentation of conclusions bydesignated rapporteurs. Highlights of the intercomparison results and the minutes ofthe group discussions form the major part of this report, and summarise the mainachievements of the workshop. It was agreed to write up the main results in threeseparate papers prepared by the respective topic coordinators and co-authored by theparticipants. These papers are to be submitted to the international literature within thenext few months.From the intercomparison results, it turned out that most of the Greenland modelsproduced very similar results. As these models only dealt with grounded ice dynamics,this clearly indicated that the previous intercomparison tests concentrating onnumerics and detecting errors in the various codes had borne fruit (topic 1). Two moreor less complete Antarctic models dealing with the entire ice-sheet/ ice-shelf systemwere presented at the meeting. An interesting difference between these two modelsconcerned the way the grounding zone is treated, which resulted in a differentchronology of, in particular, the history of the last deglaciation in West Antarctica(topic 2). The two ice-shelf models participating in the ice-shelf tests behaved nearlyidentical, thereby confirming previous results from the Ross Ice Shelf benchmark(topic 3). Perhaps most interestingly, all thermomechanically coupled models wereshown to exhibit the radial instabilities in basal temperature and ice flux firstdocumented in papers by Tony Payne, confirming that these instabilities are probablynot a mere product of the particular numerics employed by Tony in his prior work(topic 4). A concensus on how to deal with the marine ice-sheet problem andgrounding-line dynamics, on the other hand, could not be established. All modelsbehaved very differently, raising fundamental questions about the state of equilibriumand stability of the ice-sheet/ ice-shelf junction, and the reversibility of the process ofgrounding-line migration, which subjects deserve to be investigated more thoroughlyin future work (topic 5).The afternoon of the second day was devoted to an open talks session on recentdevelopments in the field of ice-flow modeling. Talks were presented on topics likedealing with small-scale features such as outlet glaciers and ice streams, thegrounding-line problem, numerical techniques, basal boundary conditions, and modelapplications to West Antarctica and the northern hemisphere ice sheets.Taking advantage of the splendidly clear weather on the third day, the 'informaldiscussions' were continued on Jungfraujoch, where we hiked to theMönchsjochhütte. The excellent views on Aletschgletscher and the surroundingmountains were very much appreciated.In all, the general feeling was that the series of intercomparison workshops heldbetween 1993 and 1997 have definitely pushed the art of ice sheet modelling a decisivestep forward. It has enabled groups and individuals worldwide to further develop andimprove their dynamic ice-flow codes. Several European groups are now in possessionof upgraded models that have been thoroughly tested under a wide variety ofboundary conditions. These models can therefore be considered as sufficiently reliabletools to better investigate ice sheets and their interaction with the climate system.