Geospatial Information Integration for Science Activity Planning at the Mars Desert Research Station Daniel C. Berrios 1 , Maarten Sierhuis 2 , Richard M. Keller 3 1 University of California, Santa Cruz 2 Research Institute for Advanced Computer Science 3 Intelligent Systems Division, NASA Ames Research Center, Moffett Field, CA USA 94035 dberrios@mail.arc.nasa.gov Abstract. NASA’s Mobile Agents project leads coordinated planetary exploration simulations at the Mars Desert Research Station. Through ScienceOrganizer, a Web-based tool for organizing and providing contextual knowledge for scientific datasets, remote teams of scientists access and annotate datasets, images, documents, and other forms of scientific information, applying pre-defined semantic links or meta-data using a Web browser. We designed and developed an experimental geographic information server that integrates remotely-sensed images of scientific activity areas with information regarding activity plans, actors, and data that had been characterized semantically using ScienceOrganizer. The server automatically obtains remotely-sensed photographs of geographic survey sites at various resolutions and combines these images with scientific survey data to generate “context maps” illustrating the paths of survey actors, and the sequence and types of data collected during simulated surface “extra-vehicular activities.” The remotely located scientific team found the context maps were extremely valuable for achieving and conveying activity plan consensus. 1 Introduction Through the proliferation of high-speed communication networks and wide-spread availability of desktop computing systems, researchers in many different fields should now be able to conduct data gathering and analysis campaigns that involve larger groups of collaborating scientists separated by great distances. However, the design requirements for computer-based systems to support these efforts are not yet fully known. There is some evidence that functions such as synchronous electronic chat and data annotation capabilities (Olson et al., 1998) and geospatial displays for locating and planning scientific data collection (Ogren et al., 2004) can be quite useful. But roles for many other functions remain to be established or elucidated. For example, the relationship between the nature of data collected (e.g., qualitative vs. quantitative), collection and analysis methods employed (e.g., automated vs. non-automated), or the domain of investigation, and the optimal design of systems supporting scientific collaboration is still unclear. In this study, we provide our experience developing and deploying a system for generating geospatial and temporal traces for scientific data through dynamic integration of semantically tagged information. For several years, the Mars Desert Research Station (MDRS) near Hanksville, Utah has been the location of planetary exploration simulation activities conducted by NASA’s Mobile Agents project (Clancey et al., 2001). Part of the research activities conducted during these simulations includes studies of the process of scientific collaboration, activity planning, sharing and review of collected scientific data, and the design and development of computer-based tools to support these processes. We have had the opportunity to participate in these investigations through our work developing and deploying ScienceOrganizer (Keller et al., 2004) during Mobile Agents MDRS fields tests in 2003 through 2005. ScienceOrganizer is a Web tool for managing contextual knowledge (Dey et al., 2001) of scientific datasets (Berrios et al., 2004) specifically developed to support the work of collaborating scientific and engineering teams. Through ScienceOrganizer, remote teams of scientists can access and annotate datasets, images, documents, and other forms of scientific information, supplying additional meta-data and interconnecting them through pre-defined logical relationships using any Web browser. Information stored thusly in ScienceOrganizer is semantically characterized along multiple dimensions, providing users with more precise “tags” with which to find data compared to traditional information storage systems. In