Exploring Geovisualization46 July/August 2006 Published by the IEEE Computer Society 0272-1716/06/$20.00 © 2006 IEEEGeologists primarily explore the Earththrough fieldwork and by analyzing thegeological record at various points on the Earth’s sur-face (see Figure 1a). They then integrate individualdata points about the Earth’s surface by means of moresynoptic analyses. To aid this analysis, they often useimage and topographic data that satellites acquire fromEarth’s orbit. One of the authors, who has decades offield experience in widely varying terrain—from recentvolcanic eruptions1to the Earth’s sea floor2to Antarti-ca’s Dry Valleys3—has demonstrated this strategy’svalue and shown how to augment it using images fromhelicopters or remotely operatedvehicles.In contrast, planetary geoscien-tists commonly work in the reverseorder: Given the time and distancesinvolved, flybys and orbital space-craft acquire the initial data fromindividual moons and planets. Insome cases, more detail might comelater through lander and roverdeployments, or—in the Moon’scase—through human explorers.These opposite approaches result ina huge difference in local analysisof the Earth and other planets.While Earth geoscientists canemploy 3D in situ strategies, mostplanetary geoscientists use static orinteractive 2D visualizations fortheir analyses (see Figure 1b). Although these 2D visu-alizations are highly developed and often effective, fortasks involving subtle 3D spatial judgments, we expectthat interactive 3D visualizations will likely lead to amore complete understanding of the data—and insome cases to recognition of features missed altogeth-er in 2D visualizations.Our goal is to enhance planetary geologists’ ability todo field work as effectively on other planets as they canon Earth. We also want to enhance the exploration ofremote and potentially hazardous Earth environments,such as Antarctica. Guided by multiple questions aboutMars’s geological evolution, as well as by future missionobjectives, we’re developing the Advanced Visualiza-tion in Solar System Exploration and Research system.4As Figure 1c shows, Adviser lets geologists virtually enterthe field by recreating remote sites using laser altime-try, camera, atmospheric, and other data. In develop-ing Adviser, we face the challenges of■ developing algorithms that interactively render a rich3D environment that combines multiple large datasources; ■ building tools that aid geoscientists in their research;and ■ running formal usability studies to evaluate the sys-tem relative to alternative approaches. Here we describe our system and present observationsbased on five case studies of its application. System overviewOur prototype Adviser implementation operates in afour-wall Cave using the model-view-controller designpattern to organize data, visualizations, and interac-tions. The Adviser system contains topography, cameradata, simulated data, and user annotation layers. Wevisualize the model using OpenGL and, in particular,the Real-Time Optimally Adaptive Meshes 2 (ROAM-2)5to render topography and camera data interactively. In terms of performance, an 8,192 × 8,192 heightfieldrenders at 80 stereo frames per second using an NvidiaGeForce 3000G graphics card. The heightfield can reg-ister multiple high-resolution inset camera images,although each inset incurs a frame-rate penalty. Theuser can toggle insets on and off to find a balancebetween information and frame rate. Geologists gener-ally display three inset images, with samples rangingfrom 1,000 × 1,000 to 15,000 × 14,000. The Cave pro-The Adviser prototype systemmakes it possible for planetarygeologists to conduct virtualfield research on remoteenvironments such asAntarctica and Mars. AmongAdviser’s interactive tools aremission-planning andmeasurement tools that letresearchers generate new dataand gain interpretive insights.Five case studies illustrate thesystem’s applications andobserved benefits.Andrew Forsberg, Prabhat, Graff Haley,Andrew Bragdon, Joseph Levy, Caleb I. Fassett,David Shean, and James W. Head III Brown UniversitySarah MilkovichJet Propulsion LaboratoryMark A. DuchaineauLawrence Livermore National LabAdviser: ImmersiveField Work forPlanetaryGeoscientistsduces an immersive effect that lets geologists feel likethey’re at the site; they achieve interaction—such as 3Dnavigation and point specification—using a trackedhandheld wand-like device. They also can use a TabletPC for 2D line and gestural input.Typically, geologists create a session in the laboratory,using their target data elements, then access the sessionin the Cave. They can also create new data within theCave and store it in a networked database that they cansubsequently access from anywhere using a Web portal.IEEE Computer Graphics and Applications 471 Different types of planetary geoscience analysis. a) Glacial geologist Dave Marchant analyzes sublimation till fabrics in the AntarcticDry Valleys. This local data will be integrated with regional air photos and satellite images to provide a broad view of the Mars-likeAntarctic glacial processes. b) Studying 2D maps of Martian topography and camera data. c) Virtually studying topography on Marswith Adviser.Related WorkMany researchers are currently working ongeovisualization systems. Following are the systems most closely related to Adviser. Desktop systemsNASA’s World Wind system (see http://worldwind.arc.nasa.gov) lets users interactively explore a wide variety ofterrain and image data. Google Earth (see http://earth.google.com) lets users look at high-resolution images,annotated data, and 3D models of some structures. TheosgPlanet viewer (see http://www.ossim.org/tiki-read_article.php?articleId=3) can visualize large geospatial datasets and import GIS data via the Web Mapping Serverstandard (see http://portal.opengeospatial.org/files/?artifact_id=5316).Immersive environmentsImmersive systems have been developed for a range ofspecific tasks. Wright and colleagues developed a system forcreating and visualizing 3D terrain for the Mars Pathfindermissions.1Powell and colleagues developed a data fusionand visualization system for rover data acquired duringMars exploration.2Although both systems have proven useful forinvestigating topographical features within a small area,we’re interested in planetary-scale data visualization andexploration. In that respect, research focused on exploringlarge data sets 3,4is