This paper might be a pre-copy-editing or a post-print ...senseable.mit.edu/papers/pdf/20060701_Mitasova_etal_RealtimeLan… · 01.07.2006 · with 6 DOF for navigation and immersive

This paper might be a pre-copy-editing or a post-print author-produced .pdf of an article accepted for publication. For the

definitive publisher-authenticated version, please refer directly to publishing house’s archive system.

Published by the IEEE Computer Society 0272-1716/06/$20.00 © 2006 IEEE IEEE Computer Graphics and Applications 55

Exploring Geovisualization

L and surface topography forms the bound-ary between the solid earth, atmosphere,

and hydrosphere. It controls air, water, and mass fluxes;redistributes solar irradiation; and, consequently, has amajor influence on physical processes, ecosystems, andhuman activities. Understanding topography—its char-acter, properties, and human interaction—is thereforefundamental for many disciplines and provides the dri-ving force for research and development of terrainanalysis and visualization tools.

For centuries, people have used 3D solid terrain mod-els—which for most are intuitive and easy to interpret—for military planning and landscape design. Newtechnologies, such as solid terrain modeling (seehttp://www.stm-usa.com) and 3D printers make modelproduction efficient and can help generate impressive3D physical models for settings where readily inter-pretable landscape representation is essential. Variousinstitutions have used static solid models to convey 3Dlandscape information to the general public. Types ofinstitutions using this approach include

■ museums—for example, the National GeographicMuseum in Washington, D.C.;

■ parks—for example, Grand Teton National Park; and■ state and local governments—for example, Sioux

Falls (see http://www.esri.com/news/arcnews/winter0203articles/inthecityof-siouxfalls.html), SanFrancisco, and Illinois (see http://www.sws.uiuc.edu/chief/gis/illinois3d.htm).

Military and firefighting groups have focused on train-ing and planning applications using solid models. Solidmodels are frequently employed for collaboration anddecision support, although the range of displayed dataand type of interaction is limited to viewing a single ter-rain representation.

By contrast, virtual terrain modeling environments arehighly interactive, allowing users to combine elevationdata with additional map layers, select a viewing posi-tion, adjust exaggeration, use selective illumination, flythrough the model, and modify other surface properties.The basic tools use a GUI and mouse to control the view,while the more sophisticated systems provide controllerswith 6 DOF for navigation and immersive VR.1

The virtual terrain modeling tools are particularlyeffective in combination with geographic informationsystems (GIS) and numerical mod-els of landscape processes.2 Howev-er, this type of visualization is aimedat a single user and the immersiveenvironments can lead to a feeling ofseparation from the real world andnausea on the part of the user.1 Foreffective group discussion, collabo-ration, and collective decision mak-ing, virtual terrain representationsrequire the development of appro-priate interaction tools and newmultiuser environments,3 such asthe combination of computer-vision-based hand and object tracking withaugmented and virtual reality.1

We introduce a concept thatbuilds upon previous independent tangible user inter-face (TUI) and terrain analysis research (see the “Relat-ed Work” sidebar on the next page) and aims at moreintuitive collaborative interaction with digital landscapedata. In addition to visualization of existing data, ourwork’s specific focus is to use terrain surface modifica-tions to explore relationships that occur between differ-ent terrains, the physical parameters of terrains, and thelandscape processes that occur in these terrains. Here,we describe various approaches to human interaction

Emerging technologies thatcombine the flexibility of digitallandscape representation witheasy-to-interpret 3D physicalmodels open new possibilitiesfor user interaction withgeospatial data. A prototypetangible geospatial modelingenvironment lets users interactwith landscape analysis andsimulations using a tangiblephysical model.

Helena Mitasova and Lubos MitasNorth Carolina State University

Carlo Ratti, Hiroshi Ishii, and Jason AlonsoMassachusetts Institute of Technology

Russell S. HarmonUS Army Research Laboratory

Real-Time LandscapeModel InteractionUsing a TangibleGeospatial ModelingEnvironment


























































































































































with terrain data, propose a tangible system configura-tion, and discuss coupling our system with the Geo-graphic Resource Analysis Support System (Grass) GIS.4

Interaction with terrain data in Grass GIS

Capabilities for visualizing topography in GIS haveevolved from 2D raster maps, contours, and simple 3Dmeshes in the 1980s to interactive 3D rendered surfaceswith shading and color maps draped over the surfacethat became standard in the 1990s. Grass GIS wasamong the first general GIS environments that offeredsurface and volume visualization in 3D space controlledby an easy-to-use GUI (see http://skagit.meas.ncsu.edu/~helena/gmslab/appdoc/sg3d_tut.html). Thesystem’s capabilities included dynamic 3D surfaces2

used to represent dynamic phenomena based on mea-sured data, such as a pollutant plume’s temporal evolu-tion or to communicate landscape process simulationresults—for example, water and pollutant flow overcomplex terrain (see more at http://skagit.meas.ncsu.edu/~helena/publwork/grasskey02/grass02talk10.html). Support for multiple surfaces and interactive cut-ting planes later added the ability to explore terrainchange using a set of digital elevation models (DEMs)

representing elevation surfaces at different time peri-ods (for example, before and after construction4), butelevation surface has been considered a static entity formost GIS applications.

New mapping technologies, especially laser scanning(for example, lidar and ladar), have dramaticallyimproved terrain-mapping efficiency and high-resolu-tion, multitemporal DEMs are now increasingly avail-able. These data provide a new and more dynamic viewof landscapes and stimulate the need for viewing, ana-lyzing, and modeling terrain change and its impactswithin GIS. Interacting with terrain data has becomeimportant for different types of spatial analysis anddesign and planning tasks. It can provide insights intothe relation between terrain surface shape changes andspatial patterns of the terrain’s derived parameters. Forexample, access to this type of data can help address thefollowing questions:

■ What is the pattern of slope around a peak or a saddlepoint, on a ridge, or in a valley, and how will this pat-tern change if the peak is removed or the valley filledduring a surface mining operation?

■ What is the relation between the terrain shape andwater flow patterns, which shape is associated with


56 July/August 2006

Related WorkRecently, several new technologies that combine the

flexibility of digital representation with intuitive 3D physical models have emerged.

In one system, solid physical models combine high-resolution imagery printed on the model surface with aprojection of selected map layers retrieved from geographicinformation systems (GIS).1 The system is static, but theuser can display various types of GIS data, such as thelocation of a floodplain, wetlands, or a planned highway,over the terrain surface and the printed imagery.

A pin-based system creates the terrain model usingmovable rods covered by a flexible latex sheet. The pins liftthe latex sheet to form a 3D model of the landscape basedon digital elevation data while an image file is projected onthe surface. Through computer control, a new surface canbe generated within 30 seconds to 2 minutes. Recently,researchers have combined this type of physical terrainmodel with a touch table for better interaction (see http://www.xenotran.com/products.htm and http://www.ms.northropgrumman.com/images/TerrainTable_FS.pdf).

Illuminated Clay consists of a flexible clay terrain model, alaser scanner, and a projector.2 The laser scans the surfacethat is amenable to physical modification by hand, and theimpact of the modification on a selected terrain parameter(for example, slope and water flow direction) is thenprojected as a color map on the surface in near real time. TheMIT Media Laboratory developed the prototype systems.

The first two technologies, with a focus on data display,have recently become commercially available. The thirdtangible approach, which offers the opportunity for asuperior combination of flexibility and cost efficiency, is atthe research and development stage. This tangible user

interface (TUI) integrates terrain representation and controlwithin a physical tangible model coupled with its virtualdigital representation.

TUIs have emerged as an alternative paradigm to themore conventional GUIs3 letting users manipulate objects inspace, thereby combining the benefits of physical anddigital models in the same representation. The earlydevelopments of TUIs date back to the early 1980s, whenFrazer explored different approaches to parallel physical anddigital interactions with his 3D data input devices.4 TheIlluminated Clay concept was directly inspired by the UrbanDesign Workbench,5 which uses digitally augmentedtagged physical objects to represent buildings that can berearranged to facilitate the process of urban design. Asimilar system has been coupled with a GIS.6

References1. C. Coucelo, P. Duarte, and R. Crespo, “Combining 3D Solid Maps

with GIS Data Video Projection,” Proc. 2nd Int’l Conf. GeographicInformation GIS Planet, CD-ROM, 2005; http://www.gison3dmap.com/index.asp.

2. C. Ratti et al., “Tangible User Interfaces (TUIs): A Novel Paradigmfor GIS,” Trans. GIS, vol. 8, no. 4, 2004, pp. 407-421.

3. B. Ullmer and H. Ishii, “Emerging Frameworks for Tangible UserInterfaces,” IBM Systems J., vol. 393, no. 3, 2000, pp. 915-931.

4. J. Frazer, An Evolutionary Architecture, Architectural Assoc., 1995.5. J. Underkoffler and H. Ishii, “Urp: A Luminous-Tangible Work-

bench for Urban Planning and Design,” Proc. Conf. Human Factorsin Computing Systems (CHI), ACM Press, 1999, pp. 386-393.

6. V. Coors, V. Jung, and U. Jasnoch, “Using the Virtual Table as anInteraction Platform for Collaborative Urban Planning,” Comput-ers & Graphics, vol. 23, no. 4, 1999, pp. 487-496.

dispersal flow, and what terrain geometry leads toflow convergence?

■ How will our example mining operation change thepattern of water flow and sediment transport?

■ What happens to the flow distribution pattern if ter-races are introduced on a uniform hill slope?

Interactive manipulation of terrain models can helpteach reading of a topographic map or a cut-and-fill planrepresented by contours. Educators can simultaneous-ly project contours over the 3D model and as a 2D mapto demonstrate how topographic contours change whenthe surface is modified. Easily adding structure andbuilding models to the DEM (essentially combining CADand GIS) is important for exploring design alternativeswhen creating neighborhoods, preparing plans forstorm water and sediment control, placing structuresfor best passive solar energy performance, and numer-ous other creative design tasks requiring human inputand analysis of alternatives.

Currently, we can modify DEMs in Grass GIS byonscreen digitizing, using a mouse and GUI. Thisapproach involves making 3D changes using an imagedisplayed on a 2D screen—for example, by outlining thearea to be modified and assigning it a selected elevationvalue. When using a mouse and GUI in such a way, thereis a disconnect between the manipulated object (repre-sented by an image on the screen) and the tool to do it(a mouse) that requires good hand–eye coordinationand 3D spatial perception. Therefore, we have beeninvestigating more intuitive approaches. A variety ofhaptic devices, such as a haptic desktop arm, allow theuser to touch and modify a virtual model, but the handof the person making the changes is still separated fromthe object, although the haptic arm lets the user feel it (see http://www.computersculpture.com/Pages/Index_Modeling.html). The haptic arm approach is cur-rently used for object modification (such as in industri-al design); the practicality of applying this approach ingeosciences remains to be explored.

To see the impact of surface modifications on terrainparameters, the user must run several additional mod-ules to perform the analysis. To enhance the GIS capabil-ities for interaction with terrain data, we are investigatingcoupling between Grass GIS and the illuminated clay5

concept.

Tangible system configuration The Illuminated Clay system uses a commercially

available laser scanner (a Minolta Vivid-900 or 910) tocapture the surface geometry of a 3D topographic modelmade of flexible clay. The 3D scanner is coupled to avideo projector that projects images onto the landscapemodel. The scanner and projector pair is housed insidea casing approximately 2 meters above the terrainmodel. A special program based on Minolta’s softwaredevelopment kit performs successive scans of the modelat approximately three scans per second—instead of thestandard scan-and-stop mode of operation. The pro-gram converts the results obtained as distance valuesfrom the scanner’s lens into a raster DEM, commonlyused in the geosciences. It then applies the terrain analy-

sis algorithms5 to the DEM and the resulting raster mapsare then projected on the landscape model using a videoprojector. The interaction loop is recursive, performingapproximately one scan-analysis-projection cycle each1 to 2 seconds, providing users with an almost instantresponse to any change manually introduced on themodel surface.

The entire set of analysis functions (for example, ele-vation, flow direction, and profile) occurs simultane-ously and all parameters are displayed as small 2D rastermaps around the physical model edges (see Figure 1)while the user-selected parameter (such as land surfaceslope) is projected over the physical model. The user canthen use any of the 2D raster maps to select an image toproject over the physical model. We can add additionalprojectors to create a virtual 3D terrain model that iscoupled with the tangible physical model and the sys-tem interface. The virtual terrain model lets the userview the modified surface and objects placed on it froma near-ground perspective or any other user-selectedviewing position.

We built the physical landscape model from variousflexible materials, such as plasticine with a metal mesh,and wooden or plastic blocks for structures. We can formthe plasticine and mesh surface using a mold createdfrom real-world digital elevation data and a 3D printer,solid terrain modeling technology, or simple foam lay-ers cut along contours. Alternatively, we could createthe surface interactively by projecting a differencebetween the DEM and the scanned model onto the sur-face and modifying it until the difference is below agiven threshold. The possible interaction with the modelincludes manual shaping of the clay surface and addingor removing model structures.

Coupling with GISWe can consider the original illuminated clay system

a TUI for terrain surface analysis because it supportscomputation of those landscape parameters that useonly elevation data as an input. To expand the system’suse beyond the topography by including additionallandscape features and processes, we have defined sev-eral levels of coupling between the 3D physical model

IEEE Computer Graphics and Applications 57

1 (a) Illuminating clay system configuration combining a ceiling-mounted3D laser scanner with a projector and a flexible clay landscape model. (b)An additional projector projects a coupled virtual terrain model on thevertical screen.

ProjectorScanner

(a) (b)

and Grass GIS. The coupling options described in the“Different Levels of Possible Coupling between GrassGIS and a 3D Physical Landscape Model” sidebar pro-vide a framework for building scalable tangible mod-eling systems, from simple, low-cost options tosophisticated geospatial modeling environments withnear-real-time response to different types of landscapemodifications.

Grass GIS,4 as one of the largest open source projectsby the size of its code, includes more than 350 modules

that support georeferenced data management, projec-tion, spatial analysis, image processing, modeling, andvisualization. Among these modules, some 20 aredirectly applicable for analysis of impacts of terrainmodification in a two-way coupling mode (see the “Dif-ferent Levels of Possible Coupling between Grass GISand a 3D Physical Landscape Model” sidebar’s secondand third options). These require only elevation dataas an input and allow us to perform a comprehensivetopographic analysis as discussed in the “Topographic


58 July/August 2006

Different Levels of Possible Coupling betweenGrass GIS and a 3D Physical Landscape Model

Based on the type of interaction between the physicalmodel surface representing elevation, its color representinglandscape attributes, and a computer system, we can definedifferent levels of coupling.

Passive physical model combined with activeimage projection

This level transfers a landscape attribute map from thecomputer to the physical model. This one-way couplinginvolves a projection of a thematic map created in the GrassGIS1 graphical window over the physical model (see Figure A),an adequate approach for small models (around 1 foot2).Image transformation to the central projection is required forlarger models and greater terrain complexity.

A One-way coupling projects a GIS map onto the solid terrainmodel.

The projected map can be static, incorporating acombination of raster and vector features (for example,combined vegetation, roads, building footprints, andstreams) or an animation representing a landscape process(for example, water flow or spread of fire). This approachoffers great flexibility for the type of displayed data andallows the user to explore the relationship between the datarepresented by a projected map and the topography in anintuitive, natural manner. For example, by projecting avegetation map, you can easily identify the associationbetween the pattern of bottom-land forest distribution andriver valleys. This type of coupling is useful for education,displays for the general public (for example, museums andparks), and other applications where solid models are usedwith GIS map layers to broaden the range of themes thatusers can explore.

Active physical model combined with activeimage projection

This level transfers elevation data from the physical modelto the computer and an attribute map from the computer tothe model (see Figure B). The model surface can be modifiedmanually and the new surface elevations transferred to thecomputer using 3D laser scanning. The elevation data areimported into the Grass GIS where digital elevation models(DEMs) and parameters computation takes place at a user-selected resolution. The analysis can take several seconds;therefore, the attribute data are projected from thecomputer to the surface with a delay (several seconds ormore, depending on the resolution and complexity ofanalysis) so the user has to wait for the result. Sophisticatedspatial approximation methods,1,2 used also for lidar data,can be applied within Grass GIS to reduce noise associatedwith the laser scanning and manual manipulation of the claysurface.

B Basic two-way coupling transfers the data from physicalterrain model to GIS and projects GIS maps onto the model.

The spatial approximation module simultaneouslycomputes a set of raster maps representing topographicparameters that can be projected on the model surface andthe small peripheral 2D maps. This two-way asynchronouscoupling permits complex spatial modeling that requiresadditional map layers stored in the GIS database and GISmodeling tools. For example, erosion modeling combinesthe land cover, soil, and rainfall data stored in the GIS withterrain parameters derived from the scanned model. Theresulting erosion-risk raster map can then be projected overthe terrain model and subsequently used to guide the designof specific soil conservation practices.

The asynchronous coupling also lets the user derive vectorlayers from the modified surface, such as new topographiccontours and stream networks as well as perform analysis,modeling, or optimization tasks that might use attributesstored in an external database. This option can also support

Color: attributesProjector

Surface: elevation

Computer:GIS


Surface: elevation3D scanner

Computer:GIS/IC

Analysis Coupled with a Tangible Landscape Model”sidebar on the next page.

Several modules combine elevation data with addi-tional map layers for more complex analysis that takesfull advantage of GIS. Soil erosion risk assessment, firespread simulation, ecosystem modeling, and analysis ofshortest path are examples of applications where thescanned terrain surface is combined with raster mapsrepresenting vegetation, soil types, climate conditions,infrastructure, or other landscape properties to com-

pute an updated image projected over the modified sur-face. Using the tangible interface, we can introduce ter-rain modifications at various locations and investigatethe sensitivity of the modeled phenomenon to chang-ing topography. For example, in the study of erosionrisk, we can compare the impact of various types of ter-rain change in vegetated areas with impacts in areaswith bare soil.

Dynamic models of landscape processes, such aswater flow and sediment transport, require special


complex numerical models representing dynamicphenomena.

Active physical model combined with activenear-real-time image projection

This option is similar to the first option, but the transfer ofelevation data from the physical model to the computer andthe projection of an attribute map onto the model happen innear real time. Users can manually modify the model surfacewhile its elevation is being transferred to the computer and anew attribute map is projected onto the surface within 1 to 2seconds, creating a user perception of instant response. Thistype of coupling was developed for the Illuminated Clayconcept and the fast response is achieved by combining thelower-resolution fast-scanning mode with simpleapproximation methods for computation of DEM and itsparameters.

Various options for two-way synchronous coupling withselected Grass GIS modules are currently being investigated.The synchronous coupling is the most unique aspect of theproposed system, as it permits visualization of the impact ofterrain modifications while they are performed so the usercan continue to make adjustments until the desired outcomeis achieved.

Active, computer controlled landscape modelwith active near-real-time image projection

This option transfers elevation data from the physicalmodel to the computer and from the computer to themodel, with an attribute map transferred from the computerto the model (see Figure C). This two-way surface couplingcould support the capabilities described in the secondoption, but also would make the surface adjustable using thedigital elevation data, a capability currently supported onlyby the pin-based systems.

C Full two-way coupling transfers both elevation and attributedata between the physical model and GIS.

Replacement of the clay surface by a specialized type ofhardware, or coupling with a modified pin-based systemwould be required to realize this capability—currently anexpensive and not yet practical option because the pin-based systems don’t allow interactive modification of thetopographic surface by hand.

This option would expand the system’s use letting usersmodify the physical terrain model based on the results ofnumerical simulations of landscape evolution due to erosionand deposition processes. It would also allow users to modifythe physical model using digital data, for example byincorporating CAD data representing a planneddevelopment that could be further explored and modifiedusing the tangible approach.

Active, computer-controlled physical model withactive near-real-time attribute input andprojection

This level would permit elevation transfer from the modelto the computer and/or from the computer to the model, aswell as accommodate the transfer of an attribute map fromthe computer to the model and from the model to thecomputer (see Figure C). This approach would support allthe capabilities described in the previous option, but alsowould use the landscape model’s RGB scan, acquired by thelaser scanner simultaneously with the elevation surface, tocapture manually introduced change in color. Thus, it wouldpermit color coding of the surface and structures to indicatethe type of land cover, permeability of structures, and otherproperties needed as input for landscape process modelingand decision making.

Considerable research still needs to be done to build thistype of system. In addition to the development of anaffordable computer-controlled physical model, suitablematerials and tools for adding the color to the physicalmodel will have to be explored and methods for fast analysisof the color image and its transformation to surfaceattributes need to be investigated.

References1. M. Neteler and H. Mitasova, Open Source GIS: A Grass GIS Approach,

2nd ed., Kluwer/Springer, 2004.2. H. Mitasova, L. Mitas, and R.S. Harmon, “Simultaneous Spline

Interpolation and Topographic Analysis for Lidar Elevation Data:Methods for Open Source GIS,” IEEE Geoscience and Remote Sens-ing Letters, vol. 2, no. 4, 2005, pp. 375- 379.


Surface: elevationShaper

3D scannerComputer:GIS/IC

attention. In this case, the evolution of the simulatedprocess changes as the user interacts with the surface.Researchers are investigating various approaches forthis type of dynamic real-time coupling, including

■ displaying the evolution of flow during computation; ■ displaying the resulting flow with increasing accura-

cy by gradually sharpening the image using anincreasing number of particles for simulation of flowby a path sampling technique;2

■ updating the flow only for the modified location andthe area downstream from it; and

■ performing the simulation using parallel computing.

We consider coupling with dynamic models one ofthe most interesting and at the same time the most chal-lenging directions in the development of a tangible GISinterface because it might allow us to add a temporalaspect to the current spatial applications. For example,we can observe water running over a dam and try to addmodel sand bags to stop it while water levels are still ris-ing. We can investigate how fast this needs to occur andwith how many bags are needed. Moreover, differentusers can modify terrain at different locations simulta-neously—for example, one user can break the damupstream from a city and several others can try to placeprotection such as models of sand bags or retentionwalls at critical locations downstream, providing anopportunity to explore complex spatial and temporalinteractions. Because users do not need obtrusive equip-ment such as glasses or complex navigation devices,they can communicate face to face in a natural way,without needing special skills such as the knowledge ofoften complicated GIS interfaces.

We will need to conduct specific usability studies toassess the effectiveness of the system for various typesof tasks; however, the system is still evolving in terms ofvarious configurations, control of attribute maps, andinteraction with GIS, so it might be too early for suchstudies.


60 July/August 2006

0.00100.25000.50000.75001.0000

(a) (b)

kg/ms

3 Significant sediment transport is observed and modeled at the test site:(a) field photo after a storm and (b) simulated pattern of sediment flowrate.

2 Elevation surface for the test site: (a) 1:2,000-scale clay model; (b) 1-millimeter-resolution DEM based on a 3D scan of a modified physicalmodel; (c) original 2-meter-resolution DEM used to create the physicalmodel; (d) 2-meter-resolution DEM based on a 2001 airborne lidar scan.The surfaces were visualized in Grass GIS.

1993

2001

0 300 meters(a)

(b)

(c)

(d)

N

Topographic Analysis Coupled with a TangibleLandscape Model

Grass GIS-supported topographic analysis includescomputation of maps and reports representing propertiesof terrain surface and their relation to natural processes.These properties include

■ elevation (contours and color elevation map);■ terrain surface geometry (slope, aspect, curvatures, and so

on);■ terrain features (peaks, valleys, ridges, dune crests and slip

faces, and so on);■ watershed characteristics (watershed boundaries and

stream networks); ■ view shed analysis, shadowing, and line of sight;■ solar irradiation pattern and dynamics; ■ profiles and routes;

■ summary statistics, volumes, areas, and histograms; and■ spatial query (values and attributes at given locations) and

measured distances.

The Illuminated Clay model approach has implementedcomputation of elevation, contours, slope, aspect,curvature, flow direction, profiles, and shade casting; weare enhancing these computations using methodsdeveloped for Grass GIS. Algorithms derived from flowaccumulation, such as watershed boundaries and streamnetworks, are more difficult to implement within theframework of two-way synchronous coupling because thecomputation requires accumulation of values over theentire modeled region. An approach that shows the resultat gradually increasing resolution might hide the delayfrom the user; currently, two-way asynchronous coupling isused. Visibility and solar irradiation might require a similarapproach.

(Cou

rtes

y Ri

char

d M

cLau

ghlin

, NC

SU)

An example applicationTo illustrate the concept of the tangible geospatial mod-

eling environment, we have coupled the Illuminated Clayactive landscape model with Grass GIS and used this toexplore various structure placement and terrain modifi-cation scenarios for sediment control and prevention oflocal flooding in a small watershed on the North Caroli-na State University experimental farm. The test area is a25-hectare site at the NCSU Sediment and Erosion Con-trol Research and Education Facility. Various methods forerosion and sediment control are developed and testedat this site and it has a sufficiently complex terrain todemonstrate the tangible interface’s functionality.

We built the 3D clay model (see Figure 2a) for ourtest landscape using the 0.6-meter (2-feet) elevationcontours obtained from the Wake County, North Car-olina GIS. The model’s horizontal scale was 1:2000with six-times vertical exaggeration. Figure 2b showsa 1-millimeter-resolution DEM based on the scannedmodified clay model. For comparison, we have alsocomputed a 2-meter-resolution DEM from the originalcontours, representing the terrain in the year 1993 (seeFigure 2c) and an airborne lidar-based DEM using thedata acquired in 2001 (see Figure 2d) to illustrate actu-al changes in topography. Field observations (see Fig-ure 3a) and hydrologic and erosion simulations6

performed in Grass GIS show significant runoff and sed-iment transport through the study site’s center (see Fig-ure 3b) and flooding in the lower parts of terrain wherea new road is located (see Figure 2d). We have used thephysical model as a tangible interface (see Figure 4a) toexplore the changes in topography using various struc-tures, ditches, and basins that could be implementedto minimize the flooding and reduce the excessive sed-iment transport.

We first projected GIS-based landscape characteriza-tion layers—such as soils, land cover, footprints of struc-tures, and roads—onto the physical model surface to

provide information about the modeled landscape’sproperties. We performed surface analysis, flow rout-ing, and erosion modeling in Grass GIS for the initial ter-rain to assess the patterns of slope, overland water flow(see Figure 4b), and sediment transport. We then man-ually modified the physical landscape model to explorevarious approaches for water flow and sediment pollu-tion control (see Figure 5).


4 Clay model placed in the Illuminated Clay environment: (a) flow direction is projected over the surface while additional parameters (elevation, contours, slope, aspect, and profile) are displayed by smaller 2D images around the worktable; (b) more detailed maps of water flowand slopes are computed for the scanned surface using asynchronouscoupling with Grass GIS.

(a)

15

0

5

10

(b)

Degrees

(b)

5 Users modify the physical landscape model by (a) creating depressions or (b) adding a check dam. Simultaneously, the impact of themodification on the spatial pattern of flow direction and slope value (represented by color) is projected over the model surface, whilechanges in other parameters appear in the smaller 2D images around the worktable.

(a)

The first set of changes involved manual modificationof a convex hill slope into a shallow basin while the flow-direction map was displayed and concurrently updatedalong with the terrain surface change. We graduallyadjusted the basin shape until flow was redirectedthrough this basin into the neighboring forested water-shed, thus reducing the concentrated flow that wascausing the flooding and sediment pollution. We thenprojected a new slope map over the modified surface tocheck whether the modifications might have introducedunwanted steep slopes that could cause stability prob-lems over the long term. We manually modified slopesthat exceeded the critical threshold, such as the redareas in Figure 4b, until the projected slope map showedstable values everywhere. We implemented a similarredirection of flow and slope adjustment on the otherside of the valley. We then imported the new DEM intoGrass GIS and re-ran the topographic analysis and flowsimulation with greater detail. The process-based modelallowed us to quantify the proposed change’s impact interms of water depth and sediment flow rates for aselected design storm.6

A second set of modifications involved the additionof check dams to restrict flow in the main valley. We usedsmall, dam-shaped pieces of clay to approximate thesestructures (see Figure 5). The tangible environment letsus move a dam up and down along the valley, add a sec-ond and a third structure, and modify the dam’s shapeand size while observing the impact on water flow pat-terns in near real time.

The flow direction map showed interruption of flowdue to check dams and a reduction in water flow ratethrough the valley, with the effectiveness dependent onthe structure’s location in relation to the rates of lateralinflow. We then imported the modified DEM containingthe selected distribution of dams into Grass GIS and per-formed process-based modeling to quantify the resultingwater flow depth and sediment flow rates (see Figure 6).

The placement of larger structures in the valley’s lowersection resulted in a more robust design in terms of a capa-bility to reduce flow rates and to withstand larger storms

than the use of a larger number of smaller structures locat-ed along the entire valley’s length. We have used both two-way synchronous and asynchronous coupling for this case.In addition, this approach let us investigate the flowdynamics and assess the conditions under which the checkdams could be overtopped by storm water.

ConclusionThe presented coupling of a tangible physical model

and Grass GIS represents a promising first step in aneffort to build tangible geospatial modeling environ-ments that will allow users to interact with 3D land-scape data using the natural human ability to work withtheir hands. We need to further research several areasto better understand the usability and potential forapplications. In particular, we need to investigate therange of scales that manual manipulation of a physicallandscape model can support and to assess the feasibil-ity of adding zoom-in capabilities at least for the cou-pling with the computer-controlled physical model. Arelated issue is the merging of the model data with thereal-world data and a tradeoff between real-timeresponse and accuracy.

In the future, we envision that large 3D physicalmodels coupled with GIS and numerical simulationsand capable of receiving real-time data from satellitesand terrestrial sensors will form the heart of the land-use management technology at different levels of government and within the military installation man-agement community. These kinds of systems will notonly help us solve day-to-day land management prob-lems, but also improve response to natural disastersand emergencies. ■

AcknowledgmentsWe gratefully acknowledge the US Army Research

Office Terrestrial Sciences Program’s support of thisresearch. We obtained the data used for the model fromthe Wake County, North Carolina GIS and the NorthCarolina flood mapping program. The site photo is cour-tesy of Richard McLaughlin, NCSU. We thank Ondrej


62 July/August 2006

0.0010.0050.010.050.10.5

Meters

N

0 200 meters(a) (b)

6 Water-depth distribution for (a) single and (b) two check dams. The output is from a numerical model implemented in Grass GIS. Related animation showing water flow evolution can be viewed at http://skagit.meas.ncsu.edu/~helena/wrriwork/tangis/tangis.html.

Mitas for his assistance with the physical model. We alsothank all of the tangible-modeling enthusiasts at theMassachusetts Institute of Technology who contributedin one way or another to this research, in particular EranBen-Joseph, Assaf Biderman, Dennis Frenchman, BillMitchell, Ben Piper, Cecilia Yu, and Yao Wang.

References 1. N.R. Hedley et al., “Explorations in the Use of Augmented

Reality for Geographic Visualization,” Presence, vol. 11, no.2, 2002, pp. 119-133.

2. L. Mitas, W.M. Brown, and H. Mitasova, “Role of Dynam-ic Cartography in Simulations of Landscape ProcessesBased on Multi-Variate Fields,” Computers and Geosciences,vol. 23, no. 4, 1997, pp. 437-446.

3. A.M. MacEachren and I. Brewer, “Developing a Conceptu-al Framework for Visually-Enabled Geocollaboration,” Int’lJ. Geographical Information Science, vol. 18, no. 1, 2004,pp. 1-34.

4. M. Neteler and H. Mitasova, Open Source GIS: A Grass GISApproach, 2nd ed., Kluwer/Springer, 2004.

5. C. Ratti et al., “Tangible User Interfaces (TUIs): A Novel Par-adigm for GIS,” Trans. GIS, vol. 8, no. 4, 2004, pp. 407-421.

6. L. Mitas and H. Mitasova, “Distributed Erosion Modelingfor Effective Erosion Prevention,” Water Resources Research,vol. 34, no. 3, 1998, pp. 505-516.

Helena Mitasova is a researcher inthe Department of Marine, Earth, andAtmospheric Sciences, North Caroli-na State University. Her researchinterests include GIS, environmentalmodeling, and geovisualization.Mitasova has a PhD in geodesy-car-tography from Slovak Technical Uni-

versity, Bratislava. She received the Excellence inDevelopment award from the Open-GIS Foundation and isa member of the Transactions in GIS editorial board. Con-tact her at [email protected].

Lubos Mitas is a professor in theDepartment of Physics, NCSU. Hisresearch interests include computa-tional methods for quantum systemsand methods for spatial interpolation,and process-based simulation of land-scape fluxes. Mitas has a PhD inphysics from the Institute of Physics,

Bratislava, Slovakia. Contact him at [email protected].

Carlo Ratti directs the SENSEableCity Laboratory at the MassachusettsInstitute of Technology. His researchinterests focus on how technology ischanging the way we live and work incities. Ratti graduated in civil struc-tural engineering from the Politecni-co di Torino, Turin, Italy and the Ecole

Nationale des Ponts et Chaussées, Paris. He also has a PhDin architecture from the University of Cambridge. Contacthim at [email protected].

Hiroshi Ishii is a tenured associateprofessor of media arts and sciences atthe MIT Media Lab where he foundedand directs the Tangible Media Group.His research interests include thedesign of seamless interfaces betweenhumans, digital information, and thephysical environment. Ishii has a PhD

in computer engineering from Hokkaido University,Japan. Contact him at [email protected].

Jason Alonso is a technical assis-tant at the MIT Media Lab, TangibleMedia Group. His research interestsinclude human–computer interac-tion. Alonso has a BS in computer sci-ence and computer engineering fromMassachusetts Institute of Technolo-gy. Contact him at jalonso@media.

mit.edu.

Russell S. Harmon is a senior pro-gram manager at the Army ResearchOffice. His research interests includeterrestrial sciences, landmine detec-tion, and environmental quality. Har-mon has a BA in geology from theUniversity of Texas, an MS in geo-chemistry from the Pennsylvania

State University, and a PhD in geology from McMaster University. Contact him at [email protected].

For further information on this or any other computingtopic, please visit our Digital Library at http://www.computer.org/publications/dlib.


Documents

This paper might be a pre-copy-editing or a post-print ...senseable.mit.edu/papers/pdf/20060701_Mitasova_etal_RealtimeLan… · 01.07.2006 · with 6 DOF for navigation and immersive