Next-generation Digital EarthMichael F. Goodchilda,1, Huadong Guob, Alessandro Annonic, Ling Biand, Kees de Biee, Frederick Campbellf, Max Cragliac,Manfred Ehlersg, John van Genderene, Davina Jacksonh, Anthony J. Lewisi, Martino Pesaresic, Gábor Remetey-Fülöppj,Richard Simpsonk, Andrew Skidmoref, Changlin Wangb, and Peter WoodgatelaDepartment of Geography, University of California, Santa Barbara, CA 93106; bCenter for Earth Observation and Digital Earth, ChineseAcademy of Sciences, Beijing 100094, China; cJoint Research Centre of the European Commission, 21027 Ispra, Italy; dDepartment ofGeography, University at Buffalo, State University of New York, Buffalo, NY 14261; eFaculty of Geo-Information Science and EarthObservation, University of Twente, 7500 AE, Enschede, The Netherlands; fFred Campbell Consulting, Ottawa, ON, Canada K2H 5G8;gInstitute for Geoinformatics and Remote Sensing, University of Osnabrück, 49076 Osnabrück, Germany; hD_City Network, Newtown 2042,Australia; iDepartment of Geography and Anthropology, Louisiana State University, Baton Rouge, LA 70803; jHungarian Association forGeo-Information, H-1122, Budapest, Hungary; kNextspace, Auckland 1542, New Zealand; and lCooperative Research Center for SpatialInformation, Carlton South 3053, Australia

Edited by Kenneth Wachter, University of California, Berkeley, CA, and approved May 14, 2012 (received for review March 1, 2012)

A speech of then-Vice President Al Gore in 1998 created a vision for a Digital Earth, and played a role in stimulating the development of afirstgeneration of virtual globes, typified byGoogle Earth, that achievedmany but not all the elements of this vision. The technical achievementsof Google Earth, and the functionality of this first generation of virtual globes, are reviewed against the Gore vision. Meanwhile,developments in technology continue, the era of “big data” has arrived, the general public is more and more engaged with technologythrough citizen science and crowd-sourcing, and advances have been made in our scientific understanding of the Earth system. However,although Google Earth stimulated progress in communicating the results of science, there continue to be substantial barriers in thepublic’s access to science. All these factors prompt a reexamination of the initial vision of Digital Earth, and a discussion of the majorelements that should be part of a next generation.

scientific communication | visualization

Digital replicas of complex enti-ties and systems are now rou-tine in many fields—in thedesign and testing stages of

aerospace engineering, in digital re-constructions of ancient cities, or even inphysiology, in which digital cadavers canreplace real ones in the training of medicalstudents. The concept of a Digital Earth,a digital replica of the entire planet, occursin Al Gore’s 1992 book Earth in theBalance (1), and was developed furtherin a speech written for delivery by Goreat the opening of the California Sci-ence Center in January 1998 (portal.opengeospatial.org/files/?artifact_id=6210). By the turn of the century, supportfor advanced 3D graphics had becomestandard on personal computers, allowingreal-time manipulation of complex ob-jects. Following the speech, the ClintonAdministration directed NASA to forma Digital Earth office, a number of pro-totype projects were begun, and, with the2005 launch of the Google Earth service,a fine-resolution, digital replica of theplanet, or at least of its surface, was withinreach of anyone with a broadband Internetconnection.The effect on the scientific community

was immediate and palpable (2). Here wasa readily accessible technology that couldbe used to present scientific data and re-sults (scientific applications of GoogleEarth were recently reviewed in ref. 3) ineasily digestible, visual form to collabo-rators and a general public that perceivedit as free, fast, and fun. Digital Earth hadrelevance to any science dealing with the

surface and near-surface of the Earth, andcould be used to illustrate and even ad-dress a host of problems facing humanity,from climate change and natural disastersto warfare, hunger, and poverty (see, forexample, the 2009 Beijing Declarationon Digital Earth; 159.226.224.4/isde6en/hykx11.html). The Chinese Academy ofSciences responded by organizing the firstInternational Symposium on Digital Earthin Beijing in 1999; the symposium becamea biennial event, and has been held inCanada, the Czech Republic, Japan, theUnited States, China, and Australia. In2006, the International Society for DigitalEarth (ISDE) was established, and bi-ennial Digital Earth Summits have beenconvened in New Zealand, Germany, andBulgaria. In 2008, ISDE’s official publica-tion, the International Journal of DigitalEarth, was inaugurated.In this concept of Digital Earth, sharing

of scientific information about the planetcould extend well beyond the scientificcommunity, to people with very limitedtechnical skills and computing resources—the vision of an information food chainextending all the way from science topublic policy seemed almost within reach.Google’s own Earth Engine project ex-emplifies this vision by making availableGore’s “vast quantities of geo-referencedinformation.” Unlike geographic infor-mation systems (GISs), which have a rep-utation for being difficult to learn, andforce users to confront the intuitively dif-ficult spatial concepts of scale and mapprojections, Digital Earth implementa-tions such as Google Earth avoided pro-

jections entirely by showing the Earth asseen from space (technically a perspectiveorthographic projection onto the 2Dplane of the screen), measured distance byusing the length of the shortest path overthe curved surface, and reduced scale tothe simple metaphor of raising or loweringthe user’s viewpoint. A fly-by, the “magiccarpet ride” of the Gore speech, thecapstone achievement for generations ofundergraduate GIS students, could becreated by a child of age 10 in 10 minutesby using nothing more than a homecomputer and a downloaded GoogleEarth client.There will always be a need for flattening

the Earth, to see the entire surface at once,albeit distorted, or to present informationon paper. However, there is no doubt thatvirtual globes had enormous advantagesover traditional maps as a means of com-municating data, information, and knowl-edge about the surface and near-surface ofthe planet. Here for the first time weresoftware environments that could overlaylayers of information by using geographiclocation as a common key, show buildingsand vegetation as complex 3D structures,pan and zoom at the click of a mouse, andvisualize extracts from petabytes of rapidly

Author contributions: M.F.G., H.G., A.A., L.B., K.d.B., F.C.,M.C., M.E., J.v.G., D.J., A.J.L., M.P., G.R.-F., R.S., A.S., C.W.,and P.W. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.1To whom correspondence should be addressed. E-mail:good@geog.ucsb.edu.

11088–11094 | PNAS | July 10, 2012 | vol. 109 | no. 28 www.pnas.org/cgi/doi/10.1073/pnas.1202383109

Dow

nloa

ded

by g

uest

on

June

14,

202

0

accumulating georeferenced informationin fractions of a second.It is now 7 years since the launch of

Google Earth, and a similar period haselapsed since the release of the earliest ofwhat is now a long list of comparable virtualglobes, including NASA’s open-sourceWorldWind (worldwind.arc.nasa.gov),Wuhan University’s GeoGlobe, the Chi-nese Academy of Sciences Digital EarthPrototype System, Microsoft’s Bing Maps(www.bing.com/maps), Esri’s ArcGISExplorer (www.esri.com/software/arcgis/explorer/), Unidata’s Integrated DataViewer (www.unidata.ucar.edu/software/idv/), and Digitnext’s VirtualGeo (virtual-geo.diginext.fr/EN/). It is also more thana decade since the Gore speech and thevision that did much to motivate theseefforts, which, in some cases, have ex-tended well beyond the Gore vision. Thisseems an appropriate time, therefore, toexamine what this first generation hasachieved, and to envision what might beachievable and desirable in the future,another 6 or 7 years from now (an earlierand now somewhat dated analysis is pro-vided in ref. 4). At the same time, thesevirtual globes, especially Google Earth,are examined from a critical scientificperspective to draw out some of the issuesthat lie at the intersection of commercialsoftware and the scientific enterprise.

Achievements of Virtual Globes toDateA common reaction to the Gore speech in1998 centered on data volume: with ap-proximately 5 × 1014 sq m of surface, aslittle as a single byte allocated to eachsquare meter results in half a petabyte ofdata before compression. Zooming downto 1 m resolution and panning across thesurface requires some clever algorithmsand data structures if it is to be achievablewith a standard personal computer at theend of a typical broadband connection,with limited capacity for local cachingof data. Virtual globes use a variety ofhierarchical tiling structures known asdiscrete global grids (and reminiscent ofBuckminster Fuller’s geodesic domes) toenable rapid zoom (reviewed in ref. 5),precompute tiles on the server to avoidextensive local computation, and usesophisticated level-of-detail managementto allow the field of view to be refreshed atvideo rates (6).One of themost attractive features of the

virtual globes for the scientific user hasbeen their capacity for extension and ad-aptation to individual needs. Google’sKeyhole Markup Language allows theuser’s 2D and 3D data to be readily su-perimposed on the virtual globe in a vari-ety of formats, opening the door to thevast number of “mashups” that have beenconstructed recently in every area of sci-

ence. Open-source packages such asNASA’s World Wind can be extendedwith new code, and Google’s ApplicationProgramming Interface allows GoogleEarth functions to be embedded within theuser’s own application.Virtual globes that are able to capture

and display elevation above or below thesurface have obvious applications inoceanography, atmospheric science, andgeomorphology. Since its initial launch,Google has also enhanced the temporaldimensions of Google Earth, allowing thedisplay of time series and the use of historicbase maps. Unidata’s Thematic RealtimeEnvironmental Distributed Data Servicesproject (www.unidata.ucar.edu/projects/THREDDS/) aims to facilitate sharing ofdata among Earth scientists using theIntegrated Data Viewer platform, andincludes a full range of 3D and tempo-ral support.Nevertheless, this first generation of

virtual globes has several limitations, eachwith implications for potential uses. Fora scientist, issues of accuracy, replicability,and documentation are likely to be ofspecial concern, although they may be ofless concern for the general public. Some ofthe more technical limitations are dis-cussed in this section, and the followingsection asks what new developments mightcharacterize a second, more powerfulgeneration of Digital Earth.

Accuracy, Replicability, and Documentation.The Earth does not conform to any sim-ple mathematical shape. The geoid orisopotential surface, best understood as thesurface formed by sea level and its imagi-nary extension under the continents, mustbe approximated by a mathematical shapeto define latitude and longitude and thus tomeasure location. Google Earth uses thespecific oblate spheroid known as WGS84,an international standard World GeodeticSystem, and, in principle, measures madeon the surface of this implementation ofDigital Earth should conform to thisstandard. Lines of latitude should growfurther apart away from the equator,whereas lines of longitude will, of course,converge on the poles.Google Earth allows the user to add

several types of figures to the Earth’s sur-face, and in some cases to drag them freelyover the surface. Dragging a circle, forexample, produces several expected be-haviors [these experiments were con-ducted with Google Earth Pro-6.0.1.2032(beta); build date, December 10, 2010]: itsradius remains constant, but its circum-ference grows longer toward the poles asthe local curvature becomes less. How-ever, some behaviors are unexpected, in-cluding a stepwise change of circle area asit is dragged toward the poles. Behaviorbecomes almost chaotic when the circle is

dragged to contain the pole itself, or tointersect with 180° longitude. The mathe-matics of the spheroid are difficult, andperhaps shortcuts are being taken, for ex-ample, by resorting to calculations on thesphere or on a planar projection when thespheroid becomes less tractable.Problems of dealing with uncertainty

may have broader implications. Positionson the Earth are the result of measure-ment, and thus have accuracies that arelimited by the measuring device. It isa long-established principle of science thatmeasurements should be reported witha numerical precision that matches theiraccuracy, and the principle provides easyaccess to estimates of inherent uncertainty.However, latitudes and longitudes areroutinely reported by Google Earth to asmany as six decimal places of degrees(corresponding, in the case of latitude, orlongitude at the equator, to ∼10 cm) ir-respective of the spatial resolution of thedisplay. Distances, for example, betweenLos Angeles and New York, can be re-ported to hundredths of centimeters, anabsurd precision given the lack of an ac-cepted definition of the distance betweentwo extended objects. Clearly the design-ers of this software lost track of an im-portant scientific principle, preferring,instead, to exploit the full numerical pre-cision of the computing system.Readily accessible virtual globes are at-

tractive bases for rapid determination oflatitude and longitude. It is easy, for ex-ample, to find the location of a feature suchas a road intersection by identifying it onthe virtual globe’s already georegisteredimage and capturing its coordinates. Anymisregistration of imagery (by tying animage to the Earth’s surface inaccurately)is inherited by any locations captured byusing it, and when new imagery is insertedin the virtual globe, any previously regis-tered feature will now be offset by anypositional difference in the new registra-tion. It is important to recognize that thisis a measurement problem, and that exactmeasurement of location is impossible.Each new registration of the base imagery,in effect, creates a new datum—a new lo-cal approximation to WGS84 that may ormay not be better than previous approx-imations. A scientist expects extensivedocumentation (i.e., metadata) on theregistration process, its temporal history,and its implications.The actual amount of misregistration of

base imagery varies over space and overtime, depending on the availability ofsmall, recognizable features with knownlocations that can be used as referencepoints, on the spatial resolution of theimagery itself, and on many other factors.Potere (7) found that the mean mis-registration of Google Earth imagery was∼40 m relative to Landsat GeoCover.

Goodchild et al. PNAS | July 10, 2012 | vol. 109 | no. 28 | 11089

PERSPECTIV

ED

ownl

oade

d by

gue

st o

n Ju

ne 1

4, 2

020

Over Santa Barbara, CA, misregistrationhas been as much as 40 m at times relativeto high-quality global positioning system(GPS) measurements. What is important,of course, is not whether misregistrationexists, as it must, but whether it is sub-stantial enough to affect a given applica-tion. Moreover, positional uncertaintyis only one of the many dimensionsof uncertainty that characterize geo-graphic data.

Achievements to Date Against the GoreVision. Despite the progress they repre-sent, today’s generation of virtual globesfall short of Gore’s vision of a DigitalEarth in several key respects. [Goodchild(8) provides a more complete analysis ofthe uses of Digital Earth mentioned in theGore speech, and compares them to thoseof GISs.] First, whereas Gore imagineda Digital Earth that would allow the userto explore backward in time by using his-torical data and forward in time by usingcomputational models to simulate futurescenarios, virtual globes remain largelycentered on portraying the Earth as itlooks at the current time. Older imagery isavailable in Google Earth for many areas,and historical maps have been made ac-cessible as mashups using KeyholeMarkup Language. Simulation-modeloutputs and time-series of imagery havealso been linked to virtual globes. How-ever, to date, it is not normally possible fora member of the general public to recoverwhat his or her neighborhood looked likeat some arbitrarily defined point in thepast, or to use a virtual globe to visualizewhat his or her world will be like, in termsof urban growth, climate change, or sea-level rise, at some point in the futurebased on the best scientific modeling.Second, whereas Gore imagined an en-

vironment in which it would be possible tostore, retrieve, and share “vast quantities”of Earth-related information, in reality,not all such data are equal in the world ofthe virtual globes. Although the user canoverlay data of various kinds on the baseimagery, he or she has no control over theprocess by which Google acquires andmakes available such imagery, and theTerms of Use impose tight restrictionsover what can be done with the results.Rather than exploit virtual globes, thestrategies developed for sharing of geo-graphic data among scientists tend insteadto follow the model of a data library [orgeolibrary (9)]: a searchable catalog ofholdings, formally structured metadata,and mechanisms for facilitating thedownloading of selected data sets (see,e.g., the Geospatial One-Stop, geo.data.gov;or the Global Change Master Directory,gcmd.nasa.gov). Geoportals (10) providea single point of entry into a distributed setof holdings, but nevertheless impose a sig-

nificant burden on the user to executea successful process of search, discovery,extraction, and use that is very differentfrom the user-friendly, visual paradigm ofthe virtual globes.Third, Google Earth’s emphasis on vi-

sualization makes it difficult to communi-cate information that is not inherentlyvisual. A virtual globe that looks like theEarth itself is intuitively straightforward,easy to navigate, and easy to understand.Topography can be shown in three di-mensions, rather than by using the con-tours and other coded techniques adoptedby cartographers; the effects of sea-levelrise can be visualized by flooding areaswith solid blue. However, other propertiesimportant to science, such as temperaturefields, the spatial variation of soil types orvegetation properties, biodiversity, or thedistributions of social, economic, and de-mographic variables, are not inherentlyvisual, and not easily conveyed in imme-diately meaningful ways. Uncertainty iseven more problematic, as there are noobviously intuitive ways of conveying itsmagnitude or implications—we are usedto seeing a single, solid world rather thana world of many conflicting possibilities.

Prospects for the Next GenerationSeveral fundamental changes in society andin the world of digital geographic in-formation have occurred since Gore’sspeech was formulated in late 1997.Mention has already been made of theadvances in bandwidth and 3D graphicsthat have enabled the standard consumerdevice to visualize and manipulate a Digi-tal Earth (Fig. 1). The supply of geo-graphic information from satellite-basedand ground-based sensors has expandedrapidly, encouraging belief in a new,fourth, or “big data,” paradigm of science(11) that emphasizes international collab-oration, data-intensive analysis, hugecomputing resources, and high-end visu-alization. It also demands improved tech-niques of search and discovery such asthose that can be readily implemented ina virtual globe.A large part of this new information is

the property of governments and corpo-

rations, and unavailable to others, orheavily restricted in its reuse. However, thecase for open data access is becoming in-creasingly compelling, given the severityand urgency with which problems must beaddressed at a global scale. The thrusttoward open data is gaining momentum inseveral countries that embrace it for rea-sons of transparency, administrative effi-ciency, and the economic potential ofreuse. They support open governmentthrough legislation and practical measures,such as the production of data in machine-readable formats and the creation ofdata portals.The ubiquity of GPS has allowed anyone

to measure location on the Earth’s surfacewith an accuracy of a few meters, and totag other information, such as photo-graphs, with geographic locations. Crowd-sourcing is now a major source of geo-graphic information (12), exemplified bythe enormously successful, worldwideOpenStreetMap project and Google’sMapMaker. New forms of citizen scienceare encouraging the general public to be-come involved in the acquisition of dataon phenology, weather, disasters, andmany other Earth-related phenomena.Many of these trends come together in theterm neogeography (13), which impliesa breaking down of the traditional dis-tinction between expert and amateur inthe world of mapping and geographic in-formation. Thanks to GPS and a range ofWeb-based services, the average citizen isable to be a consumer and a producer (a“prosumer”) of geographic information,and to make maps, such as those gener-ated by in-vehicle navigation systems, thatare valid only at an instant in time, cen-tered on the user, representing the worldas seen from above or from the ground,and useful only for the purposes of themoment. Many of these services haveproven to be effective even through theinterface of a simple mobile phone (e.g.,Ushahidi, www.ushahidi.com). The citizenas prosumer offers an entirely new visionof the role of the citizen in relation toscience: as volunteer observer, as in-telligent recipient of the results of scienceas applied to the citizen’s own

Fig. 1. Handheld consumer devices such as this tablet computer can augment a real scene with dataobtained from Digital Earth, such as these underground pipes. (Image courtesy of NextSpace.)

11090 | www.pnas.org/cgi/doi/10.1073/pnas.1202383109 Goodchild et al.

Dow

nloa

ded

by g

uest

on

June

14,

202

0

surroundings, and as an informed stake-holder in the Earth’s future.Crowd-sourcing produces information of

variable quality that lacks the structuredsampling and rigorous methods of mea-surement that scientists emphasize. Newmethods will be needed to synthesize suchcrowd-sourced information, and to assureits quality, if it is to be used with confidenceby scientists and more broadly in society. Infuture, techniques of synthesis applied todisparate data of varying quality may be asimportant as techniques of analysis havebeen in the past.These new technologies of neo-

geography offer an unprecedented oppor-tunity for science to extend its findingsfrom the laboratory not only to the pages ofrefereed journals, but also to the eyes andminds of the general public. As easy-to-usetools capable of visualizing vast amountsof information at scales from the global tothe local, the virtual globes clearly havea major role to play in this new vision ofscientific communication. Several aspectsof a renewed Digital Earth vision derivefrom this perspective. The ability to link theglobal to the local, through rapid zoom, isclearly a key aspect of the virtual globes.At the same time, it would be wise to

recognize that a simple linear model ofcommunication between scientists and thepublic is naive in several respects. Scientistshave much to learn about communication,especially in dealing with the uncertaintyinherent in all measurements and pre-dictions, as the recent “Climategate” andits aftermath demonstrated. The generalpublic can often understand concepts ofuncertainty and probability if they are ex-plained well, and made an inherent part ofsystems like Digital Earth. Research overthe past two decades has given us a richunderstanding of the modeling and visu-alization of uncertainty in geographicinformation, and of how to communicateuncertainty to the user. It is vital that thenext generation of Digital Earth replacethe exaggerated precision of the first gen-eration with techniques that convey un-certainty clearly and unambiguously to theuser. More broadly, the social scienceshave much to offer on the subject ofcommunication, the cultural differencesthat characterize the various environmentsin which Digital Earth is accessed, and theways in which data are understood andinterpreted. Vast differences exist aroundthe world in access to computer hardware,software, and the Internet, as well as in thetechnical skills and attitudes of users. Thenext generation of Digital Earth will needa strong commitment to involving thesocial sciences in its development.One of the most powerful benefits of

georeferencing—in other words, the asso-ciation of attributes with locations on theEarth’s surface—is context. Observations

made at a location can readily be linked toother known facts about that location, orabout nearby locations, in a rich con-textualization. We can distinguish betweenhorizontal context, or the context estab-lished by knowledge about nearby loca-tions, and vertical context, or the contextestablished by other things that are knownabout the same location. Both are cap-tured in the famous principle of Tobler,often termed the First Law of Geography:“All things are related, but nearby thingsare more related than distant things” (14).A virtual globe provides easy access tocontext, especially horizontal context, al-lowing inferences to be made from theconditions surrounding a location. Verti-cal context is more difficult to visualize,because a virtual globe allows only onelayer (or what one might term on a virtualglobe a “topping”) to be transparentlyoverlaid on the imagery base. Mappingtechniques such as cross-hatching mightallow for two layers, but, more generally, itwould be useful to look for more powerfulways of making the user aware of thevertical context of geographic attributes.Moreover, insights can often be gained notfrom the horizontal and vertical context ofattributes, but by searching for analogousconditions elsewhere on Earth; virtualglobes seem especially well suited to thiskind of investigation.More important, perhaps, would be the

development of ways of capturing thelinkages that exist between locations andacross scales. Maps are powerful ways ofvisualizing the properties of locations, suchas elevation or soil type, or what one mightterm “unary” geographic information.However, they are much less powerful atportraying the linkages that exist betweenplaces, such as flows of migrants or theconfigurations of social networks—in-formation that one might term binary be-cause it involves the properties oflocations taken two at a time. A map ofmigration flows between each pair of themore than 3,100 US counties, for example,would show a mass of lines that wouldbe impossible to comprehend withoutsophisticated techniques of abstractionand generalization (e.g., ref. 15), whereasa map of county life expectancy or eth-nicity is easily understood. Linkages areclearly important to our understanding ofenvironmental and social processes, andaccess to information about them shouldbe an important part of a renewed vision.This is especially true when linkagesrepresent the impacts of changing con-ditions in one location on conditions inanother location, such as downstream ordownwind.The Gore speech describes a Digital

Earth that would be capable of presentingpredictions about the Earth’s future—ofclimate change and sea-level rise, for ex-

ample, or food supplies. Such predictionsare commonly the outputs of simulationmodels, and represented in the extreme byprojects that aim to develop a DigitalEarth capable of predicting a vast array ofproperties at fine resolution in space andtime, such as Europe’s FuturICT (www.futurict.eu) or Japan’s Earth Simulator(www.jamstec.go.jp/esc/index.en.html).They might be generated in one of twodistinct ways: by running a model in itsown software environment and then visu-alizing the results by overlaying them asa topping on a Digital Earth, or by runninga model within the Digital Earth softwareenvironment itself, perhaps by using thevirtual globe’s discrete global grid as a setof finite elements. The latter might allowthe virtual-globe user to modify parame-ters or boundary conditions, to evaluatealternative scenarios. In essence, such ap-proaches would allow the user to in-vestigate how the Earth works and how itmight evolve in the future, rather than toobserve how it currently looks. In the spiritof neogeography, these approaches mightallow the user to augment reality withvisualizations from the user’s own per-spective. They would require substantialinvestment in model management, giventhe vast number of models that are al-ready available, their lack of intero-perability, and the difficulty of choosingbetween them.It would be important to incorporate

a few simple analytic functions in theinterests of supporting the evaluation ofscenarios. A subset of the functions gen-erally available in GIS would allow the userto obtain summary statistics for areas,for example, or to compute correlations.However, wholesale adoption of thefunctionality of a GIS is likely to lead to thesame kinds of inaccessibilities the averagecitizen already encounters with GIS soft-ware. Careful consideration of cognitiveissues will be needed to find an appropri-ate compromise between the scientist’sneed for advanced functionality and thegeneral public’s need for simplicity andease of use.Although it is always risky to offer any

kind of prediction of technological futures,several trends are already apparent thatshould be incorporated in a renewed visionof Digital Earth. We already have themeans to keep track of many kinds ofobjects: many vehicles, and the mobilephones of their occupants, are constantlytracked by using GPS and managed as“probes” to collect data on traffic speedsand congestion; commercial shipments areoften tracked, as are pets, parolees, andmany types of animals, in the interests ofresearch. In the future, we should envisiona world in which it will be possible to knowthe locations of everything at all times,through a georeferenced “Internet of

Goodchild et al. PNAS | July 10, 2012 | vol. 109 | no. 28 | 11091

Dow

nloa

ded

by g

uest

on

June

14,

202

0

Things.” We are also seeing a massiveexpansion in the functionality of mobiledevices, and can envision a world in whichcomputing will be ubiquitous, and in whichDigital Earth will be accessible anywhere,at any time.Although GPS positioning generally

works more effectively outdoors, andgeographic information is mostly 2D de-spite the increasing prevalence of 3Dbuilding representations in virtual globes,we should anticipate a day when it will bepossible to support georeferencing andnavigation within detailed 3D structuresanywhere on the planet to centimeter ac-curacies (see, e.g., the work of the D_CityNetwork, dcitynetwork.net and Fig. 2).Despite these advances, the problem of

archiving remains: we have, as yet, nosatisfactory solution to the long-termpreservation of the vast amounts of digitaldata already created, along with the ex-ponential growth in data that seems likelyto continue for some time to come. Muchof what we know of the history of the planetis recorded on paper, and some of it hasalready survived for centuries. However, itis very hard at this point to see how century-long preservation of digital media can beachieved, together with the means tosearch and retrieve useful information.Extensive research and development willbe needed if the records now being as-sembled for Digital Earth are to survive formore than a few years.Finally, today’s world of Digital Earth is

predominantly silent, using only the visualchannel to communicate with the user.Sound is at least as rich a medium forcommunication as vision, yet it plays al-most no part in the communication ofgeographic information. The next genera-tion of Digital Earth might make use ofaudio in several ways: by allowing the userto make requests through speech, by

storing the sounds and stories associatedwith locations, and by responding to spo-ken place names. Audio is already stronglyassociated with small portable or wearabledevices, which may in the future becomethe primary modality for access toDigital Earth.

The Way ForwardIt is one thing to describe a vision for thenext generation of Digital Earth, but itmay be quite another to imagine how itmight be achieved. The top-down optionfor implementing Gore’s Digital Earththrough large-scale public investmentstuttered to a halt shortly after the 2000US Presidential election, and in the cur-rent funding climate, a top-down initiativefor a next generation seems equallyunlikely.In reality, the first generation resulted

from advances in Earth observation; froma timely evolution of bandwidth and 3Dvisualization, the latter driven in part by thevideo game industry; from the modeststimulus provided by government funding;from the impetus provided by a US VicePresident; by the commercial possibilitiessensed by large corporations such asGoogle and Microsoft; and by the interestsand enthusiasm of the open-source com-munity. Underlying this was a sense amonga loose-knit community that Digital Earthwas both possible and desirable. We believesimilar factors can drive a next generation.The next generation of Digital Earth will

not be a single system but, rather, multipleconnected infrastructures based on openaccess and participation across multipletechnological platforms that will addressthe needs of different audiences (16). Amore dynamic view has also been pro-posed of Digital Earth as a digital nervoussystem of the globe, actively informingabout events happening on (or close to)

the Earth’s surface by connecting to sensornetworks and situation-aware systems (17).Although great progress has been made

in overcoming differences of software anddata formats in sharing and communicatinggeographic information, issues of seman-tics continue to present major challenges.Differences of mapping practice, the waysin which land is classified, and the mean-ings attached to data are being addressedby projects such as the European Com-munity’s Infrastructure for Spatial Infor-mation in the European Community(inspire.jrc.ec.europa.eu). However, theEarth’s surface is clearly heterogeneous,and any community will inevitably deviseits own ways of describing its own sur-roundings. Although remote sensing ach-ieves a degree of uniformity, other sourcesof geographic information tend to reflectthe standards and practices of their source.One has only to compare, for example, themaps of the political boundaries of theHimalayan region provided by GoogleMaps in the US with those provided byGoogle Maps in China and India (Fig. 3)to realize the impossibility of achievinguniformity when mapping the humancondition.Any effort to develop a next-generation

Digital Earth will require new governancemodels. In addition to the ISDE (www.digitalearth-isde.org), many other organ-izations, including the Global Spatial DataInfrastructure Association (www.gsdi.org),the International Council for Science andits Committee on Data for Science andTechnology (www.codata.org), the Groupon Earth Observations (and its geo-portal, the GEO System of Systems, www.earthobservations.org), the UN Commit-tee of Experts on Global Geospatial In-formation Management (ggim.un.org),and many national agencies will addressvarious aspects of the future Digital Earth.These organizations can play a helpful rolein endorsing the concept of a next-gener-ation Digital Earth, and helping to elabo-rate its vision. Along with the scientificcommunity, they can also play a useful rolein ensuring that the next generation meetsthe highest standards of scientific rigor,especially careful and detailed documen-tation of uncertainty. Perhaps there is alsoroom for a Digital Earth code of ethicsthat could set standards for behavior ina complex, collaborative enterprise.However, in the realities of today’s

economies, it seems most likely that in-novation will come from the private sector,together with volunteer initiatives cen-tered around open-source tools. The pri-vate sector is fundamentally competitive,but organizations such as the Open Geo-spatial Consortium (www.opengeospatial.org) have been able to achieve remark-able degrees of cooperation and in-teroperability. A similar collaboration

Fig. 2. A rendering of Auckland, New Zealand, using 3D models of all of its buildings and other builtstructures. (Image courtesy of NextSpace.)

11092 | www.pnas.org/cgi/doi/10.1073/pnas.1202383109 Goodchild et al.

Dow

nloa

ded

by g

uest

on

June

14,

202

0

among the private sector, academia, non-governmental organizations, and govern-ment, focused on Digital Earth, might besimilarly successful.Collaboration at this scale will also be

necessary to tackle the growing issues ofprivacy and ethics that are associated withaccess to fine-resolution geographic in-formation (18). Technologies such as GPSand radio-frequency identification (RFID)raise the possibility of a future in which itwill be possible to know and record whereanything is, at all times. Fine-resolutionimaging from space can identify objects onthe surface less than 1 m across, and im-aging from the ground can be used torecognize faces and license plates, creatinga world that Dobson and Fisher charac-terize as “geoslavery” (19). To date, effortsto protect privacy have been patchwork innature: for example, some countries nowrequire Google Street View imagery toblur faces and license plates, and mobile-phone service providers in the US arerestricted in their use of locational in-formation. Against this one must place theevident willingness of people to publicizetheir location through services such asFourSquare or Google Latitude. To date,no principle of privacy protection hasemerged that finds an acceptable, consis-tent position between the extremes ofuniversal access and universal protection.At minimum, perhaps any individualshould have a guarantee of control overhis or her own locational privacy, able toturn it on or off at will.It is clear to us that the first generation is

powerful but limited, that technology isadvancing rapidly and making possiblewhat might have been inconceivable a fewyears ago, and that humanity needs betterways of linking the scientific communityand its discoveries about the ways the Earthworks and is structured with the generalpublic and their growing concern about theplanet’s future. We would even go so far asto argue that access to scientificallygrounded information about the planet’sfuture is a basic human right; that it shouldbe available to all, independent of nationalpolicies and strategies, and in a form thatis readily understood and absorbed. Thisis, after all, the only planet we have, andwe all have a vested interest in its future.Digital Earth seems to us the most acces-sible and compelling way to organize,visualize, and use that investment.

ACKNOWLEDGMENTS. This paper is largely theoutcome of aworkshop on next-generation DigitalEarth held in Beijing in March 2011 and hosted bythe Center for Earth Observation and Digital Earthof the Chinese Academy of Sciences and the Sec-retariat of the International Society for DigitalEarth (ISDE). The authors thank LukeDriskell for hisextensive notes of the discussion. The paper alsobenefited from comments on earlier drafts frommembers of the ISDE Executive Committee: MarioHernandez and David Rhind.

Fig. 3. Three versions of the boundaries of the Himalayas, as provided by Google Maps in response toqueries from (A) the US (Copyright Google, Kingway, MapKing, Mapabc, SK M&C, TeleAtlas, ZENRIN), (B)China [Copyright GS(2011)6020 Google, Kingway, MapKing, Mapabc, TeleAtlas], and (C) India (CopyrightGoogle, LeadDog Consulting, Mapabc, TeleAtlas).

Goodchild et al. PNAS | July 10, 2012 | vol. 109 | no. 28 | 11093

Dow

nloa

ded

by g

uest

on

June

14,

202

0

1. Gore A (1992) Earth in the Balance: Ecology and theHuman Spirit (Houghton Mifflin, Boston).

2. Butler D (2006) Virtual globes: The Web-wide world.Nature 439:776–778.

3. Yu L, Gong P (2012) Google Earth as a virtual globetool for Earth science applications at the global scale:Progress and perspectives. Int J Remote Sens 33:3966–3986.

4. Craglia M, et al. (2008) Next-Generation Digital Earth.A position paper from the Vespucci Initiative for theAdvancement of Geographic Information Science. Int JSpatial Data Infrastructures Res 3:146–167.

5. Sahr K, White D, Kimerling AJ (2003) Geodesic discreteglobal grid systems. Cartography and Geographic In-formation Science 30:121–134.

6. Bar-Zeev A (2007) How Google Earth [really] works,RealityPrime Advanced Technology Research, July 3,2007. Available at www.realityprime.com/articles/how-google-earth-really-works. Accessed July 16, 2011.

7. Potere D (2008) Horizontal positional accuracy of Goo-

gle Earth’s high-resolution imagery archive. Sensors 8:

7973–7981.8. Goodchild MF (2008) The use cases of digital earth. Int J

Digital Earth 1:17–30.9. National Research Council (1999) Distributed Geolibra-

ries: Spatial Information Resources: Summary of a

Workshop (National Academy Press, Washington, DC).10. Maguire DJ, Longley PA (2005) The emergence of geo-

portals and their role in spatial data infrastructures.

Comput Environ Urban Syst 29:3–14.11. Hey AJG, Tansley S, Tolle KM (2009) The Fourth Para-

digm: Data-Intensive Scientific Discovery (Microsoft Re-

search, Redmond, WA).12. Goodchild MF (2007) Citizens as sensors: The world of

volunteered geography. GeoJournal 69:211–221.13. Turner A (2006) Introduction to Neogeography

(O’Reilly, Sebastopol, CA).

14. Sui DZ (2004) Tobler’s First Law of Geography: A big

idea for a small planet? Ann Assoc Am Geogr 94:

269–277.15. Tobler WR (1975) WINDS: A Computer Program for the

Analysis of Geographical Interactions. Cartographic

Laboratory Report No. 13 (Univ Michigan, Ann Arbor).16. Craglia M, et al. (2012) Digital Earth 2020: towards

a vision for the next decade. Int J Digital Earth 5:4–21.17. De Longueville B, Annoni A, Schade S, Ostlander N,

Whitmore C (2010) Digital Earth’s nervous system for

crisis events: real-time sensor Web enablement of vol-

unteered geographic information. International Jour-

nal of Digital Earth 3:242–259.18. Curry MR (1998) Digital Places: Living with Geographic

Information Technologies (Routledge, New York).19. Dobson JE, Fisher PF (2003) Geoslavery. IEEE Technol

Soc Mag 47:47–52.

11094 | www.pnas.org/cgi/doi/10.1073/pnas.1202383109 Goodchild et al.

Dow

nloa

ded

by g

uest

on

June

14,

202

0