This article was downloaded by: [Northeastern University]On: 06 October 2014, At: 09:02Publisher: Taylor & FrancisInforma Ltd Registered in England and Wales Registered Number: 1072954 Registered office: MortimerHouse, 37-41 Mortimer Street, London W1T 3JH, UK

Cartography and Geographic Information SciencePublication details, including instructions for authors and subscription information:http://www.tandfonline.com/loi/tcag20

A New World Geographic Reference SystemKeith C. Clarke , Peter H. Dana & Jordan T. HastingsPublished online: 14 Mar 2013.

To cite this article: Keith C. Clarke , Peter H. Dana & Jordan T. Hastings (2002) A New World Geographic ReferenceSystem, Cartography and Geographic Information Science, 29:4, 355-362, DOI: 10.1559/152304002782008387

To link to this article: http://dx.doi.org/10.1559/152304002782008387

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) containedin the publications on our platform. However, Taylor & Francis, our agents, and our licensors make norepresentations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose ofthe Content. Any opinions and views expressed in this publication are the opinions and views of the authors,and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be reliedupon and should be independently verified with primary sources of information. Taylor and Francis shallnot be liable for any losses, actions, claims, proceedings, demands, costs, expenses, damages, and otherliabilities whatsoever or howsoever caused arising directly or indirectly in connection with, in relation to orarising out of the use of the Content.

This article may be used for research, teaching, and private study purposes. Any substantial or systematicreproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in anyform to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://www.tandfonline.com/page/terms-and-conditions

A New World Geographic Reference System

Keith C. Clarke, Peter H. Dana,and Jordan T. Hastings

ABSTRACT: A nell' global georeferencing system-the World Geographic Reference System 0¥GRS)-isproposed. This system has particular advantages for location description and communication withelectronic devices, i.e., in digital environments that are shared between humans and machines. Thenew World Geographic Reference System strikes a compromise between the dominant use of numbersin established scientific coordinate systems, such as latitude/longitude, and the colloquial preference fornames, particularly names of administrative units and populated places, in everyday life. Specifically,WGRS defines a system of uniform regional grids, each lOOxlOOkm in extent, anchored on and namedby prominent cultural and/or physical features. Subsets of these regional grids, called local grids, whichare particularly adapted to smaller places, also may be defined. A location within a regional or localgrid is georeferenced by suffixing the grid identifier with a coordinate string of doued-digit-pairs thatrepresent interleaved Cartesian x-y displacements from the grid origin. A typical WGRS locator, forexample, is US.DC.WAS.54.18.28,representing a lOOxlOOm area, the southwest corner of which is 0.512of the way across (east) and 0.488ofthe way up (north) in the Washington, B.C., grid, roughly the lawnsurrounding the Washington Monument. This 10cat01~which is easily interpreted by both humansand machines, also rna)' be effectively communicated between them via computer networks using anotation, such as "wgrp://us.DC.WAS.54.18.28"in web code. The similarity ofWGRS locators (WGLs)to Uniform Resource Locators (URLs) on the Internet is intentional, facilitating their use in Web andwireless application interfaces, especially those employed in location-based service systems.

KEYWORDS: Coordinate systems, geocoding, georeferencing, mapping grids

Introduction

Geospatial technologies are ultimatelydependent upon the georeferencinggrid systems used to assign coordi-

nates to locations on or near the surface of theEarth. Cartography and geodesy have devised amultitude of such grid systems over time for vari-ous purposes, including the Universal TransverseMercator (UTM) system, the U.S. Military GridReference System (MGRS), the State PlaneCoordinate System (SPCS), the British NationalGrid (BNG), and many others. Regardless of theirexact technical specifications, all such grid systemsshare: 1) a structured terminology and explicitsyntax for coordinate references; 2) a specificmechanism for producing coordinates in rela-tion to a geodetic datum and projection system;

Keith C. Clarke is professor and chair at the Department ofGeography, University of California Santa Barbara, Santa Barbara,CA 93106 U.S.A. E-mail: <kclarke@geog.ucsb.edu>. PeterH. Dana is a research fellow and lecturer at the Departmentof Geography, University of Texas at Austin, Austin, TX 78712,U.S.A. E-mail: <pdana@mail.utexas.edu>. Jordan 1. Hastingsis a doctoral student at the Department of Geography, Universityof California Santa Barbara, Santa Barbara, CA 93106, U.S.A. E-mail: < jordan@geog.ucsb.edu>.

and 3) a unique and exhaustive system of spatialindexing to either point or areal locations. Spatialindexing permits tiling of grids according to agiven pattern over the region of interest. Tilingapproaches have varied from 6-degree longitudi-nal strips (UTM), to 6 by 8 degree blocks (MGRS),to the boundaries of states or counties within states(SPCS). Surprisingly, little cartographic researchhas focused on either identifying the benefits ofparticular grid systems or comparing them analyti-cally. Goodchild (1994) specified a set of criteria forgrids, including accuracy, authority, definitiveness,exhaustiveness, hierarchy, simplicity, succinctness,tractability, uniqueness and universality. Some ofthese criteria were analyzed further by Kimerlinget a1. (1999) and by Clarke (2002). In designingthe present World Geographic Reference System(WGRS), we have emphasized simplicity abovemost of the other criteria, especially exhaustive-ness and uniqueness.

Given a grid system, precise location references aremade by numeric offsets (e.g., eastings and northings)within its tiles. The complete georeference is typi-cally encoded in a single alphanumeric string, orgeocode, that gives both the identifier (or index) ofthe base grid and the offsets within it in some form.The length of the geocode determines the precisionof the system. In UTM, for example, the top of

CartograPhy and GeograPhic Information Science, Vol. 29, No.4, 2002, pp. 355-362

Dow

nloa

ded

by [

Nor

thea

ster

n U

nive

rsity

] at

09:

02 0

6 O

ctob

er 2

014

the dome of the U.S. Capitol in Washington occursat 325755m E, 4306467m N; Zone18 N (NAD83).Here the leading numbers define the easting andnorthing coordinates, indexed to zone 18 (North),and NAD83 specifies the datum.' Alternative recentsystems have varied both the nature of the tilingand the syntax of geocodes (Tobler and Chen 1986;Dutton 1997).

The primary benefits of encoded georeferences aresimplicity and universality, while some geocodes alsoprovide explicit indication of precision. Automatedsystems can take advantage of compression schemeswhen many geocodes are stored together (Clarke1995). However, geocodes tend to be cryptic andso are susceptible to being misstated or confused.Problems also arise with grid tile overlap and projec-tion distortion, particularly near grid edges and atthe poles. The inevitability that important featureswill fall on grid seams has been termed the "secondlaw of geography" (Clarke 1995), i.e., that things ofhighest geographic interest lie at the intersection ofseveral unfortunately tiled maps or images. Overall,georeferencing grid systems may be mathematicallysuccinct, but they are seldom user-friendly. The BNG,devised and promulgated for a relatively small region,is a notable exception and widely used.

Bycontrast to these formal, mathematical approaches,the majority of georeferencing in everyday life isaccomplished in relation to named places, i.e., wellknown cultural and/or physical geographic features(Dana and Hancock 1997). Place names also appearprominently in street addressing and routing instruc-tions, and in many legal descriptions. Postal codesare a conspicuous exception (Raper et aI. 1992) tothe human preference for names over numbers andcodes. Thus, while the UTM geocode given abovelocates the Capitol dome within one meter, so does thesimple text string "Top of U.S. Capitol, Washington,D.C., USA," which also uses 40 characters (includ-ing punctuation and spaces). The former notationrequires a map expert to decode, while the latter isaccessible to anyone with knowledge of basic geog-raphy, presuming the place name has a generaIlyaccepted meaning. Automated conversion of placenames to geodetic coordinates is error-prone, however,and often imprecise; metrics between named placesare necessarily approximate. We propose WGRS asa compromise between the two georeferencing sys-tems, mathematical and conventional, preservingthe essential benefits of each.

The intent ofWGRS is to create a firm foundationfor location description and communication withelectronic devices in the 21st century. Webegin fromthe position that the geographical reference systems

in common use (place names, cultural landmarks,street addresses) are undeniably more user-friendlythan grid coordinates of any kind. In these systems,point locationsare typicallygivenvia street addressesand/or intersections, while more general locationsare identified by the names of prominent places. Thelatter are often nested, e.g., "the church in CentralPark," sincehuman geographicbehavior isacross-scaleand multi-scale in nature (Goodchild and Proctor1997). FoIlowingthese precepts, WGRS is both col-loquial and hierarchical, intermixing mnemonic andnumeric stylesof georeferencing in a notation equallyaccessible to people and machines. The WGRSloca-tor for the U.S. Capitol dome, again with one-meterprecision as given earlier, is the 24-character codeUS.DC.WAS.54.38.58.56.29. This geocode iscompact andsuccinct, facilitating its use in print and speech asweIlas electroniccommunications that involveglobalpositioning systems (GPS),geographic informationsystems (GIS), and the wide array of emerging loca-tion-aware appliances such as ceIl-phones, personaldigital assistants, and in-car navigation systems. Inboth written and verbal forms, WGRS is specificallydesigned for ease of use by people and machines, tofacilitate their interaction on everyday geographicchores. The WGRSnotation also emphasizes localityin georeferencing, thereby encouraging its adoptionand preserving geographic heritage. This localiza-tion has the additional benefit that spatial error isminimized in practical terms, not over an arbitrarymap region, but specifically centered on places ofmaximum interest and use.

The World Geographic Reference System isthe outcome of collaboration between a private-sector location-based services (LBS)provider, UDSDirectory Corp., d/b/a g02 Directory Systems (G02;http://www.g02.com). and the National Center forGeographic Information and Analysis(NCGIA;http://www.ncgia.org)at the Universityof California SantaBarbara. Various features ofWGRS are covered byU.S. patents (Hancock and Dana 1998, et seq.). G02intends to make a significant portion of the WGRStechnology available to software developers, otherLBS providers and academic researchers on a com-munity source-license basis (Rosenberg 2000). Tofurther encourage its use, WGRS may be proposedfor incorporation in Internet and Webspecificationsdeveloped through the Internet Engineering TaskForce (http://www.ietf.org)and the World Wide WebConsortium (http://www.w3.org). Essential compo-nents oHVGRS,including its technical principles andcomputer algorithms and databases, will be postedand maintained on the Web and also publishedperiodicaIly in hardcopy forms. The G02 Web site

I The bare minimum UTM encoding is 15 digits, "325755430646718", plus 1 (bit) more for the hemisphere.

356 CartograPhy and GeograPhic Information Science

Dow

nloa

ded

by [

Nor

thea

ster

n U

nive

rsity

] at

09:

02 0

6 O

ctob

er 2

014

N

Grid Definition

providessoftware(http://www.go2online.com/webbrowser/indexGo2AddressWizard.cfm)that allowsconversionofzip codes and street addresses into WGRS locators.

This paper introduces the World GeographicReference System conceptually, formally and inde-pendently of its applications, sinceWGRSstands aloneas a unique spatial indexing system. We concludewith a discussion ofWGRS's anticipated advantagesand challenges.

the grid center-lines. The grid origin is shifted to thelower-left corner of its .00 cell by a false easting anda false northing of 50,000 m each (Figure 1). Thegrid is oriented such that its vertical center-line runsdue north-south, coincident with the local meridianat the center-point.

Locations within a grid, which are always squareareas, are notated bya string ofdotted-digit-pairs (DDP),which are interpreted hierarchically. Within eachpair, the first digit represents the lower-left x-coor-dinate (easting) and the second digit represents thelower-left y-coordinate (northing) in the hierarchy,which applies initially to the IOOxiOOgrid square.Each digit-pair subdivides its antecedent grid spacedecadally in tiers, i.e., the first pair resolves to a cell,the second pair to a sub-cell, etc. Thus, for example,the location .55 denotes the upper-right cell (lOxlOkm area) of the four central cells of any grid; whereas

.28.55 denotes a sub-cell (Ixi km area) within thegrid's .28 cell (see Figure 1).

The mnemonic basis for WGRS is the place name(PN): the official, fullyspelled-out name ofa culturalor physical geographic feature on the Earth. In gen-eral, PNs are as specified in the National Imageryand Mapping Agency (NIMA) world gazetteer ofplace names,3 which also includes their locations inMGRS notation. Place names are preferably angli-cizedusing the International Standards Organization(ISO) 8859 Extended Latin Character Set (Latin1), although multilingual versions of names can beaccommodated. For convenience in compositingstrings of names (below), a PN generally will beshortened to a place name abbreviation code (PNAC),which is any accepted abbreviation for the place thatis unambiguous within the context of that string. Inthe U.S., the Federal Information ProcessingStandard(FIPS)state and county name codes are adopted (seehttp://www.itl.nist.gov/fipspubsJby-num.htm/. docu-ments 5-2 and 6-4); analogous conventions and/orstandards apply in many countries.

Place names and/or place name abbreviation codesare compo sited hierarchically into place name strings(PNS), using dot separators similar to the DDPs ingrid locations. The composition progresses fromsmaller to larger spatial scales, left to right. Thecoarsest (left-most) element in the PNS is the ISO-3166 two-character world country code, e.g., 'us'for the United States; the intermediate elementsare country-dependent codes for administrativeunits-states, provinces, regions, and departments;the finest (right-most) element is typically a code fora populated place. For example, Washington, D.C.,might be WGRS-coded as US.De.WAS.

.00grid origin

.55grid center

.28 cellsub<ivlded

'1100km

I T -r-' I

I ~. II I

I II I r. --.-f

\..:':.j

I

~ I

"Figure 1. A single WGRS grid.

The geometric basis for WGRS is the regional grid(RG): a IOOxiOOkm Cartesian mapping frame-work, comprising one hundred IOxiO km cells,each of which in turn comprises one hundred Ixikm sub-cells, etc. (Figure 1). Using a stereographicprojection, such a grid may be readily constructedanywhere on Earth; different computational meth-ods apply to polar, equatorial, or oblique cases(Snyder 1987). WGS-84 is taken as the geodeticdatum.2 The stereographic projection is chosendespite the fact that it is not equal area, becauseit is conformal with error increasing uniformlyoutward from the central point of projection, thusminimizing error at the point to which it is applied.Alternative projections are possible.

The center-point for the grid, which is also thecenter of the projection, is placed at the lower leftcorner of the grid's .55 cell, i.e., the intersection of

.!1 •oo

2 Specific WGS·84 parameters will be selected to establish a consistent one-to-one mapping of geodetic coordinates to WGRS, andvice-versa, on all computing platforms.

3 NIMA maintains a gazetteer server with access to 3.5 million such features in a database, at http://www.nima.miVgns/htmV.

Vol. 29, No.4 357

Dow

nloa

ded

by [

Nor

thea

ster

n U

nive

rsity

] at

09:

02 0

6 O

ctob

er 2

014

Each regional grid is identified by a place namestring; dotted-digit-pair suffixesfurther refine locationswithin them. Thus US.DC.WAS.55,US.DC.WAS.55.55,andUS.DC.WAS.55.55.55represent progressively tighter focuson the center of Washington, while US.DC.WAS.34.98refers to a Ixi km area, the southwest corner ofwhich lies 11 km west and 2 km south of the citycenter. Upper tiers at the left of the location stringcan be suppressed if the context is evident; similarly,lower tiers can be truncated on the right when thedesired spatial resolution is achieved. Well formed, i.e.,syntactically and semantically correct, unambiguousWGRS georeferences are termed World GeographicLocators (WGL).The notation is intentionally similarto that of Uniform Resource Locators (URL) (WorldWide Web Consortium 2002) on the Web.

The unique value of WGRS is that its underlyingframework, that of regional grids, is predominantlyculturally defined: regional grids are generally selectedto coincide with the most populous places. The center-point of each regional grid is derived by rounding thelisted geographic coordinates, or anchor point, 4 for thedefining real-world place to the nearest 0.05 degrees(3 minutes) of latitude and longitude, typically <5km. By this definition, the "center" of a place, e.g.,its business district or cultural downtown, will alwaysoccur in or among the four central cells (.44,.45,.54,and .55)of its regional grid, and the cross (+) wherethese cells meet defines the four cardinal directionsat that point on the grid (refer to Figure I) ..

As a comprehensive example, the WGRS grid forWashington, D.C., US.DC.WAS,has the anchor point(38°53'42"N,77°02'I2"W),at the ZeroMilestonemarker(http://www.fhwa.dot.gov/infrastructure/zero.htm).Thecorresponding RG center-point is (38.90oN, 77.05°W),approximately 1 km northwest of the Mall (Figure 2).The Washington Monument5 is one of many wellknown landmarks in this grid. The following WorldGeographic Locators apply:• uS.DC.WAS-thefullIOOxiOOkm area (grid) centered

on the Zero Milestone marker, i.e., extending 50 kmN, E, S, and W from the center-point;

• WAS.55-the lOxlO km area (cell) immediately NEof the grid center; the prefix tiers US.DC.have beendropped, optionally;

• WAS.54-the lOxlO km area (cell) immediately SEof the grid center;

• WAS.54.18-the Ixi km area (sub-cell) beginning 1km across and 8 km up within the .54cell;

• WAS.54.18.28-the1OOxlOOm lawnarea surroundingthe Washington Monument;

Figure 2. Geographic and WGRS references for theWashington Monument.

• WAS.54.18.28.83-thelOx10m square approximatingthe concrete base of the '''Tashington Monumentitself; and

• US.DC.WAS.54.18.28.83.11-fullWGL for the Ixlm center area directly beneath the apex of theWashington Monument.In relation to a given regional grid (RG), any

number of local grids (LGs) may be specified, eachofwhich exists on the coordinate framework definedby that RG (Figure 3). The defining RG is referredto as the parent grid and the LGs as its children, all ofwhich are siblings. Local grids also are identified byplace names, which are syntactically indistinguish-able from those of regional grids. Hence, the WorldGeographic Locator (WGL)geocodes based on localand regional grids may be used interchangeably. Alsoby this construction, the sub-cell (second DDP) andall lower tiers of an LG-based WGL are identical tothose of the parent RG-based WGL for the sameplace, avoiding proliferation of only slightly differ-ent gridding frameworks.

Typically, local grids are subsets of the definingregional grid's cells; however, limited supersets(groupings) are also possible (refer to Figure 3).For consistency, both types of grids designate .55as their central cell. To accomplish this, the LG isshifted, if necessary, by an integral number of cellseast (or west) and north (or south) in relation to itsparent RG. In general, only a few cells of a localgrid, roughly corresponding to the perimeter ofthe locality, are treated as valid or activated; the bal-ance are un-activated, or masked. In this way, thesame point can have both a local reference and abroader regional reference, depending on the user's

4 Typically, but not necessarily its centroid; definition of centroids is beyond the scope of this discussion.5 Listed by the National Geodetic Survey as Washington Monument 1913 PID=HV4442 NAD 83 0993)-38 53 22.08377(NI 077 02

06.86378 (W). This location is described in the data sheets as "the center of the aluminum tip that surrounds the monument. This tiphas apparently been burned by lightning, as the top is about 'h inch square (1934)." In 2000, the monument was reworked and the tipwas replaced.

358 Cartography and GeograPhic Information Science

Dow

nloa

ded

by [

Nor

thea

ster

n U

nive

rsity

] at

09:

02 0

6 O

ctob

er 2

014

Figure 3. Shifting and masking of local vs. regional WGRS grids.Anchor A defines a regional grid; Band C are sibling local gridswithin it. Place perimeters are shown with activated grid extentscross-hatched.

I ·E...:

C>C>

I

Y 0

ro:'(-------- 100 km

application, although only one local grid can apply(be active)at any given time. The parent RG isalwaysconsidered activated over its full geographic extent.The activated extent of a LG may fall outside theperimeter of its parent RG, however, in border-cellcases. For example, in Figure 3, the LG cell C.55 issynonymous with its parent RG cell A.99, and hence6 cells of C fall outside A altogethel: The maximumeffectivereach ofa regional grid is thus 190x190 km,in the (unlikely)event that multiple large local gridsappear along all four of its borders.

Regional grids are designed to overlap to a limitedextent to ensure complete coverage of a given geo-graphic territory; their dependent localgrids alsomayoverlap, in populous regions substantially,both withthemselves and other regional grids. This over-tiling,which is unusual and avoided in other georeferenc-ing frameworks, does not cause problems for WGRSbecauseeach place isreferencedto a singleautonomousgrid, either a regional grid or a local grid. Where acandidate place for a local grid falls in two or moreregional grids, it is activated in only one of them.

Local grids serve to co-relate groups of places thatoccur within the ambit of a regional grid, for example,the continuum of populated places that often appears

)'

in a large metropolitan area. The localities ofAlexandria, VA,and Bethesda, MD, are identifi-able sibling LGsofthe Washington, D.C., (WAS)regional grid.Also,Washington National Airport(DCA) is a single-celllocal grid, defined as coinci-dent withthe WAS.44cell. Then US.DC.DCA.55.12.34and US.DC.WAS.44.12.34 are equivalent WGLs, ornumeric aliases, for the same spatial location.Aliases accommodate local conventions andpreferences. The administrators of DCA mayviewthe airport's service area radially from thatcenter, while airline passengers may relate tothe larger place, Washington, D.C. Residentsof Alexandria, VA,and Bethesda, MD, are freeto identifYwith those places in preference toeither the national capital or its downtownairport.Viewed from afar, these distinctions are oftenimmaterial: foreign tourists likelywill considerand georeference all four places together withinthe single ambit of the WAS regional grid.

Grid SelectionThe WGRS grids are uniformly 1OOx100 kmin extent. In populous regions, the anchorpoints for a set of grids should be chosencarefully in order to minimize the regionalover-tiling of grids, when viewed at the statelevel. In sparsely settled areas, the opposite

circumstance arises: the set of well-known namedcultural and physical features may be insufficientto define a covering set of grids, in which casearbitrary benchmarks may be chosen for anchorsas convenient. On the open ocean, and elsewhereif desired, a graticular mapping framework basedon fixed increments of latitude and longitude alsomay be used to define systematic anchors for aregular quilt of grids.

Given the global nature ofWGRS, and the plethoraof place names globally, an automated method forselecting both regional and local grids and assigningplace name abbreviation codes to them is required.One such method, demonstrated to be workable forthe United States, follows. First, a list of the larg-est populated places within the territory of interest(here California, i.e., US.CA) is extracted from theGNIS database.6 This database contains over 168,100populated places with coordinates in the U.S.; how-ever, only about 22, 100 carry current (1990 or later)population estimates. From these, duplicate and vari-ant names (e.g., City of Los Angeles vs. Los Angeles;San Buenaventura vs.Ventura) are removed, leavingabout 21,750 places, which are then sorted by state

6 The Geographic Names Information System (GNISI is maintained by the U.S. Geological Survey for the Board on Geographic Names, athttp://mapping.usgs.gov/www/gnis/.

Vol. 29, No.4 359

Dow

nloa

ded

by [

Nor

thea

ster

n U

nive

rsity

] at

09:

02 0

6 O

ctob

er 2

014

Figure 4. Automatically selected WGRS grids for US.CA (68 grids, 86percent state coverage).

and population to generate a prioritized listof regional grid candidates. These candidatesare thinned in an iterative algorithm thatrepeatedly: I) selects, as a regional grid, thelargest remaining place, based on population(or other priority status); then 2) removesfrom the list both this place and all otherplaces that fall within some tolerance of thenewly-selectedgrid perimeter. The algorithmruns until the list is exhausted. As the toler-ance is increased, both the number of gridsgenerated and their overlap increase.

California, having an area of approxi-mately 406,000 square kilometers, wouldrequire a minimum of 41 regional gridswith zero overlap, if that were possible. Infact, the top 38 grids for California selectedby the above algorithm, with 20 km overlap(i.e., excluding candidates within a central60x60 km area of each other) covered -69percent of the land area of the state. WithIII grids similarly selected, but only 2 kmoverlap, -88 percent of the land area wascovered. Byexperimentation, a 5-km overlapwas found to be near optimal, producing aselection of68 grids with -86 percent cover-age. The remaining 14 percent of the landarea occurred in a mixture of small sliversand some large blocks, particularly in thesparsely settled eastern portion of the State (Figure4). Using local knowledge of California, 20 moregrids were added, and 10 were moved or renamed.Place identity is often a matter of local knowledge(of history, landmarks, etc.) in preference to theformal census and municipal place names derivedfrom gazetteers. The final result for California wasa set of 88 regional grids that completely coveredthe State (Figure 5).

An identical treatment for New York, again usinga 5-km overlap, required the addition of only onegrid to the 36 regional grids auto-generated by thealgorithm. Connecticut was completely tiled intosix regional grids by the algorithm, without needfor editing.

DiscussionThe World Geographic Reference System isdesigned to support the convergence of geospatialand telecommunications technologies in con-sumer devices such as GPS receivers, in-vehiclenavigation :systems, location-aware cellular tele-phones, and many expected new products. Suchlocation-based services (LBS) devices necessarilytransact spatial references across a human-com-

360

puter interface via geocodes (NCGIA 2001). rIbdate, geocoding has been largely hidden frompublic view, although geocodes do appear onGPS receivers and often are printed crypticallyalong the edges of topographic maps. Increasingly,however, geocodes and digital maps are becomingconsumer products, on the Weband in burgeoningwireless devices, notwithstanding the fact that thecapabilities of LBSdevices and services far exceedthe sophistication of most potential users. Therapidly increasing number of interactive maps onthe World Wide Web-an obviouslydated estimatecites over a million maps a day (Crampton 1998)-is forcing public awareness of geographical science,including geocoding.

For all LBSapplications, simplicity and clarity ofgeocodes are critical, both because of the hand-heldand mobile devices' limited displays and keyboards,and also becauseofusers' limited attention spans andgeographic training. It seems unlikely that the gridsystems designed for cartographers, surveyors, andgeodesists can easilyaccommodate the casualuser ofgeospatial technologies, or even should. Locationalchores with LBSdevicesare often approximate, local-ized, time-sensitive, and typicallymap-free. Spatialdata that come from GPS and GIS, navigationalaids, directional signage in the environment, even

CmtograPhy and GeograPhic Information Science

Dow

nloa

ded

by [

Nor

thea

ster

n U

nive

rsity

] at

09:

02 0

6 O

ctob

er 2

014

Figure 5. Manually adjusted WGRS grids for US.CA (88 grids, 100 per-cent state coverage).

maps-all need to be quickly integrated in rela-tion to nearby cultural or physical features. Concise,mnemonic geocodes, such as WGRS provides, cansimplifYthis integration.

For example, a driver has a minor accident in ashopping mall parking lot and calls the local policedepartment to report it. The driver's cellular tele-phone with on-board GPS could present specificgeographic coordinates in one of numerous formats,all of which are error-prone to recite; alternatively,the driver could report the location as "in the park-ing lot in front of the supermarket at the FairviewShopping Center," which lacksspecificity.Bycontrast,the WGRS geocode, US.DC.WAS.12.34.56.78 (possiblyfurther shortened to WAS.12.34.56.78), is precise andsuccinct, with some valuable error checking ofboth the name string and the digit-pairs built intothe syntax. A mangled geocode would be imme-diately recognizable as such by both humans andmachines. If truncated at any point, but still wellformed, this geocode would correctly narrow thearea of search.

Another point of consumer contact with geospatialdata is,of course, the WorldWideWeb.Curiously,mostWebsearches are "spaceless:" a search isjust as likelyto return results from the other side of the country

Vol. 29, No.4

(or the world) as nearby. Yet, if geographymakes a difference, then near things shouldbe more related to each other than to thosemore distant. Most searches for goods andservices seek nearby results; achieving thisrequires network-based spatial constraintsor at least as-the-crow-f1ies distance rank-ings for the search. A WGRS geocode, likeUS.DC.WAS.55, could easily be used to focussearches geographically, especially given aWeb-mapping interface. Gazetteers based onassociationsof sub-grid-scalenames-CapitolHill, Foggy Bottom, the Mall-with WGRSequivalents also can be imagined, such asUS.DC. WAS.CapitolHill.

At this time, WGRSis a proposal for a com-munity-licensed, public standard followingthe examples of the Internet and the Web.ForWGRSto be workable, the georeferencingmechanisms and notations outlined in thispaper need to be reviewed, perhaps revised,and eventually ratified by a consortium ofacademic, business,and government interests.A single enterprise designated by the consor-tium must conduct the standards "exercise"of defining and maintaining the regionaland local grids on a worldwide basis, withassistance by local experts as necessary. Theresulting database of grids and associatedsoftware to access it also must be publishedin numerous venues-hardcopy and digital

media and, of course, the wired and wirelessWebs-forconvenient access by all.

Implementation of WGRS thus amounts to thecreation and maintenance of a globally accessiblegazetteer, not only of the WGRS grids, but alsopublic and private place names, abbreviations,populations (and undoubtedly much other spatialdata and information). This is a very large databaseundertaking, which also must be accomplished probono, probably in the private sector. The currency,privacy, and security of the data will certainly beissues; equality and equity of access to the data alsowill need to be considered.

ConclusionThis paper has outlined the conceptual back-ground and technical specifications of a newWorld Geographic Reference System. This systemoffers a computer- and user-friendly mechanismfor specifYing location, both in the real world andin facsimiles of it, i.e., in GPS, GIS, LBS and Webapplications generally. The fundamental unit ofthe system is a set of culturally anchored mapping

361

Dow

nloa

ded

by [

Nor

thea

ster

n U

nive

rsity

] at

09:

02 0

6 O

ctob

er 2

014

•

grids, which fit well with human spatial cognition.In contrast to mapping frameworks such as MGRSand UTM, which tile space arbitrarily, WGRS tiesgrids to centers of human activity, precisely wherethe need for mapping is greatest. The WGRS ref-erences are always areal, at some scale, explicitlyrecognizing the approximation inherent in geore-ferencing.

We see WGRS as a vital link between the theoreti-cal and the practical aspects of georeferencing forthe Internet age. We have used the title "A NewWorld Geographic Reference System" deliberately,since there have been many other such grid systemproposals, even one that has used the same name(cf. ASCE/ACSM/ASPRS 1994). Like a single mapprojection, a single grid system cannot solve allcartographic problems, of course. Nevertheless, ifease of use is chosen first among the factors cited byGoodchild (1994), then WGRS may facilitate a bridgebetween basic geography, location-based services andthe World Wide Web. Specifically, we look forward toa day when LBS appliances and Web pages embedexplicitly spatial WGRS references directly in HTML(or similar) textual code, for example:

<a gref = "wgrp:/ruS.DC.WAS.54.18.28.83.11">The Washington Monument </a>.

With such a geographical extension, many new andexciting personal mobility and Web applications willbecome possible.

ACKNOWLEDGMENTSS. Lee Hancock, a co-inventor on all the WGRS-related patents, contributed to this paper innumerous ways. The authors also thank ValNoronha of the University of California SantaBarbara and three anonymous reviewers.

REFERENCESASCE/ACSM/ASPRS. ]994. Glossary of the maPPing sci-

ences. ACSM/ASPRS, Bethesda, Maryland.Clarke, KC. 1995. Analytical and computer cartograPhy.

Englewood Cliffs, New Jersey: Prentice Hall.Clarke, KC. 2002. Criteria and measures for the compari-

son of global geocoding systems. In: M.F. Goodchild andAJ. Kimerling (eds), Discrete Global Grids, chapter 8. [WebBook at http://www.ncgia.ucsb.edu!globalgrids-book!].

362

Crampton, J.w. 1998. The convergence of spatial tech-nologies. CartograPhic Perspectives 30 (Spring) [http://www.nacis.org/cp/cp30/cp_essay.html].

Dana, P.H., and S.L. Hancock. 1997. A user friendly lo-cation reference system. In: Proceedings of the NationalTechnical Meeting, Santa Monica, California. The Insti-tute of Navigation, Washington, D.C.

Dutton, G.H. 1997. A hierarchical coordinate system forgeoprocessing and cartography: Working through thescales. Ph.D. Dissertation, University of Zurich. Hei-delberg, Germany: Springer-Verlag.

Goodchild, M.F. 1994. Criteria for evaluation of globalgrid models for environmental monitoring and analy-sis. Handout from NCGIA Initiative 15. See SpatialAnalysis on the Sphere: A Review,by Rob Raskin, NC-GIA Technical Report 94-7.

Goodchild, M.F., and J. Proctor. 1997. Scale in a digitalgeographic world. GeograPhical and EnvironmentalModeling 1(1): 5-23.

Hancock, S.L., and P.H.Dana. 1998.U.S. Patent 5,839,088,issued November 17, 1998: "Geographic Location Ref-erencing System and Method;" further developed inU.S. Patents6,047,236,6,202,023,6,339,744,6,356,834issued April 4, 2000, March 13, 2001, January 15,2002,and March 12, 2002, respectively.

Kimerling, AJ., K Sah, D. White, and L. Song. 1999.Comparing geometrical properties of global grids.Cartography and Geographic Infonnation Systems 26(4):271-88.

NCGIA. 2001. Specialist meeting on location-based ser-vices. Center for Spatially Integrated Social Science,December 14-15, 200 I. [http://csiss.ncgia.ucsb.edu!events/meetings/location-based/].

Raper, J.F., D.W.Rhind, and J.w. Shepherd. 1992. Post-codes: The new geography. New York, New York: Long-man Scientific and Technical.

Rosenberg, D.K. 2000. The open source software licens-ing page.[http://www.stromian.com/Open_Source_Licensing.htm] .

Snyder, J.p. 1987. Map projections-A working manual.U.S. Geological Survey Professional Paper 1395. Washing-ton, D.C.: U.S. Government Printing Office.

Tobler,W.R.,and Z. Chen. 1986. A quadtree for global in-formation storage. GeograPhicalAnalysis 18(4):360-71.

World Wide Web Consortium. 2002. Naming andaddressing: URIs, URLs, [http://www.w3.orglAddressing/; last revised 2002/07/09].

CartograPhy and GeograPhic Infonnation Science

Dow

nloa

ded

by [

Nor

thea

ster

n U

nive

rsity

] at

09:

02 0

6 O

ctob

er 2

014