Framing Guidelines for Multi-Scale Map Design Using Databases at Multiple Resolutions
Cynthia A. Brewer and Barbara P. Buttenfield
Abstract: This paper extends European research on generating cartographic base maps at multiple scales from a single detailed database. Similar to the European work, we emphasize reference maps. In contrast to the Europeans’ focus on data generalization, we emphasize changes to the map display, including symbol design or symbol modification. We also work from databases compiled at multiple resolutions rather than constructing all representations from a single high-resolution database. We report results demonstrating how symbol change combined with selection and elimination of subsets of features can produce maps through almost the entire range of map scales spanning 1:10,000 to 1:5,000,000. We demonstrate a method of establishing specific map display scales at which symbol modification should be imposed. We present a prototype decision tool called ScaleMaster that can guide multi-scale map design across a small or large range of data resolutions, display scales, and map purposes.
Introduction • At what point(s) in a multi-scale progression will data geometry be modified to suit smaller scale presentation; and apmaking at scales between those • At what point(s) in a multi-scale progression that correspond to available database must new data compilations be introduced. resolutions can involve changes to From a data processing standpoint, symbol modi- Mthe display (such as symbol changes), changes to fication is often less intensive than geometry modi- feature geometry (data generalization), or both. fication, and thus changing symbols can reduce These modifications are referred to (respec- overall workloads for the map designer. Figure 1 tively) as cartographic generalization and statisti- shows a simplified example of a few coastal features cal or model generalization by Brassel and Weibel to provide graphical distinctions between display (1988) and later by Weibel (1995). Changes to and geometry modification. The large map (Figure display and geometry are in some instances co- 1a) is shown at 20 percent of original size, with dependent, meaning that modifications of one modifications to suit presentation at this scale in type can impact subsequent changes to the other. maps b1 to e1 in Figure1. Maps b2 to e2 are enlarged We argue that in previous literature, changes to versions of these 20-percent maps showing more geometry have overshadowed display change for clearly the quality of the different solutions. Map mapping at multiple scales; and in some situa- 1b shows no modification in symbols or geometry. tions, this increases the mapmaking workload. In contrast, map 1e shows modifications to both, Cartographic managers intending to balance or with less change in symbols than in map 1c and less optimize the workload are thus faced with a vari- change in geometry than in map 1d. For example, ety of decisions, including: roads are reduced less in width in 1e than b and • Which compiled-data resolution will be used displaced less in e than d. Overall these smaller for maps at any given display scale; changes maintain greater detail in the data and • At what point(s) in a multi-scale progression require lower magnitude adjustments, reducing will map symbols be changed; overall workloads. The European community has emphasized the Cynthia A. Brewer, Associate Professor, Department of need to develop efficient workloads for scale chang- Geography, 302 Walker Building, Pennsylvania State University, ing and for generalization operations (for example, University Park, PA 16802. E-mail:
Cartography and Geographic Information Science, Vol. 34, No. 1, 2007, pp. 3-15 Figure 1. Schematized example showing geometry and display modifications applied for reduction to smaller-scale map.
data generalization, especially when the target topology, and other properties, thus most algo- mapping scales show levels of detail that are similar rithms are designed to preserve one or more of to the resolution of compiled data. We demon- these characteristics (Cromley and Campbell strate a method of establishing specific display 1990; Buttenfield 2002). The guiding principles scales at which display modification should be driving this body of work are to reduce visual imposed. We also prototype a decision tool called clutter while preserving clarity and logical con- ScaleMaster for multi-scale map design across a sistency. We review the work on generalization to small or large range of data resolutions, display show that the solutions offered to preserve clar- scales, and map purposes. ity and reduce clutter through changes in symbol design and selection (Brewer 1995; Slocum et al. 1995) have been largely ignored in the multi-res- Background olution mapping literature. Published reports that describe and evaluate and Related Literature algorithms for cartographic generalization date The process of modifying details in spatial data back forty years at least, with early work focused has been a subject of longstanding interest in on simplistic coordinate reduction of linear fea- many disciplines, judging by publications in tures and polygon boundaries. Subsequently, the Cartography, GIScience, and Computer Science emphasis shifted to “… address the interdepen- literature. Data reduction through generaliza- dent generalization of related features (i.e., the tion can effect varying and sometimes dramatic elements of a topographic map)” (Weibel 1991, p. changes in line length, local coordinate density, 172). In the past decade, generalization work has
Cartography and Geographic Information Science formalized knowledge-based rules (Kilpelainen and Sester 2004; Bobzien et al. 2005), and other 1997), context-sensitive algorithms (Jones et al. interactive cartographic applications. 1995; Burghardt and Cecconi 2003), data modeling In multi-resolution work, symbolization is typi- (Buttenfield 1995; Ruas and Lagrange 1995), and cally pre-established, perhaps based on the national software agents (Galanda and Weibel 2002). It is look of a topographic series. Thus adjustment of important to note that the majority of recent work symbols is not always discussed as a generaliza- in cartographic generalization has taken place in tion operation. However, these authors do discuss Europe. Chronologic overviews can be found in it. Bédard and Bernier (2002) explicitly involve Brassel and Weibel (1988), Buttenfield and McMaster variation in visual variables in their discussion of (1991), McMaster and Shea (1992), Meng (1997), multiple representations and their proposed VUEL Weibel and Dutton (1999), and Mackaness et al. (view element) concept for managing geometric, forthcoming). semantic, and graphic multiplicities. Rayson (1999) Another thread of relevant literature has focused proposes a method of aggregating point symbols on database support for multi-scale data manage- into stacks for military map display while zooming. ment and flexible zooming through many scales Torun et al. (2000) discuss selection decisions for and for many map purposes (see, for example, web mapping at scales roughly between 1:15K and Jones et al. 2000). Many European authors work 1:80K. They develop hierarchies of feature codes and within contexts of a national mapping agency eliminate progressively larger numbers of feature producing databases containing linked multiple code groups at smaller map scales. Their classing- representations of detailed, high-resolution objects. based method sets thresholds at which groups of These databases are referred to as multi-resolu- features are removed from a map display, with class tion or multi-representation databases (MRDB) visibility becoming more exclusive in displays at (Kilpelainen 1997; Spaccapietra et al. 2000). smaller scales. They acknowledge the interaction Updates to these detailed objects autonomously between generalization constraints (e.g., graphical propagate to all resolutions, reducing the work limits of output media and reading conditions) involved in updating maps (Stoter et al. 2004; and symbol choices for feature classes. Bobzien et al. 2005). Objects are linked through Arnold and Wright (2005) focus on the inter- all resolutions to accomplish efficient updates, action between feature updates and production so the approach emphasizes object-level rather of multiple cartographic representations from a than scale-level representations. Some researchers single detailed database. They emphasize that approach object-level generalization and update mapmakers may have varied purposes (such as using representation stamps (Balley et al. 2004). “The producing tourist, road atlas, or electoral maps) as concept of multiple representations includes both they move from raw data to high-quality finished the various representations stored in the database presentations. They contrast these more custom- and also those that can be derived from existing ized mapping goals to mapping that conforms to a representations” (Kilpelainen 1997, p.14), either standard design for a predefined scale range. They by display change (symbol redesign) or geometry acknowledge the roles of four types of cartographic change (generalization of detail). processing—selection, elimination, symbolization, Managing decisions to create consistently high- and reclassification—that other authors tend to quality products must be balanced against data de-emphasize or pass over in discussion of gen- processing workloads. This is especially important eralization. They explicitly include the influence for on-demand and on-the-fly multi-scale map- of symbolization in their workflow diagrams; they ping for the web (Torun et al. 2000; Cecconi and position symbolization as an influencing factor Gallanda 2002; Cecconi et al. 2002), mobile devices that affects cartographic and analytical represen- (Harrie et al. 2002; Hampe et al. 2004), and in- tations for products derived from a multi-scale car navigation (Li and Ho 2004; Ulugtekin et al. multi-purpose database (for example, see Figure 2004), in addition to multi-scale database manage- 14 in Arnold and Wright, p. 415). ment (Timpf 1998; Jones et al. 2000; Zlatanova Cecconi et al. (2002) position symbolization at et al. 2004). Related innovations include adaptive the beginning of a generalization process. For zooming (Cecconi and Gallanda 2002; Harrie et example, in road network mapping, symbolization al. 2002; Hampe et al. 2004), online progressive precedes all other processes: selection, morphing, vector transmission (Buttenfield 1999; Paiva et weeding, smoothing, rendering, and displacement. al. 2004; Zhou et al. 2004; van Oosterom et al. These other generalization processes would be 2006), update propagation across multiple resolu- re-run if symbols are adjusted in response to an tions or representations (Dunkars 2004; Haunert unsatisfactorily generalized cartographic represen-
Vol. 34, No. 1 Figure 2. Example guidance on representation decisions through scale based on adjacent LoDs. [From Cecconi et al. (2002, their Figure 2, p. 5), reprinted with permission.]
tation. Cecconi et al. (2002) propose methods for applied throughout the complete range of scales, adaptive zooming in web mapping that mix LoDs but typification and displacement are applied mostly (“Level of Detail” databases pre-computed from through the middle range of scales (approximately existing databases) with on-the-fly generalization 1:50K to 1:300K). Because additional operators of Swiss topographic data. They explain how the are applied, the complexity curve arcs upward in two approaches are combined using diagrams, and this range. Similar diagrams are presented for examples of their figures are reprinted here. buildings/settlements, railroads, rivers, and lakes Figure 2 shows a representation at 1:100K, gen- and each has different operator limits and selected eralized from an LoD at 1:25K, which includes LoD scales. The authors implement symbolization individual buildings. For some data layers, the at the beginning of the mapping process, and they 1:25K LoD would serve as a source for mapping suggest adjusting to smaller symbols if the other down to 1:200K. However, the authors state that operations do not resolve conflicts. Swiss cartographic convention breaks the represen- In the next section we report on an exercise that tation of individual buildings at 1:150K. Building helped us think about how display change (i.e., representations at scales smaller than 1:150K Brassel and Weibel’s 1988 “cartographic general- would be prepared from a 1:200K LoD showing ization”) may obviate the need for some types of homogeneously colored urban areas (rather than geometry change. We said earlier that many types buildings). For some types of data then, feature of display change will involve lower workloads than data are pulled from an LoD source compiled geometry change. We reiterate this argument using at a smaller scale than the demanded map scale an empirical exercise in which we only change (for example, a 1:150K building representation symbols and change feature selections to examine is derived from 1:200K LoD data), going against this aspect of workloads for multi-scale mapping. the cartographic convention of always deriving Our exercise uses existing U.S. databases for map- from larger compilation scales. We agree that this ping through a continuous range of scales. contravention will be necessary to implement map production across continuous ranges of scales from a small number of anchor databases; and The ScaleMaster Exercise we demonstrate it in empirical work described Cindy Brewer worked through a scale exercise later in the paper. with seven graduate students at the Pennsylvania Cecconi et al. (2002) expand this approach by State University in the fall of 2004, in collabo- delineating three scale ranges that summarize scale ration with Charlie Frye and Aileen Buckley at limits for appropriate processing. These bands ESRI. Charlie and Aileen provided USGS DLG are separated by vertical dashed lines through (Digital Line Graph) data sets for the Redlands, the diagram in Figure 3 which sketches levels CA, area at anchor scales of 1:24,000, 1:100,000, of generalization complexity for a road network and 1:2,000,000, along with style files for example. The preprocessed LoDs are shown as default topographic symbols. The graduate thick bars marking points along the complexity seminar project focused primarily on design curve. Selection and simplification operators are change across a range of display scales. The stu-
Cartography and Geographic Information Science Figure 3. Generalization complexity through scale with mapping derived from a selection of LoDs. Ranges of application of generalization operators are also shown through scale. [From Cecconi et al. (2002, their Figure 3a, p. 7), reprinted with permission.]
dents’ goal was to produce topographic general- with related symbol styles. For example, water reference maps that could be read on 19-inch body outlines and areas and stream lines were flat-panel screens, which were typically set at 72 represented with the same hue. More contrast was ppi. On-screen display dictates coarse resolution added if symbols were set too similar at smaller and thus constrains symbol design. Figures 4 to scales. For example, a solid gray line for a power 7 look coarse because they demonstrate what the line caused it to look too similar to road symbols maps looked like at 72 ppi as the students made and required adding a pattern to the power line. design decisions. They are shown in the Figures Point symbol sizes and line widths were reduced to at 50-percent reduction with no re-sampling, ease clutter among features. Area feature outlines retaining the pixilated onscreen appearance. were also removed, small features within a class To print them from higher-resolution exports were eliminated, and fewer feature classes were would not demonstrate the design challenge set selected to reduced clutter. A selection of mapped for the students. The students did not attempt examples illustrates the exercise (Figures 4 to 7) to include labels on these maps, which would and is discussed in more detail below. The purple also have affected symbol design. Other than box in these Figures marks the same extent on selection decisions, they did not augment the each map to highlight the amount of scale change designs with additional generalization opera- between these example displays. tions (such as simplification, displacement, and Students worked with their layers to decide collapse) that affect geometry, in order to focus scale ranges, selections, eliminations, and symbol the study on design (and complete the project in changes. They copied their layers into a master a half semester). ArcMap project. The group then gathered in class Each student studied a group of related features and critiqued the appearance of maps with layers (such as hydrography) across the three anchor contributed by all students viewed together in the scales and through continuous scale change for the master project. These were visual evaluations of ScaleMaster exercise. Each student’s map layers the maps (rather than computed evaluations). were then integrated into a reference map dis- Smaller scales often required omission of feature played at many scales. Students used multiple types when the display became too cluttered. For criteria as they made symbol modification and example, Steven Gardner used a wide red and feature selection and elimination decisions. They white dash with a black casing for state routes set examined the visual hierarchy among symbols and visible to 1:38K (Figure 4), replaced this symbol reduced size and contrast so that particular feature with a thinner version of the symbol from 1:38K types were not more prominent than suitable for to 1:50K (Figure 5), then changed to a single black a topographic general-reference map. Likewise, line at scales smaller than 1:100K (Figure 6), and features with related characteristics were designed omitted selected state routes at scales smaller than
Vol. 34, No. 1 Figure 4. A selection of symbol decisions within the ScaleMaster project at 1:15K. The Table of Contents (TOC) to the left has the road layers group expanded to show eight road_arcs layers, each associated with different symbol sets and selections used for the exercise.
Figure 5. A selection of symbol decisions within the ScaleMaster project at 1:40K. The road symbols for 38K-50K are shown in the TOC.
1:275K (Figure 7). The map subsets in Figures 4 and the group would evaluate map appearance to 7 also show changes in contour intervals and at each of these sampled scales. At the next class omission of building markers at smaller scales. meeting the group would sample a different set Map designs were examined by typing in pro- of scales through the 10K to 5M range. The class gressively smaller representative fractions for the criticized symbolization, commenting on symbols display from a range of 10K to 5M. For example, that overwhelmed others or were not sufficiently the instructor would type in a series of scales such visible, were too cluttered, were too coarsely pre- as 13K, 20K, 45K, 150K, 300K, 600K, 1M, and 4M sented, and so forth. Then the students each went
Cartography and Geographic Information Science Figure 6. A selection of symbol decisions within the ScaleMaster project at 1:90K. The road symbols for 50K-100K are shown in the TOC.
Figure 7. A selection of symbol decisions within the ScaleMaster project at 1:300K. The 1:300K map falls in a “scale gap” where DLG hydrography data were not suitable in the absence of generalization, so it is missing that layer.
back to editing their feature group by adjusting the group was designing for the entire range of visibility ranges, eliminating subsets of features, possible map scales. and adjusting symbols to integrate better through Once the cycles of symbol redesign and class all scales with the other students’ designs. Students’ critique were complete, the exercise had established decisions affected subsequent rounds of design appropriate symbol designs within ranges of map- checking, which prompted further adjustments, ping scales for each feature group (hydrography, until the group was satisfied with the overall design. transportation, and so forth). These specifications The focus was not on designing for the anchor formed the basis for creating the ScaleMaster tool scale and extrapolating from that design. Instead, (Figure 8). Feature groups are listed to the left,
Vol. 34, No. 1 t Studen Contributors: ania Lopez anuka Bhowmick Adrienne Gruver T Steve Gardner Sohyun Park Anthony Robinson T 5M scale; described below: g 2M
f 1M r erent intermediate contour symbols ff 50% brown lines (lighter) l. di m. 20m interval, 60% brown (lighter) n. change to 40m interval, 30% brown (lighter) o. change to 100m interval p. change to 400m interval q. 20% brown lines (lighter) r. 500K gaps mean data not suitable e
1:2,000,000 250K p q
a c "ScaleMaster" j 100K i n o h. multi-track RR to single track i. RR hatches longer and closer j. RR at regular USGS specification k. larger symbols, many feature types culled (e.g., bridge abutments) 50K d d 1:100,000 m d d h 24K ferent USGS data sets. Lightness changes represent selection or design adjustments to suit change smaller l k b 10K Hues represent dif a. select main rivers b. eliminate 'other' hydro polygons and set washes visible c. lighten lake outline d. class 1 & 2 roads set wider e. interstate line set narrower f. lines set all black with interstates wider g. all lines set thin and black 1:24,000 ransportation T . ScaleMaster diagram prepared to describe symbol behavior for feature classes from three DLG databases for maps ranging from 1:10K to 1:5M. Maps were general- Feature type: Hydrographic arcs Hydrographic polygons Highways Surface streets Railroads Misc. Park & forest polygons City polygons Urban area & county State Point features Contours (from DEM) reference topographic-style maps without labels and viewed onscreen using desktop computer displays at 72 ppi. Figure 8
10 Cartography and Geographic Information Science and each is associated with student names on the notes a, b, and k in Figure 8). Thus selection and right. elimination were used along with symbol design The final product each student delivered was a for the map solutions, and we consider both to set of ArcMap layers with symbol styles and visibil- be display modifications. We acknowledge that ity ranges specified, and a sketch of a horizontal some cartographers consider selection and elimina- slice through the ScaleMaster diagram. Diagram tion to be types of generalization. Our discussion compilation was organized in a fairly low-tech has emphasized the distinction between display manner. The instructor handed out graph paper modifications and generalization methods that with scales marked on a logarithmic axis. Each modify feature geometry. Feature geometry is not student then drew the ranges through which each changed when removing entire feature classes or dataset was useful and marked key design-change subclasses from the display. Thus, we consider break points. These individual lines across strips selection and elimination of features to also be a of the graph paper were then compiled into the design decision. composite ScaleMaster diagram. The format for As an aside, the term “ScaleMaster” grew from this diagram is patterned after a diagram Charlie in-class conversations and procedures. The instruc- Frye presented during a research planning talk tor named the master GIS project, to which each on multi-scale basemap data models in May 2003 student contributed layers, ScaleMaster, and at ESRI, and which developed into a core visual that filename grew to be the label for the whole example for our research discussions. exercise. Also, the class worked on the exercise Three hues are used for horizontal bars across immediately before the 2004 U.S. Presidential the ScaleMaster diagram, each indicating a dif- election and was watching the Votemaster web ferent anchor scale: blues for 1:24K, purples for site (electoral-vote.com) which synthesized poll 1:100K, and greens for 1:2M. The lightness series results and displayed predictions on a generalized for each anchor scale marks scale ranges for par- cartogram. Thus mastering maps was on their ticular symbol design settings. General descrip- minds. It is interesting to note that in an ecology tions of symbol changes are noted with letters at modeling problem, Lilburne et al. (2004) present breaks along these bars. For example, at about a framework called “Scale Matcher” for selecting 1:150K, letters a and c indicate that main rivers data scales to suit ecology problems. were selected from the 1:100K data and the color of lake outlines was lightened. Some bars stop with no further representations at smaller scales Discussion—The Consequences to indicate that students felt the feature should be eliminated to reduce clutter. For example, city of Scale Change polygons and railroad lines are no longer repre- The mapping process for the ScaleMaster exer- sented on maps past 1:300K. cise differed from designing a map at each anchor In contrast to the gaps seen for hydrography scale or at a particular intermediate scale. The and point features, the students pulled administra- emphasis was first on how each group of features tive boundaries “up-scale” from the 1:2M dataset. “behaves” through all scales and how sensitive The straight lines and smooth curves of railroads the features are to scale change. The student also allowed students to use them at much larger group systematically examined how the symbol- scales than their 1:2M source would suggest. The ized features in one layer worked together with green bars for railroads are far displaced from the other layers to produce maps at multiple scales. 1:2M position on the diagram because rail lines This was an iterative process that concluded with were not included on the smaller scale maps, but draft map designs for all scales and an overall the selection of lines offered in the 1:2M data was understanding of how differently the feature suitable through the 1:30K to 1:250K range. The groups behaved during scale change. ScaleMaster decisions were made in the context Behavior, in this context, means that symbol of a student exercise without strong requirements appearance and data geometry change across a for map accuracy and completeness, but many range of map scales. Visual expectations of the mapmaking purposes share a reduced need for Earth vary as one moves closer (Morrison and planimetric precision, especially for on-demand Morrison 1994), reflecting earth and atmospheric web contexts. processes as they become detectable at specific In addition to symbol design changes, students resolutions. Cartographers are trained to design selected feature classes and removed some sub- maps such that some processes appear more classes for the mapping exercise (for example, see prominently when creating maps at a given scale.
Vol. 34, No. 1 11 For example, Buttenfield et al. (1991) invento- data to be useful. For example, the 1:24K data ried European topographic maps to determine for hydrographic features and roads have fairly changes in symbol appearance and priorities constrained ranges that extend up to about 1:50K. for transportation, hydrography, and settlement In contrast, the 1:24K parks, forests, and city features across a range of large (1:10K) to small boundaries are useful on maps well past 1:200K. (1:250K) scales. These authors determined that Running vertically through the ScaleMaster stack terrain and natural processes are more prominent at various scales suggests that a mapmaker may be on larger scale maps, while urban limits and trans- pulling from multiple datasets to build a map at a portation networks attain higher prominence on custom scale. For example, the 1:70K vertical line medium (1:100K) and smaller scale topographic intersects with parks, forests, and city polygons maps. This also appears to be the logic underlying from the 1:24K data; hydrography, roads, point USGS topographic mapping symbols (Thompson features, and contours from 1:100K data; and 1988). railroads from 1:2M. Furthermore, data sensitivities to scale are theme Another interesting aspect of the ScaleMaster based. Mark (1991) argues that across a range outcome is a data gap across the ScaleMaster stack, of map scales, naturally occurring features (for roughly between 1:250K and 1:700K. Horizontal example, terrain data and hydrography) will change bars for some features do not cover this range more often in geometry and symbology than will because they are not ordinarily shown on refer- cultural features (roads, land use). “More often” ence maps at these scales (for example, surface means that a larger number of critical breakpoints streets, railroads, and miscellaneous transportation). or thresholds can be defined along the progression Hydrography, however, should be shown, and the of scale, where changes must be made either to gaps in the ScaleMaster stack indicate that data symbols or geometry (or to both). In part, this is in the anchor compilations (1:24K, 1:100K, and due to the nature of cultural features; for example, 1:2M) are not suitable within the range when map roads are built to a fixed radius of curvature, based design proceeds by symbol modification. This is on the turning radius of automobiles. However, not a criticism of the National Mapping Agency it is also the case that certain naturally occurring that produced these data, rather it indicates that surface processes (such as vegetation) will change hydrography data require geometry change in less often across scales simply because they are addition to symbol redesign, or, alternatively, that dependent on, and surficial to, underlying coarser maps in the range of 1:250k to 1:700K based on resolution processes (for example, soil type and DLG anchor data would benefit from inclusion moisture, water table depth, or local climate). of an intermediate anchor database. Buttenfield Essentially, the resolution at which change in the and Hultgren (2005) undertook this exercise, dependent process becomes evident is a function embedding 1:250K VMAP data into this same of the resolution of the causative processes. DLG dataset, and integrating the two thesauri. In this context, we can reflect on the utility of A similar exercise was undertaken in Poland to ScaleMaster as a tool to guide symbol redesign fuse cartographic thesauri compiled at 1:50K for multi-scale mapping. It is important to note and 1:250K (Gotlib and Olszewski 2006). Those that the patterns in this ScaleMaster outcome are authors report similar findings on discrepancies specific to DLG data and to the three anchor scales and redundancies between feature codes, and the (1:24K, 1:100K and 1:2M). They are also specific impact of military versus civilian agency missions to modifying symbology and making feature selec- on semantics of the fused databases. tions and eliminations only (i.e., not to modifying A third interesting aspect relates to scale sensitivi- feature geometry). A ScaleMaster tool built for ties of the DLG data depicted in this ScaleMaster other multi-scale anchor databases compiled by tool. Conventional wisdom is that anchor data can another National Mapping Agency would likely be suitably used for mapping across a range of differ, as would a ScaleMaster tool constructed for scales roughly two to five times the anchor. Thus, types of maps other than reference base maps. It is 1:24K data could be reduced to 1:100K, 1:100K difficult to predict which of these conditions would data could be reduced to 1:400K, and 1:2M data have the greatest impact, the anchor databases, could be reduced well beyond 1:5M. Following scale changing operators, or the map purpose. this logic, the 1:24K anchor data should be most Nonetheless, we offer a few initial reflections on sensitive to scale change, requiring the most fre- the ScaleMaster tool and on the exercise. quent symbol design breakpoints, and behaving One thought-provoking aspect of ScaleMaster is suitably across the shortest scale range. Looking the widely varied ranges at which students found at the ScaleMaster, one can see that this is not
12 Cartography and Geographic Information Science the case. Scale ranges for the 1:24K anchor data ences between ScaleMaster diagrams produced extend across a scale range of about 2x (40K), and for maps with different purposes from a larger these anchor data required few display changes database. beyond selection at the smallest scales. We interpret Formalizing and parameterizing the decisions this to mean that avoiding geometry changes will that trained cartographers make about changing severely limit the usable range of mapping scales map scale effectively would benefit organizations for 1:24K DLG data. The 1:2M anchor data should seeking to maintain consistency of multi-scale be least sensitive and suitable for mapping across design outcomes. This approach would guide the longest scale range with fewest breakpoints. In systematic production of a minimum number of the exercise this tends to be the case, with 1:2M LoDs derived from larger scale data and identifi- data layers suitable across a 10x scale range (500K cation of additional intermediate-scale data sets to 5M). It carries the most display revision break- suited to an organization’s mapping needs. The points, which are in every case symbol redesign approach also benefits organizations lacking a (not selection), implying that symbol redesign is staff cartographer, and it may lead to an improved adequate to “push” the display suitability for this quality of map products. smallest scale anchor database. The ScaleMaster concept informs decisions The scale sensitivity of the 1:100K anchor data that support effective map production workloads. should lie somewhere in between the large- and Incorporating symbol change into generalization small-scale anchors but, curiously, it does not. This activities extends the range of mappable scales for range of suitable mapping scales extends only a given database, which can reduce the volume to 1:150K, except for point features and urban of data to be maintained in-house and reduce area boundaries, at which selection and symbol data processing workloads for map production. redesign are mandated. Moreover, data must be The goal is to transform cartographic data across pulled from both other anchor resolutions to fill scale ranges and only change symbols and data as in content on landcover and some transportation much as is needed to preserve visual quality and features. In these respects, the 1:100K anchor data still meet the map purpose, while maintaining appear to be the most sensitive to scale change, the smallest-volume cartographic database. This contrary to cartographic intuition. approach generates an effective and efficient map- ping workflow that can support all the map products Summary created and distributed by an organization. We justify our argument that symbol modifica- ACKNOWLEDGEMENTS tions can offset the need for geometry change The graduate students who worked together and extend the usable range of reference base- to create ScaleMaster in Cynthia Brewer’s mapping scales (for a given set of anchor data- Cartography Seminar at Penn State were Tanuka bases) by means of an empiric exercise mapping Bhowmick, Steven Gardner, Adrienne Gruver, USGS DLG data across scales 1:10K to 1:5M. Tania del Mar López Marrero, Sohyun Park, Our results support earlier reverse engineering Anthony Robinson, and Stephen Weaver. Charlie of cartographic basemap symbols (Buttenfield Frye and Aileen Buckley at ESRI supported the et al. 1991; Paranteinen 1995; Edwardes and project with planning, discussion, data, and Mackaness 2000) showing that various carto- symbol files. The authors gratefully acknowl- graphic layers behave differently with scale edge funding by ESRI (Professional Services change. We report results demonstrating how Agreements 2003B6345 and 2003B6346). An symbol change and selection/elimination can initial version of this paper was published in the produce maps through much of this scale range. proceedings of the AutoCarto 2006 conference We present a decision tool called ScaleMaster (Brewer and Buttenfield 2006). that can be constructed for multi-scale map design across a small or large range of data REFERENCES resolutions, display scales, and map purposes. Arnold, L.M., and G.L. Wright. 2005. Analysing product-specific behaviour to support process The ScaleMaster tool that we introduce dem- dependent updates in a dynamic spatial updat- onstrates a method of establishing specific map ing model. Transactions in GIS 9(3): 397–419. display scales at which display modifications Balley, S., C. Parent, and S. Spaccapietra. 2004. should be imposed. In this case study, we exam- Modelling geographic data with multiple rep- ine a topographic basemap on-screen display. resentations. International Journal of Geographical Further work is in progress examining differ- Information Science 18(4): 327–52.
Vol. 34, No. 1 13 Bédard, Y., and E. Bernier. 2002. Supporting Cecconi, A., R. Weibel, and M. Barrault. 2002. multiple representations with spatial databases Improving automated generalization for on- views management and the concept of VUEL. demand Web mapping by multiscale databases. In: Proceedings of the ISPRS/ICA Workshop on In: Proceedings of the Symposium on Geospatial Multi-scale Representations of Spatial Data, Ottawa, Theory, Processing and Applications, ISPRS/IGU/ Canada. 14 pp. CIG, Ottawa, Canada. [http://www.isprs.org/com- Bobzien, M., D. Burghardt, and I. Petzold. 2005. mission4/proceedings02/pdfpapers/138.pdf]. Re-generalisation and construction—Two alter- Cromley, R.G., and G.M. Campbell. 1990. A geo- native approaches to automated incremental metrically efficient bandwidth line simplification updating in MRDB. In: Proceedings of the 22nd algorithm. In: Proceedings of the 4th International International Cartographic Conference, A Coruna, Symposium on Spatial Data Handling, Zurich, Spain. 8 pp. Switzerland. Vol. 1, pp. 77-84. Brewer, C.A. 2005. Designing better maps: A Guide Dunkars, M. 2004. Automated generalisation in a for GIS users. Redlands, California: ESRI Press. multiple representation database. Geoinformatics 203 pp. 2004, Proceedings of the 12th International Brewer, C.A., and B.P. Buttenfield. 2006. Mastering Conference on Geoinformatics, Gävle, Sweden. map scale: Formalizing guidelines for multi- pp. 741-8. scale map design. In: Proceedings of AutoCarto Edwardes, A., and W. Mackaness. 2000. Modelling 2006, Vancouver, Washington. Published by knowledge for automated generalization of cat- the Cartography and Geographic Information egorical maps—A constraint-based approach. In: Society, Gaithersburg, Maryland. 10 pp. P. Atkinson and D. Martin (eds), GIS and geocom- Brassel, K.E., and R. Weibel. 1988. A review and putation (Innovations in GIS 7). London, U.K.: conceptual framework of automated map gen- Taylor & Francis. pp. 161-73. eralization. International Journal of Geographic Galanda, M., and R. Weibel. 2002. An agent-based Information Systems 2(3): 229-44. framework for polygonal subdivision general- Burghardt, D., and A. Cecconi. 2003. Mesh sim- ization. In: D. Richardson and P. van Oosterom plification for building selection. In: Proceedings, (eds), Proceedings 10th International Symposium on 5th Workshop on Progress in Automated Map Spatial Data Handling. Berlin, Germany: Springer. Generalization, Paris, France. pp. 121-36. Buttenfield, B.P. 1995. Object-oriented map gen- Gotlib, D., and R. Olszewski. 2006. Integration of eralization: Modeling and cartographic consid- the topographical database, VMap L2 database erations. In: J.C. Muller, J.P. Lagrange, and R. and selected cadastral data—A step towards the Weibel (eds), GIS and generalization: methodology integrated, MRDB reference database in Poland. and practice. London, U.K.: Taylor and Francis. In: Proceedings, ICA Commission Workshop pp. 91-105. on Generalization and Multiple Representation, Buttenfield, B.P. 1999. Progressive transmission Vancouver Washington. [http://ica.ign.fr/ of vector data on the Internet: A cartographic Portland/paper/ICA2006-Gotlib_Olszewski.pdf]. solution. In: Proceedings, 19th ICA/ACI Conference, Hampe, M., M. Sester, and L. Harrie. 2004. Ottawa, Canada. pp. 581-90. Multiple representation databases to support Buttenfield, B.P. 2002. Transmitting vector visualisation on mobile devices. International geospatial data across the Internet progressively. Archives of Photogrammetry, Remote Sensing and In: Proceedings, Second International GIScience Spatial Information Sciences 35:6. Conference (GIScience 2002), Boulder Colorado. Harrie, L., T. Sarjakoski, and L. Lehto. 2002. A pp. 51-64. mapping function for variable-scale maps in Buttenfield, B.P., and T. Hultgren. 2005. Managing small-display cartography. Journal of Geospatial multiple representations of “base carto” features: Engineering 4(2): 111-23. A data modeling approach. In: Proceedings, AUTO- Haunert, J-H., and M. Sester. 2004. Using the CARTO 2005, Las Vegas, Nevada. straight skeleton for generalisation in a multiple Buttenfield, B.P., and R.B. McMaster (eds). 1991. representation environment. In: Proceedings Map generalization: Making rules for knowledge rep- of the ICA Workshop on Generalisation and resentation. London, U.K.: Longman. 245 pp. Multiple Representation, Leicester, U.K. 10 pp. Buttenfield, B.P., C.R. Weber, M. Leitner, J.J. Jones, C.B., A.I. Abdelmoty, M.E. Lonergan, P.M. Phelan, D.M. Rasmussen, and G.R. Wright. Van Der Poorten, and S. Zhou. 2000. Multi-scale 1991. How does a cartographic object behave? spatial database design for online generalisation. Computer inventory of topographic maps. In: In: Proceedings, International Symposium on Spatial Proceedings, GIS/LIS’91, Atlanta, Georgia. Vol. 2, Data Handling, Beijing, PRC. pp. 891-900. Jones, C.B., G.L. Bundy, and J.M. Ware. 1995. Map Cecconi, A., and M. Galanda. 2002, Adaptive generalization with a triangulated data structure. zooming in Web cartography. Computer Graphics Cartography and Geographic Information Science Forum 21(4): 787–99. 22(4): 317-31.
14 Cartography and Geographic Information Science Kilpelainen, T. 1997 Multiple representation and Spaccapietra, S., C. Parent, and C. Vangenot. generalization of geo-databases for topographic maps. 2000. GIS databases: From multiscale to multi- Ph.D. Dissertation and Technical Publication of representation. In: B.Y. Choueiry and T. Walsh the Finnish Geodetic Institute No. 124, Helsinki (eds), SARA 2000, LNAI 1864. Berlin, Germany: University of Technology, Finland. Springer-Verlag. pp. 57-70. Li, Z., and A. Ho. 2004. Design of multi-scale and Stoter, J.E., M.J. Kraak, and R.A. Knippers. 2004. dynamic maps for land vehicle navigation. The Generalization of framework data: A research Cartographic Journal 41(3): 265-70. agenda. In: Proceedings of the ICA Workshop Lilburne, L.R., T.H. Webb, and G.L. Benwell. 2004. on Generalisation and Multiple Representation, The Scale Matcher: A procedure for assessing Leicester, UK. 14p. scale compatibility of spatial data and models. Thompson, M.M. 1988. Maps for America (3rd ed.). International Journal of Geographical Information U.S. Geological Survey, Washington, D.C. Science 18(3): 257-79. Timpf, S. 1998. Map Cube Model—A model for Mackaness, W., A. Ruas, and T.L. Sarjakoski. (forth- multi-scale data. In: Proceedings 8th International coming). Generalization of geographic information: Symposium on Spatial Data Handling (SDH’98), Cartographic modelling and applications. London, Vancouver, Canada. pp. 190-201. U.K.: Elsevier. Torun, A., B. Köbben, and R. Lemmens. 2000. Mark, D.M. 1991. Object modelling and phenom- Processing spatial data on the Internet. enon-based generalization. In: B.P. Buttenfield International Archives of Photogrammetry and Remote and R.B. McMaster (eds), Map generalization: Sensing 33(B6): 269-78. [http://kartoweb.itc.nl/ Making rules for knowledge representation. London, kobben/publications/ISPRSamsterdam.pdf]. U.K.: Longman. pp. 103-18. Ulugtekin, N.N., A.O. Dogru, and R.C. Thomson. Meng, L.Q. 1997. Automatic generalization of geo- 2004. Modelling urban road networks inte- graphic data. Technical Report of the Institute of grating multiple representations of complex Cartography University of Hannover, Germany. 76 road and junction structures. In: Geoinformatics pp. 2004, Proceedings of the 12th International McMaster, R.B., and S.K. Shea. 1992. Conference on Geoinformatics, Gävle, Sweden. Generalization in digital cartography. Monograph pp. 757-64. Series. Association of American Geographers, van Oosterom, P., M. de Vries, and M. Meijers. Washington D.C. 2006. Vario-scale data server in a Web service Morrison, P., and P. Morrison, 1994, Powers of ten context. In: Proceedings, ICA Workshop on (revised). San Francisco, California: W.H. Freeman. Generalization and Multiple Representation, Paiva, A.C., E. Rodrigues da Silva, F.L. Leite Jr., and C. Vancouver, Washington. [http://ica.ign.fr/ de Souza Baptista. 2004. A multiresolution approach Portland/paper/ICA2006-VanOosterom2.pdf]. for Internet GIS applications. In: Proceedings of the Weibel, R. 1991 Amplified intelligence and rule- 15th International Workshop on Database and Expert based systems. In: Buttenfield, B.P., and R.B. Systems Applications (DEXA’04). 5p. McMaster (eds), Map generalization: Making Parantainen, L. 1995. Knowledge acquisition rules for knowledge representation. London, U.K.: by comparison of map series: The generaliza- Longman. pp. 172-86. tion of forest parcels on the Swiss national map Weibel, R. 1995 Three Essential Building Blocks series. Unpublished MSc thesis, Department of for Automated Generalization. In: GIS and Geography, University of Zurich. Synopsis in: R. Generaliaation: Methodology and Practice. London: Weibel. 1995. Three essential building blocks for Taylor and Francis: Muller, J.C., Lagrange, J.P. automated generalization. In: J.C. Muller, J.P. and Weibel, R. (eds.): 56-69. Lagrange, and R. Weibel (eds), GIS and gener- Weibel, R., and G. Dutton. 1999. Generalising alization: methodology and practice. London, U.K.: spatial data and dealing with multiple represen- Taylor and Francis. pp. 56-69. tations. In: P.A. Longley, M.F. Goodchild, D.J. Rayson, J.K. 1999. Aggregate towers: Scale sensi- Maguire and D.W. Rhind (eds), Geographical tive visualization and decluttering of geospatial information systems: Principles, techniques, manage- data. In: Proceedings of the IEEE Symposium on ment and applications. London, U.K.: Longman. Information Visualization. pp. 92-99, 149. pp. 125-55. Ruas, A., and J.P. Lagrange. 1995. Data and knowl- Zhou, X., S. Prasher, S. Sun, and K. Xu. 2004. edge modeling for generalization. In: J.C. Muller, Multiresolution spatial databases: Making web- J.P. Lagrange and R. Weibel (eds), GIS and gener- based spatial applications faster. Lecture Notes in alization: Methodology and practice. London, U.K.: Computer Science 3007: 36-47. Taylor and Francis. pp. 73-90. Zlatanova, S., J.E. Stoter, and W. Quak. 2004. Slocum, T.A., R.B. McMaster, F.C. Kessler, H.H. Management of multiple representations in spa- Howard. 2005. Thematic cartography and geo- tial DBMSs. In: Proceedings of the 7th AGILE graphic visualization (2nd ed.). Upper Saddle River, Conference on Geographic Information Science, New Jersey: Prentice Hall. 518p. Heraklion, Greece. pp. 269-278.
Vol. 34, No. 1 15