lessons learned in cartographic data modeling -...

1

UC 2007 Tech SessionsUC 2007 Tech Sessions 11

Lessons Learned in Cartographic Data Lessons Learned in Cartographic Data ModelingModeling

Aileen BuckleyAileen BuckleyCharlie FryeCharlie Frye

2


Collaborators Collaborators –– university facultyuniversity faculty

University facultyUniversity faculty•• Cindy Brewer, Penn State University (PSU)Cindy Brewer, Penn State University (PSU)•• Barbara (babs) Buttenfield, University of ColoradoBarbara (babs) Buttenfield, University of Colorado--

Boulder (CU)Boulder (CU)•• Robert McMaster, University of MinnesotaRobert McMaster, University of Minnesota

To start, we need to let you know that this work was done in collaboration with a number of university faculty, including:

Cindy Brewer of Penn State UniversityBarbara (babs) Buttenfield of the University of Colorado-BoulderAnd more recentlyRobert McMaster of the University of Minnesota

3


StudentsStudents

•• Graduate Research Assistants/InternsGraduate Research Assistants/Interns–– CU: Torrin Hultgren, CU: Torrin Hultgren, Kiyoshi YamashitaKiyoshi Yamashita, Jochen , Jochen WendelWendel–– PSU: Jessica Acosta, Mamata AkellaPSU: Jessica Acosta, Mamata Akella

•• Undergrad Assistants/InternsUndergrad Assistants/Interns–– Anna Harman (CU), Zachary Anna Harman (CU), Zachary TardivoTardivo (PSU) (PSU)

•• Classes (PSU)Classes (PSU)– Cartography Seminar, PSU graduate students, Fall 2005:

• Tanuka Bhowmick, Steve Gardner, Adrienne Gruver, Tania del Mar López Marerro, Sohyun Park, Anthony Robinson, Steve Weaver

– Applied Cartographic Design, PSU undergrads, Spring 2006:• Steve Cline, Rosie Daley, Emily Dux, Derek Foll, Katie Holmes, Allen

Huber, Alex Krishchyunas, Adam Naito, Joe Palchak, Ed Petronsky, Ryan Stoyek, Matt Sudac, Jessica Acosta (grad)

There were also a significant number of students involved – the ones shown here in yellow have gone on to work for us here at ESRI.

4


Regularly research meetingsRegularly research meetings

Clint, Aileen, babs, Thierry, Charlie

May 2003 in Redlands

Photos by Cindy Brewer unless otherwise noted

We met regularly starting in Spring of 2003 in Redlands.

5


December 2006 in Redlands

Kiyoshi, babs, Charlie, Aileen, Clint, Jess

Our most recent meeting in Redlands was in December of last year.

6


March 2007at Penn State

babs, Jess, Aileen, Charlie

Then this spring, we met at Pennsylvania State University.

7


AgendaAgenda

•• Define what we mean by a cartographic data modelDefine what we mean by a cartographic data model•• Describe the advantages we have found to using themDescribe the advantages we have found to using them•• Seven Lessons LearnedSeven Lessons Learned

For this presentation, we’ll start by defining what we mean by a cartographic data model, then we’ll describe the advantages we have found to using them, and we’ll end with our “Seven Lessons Learned”.

Effective use of GIS for map production requires a well designedcartographic data model. Cartographic data models are a formalization of the map’s design stored in the features and attributes of a GIS database. Cartographic data modeling requires a clear understanding of the maps to be produced and, in addition, the software used to produce the maps. Arranging information about the geographic features, attributes and map making processes in a systematic form (i.e., a data model) allows it to be shared and repurposed for other map making uses.

8


What is a What is a ““cartographic data modelcartographic data model””??

•• ““A cartographic data model is the codification of the A cartographic data model is the codification of the geographic features, attributes and processes that geographic features, attributes and processes that produce a desired cartographic product or products produce a desired cartographic product or products through specified software.through specified software.”” (Buckley and Frye, 2007)(Buckley and Frye, 2007)

A cartographic data model is the codification of the geographic features, attributes and processes that produce a desired cartographic product or products through specified software. Data models can be articulated in a variety of ways, through books of specifications, UML diagrams, organizational charts and more.

9


Required knowledgeRequired knowledge

•• The maps to be producedThe maps to be produced•• The software to be used to make the mapsThe software to be used to make the maps

Cartographic Data ModelCartographic Data Model

But cartographic data models are best articulated through the maps that are made from them.

So, cartographic data models are a formalization of the map’s design stored in the features and attributes of a GIS database. Cartographic data modeling requires a clear understanding of the maps to be produced and, in addition, the software used to produce the maps. Arranging information about the geographic features, attributes and map making processes in a systematic form (i.e., a data model) allows it to be shared and repurposed for other map making uses.

10


Advantages of cartographic data modelingAdvantages of cartographic data modeling

•• Analytical thinking about map design and the map Analytical thinking about map design and the map production processproduction process–– The map drives the software, not vice versaThe map drives the software, not vice versa–– The maps and the software drive the production processThe maps and the software drive the production process

•• Influence the primary data capture and initial database Influence the primary data capture and initial database compilation compilation –– The maps influence the data captureThe maps influence the data capture

•• Influence subsequent updates to the dataInfluence subsequent updates to the data

There are many advantages to data modeling. One is that it requires us to think analytically about map design and the map making process. In many cases, GIS users and novice map makers allow the software to drive the map design rather than harnessing the power of the software to effectively and efficiently execute a pre-conceived design (Buckley and Hardy, 2007; Buckley and Frye, 2005).

11


Example: Data model & primary data captureExample: Data model & primary data capture

For example, maps often have names for physiographic features such as mountain ranges and valleys, but these features are not normally captured in geographic databases. For cartographic purposes only, it is useful to have polygonal outlines of the areas in which these names will appear (Buckley and Frye, 2006). Knowing that this is a mapping requirement at the outset allows it to be incorporated into the initial database design and allowances can be made for capturing these data.

12


Advantages of cartographic data modelingAdvantages of cartographic data modeling

•• Analytical thinking about map design and the map making Analytical thinking about map design and the map making processprocess–– The map drives the softwareThe map drives the software–– The maps and the software drive the production processThe maps and the software drive the production process

•• Influence the primary data capture and initial database Influence the primary data capture and initial database compilation compilation –– The maps influence the data captureThe maps influence the data capture

•• Influence subsequent updates to the dataInfluence subsequent updates to the data•• Allow the map products to be supportedAllow the map products to be supported

at the outset rather than an afterthoughtat the outset rather than an afterthought

And lastly, cartographic data models includes any modifications to the database that allow the software to handle or process the data appropriately and efficiently in order to make the maps. And one of the most important things I think it allows is for the maps to be incorporated into the design from the outset, rather than being considered as an afterthought.

13


Map making techniquesMap making techniquesWhat next?

Iterative and inexactIterative and inexactnature of map makingnature of map making

Lack of historicLack of historiccodification of taskscodification of tasks

Difficulty translating itDifficulty translating itto a computing environmentto a computing environment

Incomplete knowledge Incomplete knowledge of the dataof the data

Incomplete knowledge Incomplete knowledge of the softwareof the software

Perhaps the hardest thing to embody in the data model are the techniques that the map maker uses to produce the final map design and product. In some cases, it is possible to translate the cartographer’s thinking directly into GIS data and sequences of software functions; other times the requirements are more elusive because it is difficult to formalize how a cartographer completes certain tasks. There are a number of reasons for this elusiveness, including

the iterative and inexact nature of some map making tasks,the lack of attention historically given to codification of some map making tasks, difficulties in translating the tasks to their expression in a digital environment, and incomplete knowledge of the data……and/or software required to complete the tasks.

So it’s not easy to inform in the data model with all the techniques that the map maker uses to produce the map design and product.

14


The Data Modeling ProcessThe Data Modeling Process

The overall data modeling process starts with the desired map product, which is then translated into first the basic themes of information to be shown on the map or maps. The themes are then broken down into the building blocks for the maps and subsequently become the basic elements of the data model (Frye, 2006). The model is then modified to take into account any special requirements for data handling and processing by the software. For example, polygons might have to be converted to lines so that a special labeling effect can be achieved. The robustness of the model is then checked – redundant or missing data are identified, and opportunities for data substitution are seized.

15


Seven Lessons LearnedSeven Lessons Learned

1.1. The maps drive the data modelThe maps drive the data model2.2. A complete inventory of the cartographic features is A complete inventory of the cartographic features is

requiredrequired3.3. Semantic models do not have to be exhaustive or even Semantic models do not have to be exhaustive or even

geographic in order to be completegeographic in order to be complete4.4. The data model is informed by the softwareThe data model is informed by the software5.5. MultiMulti--purpose, multipurpose, multi--scale databases are possiblescale databases are possible6.6. The cartographic data model informs the primary data The cartographic data model informs the primary data

compilationcompilation7.7. There is a There is a ““connectednessconnectedness”” of information between of information between

scalesscales

Now let’s start in with our “Seven Lessons Learned”.

16


Lesson 1: The maps drive the data modelLesson 1: The maps drive the data model

•• Start with the maps you make or would like to makeStart with the maps you make or would like to make•• Paper copies or digital displays are the best embodiment Paper copies or digital displays are the best embodiment

of your mapsof your maps’’ designsdesigns–– New designs take more time and effortNew designs take more time and effort–– Most cartographersMost cartographers’’ designs are seeded by the maps theydesigns are seeded by the maps they’’ve seen ve seen

or madeor made

•• Everything on the maps needs to be accounted for in the Everything on the maps needs to be accounted for in the databasedatabase–– Each different kind of graphic markEach different kind of graphic mark–– A featureA feature’’s graphic mark includes its symbol and/or labels graphic mark includes its symbol and/or label

Lesson 1: The maps drive the data modelThe map is essentially a reflection of the data (information about the geographic environment stored in a recognizable form) and the tools that are used to manipulate the data (software functions and human interventions that refine the outputs) (Hardy and Kressman, 2005; Frye, 2006). Designing a data model to make maps without a specific product or products in mind can be an endless and ultimately fruitless task. Without the final map product as a litmus test, it is impossible to determine if the model is accurate and complete. Cartographic data bases need not be designed for single map products; multi-scale, multi-purpose databases are more desirable as they can be repurposed for the variety of products an organization is responsible for (Brewer and Buttenfield, 2006). The data should be modeled to suit the purpose and requirements of all the standard map and analytical products as well as to support efficient management of the products and production workflows (Frye, 2006). A fallacy about cartographic data models is that they should be able to support any map or analytical product that could possibly be imagined. This is not only impractical, it also leads to superfluous content or organization of the data model that may impact its efficiency for producing the products that are identifiable and required.

17


Lesson 1: The maps drive the data modelLesson 1: The maps drive the data model

•• If a model, tool, or script is needed in the mapping If a model, tool, or script is needed in the mapping workflow, the database must contain all the input workflow, the database must contain all the input parametersparameters

•• Your GIS data may have more classes or richness than Your GIS data may have more classes or richness than your map requiresyour map requires–– Create a crossCreate a cross--walk that relates your GIS data to your mapping walk that relates your GIS data to your mapping

requirementsrequirements

•• End with your mapsEnd with your maps–– The maps you produce match the maps that drove the design of The maps you produce match the maps that drove the design of

your datayour data

Related to this first lesson are a few caveats:If a model, tool, or script is needed in the mapping workflow, the database must contain all the input parameters.Your GIS data may have more classes or richness than your map requires – if it does, you can create a cross-walk that relates your GIS data to your mapping requirements.End with your maps – that is, the maps you produce should match the maps that drove the design of your data.

18


Example: Marine Water BodiesExample: Marine Water Bodies

•• Oceans, Gulfs, Bays, etc. are essential information on Oceans, Gulfs, Bays, etc. are essential information on many mapsmany maps–– However, their locations are not typically in GIS datasetsHowever, their locations are not typically in GIS datasets–– Until nowUntil now……http://mappingcenter.esri.comhttp://mappingcenter.esri.com ArcGIS ResourcesArcGIS Resources

•• Named Marine Water Bodies datasetNamed Marine Water Bodies dataset

•• Data modelData model–– Shape Shape –– PolygonsPolygons–– Label_SizeLabel_Size attribute attribute –– to to

set label text sizeset label text size–– Label_TypeLabel_Type attribute attribute –– to to

assign Maplex placement assign Maplex placement rulesrules

An example we can show you for this first lesson is Marine Water Bodies:

Oceans, Gulfs, Bays, etc. are essential information on many maps; however, their locations are not typically in GIS datasets -- until now… you can actually find them on our new Mapping Center Web site at http://mappingcenter.esri.com under ArcGIS Resources. There you will find the Named Marine Water Bodies dataset.

The data model for this dataset includes:Shape – PolygonsLabel_Size attribute – to set label text sizeLabel_Type attribute – to assign the Maplex placement rules

19


Lesson 2: A complete inventory of the cartographic Lesson 2: A complete inventory of the cartographic features is requiredfeatures is required

•• Each kind of hydrographic Each kind of hydrographic point symbol (graphic point symbol (graphic mark) is listedmark) is listed

•• Each kind of map it Each kind of map it appears on is shadedappears on is shaded

Lesson 2: A complete inventory of the map features is requiredA complete inventory of the features on the maps is required to inform many aspects of the data model (Buckley, 2004). This includes the graphic marks (points, lines, polygons) that are symbolized to represent geographic features, all labels, and any rasters or images. Any uniquely symbolized or labeled map feature is a primary element in the data model, because each must be assigned a specific symbol and/or label. Without the ability to categorize the data by their symbological requirements (including labeling), it is impossible to handle the appearance of each unique map feature. Often, these unique map features can be identified by inspecting the map products as well as any specifications for symbology and labeling that an organization may have. For example, we developed an inventory of all the unique map features on U.S. Geological Survey USGS) topographic sheet series at scales from 1:24,000 to 1:2,000,000 through inspection of the maps and reference to the map compilation and data capture specifications (USGS, 2005).

This inventory allowed us a clear understanding of the types of features that needed to be in the GIS database and the rules required to symbolize and label them. It also allowed us to easily add or drop map features, and to group or separate them based on data handling requirements. This in essence becomes the semantic model for the map.

20


Example: USGSExample: USGS

The semantic model is then translated into the cartographic data model by identifying the geographic features that are used to create the map features and endowing those features with the appropriate geometry and attribution to allow the software to render them appropriately. For example, a city may appear as a polygon at larger scales, a piece of text at a smaller scale, and a point at the smallest scales. Knowing that the geometry, symbology and labeling changes across scales allows three (in this case) representations of the single entity to be explicitly defined the data model. If an organization makes multiple map products, each of the products needs to be inventoried. Of course the data model should be flexible and extensible enough to absorb and reflect modifications to the map content or design at any time.

But cartographic data models are perhaps articulated best in the map products that are created from them.

For example, we examined the documentation for the USGS 1:24,000 scale topographic map series. This allowed us to find out what features and text would be shown on these maps.

21



We examined the documentation for how features like springs were captured from source data – this gave us an idea of how the USGS defined a spring and it’s relationship to other features, like streams. Springs, for example, that are the source of a stream must be located at the endpoint of the stream.

22



But these specs were written before GIS was used for map making, so what they really said is the that center of the circle used to SYMBOLIZE the spring was to be located at the endpoint of the line that was used to SYMBOLIZE the stream.

23



The investigation of these specs allowed us know exactly what features and text would be shown on the 24,000 scale maps. Then we did the same thing for the rest of the maps in the series, at scales of 63,360, 100,000, 25,000 and 1,000,000 to 2,000,000. At the same time we learned about any changes in their appearance on the map, including labeling.

24


Cartographic featuresCartographic features

•• Uniquely symbolized and/or labeled graphic marks Uniquely symbolized and/or labeled graphic marks (points, lines, polygon, text) on the map(points, lines, polygon, text) on the map

This might seem tedious, but the first step in data modeling is to determine the primary elements or what we can call “cartographic features” rather than the “geographic features”.

For cartographic data models, the primary elements of the model are any uniquely symbolized and/or labeled points, lines, polygon, or piece of text on the map. Cartography call these primary elements graphic marks.

The data model includes specification of all geographic features that will appear on the maps, and their symbology and labeling.

25


Lesson 2: A complete inventory of the Lesson 2: A complete inventory of the cartographic features is requiredcartographic features is required

•• All map products reviewedAll map products reviewed•• ““Cartographic featuresCartographic features””•• Complete inventory across map productsComplete inventory across map products

So to review, Lesson 2 required that all map products be reviewed in order to identify each unique “Cartographic feature” – this resulted in a complete inventory across map products.

26


Lesson 3: Semantic models do not have to be Lesson 3: Semantic models do not have to be exhaustive or even geographic in order to be completeexhaustive or even geographic in order to be complete

•• Case 1: Physiographic FeaturesCase 1: Physiographic Features–– The initial data modelThe initial data model

•• Shape Shape –– PolygonPolygon•• Feature_typeFeature_type attribute attribute –– Type of LandformType of Landform

–– We identified over 30 kinds of landforms labeled on USGS We identified over 30 kinds of landforms labeled on USGS Topographic mapsTopographic maps

–– NoneNone of it was useful for labeling these featuresof it was useful for labeling these features–– What was important?What was important?

•• Size Size –– to set font sizeto set font size•• Shape (long, round, skinny, etc.) Shape (long, round, skinny, etc.) –– to assign Maplex placement to assign Maplex placement

rulesrules

–– The final data modelThe final data model•• Shape Shape –– PolygonPolygon•• Feature_SizeFeature_Size Attribute Attribute –– set by Python Script Toolset by Python Script Tool•• Shape_TypeShape_Type Attribute Attribute –– set by Python Script Toolset by Python Script Tool

Lesson 3: Semantic models do not have to be exhaustive or even geographic in order to be complete The feature geometry for physiographic features are typically not drawn on maps; instead their label is placed in a fashion that implies the practical spatial limits of these features. So the graphic mark is just a label. We initially assumed the label placement could be driven by feature type, as this had been the case with most other kinds of features. We tried this and it didn’t help. Instead we found that the label’s placement was best informed by following the trend of the shape, when an obvious trend was visually suggested in the shape, and if not, then a horizontal label could be placed, either inside, on, or outside the shape depending on its size. We also found that the USGS maps labeled larger features with larger type sizes, so experimented and found a reliable way to assign size classes. We took what we knew about shape types and size classes and devised a Python Script that automated the assignment of this information.

Semantic models do not have to be exhaustive in order to be completeFor this lesson, we return to the case study described for Figure 1, in which we defined polygonal extents that could be used at multiple scales for the labels of physiographic features. At the beginning of this exercise, we again inventoried the various types of physiographic features that appeared on the maps that the data model was designed to support (Buckley and Frye, 2006; Frye and Buckley, 2006). Our initial inventory included the physiographic features in Table 1.

We then realized that we had inventoried the geographic features rather than the map features. Many of these features types would receive the same labeling treatment on the map; therefore, it was not required that they be unique. None of the features would be symbolized using a graphic mark, or polygon, and the label for the feature along with the terrain depiction through hillshading and hypsometric tinting would constitute the mapped representation. We decided to reduce the data model to simply point and polygon physiographic features, since they clearly had different labeling requirements.

27


On this portion of a map, you can see the geographic features that we created to hold the text for physiographic features – the polygons are now shown on the final map, but their labels are.

28


Lesson 3: Semantic models do not have to be Lesson 3: Semantic models do not have to be exhaustive or even geographic in order to be completeexhaustive or even geographic in order to be complete

•• Case 2: Swamps, Bogs, Fens, Marshes, etc.Case 2: Swamps, Bogs, Fens, Marshes, etc.–– Regional terms Regional terms –– all are used in national mapping systemsall are used in national mapping systems–– Most locales, only need one or twoMost locales, only need one or two–– Use the Use the Domain to Table toolDomain to Table tool to make edits to our domains and to to make edits to our domains and to

make your ownmake your own–– Store your custom model using the Store your custom model using the Table to Domain toolTable to Domain tool in in youryour

geodatabasegeodatabase

Much of what we’ve published in terms of “model data” include attribute domains, which you can convert to a table (Domain to Table tool), edit—add or remove rows, change values, etc. For instance swamp, bog, fen, marsh, are not used consistently in all places in the U.S. and don’t need to be in everybody’s data model. In fact, on our example maps, they all get the same symbol.

29


Lesson 4: The data model is informed by the Lesson 4: The data model is informed by the softwaresoftware

•• Maplex Labeling ExtensionMaplex Labeling Extension•• Spatial Analyst ExtensionSpatial Analyst Extension•• RepresentationsRepresentations•• EditorEditor

•• Additional knowledge of these areas of ArcGIS gives you Additional knowledge of these areas of ArcGIS gives you the power to define mapping workflows that arethe power to define mapping workflows that are–– EfficientEfficient–– AutomatedAutomated–– RobustRobust

•• Each area has tools with potentially special requirements Each area has tools with potentially special requirements for data input for data input –– that that mustmust be addressed in your data be addressed in your data modelmodel

Lesson 4: The data model is informed by the software These aren’t the only areas of ArcGIS that are important, but the idea is the more you know, the more you can help yourself. If you don’t know that a particular task can be done in the software, you may opt to use a manual workflow, which takes more time and money; buy third party software to do the job, costing more money; or worse not do the job at all.

The data model is tailored to the softwareFollowing on the previous lesson, we continued to work with physiographic features and considered that the labeling options for ArcGIS allowed us to refine the label placement for the features. However, to correctly identify the features that would have similar placement rules, we had to modify the data model to match the software requirements. The label engine for ArcGIS can use the shape and size of the feature to identify the optimal label placement. In order to take advantage of this software functionality, we enhanced the cartographic data model to contain attributes that explicitly defined the label requirements (Frye and Buckley, 2006; Frye, 2006). This resulted in the addition of the rotated minimum bounding rectangle (RMBR) for the feature, which was used to calculate the values for two added attributes: the length-to-width ratio of the RMBR and the percentage of the RMBR that was occupied by the feature (Figure 4). Using these two attributes, five shape type classes were identified that were used to drive refined label placement rules.

This is but one example of leveraging the database in order to maximize the software capabilities and identifying opportunities for automated map production. The primary requirement for being able to do this is to know what the software can do and how the data need to be modeled in order for them to be handled (Buckley, Richards and Barnes, 2006; Punt et al, 2006; Buckley and Hardy, 2007; Brewer and Frye, 2005). This can be a challenge for cartographers who have traditionally used illustration and graphics software and are nascent to GIS use. Even seasoned GIS users can find it challenging to stay informed about software enhancements and changes in functionality in more recently released versions.

30



•• Case 1: Contour linesCase 1: Contour lines

Intermediate Intermediate DepressionDepression

IndexIndex

Index DepressionIndex Depression

IntermediateIntermediate

Types of Contour LinesTypes of Contour Lines

31



•• Case 2: Shape types for Case 2: Shape types for labeling different types of labeling different types of shapes shapes

•• Require different label Require different label placement strategiesplacement strategies

•• Shape types are based Shape types are based onon–– The way a polygon is The way a polygon is

related to its rotated related to its rotated minimum bounding minimum bounding rectangle (rectangle (rMBRrMBR))

–– the ratio of the the ratio of the rMBRrMBR’’sslength to widthlength to width

O s s i p e e L a k e

O n e i d a L a k e

L a k e G as ton

Hartwell Lake

S

us

q u e h a n n a Ri v

er

ProngedPronged RoundRound--ishish

OblongOblong

SplotchySplotchy

Long orLong orSnakySnaky

32


Lesson 5: MultiLesson 5: Multi--purpose, multipurpose, multi--scale databases scale databases are possibleare possible

•• Limited scale range the source data can be used forLimited scale range the source data can be used for•• Symbology alone can extend this scale rangeSymbology alone can extend this scale range•• Scale ranges vary by feature typeScale ranges vary by feature type

–– Make different feature classes for these typesMake different feature classes for these types

•• Geoprocessing is required to further extend the scale Geoprocessing is required to further extend the scale rangesranges

•• Workload can be reduced through careful consideration of Workload can be reduced through careful consideration of the interplay between symbolization and geoprocessing the interplay between symbolization and geoprocessing strategiesstrategies

Lesson 5: Multi-purpose, multi-scale databases are possible•Limited scale range the source data can be used for•Symbology alone can extend this scale range•Scale ranges vary by feature type

•Make different feature classes for these types•Geoprocessing is required to further extend the scale ranges•Workload can be reduced through careful consideration of the interplay between symbolization and geoprocessing strategies

33


ScaleMasterScaleMaster

Initial workInitial work

34


ScaleMaster: Initial workScaleMaster: Initial work

•• Developed diagram formDeveloped diagram form

35


ScaleMaster: Initial workScaleMaster: Initial workSymbol change and eliminationSymbol change and elimination

36



MultiMulti--purpose mapping:purpose mapping:One complete database scale, one partialOne complete database scale, one partial

37


Road map exampleRoad map example

••ScaleMasterScaleMaster

20K5K

Change hue use to themes

38


Hydrography map exampleHydrography map example

••RosieRosie’’s ScaleMasters ScaleMaster

39


Hydrography map exampleHydrography map example

••ScaleMasterScaleMaster

40


Maps at 1:5K, 15K, 30K, 100K, 200K, 700KMaps at 1:5K, 15K, 30K, 100K, 200K, 700K

•• Research Assistant Jess AcostaResearch Assistant Jess Acosta’’s designss designs

41


Different map purposes with same base dataDifferent map purposes with same base data

TopoZoning

SoilsPopulation density

42


43


44


s s –– size change size change c c –– color change color change t t –– transparency change transparency change l l –– modify label appearance modify label appearance (e.g., bold, italic, character (e.g., bold, italic, character spacing, leading)spacing, leading)e e –– eliminate layer or eliminate layer or eliminate by feature type eliminate by feature type (e.g., eliminate intermittent (e.g., eliminate intermittent streams)streams)G G –– geometry change (e.g., geometry change (e.g., new data set or new layer new data set or new layer with generalized features)with generalized features)…… p, i, o, r, f, a, x, R p, i, o, r, f, a, x, R ……

45



Current work:Current work:MultiMulti--purpose, multipurpose, multi--scale mappingscale mapping

46


Understanding workload through scaleUnderstanding workload through scale

•• The most efficient The most efficient workflows are produced in workflows are produced in GIS by leveraging GIS by leveraging changes to both changes to both symbology (display) and symbology (display) and geometry (generalization)geometry (generalization)

•• In the graphs, there is In the graphs, there is less area under curve (c) less area under curve (c) than either (a) or (b)than either (a) or (b)

•• Introducing intermediate Introducing intermediate Level of Detail (Level of Detail (LoDLoD) ) datasets further reduces datasets further reduces workload, as seen in (d)workload, as seen in (d)

Brewer and Buttenfield, 2007, Brewer and Buttenfield, 2007, CAGISCAGIS

We have learned some things about balancing workloads through scale, including:•The most efficient workflows are produced in GIS by leveraging changes to both symbology (display) and geometry (generalization)•In the graphs, there is less area under curve (c) than either (a) or (b)•Introducing intermediate Level of Detail (LoD) datasets further reduces workload, as seen in (d)

47


Lesson 6: The cartographic data model informs the Lesson 6: The cartographic data model informs the primary data compilationprimary data compilation

•• Data capture rules can Data capture rules can be written to ensure be written to ensure successsuccess

•• Example Example –– generalizing generalizing hydrography datahydrography data–– Capture stream polygons Capture stream polygons

and centerlines and centerlines simultaneously or at least simultaneously or at least with one or the other with one or the other presentpresent

–– Makes it possible to Makes it possible to substitute center lines for substitute center lines for polygons in smaller scale polygons in smaller scale maps maps –– eliminating costlier eliminating costlier and less accurate and less accurate generalization workflowsgeneralization workflows

Lesson 6: The cartographic data model informs the primary data compilationKnowing the data and software requirements leads to informed recommendations for primary data capture (Frye, 2006). Rather than retrofitting the data to allow for cartographic presentation, it is more efficient to capture the geometry and attributes required for mapping at the outset. A case study that demonstrates this lesson involves the digitization or modification to digitization of hydrological features in the Boise, Idaho area.

In this study, the goal was to develop a cartographic database that would support the production of maps at multiple scales (Brewer and Buttenfield, 2007). This required digitizing the stream centerlines so that they could be substituted for the polygons at smaller scales. It also involved splitting the polygons where they would connect to the replaced centerlines. An example is the connection of the main river polygon with the centerline that was substituted for the complex pink polygons in the river bend at the right.

48









49









50


Feature Selection (Python scripting)Feature Selection (Python scripting)

•• Select on feature sizeSelect on feature size•• Modified by dimension (polygon or line feature)Modified by dimension (polygon or line feature)

Feature Selection (Python scripting)•Select on feature size•Modified by dimension (polygon or line feature)

51



•• Select on size and feature dimensionSelect on size and feature dimension•• Modified by feature shape (round, oval, long channels, etc.)Modified by feature shape (round, oval, long channels, etc.)

ShapesShapes

All

Round

Oblong

Long Lines

Feature Selection (Python scripting):•Select on size and feature dimension•Modified by feature shape (round, oval, long channels, etc.)

Here you can see all the features.

52



•• Select on size and feature dimensionSelect on size and feature dimension•• Modified by feature shape (round, oval, long channels, etc.)Modified by feature shape (round, oval, long channels, etc.)

ShapesShapes

All

Round

Oblong

Long Lines

Here you can se the different types of features selected and modified.

53



•• Select on feature size and dimension and shapeSelect on feature size and dimension and shape•• Modified by stream attributes (major/ minor, intermittent Modified by stream attributes (major/ minor, intermittent

stream, feature name, other)stream, feature name, other)

Feature Selection (Python scripting) •Select on feature size and dimension and shape•Modified by stream attributes (major/ minor, intermittent stream, feature name, other)

54


Feature SimplificationFeature Simplification

•• Test algorithm and threshold valuesTest algorithm and threshold values•• Automate the processAutomate the process

Feature Simplification:•Test algorithm and threshold values•Automate the process

55


Feature SimplificationFeature Simplification

•• Test an algorithm and threshold valuesTest an algorithm and threshold values•• Automate the processAutomate the process•• Modified by shape (natural or cultural hydro feature)Modified by shape (natural or cultural hydro feature)

We used Feature Implication to:•Test an algorithm and threshold values•Automate the process•Modify by shape (natural or cultural hydro feature)

56


Best Practices for Data CaptureBest Practices for Data Capture

•• Many data modeling problems might have been avoided Many data modeling problems might have been avoided by by ““informed data captureinformed data capture””–– Redundancy, i.e. capturing more than one feature representation Redundancy, i.e. capturing more than one feature representation –– Maintaining consistency Maintaining consistency

Others, too; but in the interests of timeOthers, too; but in the interests of time……

……graphic examplesgraphic examples

•• Collated these into a first draft of a Best Practices report Collated these into a first draft of a Best Practices report for for AdaAda County and other local governments who capture County and other local governments who capture data for mappingdata for mapping

This led us to learn some best practices for data capture:•Many data modeling problems might have been avoided by “informed data capture”, including:

•Redundancy, i.e. capturing more than one feature representation •Maintaining consistency •There are others, too; but in the interests of time we’ll show you just a couple of graphic examples

•We collated these into a first draft of a Best Practices report for Ada County and other local governments who capture data for mapping.

57


Best Practice Best Practice –– redundancyredundancy(capture more than one representation)(capture more than one representation)

• For every stream captured as a polygon, also capture a stream centerline.

• At smaller mapping scales, when polygonscoalesce, a linear map representation isavailable.

To avoid redundancy, these were the capture rules:•For every stream captured as a polygon, also capture a stream centerline.• At smaller mapping scales, when polygons coalesce, a linear maprepresentation is available.

58


Best Practice Best Practice –– maintain consistencymaintain consistency

• Every polygon should have a single centerline.

• Capture no more polygons than centerlines.

Too many centerlines, ignores stream channel width

Better solution.

To maintain consistency, these were the capture rules:•Every polygon should have a single centerline.Capture no more polygons than centerlines.

59


Lesson 7: There is a Lesson 7: There is a ““connectednessconnectedness”” of of information between scalesinformation between scales

•• Applies to maps and globes that are used at a wide range Applies to maps and globes that are used at a wide range of scalesof scales–– Such maps are actually a series of maps in sequence; and each Such maps are actually a series of maps in sequence; and each

map must be connected with its two neighborsmap must be connected with its two neighbors

•• The generalization 5x scale rule of thumb does not apply The generalization 5x scale rule of thumb does not apply or is not equivalent to how information on maps is or is not equivalent to how information on maps is integrated through scaleintegrated through scale–– 5x rule often applies well to any single dataset5x rule often applies well to any single dataset–– Maps include many datasets so a lowest common denominator is Maps include many datasets so a lowest common denominator is

needed for any given scaleneeded for any given scale

•• Names are Names are ““scalescale--lessless””

Lesson 7: There is a “connectedness” of information between scales•Applies to maps and globes that are used at a wide range of scales

•Such maps are actually a series of maps in sequence; and each map must be connected with its two neighbors

•The generalization 5x scale rule of thumb does not apply or is not equivalent to how information on maps is integrated through scale

•5x rule often applies well to any single dataset•Maps include many datasets so a lowest common denominator is needed for any given scale

•Names are “scale-less”

60


61


62


Our AdviceOur Advice

•• Data modeling assures that you can make the maps you Data modeling assures that you can make the maps you need to makeneed to make

•• Cartographic data models enhances GIS dataCartographic data models enhances GIS data•• Cartographic data modeling enhance all other data Cartographic data modeling enhance all other data

modelsmodels•• Cartographic data modeling should be undertaken for Cartographic data modeling should be undertaken for

every GIS database designevery GIS database design•• Your GIS data will most likely end up on some kind of Your GIS data will most likely end up on some kind of

map, even if itmap, even if it’’s not your maps not your map

Out advice to you includes:•Data modeling assures that you can make the maps you need to make•Cartographic data models enhances GIS data•Cartographic data modeling enhance all other data models•Cartographic data modeling should be undertaken for every GIS database design•Your GIS data will most likely end up on some kind of map, even if it’s not your map

63


Summary: Knowledge gained from our researchSummary: Knowledge gained from our research

•• Translate the semantic model to a data modelTranslate the semantic model to a data model•• Inform primary data captureInform primary data capture•• Symbology and labeling must be in the databaseSymbology and labeling must be in the database•• Leverage the database to maximize software capabilitiesLeverage the database to maximize software capabilities•• Identify opportunities for automationIdentify opportunities for automation•• Data models are not static; they evolveData models are not static; they evolve

So to end, we can summarize a few major points:•Translate the semantic model to a data model•Inform primary data capture•Symbology and labeling must be in the database•Leverage the database to maximize software capabilities•Identify opportunities for automation•Data models are not static; they evolve

64


Thank you!Thank you!

Please fill out your surveys!Please fill out your surveys!

Thanks for your attention and please fill out your surveys.

65


mappingcenter.esri.commappingcenter.esri.com

Don’t forget – you can soon find this presentation on the new Mapping Center Web site.

lessons learned in cartographic data modeling -...

Documents