THE NATURE OF GEOGRAPHIC DATA AND THE SELECTION OF THEMATIC MAP SYMBOLS PDF

den43823_ch04_063-079.indd Page 63 6/7/08 4:25:57 PM user-s175 /Volumes/109/MHDQ022/den THE NATURE OF GEOGRAPHIC 4 DATA AND THE SELECTION OF...

den43823_ch04_063-079.indd Page 63 6/7/08 4:25:57 PM user-s175 /Volumes/109/MHDQ022/den THE NATURE OF GEOGRAPHIC 4 DATA AND THE SELECTION OF THEMATIC MAP SYMBOLS CHAPTER PREVIEW The presentation of spatial components of the graphic system. Symbol place- data using thematic maps assumes an understand- ment positions the data in geographic space ing of both the nature of the data and the symbol- while the nature of the symbol provides communi- ization necessary for thematic map design. Spatial cation clues that permit the understanding of the data is described by (1) its characteristics of loca- data distribution. The nature of symbols must tion, form, and time, and (2) its level of measure- match the nature of the data. ment hierarchy. Characteristics of location tell us Before we begin the cartographic process, about the manner in which the data are distrib- we should ask ourselves why we create a map in uted, that is, do the data distributions exist as the first place. Normally, we have a reason to points, lines, or areas. Form describes the data generate a map. Such a reason may be to com- characteristics according to its inherent nature, municate graphically something that we know such as are the data qualitative or quantitative, (personal knowledge) to someone else (public discrete or continuous, totals or derived. The char- knowledge), the reader of the map. We create acteristics of the spatial data dictate the type of maps for many of the same reasons that we take thematic map used. Data measurement (nominal, photographs, send emails to friends, or post a ordinal, interval/ratio) categorizes the data ac- short movie on the Web. We want to show some- cording to a hierarchical structure. thing or tell something to someone else about Thematic maps are designed by matching the things we’ve seen or things we know, or we want nature of the data to map symbols. Thematic map to present a map of our area. We also create symbols use point, line, and area symbolization maps to visually display geographic phenomena. for displaying the spatial nature of the data. Such A map becomes a form of communication similar symbols are matched to visual variables that are to the written or spoken word. j Cartography, like language, use rules so that both the pre- is interpreted, for example, a broadleaf or needle-leaf tree, is senter and the receiver of information understand what is based upon the experiences of the individuals. Cartographi- being communicated. Both maps and language communi- cally we use symbols to represent a tree. Some of these cate using symbology. For example, the word “tree” is a symbols may resemble a tree, replicative symbols, on the term that we associate with a vegetative form. How the word landscape while others may be a more abstract form. Whether 63 den43823_ch04_063-079.indd Page 64 7/1/08 6:55:18 PM user-s175 /Volumes/202/MHDQ022/mhden6/den6c 64 PART 1 THEMATIC MAPPING ESSENTIALS or not communication occurs is dependent upon an under- nonspatial data, also referred to as attribute or characteristic standing of the symbology by both the cartographer and the data (Meyer 1997). That is, the data are independent of their map reader. geometric relationships. The maps we create from that data- Karl Sauer’s (1956) comment that maps are the language base display the information in a spatial context such that we of geography suggests a strong link between language, car- observe its distribution which permits us to visualize pat- tography, and geography (see Chapter 1). Thus, the world of terns and to draw inferences. The data, when mapped, be- maps allows for the display of information about the milieu, come spatial data which display characteristics of location, our physical environment, whether it be above, on, or below size, or amount. the Earth’s surface, and about our cultural environment, in- The remainder of this chapter will deal with the nature of cluding socio-economic, demographic, political, and other spatial quantitative data. We will also investigate the statistical human-induced activities. These maps may also pertain to characteristics of the data and how they assist us in the selec- statistical data that are collected about a specific location or tion of which thematic map to use. Presented too is a discus- region that can vary from a small area such as your backyard sion of the visual variables used in creating thematic maps. or a much larger area such as the entire Earth (see the discus- We conclude with an examination of how statistics are used sion of scale in Chapter 1). in data analysis. The subject matter and intent of the map will dictate the map type created. General purpose maps (see Chapter 1) in- clude those topics that display a planimetric base on which we THE NATURE OF DATA map the locations of cultural and physical features, for exam- ple, buildings, parks, lakes, forests, streets, or political bound- Thematic maps are used to display the spatial qualities of geo- aries. These maps are created for areas such as cities and towns, graphic data. A distinction must be made, however, between counties, recreational areas, and other locations where we want geographic phenomena and geographic data. Data are facts to communicate the distribution and location of such entities. observed or measured from which conclusions can be drawn. State transportation departments and private mapping corpora- Attribute is another word that can be used in place of data. tions create highway maps that focus on the transportation Geographic data are selected features (usually numerical) that network of a particular area. The general purpose map is con- geographers use to describe or measure, directly or indirectly, sidered to display information qualitatively by using symbols phenomena that have a spatial quality. This information is the to communicate distributions on the cultural landscape. attribute of a sampled location that is either qualitative or quan- Thematic maps have traditionally involved the display of titative. For example, the phenomenon of climate can be de- quantitative data that are easily incorporated into a GIS or scribed in part by looking at precipitation data. The use of advanced mapping software so that we can communicate the thematic maps allows the user to understand the where, what, spatial variability of that data. As we will see as we progress and when factors of spatial data. These three represent the na- through the remainder of the book, the characteristics of the ture of all data as they are tied to location, characteristics, and data will guide us in the selection of which thematic map time. Data are gathered by geographical location, which we type to use. That is, the data dictate the map type for the will see later in this chapter is directly tied to the symbolization display of the information. form used to map that data. These locations may be at a point, Within a GIS database lies the numerical foundation from along a line, or over an area. Their spatial components are gen- which mapping occurs. The data are presented in a row and erally considered as zero-dimensional (point), one-dimensional column format in which the geographic location, whether it (line), two-dimensional (area), three-dimensional (volume), be for a specific location that can occur at a point, along a and four-dimensional (space-time continuum) (Harvey 1969; line, or within a polygon, is normally represented in rows see Figure 4.1). Location characteristics can be directly influ- and the many attributes collected for that location are pre- enced by and may vary according to the map scale and time. sented in the columns. As an example, we may collect the Most geographical study involves explanation of phenomena total population, median family income, percentage of of the first four kinds, and to some degree the fifth. owner-occupied housing, or other such data for each county An interesting feature of phenomena is that their form is in Ohio. When we view the database, its contents represent intricately related to scale and may change with the level of FIGURE 4.1 THE SPATIAL COMPONENTS OF MAP SYMBOLS. Spatial Components Point symbols are considered as zero-dimensional Zero- One- Two- Three- Four- as they are indicative only of an x–y location. Dimensional Dimensional Dimensional Dimensional Dimensional 1 2 3 Point Line Area Volume Time den43823_ch04_063-079.indd Page 65 7/1/08 10:37:14 AM user-s208 /Volumes/202/MHDQ022/mhden6/de CHAPTER 4 THE NATURE OF GEOGRAPHIC DATA AND THE SELECTION OF THEMATIC MAP SYMBOLS 65 inquiry. A city, for example, may be a point phenomenon at the Data Characteristics macroscale but can be considered to have two-dimensional qualities when examined at micro levels. At one level of Location investigation, a road may be a link between points (one- Observations of data may be found as a point, line, or area dimensional), but at micro levels the road can have the two- feature. Attributes are assigned to these locations within a dimensional, areal qualities of length and width. database and then mapped in order to depict the spatial vari- Examples of volume phenomena include landforms, oceans, ation of that data. and the atmosphere. Other geographic phenomena that are conventionally treated as three-dimensional because of their Point Data. The implication of point data is that it is applied to similarity to volumes are rainfall, temperature, growing- a specific location that has a unique geospatial coordinate set season days, and such derived ratios as population density (x–y)—such as generic x, y positions, or latitude-longitude or and disease-related deaths. Rainfall collects in a glass and is eastings and northings (which are actually y–x). Within a par- volumetric; people piled close together make up a volume. ticular topic, population for example, there will be only a single Space-time phenomena are best exemplified by succession value for each point. However, within the attribute database (for example, human settlement over time) and migration/ there may be any number of these point-data relationships that diffusion. Location characteristics can be directly influenced have the potential to be mapped. The data value is referred to by and may vary according to the map scale and time. as the attribute, the location (point) as a node or location ID, We observe these phenomena but we map data. Such data and the number of individual points as the observations of the may take on various characteristics that include location, data set. Point data may be counted, measured, or estimated form, and time. (see Figure 4.2). Examples of these are listed in Table 4.1. Observation Number Total Data Value Derived Data Value ID Information Attribute Variables Population Density (People Per Square Mile) 1990 = 10467/253.6712 = 41.2621 or 41 People Per Square Mile FIGURE 4.2 ATTRIBUTE TABLE. Observations are county-based values. The ID information is used by software to match the values to the map locations. Examples shown here include both discrete and derived data values. den43823_ch04_063-079.indd Page 66 7/1/08 10:37:21 AM user-s208 /Volumes/202/MHDQ022/mhden6/de 66 PART 1 THEMATIC MAPPING ESSENTIALS TABLE 4.1 NATURE OF DATA DEFINED BY LOCATION applied to the entire area as measured by a series of point loca- tions within that area or by a quantity thought to exist uni- Data Type Examples formly over the entire area (see Table 4.1). Point Data Number of shoppers at a mall A special case of area data is that of volume data com- Number of students enrolled at an prised of attributes that exist over a three-dimensional extent elementary school of an interpolated surface. An example of these data is the Temperature measured at the local airport total volume of precipitation that falls over a particular Dollar value of your home stream’s drainage basin. There may be a network of precipi- Value of damage caused by a tornado that hit tation gauges that are used to measure the amount of pre- your town cipitation at a point and these data are extrapolated out to the Line Data Traffic flow in vehicles per day Stream discharge entire surface area. The total volume of precipitation for a Movement of a commodity exported to particular storm event can be calculated once the interpo- various locations lated surface has been created (see Chapter 9). Area Data Population density in people per square mile Crop yield in bushel per acre Acreage of various county land-use categories Form Precipitation from a storm event Data form establishes the contextual characteristics in which the data are found. Three context forms permit the user to understand the details of the data characteristics. These Line Data. Quantitative line data are generated by the collec- forms include either quantitative or qualitative, discrete or tion of information as applied along a linear feature or path. continuous, and total or derived data. Thus a series of nodes are used to define the path and the at- tributes are used to define the characteristics of that path. For Qualitative/Quantitative Context. The form of data charac- example, transportation flow maps are generated by a series teristics is comprised of three relationships. The primary of data collection points along a street or highway. These form of the data is determined by the qualitative versus points are selected by the traffic engineer as representative of quantitative nature. The qualitative attribute of the data de- travel along a particular corridor. These can be identified by scribes the inherent nature of the feature, for example, house, the rubber tubing stretched halfway across a street and con- church, railroad, swamp (see Table 4.2a). These features nected to a counting device mounted alongside the street. By may appear as point, line, or area locations and are described determining how many vehicles pass that point in a given according to their cultural or physical form. Quantitative time period and then associating that point to the other col- data utilize numerical values to indicate the differences in lection locations, one can determine the flow volume of ve- attributes according to some measurement scale, for exam- hicles per day along a roadway. Therefore, a set of point data ple, feet above sea level, total population, or degrees Celsius. attributes is used to create line data. Data can be inferred The quantitative data form also applies to any of the loca- only along the length of the line based upon the attributes tional characteristics. determined to exist at various points along that line (see Table 4.1). Data may also be assigned to a line representing Spatial Context. This second data form is determined by its the quantity of goods shipped from one location to another. distribution. Point data is unique in form and represents a The data would apply to the length of the line without spatial value that applies to a specific node within the spatial frame- variation along the line. work. Data may be considered as unique if the node represents a single x–y coordinate location or if the node represents a Area Data. Attributes that exist over a two-dimensional ex- centroid identifying a line or area component. That is, both tent (polygon) comprise area data. In certain instances we are temperature data measured at a weather station and the total able to determine the attribute from aerial photography or population of a county or other enumeration unit are consid- other remote sensing techniques. The area of the polygon in ered as discrete data (see Table 4.2b). Frequently associated square kilometers of a lake or land in forest can be determined with data counts or totals, the data are uniquely associated to based upon the scale of the photograph or image. The polygon a singular spatial location, no matter how it is defined. Non- area is frequently generated automatically by the GIS or map- unique data are those data that are considered to be of a ping software whereas the area data are assigned, observed, or continuous form. Examples of continuous data include most measured for that polygon. The government collects statistics weather-related variables and the topographic surface. Con- that tell us how much land area is associated with a particular tinuous data are areal and exist everywhere. It is impractical land use. By knowing the total number of acres in farms and to measure these variables at every position so we utilize a the total bushels of corn produced by those farms, we can cal- sampling procedure to collect data at various locations within culate an attribute of bushels per acre. Other common examples the areal extent. These data may then be displayed as a com- of area data include population density of people per square puted continuous variable using interpolation procedures mile (see Figure 4.2). Area data include attributes that may be (refer to Chapter 9). den43823_ch04_063-079.indd Page 67 7/1/08 8:40:48 AM user-s208 /Volumes/202/MHDQ022/mhden6/de CHAPTER 4 THE NATURE OF GEOGRAPHIC DATA AND THE SELECTION OF THEMATIC MAP SYMBOLS 67 TABLE 4.2 CHARACTERISTIC FORM OF DATA a) Characteristic View Qualitative versus Quantitative Mine Tons of coal Airport Number of aircraft River Volume of water Forest Board feet of timber Farms Acres of farmland b) Spatial View Discrete versus Continuous Temperature at your home Temperature across the United States Precipitation at the airport Precipitation in the Southeast Elevation of the bridge Topographic surface c) Attribute View Totals versus Derived Total population People per square kilometer Ohio immigrants to Minnesota Ohio immigrants to Minnesota as a percentage of all Minnesota immigrants Employment in mining Mining employment as a percentage of all employment Attribute Context. This data form is characterized as either a data have been normalized to adjust for the bias of size. In total or derived form. Total data, similar to discrete used in order to communicate effectively the nature of the data, the the spatial context discussed above, may be observed, col- data are described according to its form: qualitative or quan- lected, or measured at a location so that it has a single value. titative; discrete or continuous; total or derived. These char- That value is associated only with the location for which it acteristics also play an important role in determining the was acquired. One can easily identify a total value by the fact thematic map type used to display that data. Again, the data that the data are counts or measurements identified by their dictate the map type for the display of the information. descriptor, for example, 12 degrees Celsius or 9669 people. In this context, total data may be considered as a single obser- Time vation of one variable among possibly many variables of a It is also important to recognize that all data are time spe- larger data set (see Figure 4.2). Derived data are represented cific. Therefore, time is a necessary attribute to a data set. We by an attribute definition that indicates the data have been display data on a map and we tell the map reader the subject mathematically calculated. This is frequently done to normal- matter being presented. It is also necessary to communicate ize (standardize) the data so as to either adjust for the impact the time period for which the data applies. Data may have a of area/size or to represent the data as a rate or percentage very specific time stamp, such as 1200Z, September 13, (see Chapter 5). Derived data are also used to make compari- 1944. (The Z is used to specify time adjusted to Greenwich sons, for example, the variation in crime rates per county. mean time in order to eliminate confusion between time Area plays an important role in the potential quantity that zones or the application of daylight savings time.) We find can occur within its bounds. Logically, a county of larger this specificity applying to meteorological data. We may also area should be able to hold a larger population or larger farms find data applying to a specific month, year, or decade. should be able to produce a greater quantity of crops. In order Temporal data are used either to graph sequential (longitudi- to make these data comparable, we must normalize the data nal) changes in values for a location or to determine magnitude adjusting for the area differences. This is achieved by divid- of change between time period one and time period two. Fre- ing the area into the total population or into the total quantity quently, this magnitude is expressed in terms of a percentage of crops produced. The resulting values are represented as change since the initial date. For example, the change in popu- people per square kilometer or in bushels per acre. Similar lation between 1960 and 2005 may be mapped as a percentage logic applies to data characterizing a sector of the population. change between the two time periods. Change values may be Expressing data in the form of rates or percentages normal- either positive or negative, depending on the population charac- izes the data to a specific size of population or a number per teristics over time. Temporal data may include only a beginning 100. Table 4.2c provides examples of these data forms. and ending date as just described or the data may include a va- The data characteristic of form is utilized in order to riety of time periods. Even though time is a continuum and thus define its spatial association and/or to indicate whether the uniform, the collection periods may not be uniformly spaced. den43823_ch04_063-079.indd Page 68 6/7/08 4:26:03 PM user-s175 /Volumes/109/MHDQ022/den 68 PART 1 THEMATIC MAPPING ESSENTIALS Whatever the time period, we are obligated to communi- The transformation of data when changing to a larger cate this in the map’s title. The date associated with the data scale presents different problems. Data that are collected source does not necessarily communicate the date of the for a larger area in order to produce the small-scale map data. For example, in many statistical publications one can will appear highly simplified when retained for the depic- find historical data that can be mapped. Only the data year tion at the larger scale. A disaggregation of the data re- and not the publication year should be communicated to the quires a greater effort. Census data collected at the county map reader. Temporal data may also cover a period of time, level may also be available at the census tract level. If the multiple years or multiple decades. Again, the communica- data cannot easily be disaggregated, a return to the original tion of time sequence displayed is what is important. data source may be required in order to acquire more ap- propriate data for the new map scale. This may require Data Transformations revisiting the original data source for acquisition of more refined data. The cartographer will frequently modify the data prior to its display. Data collected or imported may be appropriate as Form nonspatial tabular data and yet not be ready for cartographic applications. In order to prepare the data for mapping, the A second type of transformation of the data may be associ- cartographer will manipulate either the attribute data or the ated with the data form. Data that are collected at a series of geographic space to which it applies. Three forms of trans- random sampling points may be used to transform the data formations which may occur prior to mapping include the from discrete point data to the display of the data as a con- modification of scale, form, or geometric boundary. tinuous variable. Such a transformation is common when mapping surfaces. Data that occur at random locations will Scale require a transformation to continuous data via interpolation (see Chapter 9). The first is the modification of the data based upon changes Data form transformation may also involve the conver- in map scale. Data that may be appropriate for a large scale sion of totals to rates or percentages. This transformation is map, such as 1:24,000, may not be appropriate for a small important to adjust for the impact of size and area as ex- scale map, such as 1:1,000,000. If we are converting the plained above. In order for spatial comparisons to be made, scale of display from the large-scale to the small-scale map, this transformation process is frequently required depending the data may need to be aggregated. Thus, data that origi- upon the thematic map type being selected. nally was displayed according to census tracts may require aggregating to county values at the smaller scale. Neighbor- hood statistics may be combined to represent city statistics. Boundary Changes The transformations of the data as a result of changes to a One problem frequently overlooked when comparing data smaller scale are easily done by merging the data into a sin- in a temporal sequence is that the boundary of an enumera- gle value. The data transformations include a change in the tion unit may vary in different time periods. This is com- definition of the data location. Data that originally was ap- monly referred to as the Modifiable Area Unit Problem plied to a point location may now be aggregated and applied (MAUP) and is a common occurrence when using U.S. to an area location (see Figure 4.3). Census tract boundaries (Openshaw 1984). Just as impor- tant as it is to compare data of the same data form, it is equally important to compare data covering comparable areas. The manner in which the Census Bureau defines cen- Larger Smaller sus tracts and their boundaries often changes from one de- cennial census to another. Some boundary changes are mi- A. nor, creating sliver polygons when the boundaries are overlaid. Other times the changes are significant. You will even find situations where the boundary remains the same but the census tract number is changed so that the tract number may appear in sequential censuses but applies to B. different polygons. As a cartographer, you must be aware that such boundary definitions and locations may be differ- ent between the dates of comparison. The simplest ap- Smaller Larger proach in addressing this problem is to aggregate the data to a smaller scale or county level in this example. Other more complicated approaches involve the use of spatial FIGURE 4.3 DATA AGGREGATION. analysis modeling (Green and Flowerdew 1996) and/or re- As the scale decreases, data should be aggregated in order to mote sensing techniques for the adjustments of boundary maintain clarity. changes (Holt et al. 2004). den43823_ch04_063-079.indd Page 69 6/7/08 5:49:24 PM user-s175 /Volumes/109/MHDQ022/den CHAPTER 4 THE NATURE OF GEOGRAPHIC DATA AND THE SELECTION OF THEMATIC MAP SYMBOLS 69 DATA MEASUREMENT several minor ports. The definition of major and minor may be determined by the number of ships serviced in a S. S. Stevens, a noted scientist and psychologist, has said year or by the total tonnage of goods that pass through the that measurement is the “assignment of numerals to things port annually. so as to represent facts and conventions about them” (Harvey One interesting feature of the ordinal class is that we can 1969, 306 ). It has been customary in recent years for geog- attach any numerical scale to the ranking without violating raphers to classify the ways they measure events into catego- the underlying structure. If we know that the order is K L M, ries of data measurement. Measurement is an attempt to we can have either K = 5, L = 3, M = 1, or K = 500, L = 300, structure observations about reality. Ways of doing this can M = 1, while retaining the original representation. Remem- be grouped into four levels, depending on the mathematical ber that in ordinal measurement we do not know how much attributes of the observed facts. A given measurement sys- difference separates the events in the array. For example, tem can be assigned to one of these four levels: nominal, several geography students spend a summer touring 50 large ordinal, interval, and ratio, listed in increasing order of so- cities throughout North America. After they return, their phistication of measurement. Methods of cartographic sym- professor asks them to rank the cities based on their appeal, bolization are chosen specially for representing geographic from best liked to least liked. In this array, a city’s position phenomena or data at these levels of measurement. The mea- is known in the overall ranking, but it is not possible to dis- surement scale will also play an important role in identifying cern how much it differs from those ranking above and the thematic map type to be used. below it. Other examples of the ordinal measurement are social class, social power (more, less), and agreement (strongly Nominal agree, strongly disagree). We may also utilize the ordinal class in refining a nominal description based upon impor- Nominal scaling is the simplest level of data measurement tance of the category. Whereas, nominally we identify a lin- (Taylor 1977), sometimes considered a qualitative mea- ear feature as a road, we use the ordinal measurement to surement to be descriptive; answering the question of what differentiate between a major road or minor road, an inter- is being mapped. An example would be the nominal iden- state highway or a national highway. Other examples would tification of wheat regions, corn regions, and soybean include the distinction between a seaport versus a major sea- regions. Each crop region is distinct; arithmetic opera- port or a forest versus a mixed deciduous forest. Again, the tions between the regions are not possible at this level. characteristic that defines ordinal apart from other ap- Political party affiliation (Democrat, Independent, Repub- proaches is that we cannot discern magnitude differences lican), sex (male, female), and response (yes, no) are other between the observations, only a hierarchical distinction examples. At this level of measurement, mathematical of categories. operations cannot be performed between classes. Equality/ inequality between groups according to their classifica- tion or identification or the dominance spatially of one Interval group versus another may be ascertained using this data At this measurement, we can array the events in order of rank measurement level. and know the distance between ranks. Observations with numerical scores at the interval measurement are important in geographical analysis because data at this level are needed to Ordinal perform fundamental statistical tests having predictive power. The underlying structure of ordinal measurement is a hierar- Units on an interval scale are equal throughout; that is, one chy of rank. Objects or events are arranged from least to degree on the Fahrenheit scale is assumed to be the same most or vice versa, and the information obtainable is of the regardless of whether it’s at 22–23 or 78–79 degrees. “greater than” or “less than” variety. Ordinal measurement Magnitude scales at the interval level have no natural ori- provides no way of determining how much distance sepa- gin; any beginning point may be used. The classic example rates the items in the array. is the Fahrenheit temperature scale. There are no absolute There are several types of ordinal measurement. In values associated with interval measurement; they are rela- complete ordering, every element in the array has its own tive. In the interval approach, units are agreed upon by re- position, and no other element can share this position. This searchers and are assumed to be standard from one set of kind of ordinal measurement is considered relatively conditions to another. Variables at the interval scale do not strong because statistical observations about the ranking have absolute zero as a starting point. are possible. This is not the case with the second main Try this experiment for an example. Place a yardstick type, weak ordering, in which elements can share posi- next to a tabletop, but not touching the floor. You can slide tions (called paired ranks) along the ordinal continuum. In the yardstick up and down relative to the tabletop, but as it is the first instance there may be only one major seaport not touching the floor, there is no absolute height for the ta- along a coast while in the second instance there may be ble, only a relative height as shown by the sliding yardstick. den43823_ch04_063-079.indd Page 70 7/1/08 8:40:53 AM user-s208 /Volumes/202/MHDQ022/mhden6/de 70 PART 1 THEMATIC MAPPING ESSENTIALS Ratio type. General reference maps utilize data that are qualitative with either a nominal or ordinal measurement that occur at Like interval, ratio measurement involves ordering events discrete locations spatially. The thematic map types intro- with known distances separating the events. The differ- duced in Chapter 1, which are presented in this table are ad- ence is that ratio magnitudes are absolute, having a known dressed in detail within the chapters of Part II of this text. starting point. This scale of measurement has a zero (ab- Thematic maps display quantitative data that primarily utilize sence of a magnitude) as its starting point. The Kelvin ordinal or interval measurements of discrete values of totals. absolute temperature scale is an example here. Other ex- The exceptions to this generalization are both the choro- amples include weight (20 lbs is twice as heavy as 10 lbs), pleth and surface maps which display derived and continu- and distance (100 miles is twice as far as 50 miles). ous data, respectively. The characteristics of the data dictate Elevation above sea level is another ratio scale where the the thematic map type that should be created. average elevation of sea level is set as zero. The ratio ap- proach is important to geography because more sophisti- cated statistical tests can be performed using this level Map Symbols of measurement. The selection of symbols is one of the more interesting tasks In the yardstick/tabletop example, if the yardstick is placed for the cartographic designer. The choice is wide, and no on the floor, then the height of the tabletop can be measured firm rules prevail. Symbol selection, however, is increas- relative to a zero starting point (the floor). If we chop off its ingly based on a compelling system of logic tied to both the legs, we know exactly how much shorter it has become. type of geographic phenomenon mapped and certain graphic The amount of information that can be obtained, statistical primitives or variables. confidence, and predictive power increase as one progresses Symbols are the graphic marks used to encode the the- from nominal to ratio measurement. Cartographically, we matic distribution onto the map. From a vast array of sym- view interval and ratio as being equal. The thematic map bols having different dimensions, the cartographer selects generated from interval data will utilize the same design the symbol that best represents the geographic phenomena. techniques if it was created using ratio data. Fortunately, the task is reduced somewhat by controlling factors, such as cartographic convention and the inability of most map readers to easily understand the more complex DATA: THEMATIC MAP symbols that might be chosen. RELATIONSHIPS The three generally recognized cartographic symbol types are directly tied to the data characteristics of location. The nature of the geographic data described above has a di- These point, line, and area symbols continue to be the stan- rect relationship on the thematic map type selected and thus dard in thematic mapping. The use of GIS and advanced the cartographic symbolization utilized. Table 4.3 identifies mapping software has given the cartographer the power to the data characteristics and their associated thematic map generate maps in either two- or three-dimensional design TABLE 4.3 DATA AND THEMATIC MAP RELATIONSHIPS Characteristic View Measurement Spatial View Qualitative Quantitative Nominal Ordinal Interval/Ratio Discrete Continuous X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X 1. Other data restrictions exist that should be observed and will be detailed in their constituent chapter. 2. Densities are inappropriate for this map type. den43823_ch04_063-079.indd Page 71 7/1/08 8:42:02 AM user-s208 /Volumes/202/MHDQ022/mhden6/de CHAPTER 4 THE NATURE OF GEOGRAPHIC DATA AND THE SELECTION OF THEMATIC MAP SYMBOLS 71 (see Figure 4.4). Maps that use the three-dimensional design reacts to the individual components. These components, display either volumetric data or surface phenomena. whether observed individually or in concert with each other, There is a logical (and traditional) correspondence between serve to communicate the topic to the reader. These symbols geographic phenomena (point, line, area, and volume) and serve as visual variables from which the reader gathers infor- the employment of symbol types (point, line, area). Of course, mation and interprets the map. Bertin described the use of the match is not a convenient one-to-one correspondence. symbols as “components of the graphic system [to] be called Point data (for example, cities at the appropriate scale) are ‘visual variables’” (Bertin 1983, 42). Two of the variables customarily mapped by point symbols such as dots, or scaled are, in his words, the planar dimensions. That is the X and Y circles. Roads, which are linear phenomena, can be mapped positioning in two-dimensional space. Such variables are by line symbols; geographic phenomena having areal ex- used to represent the location of objects whether they be tent (lakes, countries, nations) can be mapped with areal represented by point, line, or area symbols. In the nominal or symbolization—solid fills of varying hue, intensity, or ordinal measurement, they are used in a general reference value, or sometimes patterns. Geographical landforms have map to identify location. been mapped by a form of area symbolization called hyp- The remaining six variables are size, shape, orientation, sometric tinting: areas between selected elevation boundar- texture, saturation, and value. These variables allow the car- ies are rendered in various color shades. Elevation, also a tographer to provide additional information that serves to 3-D form, is mapped by contours of elevation, which are display more complex spatial settings or quantitative data in 2-D line symbols. The use of three-dimensional mapping the form of thematic maps (Bertin 1983; see Figure 4.5 and has become a standard form of map symbol in which to Color Plate 4.1). represent volume phenomena (see Figure 4.4). The three symbol types may be matched to the nature of Size the data, both characteristics and measurement levels, to pro- duce a valuable typology of map symbols (see Figure 4.4). Size is used to imply relative levels of importance. Propor- The selection of map symbolization has been made easier by tional symbols are used in which the size of the geometric the inclusion of a wider array of symbols in GIS and map- form is scaled proportionally to the data. Proportional cir- ping software. Clipart and other sources of pictoral symbols cles, squares, and other geometric forms as well as irregular have widened the options of the cartographer significantly. shapes like cartograms are frequently used to portray varia- Symbolization choice is enhanced through careful consider- tions in data through the use of varying size of the symbol. ation of the visual variables. Line thickness is also a representation of size and thus com- municates the nature of flow data. Visual Variables Shape The way we use symbols to display spatial phenomena is to Shape is used to “(1) reveal similar elements, and therefore, create a graphic scene in which tshe reader observes and different elements and (2) to facilitate external identification, Attribute View Symbol Chapter in Totals Derived Type Dimensionality Map Type Text General Purpose: Point 0 Reference Line 1 Reference Area 2 Reference Thematic: X Area 2 Choropleth 6 X Point 0 Dot Density 7 X X1, 2 Point 0 Proportional 8 X X Area 2/3 Surface 9 Cartogram: 10 X X1, 2 Area 2 Value-by-Area X X1, 2 Line 1/2 Flow 11 den43823_ch04_063-079.indd Page 72 6/7/08 4:26:04 PM user-s175 /Volumes/109/MHDQ022/den 72 PART 1 THEMATIC MAPPING ESSENTIALS Qualitative Quantitative By Location and Nature By Location and Amount Point Symbols Ship Wreck Presence School Church Length Building Mine Area Bench Mark Airport Towns and Cities Volume Line Symbols One-Dimensional Political 06 10 Boundary Relative Parallels and Position 3 Meridians 997 100 1000 Railroad Two-Dimensional Width River Road Area Symbols 20 30 40 50 60 Bounded Lake By Isolines Marsh Sand Bounded By Forest Non-Isolines FIGURE 4.4 THE COMPLEXITY OF MAP SYMBOLS RESULTS FROM THE NATURE AND CHARACTERISTICS OF THE DATA. Source: After Anonymous 1944. den43823_ch04_063-079.indd Page 73 7/1/08 8:42:08 AM user-s208 /Volumes/202/MHDQ022/mhden6/de CHAPTER 4 THE NATURE OF GEOGRAPHIC DATA AND THE SELECTION OF THEMATIC MAP SYMBOLS 73 FIGURE 4.5 VISUAL VARIABLES. Source: MacEachren, 1994. Used by permission. Point Line Area Size Value Hue Saturation Orientation Shape Arrangement Texture Focus through shape symbolization” (Bertin 1983, 95). Thus, we reader that those objects are unique and belong together as a select point symbols to represent a graphic display of the group. As similar orientation creates an essence of order and actual entity being mapped. For example, we use squares thus similarity, misalignment of symbols can achieve the op- with a small flag on top to represent schools, those with posite affect so that objects stand out on the map landscape. crosses to represent churches, crossed pickaxes to represent Special care should be taken when designing graphs in mines, and countless other symbol shapes that we have which a fill is used to help differentiate the data variables, grown to recognize by cartographic conventions. Carto- especially in bar graphs. Schultz (1961) found that the use of graphic traditions have established certain shape-object rela- diagonal line patterns can actually create a perception that tionship. For example, the use of a star to represent a capital the bars tilt as opposed to their actual parallel construction. or squares to represent buildings is recognized by the most Thus the use of diagonal lines should be avoided, consider- novice of map users. Linear and area symbols are also recog- ing the other fill options available. nized by their inherent shape in the process of representing roads, lakes, and so on. Texture As the complexity of a map increases and we become chal- Orientation lenged to find different visual designs to use as symbols, we The order of things on the landscape should not be changed to can select different textures (or patterns) for the symbols or increase the interpretation of information. The orientation of as an overlay to color in order to increase the number of building symbols and symbols of other structures should be symbol options. Frequently, texture is used in the area sym- such that it represents their actual position as closely as pos- bolization to help communicate areas where the landscape is sible. However, when we are using symbols in a more general somewhat coarse or smooth. The use of aerial photographs sense to represent an idea of the existence of objects, the ori- in a map layer helps to create this texture of the surface. The entation of those symbols creates a perception by the map upper surface of a forest canopy is coarse when compared to den43823_ch04_063-079.indd Page 74 6/7/08 4:26:14 PM user-s175 /Volumes/109/MHDQ022/den 74 PART 1 THEMATIC MAPPING ESSENTIALS a plowed field. There are many examples of how texture is TABLE 4.4 SOURCES OF ERROR ON THEMATIC MAPS found on the landscape. We can also design symbols through the use of texture. The variation in line or dot density per SOURCE ERRORS linear measurement along with the variation in dot or line Source Map Error size will create a (visual) texture variation. The use of clipart Existing Maps for patterns provides the cartographer with a large variety of Scale Detail symbol options that can be used in area symbols. Most GIS Accuracy and mapping software provide for the importation of sym- Projection bols and patterns of differing textures to use in the map’s Currency design. Data Entry Error Boundary Data Download of inaccurate boundary Saturation and Value Digitizing Bertin’s use of saturation and value are similar to the color Manual model of HSV described in detail in Chapter 15. Hue is the Heads-up name we apply to a particular color. Red, blue, or green are Attribute created by a portion of the electromagnetic spectrum with Incorrect entry wavelengths that separate these colors out from one another. Inappropriate link to ID We have since applied these names to the colors created. We Incomplete data can use hue to represent a large variety of symbols. Certain PROCESSING ERRORS hues, through the establishment of certain cartographic con- Numerical Rounding in Computing ventions, are reserved for symbolizing specific features. Blue Data Classification for water, red for roads or areas of building omission, green Data Transformation for vegetation are just a few hues that have become standard Interpolation Technique Inappropriate Use of Algorithm practices in symbolization. When designing thematic maps, the selection of color is one of the most exciting activities CARTOGRAPHIC DESIGN ERRORS available to the cartographer. Brewer (2005) provides a com- Thematic Map Type plete overview of the process of color selection. Scale Projection Saturation is thought of as a level of brightness of the hue Generalization where value is considered as a sequence of steps between Symbolization light and dark. The combination of these two variables Use of Color provides the cartographer with the capability of designing quality maps using either grayscale or a color model. The software utilized provides the cartographer with the choice of millions of color options based upon hue, saturation, and value. These colors are easy to use in a virtual environment Source Error but may become more of a challenge in the hard copy or Source errors are those errors that are found in the data col- printed environment. The cartographer should not rule out lection, compilation, and date entry procedures (Beard 1989). the use of grayscale in the design of maps. Many profes- As a mapping project begins, a search is conducted to deter- sional journals are just now beginning to publish color graph- mine the existence of maps (both digital and paper) for the ics. The grayscale map is still a most effective means of study area and data topic. Often these maps are used as refer- map design. ence or a base on which other information is compiled or The interpretation of these visual variables and how they placed. The use of base maps is quite common in cartography are used to achieve a visual hierarchy in map design is dis- providing foundations for exploratory design and layout (see cussed in Chapter 12 of this text. The fundamentals of figure- Chapter 1). ground association are achieved by the careful selection of the Caution must be taken in the selections of existing maps visual variables and the manipulation of their components. for use as a layer in the generation of a new map. These maps are limited in scale, detail, accuracy, projection, and the date to which the information applies. These limitations may inject CARTOGRAPHIC ERROR data and location discrepancies created by the generalization level required by a particular map scale. Perhaps the map pro- There are many places where inaccuracies can enter in the jection was inappropriate for the data being displayed or even mapping process. The kinds of error that are encountered in created without a geographic base. Using maps created by creating a map include source error, processing error, and others often perpetuates error and thus the map’s lineage must design error. These errors are discussed in this section and be considered. Careful evaluation of the map’s accuracy is es- outlined in Table 4.4. sential before making use of such a foundation. The date of den43823_ch04_063-079.indd Page 75 7/1/08 8:43:11 AM user-s208 /Volumes/202/MHDQ022/mhden6/de CHAPTER 4 THE NATURE OF GEOGRAPHIC DATA AND THE SELECTION OF THEMATIC MAP SYMBOLS 75 SIR JOSIAH STAMP When using data collected by others, we urge the reader to cube root and prepare wonderful diagrams. But you must never heed the words of Sir Josiah Stamp (1869–1941) of England’s forget that every one of these figures comes in the first instance Inland Revenue Department, who observed: from the village watchman who just puts down what he damn pleases. The government are very keen on amassing statistics. They collect them, add them, raise them to the nth power, take the Source: Stamp 1929: 258–59. the map may be too old for use in a current study. Both physi- be taken so that values of zero are not interpreted the same as cal and cultural features may be left off the map as a result of missing data. its age. The caution applies to both hard-copy maps and those in digital form. Assuming that the map is correct doesn’t nec- Processing Error essarily make it so. The federal government creates topographic maps at vary- Processing errors can result from the cartographic transforma- ing scales that include all these features plus a set of contour tion of data as a result of changing scale, projection, or data lines helping the map reader interpret and visualize the undu- form. Line simplification and data classification are techniques lations in surface topography. These maps are created so that involved in generalization (see Chapter 1). Cartographic gen- any location or elevation meets the National Map Accuracy eralization based upon scale, such as the removal of islands or Standards and thus gives us a base that is geometrically ac- the straightening of sinuous lines, often produces errors that curate. These maps have served as base maps on which car- are unknown to the map user (Clarke 1990). Classification tographers have added other layers of information. that overly generalizes the data will tend to reduce the spatial For data that is downloaded from government sites, educa- patterns that exist in that data. Careful consideration of clas- tional institutions, or other online sources, checking the meta- sification methods is crucial (see Chapter 5). data (data about data) can be one way to ascertain the source When utilizing computer software for data transforma- map’s qualities. Most correctly documented spatial metadata tion, the values of the raw data must be considered. Simple includes information such as its spatial reference (that is, map formulas used in spreadsheet software that permit the projection and coordinate system), an assessment of data conversion from total to derived data often provide results quality (including the lineage of both spatial and nonspatial with seven to ten decimal values. To include this level of attributes that the data may contain), and contact information precision will generate a false perception of accuracy of the about the data set, as well as other information. data when used in mapping. Numerical rounding of such Error can also enter via the data entry process. This pro- values to a single decimal equivalent permits the use of such cess includes the acquisition of digital boundary files and calculations without falsely implying greater accuracy. We associated attribute data. The source of the boundary file suggest that an increase of only a single decimal value over must be carefully considered when downloading from the that of the original data be used. Web (see Federal Governmental Agencies, page 76 ). Only Transformation from discrete to continuous data requires reputable sources should be considered. The production of the use of interpolation in order to generate the database. new boundary/location files via manual or heads-up digitiz- Care must be taken to use an appropriate algorithm for that ing also generates the potential for errors in geometry of the interpolation (Lo and Yeung 2007). A discussion of such map or subsequent creation of erroneous polygons that are conversions is provided in Chapter 9. nonexistent. The accuracy and precision of the instrument and the experience of the operator are also sources of error in Cartographic Design Error the map geometry (Burrough and McDonnell 1998). Careful editing must be included in the project to be certain that Once the data have been collected and processed, the car- errors of omission or commission are not included. tographer should take time to examine the data attribute The keying of attribute data into a database can create er- table for inconsistencies and completeness. Following that rors that are carried throughout the data analysis and data examination, the cartographic design process begins. The display phase of any project. Entry errors that include enter- first potential for design error is in the selection of the ing the wrong attribute, transposing numbers in a data value, wrong thematic map type based upon the data. As sug- or incorrect attribute-ID association may cause the map to gested, the data dictate the map type for the display of spa- display spatial patterns that do not actually exist. Care must tial data. Careful consideration of the nature of the data den43823_ch04_063-079.indd Page 76 6/7/08 4:26:14 PM user-s175 /Volumes/109/MHDQ022/den 76 PART 1 THEMATIC MAPPING ESSENTIALS will serve as a guide in the selection of the appropriate from either reputable government agencies, such as the U.S. thematic map (see Table 4.3). Census Bureau or the U.S. Geological Survey, who offer The cartographer must make many decisions when it these products, or from the software vendors themselves. comes to map design. Each of those decisions has the poten- tial for introducing map error whether it is factual on the map or one of incorrect perception by the map user. Choices U.S. Census Bureau in scale, projection, generalization, symbolization, and color In February 1989 the U.S. Census Bureau released the first are based upon traditions and conventions in cartography. To TIGER/Line (Topologically Integrated Geographic Encod- vary widely from such traditions may cause a decrease in the ing and Referencing ) files, called the Prototype TIGER/Line successful communication of the theme presented. An ex- files. The TIGER/Line files provided the first seamless na- amination of the best choices to be made in these areas are tionwide street centerline coverage of the United States and presented in both Part I and II of this text. Puerto Rico. Over the past 17 years, based upon user re- Not all errors can be eliminated, but every conceivable quests for additional data content, the TIGER/Line files have means must be employed to reduce error to tolerable levels. grown from a file containing six record types to a file con- Each mapping method discussed in this text has its own error taining 19 record types. sources, based on its unique way of mapping and symboliza- With the modernization of the Master Address File tion. The thematic map designer must learn to take these (MAF) and TIGER systems, the Geography Division will, error sources into account. in 2008, begin releasing TIGER spatial data in the following formats: Shapefiles DATA SOURCES TIGER/GML™ The Census Bureau also will make available the TIGER Data (GIS) Clearinghouses spatial data over the Web: Traditionally, the cartographer has had to access hard copy WebTIGER™—A Web Feature Service (WFS) of data by going to the library or by contacting local, state, interface allowing requests for geographic features or federal government agencies and then manually entering across the Web. It uses the XML (Extensible Markup that data into the mapping program. Currently, GIS and Language)-based GML (Geographic Markup mapping software come with a large data supply as a part of Language) for data exchange. the program. These data are frequently acquired from the Web Map Server (WMS)—A WMS producing maps 1990 or 2000 U.S. Census of Population and may be used to of spatially referenced data dynamically from TIGER produce a variety of thematic maps. With the rapidly ex- in PNG, GIF, and JPEG formats or as vector-based panding Internet, access to data is faster and easier than graphical elements in Scalable Vector Graphics (SVG). before. Most state governments provide GIS data clearing- Available with the 2003 TIGER/Line files contain up- houses from which data can be downloaded. Google dated national ZIP Code Tabulation Areas (ZCTAs) reflect- searches provide many links to data for almost any desired ing the October 2002 U.S. Postal Service ZIP Codes (U.S. topic. The U.S. federal government provides data access Census Bureau Website 2008 see Table 4.5). from agencies, such as the Bureau of the Census (popula- tion and housing, manufacturing and agriculture), the Geo- logical Survey, and the Department of Agriculture, to name U.S. Geological Survey a few. Most data warehouses allow for free access to data The U.S. Geological Survey produces a variety of digital from the public domain. vector and raster files that can be used in the generation of maps. The USGS Digital Cartographic Data include products such as DLG, DRG, DEM, DOQ, and NHD files. Federal Governmental Agencies Table 4.5 provides a summary of files scale and content. Not only are attribute data available via the Internet, one can These products are available from the USGS website also acquire digital boundary files for practically any loca- (USGS 2008). tion. Most GIS and mapping software contain basic boundary files with associated attribute data for the countries of the Federal Information Processing world, as well as many of their internal political units, such as Standards (FIPS) states and counties. Frequently, these boundary files are of a proprietary nature or they are not readily available for a loca- Most software utilizes a field (column) in the database to link tion to be mapped. Care should be taken when utilizing the data to the location for which it applies. Frequently, this boundary files acquired from remote sources. The quality and field will be one either containing the name of the associated reliability of the digital boundaries may not be of the highest observation or the associated Federal Information Processing standards. We recommend that you acquire boundary files Standards (FIPS) code. Every country, state, city, metropolitan den43823_ch04_063-079.indd Page 77 7/1/08 8:43:19 AM user-s208 /Volumes/202/MHDQ022/mhden6/de CHAPTER 4 THE NATURE OF GEOGRAPHIC DATA AND THE SELECTION OF THEMATIC MAP SYMBOLS 77 TABLE 4.5 EXAMPLES OF DIGITAL BOUNDARY AND DATA FILES AVAILABLE FROM THE U.S. CENSUS BUREAU AND THE U.S. GEOLOGICAL SURVEY U.S. Census Bureau: Vector/Raster: Topologically Integrated Geographic Encoding and Referencing (TIGER®) System. Content of Shapefiles: Blocks Block Groups Census Tracts Counties County Subdivisions Places Urban Areas Congressional Districts States And many more U.S. Geological Survey: Vector: Digital Line Graphs (DLG) are digital vector representations of cartographic information derived from USGS maps and related sources. Scales Available: 1:24,000, 1:100,000, and 1:2,000,000 Layers Available: Public Land Survey System (PLSS) Boundaries Transportation Hydrography Hypsography Non-vegetated Features Vegetation Survey Control and Markers Manmade Features Raster: Digital Raster Graphics (DRG) is a scanned image of a U.S. Geological Survey (USGS) standard series topographic map, including all map collar information. The image inside the map neatline is georeferenced to the surface of the Earth and fit to the Universal Transverse Mercator projection. The horizontal positional accuracy and datum of the DRG matches the accuracy and datum of the source map. The map is scanned at a minimum resolution of 250 dots per inch. Digital Elevation Models (DEM) were comprised of scanned topographic map series. As of November 2006, the DEM was no longer offered by the USGS as it has been replaced by the NED. National Elevation Data sets (NED) has seamless elevation coverage for the United States with a resolution of 1arc second. Elevation is provided in meters. area, or other identifiable unit for which data are collected are designation. If this is the order in which data are provided, assigned a FIPS code. Country codes are comprised of two- the data will be in neither alphabetical order by state name letter designation, U.S. states and territories are designated by nor by county FIPS code. Many databases are created using a two-digit number, and U.S. counties are designated by a an alphabetical order by county name. The cartographer is three-digit number. cautioned that data errors can occur when combining data sets if the states contain counties that begin with “Mc.” Table 4.7 displays a select list of counties for the state of Potential Problems Tennessee (state FIPS 5 47) and those counties whose It is the cartographer’s responsibility to utilize accurate data name begins with “M.” The FIPS code order for these in a responsible manner. Caution is advised when copying counties place McMinn and McNairy counties at the columns of data into a database. The order in which the states beginning of the “M” list (county FIPS of 107 and 109, appear within a column will depend upon the manner in respectively). Many data sets are produced alphabetically which the alphabetical list is created. Table 4.6 provides a with these counties occurring farther down the list, after list of select states ordered alphabetically by their two-letter Maury County (119). Care must be taken when importing den43823_ch04_063-079.indd Page 78 7/1/08 8:43:29 AM user-s208 /Volumes/202/MHDQ022/mhden6/de 78 PART 1 THEMATIC MAPPING ESSENTIALS TABLE 4.6 FEDERAL INFORMATION PROCESSING Green, M., and R. Flowerdew. 1996. New Evidence on the Modi- STANDARDS (FIPS) CODES FOR SELECT STATES. fiable Area Unit Problem. In P. Longley and M. Batty (eds), Spa- The States Beginning with the Letter “A” in Order by Their tial Analysis Modelling in a GIS Environment (41–54). New York: Two-Letter Designation. Wiley. Harvey, D. 1969. Explanation in Geography. New York: St. Martin’s 2-Letter State State Press, 293–98. Designation Name FIPS Holt, J., C. Lo, and T. Hodler. 2004. Dasymetric Estimation of AK Alaska 02 Population Density and Areal Interpolation of Census Data to AL Alabama 01 Compensate for Census Geography Changes, Metropolitan AR Arkansas 05 Atlanta, 1980–2000. Cartography and Geographic Information AZ Arizona 04 Science 31 (2): 103–21. Lo, C., and A. Yeung. 2007. Concepts and Techniques of Geographic Information Systems, 2nd ed. Upper Saddle River, TABLE 4.7 FEDERAL INFORMATION PROCESSING NJ: Prentice-Hall. STANDARDS (FIPS) FOR SELECT COUNTIES. MacEachren, A. 1994. Some Truth With Maps. Washington, DC: Examples are from the State of Tennessee. (a) FIPS Code Order. Association of American Geographers, 129. (b) Alphabetical Order. Meyer, T. 1997. NCGIA Core Curriculum in GIScience. Openshaw, S. 1984. The Modifiable Area Unit Problem. Norfolk, (a) (b) VA: Norwick. State County County* Alphabetical FIPS FIPS Name List Sauer, K. 1956. The Education of a Geographer. Annals of the Association of American Geographers 46: 287–99. 47 107 McMinn Macon Schultz, G. 1961. Beware of Diagonal Lines in Bar Graphs. The 47 109 McNairy Madison Professional Geographer 13 (4): 28–29. 47 111 Macon Marion 47 113 Madison Marshall Stamp, J. 1929. Some Economic Factors in Modern Life, 47 115 Marion Maury 258–59. 47 117 Marshall McMinn Taylor, P. 1977. Quantitative Methods in Geography: An Introduc- 47 119 Maury McNairy tion to Spatial Analysis. Boston: Houghton Mifflin. 47 121 Meigs Meigs U.S. Census Bureau. 2008. http://factfinder.census.gov. Accessed 47 123 Monroe Monroe January 12, 2008. 47 125 Montgomery Montgomery 47 127 Moore Moore U.S. Geological Survey. 2008. http://nationalmap.gov/gio/status. 47 129 Morgan Morgan html. Accessed January 12, 2008. *State alternatives for county: Alaska—Borough; Louisiana—Parish GLOSSARY data into an existing database. If the existing database and area data attributes that may be applied to an entire area as a the one being imported are arranged differently, the com- quantity that is assumed to exist uniformly (rightly or wrongly) bined database will have incorrect data associated with over an entire area; examples include people per square kilometer these counties. and bushels per acre attribute data values or characteristics associated with a point, line, or area REFERENCES continuous form variables that are areal and exist everywhere, such as temperature or barometric pressure; these values are often Anonymous. 1944. A Proposed Atlas of Diseases. Geographical interpolated from sample observations that lie within the area of Review, Vol 34 (4), 642–52. study Beard, K. 1989. Use Error: The Neglected Error Component. data form contextual characteristics that include identification as Auto-Carto 9 (Ninth International Symposium on Computer- either quantitative or qualitative, discrete or continuous, and total Assisted Cartography) 808–17. or derived data Bertin, J. 1983. Semiology of Graphics. Madison, WI: University data measurement an attempt to structure observations about of Wisconsin Press. reality; observations can be grouped into four levels, depending on Burrough, P., and R. McDonnell. 1998. Principles of Geographi- the mathematical attributes of the observed facts, as either cal Information Systems. New York: Oxford Press.