Using Maps to Display Geographic Risk, Personal Health Data, and Ownership
Summary and Keywords
Fundamental structural features of risk maps influence how health risk and burden information is understood. The mapping of health data by medical geographers in the 1800s has evolved into the field of geovisualization and the use of online, geographic information system (GIS) interactive maps. Thematic (statistical) map types provide basic principles for mapping geographic health data. It is important to match the nature of statistical data with map type to minimize the potential for communicating misleading messages. Strategic use of structural map features can facilitate or hinder accurate comprehension of health risk messages in maps. A key challenge remains in designing maps to communicate a clear message given the complexity of modern health risk burdens. Various structural map features such as symbols, color, grouping of statistical data, scale, and legend must be considered for their impact on accurate comprehension and message clarity. Cognitive theory in relationship to map comprehension plays a role, as do insights from research on visualizing uncertainty, future trends in developing predictive mapping tools for public health planning, the use of geo-social and “big data,” as well as data ownership.
This article examines structural message features of maps and how structural features are used to convey health risk in maps. The role for risk maps will only grow in time as public health threats around the globe increase and communities collaborate to address threats. Despite the proliferation of geographic digital maps including geovisualization of health risks, systematic evaluation of the design, portrayal, and impact of these map features is lagging, resulting in a lack of guidance on evidence-based practices (Kinkeldey, MacEachren, & Schiewe, 2014).
Yet despite evidence-based guidelines, enthusiasm for producing online maps to depict health risk information graphically (and cartographically) continues to grow with advances in information technologies. Geographic information system (GIS)–driven technologies produce spatial and temporal health risk information as maps. Public health agencies such as the Centers for Disease Control and Prevention (CDC), the National Cancer Institute (NCI), and the Department of Health and Human Services (DHHS) use internally and make available to the public health risk maps on a range of topics. Health risk maps vary from cancer to injury risk (e.g., motor vehicle, poisoning, fires), non-communicable chronic disease prevalence (e.g., heart disease, diabetes), infectious disease (e.g., West Nile virus, Lyme disease), to climate change health impacts (e.g., county vulnerability profiles). Public health program planners increasingly turn to maps as a planning and evaluation tool to make resource allocation decisions and communicate with the public and policymakers about local health burdens. How well lay and public health professionals accurately understand these forms of communication is unknown. Croner and others report that online maps are increasingly becoming an information source that consumers turn to (Croner, 2003; Lindenbaum, 2006; Severtson, 2015).
Online map literacy encompasses specific abilities. Cartographic research has conceptualized map literacy with attention to improving map design and the transmission of intended messages. MacEachren (1995) notes that earlier efforts in geography on the communication of maps (before the digital, interactive era) were misguided in that maps do not “contain” and “transmit” their messages to users but stimulate ideas and inferences by interacting with prior beliefs of those users. Other research approaches to online map literacy place emphasis on the map user rather than on map design. Such conceptualizations elucidate understanding about cognitive processes associated with map use, prior experience, spatial abilities, and memory. A user-centered approach has received some criticism with claims that this research cannot translate into valid map design practices because its focus is on one part of a complex process. MacEachren (1995, p. 12) argues that maps should be studied as “many potential representations of phenomena in space that a user may draw upon as a source of information or an aid to decision-making and behavior in space” and that “the map user’s interaction with the map should be viewed as a complex information problem.”
A communication perspective raises the need to minimize the potential for consumers of health risk maps to draw inaccurate, inappropriate, or misleading conclusions. Other disciplines in addition to communication that inform promising approaches to online map literacy include instructional design, cognitive psychology, geography, human computer interactions (HCI), sociology, aesthetics, architecture, and interface design. Visualizing federal population health statistics that communicate a clear message while avoiding misleading messages presents its own set of challenges. Cartographic best practices for map design are available at the CDC. This resource provides tips for creating maps for public health, describes map elements, describes how data are typically classified, shows types of thematic (statistical) maps, and recommends resources. Generally, however, evidence-based guidelines are underdeveloped (Clarke, McLafferty, & Tempalski, 1996). The evolving field of geovisualization brings an entirely new dynamic to health risk mapping, presenting an opportunity to augment the function of risk maps as decision support tools as well as to communicate dynamic risk information.
Historically, it was with the availability of national-level statistics that cartographers and statisticians began mapping health data such as diseases and sanitation. The most famous example is of Dr. John Snow, who demonstrated using a geographic health risk map the water-borne origin of cholera by plotting cholera-related deaths in London during the most severe 1854 epidemic. In addition to disease, he plotted the city’s water pumps and drew concentric circles to determine that the areas with the highest concentration of cases were within close proximity to the Broad Street pump (Musa et al., 2013). Removing the pump led to an almost immediate end to new cases of cholera. These early mapping techniques proved useful in elucidating geospatial correlates of disease. While early medical geography was largely a descriptive science, the rapid evolution of GIS and geocoding technology has led to a shift toward an emphasis on analytical science and hypothesis testing as well as the production and support of health risk maps that function as decision support tools in public health planning and disaster response. GIS allows policymakers to better assess potential risk factors and prevent disease. Using appropriately designed health risk maps offers an invaluable approach to identify and map medically vulnerable populations, health outcomes, risk factors, and the relationship between them to address and tackle worldwide health care. Health organizations visualize, analyze, interpret, and display multivariate attributes contributing to complex health phenomena using GIS tools and Big Data (Musa et al., 2013; Parrott, Hopfer, Ghetian, & Lengerich, 2007). Next, we provide an introduction to the types of geographic risk maps most often used in public health planning.
A Primer on Traditional Cartographic Risk Maps
Traditional static maps fall into two broad categories: general-purpose/reference maps and thematic (i.e., statistical) maps (Brewer, 2016; GeoSWG, 2012; Slocum, McMaster, Kessler, & Howard, 2009). Reference maps may include maps that illustrate political boundaries (e.g., counties, states), census tract data, and topographic or physical maps. Thematic maps, on the other hand, illustrate the geographic distribution of a phenomenon, and these map types are often used to visually communicate health risk data such as mortality rates or disease diffusion (GeoSWG, 2012; Slocum et al., 2009). Thematic maps can be used in at least three ways: to provide specific information about particular locations, to provide general information about spatial patterns and relationships, and to compare patterns on two or more maps (Slocum et al., 2009).
Types of Thematic (Statistical) Maps
Choosing the appropriate map type to communicate health risk derives from multiple factors, one of them being the type of (statistical) data being mapped (e.g., whether data consist of count data or rates, ranges or single values) (Krygier & Wood, 2011; Slocum et al., 2009). For density and rate data, choropleth maps are often the appropriate choice (Krygier & Wood, 2011; Slocum et al., 2009). Choropleth maps use different shades of a color for data, which are grouped into distinct geographic boundaries (i.e., enumeration units) such as counties or states. When the purpose of a map includes communicating a good visual impression of change over space, then using a choropleth map is a good choice. Caution must be taken, however, to avoid giving false or spurious impressions of abrupt changes at geographic boundaries when there are none in the data being represented visually (Brewer, 2006; Brewer & Pickle, 2002). Choropleth maps use ranges per shade (typically using standardized data that takes the area of the enumeration unit into account). While various color schemes can be used, a logical scheme includes choosing darker shades of the same color to represent higher density or rate data. Choropleth maps (see Figure 1) are appropriate to use when values of a health phenomenon change abruptly at enumeration unit boundaries (e.g., state health data), when the map designer wants map users to focus on typical values for units, or when emphasizing how one county or state compares with another.
Dasymetric maps are an alternative to choropleth maps. Instead of mapping data so that geographic regions (enumeration units) are uniform, ancillary information is used to model the internal distribution of a phenomenon. Dasymetric maps can be used to identify fine-scale risk patterns of infectious and chronic disease and associated socioeconomic or environmental risk factors (Barrozo, Perez-Machado, Small, & Cabral-Miranda, 2016). For example, dasymetric maps are often used for mapping population density data. These types of maps may be used to improve health outcomes in a megacity. Dasymetric maps can provide a more accurate and detailed distribution of population data (compared to choropleth maps, which show uniform distributions within enumeration units) by using ancillary information to spatially disaggregate population within administrative units.
Mapping totals (raw data) or count data are best illustrated with graduated or proportional symbol maps, or dot maps—another type of map representation. Graduated symbol maps are constructed by scaling symbols in proportion to the magnitude of data occurring at point locations (reflecting numeric risk, e.g., size of circles) (Brewer, 2016; Krygier & Wood, 2011). In dot maps, there are classified and unclassified legends. Classified legends use a range instead of a concrete number per symbol and are easier to understand, while unclassified legends use concrete numbers, which are more difficult to match accurately (Slocum et al., 2009). Dot maps are ideal for representing total values and for representing sparse data or phenomena (Brewer, 2006). They can be helpful for quickly giving a sense of density or distribution of data. Dots usually have a value assigned to them in a legend; for example, one dot equals 100 cases. When using a computer program to map dots, it is best to break geographic regions into smaller sections so the dots accurately cluster where appropriate. Without this step, dots may randomly cluster where they should not, giving a false impression and potentially misleading map readers about concentrations of a health risk that are spurious rather than real.
Finally, isopleth or contour maps are used to depict smooth, continuous phenomena (e.g., ozone air quality) (Krygier & Wood, 2011; Slocum et al., 2009). Isopleth maps are created by interpolating a set of isolines between sample points of known values. They are a special type of isarithmic map, used as an alternative to choropleth maps when the data are part of a smooth, continuous three-dimensional phenomenon. Like choropleths, isopleth maps require standardized data (e.g., dividing raw totals by enumeration units). Unlike in choropleth maps, data in isopleths maps are not grouped to a predefined unit like a city district. Isopleth maps are ideal for showing gradual change over space (temperature maps showing heat waves and impacts on mortality would more appropriately be depicted as isopleths maps since temperature changes do not abruptly change with state boundaries). The choice of thematic map type is for the most part dictated by the nature of the underlying statistical data being mapped. Beyond considering the nature of the available data, a map’s purpose must be considered for constructing a clear message.
Using Maps to Communicate a Message
Often, we, as communicators, use maps to tell a story. Basic steps to designing maps with a clear message involve asking several questions. Slocum (2009) suggests the following when designing and using a map:
1. Consider what real-world distribution of the phenomenon look like.
2. Determine the purpose of the map and its intended audience.
3. Collect data appropriate for the map’s purpose.
4. Design and construct the map.
5. Determine whether users (the intended audience) find the map useful.
Constructing the map involves selecting appropriate symbols (using a dot rather than a choropleth map) but also selecting and positioning map elements (e.g., title, legend, source) so that the map is visually appealing and clear. Designing and constructing a map involves assessing questions such as:
1. Will the map be used to portray general or specific information?
2. What is the spatial dimension of the data (points, lines, or areal)?
3. At what level are the data measured (nominal, ordinal, interval, ratio)?
4. Is data standardization necessary?
5. How many attributes are to be mapped?
6. Is there a temporal component to the data?
7. What are the characteristics of the audience (general public or professionals/experts)?
Careful consideration of these questions will result in the creation and communication of a clearer map with an intended message. Many structural map features come into play to increase accuracy of understanding and interpreting risk maps. The types of symbols used to represent spatial data play an integral part in clearly communicating messages in maps.
Symbolization in Maps and Its Relevance for Communication
Concepts from the field of semiotics, the study of signs, explain how images convey meaning. A sign refers to something other than itself. Iconic signs (iconicity) support comprehension by resembling familiar objects or ideas (Chandler, 2007; Severtson, 2015). Map symbols fall on a continuum of iconicity from arbitrary symbols (non-iconic) to pictorial icons that resemble real-world objects (on the highly iconic continuum) (Robinson & Petchenik, 1975). Cartographers have formalized symbol-referent relationships by developing typologies of symbol categories and rules for matching these symbol categories with geographic features (referred to as map syntactics) (MacEachren, 1995). These typologies are devised on the basis of evidence concerning human vision and visual cognition. Map sign systems are intended to match with categories such that users are able to distinguish categories that are similar to one another from those that are different and distinguish, for example, global from local patterns. Bertin (1983) is recognized as the first cartographer to formally propose fundamental visual variables to include geographic position on the plan, size, (color) value, texture, color (hue), orientation, and shape. Linguistic codes (through map legends) serve the role of “interpreter” between the unique semiological system of individual map and the cultural universal system of language. At the level of an entire map, Bertin’s approach to symbolization allows map designers to build the appearance of similarity or difference across kinds of information. Consideration of map symbols and semiotics familiarizes not only the rules by which mapmakers should design maps but to recognize that map users bring meaning to interpreting maps.
To support comprehension of maps, visual variables that symbolize data should conceptually match the level of measurement depicted in data (Krygier & Wood, 2011; Slocum et al., 2009). For example, the symbol size used (when using graduated symbol maps) should enable the map user to visually deduce numerical relationships (e.g., larger symbols reflect larger magnitude). Another example is using shading of one color from light to dark for continuous data (with darker shades symbolizing higher density or prevalence). Yet another example might include careful consideration of the type of data being mapped. For smooth versus continuous data (e.g., yearly health data), a contour rather than a choropleth map is appropriate. Symbol choice in pictographs should always aim to support comprehension. Visual variables for qualitative phenomena often use pictographic symbols, which are intended to look like phenomena being mapped.
Pictograms make use of shape for data representation. Pictographic symbols can be understood on some level without consulting a legend. For example, if there were a map that displayed wildfires in the last year, the reader of the map would understand that wildfires are being mapped if a fire symbol is used in the locations where there was a wildfire last year (Slocum et al., 2009). In pictograms, using symbol conventions for a map topic can effectively communicate the map message with the caveat of not using too many. When using multiple symbols, using hue with shape is a helpful way to distinguish qualitative features. Compact shapes are easier to associate with locations than tall or wide shapes. Symbols should always be readily identifiable and look markedly different from one another. If the function of symbols is to be seen as groups, use color and lightness to help identify symbol clusters and always test shapes and pictograms being used with intended audiences (Brewer, 2016).
Maps excel at showing relationships between data distributions. Therefore, symbols can represent multiple variables by combining perceptual dimensions of color—hue, lightness, and saturation—in arrangements together with shape.
The Use of Color in Static and Interactive Online Health Risk Maps
Color in maps engages audiences and peaks interest. Color can be used strategically to bring attention to different patterns of data (Brewer, 2016; Brewer, MacEachren, & Pickle, 1997; Retchless, 2014; Severtson & Vatovec, 2012). Choosing colors to symbolize data goes beyond considering what colors might be related to map topic (e.g., blue for bodies of water, brown for pollution). Professional tools can guide choosing color combinations so that they match the purpose of the map and the data being portrayed. A free color guide tool, ColorBrewer2.0©, is available. This color aid tool, developed by geographer Cynthia Brewer, has been integrated into most GIS graphics programs, in particular ESRI, the leading online GIS mapping tool.
Colors display differently depending on the presentation medium. Consequently, when choosing color combinations, consideration of the medium plays an important role. GIS and graphics software offer a variety of color systems for specifying color in maps such as HSV (hue-saturation-value), CMYK (cyan, magenta, yellow, black), and RGB (red, green, blue) (Brewer, 2016; Peterson, 2012). The HSV system makes partial use of the human visual perceptual system, whereas CMYK is the language of graphic arts (and printing maps). When designing a screen display (e.g., for a computer graphics), RGB is the main color system. Map designers who produce quality maps consider the following aspects: (a) perceptual dimensions of hue, lightness, and saturation, (b) perceptual color systems and their relationship to HSV and color mixture cubes, and (c) how to mix color to create map symbols using CMYK and RGB (Brewer, 2016).
Hue, lightness, and saturation, three perceptual dimensions of color, can be used strategically to emphasize statistical data. Hue is the perceptual dimension we associate with color names (e.g., red, yellow) and is defined as the dominant wavelength of light making up color (Brewer, 2016). In thematic maps, hue is used for analytical purposes to highlight numerical rates or proportions showing differences in magnitude and concentration by geographic area (e.g., areas that lost population shown in orange versus areas gained population shown in purple). When using color in this analytical way, it functions as a symbol to represent a phenomenon.
Color can be an asset or disadvantage when it comes to creating maps. Thoughtful use of color (hue) can substantially increase comprehension (Brewer, 2016; Brewer et al., 1997; Brewer & Pickle, 2002; Retchless, 2014). By contrast, use of too many hue shades (e.g., more than seven) should be avoided since too many can decrease comprehension (Brewer, 2006; Severtson, 2013; Severtson & Myers, 2013). More complex rules for choosing color combinations are provided in detail by ColorBrewer 2.0.
Other dimensions of color such as shades or lightness (of one color) also come into play for accurately interpreting geographic health data. Variation in lightness is the most important of the three color dimensions—hue, lightness, saturation—for representing quantitative data. Lightness is frequently used to represent a ranking within mapped data (Brewer, 2016; Severtson, 2015). Light colors typically are associated with low values and dark ones with high values. The lightness change communicates the low-to-high sequence or change in values. Lightness can be changed using the same hue or different hues.
Saturation is a measure of the vividness or intensity of color. Changing saturation together with lightness impacts contrast (if this is a desired effect). Saturation changes can be used to reinforce or augment lightness changes. Colors that are accidently more vivid (greater saturation) will stand out strongly from others, which can be unintentional yet misleading for interpreting maps. Inappropriate emphasis on map categories that are not important is a phenomenon that needs to be attended to. This translates to caution in choosing colors for large map areas; large areas of high saturation will dominate the look of the map. Strong saturation is appropriate for highlighting small features of a map (Brewer, 2016).
For clear map communication, perceptual structuring of colors should correspond with logical structuring inherent in data. Data can be sequential, diverging, or qualitatively arranged (Brewer, 2016). These color schemes are used to appropriately symbolize types of data. Sequential schemes reflect when lightness is used to represent ordered data. For example, a map showing forest fire hazard risk might represent high fire risk in a dark red color with lighter red reflecting lighter fire severity risk. This is a sequential scheme. Combining hue and lightness offers a strategy for providing contrast to help readers distinguish differences. Careful use of hue and lightness can make maps showing change easier to understand by using diverging schemes. Diverging schemes place equal emphasis on extremes and on rates near the median and are therefore useful for bringing attention to high and low extremes of data. Advantages of using a diverging color scheme include bringing attention to clusters. While a sequential color scheme may emphasize the range of values that exist, a diverging scheme may emphasize how extremes differ from the midpoint value. Finally, qualitative color schemes represent map features or categories that are not ordered (e.g., different languages spoken in a neighborhood or types of government spending such as military, education, health care). A well-designed qualitative scheme will not suggest that data is ordered or that one category is more important than another (Brewer, 2016). When using qualitative color schemes, hues used should have similar contrast across the spectrum. Whether using sequential, diverging, or qualitative color schemes to show changes in health data, investing time and pilot testing color schemes is worth the extra effort (Brewer et al., 1997).
Audience characteristics come into play and need to be factored in when producing color risk maps. For instance, those who are older or color-blind may have a difficult time interpreting colors in maps. It is estimated that at least 8% of men and fewer than 1% of women in the United States are color-blind–red-green color-blind (Krygier & Wood, 2011). For these audiences, using multiple hues can become problematic. People who are color-blind can see many hues, but there are predictable groupings of hues that will be confused with each other (they appear often as shades of beige). The extent of color confusion depends on the severity of a person’s color vision deficiency. The range of hues from red to orange, brown, yellow, and green may all look similar if they are similar in lightness. For these audiences, certain color groupings will be confusing, such as spectral (rainbow) and “stoplight” (red-yellow-green) schemes. Hue pairs that work well as anchors in diverging schemes for color blind people include red-blue, red-purple, orange-blue, orange-purple, brown-blue, brown-purple, yellow-blue, yellow-purple, yellow-gray, and blue-gray, and, if using a spectral (rainbow) scheme, avoiding greens (Brewer, 2006). In any application where maps are to be used by a wide range of readers, color schemes should accommodate the color-blind. ColorBrewer 2.0 offers advice on color schemes that work best for the color-blind. Older map viewers, who according to the 2010 U.S. Census make up at least 16% of the population, may benefit from viewing maps produced with saturated colors in which the greater intensity and vividness provides greater contrast.
While color as structural message feature plays a role in influencing impressions of map information, data classing is another basic decision that affects how information in a map is comprehended.
How Statistical Data are Grouped and Cartographically Represented
There are numerous methods for grouping or classing data. Data classing has the potential to create dramatically different visual structures/patterns of data. GIS and mapping programs offer a selection of techniques that often include grouping data by quantiles, equal intervals, or the Jenks optimized method. Other choices include classing by standard deviations and minimizing differences across boundaries. Quantile classing assigns the same number of enumeration (geographic) units to each class. By contrast, equal interval classing breaks the data range into equal segments for predictable and equal class ranges. Jenks methods minimize variation within classes and maximizes variation between classes (Dent, Torguson, & Hodler, 1999). With this approach, enumeration units that share a color are statistically more similar to each other than other units in other color classes. Cartographers most commonly choose a Jenks method for their first look at data in maps (e.g., when exploring etiology of phenomena). Epidemiologists by contrast typically choose quantile classing to view different types of standardization and age adjustments (Brewer, 2006).
Designing Maps for a Purpose
Brewer (2016) discusses the importance of designing maps for a purpose. This encompasses consideration of what information will be mapped, who is the audience, how the map content will be coordinated with accompanying text and graphics (e.g., inlays of smaller maps or other graphics); size, prominence, and placement of title, subtitles, and legends; orientation (e.g., showing North); and use of white space. With this in mind, a map’s purpose determines which elements are the most important and should be displayed most prominently in the visual hierarchy (Herrmann & Pickle, 1996). Title and key features of the map are highest in the visual hierarchy and should be displayed most prominently, while supporting information (also referred to as marginal elements) are lower in the visual hierarchy and should take balancing white space into consideration (Brewer, 2016). The map title should present the basic topic and invite the reader to investigate further. Layout of title, subtitles, legends, and scale as well as any smaller-scale location maps or inset maps showing detail need to be tested for how they are organized to avoid a cluttered and confusing map. A best practice is to use only one or two font types on a map, applying various combinations of style, size, and color to achieve a visual hierarchy (GeoSWG, 2012). Care should also be taken to test whether font types are legible against background tints.
Considerations of Culture When Designing Maps
Color may have cultural, personal, or emotional meanings. A communication perspective emphasizes the need to take the intended audience into consideration. Colors can culturally symbolize various physical features. However, colors can have different cultural connotations depending on a world region. For example, black symbolizes mourning in the West, white symbolizes mourning in China, and brown symbolizes mourning in India (Krygier & Wood, 2011). Depending on the area of the world, the direction in which reading occurs is how that audience tends to read a map. For example, if the primary audience is from the West, most likely they will read the map from left to right, top to bottom (Krygier & Wood, 2011). Marginal map elements such as scale bars and direction indicators orient the map reader regardless of culture.
Groupings of color allow maps to be interpreted for a general overview as well as by category. One way to choose color for qualitative categories is to use colors that already associate with a category (e.g., red or orange for fire). But caution is to be taken for implicit colors to function as symbols of categories. The meaning of colors differs depending on a culture and on individual characteristics. Be alert to color associations that may be offensive. For example, when mapping ethnicities, avoid using literal uses of color to associate with ethnicities (e.g., black for African American, yellow for Asian, red for American Indian). Such a superficial and exaggerated emphasis on skin color associations for groups is likely to offend readers. Use an abstract set of hues for mapping race groups.
Map scale has an important role to play in map design. Map scale can refer to geographic scale, cartographic scale, or data resolution. Geographic scale refers to area covered, while cartographic scale refers to the relationship between a map and Earth distance (referred to as representative fraction, or RF, in geography), and, finally, data resolution refers to the granularity of data (or level, e.g., acquiring data at various resolutions such as census blocks).
Map scale can be communicated in three ways: as a graphic bar scale, a verbal scale, and an absolute scale (RF) (Brewer, 2016). The graphic bar scale remains the most accurate in dynamic, online map environments, in which map size varies with zooming, screen resolution, and reproduction. For thematic (statistical) maps, simple scale bars that encourage approximate distance estimates and do not distract attention from the map message communicate effectively. A good scale presents rounded units that the map reader can easily use. For online environments, with interactive maps that zoom, a competent web service shows changes in scale bar lengths as scale become dramatically larger or smaller. Especially in the context of online maps, as a map zooms out to cover a larger area, some features need to disappear or change size to keep the map from getting too cluttered. Map designers nowadays have tools that help in deciding which features to include at given scales and how, for example, thickness of lines and size of symbols and text change to align with keeping maps legible.
Legends present information that allows the reader to interpret the risk visually and geographically portrayed in the map and to understand its symbols. Title usually include what, where, and when components, while the legend assists with interpreting base information, explaining map symbols, data ranges, and the level at which data have been aggregated (Brewer, 2016; Krygier & Wood, 2011). The basic types of thematic maps have fairly conventional legend content. GIS software provides some of the content within legend tools. How legends need to be designed to communicate effectively may differ depending on the type of map.
For choropleth maps (maps that use shades of color to denote a phenomena), legends state the data range for each color or pattern used. Other decisions for choropleth maps include whether to (a) round numbers for breaks or within labels, (b) increment labels (e.g., 0–10 and 10–20 or 0–9 and 10–19), (c) use the word “to” or a dash, (d) label breaks between classes with single numbers rather than a class range, (e) order classes with the highest number at the top or at the bottom of the legend, (f) label ranges with the actual value represented by the symbol with gaps in between ranges, (g) use statements such as “fewer than 100 people” or “more than 150 percent” for extreme ranges, and (h) add annotations to explain the meaning of class breaks or to assist in map interpretation (Brewer, 2016).
For maps that depict qualitative data, it is important to provide descriptive labels for each legend color or pattern. Creating logical groupings for related categories assists in map reading. For area symbols, colors should be presented in the same way in the legend as they appear on the map. For example, if map polygons (designated geographic area units) are used in the map with outlines, then the same outline color and weight should be used in the legend. Attention should be made to represent colors exactly the same way in the legend as in the map. This includes paying attention to spacing between colors and duplicating exactly how the colors are represented in the map—whether they are differentiated by space or by a colored line (Brewer, 2016).
For dot density maps, legends define the amount that one dot represents (how concentrated in a given space). Dot map symbols’ purpose is to illustrate density. Dot maps are improved when one can show density examples in the legend (Brewer, 2016; Slocum et al., 2009). Density boxes may need to be manually constructed. For example, in the legend it could state that one dot represents 80 people and then provide sample density boxes. For proportioned symbol maps, it is best in the legend to present smallest and largest symbol sizes seen in the map (e.g., circles) and include one or more intermediate size symbols in order for the map reader to interpolate between extremes. If graduated rather than proportioned point symbols are used, this should be clearly stated in the legend—that sizes do not represent precise data values. Segmented symbols such as pie charts or bar graphs require custom legend design. Two-part legends can facilitate explaining multivariate symbols used in maps. For example, a legend can explain multivariate relationships by showing both segmented bar graphs and what the different heights mean and then also show what different colors represent.
Legends are used to explain symbols to a reader. Within legends, the symbol should be on the left and the explanation should be on the right (Slocum et al., 2009). This is because most people read maps from left to right and seeing a symbol first helps the reader understand the map. A heading on the legend helps the reader understand the theme of the map. At the bottom or below the legend should be a citation of the source of the data mapped, formatted in the following way, “Source: Citation.” Legends together with the title and overall spacing of the basic elements as well as use of white space contribute to the overall visual hierarchy and sense of balance, which facilitates cognitive processing of maps (Slocum et al., 2009).
Cognitive Processing of Risk Maps
A well-designed map that communicates effectively will avoid cognitive overload (Oviatt,Coulston & Lunsford, 2004; Paas,Renkl & Sweller, 2003). Dual-coding theory (Paivio, 1990) was the first seminal work to propose that visual information processing occurs distinctly and independently from cognitive systems processing verbal information in working memory (conscious, short term memory). A substantial amount of empirical research has supported this theory since its claim was first made. Criticisms of Paivio’s work questioned whether dual-coding applies to processing more complex information, since his work on visual processing was mostly of simple tasks. Dual-coding rests on the belief that there is limited working memory with strategies focused on how to expand working memory. Sweller’s cognitive load theory (CLT; Paas et al., 2003) makes explicit the assumption of limited working capacity in Paivio’s dual-coding theory. Much research on map-based message design has been driven by the assumption, and even paradigm, of limited working memory. Thus, underlying efforts to design structural map features should seek to eliminate any avoidable load on working memory. Paivio’s dual-coding research and Sweller’s CLT have since been refined and extended by several researchers.
Notable research in the recent decade and in map design as it pertains to health includes that by Severtson, who has laid out an integrated representational and behavioral framework for understanding how structural features of health risk maps may influence audience interpretation and decision behavior (see Figure 2) (Severtson, 2013; Severtson & Burt, 2012; Severtson & Vatovec, 2012). People derive meaning from what they see. For a visual representation (a map), visual cognition involves how seeing an image is related to the meaning derived from what is seen. This process is explained by “top-down” and “bottom up” information processing (Severtson, 2013; Severtson & Burt, 2012). “Top-down” processing is when a reader uses his or her perception, thought process, prior experiences, and memory to interpret a map (in a more deliberate way). For example, one can see a local map of hazards and deliberately look for hazard near their home (e.g., how close this risk is to where I reside, how personally relevant this risk is to me). “Bottom-up” processing occurs subconsciously when people interpret a map (i.e., without conscious effort). This process uses perception, which is a reader’s initial reaction to map symbols with little thought (Severtson, 2013; Slocum et al., 2009).
An illustration of “top-down” processing is captured in a study of Florida tourists’ exposure and reaction to hurricane warning messages (Villegas et al., 2013). Tourists use their location, hurricane traits, knowledge, and previous experience with hurricanes and other tourism-related factors to evaluate their personal risk and need to evacuate. As the likelihood of evacuation increases based on data shown in the map, fear and risk perception increase, especially for those by the coast. The fact that study participants took a more calculated approach integrating that they were in the coastal area suggests a top-down processing approach. People consciously take into consideration hazards around them rather than subconsciously understanding the map (Villegas et al., 2013).
Bottom-up processing, by contrast, occurs because our visual system is neurologically connected to cognition areas in the brain such that “seeing” is literally linked with “knowing” and enhances comprehension by freeing short-term memory for other processing needs. Preconscious bottom-up processing occurs without apparent cognitive effort (Severtson & Myers, 2013). Bottom-up processing is likely influenced to a greater degree by the visual salience of structural map features than top-down processing. Bottom-up processing is also likely automatic and passive and driven by emotional responses and prior experience. In bottom-up processing, perception starts at the sensory input or the stimulus. Thus, properties of visual features can support comprehension (Cleveland & McGill, 1984, 1985).
Data Ownership in Mapping Population Health Data
Federal, state, and municipal public health agencies create and produce decision support maps from health data registries to aid in not only epidemiological surveillance and monitoring, but also resource allocation decision-making and generating hypotheses related to public health threats. Legislation regulates data acquisition, maintenance, security, use, and disclosure of registry data. Entities carry the responsibility of abiding by the Privacy Rule when creating, using, and mapping health registry data (Gliklich, Dreyer, & Leavy, 2014; Van Panhuis et al., 2014). The Privacy Rule encompasses the Health Insurance Portability and Accountability Act (HIPAA) and its implementing regulations that address data ownership and disclosure. The key for map producers is to visually include data sources in maps for credibility purposes (Hopfer, Chadwick, Parrott, Ghetian, & Lengerich, 2009; Parrott et al., 2007).
Within the United States, only one state, New Hampshire, has laws that state health data belongs to the individual (“Who Owns Medical Records: 50 State Comparison,” 2015). Twenty-one states have laws that health data belong to the hospital or physician that owns the record. The remaining 28 states do not specify who owns the health data. At the current time, while such claims of ownership are plausible, none have been legally tested or recognized. Entities that could claim ownership include health care providers, insurance plans, funding agencies for registry projects, research institutions, and government agencies.
With increasing collection of “big data,” the field of geovisualization brings an entirely new dynamic to health risk mapping, presenting an opportunity to augment the function of risk maps as decision support tools to communicate dynamic risk information. While we do not have the space for a section on big data in this article, we do comment on the use of big data for health as an emerging field as well as the increasing use of social media data to map dynamic, participatory grassroots or “on-the-ground” data for public health applications.
Geo-Social Data in Online Risk Maps
The interactivity of online maps presents an opportunity to communicate dynamic and changing risks such as wildfire, hurricane, or disaster response and many more types of response but to map input from local participants who may be experiencing public health threats directly (Parrott et al., 2010). Interactive online maps make room for incorporating and communicating local, on-site knowledge. Incorporating local voices and local expert knowledge not only improves information (or decreases/distorts accuracy) but, importantly, has the potential to increase trust in risk maps when used to communicate anticipated risks to communities (Parrott et al., 2010; Shook & Turner, 2016).
Social media offer an alternative data source that can reveal spatio-temporal differences in perception of risk threat to improve risk communication systems as well as a way to build social capital (Shook & Turner, 2016). For example, Shook and Turner (2016) leveraged Twitter data to map response to an extreme heat event in the Northeast United States as it unfolded. In this context, geo-social data allow for the spatial and temporal reconstruction of individual-level reaction to events whether it is a heat event, a snowstorm, or a wildfire. Addition of geo-social data augments the risk map with localized response to which public health response efforts can tailor their efforts. This calls attention to the multiple functions of online interactive and dynamic risk maps not only to communicate in one direction (from public health agencies to the public) but also to receive, integrate, and respond specifically to community messages and involvement in collaborative public health responses to dynamic public health events (Luo & MacEachren, 2014; MacEachren, 1995).
The use of annotations (e.g., photographs as notes) on geographic risk maps can also function as geo-social data but also, more specifically, as conversational collaborative annotations in maps to update information. Furthermore, annotated information offers a strategy for tailoring spatial information to suit the needs of different audiences. Community interactions with hurricane evacuation maps (Allen, Sanchagrin, & McLeod, 2013) or volcano eruption evacuation emergency maps (Haynes, Barclay, & Pidgeon, 2007) have been shown to facilitate spatial orientation and aid in the comprehension of risk communication particularly for “lay” audiences who are less versed in reading technical thematic risk maps. Photographs on maps have been shown to function as conversational annotations among collaborators and communities responding collectively to a risk event (Hopfer & MacEachren, 2007). Annotated photographs can effectively communicate spatial risk by aiding map readers in location orientation (Haynes et al., 2007). For individuals less familiar with reading geographic risk maps, annotated photographs have the potential, if readily recognizable, to facilitate orientation of the map user and draw attention to the map. Such visual conversational annotations (structural map features) have increased effectiveness of risk communication efforts by facilitating map readers’ ability to rapidly evaluate and comprehend efficient and safe evacuation routes for anticipated flooding (Villegas et al., 2013) or erupting volcanoes (Haynes et al., 2007).
Social processes and responses that take place within particular “geographic” contexts and map the dynamic information flow can propel or hinder ideas and behaviors in an interactive map. Maps as visual analytic tool can reflect shifts of influence from geographically and socially close community interactions to networks that are socially near but geographically dispersed (Luo & MacEachren, 2014). For example, while maps can visualize an outbreak that is specific to a geographic area, future predictive tools will be able to integrate social information that facilitates adaptive responses. These interactive online maps that feature geo-social data may facilitate the two-way communicative response needed for addressing acute events (e.g., hurricane or fire evacuation) but also chronic public health problems to improve community public health (e.g., preventing obesity). Physical activity (PA) neighborhood walking maps increase daily walking by providing suggestions for walking paths and informing community members that these paths are safe (McNeill & Emmons, 2012). Interactive maps also alert communities to the potential risks of looming unanticipated sudden health threats like the potential of a gas leak explosion. Such a map can be augmented with geo-social data providing localized, real-time updates. The increasing use of online interactive maps consumed by the public is evident by media outlets providing educational features on the topic. An example is the New York Times special section, “All Over the Map: 10 Ways to Teach About Geography”. The news agency illustrates new applications of geo-social data to bring attention and communicate about population health risks.
Risk Maps for Public Health Planning
Louis and colleagues (2014) review a typology of risk maps for public health planning. Risk maps are categorized along a continuum of descriptive, validative, predictive, and early warning system (EWS) maps. While descriptive maps are useful for identifying high-risk hotspot areas already affected, predictive maps anticipate affected areas assisting in capacity planning to mitigate adverse public health effects. The majority (73%) of maps described in the literature—planning for dengue risk—functioned as descriptive maps. These maps describe retrospective geographic occurrence of a phenomenon. Descriptive maps help public health officials estimate geographic hotspots or an increased occurrence of a health risk over time. Fewer researchers (reported studies) used validated risk maps (35%), with an equal amount (35%) using predictive maps to assess dengue risk for the population (Louis et al., 2014). Observed and predicted data are modeled in these maps for evaluating, predicting, and planning for future geographic dengue hotspots. Only 11.5% of risk maps used EWS-establishing criteria for early recognition of an anticipated disease outbreak with direct public health application. Frameworks for evaluating early warning systems (Murphy, Packer, Stevens, & Simpson, 2007) as well as communication functions (Ebi, 2014) serve to facilitate comprehension in maps. Accompanying the complex modeling used in EWS risk maps is the increased amount of uncertainty accompanying the data.
Visualizing uncertainty and understanding how uncertainty visualization influences reasoning and decision-making is a current active research area among scholars in the field of geovisualization (and has been for several decades) (Kinkeldey, MacEachren, Riveiro, & Schiewe, 2015; Kinkeldey et al., 2014; MacEachren et al., 2005). Uncertainty has been an umbrella term for concepts that include inaccuracy, imprecision, ambiguity, vagueness, subjectivity, and error. One of the challenges is to better understand the effects of visually depicting uncertainty in maps on reasoning skills. Generalized findings and practices are still rare (Kinkeldey et al., 2015). The first lesson is that ignoring uncertainty can have severe consequences when the map is used for public health response decisions and users assume information in the map is visually depicted with certainty. A second lesson, although the evidence is tenuous, is that visually communicating uncertainty in maps has the potential to increase trust in results (Fisher, Popov, Drucker, & Schraefel, 2012). Even relatively simple representations of uncertainty using error bars progressively updating over time allow analysts to trust their decision points to a greater degree (Fisher et al., 2012). In some research, visualizing uncertainty has not impacted decision accuracy but has impacted decision speed—increasing it (Finger & Bisantz, 2002). Graphic displays of uncertainty outperformed textual (verbal) descriptions of uncertainty (Kirschbaum, MacEachren, & Schiewe, 2014). Features portraying uncertainty in maps may matter more with complex issues like cancer risk from air pollution (Kinkeldey et al., 2015; Severtson, 2015; Severtson & Myers, 2013). MacEachren et al. (2005) lay out a typology of uncertainty frameworks visualizing various types of uncertainty. Among the examples, they outline uncertainty depiction using the metaphor of fog obscuring one’s view. Data points with less certainty are depicted visually as less clear. Assessing the utility of visual uncertainty portrayal for decision-makers using risk maps is an ongoing active research program and challenge.
Visualizing Policy Choices and Consequences
Visualizing the consequences of varying public health investment and capacity choices may help make clear potential outcomes of complicated decisions and choices for policymakers and the public. For example, Climate Central (CC), an independent organization of scientists and journalists, offers an online mapping interface that provides community risk profiles in the form of maps. CC visualizes the impacts of policy choices, i.e., impacts of unchecked pollution on health. Such mapping organizations provide regional summaries that integrate key findings, methods, and interpretation in risk maps. CC maps show how community vulnerabilities may intersect with social vulnerabilities and population densities, showing how threats to health may differ from place to place (Tebaldi, 2014). State health departments increasingly adopt online map tools not only for internal resource planning and decision-making but also for communicating with the public and policymakers.
New diseases and epidemics will continue to emerge and spread through the world’ s population. The use of risk maps as communication tools to address global and local emerging public health challenges continues. Structural map features such as color choice, data classing, symbol types, legends, visualizing levels of uncertainty inherent in data, and annotated geo-social data play an important role in communicating clear messages of these dynamic health risks. Structural features play an important role in increasing comprehension of map messages and must be carefully considered given their impact, especially in public health planning and resource allocation decision-making. A primer on traditional cartographic thematic maps reviews the importance of matching data type with map choice. Moreover, an important take-home message from reviewing color choices is the importance of investing time and pilot testing color choices for an intended map purpose, message, medium, and audience. A carefully selected cartographic framework can guide prioritization of communication variables when visually depicting health risks to which an audience responds and with which it interacts to generate both research ideas and practical guidelines (Kostelnick, McDermott, Rowley, & Byunnyfield, 2013; MacEachren, 1995; MacEachren, Brewer, & Pickle, 1998; Severtson, 2013). Cartographic principles grounded in cognitive and semiotic principles guide map designers in communicating visual geographic risk messages (Brewer, 2016). Important in communicating risk is acknowledging the integrity of available data being visualized as well as attending to how missing data is acknowledged and represented (GeoSWG, 2012).
Maps used as surveillance, planning, or predictive tools will continue to be used and hold great promise (Hopfer et al., 2009; Parrott et al., 2007) but will only be as valuable as the quality of the data being used to produce maps. There will continue to be a need for increasing resources devoted to collecting the data needed to accurately map and predict anticipated public health threats, whether communicable or non-communicable, as well as acute response demands. The novelty of contemporary interactive maps is their ability to visualize dynamic (changing) information and to visualize the consequences for populations of potential policy choices (CDPH, 2016; Gould & Rudolph, 2015). State health departments planning for mitigation and adaptation to public health threats can and are developing community- and county-level vulnerability profiles (CDPH, 2016). Any design of structural map features needs to consider whether goals include distinguishing global from local specific risks in maps, comparing different maps, or understanding geographic/spatial changes visually depicted in maps. Furthermore, designers of maps for public health surveillance and planning need to understand the consequences of visually depicting structural features in varying ways to show changes over time across geographic space (Parrott et al., 2010, 2007). Pilot testing how the intended audience understands, receives, and responds to the primary intended message of a map remains an important consideration. Is the map product for a state report, made available online for the public at large? Will it be printed or online? Is the product best presented as static or interactive? These are considerations that factor into how map design questions are ultimately approached.
Further reading on evaluating future directions and considerations of structural features of risk maps are recommended. For theoretical perspectives on basic map design elements, Cynthia Brewer’s Designing Better Maps (2016) covers many fundamental structural elements of map design, including considerations of map purpose, audience, medium, visual hierarchy, color choice, symbols, legends, and much more. Fundamental cartographic principles are reviewed in classic textbooks such as Thematic Cartography and Geovisualization (Slocum, McMaster, Kessler, & Howard, 2009).
Additional cartography textbooks also include Cartography (Dent, Torguson, & Hodler, 1999) and Making Maps (Krygier and Wood, 2011). Finally, for theoretical musings on the various roles of map function (e.g., as communication tool, as process for generating etiologic hypotheses, as medium, as evidence form), a good read and seminal text is Alan MacEachren’s How Maps Work (1995). More recent theoretical frameworks of geovisualization in the context of climate change risk include Kostelnich, McDermott, Rowley, and Bunnyfield’s A Cartographic Framework for Visualizing Risk (2013), Luo and MacEachren’s “Geo-Social Visual Analytics” (2014), and Severtson’s integrated representational and behavioral framework (Severtson, 2013; Severtson & Vatovec, 2012).
Links to Public Health Risk Maps
Brewer, C. (2016). Designing better maps: A guide for GIS users (2d ed.). Redlands, CA: ESRI.Find this resource:
Krygier, J., & Wood, D. (2011). Making maps: A visual guide to map design for GIS. New York: Guilford.Find this resource:
MacEachren, A. M. (1995). How maps work. New York: Guildford.Find this resource:
Monmonier, M. (2014). How to lie with maps (2d ed.). Chicago: University of Chicago Press.Find this resource:
Shah, P., & Miyake, A. (2005) The Cambridge handbook of visuospatial thinking. Cambridge, U.K.: Cambridge University Press.Find this resource:
Slocum, T. A., McMaster, Kessler, F. C., & Howard, H. H. (2009). Thematic cartography and geovisualization (3d ed.). Upper Saddle River, NJ: Prentice Hall.Find this resource:
Allen, T. A., Sanchagrin, S., & McLeod, G. (2013). Visualization for hurricane storm surge risk awareness and emergency communication. In J. Tiefenbacher (Ed.), Approaches to disaster management—Examining the implications of hazards, emergencies, and disasters. InTech.Find this resource:
Barrozo, L. V., Perez-Machado, R. P., Small, C., & Cabral-Miranda, W. (2016). Changing spatial perception: Dasymetric mapping to improve analysis of health outcomes in a megacity. Journal of Maps, 12, 1242–1247.Find this resource:
Bertin, J. (1983). Semiology of graphics: Diagrams, networks, maps. Madison: University of Wisconsin Press.Find this resource:
Brewer, C. A. (2006). Basic mapping principles for visualizing cancer data using geographic information systems (GIS). American Journal of Preventive Medicine, 30, S25–S36.Find this resource:
Brewer, C. A. (2016). Designing better maps: A guide for GIS users. Redlands, CA: ESRI Press.Find this resource:
Brewer, C. A., MacEachren, A. M., & Pickle, L. W. (1997). Mapping mortality: Evaluating color schemes for choropleth maps. Annals of the Association of American Geographers, 87(3), 411–438.Find this resource:
Brewer, C. A., & Pickle, L. W. (2002). Evaluation methods for classifying epidemiological data on choropleth maps in series. Annals of the Association of American Geographers, 92(4), 662–681.Find this resource:
CDPH. (2016). Climate change and health profile report. Davis, CA: California Department of Public Health.Find this resource:
Chandler, D. (2007). Semiotics: The basics (2d ed.). London: Routledge.Find this resource:
Clarke, K. C., McLafferty, S. L., & Tempalski, B. J. (1996). On epidemiology and geographic information systems: A review and discussion of future directions. Emerging Infectious Disease, 2, 85–92.Find this resource:
Cleveland, W. S., & McGill, R. (1984). Graphical perception: Theory, experimentation and application to the development of graphical methods. Journal of American Statistical Association, 229, 828–833.Find this resource:
Cleveland, W. S., & McGill, R. (1985). Graphical perception and graphical methods for analyzing scientific data. Science, 229, 828–833.Find this resource:
Croner, C. M. (2003). Public health, GIS, and the Internet. Annual Review of Public Health, 24, 57–82.Find this resource:
Dent, B., Torguson, J. S., & Hodler, T. W. (1999). Cartography: Thematic map design (6th ed.). Boston: McGraw-Hill.Find this resource:
Ebi, K. L. (2014). Health in the new scenarios for climate change research. International Journal of Environmental Research and Public Health, 11, 30–46.Find this resource:
Finger, R., & Bisantz, A. M. (2002). Utilizing graphical formats to convey uncertainty in a decision-making task. Theoretical Issues in Ergonomic Science, 3, 1–25.Find this resource:
Fisher, D., Popov, I. O., Drucker, S. M., & Schraefel, M. C. (2012). Trust me, I’m partially right: Incremental visualization lets analysts explore large datasets faster. Paper presented at the Proceedings of the 2012 Conference on Human Factors in Computing Systems, New York.Find this resource:
GeoSWG. (2012). Cartographic guidelines for public health. Atlanta: Centers for Disease Control and Prevention.Find this resource:
Gliklich, R. E., Dreyer, N. A., & Leave, M. B. (2014). Principles of registry ethics, data ownership, and privacy. In Registries for evaluating patient outcomes: A user’s guide (Internet). 3d ed. Rockville, MD: Agency for Healthcare Research and Quality (AHRQ).Find this resource:
Gould, S., & Rudolph, L. (2015). Challenges and opportunities for advancing work on climate change and public health. International Journal of Environmental Research and Public Health, 12, 15649–15672.Find this resource:
Haynes, K., Barclay, J., & Pidgeon, N. (2007). Volcanic hazard communication using maps: An evaluation of their effectiveness. Bulletin of Volcanology, 70, 123–138.Find this resource:
Herrmann, D., & Pickle, L. W. (1996). A cognitive subtask model of statistical map reading. Visual Cognition, 3, 165–190.Find this resource:
Hopfer, S., Chadwick, A. E., Parrott, R. L., Ghetian, C. B., & Lengerich, E. J. (2009). Assessment of training needs and preferences for geographic information systems (GIS) mapping in state comprehensive cancer-control programs. Health Promotion and Practice, 10, 579–587.Find this resource:
Hopfer, S., & MacEachren, A. M. (2007). Leveraging the potential of geospatial annotations for group decisions: A communication theory perspective. International Journal of Geographic Information Systems, 8(21), 921–934.Find this resource:
Kinkeldey, C., MacEachren, A. M., & Schiewe, J. (2014). How to assess visual communication of uncertainty? A systematic review of geospatial uncertainty visualization user studies. The Cartographic Journal, 51, 372–386.Find this resource:
Kinkeldey, C., MacEachren, A. M., Riveiro, M., & Schiewe, J. (2015). Evaluating the effect of visually represented geodata uncertainty on decision-making: Systematic review, lessons learned, and recommendations. Cartography and Geographic Information Science, 1–21.Find this resource:
Kirschbaum, C. A. M., MacEachren, A. M., & Schiewe, J. (2014). How to assess visual communication of uncertainty? A systematic review of geospatial uncertainty visualization user studies. The Cartographic Journal 51, 372–386.Find this resource:
Kostelnick, J. C., McDermott, D., Rowley, R. J., & Byunnyfield, N. (2013). A cartographic framework for visualizing risk. Cartographica, 48, 200–224.Find this resource:
Krygier, J., & Wood, D. (2011). Making maps: A visual guide to map design for GIS (2d ed.). New York: Guilford.Find this resource:
Lindenbaum, S. (2006). GIS/Mapping: Introducing the case studies. San Francisco: Open Society Institute.Find this resource:
Louis, V. R., Phalkey, R., Horstick, O., Ratanawong, P., Wilder-Smith, A., Tozan, Y., et al. (2014). Modeling tools for dengue risk mapping—a systematic review. International Journal of Health Geography, 13, 50.Find this resource:
Luo, W., & MacEachren, A. M. (2014). Geo-social visual analytics. Journal of Spatial Information Science, 8, 27–66.Find this resource:
MacEachren, A. M. (1995). How maps work. New York: Guilford.Find this resource:
MacEachren, A. M., Brewer, C. A., & Pickle, L. W. (1998). Visualizing georeferenced data: Representing reliability of health statistics. Environment and Planning A, 30, 1547–1561.Find this resource:
MacEachren, A. M., Robinson, A., Hopper, S., Gardner, S., Murray, R., Gahegan, M., et al. (2005). Visualizing geospatial information uncertainty: What we know and what we need to know. Cartography and Geographic Information Science, 32, 139–160.Find this resource:
McNeill, L. H., & Emmons, K. (2012). GIS walking maps to promote physical activity in low-income public housing communities: A qualitative examination. Preventing Chronic Disease, 9.Find this resource:
Murphy, K., Packer, C., Stevens, A., & Simpson, S. (2007). Effective early warning systems for new and emerging health technologies: Developing an evaluation framework and an assessment of current systems. International Journal of Technology Health Care, 23, 324–330.Find this resource:
Musa, G. J., Chiang, P. H., Sylk, T., Bavley, R., Keating, W., Lakew, B., et al. (2013). Use of GIS mapping as a public health tool—from cholera to cancer. Health Services Insights, 6, 111–116.Find this resource:
Oviatt, S., Coulston, R., & Lunsford, R. (2004). When do we interact multimodally? Cognitive load and multimodal communication patterns. Proceedings of the 6th International Conference on Multimodal Interfaces, 129–136.Find this resource:
Paas, F., Renkl, A., & Sweller, J. (2003). Cognitive load theory and instructional design: Recent developments. Educational Psychologist, 38, 1–4.Find this resource:
Paivio, A. (1990). Mental representations: A dual coding approach. New York: Oxford University Press.Find this resource:
Parrott, R. L., Hopfer, S., Ghetian, C. B., & Lengerich, E. J. (2007). Mapping as visual health communication tool: Promises and dilemmas. Health Communication, 22(1), 13–24.Find this resource:
Parrott, R., Volkman, J. E., Lengerich, E. J., Ghetian, C. B., Chadwick, A. E., & Hopfer, S. (2010). Using geographic information systems to promote community involvement in comprehensive cancer control. Health Communication, 25, 276–285.Find this resource:
Peterson, G. N. (2012). Cartographer’s toolkit. Fort Collins, CO: PetersonGIS.Find this resource:
Retchless, D. P. (2014). Communicating climate change: Spatial analog versus color-banded isoline maps with and without accompanying test. Cartographic and Geographic Information Science, 41, 55–74.Find this resource:
Robinson, A., & Petchenik, B. B. (1975). The map as a communication system. Cartographic Journal, 12, 7–15.Find this resource:
Severtson, D. J. (2013). The influence of environmental hazard maps on risk beliefs, emotion, and health-related behavioral intentions. Research in Nursing & Health, 36, 330–348.Find this resource:
Severtson, D. J. (2015). Testing map features designed to convey the uncertainty of cancer risk: Insights gained from assessing judgments of information adequacy and communication goals. Science Communication, 37, 59–88.Find this resource:
Severtson, D. J., & Burt, J. E. (2012). The influence of mapped hazards on risk beliefs: A proximity-based modeling approach. Risk Analysis, 32, 259–280.Find this resource:
Severtson, D. J., & Myers, J. D. (2013). The influence of uncertain map features on risk beliefs and perceived ambiguity for maps of modeled cancer risk from air pollution. Risk Analysis, 33, 818–837.Find this resource:
Severtson, D. J., & Vatovec, C. (2012). The theory-based influence of map features on risk beliefs: Self reports of what is seen and understood for maps depicting an evironmental health hazard. Journal of Health Communication, 17, 836–856.Find this resource:
Shook, E., & Turner, V. K. (2016). The socio-environmental data explorer (SEDE): A social media-enhanced decision support system to explore risk perception to hazard events. Cartography and Geographic Information Science.Find this resource:
Slocum, T. A., McMaster, R. B., Kessler, F. C., & Howard, H. H. (2009). Thematic cartography and geovisualization (3d ed.). Upple Saddle River, NJ: Prentice Hall.Find this resource:
Tebaldi, C. (2014). Climate central. Available at http://climatecentral.orgFind this resource:
Van Panhuis, W. G., Paul, P., Emerson, C., Grefenstette, J., Wilder, R., Herbst, A. J., Heymann, D., Burke, D. (2014). A systematic review of barriers to data sharing in public health. BMC Public Health, 14, 1–9.Find this resource:
Villegas, J., Matyas, C., Srinivasan, S., Cahyanto, I., Thapa, B., & Pennington-Gray, L. (2013). Cognitive and affective responses of Florida tourists after exposure to hurricane warning messages. Natural Hazards, 66, 97–116.Find this resource: