This README.txt file was generated on 2020.02.29 by Carolyn S. Fish ------------------- GENERAL INFORMATION ------------------- Title of Dataset: Climate Change Maps in the US Media 2012-2017 and Content Analysis Author Information (Name, Institution, Address, Email) Principal Investigator: Carolyn S. Fish, PhD ORCID 0000-0002-2366-5035 University of Oregon Department of Geography 1251 University of Oregon Eugene, OR 97403-1251 cfish11@uoregon.edu Date of data collection (single date, range, approximate date): 2017.08.11 to 2018.04.24 Geographic location of data collection: University Park, PA Information about funding sources or sponsorship that supported the collection of the data: This research is based upon work supported by the National Science Foundation Doctoral Dissertation Research Improvement Grant under Award No. 1735747 and the Society of Woman Geographers Pruitt Fellowship. -------------------------- SHARING/ACCESS INFORMATION -------------------------- Licenses/restrictions placed on the data, or limitations of reuse: Recommended citation for the data: Fish, C.S. (2020) Content Analysis of Climate Change Maps in the US Media. ScholarsBank. DOI:10.7264/58x2-g466 -------------------- DATA & FILE OVERVIEW -------------------- File list (filenames, directory structure (for zipped files) and brief description of all data files): The folder in which contains this README includes 4 files: 1. CodedMaps_022920.csv 2. README.txt 3. SecondCoder_022920.csv 4. Screenshots.zip The Screenshots.zip file includes 240 screenshots of maps which are described in the CodedMaps_022920.csv. Relationship between files, if important for context: 1. CodedMaps_022920.csv: This data is described in this README 2. README.txt: This file 3. SecondCoder_022920.csv: File containing coding by second coder. 172 of the maps were coded for the general codes and other extraneous information, 240 the maps were coded for the vividness codes. Where the second coder did not code, NULL is input in the cell. 4. Screenshots.zip: The Screenshots.zip file includes 240 screenshots of maps which are described in the CodedMaps_022920.csv. Additional related data collected that was not included in the current data package: Some other data was collected about the maps but it was not included in this (or any) publication and was not included. If there are there multiple versions of the dataset, list the file updated, when and why update was made: This 2020.02.29 dataset is the final dataset included in publication. Previous versions of the data included more data that was not included in this (or any) publication. -------------------------- METHODOLOGICAL INFORMATION -------------------------- Description of methods used for collection/generation of data: The full description of the methods can be found in: Fish, Carolyn. 2018. ÒThe Cartography of Climate Change in the United States Media.Ó University Park, PA: The Pennsylvania State University. The following is copied and pasted from this dissertation. THE SAMPLE Between early 2015 and late 2017, a group of four undergraduate interns under my direction collected maps of climate change. The goal was to find maps in the print and online media published between January 2012 and December 2017 that illustrated climate change causes (e.g., CO2 production and movement), impacts including everything from temperature and precipitation changes to glacial melt and sea level rise, as well as maps which illustrated the geographic disparities in public opinions about climate change. To be included in the sample of maps, the map needed to clearly indicate that it illustrated climate change in the title, legend, or map notes, or the article needed to mention the term climate change in the text. The print and online media included in the search did not include maps on personal blogs, maps from peer-reviewed articles, maps in reports for lawmakers (e.g., the IPCC reports), or government agency maps unless these were reproduced in print or online media sources. The sources that were included were newspapers (e.g., The New York Times and The Washington Post) and magazines (e.g., National Geographic Magazine), as well as new digital media (e.g., Buzzfeed, Mashable, etc.) The sample did not include maps on TV because these maps are often not thematic and instead are simple locator maps with little climate data. These restrictions also limited the volume and complexity in locating and archiving these sources. The maps were located through Internet searches (Google and Twitter), National Geographic Magazine repositories, The New York Times website, The Washington Post website, The Los Angeles Times website, the PressReader database, the Associate Press (AP) Image Database, and The New York Times Historical Database. I used the following search terms across all of these websites and databases: Òclimate change,Ó Òclimate change map,Ó and Òglobal warming,Ó as well as more specific terms such as: Òsea level rise,Ó Òsea ice,Ó Òglaciers,Ó Òflooding,Ó Òtemperature change,Ó and Òprecipitation change.Ó I also used more general search terms including ÒclimateÓ and Òenvironment.Ó PressReader, The New York Times Historical, and the Associated Press (AP) Image databases were available through my university library subscription. PressReader is a subscription based service to which libraries can subscribe and allows patrons to browse in full-color the past 90 days of over 6,000 periodical publications from around the globe. Within PressReader I focused on publications in English from the United States from cities with major newspapers. Since this resource only contains the previous 90 days of content, I looked at this source three times: in September 2016, January 2017, and April 2017. The New York Times Historical database contained every article from The New York Times from 1851 to 2013 in full print page form including graphics. However, this database only contained the high contrast black-and-white (no greyscale) versions of the print articles. In this database, I only looked at articles and maps from 2012 and 2013 because the database did not contain articles after 2013 during the data collection period. From the identified articles, I searched for the full color images online on The New York Times website since often the pure black-and-white scans of the maps were unreadable. The AP provides news stories to other news organizations. Generally small local newspapers do not produce their own stories and graphics on larger international topics, such as climate change. Instead these smaller outlets rely on the AP for broad non-local stories because the AP has the resources to write these stories with the goal of dissemination through smaller local news organizations. Thus, the graphics from the AP database served as a representation of a wider range of sources. I also had access to every map published in National Geographic Magazine during the time period of interest, and I identified the maps during that period which illustrated climate change. Finally, after the initial set was compiled, maps within the same article which had the same design and topic and only illustrated different geographic areas were only coded once in the analysis. For instance, if one website contained maps of sea level rise for five US cities, this was coded as one map because the topic and map design was the same, even if the geographic area of interest in the five maps were different. The final set of maps amounted to 242. CONTENT ANALYSIS I analyzed the maps with content analysis. Content analysis is a systematic method for examining and comparing symbols of communication (Rose 2012). A set of codes is identified and these codes offer a systematic lens by which to examine themes (Krippendorff 2013). This type of analysis has typically been used for analysis of text, but has recently been expanded to maps to derive common themes (Muehlenhaus 2011, 2013) as well as best practices (Kessler and Slocum 2011, Roth et al. 2015). The goal of this content analysis was to understand who produced the maps, where they were reproduced, what aspect of climate change they illustrated, what types of design they used (type of map and visual variables used), what location and extent they showed, and the extent to which each map was vivid through a Likert scale rating by two coders (see Table 2). Once the full set of 242 maps was compiled, I established a coding scheme. The coding scheme consisted of general codes (Table 1) and vividness codes (Table 2). The general codes were important for understanding the content included in the maps related to: 1) the publication location, 2) producer, 3) date of publication, 4) use of dynamic map designs, 5) type of map design. The vividness codes were established to analyze for the aspects of the maps which aligned with the aspects of vividness from Chapter 3 which are: 1) visual salience 2) visible change over time, 3) color use which aligns with cultural and emotional conventions, 4) best practices for cartographic design, and 5) novel designs. These aspects of vividness were based on themes from a series of interviews conducted with media cartographers and graphics editors at major media organizations and government agencies who produce maps of climate change. I first coded the maps based on the general codes using a Google form. I typed in responses for the short answer codes (Location and Producer), and selected from a set of potential answers for the multiple-choice codes. For the date code, I entered a six-digit date. I coded based on the vividness codes in a second round of coding. Table 1. List of the general codes and how they were collected. CODE DATA TYPE Location Short Answer Producer Short Answer Date Date How dynamic is the map? Multiple Choice Type Multiple Choice The vividness scores were coded on a 5-point Likert scale where the highest score was assigned if the map fully implemented a particular vividness aspect, and the lowest score was assigned if the map did not engage with a particular aspect. Table 2. List of the vividness codes and explanations. CODES RATINGS (5-POINT LIKERT SCALE) Visual Salience Is the important thematic data made salient in the map? Change over Time Does the map show change over time to make climate change tangible? Novel Design Does the map use a novel design style? Color Use Does the map use saturated colors that align with color connotations? Projection Does the map use an appropriate projection for the data? Symbolization Does the map use visual variables that are appropriate for the data? Legend Design Is the legend clear? Layout Is the layout design balanced? NON-METRIC MULTIDIMENSIONAL SCALING (NMDS) Because the result of the vividness ratings was a combination of the nine vividness aspects, it was important to analyze these codes with a method designed for analyzing multidimensional data. I used non-metric multidimensional scaling (nMDS) in the vegan package in the R statistical package to identify clusters of maps based on vividness codes. nMDS is a visual ordination method used for understanding and explaining the interaction between variables. It is often used by ecologists for understanding species distributions (McCune et al. 2002). As a visual ordination method, nMDS graphics are meant to be read and interpreted visually. The multidimensionality of the input data is scaled to reduce the dimensions, and the resulting dimensions in the nMDS plot are arbitrary. In this case, the nine vividness attributes were scaled to 2D space. This type of analysis does not show the most and least vivid maps, instead it allows readers to see how maps cluster. Describe any quality-assurance procedures performed on the data: The maps were coded by two coders trained in the coding scheme. Both coders coded every map for the general codes and the vividness codes. I measured CohenÕs Kappa and percent agreement to assure interrater reliability. Percent agreement accounts for the differences in coding, while CohenÕs Kappa accounts for agreement that could be expected by chance (Landis and Koch 1977). The results from the interrater reliability measures are illustrated in the Table below. Many of the codes had very high interrater reliability agreement in the Landis and Koch (1977) ÒAlmost PerfectÓ range. Other codes had lower interrater agreement because there was more variability because the categories had more potential options (e.g. Type) or because the code was more subjective (e.g. Visual Salience). CODE KAPPA PERCENT AGREE MISMATCHES LANDIS & KOCH Location 0.98 0.98 3 Almost Perfect Producer 0.75 0.76 44 Substantial Date 0.95 0.91 13 Almost Perfect Dynamics 0.60 0.94 12 Moderate Type 0.57 0.66 64 Moderate Visual Salience 0.60 0.48 123 Moderate Change over Time 0.74 0.53 111 Substantial Novel Design 0.56 0.55 107 Moderate Color Use 0.82 0.71 69 Almost Perfect Projection 0.87 0.90 23 Almost Perfect Symbolization 0.85 0.96 9 Almost Perfect Legend Design 0.89 0.76 58 Almost Perfect Layout 0.90 0.75 59 Almost Perfect Topic 0.90 0.81 45 Almost Perfect People involved with sample collection, processing, analysis and/or submission: The four undergraduate students who assisted in data collection were Nicole Rivera, Julia Higson, Han Yu, and Peter Ryan. Peter also served as the second coder and assisted in writing R code which was used to evaluate interrater reliability and nMDS. -------------------------- DATA-SPECIFIC INFORMATION Number of fields: 32 Number of cases/rows: 240 Variable list, defining any abbreviations, units of measure, codes or symbols used: FIELD NAME, EXPLANATION MAPCODE, Code of map which corresponds to the file name for each screenshot SALIENCE, Likert Scale (1-5): Is the important thematic data made salient in the map? LAYOUT, Likert Scale (1-5): Is the layout design balanced? PROJECTION, Likert Scale (1-5): Does the map use an appropriate projection for the data? SYMBOLOGY, Likert Scale (1-5): Does the map use visual variables that are appropriate for the data? COLORUSE, Likert Scale (1-5): Does the map use saturated colors that align with color connotations? VISIBLE.CHANGE.OVER.TIME, Likert Scale (1-5): Does the map show change over time to make climate change tangible? LEGEND, Likert Scale (1-5): Is the legend clear? NOVELTY, Likert Scale (1-5): Does the map use a novel design style? TOTALVIVIDNESS, Total of the eight Likert Scale ratings (salience, layout, projection, symbology, coloruse, visible.change.over.time, legend, novelty) PRODUCERDETAILED, Short Answer: Who produced the map? Maybe the map is on Buzzfeed, but NASA made it, NASA would be the response PRODUCER, Producer field where only major producers are listed instead of where the data originally came from as well PRODUCERGENERALIZED, Producer field where smaller producers are combined into "other" category DYNAMICS, Multiple choice: Is the map static, interactive, animated, or both interactive and animated? TYPE, Multiple Choice: Is the map: raster, choropleth, filled isoline, isoline with no fills, point symbol, proportional line, proportional point symbol, reference, unsure, other? MAPTITLE, Title of the map or title of the article if the map does not have a title URL, URL to article with map LOCATION OF MAP, Short Answer: Where are you viewing the map? (e.g. website, newspaper, etc) MAP DESCRIPTION, Short Answer: Description of the map DATE, Date of publication of the website/article where the map was published CLIMATE CHANGE EFFECT, Short Answer: Climate change effect shown in the map DATA SOURCE, Source of the data in the map, if listed SCALE COVERAGE, Is the map of: local area, sub-country (not the lower 48), lower 48, country-wide (including Alaska and Hawaii), sub-continent, continental (also Arctic), hemispheric, global CONTINENT(S), Short Answer: North America or list only if not global REGION(S), Short Answer: United States or list only if not continental (e.g. West Africa, South Asia, etc) COUNTRY(IES), Short Answer: United States or list only if one or two countries are the focus SUBNATIONAL REGION OR LOCALITY, Short Answer: Lower 48 or list only if map has a subnational focus (e.g. US Eastern Seaboard, northern Senegal, etc) LEVEL OF MEASUREMENT, Multiple Choice: nominal, ordinal, interval, or ratio VISUAL VARIABLES (USED CORRECTLY), Multiple Choice (can choose more than one): pick from list of visual variables which are used CORRECTLY VISUAL VARIABLES (INCORRECT), Multiple Choice (can choose more than one): pick from list of visual variables which are used INCORRECTLY PROJECTION, Multiple Choice: Azimuthal, compromise, conformal, equal area, unknown