Exploropleth: exploratory analysis of data binning methods in choropleth maps

AI summary

Overview

Research area: Human-Computer Interaction, geovisualization, and cartography — specifically the design and evaluation of interactive tools for choropleth mapmaking.

Technical level: Intermediate. No machine learning or heavy statistics is required to follow the paper, but readers benefit from familiarity with cartographic concepts such as data classification, bin breaks, and small-multiple map layouts, plus basic awareness of GIS tools like ArcGIS and QGIS.

Scope: The paper presents Exploropleth, an open-source, web-based tool that displays sixteen established data binning methods side by side so users can compare, customize, and export choropleth maps, and reports feedback from interviews with sixteen cartographers and GIS experts.

What This Paper Is About

Choropleth maps color geographic regions to show the magnitude of some attribute, and before drawing one, a mapmaker must bin (classify) continuous values into a small number of groups. Different binning methods — natural breaks, quantile, equal interval, and so on — produce different groupings from the same data, so the same dataset can look dramatically different depending on the method chosen, and a poorly considered choice can produce misleading or false spatial patterns. The paper's goal is to give mapmakers of all experience levels a single, accessible entry point for browsing, comparing, customizing, and exporting the outputs of many binning methods, and to learn what practicing GIS experts think of such a tool.

Key Contributions

1. Exploropleth, an open-source, web-based geospatial visualization tool that lets users browse a catalog of established binning methods and interactively compare, customize, and export maps. It is publicly accessible at https://exploropleth.github.io .
2. Faceted browsing of multiple binning methods in one view — sixteen methods rendered as small multiples in a grid, so users can drag to reorder, hide, expand, customize, and export individual method outputs without leaving the view.
3. On-the-fly administrative unit reclassification, including a "Paint Mode" in which a user selects a target bin color and clicks individual administrative units to force them into that bin, with the tool recalculating bin counts and intervals in real time to satisfy the constraint.
4. Findings from an evaluation with sixteen cartographers and GIS experts on how Exploropleth can help choropleth mapmakers and readers, plus a supplemental comprehensive overview of binning methods including descriptions, references, examples, and usage guidelines.

Main Findings

• Multiple methods in one view was the standout value. All sixteen experts appreciated having multiple binning methods in one view, vindicating the authors' initial assessment of the need. One expert (E3) said, "Nothing better than this can happen for GIS experts & coders," and E8 said, "It's amazing just how much you can change a map by changing their binning method."
• The tool addressed a concrete workflow gap. E14 described the alternative as having to redefine and replace bins and export them one at a time, adding, "I can't imagine doing that for [sixteen] different types of binning methods. So having this is really cool." E5 called it "tedious to one-by-one explore the binning methods that are not supported by ArcGIS," concluding that "having them in the same place is phenomenal."
• It lowers barriers for non-programmers. E16 noted struggles with existing Python libraries "because [they are] not a computer programmer," and E2 described Exploropleth as "a useful, lightweight tool that is intuitive and easy to maneuver."
• It may alleviate organizational constraints. E3 noted their organization is "technically sound in creating maps but have limitations in terms of available [funding and personnel]," and said "There is no open source tool like Exploropleth so what you have done is excellent."
• It supports open-ended exploration and learning. E1 said, "I'm [going to] mess around with this on my own, as this is really cool"; E4 said, "I don't know what I want to click on next. I just like being able to see so many [binning methods] all at once"; and E10 said, "My knowledge was enhanced today."
• Familiarity with methods is uneven. Among established methods, experts used pretty breaks and manual interval most often, followed by natural breaks, standard deviation, quantile, and unclassed. Many were unfamiliar with ck-means (E4, E15), head-tail breaks (E3, E4, E15), and box plot (E1). Reasons given included personal preference ("I just use the ArcGIS default, natural breaks" — E16) and organizational policy ("it is a practice to use standard deviation here" — E1).
• The tool can educate as well as assist. The authors found the tool has potential to educate students as well as mapmakers with varying levels of experience, and experts related that comparing methods would encourage more thoughtful method choice and help avoid "unintentionally lying with maps."
• Method choice visibly changes the map. By definition, the quantile output shows an approximately equal distribution of blue and green shaded counties, whereas maximum breaks appears mostly blue, likely because of outliers.
• Consensus across methods can be quantified. In the Combine View, the tool tracks each administrative unit's bin index across eight binning methods, all configured with six bins. For Scott County, Kansas (life expectancy = 80.85 years), the bin indices were equal interval (4), quantile (6), maximum breaks (2), natural breaks (5), ck means (5), geometric interval (4), percentile (4), and box plot (5), giving a most frequent bin of 4 with frequency 3 — making it one of the most inconsistent counties. Oglala Lakota County, South Dakota (life expectancy = 62.44 years, the lowest in the U.S.) was among the most consistent at binID = 1.
• Experts identified integration opportunities. They described ways to fold Exploropleth into existing mapmaking workflows, including exporting bin breaks to recreate in other tools and exporting image snapshots for reports.

Methodology in Plain English

The authors began by surveying cartographic literature and software documentation to build a catalog of binning methods, which they organized into six high-level categories based on how each method computes bins: interval-based, statistical, iterative, spatial, human-centered, and other. Exploropleth currently supports sixteen methods spanning five of these categories (human-centered, interval-based, statistical, iterative, and other). They compared existing options — ArcGIS offers seven binning methods, QGIS and GRASS GIS five each, Mapbox Studio four, the ArcGIS Maps SDK six, PySAL ten, tmap nine, and the BinGuru JavaScript library 18+ — and derived design goals to guide the tool.

The interface was designed with four tabbed views (Browse, Compare, Combine, and Create) plus a shared Data and Configurations view. The default number of bins is n = 5, aligning with prior recommendations in the literature, and the tool ships with color palettes filterable by web-, colorblind-, and print-friendliness. Exploropleth is implemented with the Angular framework, uses BinGuru for binning methods, Vega-Lite and D3 for visualizations, and color palettes from ColorBrewer, d3-scale-chromatic, and Vega. An early version was shown to two pilot users (second-year PhD students studying data visualization), and design alternatives — synchronous zooming and panning across the sixteen maps, a single large canvas DOM element, and alternative legend treatments for bin sizes — were prototyped and discarded based on feasibility and feedback.

For evaluation, the authors interviewed sixteen cartographers and GIS expert practitioners and researchers with self-reported mapmaking experience between 3 and 33 years (median 15 years, mean 15.43 years), drawn from 13 government organizations, NGOs, and federal agencies, with eleven based in the U.S. and two in India. Sessions ran remotely over Zoom for twenty-five to thirty minutes with no compensation, and the study was approved by the Georgia Tech Institutional Review Board (Protocol Number: H22470). Each session included roughly ten minutes of screen-shared demonstration, an invitation to interact and think aloud, and roughly fifteen minutes of structured discussion. Audio was transcribed, segmented, and analyzed with open coding, constant comparison, and theoretical sampling.

Why This Matters

Impact on research. The paper argues that curation and communication of the wide established cartographic literature on binning should have a facile entry point for both experienced and new mappers. It positions Exploropleth as advancing the state of the art through faceted browsing of multiple binning methods in one view and on-the-fly reclassification, and it contributes qualitative evidence about how practitioners actually reason about classification choices — a topic that the paper notes has been critiqued and revisited for over 50 years without a single ideal method emerging.

Real-world applications:

• Public health and policy. Experts in the evaluation worked with organizations including the World Health Organization and U.S. federal agencies such as the Department of Agriculture; one usage scenario describes a government official analyzing county-level life expectancy statistics to inform future policies.
• Urban planning and demographics. A second scenario covers a mapmaker preparing a choropleth of U.S. county population density who needs to highlight densely populated urban counties.
• Government and NGO reporting. Experts noted limitations in available funding and personnel at their organizations, and the tool's ability to export bin breaks and image snapshots supports reporting workflows.
• Teaching and cartographic education. An explicit usage scenario follows a student in an introductory GIS class using Exploropleth in a lab to learn about binning methods, and experts believed it could educate mapmakers and readers of varying experience.

Industry relevance. Existing commercial and open-source options are incomplete for this task: the paper reports that ArcGIS offers seven binning methods, QGIS five, Mapbox Studio four; that these tools sometimes omit newer or less common methods such as head-tail breaks; and that they do not support side-by-side comparison of methods or client-light exploration. Exploropleth's bundling of sixteen methods with client-side interactivity and exportable TypeScript source and Vega-Lite specifications targets that gap.

Future Directions

1. Deeper integration into existing workflows. Experts identified opportunities to integrate Exploropleth into their mapmaking processes, and one scenario shows exporting bin breaks to recreate them manually in ArcGIS. Tooling that closes that loop further is an open direction.
2. Broadening method coverage. Exploropleth supports sixteen methods across five of the six identified categories; the spatial category — including equal area breaks, shared area, minimum boundary error, and boundary accuracy indices — is documented in Table 1 but not listed among the methods currently supported.
3. Investigating educative effects. The authors report that the tool has potential to educate students and less experienced mapmakers, but the study design (short remote sessions with experts) does not measure learning outcomes; whether comparison in one view actually changes method choices over time is not established.
4. Addressing misuse and propaganda risks. The Create View's Paint Mode carries the in-tool warning, "We recommend using this feature only for educational purposes," acknowledging that custom bins can be used to produce misleading maps. How to support legitimate narrative-driven classification while guarding against manipulation remains an open question.

The paper content provided does not report which additional features the interviewed experts requested in response to the question, "What additional features would you like to see in this tool?"

Target Audience

Cartographers, GIS analysts, and geospatial practitioners who produce choropleth maps and must choose a classification method; visualization and HCI researchers interested in comparative, small-multiple, and direct-manipulation interfaces; instructors and students in GIS and cartography courses; and developers building open-source mapping libraries, who may want to reuse the BinGuru-based implementation or the tool's export formats.

Authors’ abstract

When creating choropleth maps, mapmakers often bin (i.e., group, classify) quantitative data values into groups to help show that certain areas fall within a similar range of values. For instance, a mapmaker may divide counties into groups of high, middle, and low life expectancy (measured in years). It is well known that different binning methods (e.g., natural breaks, quantile) yield different groupings, meaning the same data can be presented differently depending on how it is divided into bins. To help guide a wide variety of users, we present a new, open source, web-based, geospatial visualization tool, Exploropleth, that lets users interact with a catalog of established data binning methods, and subsequently compare, customize, and export custom maps. This tool advances the state of the art by providing multiple binning methods in one view and supporting administrative unit reclassification on-the-fly. We interviewed 16 cartographers and geographic information systems (GIS) experts from 13 government organizations, non-government organizations (NGOs), and federal agencies who identified opportunities to integrate Exploropleth into their existing mapmaking workflow, and found that the tool has potential to educate students as well as mapmakers with varying levels of experience. Exploropleth is open-source and publicly available at https://exploropleth.github.io.