A Large-Scale Expert-Annotated Dataset of Thematic Maps for Layout and Color Analysis (1990-2025)
This study presents a large-scale, expert-annotated dataset of 48,453 thematic maps from 1990 to 2025, constructed via an integrated computational workflow, to systematically analyze the evolution of cartographic design patterns, revealing distinct stylistic differences between journal and web maps while providing a foundational resource for empirical aesthetics and intelligent map design.
Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
Maps are more than just pictures of the world; they are carefully constructed arguments that guide our eyes and minds through complex information. When a cartographer draws a map, they are not merely placing dots and lines on a page. They are making thousands of tiny decisions about where to put a legend, how bright to make a color, and how to arrange a title so that the viewer understands the story being told without getting confused. For decades, experts have debated these choices, relying on personal experience and small, hand-picked examples to decide what makes a map "good." However, until now, no one has been able to look at the entire history of modern mapmaking to see how these visual choices have actually changed over time or how they differ between a scientific journal and a website.
A team of researchers has now taken a massive step forward by building a digital library of nearly 50,000 thematic maps—maps that show specific topics like population, climate, or election results—published between 1990 and 2025. They did not just collect these images; they taught computers to read them, identifying every single component, from the main geographic area to the small text labels and color bars. By analyzing this enormous collection, they have uncovered the hidden rules that govern how we see maps today, revealing that while the tools have changed, the fundamental need for clarity remains the same.
The researchers started by gathering maps from two very different worlds: academic journals, where scientists publish their findings, and the web, where organizations and governments share information with the public. They collected 32,185 maps from journals and 16,268 from authoritative websites, creating a dataset that spans thirty-five years of design history. To make sense of this mountain of data, they developed a computational workflow that acts like a highly trained pair of eyes. First, the system scanned thousands of documents to find the maps. Then, it used artificial intelligence to filter out anything that wasn't a true map, such as a photograph or a simple chart. Finally, the system broke each map down into its parts, identifying the main map area, the legend, the title, the scale bar, and other elements, while also measuring the colors and the layout of every single sheet.
What they found is that despite the explosion of digital tools and the shift from paper to screens, thematic maps have remained surprisingly consistent in their core structure. The most dominant feature of any map is, naturally, the map itself. In almost every single example, the main geographic area takes up the majority of the space, serving as the visual anchor. The next most common element is the legend, which explains what the colors and symbols mean; this appears in about 61 percent of all maps. Titles and scale bars are also frequent, but other elements like north arrows or inset maps appear much less often. The researchers discovered that most maps are actually quite simple, containing only one to four distinct visual components. This suggests that mapmakers generally avoid clutter, preferring to keep the design focused so that the viewer can grasp the information quickly.
The way colors are used has also followed a steady, functional path. The study shows that maps are overwhelmingly dominated by neutral colors, specifically white and gray, which together make up more than 70 percent of the visual space in an average map. These light, quiet backgrounds act as a canvas, allowing the colored data to stand out clearly. While the specific shades of blue, green, or red used to show data have become slightly more varied over the last few decades, the overall strategy has not changed: keep the background bright and the colors distinct but not overwhelming. The researchers noted that maps rarely use a chaotic mix of many different colors; instead, they typically stick to a small palette of three to seven colors to ensure the map remains easy to read.
However, the study did reveal a clear split in how maps are designed depending on where they are published. Maps found in academic journals tend to be more conservative and focused on the geography itself. They often place the main map in the exact center of the page and rely heavily on traditional tools like scale bars and north arrows to ensure scientific precision. In contrast, maps designed for the web are more self-contained and narrative-driven. Because a web reader might not have the surrounding article text to explain the map, web maps include more titles, legends, and explanatory text directly on the image. They are also more flexible in their layout, often arranging elements to tell a story rather than strictly following a grid. While journal maps prioritize spatial accuracy, web maps prioritize immediate understanding and visual communication.
Over the thirty-five years covered by the study, the researchers observed a gradual shift toward maps that are more self-explanatory and visually diverse. In the 1990s, maps were often very plain, relying almost entirely on white and gray backgrounds with very few colors. Today, while the background remains light, maps use a wider range of colors and include more elements to help the viewer understand the data without needing to read a separate article. This evolution suggests that as maps have moved from the printed page to the digital screen, they have adapted to become more independent visual objects. They have become better at telling their own stories, packing more information into a single image while still maintaining the clear, organized structure that makes them useful.
This massive dataset does more than just describe the past; it provides a new foundation for the future of mapmaking. By having a clear picture of what works and what doesn't across thousands of examples, designers and computer systems can now learn from these patterns. The study suggests that good map design is not about following rigid, old-fashioned rules, but about balancing information with visual simplicity. Whether for a scientist analyzing climate data or a citizen checking election results, the goal remains the same: to present complex information in a way that is instantly clear, visually stable, and easy to understand. The researchers have shown that even in an age of artificial intelligence and rapid change, the most effective maps are still those that respect the viewer's need for clarity.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.