A Better Way to See the World
How NTI and Forum One Used a Tile Grid World Map to Increase Clarity, Fairness, and Impact of Its Data Display
This post was originally published on Forum One’s site as Using a New Tile Grid World Map for Displaying Impact
Nuclear Threat Institute (NTI) just launched the fifth edition of its signature Nuclear Security Index, a detailed research study that evaluates 176 world countries on how they are contributing - or not - to the security of nuclear materials. The Index tracks security across four major goals (including a new goal concerning radioactive sources), though the nature of the goals and the differing nuclear capabilities of countries means that no country is tracked across all four (for example, the security goals for a highly nuclear capable country like the United States are different than those for a country with just nuclear power or research facilities, like Morocco, and are different still from a country with no nuclear materials or facilities, like Tunisia).
With an objective to show Index results across this wide array of countries, a data-encoded world map was a key feature for the Index website. In fact, a color-coded map had been a core element of previous editions of the Index site. As we considered the data and the messages of the Index, though, the NTI and Forum One teams decided that we could do better than the past iterations. In this post, I will explain how we created a new world-level tile grid map to show country results in a clearer and more just way, thus helping NTI’s overall messaging for the Index.
The Prior Map: A Choropleth
The prior map was a choropleth map. Geographic shapes are presented as they are, given a selected map projection, and color coded to match a set of data. It is fair to say it is still the most common approach for displaying data, especially at the world level. Below is the view for those countries with nuclear materials, where the goal is to ensure security of those materials from theft by bad actors.
Such a map is well-understood by most audiences - they know what the shapes are (i.e. there is good geographic fidelity), understand how to interact with them, and, in general, can make sense of how to read the encoding. It is certainly tried, but is it true? We thought not.
Choropleths have some notable disadvantages:
Geographic distortion and justice - The map NTI used is a particularly bad Mercator projection, so the extreme latitudes, in practice, the north, are quite distorted and appear much larger in size than they actually are. Distortion is an issue for any choropleth map, and has vexed cartographers for centuries. One way or another, though, the most common projections tend to exaggerate the Global North at the expense of the Global South. Mitigating options exist, but they are less standard. We wanted a display that would treat all countries equally, both for justice and because all countries need to be equally engaged with security issues.
Differing country size and data distortion - Regardless of projection, the bigger issue is that countries have vastly different sizes, and this can result in some countries visually dominating the map - looking at you, Russia, Canada, US, China, and Australia - while physically smaller countries fade, or even totally hide. This is a critical issue with this data. These large countries are important, but, in the map above, they are all in the top two ranges. They are not where attention MOST needs to be. Israel, however, has lower results, and does deserve attention. It is, though, physically small, and so it barely appears on the map. Even highly problematic North Korea’s red color does not draw appropriate attention because it is a physically small country. In this case - and, likely, in many others - the size of the country is of no importance to the results, so the resulting display distorts the interpretation of the data.
Country size and interactions - Small countries have another issue. For a map that is to also serve as a gateway to country detail pages, as this one is, physically tiny countries become hard to select. Those with imperfect dexterity would need to zoom to access many countries around the size of Israel, but getting to countries in the data set like Luxembourg, Tonga, and Barbados would be quite a trick, and often requires hacks to the map or deep zooming.
Tile Grid Maps: Promises and Limitations
The problems above are increasingly being solved at lower levels of geography with tile grid maps, a style of which I am a notable fan. Tile grid maps at the country level are quite common now, e.g., by FiveThirtyEight or The Guardian, and some leverage the style for interesting, layered data displays (though I still like my US tile grid map better than the ones commonly used). These displays often still run into awkward balances between geographic fidelity and the true tile layout; witness the US maps that have Washington DC adjacent to North Carolina and Ohio or like.
This limitation - which comes from large differences in sizes between geographic entities - has proven too much of a burden, it seems, to take the model to the world level. A few folks, most notably, the talented Jon Schwabish, have offered a pure world tile grid map.
While I applaud these efforts, and they carry the advantages of the tile grid approach forward to the world level, I just don’t think they work. There is just too little geographic fidelity. Only the most careful reader can pick out the world in this view, and too many relationships are at odds with each other.
With tile grid maps holding so many advantages, but countries being at such different sizes, there is a clear conundrum. Is there a way forward? We believe we found one.
The Solution: A New World Tile Grid Map
The “true” world tile grid above falls short because it does not maintain sufficient geographic fidelity. This happens because of the large countries not taking up their usual space. Most of North America, by area, is the three countries of Canada, the US, and Mexico. In the true tile grid, these become three squares, a small purple area partially visible in the image above. Without the expected shapes and relationships in space, you can’t tell what the squares represent. The same is true in Asia, without the usual mass of Russia, China, Kazakhstan, and others. You can’t lose the basic overall shape if you are to maintain geographic orientation.
The solution we developed for the NTI Index site, then, creates a grid-based set of landmasses that evoke the continents as we know them. North America shows Hudson Bay, the Gulf of St. Lawrence, the Great Lakes, Florida, the Yucatan Peninsula, and Alaska. Europe includes the Baltic, Adriatic, and Aegean Seas, and the Scandinavian, Iberian, Italian, and Balkan Peninsulas. Asia includes its most notable seas, bays, subcontinents, and peninsulas. The shape is rough, but it evokes what we see when we look at a world map. We even tried to ensure the landmasses were more equitably sized.
We then place one, equal-sized square per country within the landmasses. In some cases this fills all the available space, as in Eastern Europe or West Africa. In other cases, countries float near the centroid of their available space, as in North America, South America, or northern Asia. Each square - country tile - is close to its “proper” place, though, and all are equally sized.
To help readability, we made two other decisions. We kept all countries visible, but grayed some out if they are not in the selected goal data set. This helps orientation and helps readers to understand that only some countries are included in each set. We also used a blue gradient for the data coding to make it clearer where countries are on the scale. This also ensures the data will be clear to those viewing it in contexts where color is not available.
To summarize reasons this world tile grid map works better than the prior choropleth:
Each country has equal visual weight. That means that no “data point” stands out for reasons other than its encoded data (i.e., because of its size). Israel is now readily noticeable among the countries with nuclear materials, and North Korea is as conspicuous for its low rating. There is no distortion in the core data.
There is geographic equity. The ability for Brazil, Senegal, and The Seychelles to all stand on equal terms and not be highlighted or hidden by size, lets them each justly shine as countries doing better on the same echelon. In general, the Global South stands out much more than it typically does in a choropleth, and one can find the notable success stories there, encouraging those countries and those around them.
There is easier navigation. It is as easy to access the Israel page as the Russia page because each country has an acceptable tap target at the initial presentation size.
There is still enough geographic fidelity to stay oriented. While the map is blocky, the fact that it includes major elements of the shape of continents means people can tell what they are looking at and find countries in which they are interested. Beyond that, greater geographic fidelity is not needed; the true geospatial relationship doesn't matter for this dataset.
There is also a readily accessible table-version of the data for ranking, viewing details at once, and serving needs of accessibility, but that’s a topic for another post.
Using a World Tile Grid Map
Thi data visualization approach would be most appropriate for use when:
The data set is world-wide (though it could be scaled to regional use too);
The data needs to show equally for all countries, and/or physically small countries need to carry equal weight with large ones;
True geospatial relationships are not important to data interpretation and understanding;
Use of the map as a navigational element is important.