Coronavirus coxcombs to waffles

You can enable subtitles (captions) in the video player

Thanks for joining us, Ken. I wonder if you could tell us something a little bit about yourself and your background and interest in maps.

I guess I call myself a professional carto nerd. So I have a background in academia. I now work at Esri, in the United States. I make maps, talk about maps, write about maps, teach about maps.

So there's a good chance that when people look back on this crisis that a map might be one of the most iconic things that they've seen about it.

Yeah, I agree. Certainly, we've seen that one of the most popular places that people are going for has been the Johns Hopkins University dashboard, and getting in the region of a billion clicks a day. And so why are people going there? Partly because it's the first, because it's had a strong web presence. It's basically become viral itself. And people are going to it, not necessarily for the map, but for the tabular data, for the country comparisons, for some of the graphs.

And crucially, they've been curating the data behind the scenes. So a lot of people are going to the dashboard as a way of accessing the data itself. And then they go away and all sorts of other things with that.

Well, with that level of traffic, it must be on its way to being one of the most viewed maps ever.

I think so, by a long chalk. You know, I know colleagues that are researching the idea of viral maps, and how does a map go viral. And this is just going to go off the charts, quite literally. There's all sorts of other examples we could talk about, about viral maps, but this is on another level altogether.

If we think about the crisis that we're in at the moment, the coronavirus crisis, maps have obviously been... you've said, like, you know, people are seeing more and more maps about. And that's definitely the case with the crisis. What's your take? I mean, presumably, for someone who loves maps, seeing more of them is a good thing.

Any time we have a global situation, or even a national situation, people's natural curiosity is to want to look at what's happening, and to learn and understand. And certainly, in this current crisis, it's also the case that we want to look after ourselves as well. We want to lower our risk. We want to understand where a country is heading, or a town is heading.

If you're going to go to the World Health Organization website, or you're going to go to Public Health England, or Centre for Disease Control over here in the States, you know, you would hope to see pretty authoritative maps. That's not necessarily happening.

Some people might go to websites expecting to see reliable, trustworthy maps, and maybe aren't getting them. What sort of things should people be looking out for?

I think, really, what they should be looking out for is what the map is trying to say to them as much as what it is actually saying to them. You know, a lot of people are just going to get the face value - what does this map show?

So if there's an awful lot of red splotches across your map, you know, what does that do? Well, a person's looking at that and seeing huge swathes of red splotches. And it's maybe fear, it's maybe alarm, it's maybe panic. It's maybe all sorts of emotions that it stirs. But if you step back from that a little bit and think, well, what actually is being shown here, you can begin to perhaps appreciate where there might be a little bit of error or uncertainty in the data.

So there's an example of this on your own blog, which is the map that you made of coronavirus in China, which is what's known as a choropleth map, where each area is shaded in according to a colour based on the number of cases in the region. And you made that map to show people why that's a bad map. Can you step us through that? Why is that such a bad map to look at?

Because the choropleth technique, fundamentally, is one where you take geographical areas as they exist, so the units at which data is reported. And in this case, it was aggregated cases of coronavirus in provinces. And that's the number. It sort of assumes a homogeneity across the entire province, or the entire region, which doesn't exist, because the virus is much more specifically location based.

The problem is with the technique more than anything else. So the choropleth technique demands that data be converted into a rate, or a ratio, or a proportion, or a percentage. And that's because of the way in which that particular technique mediates the map message. So if you map totals - and, you know, you sort of have a fairly good colour scheme of a light through to dark, which we see as less and more. And that's a perceptual thing. That's the eye brain system. That's, you know, light is less, dark is more.

And that's what we see, and that's what our brains then get told. And that's what we perceive in terms of cognition.

On the map that you made, you can see that, because in Hubei province, you've got the same colour as the neighbouring provinces. But when you look at the actual numbers underneath that map, of course, that was only at that stage of the pandemic, which was on the 24th of February. Really, the numbers only apply to Hubei province.

You're dealing with largely varying geographical areas. And you're dealing with fundamentally different population densities within those areas. So you often get very large, sparsely populated areas, and very small, densely populated areas. So we see light to dark, perceptually, as sort of quantity. But really, what we're seeing is intensity.

So we read a choropleth map as one of showing these different intensities, so you can maybe think of that as, what is the probability of encountering the phenomenon in that area? If there's 10 people in an area of 1,000, that's very different to 10 people, 10 cases, in an area that has one million people. So if you were just mapping the tens, they'd be the same colour. There are two very fundamentally different intensities there. The chances, I guess, of encountering the phenomenon are totally different.

So if you put totals on a choropleth, people are going to read it and get a misguided view of the actual situation. And this isn't anything new. The guy who invented choropleths decades and decades and decades ago warned against this. So it's a known misuse, if you like, of that technique.

And the one thing I would say, whenever I talk about this, about the idea that you have to convert totals to a rate is people assume, well, he doesn't want totals on a map. No, that's not what I'm saying. All I'm saying is, if you want totals, just use different mapping technique.

So in the context of coronavirus, totals is obviously one of the things that people are very interested in, you know, total number of cases, and total number of deaths. What sorts of maps should they be looking out for, that they can rely on, if they are interested in information?

You're right, because a lot of people, including epidemiologists, they want totals because they can use totals for planning. How many beds are we going to need? And where do we need those beds? You know, that's a crucial, key thing.

In terms of a different map type, you could use a proportional symbol map type, where the size of the area of the symbol is scaled to the total. And again, our eyes and our brains can figure that out, because we see smaller symbols is less, and large symbols as correct. But if you think about it, we're ignoring the real geography now, OK? We're just placing a single symbol of a consistent shape that represents the geography that's underlying.

So the symbol takes on the scalability, if you like. And each symbol will be scaled correctly in relation to each other's symbol. So we get a correct visual interpretation of quantity. That map type has other problems as well. And if there was one single map type that could do the job, you know, we'd be using it. You could perhaps use a dot density technique, where you seed a number of dots within an area, and it creates a visual density of less or more. That's a perfectly appropriate technique.

I've seen some excellent examples where people have kind of tilted the map. And it's not really 3D, but it's sort of a pseudo 3D with peaks. And so in the US, particularly, where you've got a lot of very sparsely populated counties, and then you have the huge peak of New York City in terms of cases and deaths. That's a fairly nice technique, because it allows you to focus on a clear, single signal within the data, but without having to do too much with all the other data that makes it very difficult to see that otherwise.

Do you think that there's a risk that maps, particularly the ones that you've outlined that might be misleading, that people tend to trust them because they are maps, that people trust maps implicitly, because they look so authoritative?

Yeah, and then that's always been the case as well. If you show a map to somebody, why would they have any reason to mistrust? It's often presented in a very engaging style. It's not difficult to make a map look pleasing and look authoritative. And you know, it has titles, and source text, and all the map stuff on it. You would have no real reason to not believe it.

And I guess that goes back to my earlier point of, you know, that's where you have to consider very carefully where you get your information from, because that, in a large part, plays a big role in how much you can trust what it is you're seeing. And you only have to go back to various conflicts throughout history to see all sorts of propagandist and persuasive mapping.

Without getting this conversation, too political, you know, we've seen graphs at press conferences, and news conferences, and maps. And yeah, you do question whether there's a slight political motive behind some of them, given that - at least the country I'm in at the moment is headed towards an election. And there's nothing so bad for a political election campaign as a total pandemic that's destroying your economy and all sorts of other things, you know.

People might be familiar with the maps in terms of the geography that they portray, but they're less confident in terms of the layers of information that you might put on it. And so something like coronavirus, and a pandemic map, really is a challenge to people's map literacy, because it's not something that we've necessarily seen before. So this is a time for people to be learning how to read these maps. At the same time, it's actually being sort of bombarded with them.

People are doing that. They are learning as they're reading. And they're coming across a geographical phenomenon that is highly localised. You know, it's person to person contact. It's small clusters in small locations. Yet what we're actually seeing on most of the maps is data aggregated up massively. Certainly, in the first few weeks, it was really only at a country level, and it was only really China that had sub country level breakdowns. And even that was provinces.

What you end up with is a mapping technique, such as proportional symbol, where people see one single dot representing maybe a county or a state. And they're making assumptions about what that dot means. Does it mean that's where somebody lives, or who has this? No, it doesn't. It's just purely an artefact of we need somewhere in the area, it's usually the centroid, to basically position the symbol. So it's a purely cartographic artefact that can be misconstrued very easily.

You've done some mapping with coronavirus data using some older cartography techniques - so, the coxcombs from Florence Nightingale, and also the waffle plots. Do you want to talk us through what those sort of maps might bring to the readers?

I've always been fascinated with coxcombs, because they allow you to map space and time together. And that's not easy in a static map. So, the ability to put segments that indicate different days in a circle, so they total up a number of days within each segment to be able to identify the total number of cases, the number of new cases, the number of deaths, and the number of people who've recovered as a sort of proportional segment.

It gives an awful lot of information in a single visual. Is it easy to read? Not necessarily. I think you do need to be very careful about providing a good legend and explanation, and make sure you haven't got all sorts of scary overlaps which really complicate the picture.

And the waffle grid is the same. Instead of trying to use colour, or shade, or size, a waffle grid is based on repetition. So it's a square, it's a Belgian waffle, that's really what it is. It's how many squares have you got in the waffle. And technically, you can count them, and say, oh, OK, we've got 1,000 dots. Or we might make each dot equivalent to 100 cases, so OK, 10 dots, 100 cases - yeah, that's 1,000. We're very good as humans, counting. And we're very good at looking at repetition. And that repetition is a fairly minimally invasive cartographic technique to get people to recognise.

Is there a sense that people rely too much on the software to determine how they present information on maps?

Totally. Yeah, and that's a one-word answer. I mean, we can thank Google, in 2005, for bringing web maps to the public. And then it wasn't so long after that the guy who built Craigslist found a way to mash up Craigslist data on a Google map. And instead of sending him a cease and desist letter, Google sort of pivoted and said, well, hang on a second. People might actually want to do this.

And so we start that whole mechanism for the majority of people, for the majority of mapping needs, having four, five, or six pretty fundamental techniques is enough.

One other thing that I thought was interesting, from watching what you've been saying on social media, is how open people are when they make maps to feedback and criticism of those maps. You've done a little bit of that, working with companies during the crisis.

Basically every day I wake up, it's a direct message from somebody I've never heard of saying, you know, I've made this map. What do you think of it? You know, I had one this morning. I won't name the person, but I ended up saying I hate it. You know, if I didn't know you, I know I'd be a little more polite. But no, please don't go down this route. And he ended up saying, well, maybe I do in 3D.

And I just said, no, you don't need to do 3D. You're over complicating this. It doesn't need to be that complicated. And then you get into conversations where people are iterating and iterating. And you end up sort of sitting back and thinking, you know what, it's nice. One person has improved their maps. And that's possibly helped tens or hundreds or thousands of people be able to read that map and understand what it's trying to say a lot more easily.

So people like Beck, for example, had several iterations of what would end up being a hugely influential map. So it's difficult to come up with this idea that you can just produce the perfect map in a vacuum.

Yeah, it doesn't happen. And even for me and my colleagues, we make a map, and we share it, pass it around. And someone will always point out some way that you can improve it. So, and you're right about Beck. I mean, his first scribble was in about 1931. And he was still iterating the map up until his final one in about 1960. Yeah, that's 30-odd years of basically messing around with the same map.

And to bear, you know, there's a lineage. You can see similarities between the two. But in other ways, they're actually fundamentally different in design. And so clearly, you know, he was happy just to keep iterating. And I think that's an important point to make, that iteration and feedback and redoing things is absolutely critical to weave in. Advice, comments, improvements - knock out a few errors here and there. And you end up with better work, hopefully, in the end.

You mentioned another phrase that I must imagine feels like fingernails on a chalkboard - 3D. Is there ever a case for a 3D coronavirus map?

I'm not sure I've seen one yet. I think 3D is really useful when you encode something important in the vertical dimension. There's that dimension. Now, typically, 3D is used for urban planning and all sorts of other things. And of course, that helps, because you can see the world in 3D. But for thematic data, what do you put in that third dimension?

OK, so you've got x and y on the two sort of base axes. Now, I think if you put time on the vertical axis, you can come up with something quite useful. And if you can plot moments in space and time, then you basically get a spacetime cube. It can get messy very, very quickly, because you've now got not just a single plane that you're looking at, but you're looking through a cube. And you have perspective to deal with, and occlusion, and forshortening.

So basically, 3D makes the job of communicating something simply a lot harder.

If you've seen these maps by Mathieu Rajerison, of France, is that the sort of thing that you're talking about, with the time series extruding from the areas of France?

Yeah, so he's kind of done a pseudo 3D. He's basically got a map of France and put a line graphs, line charts all over it. And I love the way he's got them to animate as well. And I think it's a beautiful example of an interesting, modern take on putting charts on a map. That animation does help people, also, to see speed, trend of a curve, rate of increase or decrease, plateauing of an epidemic curve.

He's actually organised it so the foreground charts do sit a little bit across the top of the ones in the background. So you get this impression of depth in the map. It's not a true 3D map, because for that, you'd be, he'd really tilt France on its side. It's a very good example of a compromise. You know, he's trying to put a little bit of 3D-ness across a 2D map. And that does tend to be a nice, intermediary solution.

Escaping Flatland, as Edward Tufte might put it.

Oh, I didn't know we were going to quote Tufte. All right.

One of the things that's really intrigued me about things that you've said recently, Ken, is the idea that sometimes the answer is just not to make a map.

Yeah, I think you have to make a decision at the outset. What's the best way of communicating the data to somebody? What are people interested in, with this particular pandemic? And we've heard the phrase flattening the curve everywhere. You know, that's become a real campaign, almost. Can you show flattening the curve on a map? I'm not entirely convinced you can, easily. I think you have to get into a lot more complex graphic wizardry to do that.

I mean, the Nightingale coxcombs does it to an extent, but you really have to look hard to find it. So maybe a line chart is better at communicating more useful information at the moment than a map does. Usually, with a map, we want to compare one place to another. Does my home have a similar rate as somewhere else? You know, how is my community affected, how is my country faring?

Do we really need to know that? I'm not so sure. I think, how is the rate of speed of infection going home, and is it doubling every day, or have we started to see a flattening of the curve, is far more elegantly achieved with a simple line chart than a map at the moment, particularly the maps as they are at the moment. If you look at a global map, it's just a global map of population, with large splotches where there are a lot of people, and less splotches where there are not so many people.

I'm not entirely sure what help that is at the moment.

So mapping it when it was localised makes a little bit more sense than mapping it when it was everywhere.

Yeah, I mean, I'm interested in what's going to happen with this partnership with Apple and Google, and tracing people - to be able to have a personalised map on your phone, that's centred on you, so that you can go out into your community, perhaps when stay at home orders are lifted. And you can get a sense of what's going on around you. You know, is the local supermarket populated with a lot of people who have been through infection or not? And then you make a decision. You know, you make your own decision.

And that's kind of hyperlocalised mapping. It's useful mapping, because it gives you information upon which you can base a decision. In terms of global or national maps, I think we're going to see more useful maps post-pandemic. You know, when we go back and start to analyse spread, and be able to understand, perhaps, the pattern of spread between different communities more readily.

Just going back to your point on the contact tracing. Does that drag cartography into the privacy debate?

Yeah, but I think cartography has always been on the edge of that. I mean, the question has to be asked, I guess, is if you don't have complete, full co-operation of everybody, then your data is also going to be full of holes and errors and uncertainty. But I think cartography, and these sort of mapping companies, are always trying to push the envelope, and to say, how can we make maps more useful to people? Because that, ultimately, is the point of a good map. It's to be able to give them something useful, and allow them to ask more questions of the map, and of the data that sits behind the map.

Do you think in any... that people's relationships with maps will be different after this crisis?

I guess I hope so. They're seeing these kind of maps as fundamental to their everyday lives, just as we see the weather map. You know, people are very happy to look at a weather map every day, because it tells them something useful. Maybe they start to see them as, perhaps, windows into data that surrounds them, surrounds their lives. I mean, we all have a phone in our pocket. And it's giving us data every day, as well as giving others data that they can use every day. We're a data driven society.

So a natural curiosity about all of that data is only going to increase, I think, and maybe maps are the way to allow us to communicate it. This goes back to school, I'm afraid. We learn numeracy, articulacy, literacy. And the very famous UK geographers, William Balchin and Alice Coleman, in the 1950s and 60s, termed this phrase graphicacy. It's, like, the fourth component of human conversational language. It's the ability to communicate with a visual, with a graphic, with a map.

And we don't really learn that at school. You know, maybe we do have to go back to start thinking of this moment as a point in time when geography does get back to becoming more mainstream. And then we perhaps might end up with a less graphically illiterate population, or at least the population that has the ability to question what they're seeing, and, you know-- all those red dots on a map, oh, no. Just, what does it really show? And I think that would be great.