Posts Tagged ‘bad generalization

29
Jun
10

The Vanishing Kingdom

Yesterday evening, I was having a conversation with one of my roommates about Beaver Island, which lay in the north of Lake Michigan. It’s a sizable chunk of land with some interesting history. It was, at one point in the 19th century, home to a kingdom inhabited by a breakaway Latter-Day Saints sect, until the US government facilitated the assassination of its eccentric ruler and the ejection of the Mormon settlers. While mentioning the island to my roommate, I pulled up Google Maps in order to show him where it is. Except it wasn’t there. An entire archipelago, in fact, was missing from the map. Compare the satellite photo to the map and note the difference:

Screenshots from Google Maps, 6/29/10

Perhaps more amusing is the fact that, when you zoom in sufficiently, the road network for Beaver Island (which has a population of about 650, according to Census estimates) still appears.

Screenshot from Google Maps, 6/29/10

Now, I don’t know how the sausage is made over at Google, but I’m guessing it’s a mostly automated process, given the magnitude of their undertaking. And this is what happens when you let computers keep running with insufficient oversight. This is not exactly a tiny island — it’s 55 square miles, and given how large of a scale Google lets you zoom in to, it’s not something that should be left off. Whatever algorithm they’ve used to generalize their data, it’s in need of tweaking. It’s leaving some smaller islands, but eliminating larger ones. Note the smaller Manitou Islands in the south of the first images above, marked as Sleeping Bear Dunes National Lakeshore. Despite being uninhabited and smaller than Beaver Island, they made it on the map. One of them is rather terribly distorted, however — the polygon is way too simplified for the scale.

It’s been said over and over again, but it’s still worth hearing: be careful when using Google Maps and its cousins. There are very few human hands in their creation, and not enough of the scrutiny required to prevent gaffes of this magnitude. Of course, you should be careful when using any map; once humans start making the data and design decisions rather than computers, major geographical errors may become infrequent, but more insidious problems crop up, as we discussed a few months ago.

This is also where learning lots of random geographic facts can be handy. It’s easier to catch the omission of Beaver Island if you know ahead of time that it exists. This is how I justify spending way too much time on Sporcle taking geography quizzes — it will hopefully make me less likely to make an error like the above.

The lessons from today’s map are obvious, but it’s always good to be reminded from time to time of the importance of careful editing. And the end result is a bit amusing here.

One Nice Thing: At least they’ve got a form on the page which I can use to report this error.

Tomorrow marks one year of blogging here on Cartastrophe. I really wasn’t sure that this experiment was going to last more than a few months, but your comments and emails and support have kept things lively. I appreciate your coming along for the ride. To all who have sent submissions: thank you. I don’t use all of them, but I appreciate everyone keeping an eye out and thinking of me, and hope you will keep doing so. This blog has been great for my own growth as a designer, and I hope that you have gained something from it, as well.

Finally, it comes to my attention that there’s another blog out there in a similar vein to my own. If you’d like a double dose of map critique, have a look at Misguided Maps.

04
Nov
09

Finding a Doctor in Non-Geographic Space

I sometimes feel bad about this, but I’m going to let a map I just received today jump to the front of the queue, as it was in the right place at the right time. This one comes from my colleague Tim Wallace, who was sitting next to me filling out a health insurance form.

UWHealth

From the 2009 UW Health Directory, located at http://www.unityhealth.com/apps/FindADoctor/

It may surprise those of you who are not from the area to learn that Dane County, Wisconsin, does not look very much like that. The rigid lines of the Public Land Survey System, combined with the Wisconsin River, have left us with something that looks more like this:

Dane_county

Taken from Wikipedia article on Dane County.

Now, that being said, the fact that this map has a very high level of generalization is not inherently problematic. There are lots of very generalized, even cartoonish maps out there.  It depends on the purpose and audience of your map, and sometimes fine details are not important. But I think they significantly overdid it here. It looks rather comic, and this clashes with the professional application (a healthcare provider directory) to which it’s being put. It seems haphazard and only loosely related to reality, and that makes me doubt the rest of the information that goes in to it.

The real lesson from today’s map, which applies to a lot of maps out there, is that it should not have been made. It conveys no useful spatial information. I can’t use it to figure out how to get to any of the UW clinics, and while I can sort of tell where clinics are in relation to each other (the heavy generalization makes that an estimate at best), that’s not really useful information unless you choose your health clinic based on its proximity to other clinics or cities. “I’m sorry, the village of McFarland has asked that I stay at least 10 miles away at all times, so I’ll need a clinic an appropriate distance away.” So, barring bizarre circumstances, I cannot tell how having a map is better than having a table of clinic locations. A table would, in fact, be significantly more useful, because you could use the clinic address to actually figure out how to get there. Not everything that has a location needs to be mapped, or can be usefully mapped.

Also, if you compare the two maps above, you will notice that the city dots for the clinic map don’t really bear a lot of resemblance to where those communities actually are in Dane county. And, in case you’re wondering, I spot checked this against the addresses for the clinics in Belleville and Verona, and it still did not match. The dots have only the loosest connection to reality. If geography really means so little here, why make a map? Again, a table would be better. The only reason the author of today’s map can get away with this level of generalization and haphazard dot placement is because the map is scarcely conveying any geographic information.

The labeling could be better, though it’s not horrible. It needs more consistency in how far the label is positioned relative to the dot. Look at Cross Plains vs. Waunakee. Why not put them both directly under the dot center, if that’s what you’re going for? And this is not to mention the lack of corner positions. It is conventional, and, as I have often been taught, rather easier to read, if you put the label in a corner position — that is, up and to the right of the dot, or down and to the left, etc. Fitchburg is also ambiguously placed — the label is about the same distance from two different dots. Would have been just fine if it were off to the left of the dot. Now, to be fair, not all labeling can be ideal, because geographic realities get in the way. But, the author of this map does not appear to have been strongly tied to geographic reality anyway, so I’m not sure if that’s an excuse in this case.

This map does not need a legend.  A good map title should tell you what the map is about; since this map is about only one thing, if it’s well-titled I should be able to figure out what the red dots mean. If I’m looking at a map that’s titled “2009 UW Health Clinic Locations,” I’m not going to mistake those red dots for bowling alleys. Or cheese factories. Also, the dot in the legend is not only larger than the dots on the map, but a different shape. The ones on the map are more elliptical.

I was a little confused about the statement in the legend that says the communities of Black Earth and Cambridge are excluded — both are in Dane county. Skimming the report a bit, I think it’s because they’re not part of the UW Health Network. In which case, of course they wouldn’t be on the map — why would they bother to mention that? Chicago is also not included on this map, but they forgot to mention that one. On the other hand, if they are in the network, I have no idea why they would be left off. Maybe the author is no longer allowed to make maps of Black Earth.

In the end, this looks to me to be a classic case of “a table is too boring, let’s make a map!” But a table would have been a lot more useful for people who actually want to find a doctor.

One Nice Thing: The author set the county label in a different type than the city labels, strongly distinguishing them from each other.

05
Aug
09

Concerning the Value of Human Life

Today’s effort is brought to my attention by one of our readers, Robin, who previously alerted me to the problems of Moon Maps. This map appears in a post on the OKTrends blog, where the authors analyze the geographic distribution of how people answer various questions on the dating site OKCupid. We’ll take one of them as an example:

human_lives_map

Obtained from http://blog.okcupid.com/index.php/2009/07/13/sweet-ass-american-trends/. Legend, title, and map combined by author into one PNG file.

Side note: the darkest red, for the highest percentage “No,” shows up as a single pixel at DC. So, I cannot claim that the map doesn’t follow the legend. Though, the DC dot is so unnoticeable it might as well not be. Putting an area on a map which is too small to be seen is a serious problem. Better for the authors to include an inset or two, showing the smaller states in a more magnified fashion (as we say in the business: at a larger scale).

There’s a fundamental error that a lot of people are prone to making with these maps (and, in fact, I didn’t notice myself doing it until I’d spent quite a while looking at these). If you look at the map above, or any of the others in the OKTrends post, there is always one state at each end of the scale. In the above example, Nevada is the brightest green, matching the far “Yes” end of the scale, and DC (thought you can scarcely see it) is the darkest red, matching the “No” end of the scale. This does not mean that 100% of the people in Nevada answered yes and 100% of those in DC said no. It just means that more people answered yes in Nevada than anywhere else. Maybe in DC 2% of the population said “yes,” and in Nevada 5% did, with all the other states in between. But this legend makes it appear as if DC uniformly said “No,” and Nevada “Yes,” with the vote being split in all other states. It is unintentionally misleading, which is the most tragic cartographic sin. The authors wanted to convey some information in good faith, but their communications became twisted and false.

The Yes-No labels on the legend are not the only problem. It’s the color scheme. It’s a diverging color scheme – the green and the red are opposite ends, and states shade toward one end or the other. There are many good reasons to use these schemes, but in this instance, it just makes it look like there are solid No states and solid Yes states, instead of states where a handful of people said yes, and those where a slightly larger handful of people said yes. The color scheme should be one hue, changing in lightness (say, from a pink to a dark red), and the map should show % yes votes (or % no) votes only. This still lets you pick out trends – Nevada has more Yes votes than other states, but it doesn’t mislead you in to a panic about the amoral citizens of Nevada and how they disregard human life. You can tell that “yes” is unpopular, but less so in certain areas.

Another reason to ditch the red-green scheme, which the authors use uniformly throughout their maps: It doesn’t always make sense when the options are other than yes and no. Here’s a map from later in their post:

Obtained from http://blog.okcupid.com/index.php/2009/07/13/sweet-ass-american-trends/. Legend, title, and map combined by author into one PNG file.

Obtained from http://blog.okcupid.com/index.php/2009/07/13/sweet-ass-american-trends/. Legend, title, and map combined by author into one PNG file.

These two colors, green and red, have certain connotations of positive and negative in much of the english-speaking world, so here, the scheme is a problem, unless you’re going with an implied value judgment where “Right to Bear Arms = Good” and “Right to Vote = Bad.” The people on the OKTrends blog don’t quite strike me as that type. Nor the type to greenlight an unequal valuation of human lives. And, of course, the legend does again make it look like Idaho is 100% behind giving up the right to vote. I’m guessing (and it is just guessing) that if you looked at the data, the vast majority of respondents in each state said they’d rather keep the vote and lose their guns. But Idaho happened to have the most people who were more fond of guns than voting (hard to hold that against them, given the political climate). Probably just a few percent. But people will come away from this map with the sentence stuck in their mind “Idaho strongly prefers to lose the right to vote.” I have belabored the point, so let’s move on.

Note that there are two states in the lower right corner which are not actually filled in with any color. Thanks for playing, Hawaii and Alaska! It’s OK — they’re just happy to be there, given how often they’re left off of US maps.

The shapes of the states are poorly generalized, and may in fact have been drawn freehand. Possibly just traced loosely, instead. But it has that feeling to it. Many of the states look slightly-to-moderately wrong. Look at my beloved Michigan – it’s become an amorphous blob. Saginaw Bay is entirely missing. It’s a sizable body of water. Like, nearly as big as some states. Also, I’m pretty sure Ohio doesn’t look quite like that. And Maryland seems to have taken over part of Virginia. Wyoming, though, is just as rectangular as ever.

I would be tempted to complain about the map projection, but since I think it’s freehanded, it doesn’t exactly have one.

Note the line around Michigan, showing the water border between the US and Canada as it passes through the Great Lakes. Except, the line only shows the border as it passes through two of the lakes. The actual border passes through two additional ones, north of Ohio, Pennsylvania, and New York – but you don’t see that section here for some reason. So, we have part of a Canadian border. And no Canada, by the by.

The map suffers from the dreaded island effect, where the US looks like it’s floating off in the ocean with no land nearby, and Canada looks like a great and forbidding sea, much like the Gulf of Mexico or the Atlantic Ocean. I might not call this a problem for this map otherwise, but putting a portion of the Canadian border in, and then not actually drawing Canada, makes this a serious issue.

Nitpicky: Both maps have a few artifacts here and there, like in Louisiana in the bottom map. A few pixels the wrong color.

One Nice Thing: It’s an unclassed choropleth, which I rather like. Most choropleth maps organize data into classes – say, one color will represent all states with yes votes between 0% and 5%, and the next color will be for states from 6% to 10%, etc. — the states are grouped. In an unclassed choropleth, such as those above, each state is given a color in exact proportion to the number of votes for one choice or the other. A state with 1.1% yes votes gets a different color from one with 1.2%, because they have different values. They’re never grouped together into the same color, as they would be in a classed choropleth. There’s a whole debate as to which is better, classed or unclassed, and I will not get into it here. I will just say that I am a fan of the unclassed choropleth, so I think it’s a Nice Thing.

Addendum: I am a bit embarrassed that I didn’t notice this at first, but these maps are very unkind to the millions of people with red-green colorblindness, as several commenters on the original OKTrends post mentioned. Here’s an approximation of what a red-green colorblind person sees when they look at one of these maps:

Run through http://colorfilter.wickline.org/ with protanopia filter

Run through http://colorfilter.wickline.org/ with protanopia filter

Designers cannot ignore such a vast population, and must take color vision anomalies into account.

03
Jul
09

The Town Spreadsheet

The Town of Blooming Grove, Wisconsin. A magical, pixellated land full of jagged lakes and rambling, rustic acres:

Setting aside for a moment the fact that the map has about as much detail as a Pac-Man level, I want to point out that this map is actually provided by the town as a Microsoft Excel spreadsheet. I am not kidding.

Look, the town population is only about 1700. That’s smaller than a lot of high schools. I can understand that they don’t have a trained cartographer on staff, and when someone decided to make a map, they were not going to be using the latest and most advanced tools that professional cartographers use. But if you want to make a map with giant pixels, try MS Paint (or your preferred equivalent). It is, in fact, significantly harder to make a blocky pixel map in Excel. It’s a spreadsheet program. It’s designed to crunch numbers and show you if the town can afford this year’s hayride. It takes some doing to get pictures out of it.

I have no idea what “Freeway Manor” is, but it sounds like it’s peaceful and has a great view. Is it a subdevelopment? A neighborhood? Not really clear on that. A legend of some sort might be handy, but all we get is “Anything in Green is the Town of Blooming Grove.” Anything. Any lush, verdant paradise you encounter? That’s Blooming Grove.

I think I’ll just stop right there, and leave the rest of the criticism and commentary to you, gentle reader.

One Nice Thing: Making a map in Excel is certainly an interesting challenge, and this person has managed to construct what is indeed, arguably, a map. For a first attempt, not bad. At, least I hope it’s his/her first attempt at maps ever. And that he or she hasn’t ever seen a map before. Please, please let that be true.

Thanks to my boss, Tanya, for pointing this one out to me.




Follow

Get every new post delivered to your Inbox.

Join 72 other followers