In the first post in this series, there were some great recommendations on how to change the land footprint graph. In my original version, I graphed the number of blocks in our region that fall into a given land area per person. The suggestion was to instead plot the number of people living at a given land area per person. I’ve taken it a step further and plotted percentage of people living at a given land area per person. For example, 0.52% of people in our region have a land footprint of around 600sf per person, which corresponds with the Capitol Hill picture below (the bottom left). Graphing using the percentage of total population will allow me to compare regions of different size on the same graph, which will be useful in my next post in this series.
Matt,
Perhaps this is better for an email, but what are your data sources? I’d love to dig at them myself…
Just the good old US Census Fact Finder. First choose the 2010 SF1 100% data from Topics, Datasets. Then use the Geographies tool to grab only the counties you’re interested in (Name, Block, Within State, then add county groupings of blocks to your selections). Then I choose what data I want from this set – all I need is “Geographic Identifiers” from the last page of search results (it has population, size of block, etc.) but there’s a lot of other information included in the census (I also recommend the ACS data for even more including commute patterns). Then download as a massive excel file and play with it from there.
Go matt, go! This is a nice way of getting the pitcure across.
The article in the Seattle Times Northwest magazine section on the Bullitt Center took aim at density in pretty slipshod way, conflating several measures while suggesting that maybe Seattle was nearing its limit for a place that mets our biophilic needs. I think being able to show how places we know to provide a high quality of life also can have a denser footprint could help to counter that sort of fuzzy-minded opposition to urbanization.
Or not.
Once you add value judgements to data, any conclusion is possible.
Although linking typical QOL metrics like physical and mental health, crime rates, education, cost of goods to Land Footprints would be useful.
I applaud the use of this unit of measurement, land footprint.
I’d like to see this as an integral, so you could line up percentile, ie, the median is X, the bottom 10% are these, the top 10% are X, etc.
I’ll have to dig around.
I found it interesting that in the process of setting up this data to fit my needs, I’ve effectively recreated a basic tool of statistics: the density curve (an idealized version would have the area under the curve equal to 1). I did not intend this, but I think I’ve re-learned a bit about statistics in the process.
If it would be useful to have those numbers I can easily find the median, as well as any useful markers (such as the top/bottom 10%, or even a standard deviation). In the next post I’m not going to use individual points, but instead a smoothed curve for each city (I belive I averaged over 5 or 7 data points).
I would recommend an Savitzky-Golay filter over simple averaging; tends to give a really good smooth that mostly preserves the peak widths.
Excellent. I hadn’t heard of SG filters, and I did notice some if the peaks weren’t as high after smoothing. It’s probably too much work for a minor correction, but I appreciate the tip.
This is cool, it seems like you are on the right track, but this chart is almost too abstract to stand alone. The immediate follow up question for me is how many cumulative blocks are at each level. Could you maybe plot vertical lines on the chart grouping quartiles? The way it is now, it looks like you could make an anti-density argument that people really want to live in the “long tail.”
I certainly could mark the quartiles, but I’m not sure how useful it would be. Though I’d be happy to discuss it. I don’t understand the anti-density argument you’re proposing.
I think what Mike might be saying is someone could look at that graph and say (incorrectly), “look at all those dots to the right of 10,000 sqft! People really want to live on huge…tracts of land.” Except the dots represent blocks, not people.
If you are going to put some quartile lines on that, perhaps it should represent a percentage of the total population. That would counteract that incorrect initial impression. I would wonder where the line might be for 50% of the population. Does 50% of the population live to the left of the 5,000sqft marker? 10,000?
@NSBill My hat off to you for working Monty Python into a land footprint discussion.
I just realized something that will help: there are just as many dots between 0 – 5,000 as there are between 25,000 and 30,000. Each group of 100 person/sf is represented by a dot. There’s a dot at 100, a dot at 200, etc.
“Where” people live doesn’t equate to “want” to live. You could do a similar plot for dollars per hours worked and the peak almost certainly wouldn’t be what people “want” to earn. Two areas may have an identical land footprint but be polar opposite in desirability and sustanability. The whole notion of “land footprint” seems of dubious value as any sort of stand alone metric.
Yeah, this chart gives a great feel for how the entire city is laid out and many blocks exist at each “density level” (i.e. a lot of our city land is at >10,000sqft/person!). I think this chart actually does a better job of doing that than the last chart, for the same reason this chart makes it hard to perceive how the people are spread out across the city:
Do all those .2 percent dots? I hope not, but there’s no way to instantly tell. I think taking the integral would get you there.
I’m not trying to be critical, but constructive!
Also, does each dot represent the same amount of actual land area?
Correction: *Do all those below 0.2% dots add up to more than the above 0.2% dots?
This is a really cool graph, and I can’t wait to see more cities. Maybe I’ll put some together myself :).
This is fantastic. Please keep going.
I would also add my name to the list of people asking for a breakdown (preferably with visuals).
NSBill: That is pretty much the point I was trying to get after, thanks for clarifying.
Matt: Does that make sense? It just looks like there are an awful lot of blocks to the right of 10000, so many in fact, that I can’t intuitively compare the percentage of population that lives above/below a given RLF threshold.
Yes, I get the issue now. As it changes to a line in my next post that issue should go away.
As for trying to guess how much of the population lives at any point, here’s a hint: since the total area under the curve equals 1, the proportion of area under any given curve section compared to the total is equal to the percentage of the population living in that condition. Take the 15,000 point. Almost the whole volume of the curve is to the left of that point, meaning a vast majority of all people live more densely than that.
Ok, I guess it makes more sense when you explain it that way. Thanks.
Ahhh…ok…I see how that works. The density curve as you were explaining earlier.