Toronto Density Map, 2006

[scarberiankhatru]

Can you change the colours manually or do you have to pick a set? You should pick less washed-out colours because the boundaries between blue and yellow are easily lost. If you had 5 or even 6 zones instead of 7, you could go with old standby schemes like white-yellow-orange-brown, or green-yellow-orange-red. Or you could keep the general colour scheme you have now but make a deeper blue the lowest zone, and use a bright pinkish-magenta for the top.

I should have also considered dot density, but I've never been a fan of that style of map.

Dot density just guesses where the dots should go within each tract, though.

This was a quick map intended for those already familiar with the geography of the Toronto area (hence no street names or labels), who would hopefully be familiar with the ravine and parkland system. I'll see if I can find parkland data to add to this map, but parkland does still reduce the density of a district.

You could take out parks, but that would leave industrial land, transportation land, utilities, schools, etc. Everything reduces density...everything except residential land uses. You could always use a property map and colour every residential lot in each tract the appropriate density zone colour, which would basically reduce "residential density" to a visible function of household size, backyard size, and number of storeys...I don't know how much work that would be, though.

 

[CDL.TO]

Can you change the colours manually or do you have to pick a set? You should pick less washed-out colours because the boundaries between blue and yellow are easily lost. If you had 5 or even 6 zones instead of 7, you could go with old standby schemes like white-yellow-orange-brown, or green-yellow-orange-red. Or you could keep the general colour scheme you have now but make a deeper blue the lowest zone, and use a bright pinkish-magenta for the top.

I can set the colours manually. The software package I am currently using seems to have a odd propensity for pastel colours in their included sets.

Dot density just guesses where the dots should go within each tract, though.

It's not a guess, it's intentionally random. Hence why I don't like it... you often end up with false patterns.

You could take out parks, but that would leave industrial land, transportation land, utilities, schools, etc. Everything reduces density...everything except residential land uses. You could always use a property map and colour every residential lot in each tract the appropriate density zone colour, which would basically reduce "residential density" to a visible function of household size, backyard size, and number of storeys...I don't know how much work that would be, though.

I wouldn't take out parks, it would change nature of the data. However, I could add the parks on top (maybe as a dark green outline) as a reference. I don't even want to think of how much work it would take to do that second idea manually, and I can't think of any existing data sources with that info. Overall, these suggestions would end up with a map of built form, not of density.

 

[urbanfan89]

It will be interesting to compare with other cities in Europe and North America and Asia. So we know our position in the world.

 

[scarberiankhatru]

I can set the colours manually. The software package I am currently using seems to have a odd propensity for pastel colours in their included sets.

I just pulled out some of the maps I had to make for my one GIS course and noticed I used a blue-scale map to show total population of each Toronto CMA census tract...blue-scale, of course, is a dubious choice, but all the others looked like crap and everyone in the class was too scared to do anything other than explicitly follow the instructions we were provided with, (like pick our own colours). Other maps were a questionable shade of lime green.

It's not a guess, it's intentionally random. Hence why I don't like it... you often end up with false patterns.

Intentionally random guesses

I wouldn't take out parks, it would change nature of the data. However, I could add the parks on top (maybe as a dark green outline) as a reference. I don't even want to think of how much work it would take to do that second idea manually, and I can't think of any existing data sources with that info. Overall, these suggestions would end up with a map of built form, not of density.

Yeah, there's no point taking anything out unless you're going to take it all out...I just kinda thought you were going to remove parkland from the way you said it. It would isolate built form, but one could gather together variously sized areas of this built form data (like, conveniently, census tracts) to create an even more useful density + built form map.

Don't even think about making it, though...it's just a "wouldn't it be neat if" map, like a flow map that showed exact ridership along the full length of every TTC route, or a city-wide walking distance to Tim Horton's map (but one based on real-life walking data, not simple radii). If I had the money to invest in real estate I'd want a McMansionization map, showing the site of every McMansion in the whole city for scouting purposes. Or a series of maps that showed housing price increases over time overlaid with Starbucks locations.

 

[CDL.TO]

"I just pulled out some of the maps I had to make for my one GIS course and noticed I used a blue-scale map to show total population of each Toronto CMA census tract...blue-scale, of course, is a dubious choice, but all the others looked like crap and everyone in the class was too scared to do anything other than explicitly follow the instructions we were provided with, (like pick our own colours). Other maps were a questionable shade of lime green."

Well, technically your classmates were correct (when you're dealing with a picky cartography professor). When it's a continuous variable, it should be a scale of one colour other than blue, which is reserved for water. I took a leap of faith and assumed that the members of this forum knew that Toronto isn't an island.

"It would isolate built form, but one could gather together variously sized areas of this built form data (like, conveniently, census tracts) to create an even more useful density + built form map."

That would be a good map and not to hard if I can get property footprint data from somewhere like the city.

"like a flow map that showed exact ridership along the full length of every TTC route"

I tried to do this for the subway, but it seems that no one has this data. Not even the TTC knows what the loads are on the different segments of subway.

"a city-wide walking distance to Tim Horton's map (but one based on real-life walking data, not simple radii)."

Pretty easy if you have the right software, such as one used for traffic modelling.

"Or a series of maps that showed housing price increases over time overlaid with Starbucks locations."

That one would be easy to do. Too bad I just don't have the time.

 

[CDL.TO]

EDIT: The maps originally in this post were created using an incorrect methodology.

-------------------------

I just updated the original map in the first post with (what I hope is) a better colour scheme.

Here are two more maps showing the same Census density data in a different way. With these maps, data has been normalized to average out the results. Each district in these maps represents the average of density of all districts within 1km of the centre that district.

 

[Edward Skira]

I didn't realize Lawrence and Kingston Road was so dense.

 

[scarberiankhatru]

There's a few strange bits on these newer maps, in Scarborough, at least; density that's been "moved" by reworking the data (the Galloway area is one such area that's red when it probably shouldn't be, not denser than the Markham & Eglinton area, for example. Maybe within 1km it just nicks enough denser zones to average out that high. Another area is the Steeles/Midland/Finch/Kennedy block, which is almost uninhabited but has absorbed the density in neighbourhoods immediately east and west). It's interesting to see how the Spadina extension plus a DRL would hit almost every 'red blob' in the city.

Well, technically your classmates were correct (when you're dealing with a picky cartography professor). When it's a continuous variable, it should be a scale of one colour other than blue, which is reserved for water. I took a leap of faith and assumed that the members of this forum knew that Toronto isn't an island.

My classmates also chose the blue-scale, or the green-scale or the red-scale...the blue was the only one that didn't have washed out or nearly identical colours throughout the scale, so I found it easier to work with. It was a rushed summer school assignment and I didn't have the time or the confidence in the software (which was wont to crash) to tweak, say, the red-scale to make it legible. The prof didn't care which colour I picked...but he did take 0.5 marks off for an incorrectly sized margin.

 

[Glen]

Looks like a subway across Finch would have served more people than the Sheppard line.

 

[scarberiankhatru]

Looks like a subway across Finch would have served more people than the Sheppard line.

Looks can be deceiving...

 

[CDL.TO]

This whole thing is a bit of a thought experiment of mine that I'm sharing with the forum to hopefully spark discussion. These maps were a suggestion of a coworker.

The problem with the method used to make the two most recent maps is that's it's based upon simple averages. So say you have a tiny district consisting of two apartment buildings of 100,000/km2 which is adjacent to large district containing parkland that has only 1,000/km2. The average of the two is 50,500/km2... but in reality that's nowhere close to the reality.

This is causing a major skew in those suburban nodes that contain some large apartment blocks like Lawrence-Kingston, Jane-Finch, and around the Allan.

It's not useless, but keep the methodology in mind and take it with a grain of salt. When I have the time, I'll remake the map with weighted averages that will correct this problem.

 

[CDL.TO]

I corrected the improper methodology used above, and re-made the map. This map accurately shows density with each zone showing the average of all zones within 1km.

The classes are a bit different on this one. Based upon the suggestion of Scarberian, the highest colour class was split in two. So the lowest 5 colours each represent 16.9% of observations, while the top 2 colours each represent 8.4% of observations.

This methodology was successful at averaging things out. The densest zone in the city has a density of 620,220/km2, but when averaged with surrounding areas the densest zone has only 24,144/km. Most of the densest zones are in and around St. Jamestown.

Probably the biggest issue is some skew in isolated locations. For example, the unpopulated Sunnyside beach was affected by its proximity to Parkdale and was labelled as high density and many developed areas adjacent to farmland got completely wiped out.

 

[scarberiankhatru]

Great map...it really shows which neighbourhoods have hordes of residents, and it shows what a useful line the DRL would be, too. If you have time to keep revising it to get it 100% blip-free (if any more can be done), go for it. I wonder what it'd look like with the 1km buffer changed to 500m or 2km.

 

[Urban Shocker]

CDL.TO: I like your new, more effective colours.

The hot red advances and the eye goes to it first, then to the pink, then to the orange. Everything's nicely differentiated. The colour sequence tells the density story, and the statistics confirm it.

But I wonder if the last two colours should be reversed? The dark blue is used for such a large area, and is considerably stronger tonally than the icy pale blue - and I find my eye drawn to it after pink/orange.

Or perhaps it could be lightened up a bit? There's plenty of differentiation between it and the pale blue.

 

[scarberiankhatru]

But I wonder if the last two colours should be reversed? The dark blue is used for such a large area, and is considerably stronger tonally than the icy pale blue - and I find my eye drawn to it after pink/orange.

That's a good thing...it means the densest and least dense areas jump out, as they should.