Thursday, October 9, 2014

Bad habits of North American cities: magicking distance from proximity

One fundamental principle of 20th century planning has been use separation, the idea that buildings used for certain purposes must be separated from buildings that have different purposes. Some of these usual uses are residential, commercial, industrial, institutional. Often, cities will have absurdly big single-use zones that are square kilometers in size.

Nowadays in Montréal, we have a metropolitan plan to increase density (simplistically defined as residential density) which results in residential zones taking much less place than before and therefore being much closer to commercial zones than before. Said like that, it seems like a very good thing, after all, density requires proximity to really achieve its potential. Yet, the mentality of use separation and car-centric planning still leaves its mark on newer neighborhoods.

How? Well, use separation in North America has become an art over the years. It's not just about keeping X away from Y in distance as a bird flies, it is also about deploying the magical art of making that which is near, far. An art that employs the street network to make direct paths as rare as possible.

Here is one example of this, in Laval, a suburb to the north of Montréal, near the Sainte-Rose commuter rail station.

You'll note that I identified the grocery, which I personally consider the most important store in a walkable neighborhood. In fact, if I were to make a checklist of criteria for valid TODs, I'd put in at number one:

"There is a supermarket or grocery at the station or within easy walking distance of it, with high-density residential areas within 5-minute walking distance, otherwise AUTOMATIC FAILURE"
Anyway, this area is pretty typical, with a commercial arterial street and residential side streets.

Just to the west of the commercial area, there is a new residential development with quite respectable density, which includes:
Relatively small single-family detached housing

Older semi-detached housing

New town houses built in groups of 3 or 4
Newer semi-detached houses
And some condos to top it all off
So this is some respectable density, and some of it within 10-20 meters of the grocery, great, right?

Except look at that street grid again, there is no direct link between this new neighborhood and the grocery. There is an unbroken fence all along the border between the commercial zone and the residential zone. So all the people who live in the high-density neighborhood have to take a detour by the north, through an inhospitable parking lot or on the sidewalk of a large street with poor enclosure and no buffer from cars to get to the grocery.

You have some housing that are built right next to the grocery, that they can probably throw a baseball on it, but there is a minimum of a 300-meter detour (1 000 feet), or 3-4 minutes of walking distance, in order to get to the front door.

This is a combination of two things.

The first is the psychological obsession with separating uses, even when they are right next to each other, we often build a fence or a barrier between the two to impose detours and make what is near, far. To make sure that they do not mix, no matter how near they are.

The second is a classic street network built all around the car, which may have sidewalks but which is still built with the car in mind. We can call it a "stream to river" model, with the small, sinuous residential streets going nowhere except onto collectors, which are a bit straighter with less stops, which flow onto arterials which have much higher speeds, a limited amount of intersections, most of which with traffic lights to favor speed and fluidity. This street network channelizes traffic onto high-speed arterials and keeps it off the residential streets. So the residential streets are like small, slow streams that flow into progressively faster and larger rivers. And as the arterials are where most of the traffic is and locations around it are the most quickly accessed locations, that's where businesses will want to settle.

The goal of this model isn't to reduce distances. That is inconsequential to the street network. What it seeks to optimize is vehicular speed, residential streets flow quickly onto collectors which flow onto fast arterials.

And indeed, here is an example of this at work, this is the quickest path as calculated by Google Maps:
1,3 km and 3 minutes
And this is the most direct one:
950 meters and 4 minutes
In many cases, the direct path doesn't even exist, like in the case of the newer neighborhood to the west.

So this street network actually works relatively well: people can still get where they're going fast thanks to the high-speed arterials (even when they have the same 50 km/h or 30 mph limit as residential streets), and traffic is turned away from residential streets. So you have speed and traffic-calmed residential streets, what's the problem?

The problem is that pedestrians and cyclists don't go faster on the arterials than on the residential streets. So these detours which don't matter to cars because they can get higher speeds have no benefit to pedestrians or cyclists, they walk or bike at the exact same speed no matter where they are. Worse, the high-speed arterials are often very wide, leading to very uncomfortable walking environments, or biking environments, unless bike paths are provided. Arterials also often lack crosswalks, allowing crossing the street only at rare intersections, where pedestrians need to wait one or two minutes for their signal after pushing the button. This forms a barrier to pedestrian trips in and of itself. Crossing even a moderately sized arterial is often the equivalent of walking another 100 meters or so (300 feet).

So this "stream-to-river", street hierarchy street grid is terrible for pedestrians and cyclists. It seems like a good idea only to people who assume that every trip is made in a car and that other people don't exist, or don't matter if they do exist. And talk about a self-fulfilling prophecy, on such a street network, most trips WILL be made in cars because that's the only mode of travel that is really advantaged by the street grid. Buses that follow the arterial can also be faster, but buses are not good for short trips because of the wait time, the fact is that you're usually better off walking to your destination if it's under 1 kilometer rather than taking the bus.

The result

So what's the result of this? We do have proximity between high-density housing and neighborhood stores, but that proximity is "car proximity", not "pedestrian proximity". It would have been the easiest of things to open up a pedestrian path to the front of the grocery, allowing all residents of the neighborhood to travel pleasant residential areas up to the grocery rather than taking a detour through a long parking lot.

A proper grid is generally best for pedestrians, if you already have a suburban-style street network, opening pedestrian paths to replicate a grid (or quasi-grid) for pedestrians would be a great idea.

Just for comparison's sake, here is a supermarket in a residential area of Obihiro, in the northern island of Hokkaido in Japan. Why Obihiro? Because it's one of the most car-oriented cities in Japan, but even there, walking, even if not that great, is certainly a viable alternative.
The supermarket is marked by the red bubble. Note first how compact its parking lot is and how they integrated the supermarket in the residential street grid rather than creating barriers between uses.
The contrast is striking, whereas in Laval the closest house to the grocery was about 300 meters away, in Obihiro, there are maybe 300 or 400 houses within 300 meters of distance of the grocery's front door, people who are very likely to walk there. They even added a pedestrian path in the rather longish blocks to the south of the grocery, even if it's hard to see on this image. And these are mainly single-family houses of no higher density than the new developments in Laval, in fact the condos are roughly 2-3 times as dense.

Density matters, proximity matters... and street network design matters. Unfortunately, many planners in North America have yet to shake off this mentality of separating uses and of putting barriers between them if they can't simply use distance as a buffer.


  1. Self-fulfilling prophecy is absolutely correct. It's a vicious cycle where the more it sprawls, the more car-dependent the citizens become, and the more they use their cars, the less they care about sprawl. Of course, we now know it's unsustainable, and in a decade or so we will end up with huge tracts of abandoned suburbia.

  2. Dendritic (branching, funneling, tree-like) is the word you're looking for with the "stream to river" analogy. Hierarchical works too.

    1. Thanks for the tip, I knew there was a word, I just couldn't remember it for the life of me.

  3. I think part of the reason street networks are designed in such dendritic and disconnected fashion is the huge (and IMO somewhat justified) fear of cut-through traffic. Car traffic sucks: it's noisy and dangerous and nobody wants it in their front yard. But in the previous Standard North American Design Pattern, the streets were designed as a uniform grid, with every street being equal and thus any street having a potential for cut-through traffic. There's an alternative though: the Really Narrow Street, inherently designed to keep traffic slow and low-volume. And because it's smaller and cheaper, you can have more of them and actually have a more connected grid for the pedestrians (down to ped-only shortcuts if you want).

    1. I would add that shortening blocks on their longer dimension has the benefit of 1) increasing porousness which decreases the road travel distance between two points and 2) force cars to slow/stop more often, reducing the max and average speeds.

      I also think one-way streets create friendlier neighborhoods as long as getting around by car isn't *too* circuitous.

    2. I also meant to add that organic, non-grid layouts helps calm traffic, by forcing drivers to periodically alter their vector, rather than blowing through an endless grid of small blocks.

      Long viewing distances with lots of stops/yields in between can cause driver anxiety, making the them focus on their far-off destination, whereas organic, shifting vectors force the driver to focus on the immediate surroundings, (like pedestrians!).

      They're also a lot less boring than grids.

    3. Curving streets do contribute to lower speeds, but curving streets also have the effect of confusing pedestrians and cyclists and making it harder for them to orient themselves.

      My preference is to a grid of small blocks and narrow streets, the presence of many intersections, poor visibility and enclosure is sufficient to deter most through traffic.

    4. I agree that preferencing curvy streets is risky and usually just leads to the "dead worm" cul-de-sacs of suburbia. It's true that they're superficially pleasant because your view/vista is always terminated on something other than the haze in the distance, but that's a very poor way to do it. It’s also somewhat of a windshield perspective on the situation. Pedestrians and cyclists are much more likely to be looking at the things abutting the street (shops, windows, other people, etc.) or in the more immediate vicinity than staring out in the distance. Any street with a good tree canopy can also close itself in to the extent you don’t need to worry much about what’s so far away.

      Granted the historically organic street "grids" of pre-industrial cities weren't really curved either, more like kinked. Curves were more difficult to build with stone curbs and hand-laid paving, not to mention trying to build the actual buildings to a curved frontage, so they generally avoided it whenever possible, making long tangents with as quick a transition to the new direction as possible. Because of the incremental growth pattern over time the various short blocks and funky intersections just kind of happened as they happened.

      Today, I see some new urbanists trying to recreate this sort of medieval street grid, but it rings hollow to me when it's done in a corn field. It's no more abstract and derivative than a strict grid or some curvilinear cul-de-sac development. If there's some pre-existing trails or terrain to work with, great, but trying to plan something that's supposed to look organic is just disingenuous and forced.

      There's other ways to get a well-connected street network without having to resort to a grid anyway. Early suburbs like Riverside, Illinois or Glendale, Ohio have excessively curved streets, but they all connect to one another. They’re still confusing, but at least you don’t have to turn around on some dead-end if you get lost. Then there’s hybrid grids/curves which form something of a spider web of streets like you see in New Orleans which follows the curve of the Mississippi River. More formal treatments that still create focal points would be L'Enfant’s plan for Washington DC, Haussman’s Paris, or even the Burnham plan for Chicago, all of which overlaid a network of radial streets on top of a grid (more gridded in DC and Chicago compared to the preexisting medieval Paris) to provide not only better movement options but several focal points for significant buildings and monuments. Even a strict grid can be significantly tempered by the introduction of squares at intersections such as in Savannah, Georgia.

    5. I didn't mean curving streets as in N.A. dendritic suburbs, but more the 'chaotic' suburban mesh as in Japanese cities and older European (medieval) cities, where all streets are well-connected to each other, and dead-ends (let alone cul-de-sac) are basically unheard of.

      Probably their biggest drawback is that traditionally they could get confusing - or at least get you turned around - if you're not familiar with the area, but this is increasingly moot in an age of nearly ubiquitous GPS.

    6. An alternative to a chaotic mesh/cellular block style is, if you're intent on lots of parallel streets, to use lots of T intersections: this also enforces traffic calming, but keep streets from being long, monolithic corridors.

      I actually developed a grid-like layout that I *think* is unique: (the population figures don't include intermixed small and medium sized businesses, green areas/parks, etc; 1 pixel = approx. 3 sqft/1sq m)

      Every road on it is one-way, and the flow is designed such that a driver has to slow/stop frequently, but never has to make a turn angle greater than 45 degrees. Additionally, the diagonal connecters add some leeway for longer vehicles down these roads, while maintaining a tight grid.

    7. Zeph: your layout is basically a grid that's been retrofitted for traffic calming. I can't say that I've seen anything like this, but I'm sure someone in Europe has tried it, because the "slightly bent grid" is a pretty easy way to retrofit traffic calming onto a standard grid. I have to wonder, though, have any designers of street grids tried to use video-game-style procedural generation to produce more "organic" grids? That seems like the only approach other than accretion that might produce a plausible result.

    8. @arcady: There are programs like CityEngine which allow for complex procedural city generation from a large set of parameters.

      Re. my street grid, I actually figured out shortly after I worked it out, that various points could be stretched and shrunk to make a more organic layout, and if you don't push and pull too hard, it can retain the approx. 45 degree turn in many cases. If you make a block overly large, adding pedestrian paths through them helps alleviate non-car travel distance issues.

    9. Zeph, that's an interesting concept, but I think providing a few straight streets, if only one very 400 meters or so (1400 feet) is a good idea if only to get surface transit line to be able to provide reasonably fast service. Though I criticized the street hierarchy in the post, I don't think the idea of creating arterials is all bad, I mainly disagree about a street network that eliminates porosity for pedestrians because the planners thought only about cars. Avoiding arterials that are too wide and too fast, creating barriers splitting cities is also important.

      The Canadian Mortgage and Housing Corporation (a public entity) has also started recommending a new street pattern they call the "fused grid" which is interesting. They fused the traditional grid and the suburban street pattern so that cars have to go to collectors, but pedestrians have in effect a grid because they use parks and pedestrian paths to cut residential streets inside blocs.

      There is also what I call the "quasi-grid", which is a grid where the direction of blocs are inverted so that many streets are relatively short and do not extend into the horizon, with only the collectors and arterials being completely straight. But to do this, you need to have short blocs.

      For instance, Sapporo has them in its farther areas. Look at the link:

      You'll see that most streets end in a T intersection every 300 or 500 meters, only collectors and arterials are straight all the way. So it looks from a grid from up high, but when you look at the detail, you realize there are plenty of short streets around.

    10. @sinval84, That's funny, I saw that grid design just this evening. I don't think Im sold on it, at least not as they show it there.

      I assume the grids between houses are property lines? Those huge back yards create some large distances between homes, and a number of the houses are on arterial streets. Even with good sidewalks, they are a long distance from the commercial buildings on the other end of the grid block. Furthermore, for a shortest-distance route, you end up having to walk thru two parks/green areas.

      For that matter, why are there a bunch of parks AND huge back yards? That seems like a huge waste of space. Don't get me wrong, I like parks, but if you do both, you're edging back towards suburban, not modern urban.

      Personally, I like the layouts in Hokkaido's cities least of all Japan's layouts, being too rigid, leading to long-distance views that feel more North American:

      But at least they're making porous street instead of "cul-de-sac streets with vague pedestrian concessions".

      It's like these north american (US *and* Canada) developers have never seen Japanese cities before (maybe they haven't!)

      Ultimately, I think that plans like the Fused Grid are over-engineering the problem. It's trying to tackle modern urban development, where large tracks of land are built all at once. They're built from an aerial perspective - as if anyone besides airplane passengers ever see it that way - with literally a "Master Plan". I think the Japanese have absolutely nailed it in most respects, at least when it comes to land use, zoning, and piecemeal development. It's a wabi-sabi approach to living that is about as far as you can get from our western approach.

    11. I actually think there's nothing wrong with the concept of street hierarchy, it's just the North American design doesn't extend far enough down: the smallest useful street is the Absurdly Wide North American Street, 40 feet curb to curb, which is wide enough to serve as at least a collector road. But in fact, the smallest useful street can be as narrow as three feet (an alleyway, or a walking path) for a pedestrian, and maybe 10-15 feet if you want cars to get through. Being able to throw those into the mix makes the hierarchy that much more tolerable. And given limited budgets for maintenance, wouldn't the best use for your limited square feet of pavement be in stretching them as far as possible, distance-wise, to provide the best access to the most area?

    12. There's really two ways to treat street hierarchy so it's important to distinguish between them. While hierarchical and dendritic can be used synonymously like I did earlier, that isn't always the case. A fully connected pre-WWII street grid can still have a hierarchy of major and minor streets, boulevards, avenues, side streets, alleys, etc. The important thing is that the only real distinction is width and prominence. Conversely, in the post-WWII dendritic system you have side streets, collectors, distributors, and arterials, all of which only funnel upwards to the higher order streets so there's little or no through traffic on the residential and side streets and only some on the collectors and distributors.

      The historical street grid which has a lot of connectivity and thus options and redundancy still doesn't have a lot of trouble with cut-through traffic because the side streets are narrower, sometimes full of parked cars, and usually have stop signs at every intersection. I'm thinking of Chicago neighborhoods specifically in this context, but there are a lot of examples. The point is that you could go "across town" even on some lowly back alley in many cities, it would just take longer. However, when the streets are too wide (which they usually are) then what little cut through traffic there might be is usually speeding, and this is more where the problem lies as opposed to traffic volume per se. However, when streets are blocked off to try to make it a more dendritic type of system, the problem starts to spiral out of control because now the major streets become more congested due to a lack of options and alternatives, so the pressure to cut through on the other streets that are still open increases, thus making it more likely the residents of those streets will try to close off access as well. Later, rinse, repeat. You see this in the newer subdivisions around Phoenix, Los Angeles, and even Chicago where there's a pre-existing connected grid of major streets (formerly farm roads) but the new developments within those blocks are mostly self-contained and don't connect through like the pre-war developments did.

    13. @arcady: I don't disagree with street hierarchies either, and agree with what you say about sizes. My sample street grid was more just an example, an exercise in potential density, and using a shopping mall and its parking lot to contrast.

      Realistically, for that size, I'd probably put 2 east-west collectors (1 lane each direction) and 1 running north-south, all of them running into Mall Road.

      Japanese suburban streets are still hierarchal but 1) their smallest streets are narrower and more porous, their connectors are much narrower (typically 1 lane in each direction), and even most arterials aren't much bigger until you get into city centers (e.g. downtown Tokyo) or actual highways. Contrast with North American cities where arterials are highway size, frequently 3+ lanes wide.

      @Jeffrey: I agree on your points too - the dendritic layouts miss the forest of the trees: they focus on less traffic, though not necessarily *slower* traffic, where many dendritic subdivisions have wide, long arcing streets which encourage (or at least don't discourage) speeding. The net effect is that fewer cars overall go along there, especially as you get to the end of a cul-de-sac, but it ends up prohibiting any sort of street activity other than driving, because while the rate of cars is low, when they do come, they're not going slow, and prohibits lots of alternate transport methods by increasing distances between start and end points.

      The long arcing streets also try to reduce the "long endless street" effect, but because there are no stop signs except at junctions/intersections, all they end up doing is creating long wide street with poor visibility. Combined with on-street parking which creates more visibility obstructions, huge SUVs everywhere, which are typically taller than most people, the street becomes a hostile car-only course where pedestrians have to assume drivers don't see them; bicyclists have to pretty much ride on the sidewalks to avoid speeding cars.

      Speed bumps are an attempt to lowering speeds, but I feel they end up just annoying most drivers.

    14. I would argue that there are not two but three types of street patterns: post-war dendritic (cul-de-sac hell), pre-war grid (theoretical cartesian perfection), and Traditional City accreted (historical hodgepodge). And the Traditional City accreted network actually tends to be relatively hierarchical, compared to the 19th century grid, with a much wider range of street widths and lengths. Its main problem was that it was hard to adapt to massive growth, as streets gained much more demand than they were originally built for, which has to be fixed either by some central planning providing wider rights of way ahead of time, or expensive demolitions later.

    15. @arcady: Again I agree - the Traditional City (TC) layout has been the primary way of building for probably 95% of human history. Like you say, increased street demand tends to force wider streets.

      The usually 'obvious' modern solutions is to increase density, and also to widen streets to accommodate the increased area traffic. We see this in Chicago, NYC, etc.

      But IMO the *better* solution is the TC one, distributing outward instead of concentrating upwards, so that a large but not cramped population/usage density is retained.

      (In TCs, this was a necessity since there were practical engineering limits to building height (not including cathedrals etc. which were large, expensive, long-term endeavors). With a maximum population density reached, foot traffic was still the primary mode of transportation, and if you made the streets too large, you lower the number of buildings people could visit.)

      It is, or can be, a modular design so that rather than concentrating businesses in one location, you may have a number of the same or similar business in multiple locations, each more easily reached by foot.

      For example, one Walmart megastore servicing a region 20 miles in radius, versus a number of smaller markets, hardware stores, etc. distributed over that same region. The megastore forces everyone to concentrate in one area for their needs, and forces them to drive out of their way. A layout with lots of smaller stores distributed more evenly and repeated where necessary lets people make more frequent but shorter trips, e.g. walking to a corner market for some milk or ingredients for that night's dinner, or to a hardware store to get a sack of nails.

      In very rural regions, populations are spread unecessarily thin so having a distributed set of repeated stores can become untenable without the collective usage to support the store, and this is there the Walmart-for-everything situation seems to thrive, killing off smaller local businesses.

  4. Lots of interesting stuff here. It's comforting to see a traffic engineer that gets it.
    Here are my musings on the evolution of Denver's street grid and the origin of curvilinear streets: