Infill Housing – Rescrambling The Omelette

Is There A Theory Hiding In The Urban Designers’ Corner?

What does a ‘good’ environment look like, no matter how we get there? The public domain of the street is our primary space for social interaction and the development of community value. Sure, that doesn’t happen at densities below 12hhu/ha net (trad suburbia usually sits around 10), so a denser suburbia is something of an imperative. But it also doesn’t happen because of the way we currently configure our residential streets. We treat the whole road reserve as a moving vehicle environment, and then as an apparent afterthought include a few minimum extras like pedestrians, trees, pipes, and so on.

The theory we’re after is called ‘Link-Place’. Every street has a % of Link function and a % of Place function. The ratio changes according to what’s happening in the movement network. A motorway is 0% Place (of course); a quiet residential street is mostly Place. When we make those residential streets for cars, Place is surrendered to Link. When vehicles take over take over the Place space, moving through and parking in the pedestrian margin as a priority, then residents stop using the space. They defend themselves with thick hedges and high walls, isolated behind solid locked gates with intercoms. Social life fails.

Single-lot infill creates a lot of rear lots that have no interface with the public domain. In an already bad environment, infill dumps 5—10x more vehicles per original lot into what should be social space for the community. So even fewer people engage; more people use their car for ever shorter trips to avoid the spatial conflict. Apart from the psychological impacts, infill degrades the neighbourhood in many other ways as well – safety/accident rates, rubbish collection, loss of vegetation cover, increased crime.

So we need two things – a higher density (30hhu/ha net) is a good target to start with, and we need, absolutely, a reconfigured public realm to go with it. Infill only supplies one of those, and it makes the other much worse. How much worse? Now we get to find out.

Finally, We Get To Test The Theory

A useful study case was provided as part of a 30ha greenfields housing application in a rural town that was also zoned to allow for infill housing. The application included a Residential Subdivision Performance (RSP) analysis of a 30ha area of existing suburbia within the town that was typically configured in terms of context, proportion, grain, open-space, etc. It was a relatively simple process to pick the most developable 10% of properties and fill them in, then run the RSP again to see what changed.

(RSP is a tool for measuring urban quality and functionality that operates at the level of a subdivision. More info on how it works can be found here.)

Although it’ll ruin the suspense, to save you having to wade through to the end of this text, here’s the simple result:

• • Infill resulted in a 5-point reduction (25% worsening) in overall urban functionality/quality
  • Infill resulted in a 17—18-point reduction in onsite and local function/quality
  • Reduction in function/quality is directly related to the number of new dwellings created, not the number of lots subdivided
  • As a rule-of-thumb, overall urban functionality/quality of the whole community is reduced by 1 point for every 30 new infill dwellings created

Oh well, urban designers 1; council 0.

The Analysis Breakdown

What did it look like in more detail? The most feasible lots for infill were selected, and they were feasibly subdivided to maximise yield. A good idea of where they were can be seen on the data visualisation for ‘Value’ (below). Two RSP criteria are responsible for all of the functionality decline – Frontage and Value.

Frontage

Checking the comparative Frontage data visualisation (below) the score is punished by only one field value – ‘Rear Lot’. How rear lots affect urban functionality was touched on above.

Almost a third of the properties in the existing subdivision are rear lots. This is on the high side compared to many subdivisions, and since this area contains some higher urban values (eg the coastal reserve) it is the rear lots that pull the overall RSP score down to 19. In the infill scheme, all the 198 new lots (2/3 total increase) are rear lots, reducing the frontage score by 18 points (11>-7). If you were in the business of assessment, you’d say that was a massive drop.

Value

Referring to the comparative Value data visualisations (below), although nearly 200 new lots have been created, they are all in the ‘Low Value’ data field. Again, a massive drop of 17 points (5>-12). Originally 12% of lots were low-value; after infill, 53% of lots are low-value. You have to ask, is it worth it?

At this point it’s worth noting that in the RSP, Value doesn’t mean the $value of the property – although that is one factor. Instead it’s an aggregate measure that captures the actual value of several criteria as well any ‘affective value’ – how a lot’s properties affect other things. For example, they may devalue open-space quality and function by changing spatial relationships; higher density may improve some matters at the neighbourhood level; and so on. In this way Value functions as something of a multiplier to make the effect of complex performance interactions – for better or for worse – more visible.

Using The Results

Finishing up on the scores, the RSP showed an overall 5-point drop in urban quality and functionality (a 25% reduction in the existing urban quality and functionality) within a 30ha suburban neighbourhood after infill (see image below). We were interested to see if this could be converted into a simple rule of thumb relevant to decision-making at the planning level. 198 divided by 5 points – is that 40 infill lots equals a 1-point performance drop? Or is it 198 lots per 30ha? Therefore 1.3 lots/ha equals a 1-point drop? And so on, confusingly.

We ran four more iterations of the model, two with lower infill levels and two higher. Data was generated and compiled for frontage, value, score, %, differentials, yield, and then graphed in a number of ways. Whatever the combinations, a distinct and consistent pattern emerged regarding the distribution of performance reduction. The rate of performance change is much faster initially before continuing to decline at an ever-slower rate. For instance, performance might decrease 4 points for the first hundred infill lots and then only decrease 2 points for the next hundred.

Congruence mapping (see image below) of the data permutations shows the general consistency of the change gradient. 0,0 is the intersection point almost exactly on the halfway line for the number of lots and the 2/3:1/3 line for the change in gradient. The single best fit for the mapping is, unsurprisingly, the Rear Lot data field from the Frontage category.

The change in the rate of degradation is intuitive, as well as being represented in the data. The intuitive aspect is that the effect is cumulative. One infilled lot in a neighbourhood isn’t noticeable. A few more are, but any effect is unlikely to be picked up at a human scale. A lot more and it feels like a disaster. Similarly for overall performance – a 1-point reduction is hardly noticeable, but a 5-point reduction is.

What the data shows is that the first 100 infill lots are responsible for one third of the degradation. The second 100 infill lots degrade the environment at double the rate. That’s significant. A slow start, then the avalanche.

Yes, all very interesting, but what we were after was something simple and useful. It could easily afford to be approximate. A rule-of-thumb. A rough guide. To that end, taking into account the much stronger effect at the beginning of the infill process, we worked around all of the numbers and relationships and came up with:

30 new infill lots will reduce suburban quality and functionality by 1 point (of a theoretical maximum 100-point score).

The effects we’re talking about here reside in the immediate neighbourhood. The 30ha area seemed like a good figure to us for an immediate neighbourhood, but maybe 50ha or 60ha aligns better with the real world? It would be easy enough to determine a class of mesh-block for this purpose, in which case the numbers above might change a bit, although we don’t anticipate by too much.

Is there a low limit that’s not being picked up here because of the limited sampling? Meaning, are 10, 20, or 30 infill lots not too bad within 30ha, but we shouldn’t allow as many as 100? A problem for Planning is how you stop the infill roller-coaster once it’s underway? Planning, a practice of ‘betterment’, should never condone – let alone encourage – the diminution of public good.

Arguing the fine points is all a bit moot when what’s really been shown here is that single-lot residential infill is, measurably, not a good thing. However, intuition coming in again, at some point, or rather, at some size, the parent lot will start to acquire the properties of true brownfields redevelopment when negative urban effects can be mitigated. What is that size? 2,000m²? 3,000m²? Now that’s research worth doing.