Tag Archives: Megastructures

Feasibility Study

Here’s where I left it in Sky Rectangle – a proposal for interlocking back-to-back apartments arranged in rows half-stacked and half-terraced, with both the apartments and access corridors illuminated and ventilated by 4m x 8m shafts open to the sky and corridor.

In that post I mentioned how the images above are just my impressions of what I expect the level of illumination to be. Rendering parameters can be adjusted and photographs can be manipulated. The only way of being certain would be to construct a full-scale model for some chosen latitude and observe it for a year. I didn’t do that. However, I was recently in an office four floors (approx. 12 metres) from the top of a lightwell measuring approximately 4 x 8 metres. This is what it looked like from the inside.

And this is what the bottom of the lightwell looked like from the street-level driveway entrance perhaps 20 metres away. In both sets of images you’ll have to allow for a certain amount of ambient light coming from behind just as my proposal will have a certain amount of sideways light contributing to the ambient.

This level of illumination looks sufficient but then, these photographs were taken about midday (Sun alt: ≈ 73°, azimuth 7°) in midsummer (Jan. 31) in Perth, Western Australia (lat. -31.95°. long. -115.86°).


And this is what that proposal is going to be inserted into – five levels of shopping mall floor-plate, each with an area of approximately 35,000 square metres and with the long sides oriented approximately east-west. Within that floor-plate are atriums, openings for escalators, and cores for elevators, fire-escapes and utilities. There are also double height cinemas, a double height ice-skating rink and a swimming pool, all of which I’ll ignore. Conditions on the periphery are not uniform. Unusually, the north and east side also have three levels of external deck access while the western corner has five levels of externally linked terraces. I’ll ignore these special features because I’m attempting to derive general principles.

Basically, the problem is one of inserting a regular spatial system into a structural one that’s regular only in parts. I don’t know how this is going to turn out. Disjunctions between the two systems will occur on the periphery, around the atriums and around the cores where they ought to be of most benefit to the most people.

• • • 

It was a worthy idea, and all the disjunction space was indeed around the periphery, the atriums and the elevators but, as you can see from the Level 2 layout below, there was just too much of it. I was only getting about 100 dwellings per floor. [At this time, I was still assuming an ice-skating rink in the middle on the lower side.]

The area of the floor-plate minus openings is about 30,000 m2. The footprint of each pair of apartments is 128m2, but another 64m2 needs to be added for the access corridor (on one side) as it is be shared by adjacent group of dwellings. The total built footprint for 100 units is therefore 12,800 – a little over one third which is not great. This is largely the result of the basic (paired) residential unit being 8m x 32m. This length creates much disjunction, particularly at the short ends and around elevators. Having occasional unpaired units 24m long instead of paired ones at 32m length also reduces efficiency. Staggering alternate floors doesn’t make much difference as what one gains on the even-numbered levels is lost on the odd-numbered ones. Despite that, or perhaps because of that, the overall environment doesn’t look bad.

This proposal has the views out that I’d envisaged but the best possible situation shown in the image below doesn’t occur often enough or for long enough as the longest “street” has only four units each side.

These apartments are fundamentally different from those in conventional towers. They’re more like townhouses because, although the voids are the primary source of illumination and ventilation, they also function as buffers between the private space of the units and the shared space of the streets.

This next plan is sufficient as proof of concept but I’d like to increase the floor-plate efficiency from one third to at least one half. There are about 600 units here and this doesn’t seem enough for the project to go forward. I’ve identified four problem spots and each of them are caused by the long basic unit encountering an edge or an obstacle. Alternate levels will have the same problems but in different places. Not unrelated is the fact that getting more units in is a bit like designing a car park in that the more double-loaded roads you have, the more efficient it will be. Even looking at this image below, there clearly aren’t enough of those double-loaded access streets shown in the image above. The vertical pattern of streets should be more apparent.

1 The levels above and below have the best fit for the units and this edge condition is the direct result of staggering alternate floors. This can be solved by fitting half-bay (approx. 4m x 8m) two story units between the voids. These units would have no outdoor space save for the void outside their habitable room windows.

2 A different variation could make use of some of the space to the left of the elevators between situations 1 and 2. Nothing can be done about the 16-m spaces either side of the elevators in 2. It might be helpful to have a variation with the standard for unit one half and shortened (by one bay) for the other, interlocking half.

3 This area is isolated by voids and is not large. This suggests breaking the pattern of voids and having standard three-bay units as well as specially designed two-bay units stacked every level.

4 The same might have to be done for all of this sector as none of its areas are large.

The differing overall lengths of these variations account for instances where there are sufficient columns but insufficient length for five-bay paired unis. These variations tidy up the edges and fill in some of the spaces (that would still be necessary even if the unit and corridor positions didn’t alternate). There’s no point rotating the direction of voids and units 90° because larger (longer) areas won’t occur anywhere except the leftmost edge in the above image. Moreover, the streets would be terminated by elevator cores while the same number of variations would still be needed.

On levels 4 and 5 is a two-storey ice-skating rink supported by beams on level 3 but it now makes sense to build over this to provide approximately 24 additional units. Each of the elevator and stair cores is in reality different but I want to keep them generic at this stage and not indicate the internal arrangement. My only conditions are that there be four metres of space on all sides and that their positions not be hidden any more than they currently are.

[In the existing mall, cores on the west side occur along internal corridors that link to the external access corridors. The outward-facing units and the inward-facing units are serviced by a central corridor Similar transverse routes occur on the east side except they terminate at service corridors along the eastern edge.]

For now, I’ll design those more compact variations that ought to squeeze some more efficiency out of the floorplate.

If I don’t reach 50% efficiency, then I’ll scrap the staggering and instead design a typical floor along the lines of the above with units facing outwards, units facing inwards towards the atriums and inner units lit and ventilated only by light wells. The units will still follow the column grid but in the other direction. As a solution, I don’t like having a horizontal hierarchy of light as well as a vertical one. If I adopt this arrangement then, in most instances, the ends of these inner access corridors will have elevator cores and, because of that, indirect light from both inside and outside.

Rather than pursuing that Plan B, I did this.

The yellow units are the main unit type but there are now also five other variations. There will be about 180 units per level which is almost 1,000 not counting the basements. I can still see five instances where paired half-bay units could fit but it’s approaching as good as it gets for its premises. It is now easier to see the long double-sided streets that make most efficient use of space.

I won’t bother doing something similar for the alternate floors as, in the above, you can see there are three different conditions for how the units “meet” the elevator cores. Although the situations for the same core will differ, the principle of using the variations to account for shortfalls in length will still apply, only in different positions.

Another option is to start again, and use this study and the one before as the basis for an optimized Mat-rix House with a primary structure FF of 5.5 m and a column grid of 5.5 x 5.5 m. This would (just) be large enough to to arrange a car park (if required) and would also allow either steel or mass timber structure. This structure would be more regular and, since there is no need for atriums to bring light and air, or to add “incident” for the sake of it, elevators and stairs would best placed on the periphery. This will be for some other time.


The demonstration mall is on the left in the image above, and has approximately 1,000 two-bedroom apartments. The two levels of basement car parking extend two bays past the building’s east and west sides so, if B1 and B2 were opened to the outside by removing the slabs in these extensions, it would be possible to have another 400 additional units, bringing the total to 1,400.

To the right of the mall is a typical high-rise residential complex with each half of a paired tower having 4 apartments x 30 storeys = 120 apartments. Let’s say the mall’s width is equal to three and a half tower halves, and that its length is equal to four rows of three half towers. There are approximately 1,440 units.

The spatial efficiency is about the same but the mall has six 5.5 metre levels (approx. 30 metres) and already exists, while the generic towers have thirty 3.0 metre levels (approx. 90 metres).

The cost of purchasing an abandoned mall and inserting lightweight apartments into the existing megastructure has to be weighed against the (financial and environmental) cost of acquiring and clearing a similar area of land and building anew. Purchasing an abandoned mall, demolishing it and then building a conventional residential development makes no environmental sense and is unlikely to make any financial sense either.

• • • 

What is a Megastructure?

The short-lived Japanese architectural movement of Metabolism is, Rem Koolhaas noted, notable for being the only architectural movement that didn’t originate in a Western country. I can’t say this is wrong, but I don’t feel it’s all that true either. Metabolism wasn’t exactly global and who’s to say local architectural movements are happening everywhere all the time? It’s just that we don’t hear about them and, if we do, it’s because they’ve been brought to our attention as a kind of uncritical de-regionalism/homogenization of everything. Metabolism was unique to 1960s Japan. We just made it part of our history, effectively neutering it until Archigram came along and did the same for us in metal.


City in The Air Arata Isozaki


Fun Palace Cedric Price


Plug-in City Archigram

Other than its provenance, Metabolism was defined by two more things. One was megastructure and the other was the notion of growth and changeability. Isozaki’s huge steel spaceframe at Expo ’70 was more Metabolist because of its author than its megaframe. More Fun Palace than Metabolist, it was big structure but not Metabolist megastructure akin to tree trunks serving useful parts such as accommodation with services and a structure at the same time. As ever, the joy of megastructures was the impossibility of them ever being possible. Trees have no problem delivering nutrients to growth areas via a stabilizing structure but mammals do and buildings do. Putting everything inside an exoskeleton works for crustaceans but wouldn’t do for Metabolists as it couldn’t show the potential for growth and change. There’s also the problem of it looking too much like those conventional space-enclosing shells known as walls. The charm of Metabolism may have stemmed from the conceptual incongruity of its two defining characteristics but it also made it impossible to build anything other than representations of them. It was very 1960s in that respect. Expo ’70 is said to have been Metabolism’s swansong. That’d be about right. Ten years seems to be the best-by of any of those representations we call styles.

Although the idea of having services pass through mega structure was a stupid one, the other idea of a structure that allows for units of building volume to be added or replaced wasn’t much better despite buildings often needing to have units of building volume added or replaced. The problem here is one of redundancy. How much structure is going to be built to support and service an arbitrary amount of additional building volume? We can design in some structural redundancy for the additional dead load but we’ll also have to add a bit more if that load is going to be a live (changing) load. It’s a problem. Another conceptual niggle with the growth analogy is that with trees, the structure becomes more massive as the tree grows. This doesn’t happen with buildings (or with molluscs). Sure, Metabolist buildings could be (theoretically) extended with additional structure supporting additional modules of building volume but the building is now less like a tree and more like bamboo that propagates underground. This inconvenient absurdity is probably why in 1962 Izozaki produced another City in the Air proposal with incremental growth for both accommodation and the structure to support and sustain it. However, even if the building is extended horizontally, the vertical cores still need to be designed for an unspecified amount of additional load. City in The Air V2.0 is only slightly more realistic.

By 1969 megastructures were huge. They’ve since gone out of fashion but Paolo Soleri’s megastructure cities outlined in his book Arcology: The City in the Image of Man were truly impossible, visionary stuff. This next image helpfully includes The Empire State Building for comparison. What all this concrete did other than put a city up in the air was never clear. In retrospect, we might see it as a a “touch the ground lightly” move but this would be tempered by our knowledge of the planetary impact of so much concrete for so little purpose.

This has always been the contradiction with megastructures. They are first of all, structures designed for an arbitrary and unspecified amount of building volume to be added. Or are they? Maybe a megastructure is more about attitude than potential, and its primary function to support itself first and foremost? What we do know is that any structure that doesn’t work to capacity is a waste of materials and, perhaps sensing this contradiction, Soleri saw infrastructure such as dams as megastructure as having a satisfying amount of concrete and then proceeded to adorn it with botanical centers, greenhouses and other visionary stuff.

They were examples of hanging program onto an infrastructure for which a justification is assumed. It’s still naïve compared to BIG’s megastructure which monetizes airspace in those mega-infrastructures known as bridges. It assumes surplus structural capacity and that the details can be sorted out later.

If ostensibly practical megastructures such as dams and bridges have a habit of staying just as visionary as the purpose-unbuilt visionary ones, then we’re going to have think about what it is about megastructure proposals that makes them so appealing? Consider this next photograph. Is it a megastructure? It is big and it exists but it doesn’t have the visionary romance we associate with megastructures. The program it was designed for never eventuated but there’s nevertheless the potential for growth and change? I don’t think it matters. It’s all concrete not doing much and China has at least 300 massive structures such as this one which is the New South China Mall in Dongguan. What we see in this photograph is about 20% of it

These are big structures with potential but no purpose. Thinking just in terms of the amount of concrete that went into the building of structures like these, it’s an architect’s duty to repurpose them and extract some utility from all this concrete that’s already been manufactured and poured. The difficulty stems from the fact that these are specialized structures optimized for one purpose only.

In that way they’re a bit like any other highly specialized structure such as aircraft carriers that are excellent at providing a place for military aircraft to take off or land at sea, but not very useful for anything else. There’s not much you can do with a decommissioned aircraft carrier.

So yes, this post is just me arranging the the furniture here for a demonstration project to convert a shopping mall into housing. This will of course be proof-of-concept and it will involve various assumptions that will hopefully be realistic. I don’t have data or dimensions for New South China Mall so I will be using the mall I mentioned in the Mallville post as my demonstration megastructure. The mall is approximately 350 metres long and approximately 70 meters wide on average. China’s second aircraft carrier The Shandong is 305 meters long and 75 meters wide. (For reference, the USS George H.W. Bush is 332 m long x 42.8 m wide.)

Far from being abandoned, my demonstration mall is completed and fully-functioning neighborhood mall. From memory, about 50% of its retail space is food and beverage, approximately 25% is children’s after-school activities and the remainder is everything else. There are two basement parking levels, five levels above ground, and a rooftop garden.

The mall is unusual in having ten entry points at ground level, external access to all levels on the south-west side (1) and two levels of external gallery access and shops on the east side. (2) One corner has direct access to B1. (3) There are three IN ramps and two OUT ramps. All this permeability would improve liveability were it to actually be converted into housing, but I don’t expect it to make any difference to actually fitting the housing into it.

I used store directory boards and fire escape plans to create a working model.

This image of the B2 car park gave column positions. There was much variation for both the positions and the thickness of the columns, particularly around the atriums. Most of this was undecipherable using the positions and sizes of columns in the car park levels but occasional more detailed information was used for cross checkin . These next two images of B1 were vital for scaling. The basic column grid is 8.4 meters x 8.4 meters.

Car park beam depths ranged from 50 cm to 70cm but longer beams on large columns span the upper level spaces between atriums to produce column free circulation corridors on all upper levels. For the working model, I will assume that floor slabs and supporting beams are accommodated within a 1-metre thick “slab” and, for the time being, will assume an additional headroom of 70cm in places where there are no beams. A model of an assumed structure isn’t any more accurate despite being modeled in detail.

The mall levels around the atriums will have cantilevered beams as well as hidden transfer beams to deal with local events. Some rationalization and simplification would be necessary and, in order to continue, a regularized column grid of 8.4 m x 8.4 m was used but keeping the one change of direction. Floor levels of the actual mall varied between approx. four metres high for B2 to six metres high for B1 (because of sewage and water supply) and generally five meters high for all the other levels. I used a uniform FF height of 5.5 meters.

Despite the change of use, I will keep the elevator cores and fire escape stairs. I’m imagining a low-energy building but won’t make a decision on whether or not to keep the escalators functioning. Facades will be removed and the atriums naturally ventilated. Services and utilities will need a rethink.

• • •