Brief update: I’m pleased to announce that I’ve received an Emergent Ventures grant to continue my work on Construction Physics! This will help some work on some new projects that I’ll be able to say more about in the coming months.
In the US, our buildings tend to last somewhere in the neighborhood of 50 to 100 years, depending on the type of building (residential buildings tend to last longer than commercial buildings). This means every year, something like 1 to 2% of our buildings get torn down or demolished.
For example, New York City has around 1 million buildings , and gets around 1500-3000 permits per year for building demolition. (an astute reader points out I’m off by roughly a factor of 10 here. For other sources of the 1-2% number see “How Buildings Learn” or this post on home lifespan.)
In a world of sustained economic growth, there’s an economic calculus for this sort of replacement cycle, since the present value of any extra building life would be less than the cost to purchase it. But this cycle of replacement is relatively modern - medieval houses would often last for centuries, and there are examples from around the world of buildings that have lasted for many hundreds or even thousands of years while remaining in use - The Pantheon, Aula Palatina, Brihadeeswarar Temple, Verona Area, Chartres Cathedral are a few examples. The most central examples of long-lived buildings are large, culturally important buildings that required huge investments to create and maintain, but there are plenty of examples of more modest buildings with long lifespans - England has nearly 30,000 homes built prior to 1700. Clearly we can construct buildings to last much longer than they currently do.
So it’s an interesting thought experiment - what would a modern single family home designed to maximize service life look like? What would we have to do to get an expected lifetime of, say, 1000 years? How would it differ from the houses we’re used to seeing? Let’s take a look.
Achieving a long service life
One option to extend lifespan would be to just copy the design of a house that has already survived for centuries - in effect, we could build them like they used to. This seems to be what projects like Hope for Architecture are doing, by building with traditional load-bearing brick masonry. But there’s a few drawbacks to this approach.
For one, most older buildings didn’t survive - England’s 30,000 surviving pre- 1700 houses are perhaps 3 to 5% of the homes that existed in England in 1700 . Ironically, Hope for Architecture illustrates this, as building the project required tearing down an existing house.
And even traditional design is unlikely to be sufficient to get you to 1000 years - Historic England lists just 225 houses built prior to 1400 AD, just 26 built prior to 1200 AD, and none built prior to 1000 AD. So we’ll need a different approach.
One way to think of a building’s lifespan is that it’s a function of the things that can cause a building to be torn down. The more potential sources of failure there are, and the likelier they are to occur, the shorter a building's life will be. So to extend our houses’ life out to 1000 years, we need to remove as many potential sources of failure as we can.
We can, roughly, divide potential sources of failure into two categories :
Discrete destructive events that will destroy our house. Things like natural disasters (fires, floods, earthquakes) fall into this category.
Decay processes that gradually make the house less and less appealing over time. These can be both physical processes (a roof gradually wearing out) as well as “cultural processes” (a house’s design slowly going out of style).
We’ll also need to consider how the likelihood of these types of failure might change over time. Sources of failure are a function of the surrounding environment, and we can be sure that the surrounding environment will change over the course of 1000 years. The areas at the greatest fire risk today, for instance, might not be the areas of greatest risk in 500 years.
Natural disasters and discrete destructive events
The most obvious way a building can be destroyed is by some particular destructive event - a fire, a tornado, an earthquake, etc. If we want our house to survive for 1000 years, it will need to be capable of withstanding various natural disasters and extreme weather events.
There are roughly four categories of events to consider:
Extreme water damage (floods, tsunamis)
Extreme wind damage (hurricanes, tornadoes)
Earth movement (earthquakes, but also things landslides).
Of course, over 1000 years our house might encounter any number of destructive events - a tree could fall on it, a bomb could be dropped on it, a nearby gas main could explode, etc. But designing our house for the most common natural disasters should mostly cover us for the “long tail” of other types of disaster as well - there’s only so many ways that to inflict damage on a house, and natural disasters cover most of them.
We have two options for addressing these sorts of events - designing our house to be strong enough to survive them, or building it somewhere where they’re unlikely to occur. For extremely long-term survival, our best bet is to pursue both strategies simultaneously.
Historically fire has been one of the most reliable ways a building gets destroyed, and many existing old buildings survived a fire at some point. The most straightforward method of fire prevention is to build our house out of non-combustible materials, at least for the structural frame and the exterior elements - metal, concrete, masonry, stone, or compressed earth instead of wood or plastic. This is common sense, it’s the approach building codes use to reduce fire risk, and it’s empirically what seems to work - browsing through England’s pre-1700 houses reveals that the majority of them are built of stone or otherwise non-combustible construction .
We’ll also want to use large, heavy structural elements wherever possible, since even noncombustible materials may fail in a fire if they are thin sections. And we’ll want to use design details that reduce the risk of the house catching fire, such as those provided by various ignition resistant construction guides.
We can also dive deeper, and look at the specific sources of fire damage. In our era of strict building codes and responsive fire departments, most fire damage comes from two sources - wildfires, and fires started inside the home.
For wildfires, other than using ignition-resistant construction details and materials, our best bet is to simply build our house somewhere where wildfires are unlikely to occur. This means building in a more urban area, away from the wildland-urban interface.
Fires started inside the home are harder to predict and avoid. The two most common causes of house fires are cooking and portable heaters, which don’t seem especially avoidable at the level of house design. We could perhaps build our house somewhere where portable space heaters are less necessary, but this strikes me as a sort of brittle optimization.
Designing for fire resistance is important, since it’s hard to predict the areas likely to see a fire far into the future - over 1000 years, the sources of fire damage could potentially change significantly. Today, highly urban areas see lower fire damage per capita than rural areas, possibly due to better fire department funding and response times. But in a world without fire departments (or enforced building codes), clustering near other buildings probably increases your risk of fire damage (since a fire started anywhere is more likely to become a fire that spreads everywhere). Locating in a large urban area also potentially puts your house at greater risk if war were to break out (not something we tend to think about when designing our buildings, but very likely to occur sometime over the next 1000 years).
The same is true for wildfire risk. The wildland-urban-interface, the area at greatest risk for wildfire, is a function of where other people decide to build (or not build). And over 1000 years we could see historically wet areas become arid and fire prone, or deforested areas grow back, or a change in forest management practices that make wildfires more or less common.
Compared to fire damage, designing for wind damage is a little bit simpler. The potential severity of the fire will be a function of how the building is designed, but the severity of a wind event is completely independent of any design choices we make. And unlike a fire, where it’s difficult to avoid some damage if one occurs, it’s possible to avoid wind damage by designing for sufficiently high wind speeds.
Rather than try to predict the upper bounds of storm wind speed in a given area (taking into account that our 1000 year design life will demand a longer mean recurrence interval than would be typically used), we can simply take an upper bound on what wind speeds are even possible. The highest ever recorded surface wind speed was just over 250 miles per hour, which coincidentally is the design wind speed recommended by FEMA’s P-361 guide for hurricane and tornado safe rooms. So the most straightforward way to proceed is to just design for a 250 mph wind event, and make our entire house a safe room.
(For even stricter requirements, we can go with the US Department of Energy design criteria for power plants - power plant design isn’t based on service life per se, but on minimizing the probability of failure, which for our purposes is the same thing. This criteria raises the design wind speed to 300 mph, and adds some more stringent missile impact requirements.)
Of course, this is easier said than done - not only are the forces from a 250mph wind event extremely high (there’s a reason very few buildings are built to these requirements), but designing a building to survive it likely makes it harder to modify in the future. A roof designed to survive 250mph is a roof that’s hard to replace in the future, and you’re likely to be replacing the roof much more often than you’ll be encountering a tornado. It may be prudent to consider less strict design requirements, such as the Miami-Dade high velocity hurricane requirements, for some portions of the house.
Flooding and water damage
Water damage is another fairly reliable way to destroy a house - nearly all modern building systems are susceptible to corrosion caused by extended exposure to water. And flooding is one of the most likely natural disasters to occur.
Our best bet here is avoiding it, by building our house somewhere unlikely to see a flood. For this we can use tools like FM Global’s Global Flood Map. But this will only get us so far - this (and most other maps) max out at a 0.2% annual chance of recurral, or a 500-year flood. But this would give us just a 13.5% chance of avoiding a flood over the next 1000 years. And flood risks in a given area are likely to change significantly in the future.
A better option would be to use the flood estimation tools developed for choosing nuclear power plant sites, which is based on the maximum flood that could occur in a given area, and takes into account things like floods caused by dam failure.
A common sense approach should help here - we should build our house somewhere away from the coast or other major body of water, and on a site where water is unlikely to accumulate, such as on top of a hill or other high point. We could also use flood-resistant design details that prevent water intrusion from the sides or below (making our house into something like a bathtub), or using some type of elevated construction.
Seismic and earth damage
Earthquakes and seismic events are another difficult category of damage to deal with - they have the potential to be so destructive that most building code strategy is based around preventing loss of life, rather than building damage.
We can screen off most of this risk by building outside of high seismic zones (FM Global once again provides a helpful map). But it’s hard to completely remove the possibility of an earthquake, so for the remaining risk, we’ll want to build to standards for buildings designed to survive earthquakes and keep functioning. In the US, this means designing to Risk Category IV, the same standard that things like hospitals, fire stations, and power plants must be built to.
Like with wind damage, designing for earthquake resistance is something off a tradeoff. Seismic forces go up the heavier your building is, so unlike for wind or fire damage we’d ideally want as light a building as possible in an earthquake. And the systems you’d likely choose for seismic resistance aren’t necessarily the ones you’d choose for long lifespan. Traditional, unreinforced masonry, for instance, can survive for centuries, but is extremely susceptible to collapse during an earthquake. Conversely, heavily reinforced concrete can perform well in an earthquake, but is so susceptible to decay that it’s a poor choice for building optimized for design life.
Stringent earthquake design also involves ensuring load path continuity and system ductility by tying all your building elements together with seismically rated connections. This adds safety, but it also makes it harder to change your building in the future. So we’ll need to consider the relative risks and costs.
Other earth-based damage, such as landslides, are harder to mitigate against. Our best bet here is avoidance - landslide hazard maps, such as California’s, are useful in this regard.
(We’ll continue next week with Part II, where we discuss dealing with decay-based failures!)
 England’s population in 1700 was approximately 5.2 million. Conservatively assuming an average household size of 8 gets us 650,000 homes total.
 Strictly speaking, these are probably more accurately modeled as part of a single function, perhaps something like a power law distribution, where a house has a small chance of encountering a very high amount of damage and a larger chance of encountering a small amount of damage. There are probably more fires that cause a small amount of damage than there are that completely destroy the house, for instance. But from an intervention standpoint it's useful to treat these as two separate categories.
 Many of the wood ones likely burned in the Great Fire of London
Thanks Brian, and congratulations on your grant. Since we're in an area of architecture that's sort of my specialty I'm tempted to tie up a lot of server memory for my comments. For the sake of brevity, I'll limit myself to just a few comments:
-Stewart Brand's comment "Whomever builds with wood, builds a shack" actually doesn't correspond with the longevity of residential structures. Pure masonry structures, i.e. pyramids/Parthenon/Pantheon don't represent the majority of homes that have exceeded the 2 century mark. Wood is good if you keep it pretty dry and manage fire risks suitably. The verdict is still out on the American stick-framed traditions. My anecdotal observation is that the majority of wood framed homes will still be in use a century from now---but see my final comment.
-Location factors are probably more critical than architectural features when it comes to absolute longevity, but it's worth looking at the trajectory of early American home styles for some reference points about decisions that contributed to success. The Josiah Quincy house in Quincy, Massachusetts serves as a good example of "terminal design" with respect to good site selection, robust construction, sensible ceiling heights, and functional room sizes. The "Colonial Revival" succeeded as the dominant American house style because it employed these principles during the post-war building boom.
-The "learning process" of buildings refers to the fact that the maintenance and renovation costs have to be considered on a timespan that exceeds the average tenure of the homeowner. Critical structures i.e. foundations & framing stay relatively unchanged for centuries, but roofs, siding, mechanical systems, and interior finishes get replaced on 20-40 cycles. I would speculate that over a hundred year period, several generations of homeowners will put double or triple the initial construction cost of the house into the "sustenance" of the property.
My question: is it a good idea to build a house that lasts for 1000 years? I think the ideal building would last 1000 years if there is the will to maintain it, but would also be easy to disassemble and leave behind very little impact once it is no longer needed. I'm really interested in what buildings do or don't get maintained over time and why. I wonder how much of it has to do with the building itself (ease of maintenance, flexibility, aesthetics in context with changing tastes) and how much of it has to do with larger forces (economic forces, the context in which it is built, is the program/use for which it is designed still relevant). I would guess that a single family suburban tract house is not going to last 100 years both due to the quality of construction but also because the pattern of development in which it sits will not be sustainable or desirable forever. So it's not just a question of the building itself but the larger context.