Picture a spillway gate that doesn’t just release water from an overflowing river, but manipulates sediments to create new streams, islands and wetlands. And imagine that the gate does this autonomously, guided by ecological data and shifting needs — essentially allowing nature to “evolve.” Computational landscape architect Bradley Cantrell is figuring how to do this by applying environmental sensing, machine learning, predictive modeling and robotics to environmental engineering.
The TED Blog asked Cantrell to talk to us about his ideas, how they would work, and how computational landscaping may change the relationship between human beings, machines, and nature.
What is our current relationship to the natural environment, and how do you envision changing it?
Right now, human beings are really good at saying, “We want this river to move very quickly, and we want it to always be predictable.” So we can engineer a predictable river. Take the Los Angeles River, which is a simple example. It’s basically a concrete channel. We’ve taken all the unpredictability out of it because it used to jump its banks and flood a large part of the Los Angeles River basin. We said, “We want it to be within this 20-meter-wide zone and to never move, and we want it to always run at the same velocity so it never backs up and floods anything.”
But that’s not the way an ecosystem or river works. It actually has a whole range of behaviors. We currently don’t allow these systems to have a range of behaviors. I would like to change this so that our infrastructures allow the creation of evolving and changing ecosystems.
Where does the idea of computational landscape architecture fit in?
Computational landscape architecture is the idea that, using computing and machine learning, we can build physical infrastructures and natural landscapes that relate symbiotically with our cities and natural systems.
In theory, what we’re doing is embedding the complexity that exists in natural ecological systems into our own manmade environments. We do this by feeding computers data from natural historical records. So, for example, you might have a set of records about how a particular ecosystem performed, such as the behavior of a river’s water levels and velocity. Then you might have a series of predictive models, about how sea-level rise due to climate change will affect this local ecosystem, for instance. These predictive models are used to develop a computational logic which allows them to make autonomous decisions about how it uses infrastructure — like spillway gates — to prevent possible problems.
This means that computers end up having a life of their own, within our design goals. Machine learning can be compared to how we make decisions: we make choices for the future based on data from experiences we’ve had in the past.
In your talk, you offered the example of the Mississippi River, for which you’ve prototyped a computational infrastructure. Walk us through the process of how it would work.
The example I most often give is a system of spillway gates that, instead of simply allowing river water to flood a lake when it gets too high, precisely controls the flow of water to create landscapes that benefit biodiversity or protect cities from storms — and does so in an automated way.
The Mississippi River has always jumped its banks. If you look at the shape of Louisiana, its shape is the result of the sediments in the floodwater building out land. Left to its own devices, once the river finds its longest route, it jumps its bank and tries to find a shorter route. Rivers naturally do this kind of cycling.
In the last 100 years or so, we’ve built levees all the way down the river. If you look on a map, you’ll see the Mississippi River now has this really long route, and it’s just continued to build and build and now its dumping dirt off the continental shelf into deep ocean water. There’s actually a shorter route for the Mississippi river: it naturally wants to jump its banks and go down what’s called the Atchafalaya Basin. The US Army Corps of Engineers built a structure where it wants to jump, forcing it to go the long way.
Why? Is it because people are there?
No. The Atchafalaya Basin could easily be flooded, with few consequences. But New Orleans sits further downstream, and if you change the route of the Mississippi River, suddenly New Orleans becomes completely irrelevant in terms of a city. It would be sitting out in this exposed area with no river next to it. So for the sake of commerce, we still want ships to come through. There’s all kinds of mega-engineering going on to keep that river the way we want it to be.
The problem was, in the past the levee would just break down in certain areas and start to flood out into the bayous. This happens whenever the river is very high. It breaks free in certain areas and it just floods, and all of this dirt carried down from Iowa and St Louis dumps out into the area, basically replenishing the land there. People have said, “Well, we should just begin to build massive gates on the river, and whenever the river gets to be too high, we’ll relieve the pressure by flooding these areas.” So the safety aspect is already in place. There are two spillways: one of them floods the whole Atchafalaya Basin, and the other one floods Lake Pontchartrain, north of New Orleans.
Those projects were both built in the ’50s and ’60s, right after they forced the river into its current configuration. But these solutions haven’t been about pushing water into land we want to build. They’re pushing it into places that we then have to go back and dredge out so that the Ponchartrain can still be a lake, and the Atchafalaya Basin can still be a river.
Above: Watch Bradley Cantrell’s spillway gate “print” a landscape by controlling the flow of sediment-laden water.
With your solution, what would happen?
I’m adding a layer to this. Let’s say we go ahead and open up these spillways in a range of new locations that are already being proposed. What if each of those spillways had a whole range of things it could do, rather than simply flooding or not flooding? And how can we speed up or slow down the velocity of the water coming through? The answer is by opening these gates in different sequences. Think of the way you put your finger over a water hose. When we slow the flow down, sediments fall out of it, and by speeding it up, it carries sediments further, or breaks obstacles down and pushes beyond them.
So just using those two mechanisms, we plan to push the water and the dirt to go where we want it to go. If we have control over the land-formation process alone, we can start making choices about whether ecosystems should evolve in a certain way, and we can help nudge things in that direction. Once the system is fully functioning, it would form landscapes on its own, but it will have had our curatorial help.
To read the full interview, visit the TED Blog >>>