(Unedited) Podcast Transcript 591: The Urban Pulse

This week on the Talking Headways podcast we’re joined by Yale’s Karen Seto and University of Connecticut’s Zhe Zhu to talk about their study: The Urban Pulse. We talk about using frequent time series satellite data to measure change in cities and the processes of urbanization.

The Urban Pulse – Yale School for the Environment

The Urban Pulse: Diagnosing the urbanization process as spiky, cyclical, and asynchronous – PNAS

Listen to this episode first at Streetsblog USA

Find all of our episodes in the archives including this one.

Below is a full unedited but AI generated transcript of the episode:

I’m trained as geographer as well. I do a lot of remote sensing, time series analysis.

Um, this is back in 1992, before Google Earth, before iPhones. I was in graduate school, and in an earlier life, I wanted to be a diplomat. And I took a course in geography. I saw a satellite image, and it literally transformed my life, and I became a geographer.

I was born in Hong Kong, and it was an image of Hong Kong. And the thing that I remember most is being able to see visually from space the difference between Hong Kong and China in terms of the land use, the shapes, the configuration of the built environment, agriculture. You could see so many different things that were, you know, just how the two systems were different.

The second thing get me attracted probably is because of Karen. Uh, I basically has followed Karen’s work a lot, and when I see some of her papers showed up, like the urban teleconnections research papers, I get excited. And I come back to, um, Karen invited me to visit her sometime when I was at Texas Tech University to write an urban review paper.

And the more I wrote about it, the more I get excited about urban. Yeah.

I graduated with my PhD in the year 2000. And Zhe, when did you finish your PhD? Uh,

was our PhD advisor. He’s, uh, recently retired from geography at Boston University.

What made you all interested in understanding how cities change?

And so this is back in the ’90s when if you had, like, eight or 10 images, that was considered a time series. That was like, wow, we have a lot of images, if you had one image per year. But I was never really satisfied with the idea that you could capture what’s happening in a city and the changes that are happening with just one image in a single year.

So that’s a little bit of background. As part of Zhe’s PhD, he developed an incredibly novel method to take the entire archive of Landsat satellite images, like every image that’s ever existed, and to identify change. And Zhe’s PhD work was he developed that methodology to study changes in forests. When I read the paper, I thought, “Aha.”

It was, like, literally a light bulb moment. How can we adapt this to understand urban change at scale, like, for lots of different places over a very, very short time period? So when I say short time period, it’s, like, daily or… Right. So that’s how we started talking about working together. We had collaborated on this review paper, but that wasn’t our own intellectual work in terms of, like, developing algorithms.

So it was really through a process of discussion where we said, “Well, this is a big unknown. Like, how are, how are cities changing, and how would we actually characterize change?” So that’s how it all got started, and that was probably, like, four years ago, Zhe? I’m trying to remember.

Uh, my favorite data sets were usually the USDA aerial photography that was able to overlay to make maps for transit planning. But it’s fascinating to think about how much data there is now. You talk about the satellite, you know, is interesting if you had one a year, but now you can get them daily. Like, when did that change that you could actually access this amount of information, an overwhelming amount of information, from these satellites from the data that was available?

In the ’80s, $2,000. No,

It’s crazy how much that’s advanced. It’s interesting that I was reading the notes from the paper and some of the articles that were written about it. You all often framed it as medical, looking at cities as patients, seeing the changes as pulses, as electrocardiograms. What’s the importance of talking about a more clinical thinking, uh, when it pertains to urbanization?

There’s so much satellite data and, and spatial data. There’s actually so much more data around about us, too. We’re wearing wearables, you know, we’re tracking our sleep and, you know, CGMs. I mean, there’s so many different types of data that give us different metrics about our health. There’s no one indicator.

And it, it became clear that the health analogy is so apropos to urbanization because cities are constantly changing. Like every single minute they’re changing and they’re changing at various scales. You know this, my neighbor is putting on a second floor, right? Um, we’re digging up our backyard to put in trees.

I mean, cities are not static places and the idea that we use a single metric, GDP, population to characterize cities is just fundamentally so simplistic. So Zhe and I said, “How can we take all these new data to rethink how we can characterize the process of change?” ‘Cause that’s what urbanization is, right?

It’s this process of change. And as we stated in the paper, urbanization is not one dimension of change. It’s actually multiple dimensions, right? There’s the population part that’s changing. There’s the built environment, there’s the environment, there’s the economy and that there’s no one single metric.

And so we came to this idea of the urban pulse because we thought it is very much like the human body. There’s not one metric to describe your health or your condition. Um, we have to use multiple dimensions.

So, you know, I just gave the example of my neighbor putting on a second floor, me digging up my backyard. Those are different pulses of activity that we can observe from space. So that’s probably the biggest, like, aha, is that we know inherently, I mean, you’re an urban planner, like, you know that there are different neighborhoods, different blocks have different vibes, different feelings, right?

Different pulses, right? Different levels of activity. We can actually measure that. And so that’s probably one of the biggest ahas. I think the other that, um, was a, kind of a big finding that we, again, we knew inherently, but to see it in the data. I couple my remote sensing work with field work, so I’ve done field work in many, many countries, and it’s clear that urbanization doesn’t happen in this linear fashion.

It happens in fits and spurts, right? It’s like investment or, um, the government focuses on this special economic zone, and there’s, like, a lot of investment and development there, but then this area is a little bit sleepy. We can see that. Like, urbanization is not linear, and it does happen in fits and spurts all throughout the entire city.

So that, I think, was another really big aha, in that often we compare cities to each other- And we actually don’t look at what’s happening within a city, which I would say is, like, another big insight or aha for us is we could see in all the six cities that there’s so much variation within a city. So that was really exciting.

You can clearly see there’s major differences from how the policy going to change. Another thing I want to add is that what we measure, this pulse, is a process of this urbanization. A lot of the time, if you’re looking at the, those remote sensing of urban is the outcome, okay? There’s increase in impervious surface, there’s urban expansion, there’s the intensification.

But what we see is actually is there’s some activities going on for the cell of pixels. So it’s more like the flow, the energy, the, the agent that are changing, transforming this piece of land. It’s like the blood in the pulse, right? It- we’re, we’re not measure we’re gaining weight, we’re, we’re losing weight.

We’re, we’re seeing the heart rate measured by this pulse, and that’s this very similar thing. W- and that’s why we, we link with– we, we’re using the construction amounts and intensity to measure the pulse, urban pulse, instead of the, like, how much impervious surface, how much concrete is there. So that’s one of the thing I, I think quite interesting to mention.

About, like, every three days or less than three days, you’re going to get a one observation based on how cloudy your locations are. And the data is trans- is remove this atmosphere influences and remove this bi-directional impact when you view this from different location, different direction that normalize to, to the native view.

So they’re very consistent if there’s no change. When there’s change, we can see it based on this time series

And then where this change happened multiple times, we can know, okay, that’s basically a real change instead of noise. And then the- that’s the first step, we, we call the land surface change monitoring or disturbance monitoring. The second part is the, where the deep learning, machine learning came in. We need to classify these breaks, the changes we ob- observe, to making sure, okay, this is real, a construction, instead of some kind of, like, deforestation or some kind of other kind of non-construction related change.

Uh, the reason why we’re using deep learning is because the human patterns are usually very unique, so we are combining not only the time series, but also the spatial pattern all together to classify, okay, this is really a construction change instead of some other kind of, for example, a fire or some other change happened on the land surface.

Zhe’s method that he’s describing as this continuous analysis is it’s taking the ostensibly raw data with minimal correction to see the signal, and then once there’s a change in the signal, you can say, “Oh, what is that class, given that there’s this big jump?” So the methodology takes change as an inherent part of the signal, right?

So there’s, uh, he mentioned phenology, right? If there’s a lot of… If there’s a certain pattern, you see that as normal for this particular area.

Like, what are some of the things that popped out to you?

Like, COVID happened, and there were shutdowns, and this one neighborhood, everything stayed the same. So we, we did some research on that neighborhood, and certainly it corroborated what the data showed. It was surprising.

We see that in a lot of places.

I’m curious what the pattern looks like in a time series that’s only three days. Is it just, like, little places popping, popping here and there and there? Or, like, can you even look at it from that video perspective? Is it only available to look at it the way that you are doing it through the data? I like that graph that you had.

It looks like you should actually sell that as a painting or, like, something to hang on people’s walls.

And so the pulse analogy here works really well, ’cause you’re getting a pulse of every pixel. But it would be difficult to visualize this for every single place or an entire city. You’d have to think about ways to either aggregate some of the data, which then in some ways goes against the idea of the pulse, right?

Now you’re gonna aggregate the data across, you know, multiple pixels. Yeah, so I, I think the visualization is a bit of a challenge. I think it’s easier to do this for neighborhoods. Like, if you were to say, you know, “How does this neighborhood in San Francisco vary from this neighborhood?” You could get the pulses of those pixels that fall within the different neighborhoods.

That’s essentially what we did for those six cities. It’s like looking at an EKG, right? It’s like looking at an EKG.

And you could argue that, well, if there’s no building permit, everything else is the same, but there are probably places that are changing without a building permit. So with the satellite data, we have wall-to-wall coverage for a city. And so, yes, we can capture the place that has the building permit, as well as all the places that don’t have building permits.

I think that’s one. Um, I think the other is going back to this concept of a, a pulse. It’s… You know, a building permit is- A household, an individual going in applying for approval to make change. But there are lot, probably lots of other changes that are already happening that don’t require a building permit.

And so this is giving you a sense of the changes, different types of changes, different types of, you know, magnitude changes across the entire city. It’s just a lot of information. I think it’s incr- would be incredibly useful for planners at the city level, and I think as well as perhaps like regional or state level even to understand.

You can compare between cities, but also compare within a city, like policy interventions. You know, we talked about TOD earlier, right? So does it, this policy intervention, does it affect these neighborhoods the same? Does one neighborhood respond differently to economic incentives than another neighborhood?

The pulse would be able to tell you that.

If you’re looking at a neighborhood that you did have an intervention with, that you did have an investment policy about, and then you could tell whether that, you know, changed versus another one, how do you get like a policymaker to say, “This was working,” or, “This isn’t working,” based on the data that you all are putting out?

And then, then what might be the answer, uh, or what might be the response based on, you know, information they’re gleaning from the pulse?

Like being able to visualize that, to characterize that, I think we don’t have any way to do it currently. So that is already a big advance in my mind for a policymaker. Like, okay, you put in a road or, you know, whatever the policy is, how does it change the pulse across the city? I think that’s probably one of the biggest insights.

Then your second question about, well, what do they do with that information, that’s above my pay grade- … about, you know, what they would like to do, because arguably there are some decision-makers that want differentiated responses to the same policy. That is part of what they would like to do. It’s not about equitable or even distribution of investments or even outcomes.

That’s about a few thousand cities, and it’s so hard to picking. They’re similar. There’s some differences, but they’re similar. And then I think, uh, Karen told me that, uh, global s- city from a variety, different kind of places, and the city, some are big, some are small, some of them are, uh, in different kind of developed stages can be a better.

So we basically visualize the, the different kind of cities, global cities for based on their, uh, different kind of urban pulses. We- I think we computed the principal component for those urban pulse to see how different they are. And later on we came up with those cities because they are different.

They’re relatively different in the urban pulse, and they are also a stratification of different kind of economic and social status.

And at one point I asked Zhe, like, I forgot which city I asked you to do, I said, “You know, could you c- calculate the pulse for this other city?” And it was like, “Whoa, this is so different from what we’re seeing.” Yeah. And that led to us looking at a number of other global cities. And we chose them because they are so different economically, the layouts are so different, and the way in which they’re being built, right?

You’ve got Dubai with much more top-down, large scale investment. Lagos, which is very, very bottom up, and very different type of investment. And then Shenzhen, which is master planning at the national level. So that’s … And, and the other thing is we chose cities that we knew. We, we chose cities that we knew something about.

So one of our co-authors, uh, lived in Seattle for many years. I did my PhD work on Shenzhen, right? So we wanted to choose places that we had some understanding of.

Just even just going around a couple of places, I felt it. I felt the energy of it just being built, and it was fascinating to see that.

We wanted to be able to capture that with these data.

Uh, I’m curious what you could do if you had this data for every place in the world and were able to analyze it.

I mean, in all seriousness, we think that this could be an incredibly useful policy tool at the local level for a city, for a, a country. I think that we’re also just at the, like, at the such the early stages of understanding what this can actually tell us. I can’t remember if we said this in the paper, but, you know, the human pulse has, has been known for a long time, but we’re still discovering what we can learn using the human pulse.

Like, for example, we know that- Babies have a much faster pulse, and we know that our pulse sleeping is different than when we’re exercising, right? And in recent years, people have started using the pulse as an early diagnostic for all sorts of things. And so the next step would be can we… I- we can, more like when we um, get funding to develop the pulse for lots of places, we can really start understanding how it can be used as a diagnostic.

When we analyzed the data, it was very clear that it could be used as an early warning. I mean, we had a lot of discussion about that, Zhe and I. Like, “Oh, this, like the human pulse, this could tell us when a city or a neighborhood in a city is undergoing stress, is not responding to the economic incentives or policies, and this neighborhood is doing really well.”

Going back to what we were saying earlier, not every single policymaker is a benign policymaker, right? Not every policymaker is like, “Oh, we want every place to thrive.” If you’re a policymaker that wants every place to thrive, you could use the pulse to try to achieve that. If you’re a policymaker that actually doesn’t want places to thrive equally, you could probably use the pulse in this capacity as well.

For example, if you want to purchase a house, uh, in a certain area, you want to purchase house in a more livelier neighborhood instead of place has been quiet for very long. This is gonna to be a global data set for you to see every piece of land, how livelihood h- how people are transforming the concrete, the, the land on a almost daily basis.

It can be updated on a sub-weekly basis. I

Or places that are declining.

So- You could see is this as vibrant as it was 10 years ago, and is it as vibrant as other parts of the city?

But I’m thinking of, like, a place like Austin, where I went to school. Um, obviously it’s grown and, and fallen based on its construction and the booms and the peaks and the valleys and all that stuff. Um, but there’s also been a, a cultural difference. Uh, Boise, Idaho, um, some of these places, Portland, Oregon, Seattle, um, San Francisco even, like there’s just a cultural undercurrent to the physical change that actually happens.

We’re not comparing my heartbeat to your heartbeat, right? We’re comparing my heartbeat to my regular heartbeat. So we can look at the pulse or the, the urban EKG to compare how different neighborhoods are doing, and we s- again, surmise that we can infer these cultural changes. The other thing we also talked about and could develop is a nighttime pulse, right?

There’s a daytime pulse, and then there’s also a nighttime pulse. So some of these cultural changes, like you’re saying in, in Portland and Austin come with more activity at night as well.

Like, we could see that, like these EK– I’m gonna use the EKG analogy, like EKG varies by neighborhood and by city. Now, here’s this global shock that’s affecting every place. Like, oh, we could see that some cities bounce back right away, and others never came back to the same pulse. Some neighborhoods never returned to that pulse before.

So it was a natural experiment for us.

And we want this to be usable and useful to policymakers and decision-makers, and o- other audiences like, you know, investors. Um, cities have energy because people invest in them. So, you know, our ultimate goal is to develop this for thousands of cities worldwide and to make it freely available to people who are interested in urbanization.

This paper is the start, and- Need decades to really figure, figure it out.

And as we were writing this and, and analyzing the data, we thought, “Oh, what we really need is an epidemiologist. We need someone who is a, a transportation engineer,” right? We need people with different lenses to understand urbanization so that they can help us interpret the pulse. So we have to make this for thousands of cities, and then we need to get it in the hands of other researchers and practitioners and investors.

Thinking about, like, a Shenzhen where they’re building tens of subway lines and what that means for the city, or Lagos where the development’s much different and the transition to a, a major city. They’re a huge city now, but I’m just thinking of, like, the way that their transportation system isn’t as developed as some other places.

Those might be interesting ways of thinking about how nations, how regions, how cities create value for their citizens, uh, over the long term. So it might be an interesting way to look at it.