Renforcer la densité urbaine et les espaces verts tout à la fois : le cas du Danemark

Today, the talk will cover a case from Denmark about density.

To begin with, I would just start with a simple statistic: by 2050, we know there will be an additional 2.2 billion people living in cities. And to accommodate such high population growth, can you imagine how much land we are taking every single day to construct new buildings? It is the equivalent of 20’000 American football fields every single day.

1 The impacts of the way cities are built

How these cities are built will have significant implications for the environment.

Figure 1

For energy consumption, we can see on this figure that there are different kinds of cities around the world. And there is a big contrast between European cities and North American cities. There, you can see European cities have much slightly higher density, but much lower transportation related to energy consumption. This made sense for me when I was living in Denmark. Every day, I was walking, biking to the grocery store, it was so easy. But now in Seattle, without a car it is impossible to reach a grocery store from where I live.

Figure 2

The way land use and density are planned definitely have a significant impact on greenhouse gas emission. How cities are built also has an impact on biodiversity. This study showed us that in New York City, when the density of build area is higher, the genetic biodiversity is lower. This means that if we think of our cities also as urban ecosystems, the way they are built also impacts different kinds of species and biodiversity.

Well, it also has an implication on human well-being.

Figure 3

If we think about the thermal environment, here is a study which provides three different scenarios: we can either infill the city, grow it up or expand it outwards. In one of these scenarios, for example, the most severe urban heat island effect was stimulated in Berlin by replacing green spaces with buildings in all infill spaces.

2 Cities, densities and mental health

It is not surprising to us that the way cities are built and urban density have implications for cardiovascular diseases and respiratory disease. It is less intuitive for mental health. Although it is a very common human experience. We know that one in three people, in any stage of life, will be affected by any kind of mental disorder. But at the same time, we also know that depressive disorder is 39% more prevalent in urban areas than rural areas.

There have been two very simple theories behind why urban areas have higher risk of depression. One indicates that if the city is denser, you will experience greater risks because of overcrowding, because of the lack of green space. While another hypothesis says that if we have a larger population, there is lesser risks because there are more opportunities for social interaction. These two hypotheses seem to be contradictory.

Figure 4

But if we look at an aerial photo to encompass a city, the density is not just limited to two dimensions. In a downtown area like this one, it has more high-rise buildings and a high density. While this other area has similar building density, but also much lower buildings. In the third case, if we move further outskirts, it is still residential. It is just like much more green space and trees on the street. To provide the same amount of population growth, we can have two different kinds of built environments. One is high-density low-rise. The other one is low-density high-rise. And in the past few years, my colleagues and I have been working on this simple question, about whether higher building density and higher building height will have a negative impact on mental health.

Figure 5

With two dimensions, horizontal and vertical, this seems to be very simple. But if we translate this into our real landscape, then you can see other kinds of neighborhoods that we are familiar with. It is not just a rural area and urban area. There are also other kinds of design in different places.

Figure 6

So if we are going to answer this question — the one of how urban form is associated with the risk of depression — then we actually need to first, further know what is the built environment and what is the natural environment. And we need methods to measure them, because environments are always changing. They are actions being taken. We need quantification that can help us observe the environment across time. Let me show you these methods.

3 Remote sensing and machine learning models: powerful ways to explore patterns in cities development

Remote sensing gives us access to satellite images over the past three decades that can help us understand how land cover and a built environment have been changing.

Figure 7

Through my research, I develop machine learning models. If we think about it in a simple way, it is basically taking images and taking trending data that can be any attribute that you are interested in. It can be green space, building height or building density. And by using different models, we’re able to predict urban form over time. This model was developed in Denmark, but we also apply it to other European cities, including Paris, and show decent results.

Figure 8

So now, if we have this kind of information, we can link it to population health data to further understand how different urban form is associated with depression.

For mental health, for each single individual who was born in Denmark after 1955, we follow up their medical records. For the adults above 25 years old, there were in total 75’000 Danish people who were diagnosed with depressive disorder. And at their first episode of depression, we went back five years to look at what kind of environment they were living in (a 250 meter buffer) to check the type of urban form which was around their homes.

Figure 9

Using statistical methods, we can further validate what kind of urban forms are associated with depression.

In our model, our baseline is rural area, low density, low rise. And the other five types of urban forms at the left hand side of the image are different building density and different building heights. If the idea that urban areas, in general, always have higher risks than rural areas, is true, then all the dots should be at the right hand side because the baseline one is the comparison with rural areas. This was true for some cases. Globallyn, median density, low rise neighboorhoods hold the highest amount of risk among all kinds of urban form.

Figure 10

There were also exceptions, for example, the low density, high rise form, which holds lower risks compared to rural areas. If we further adjust for social economic factors, like comparing people of the same income level and same education level, we find out that actually downtown, the high density, high rise area does not present a significant difference with rural areas. Why does that happen?

Figure 11

We tried to map urban form and risk geographically. This map shows us that for the lowest risk category, the low density, high rise form, usually caracterizes neighborhoods that are in front of large green spaces, canals, or the ocean. While the other category, the median density, low rise, holds a higher risk. It is the area where we can see the dark outline at outskirts, an area where we can find mostly single family housing and which are more car driven.

4 Developing health promotion neighborhoods taking advantages of remote sensing and machine learning technologies

Figure 12

To come back to the question that we asked in the beginning — how does urban form correlate with depression? —, we find out that urban areas might not just be the caprice. It might not just be a single problem. Actually, in urban areas where there are green spaces, spaces to gather, get a beer, and hang out with friends, people are happier. So this figure at the right hand side shows that in Denmark, the areas where population grows the most are also areas where vegetation increases the most.

Urban, rural, or natural, artificial, these binary thoughts might not be contradictory. It can co-exist if we conceive them well. We also talked a lot, during this symposium, about affordable housing and accessibility to this kind of health promoting neighborhoods. It is not affordable for everyone. And sometimes when the central area gets upgraded, it also moves people out to other places if they cannot afford a house because of the price rise.

I think this tool, like using remote sensing and machine learning to measure the built environment, provides us a kind of method where we can built up evidence about what matters for local areas. We can further use them as a tool to convince policy makers.

Figure 13

To conclude, I just want to briefly show you a few slides about the potential of using remote sensing for sustainable urban development. We are talking a lot about the urban environment, about different metrics. But eventually, they are affecting people’s behavior and how we are exposed to different environmental characteristics like air pollutant, thermal comforts: those will affect our public health outcomes.

Figure 14

We are living in an era where, in the past decade, a lot of satellite data, like I am using right now, were not publicly available. But most of the datasets now are available. And they help us observe the world in both global nodes and global cells. This data today, the one we use for measuring building density and height, is optical data, the one that creates satellite images, that you can access through Google map. That kind of data is called optical data.

Figure 15

But there are other kinds of sensors and signals which can tell us other information, such as air pollution or temperature. And those observations are wall to wall, like across space and over time, which can help us to measure the impact of specific policies. This kind of question can and should be further investigated.

Figure 16

I just want to provide three key areas that I think are very promising for students and for urban planners to use, because we are in an era where open science is the emerging trend. Now when we publish a paper, we also need to publish our method and code. So when you download a paper, there is a section called Code Availability, and you can access the method and use it for yourself. Google Earth Engine can help you collect data, analyze data, in just a few seconds, because the cloud computation is now free. And the third thing is that if you want to run machine learning or deep learning models, it can be very time consuming, like running days and days. But, you can use GPU, the supercomputing resource, which is also a publicly available resource.

Figure 1: Figure 2: Figure 3: Figure 4: Figure 5: Figure 6: Figure 7: Figure 8: Figure 9: Figure 10: Figure 11: Figure 12: Figure 13: Figure 14: Figure 15: Figure 16: