Digital Twins for Climate Action: A Singapore Case Study

In this episode, host Stephen Goldsmith speaks with Dr. Heiko Aydt, head of the Digital Twin Lab at the Singapore-ETH Centre. They discuss how digital twins are transforming urban climate research and policy, with insights drawn from the groundbreaking Cooling Singapore project. Dr. Aydt explains the development of the Digital Urban Climate Twin, strategies for testing policy implications through modeling, and navigating cross-sector collaboration. 

Listen here, or wherever you get your podcasts. The following is a transcript of their conversation.

Stephen Goldsmith:

Welcome back. This is Stephen Goldsmith, professor of urban policy at the Bloomberg Center for Cities at Harvard University with another one of our podcasts dealing with technology cities and states. And today we have a very accomplished guest, Dr. Heiko Aydt, who has, you have so many titles that will take our whole podcast to read all your titles, so I'll summarize a few of them.

You're with the ETH Zurich Singapore Center, you're head of the Digital Twin Lab there, an issue of great interest to our listeners. You work on digital urban climate issues including Cooling Singapore, and you're a co-investigator of Circular Future Cities at FCL Global. Quite a few titles. We're honored to have you. Thanks for being with us today to discuss your work.

Heiko Aydt:

Thank you, Steve for having me, and thank you for the kind introduction. It's a pleasure to be here today.

Stephen Goldsmith:

We have an audience that cares about digital ways to improve urban areas, national areas. I know you're an expert in climate, but our interest is the intersection between how to accomplish those goals more creatively through the use of digital tools, and you're one of the foremost experts in the world on this subject. Give our listeners just a quick introduction to your background before we get started, would you please?

Heiko Aydt:

Sure. So I'm a computational scientist by training and by that I mean I'm basically using computational models to do science. I also have a background in software engineering of distributed systems, so basically a computer science background.

For the last decade or so, I worked on building models of urban systems. For example, in the past I was part of a group that developed a high-performance agent-based traffic simulation, and the purpose for that was to study the impact of electric vehicles in cities.

Since 2017, I've been working on the Cooling Singapore project at the Singapore ETH Center, and the focus of this project is the urban climate and specifically the study of the urban heat island effect. In 2020, we started building the Digital Urban Climate Twin for Singapore, and that led then later to the foundation of the SEC Digital Twin Lab in 2023.

Essentially, the Digital Twin Lab is a spinoff of Cooling Singapore and the lessons that we've learned there is that digital twins can be used to help policy makers to make better-informed decisions. And unlike the Cooling Singapore project, which focuses primarily on the urban climate, the Digital Twin Lab really aims to build technology and applications for digital twins and leveraging the experience for Cooling Singapore, but not just on the urban climate, but for other urban topics as well.

Stephen Goldsmith:

Thanks for that. That was quite helpful. Let's go into Cooling Singapore projects a little bit more. It started back in 2017. Tell us a little bit about how you're using digital twinning to address heat islands in Singapore, ways that may be applicable to other large cities and states around the world.

Heiko Aydt:

Yeah, so the Cooling Singapore project started out kind of three questions. First of all, we wanted to better understand the urban heat island effect and its impact on the city. The second was to understand what can be done about mitigating urban heat. And the third one was really to think about tools that enable planners to make better informed decisions.

So the project started with the question, "Where are the hotspots?" And our approach was to use a model-based approach. So we basically had a computational model that simulates the city with all that's in it, the traffic, the buildings, the vegetation, power plants, industry. And if you run this model, we get an idea of where the hotspots are. Then the question of course then naturally is, "Well, what can we do to reduce the urban heat island effect and what can we do to reduce these hotspots?"

So the new talks were really to compile a catalog of about 80 or so mitigation measures. These included measures like adding more shading, improve natural ventilation, add green facades, reduce reliance on air conditioning in order to reduce the waste heat that's being released in the city and many more. Now with this catalog in place, then the next question became, "Which of these 80 or something measures should be applied? And also where? Also, what combination of measures might be giving us the best results? What is the expected impact of implementing one measure or another and what works better?"

So we realized that we cannot answer all these questions by manually setting up a model every time we have a new scenario. That would just simply not scale. At the beginning of the Cooling Singapore project, we basically had a very manual process in place. We had a number of experts, we had a domain expert for traffic, a domain expert for urban design, and they would all come together and basically build these models and then we run a scenario. The problem is every time we have a new scenario, then people have to come together and create a new scenario, tweak the models so that it would represent this new scenario and that doesn't scale very well.

So that is when the idea of the Digital Urban Climate Twin was born. The Digital Urban Climate Twin was basically consisting of a number of components. The first one is the models themselves. So we again have different domain experts. They build these models, but they also build them in such a way that we can parametrize them for different scenarios. Then the next step was building applications around these models, graphical applications that could be used by the intended end users. So the ultimate goal was that planners and decision makers could basically use these tools to create their own scenarios and then run these models without having to resort to us as the researchers to do it for them. And the idea here was that this would scale.

Stephen Goldsmith:

A couple of more questions on this. First of all, is the catalog of data elements, is that a publicly available catalog? Could our listeners, our cities get access to that?

Heiko Aydt:

Yes, it's a technical report and it's been published on the ETH Research Collection website. I can make the link available after this interview.

Stephen Goldsmith:

That's great. We'll post it on the Data Smart City Solutions site at Harvard, and I'll explain why we're so interested in that in just a second. One more question about your last answer. What's the relationship between the project and the government? Are you an academic center providing services to the government? What's the relationship?

Heiko Aydt:

That's an interesting question. So the Singapore ETH Center is what I sometimes call a research outpost of the ETH Zurich in Singapore, and the focus here is really to do research. We don't do teaching, so we are affiliated with the university, we are affiliated with ETH Zurich, but there's not much teaching going on here in Singapore. The focus is really purely research.

The research projects themselves are typically funded by the Singapore's National Research Foundation, and that's also the case for the Cooling Singapore project. We had multiple funding grants for the Cooling Singapore project. It has now officially ended in 2024, and it started in 2017, as I said, and the funding came from the National Research Foundation.

Now, the interesting thing about the Cooling Singapore project is that from the very beginning, we worked closely with government agencies, and that's perhaps something that set us apart from other projects because a lot of the research is often done in, let's say in an academic way where the focus is more on publishing academic papers.

In our case, there was from the very beginning, a close interaction with the government agencies, various planning agencies, and that was actually quite interesting because they gave us a lot of feedback and a lot of input, and I guess it kept us on the ground. It kept us grounded in reality and not basically end up in an academic ivory tower.

There was some academic research going on and my colleagues did publish papers as well. My group in particular, actually, we focused more on the development of the digital twin, so that was more the software development process and translating the research into something that can be executed, I guess. So it was a bit of a mix between academic research and applied research.

Stephen Goldsmith:

How does mapping play a role in the visualizations and in the analysis, in the construction of the digital twin itself?

Heiko Aydt:

Yes, in this case, for this kind of digital twin, it's an urban digital twin. So it's all about the city it's all about maps, mostly heat maps, really. So our simulations produce a lot of data, particularly the urban climate models. They produce gigabytes of data. It's typically things like air temperature, relative humidity, wind speed, wind direction, and a lot of other things.

In order for someone to understand what is going on, the best way to work with this is to visualize it. We talk about heat maps in order to map where the urban heat island hot spots are, where the wind corridors are. All these are layers essentially that you can put into a map-based graphical application, and that's essentially what we did with our explorer application. So the explorer is a browser based application. It uses maps and we have multiple layers that we can visualize with the data that comes from the simulations. Very importantly, I guess, is also that we can compare. So if you run two scenarios, you'll be able to compare them and show the difference.

Stephen Goldsmith:

This is quite fascinating. Now we have a project at Harvard called the Community Data Health Initiative. It's an initiative sponsored by a US foundation called Robert Wood Johnson Foundation. We're working with different US cities on the issues that you're working on, but I think US cities are not as technically sophisticated in the digital twinning part of their work. So just kind of distill, if you would, from what you've just been talking about and what you've done, which aspects might relate most to the work of US cities or other cities around the world that don't have quite your level of expertise.

Heiko Aydt:

I mean, you don't need the digital twin in order to know what you can do in order to mitigate the urban heat island. The mitigation measures that we've listed in our catalog could also be applied elsewhere, and I think urban planners around the world, they do have an idea of what they need to do in order to address things like urban heat problems. So a lot of the stuff that we've done in Cooling Singapore can certainly also be applied in other cities around the world. There are some unique differences though, because Singapore is in the tropics, so it's hot year-round, there's high humidity. So certain measures that we focus on here may not be so important in other places, and vice versa.

Stephen Goldsmith:

So if we step back, you're doing the modeling. Just before we let you go, talk to us a little bit about where the data sources come from. Are they mostly sensor data? Do you have traffic data? Just help us kind of see through your words what you're doing.

Heiko Aydt:

So we use a variety of data. For example, we use urban geometries, building information basically. Where are the buildings, the building footprints, height information, to get a rough idea of the geometries of the buildings. We use also climatic data that comes from measurement stations that are placed in Singapore. There's a number of locations where these weather stations are, and we take the data primarily to validate the model.

There's other data, for instance, from the Urban Redevelopment Authority, the master plan, which is also publicly accessible. Generally speaking, we have been using publicly available data to build this Digital Urban Climate Twin, but we've also always focused on building the ability to import data. And the reason here is that it's easier for us to work with public available data rather than using confidential data. Especially for the researchers this is important because they want to publish their work, but we always emphasize that we show that the Digital Urban Climate Twin works, and then if you have more accurate data or a richer data set, there's features that we provide that you can upload your own data and basically then presumably also get better output, better results from the simulations.

Stephen Goldsmith:

Is there a Singaporean regulatory agency that regularly uses your data for permitting or planning or licensing?

Heiko Aydt:

So the project ended in August 2024, so that's fairly recently. And we have now entered a phase, a 12-month period in which we are transferring parts of this Digital Urban Climate Twin selection of models into the agency's environment, into the government environment. They have certain requirements as to software that they can run in their systems and their models and where they run. There are certain restrictions, they're more restricted than what we can do in the research sector. So over the next 12 months, we are transferring the models and the software and basically installing it in their environment so they can use it and run their own scenarios with their own data.

Stephen Goldsmith:

If you're giving advice either to other researchers or, for our purposes, to city or national officials in the US about how to reach this level of technical sophistication, if the goal is to create healthier cities by lowering heat, where would they start? How would you recommend that they begin in order to make the most impact?

Heiko Aydt:

I think you begin by getting everyone around the table, and I think this is something in Cooling Singapore that we somehow managed to get right somehow. It was messy. We reached out to a number of government agencies. There were also of researchers involved. People speak different languages. I sometimes use the example of a model. When I talk to an architect and at the Future Cities Lab, there's lots of architects. Some of my colleagues are architects. When I talk about models, they have a different idea of what a model is. For me, it's a computational thing for them, it's like a physical model that they put somewhere. So even language is often a barrier, and I think a good beginning is get everyone around the same table and establish a common ground in terms of our understanding what the problem is and what can be done.

I think especially the connection between the researchers and the practitioners, I think that was very useful. The practitioners give us a lot of input, very, very valuable input that researchers often don't think about. And likewise. So we, as researchers, have often sort of a bit of a freer mindset. So we think about solutions that perhaps the practitioners haven't thought about yet, and sort of like a back and forth kind of discussions, and you have to repeat the message several times, but at some point I think this is really valuable and people get the point, and then at that point I think becomes a very fruitful collaboration.

Stephen Goldsmith:

Your work is fascinating, and I know we could ask you hours of questions, which we promised that we wouldn't. But as we leave, what advice do you have to US officials, city officials in particular on utilizing digital twins and data modeling in order to inform decisions? I mean, we've been talking to the cities with which we work. Where do you get the best impact in children's health with respect to green corridors or trees or the design of roads? And so if you thought about all of the elements that go into your digital twins that are now being transferred to the Singaporean government, where can we get the most, the US phrase is bang on the buck? Where could we direct them that would make the most difference?

Heiko Aydt:

On the technical side, I think work on standards for data formats, for instance. Think about the processes. Researchers working in one domain, and then the application is in a very different domain. So I think it's worthwhile to think about the process. How if a researcher builds this model that is really interesting and can do a lot of things, how do you get this into the hands of the practitioners? That needs to be a process in place. And ideally, you do this with the minimum of red tape, right? It doesn't help if someone has a model and it takes five years to get that into the hands of someone who can potentially do something useful with it. So standards can help procedures to reduce, I guess, the administrative burden.

Get expectations clear from the beginning. Researchers often have an idea of what this thing might be in the future, but practitioners don't necessarily see that. So work with mock-ups to get them a better idea to visualize what this might look like, what they can expect.

The other thing is, especially when it comes to model-based work, "All models are wrong. Some are useful." There's a quote from a guy called George Box, and that in many ways is right. So you have to be careful when you use these computational models and scrutinize the results that you get. So you still need the domain experts that sort of interpret and help with the understanding of what we see there in the results. So I think it's really that kind of collaboration between the researchers and the practitioners.

Stephen Goldsmith:

So let me just close with one last question. There are a lot of different stakeholders have different levels of sophistication. We have communities that obviously care, we have researchers that are talented like yourselves, we have city officials. So how do you think about the language and communication of this data across various levels of sophistication in the stakeholder groups?

Heiko Aydt:

Yes, language is obviously very important. As researchers, we often tend to speak a very technical language. The language is often depends on what domain you are in. If you're a climate expert, you speak the language of climate experts. As an architect, you speak a different language. Basically, we publish and we speak the language of our peers. But when we try to communicate science, we need to focus on the target audience. If you speak to stakeholders, government representatives, planners, they speak a different language than people in your neighborhood.

We did run citizen engagement sessions in the earlier parts of Cooling Singapore to also bring this topic of urban heat and heat stress closer to people in the communities and help them to understand also what kind of potential options there are, what they can do in order to mitigate things on their individual level. And then again, you talk to government agencies and there you have to, of course, use a different language. It takes a bit of getting used to it as researchers, but I think it's absolutely crucial to get everyone really understand and on the same page.

Stephen Goldsmith:

This has been just a fascinating conversation. I'm quite excited about the opportunities to transfer of your learnings to the cities and federal officials, which we work in the US. Stephen Goldsmith, professor of urban policy at the Bloomberg Center at Harvard University. With great appreciation for our guest today, Dr. Heiko Aydt of the ETH Zurich Singapore Center who runs the digital lab there. Heiko, your research is fascinating. I think it holds so much promise to help people around the world. Thank you so much for your time today.

Heiko Aydt:

Thank you very much, Steve. Thank you for having me.