Fellow Spotlight: Christina Last

Our Fellow Spotlight series focuses on showcasing the work of our Fellows. This time, I spoke with Christina Last, who recently joined Subak’s latest global Fellowship cohort. Christina’s project will develop a machine learning model that predicts the concentration of short-lived climate pollutants (PM10), which can be used to identify and build evidence to reduce pollution and emissions.

Christina Last is a postgraduate student at MIT, where she specialises in building machine learning solutions to understand our built environment. Christina has collaborated with a number of city governments and international organisations on software and data science, in both developed and rapidly developing urban environments in the US, the UK, and Vietnam. She has led various international research projects, most recently as a Senior Data Scientist collaborating with UNICEF to model air quality during COVID-19 lockdowns using machine learning.

Introducing Christina 

Maisie: Hi Christina, thanks for joining me! To start, do you want to introduce yourself and your Fellowship project?

Christina: Thanks, Maisie. I'm Christina Last and I am a Subak Fellow this year. I also now have a bunch of other credentials to my name, thanks to applying to a master's in the US, concurrently working alongside the Fellowships I’m doing, including being a Fulbright Fellow and a British Schools and University Foundation Fellow! It's really exciting, because I get to publish my work through loads of outlets and spread awareness of my call to action. 

So, the work I'm doing with the Subak Fellowship is primarily oriented around my goal of improving the quality of life for future generations. My standpoint is that broadly the global West has been pretty successful in terms of improving the quality of life for future generations. That's been facilitated through economic growth, real wage growth, and families have been able to sustain themselves without the whole family needing to chip in. That has freed up the time for children to invest in education, which has had long run productivity gains for those economies, which has broadly secured a safe future for younger generations. But there's been a trade off to that. The economic and social benefits that have been experienced have come alongside an environmental sanction that's been placed on other countries that haven't been able to advance in the same kind of structure or at the same pace. So, those negative environmental externalities are what I've looked at, broadly outside of developed economies, and the unequal growth that has produced profound climatic effects, as well as health effects, on those nations.

Linking this in with the Subak Fellowship, large amounts of CO2 emissions and powerful greenhouse gases driving climate change have often been produced alongside harmful particulate matter. This is damaging for the environment, but also for health. When we see young children dying prematurely, about one out of every four of those deaths is related to environmental risks, a lot of which is from acute respiratory illnesses. My Subak Fellowship tries to tie in both the climatic and health effects of this challenge. I'm looking at modelling a particulate matter particle called PM10, which is an atmospheric particle small enough to get into children's lungs, but also very damaging to the climate in terms of its warming effects. So, I'm building a global model using machine learning to model PM10, and model exposure in certain regions of the world where we see high child population density, and finally model it globally so we can get an estimate of how much we're seeing this particle, and what the sources are as well.

Maisie: You refer to yourself as “an interdisciplinary data scientist”, working at the “intersection of quantitative social science, technology and the built environment”. How did you get into this space, what was the journey so far?

Christina: So, I alluded to the fact that I'm going to use advanced statistical and computational techniques for the Fellowship, but my background is not programming by education - I definitely came into that set of skills by teaching myself and jumping at opportunities to learn. My educational background is actually geography. Therefore, I've always thought about new innovations or technology through the lens of its impact on the environment and society.

The second reason is that, as a geographer, I never really knew where to place myself in a profession, so I've worked in multiple industries. I started out working in engineering firms, specifically looking at helping engineers, but also urban planners and urban designers that master plan anything from a neighbourhood scale to large cross-state, regional infrastructure projects. Working at that scale, you get a broader idea of why building cities and improving cities is so important for innovation, for economic growth, and for environmental sustainability.

Finally, I became interested in machine learning. I'll define that as using advanced statistical methods to learn complicated relationships between features, which are areas of your data that help you predict a certain outcome. And not because that's just interesting by itself, but I know that these questions are really hard to answer with just one or two variables. Questions around how climate change is impacting certain communities are not straightforward to measure. Machine learning gives you a really interesting set of tools to be able to understand the complicated interplay of relationships between different predictors for some of these outcomes.

Addressing the interconnected climate & community health crises with data

Maisie: Could you tell me about a project that you've led in the past that was your favourite, and why?

Christina: I recently partnered with an organisation called the Environmental Reporting Collective. They funded me and a foreign investigative journalist to report on some environmental damage happening in the Iraqi Kurdistan region. We wanted to take a look at gas flaring, which is understanding excess gas that is released into the atmosphere, usually during the summer months when it's not needed. This pollutant has a twofold impact on both the climate and community health. There's no ground truth data for this, or for flaring incidents across the world. We used machine learning to identify these flares from open satellite imagery. To do this, you use satellite imagery and look at the pixel values at nighttime. When you can see very high red and orange pixel values, isolated from the other darker pixel values, then you can kind of assume it’s a fire. Also, by putting in other data sets, you can identify whether there is a temperature reading that correlates to the same location. Then, you could bring in some open source geospatial data like OpenStreetMap, to understand if there are tags close to that specific pixel that you've identified, indicating a nearby oil or gas refinery.

Now, I mentioned that it's both a climatic and a problem for community health, because gas flaring is releasing methane into the atmosphere, which is a highly potent greenhouse gas, but it also releases a bunch of volatile organic compounds with it. So we were looking simultaneously at how much methane was being produced from the flaring, but also how much particulate matter was being released in the vicinity. That was all done remotely from the UK, and I built a pipeline to identify the flare locations, and to calculate the average methane and particulate matter within a buffer region around the flares. 

Then what we did was we looked at flaring hotspots. Is there anywhere where these are occurring? And are they actually near communities? We identified a few areas and actually went to Northern Iraq to go and speak to local stakeholders in the community: doctors, health professionals, local governors, Syrian refugees and Yazidi communities. We wanted to validate our assumptions - are these people actually being impacted? In turn, we could elevate their voices. It was a really great experience working with some incredible on the ground freelance journalists (Stella Martany, Alannah Travers and Tom Brown). 

We wanted to make sure that the voices came out in the report. Consequently, the report won't be focused around machine learning. Rather, what really motivates people to solve these challenges is hearing from Syrian refugees, that their water is polluted with oil and that seven children have died of cancer on their block in the last month. It's so harrowing - this really needs to be the centre stage, not necessarily the methods. 

Maisie: Working on such harrowing issues and being at the intersection of different areas, could you tell me about one area of your work that you find particularly challenging? 

Christina: I think they all have their own challenges. The one that I feel most apt to address is availability of open data, which directly speaks to Subak’s call to action. I really think that there needs to be more initiatives that are data first, and more specifically, open data. Open data is really hard to find on some of these topics. I gave the Iraq example - there's no ground truth set of flares published by oil companies, for clear reasons. But this makes it really challenging to understand the impacts climatically and for health, and so we have to revert to other kinds of proxies, or building models to help us identify approximate probabilities of flaring incidents. 

So, this challenge is where a lot of my work stemmed from, and I'm working as a Software Sustainability Institute Fellow this year as well, who are primarily focused around open source code and good practices in developing open source code. This challenge is something that I feel really passionate about, especially when it's addressing social good.

I’d like to caveat that by saying that there are loads of challenges with my work - but availability of open data is one I think that I can influence the most. There are probably harder challenges, like getting senior leadership on board. That's just something that I don't have an outlet to do at this point in my career.

Cities as a driver of the climate crisis - but also the solution 

Maisie: Cities are quite prominent in the work that you do broadly, and in your Fellowship project as well. Can you tell me a bit more about why cities matter generally, and also in the context of the climate crisis? 

Christina: I think cities are the reason climate change is happening. I'll unpack that a little bit - I fundamentally believe cities are the reason why modern society came to be, the reason why we have the ability to build spaceships and machine learning models and advanced infrastructure or systems that mean that we can just flush a toilet. We innovate because we live close to one another. We see the best technology coming out of the world, not from rural communities, but from places like the Bay Area, or, if we want to even go back further in time, London in the early 19th century was the innovative hub of the world. If anyone wants to challenge me on that, then they'll have to write in the comments!

But cities obviously have the negative externality side of the coin as well. If we innovate, via cities, then cities are ultimately the cause of the environmental harm that has been an externality of significant economic growth. But, the way that we live in cities, and build innovations within cities is also the best way to target some of those negative externalities. The caveat is that, obviously, the externalities aren't also limited to the cities. So there needs to be policy mechanisms through which you can attribute externalities, with some sense of causality, to certain cities, to certain communities, to certain nations. We have a good way of doing that so far, but it's not perfect. It's also really hard because supply chains now are inherently global. The ownership of some of those is much more fuzzy and distributed. So, even though cities are the root cause of all this change, the agents of play are much more diverse. 

Being a rung on the ladder of collaboration 

Maisie: You're saying that innovation essentially comes from people being together and working together. In your work with different communities and different stakeholders, do you see collaboration as an important tool in tackling the climate crisis?

Christina: Collaboration is important to innovate in any area, and we're going to need to innovate to tackle climate change. I don't believe in the degrowth theory around the only way is backwards to address climate change - that's definitely not been the case with other technologies that we've built. I don't see climate change as any different, although obviously, the timelines are a little bit more scary. It's also important because we both know that climate change is such an interdisciplinary field itself. It's not just climate scientists, we have to bring on a lot of people into the conversation, because it's not just about how the temperature is warming the earth, it’s about where is the distribution of the effects happening the worst, and who are the people who are experiencing that?

The second reason why collaboration is important is data - we have so much data being generated around assessing metrics of climate change. So, bringing in data from diverse sources is really important, and we're not going to know where those sources are, or how to get access to them. Unfortunately, people are still the biggest blocker in that field, unless we connect with the communities who know where that data lives. So, I'm happy to be a rung on that ladder as well.

Maisie: I love that metaphor of being a rung on the ladder. So, I'm aware that you've only quite recently joined the Subak Fellowship, but could you tell us why you decided to apply, and also the help that you have had so far from Subak?

Christina: Subak seems well situated for my life circumstances as well as the interesting research questions that I want to ask. The Subak Fellowship in my conception is that you can apply and then launch into the Accelerator programme with that. Although I’m not quite there, I’d love to get an understanding of what it could look like through the Fellowship and also talk to people in the community who are starting out new climate ventures and not-for-profits. I want to plug that recently me and a co-founder who are working on air pollution monitoring - so as well as on the Subak side - I’m working on this pollutant that has incredibly damaging both climatic and health effects. We’re also modelling some more health oriented particulate matter distribution globally. We just got accepted onto the Unicef Innovate Venture Fund so we’re really excited about that work too and happy to build a lot of that work into the Subak Fellowship too, and I can’t wait to apply for the Subak accelerator later on with this work too. I’m excited about the journey that Subak offers, basically!

In terms of the help I’m getting from the Fellowship, it’s been really exciting - I’ve been connected up with my coach, Laurence, who has been so helpful in putting me in touch with the correct people around the Fellowship, catching up on progress, discussing ideas, and also suggesting open data sources and resources. It’s exactly what I need - this person who connects all the dots and helps you get to the right information. He’s been fantastic. 

Maisie: Looking forward to the continuation of the Fellowship, do you anticipate any challenges in the work itself? 

Christina: Yes, there’s always challenges. What I really struggle with is being able to translate the data that you generate into an impact - or at least, get individuals or communities who are invested in the issue to know that this is now a resource you can use to advocate for your cause. Being forced to think in that way is not as natural for me, even though it’s ultimately what I care about, not building machine learning pipelines for the sake of it - I care about it being translated into something that is useful for a community that is at risk of the harmful pollutants that I’m focused on.

The Fellowship is just starting, so I’m not quite at that stage yet, but I anticipate this challenge. For example, in some of my other work I’m working with UNICEF Belize who aim to integrate some of the data we are generating in schools, to both help pass an exam and write reports of the pollution they get exposed to in their areas, and they take home indoor air pollution sensors. Leading with how we’re going to use this to raise awareness, rather than leading from “well this is a really cool modelling technique”, is much more exciting and impactful - but coming from a computer-loving technical gen-z background this is harder for me! 

Maisie: Finally, how can our readers stay up to date and learn more about your work? 

Christina: You can stay up to date by following me on Twitter. I also have a website. You can read more about Christina’s work on flaring here.