Sustainability in relation to Statistical Science

Stephanie J Dickinson: Hello, Professor Jim Griffin.
This is the first interview that I've done, which I hope will be part of a series called Sustainability in Relation to Statistical Science. My name is Stephanie Dickinson. I am the Statistical Science Green Champion, and that is for the Department of Statistical Science at UCL. I am interviewing Professor Jim Griffin about his work in relation to sustainability. Let's get started.

How is what you do Environmental?

Jim E Griffin: Yeah. So I wanted to talk today about my work on the environmental DNA. In environmental DNA, people collect samples from the environment. Things like water, soil, or even air. Then in the lab they can do an analysis, using PCR to to try and find which kind of trace elements of DNA for particular species. You can either do that in terms of a single species, you can say you kind of have some signature DNA signature of a particular species. Or you can look at multiple species, you can just get lots and lots of samples and kind of try and test again. There's lots of different possible species. And so that's kind of an emerging technology that people have been looking at last 10, 15 years.

And and so the idea is that it's much cheaper than other kinds of ways of monitoring species. So traditionally, people would just go and count birds, or observe particular types of species, or maybe they would go to the same place again and again, and try and see whether they see the same species is there or not. But those things are quite expensive. This involves employing someone. So being able to to use these kind of technologies is is increasingly attractive to to researchers who don't have limitless budgets, and so that is interesting from a kind of ecological point of view.

But also statistically, it throws up some kind of statistical challenges because the data involves errors. So you can have false negatives. So you don't observe particular. Species in your sample, because maybe they're not actually present there when you went and collected your sample. Or maybe they're there and in small amounts. And they don't actually show up when people would do the analysis. And that's quite complicated in some ways. And so that leads to challenges. 

Also one thing we're interested in is as well as understanding whether the species is there, trying to understand how much of the species is there. For various reasons, to do with the way that the data gets analysed, it's quite difficult to make that link. But but there are some possibilities. So yeah, that's the kind of environmental aspect of my work.

Stephanie J Dickinson: Thank you very much. Very, very complex, but a lot of that makes sense in many different ways, especially the efficiency in cost. I think it shows where perhaps we're heading in types of research, so that's that's really interesting.

Jim E Griffin: There's other types of measurements that people take. They can take kind of acoustic measurements, and listen for different types of birdsong or different noises in, say the jungle, that different animals are making different calls, and then and then they try and understand them, which species are present. So people are looking at all. There's other ways to monitor things, and they're trying to look at all these different methods and try and combine the results to really get a much clearer picture of what's getting on. And you know it doesn't. Often people go and they disturb in some way the environment, right? Maybe the environment doesn't behave in the same way as normal. If you have put someone in it. So these kind of remote sensing, it's going to become increasingly important.

Stephanie J Dickinson: That's that's really positive. To understand how people are doing things differently, to make sure that they're not causing a mistake. Because they're making the environment react differently to them. That's really interesting and positive. And also for me, it seems quite inventive to think of all of these different ways of measuring where the animals are and what they are. 

Jim E Griffin: Yeah, yeah. Ecologists are very good at this. So I'm not sure that's a statistical thing. But yeah, colleges have kind of come up with all sorts of different ideas, because they understand very well how the species behave right, and and what that characteristics of that species that maybe you could try and observe in someone.

Stephanie J Dickinson: Really interesting. Right? Thank you. Let's let's go to question 2. What process will your work benefit?

Jim E Griffin: I think it's just about monitoring really. Initially, the work came out of people who want to monitor great crested newts. Great crested newts turn out to be an important species for understanding. It's the way that government and various agencies working, trying to understand the effects. And they're a protected species. They're particularly important. And when people want to develop land, they have to worry about what to do with it if there's great crested newts there. It's that they were interested. That's how it started, but also being able to look at many species. It then lends itself to ideas like biodiversity and understanding changes in biodiversity. Then you can get into much more complicated analysis in terms of understanding the communities. That's interesting whether an animal there is there or not. But also it's interested in how they interact right for ecologists. That's a very important area of research. Also just if people are going to try and intervene in some way, either by by building something or by trying to reintroduce species or trying to, what's the word to try and kind of regreen areas? Then it's important to understand how those will affect the community rather than just a specific species, because these things obviously don't work in isolation, they work this with complicated processes going off.

Really, those 2 sides, I think are the main practical outputs of this.

Stephanie J Dickinson: That's really interesting makes me think a lot about housing development and how we need to be very careful about where we're building new homes and communities of humans. Obviously, we need to pay attention to the other communities that exist there already. And so yeah, I can see how that research would help us to think a little bit more about that and hopefully make the right decisions.

Jim E Griffin: Yeah, of course. I think government is always setting up regulations around around building and the environment agency. And people are looking at how to to regulate things. Having a better understanding of what you should be measuring when you regulate, is an is an important thing. And that that area of research is very practical for for doing that.

Stephanie J Dickinson: Let's move on to question 3. Why did this need to happen?

Jim E Griffin: So it began, as I said, with some people who are interested in great crested newts. When I worked at the University of Kent, there's a lot of interest in great crested newts, because there's a lot of great crested newts in Kent. They were interested in using this new technology of environmental DNA and trying to understand how reliable it was really. And and so they became aware that there's these different types of error. You can have false negatives that you don't see the species, but also you can have false positives that sometimes either you get contamination of your sample, and you start detecting species that can't be there. But also in reality they don't really find species. What what they find is some kind of DNA signature which they then call a species. And so then there's some possibility that through doing that you don't match up these DNA sequences with the species properly. And so there's some potential for false negatives and false positives.

Somebody in my department was working with this group. And then she came along with this kind of problem, and it seems to me that it was an interesting statistical problem where there was a kind of natural Bayesian statistical approach to doing things. So in Bayesian statistics you want to use prior information, and the prior information we could use, there is that you kind of expect to see more true positives than false positives.

So if you're kind of doing your your experiment reasonably well, that's what you would hopefully expect to see right. And so we could use that information in the analysis that that was enough in a way to help us to be able to understand this data and overcome some of the statistical challenges of this data.

Then they are able to go forward and and use this model to try and understand how reliable using environmental DNA was for great crested newts, and also to understand how you should do the monitoring. So should you kind of go to the same site multiple times, or should you be going to different sites? And should you do replicates within the lab of your your study, and how many replicants you should do? They could answer questions of the study design as well. So that's what came out of it, really and then and then that led on to to working with other people who are interested in environmental DNA.

Stephanie J Dickinson: Great. Thank you. What are your reasons for doing this work?

Jim E Griffin: I think a lot of my reasons are statistical. Really, it seems to me an an interesting statistical challenge. I'm always interested in interesting statistical problems. And so, as I said, the the original work had the natural Bayesian solution, and as a Bayesian statistician that was nice to be able to to find that and then subsequent workers looked at looking at multiple species and trying to model to a certain extent the process that happens in the lab. So when they do the analysis it goes through various stages, and and that then introduces biases, I guess, really, and maybe errors into the results. That come out of each of the stages of this analysis in the lab. We wanted to adjust for those things in a model. That's kind of interesting modeling challenge. But it's also nice that it has an important practical application in the environment. And and it means that people are very interested in in what you're doing, because they're very interested in getting the results and getting reliable results. So they can make a scientifically justified conclusion.

Stephanie J Dickinson: Yeah, that seems like the the ideal interdisciplinary collaboration between, you know, people who are doing kind of good, worthy sustainability work and statisticians who really, really like the statistical side of of what you're doing.

Jim E Griffin: Yeah, I think I think it's one of the one of the the beauties of being a statistician, that you're able to work in in different areas, and work on very different applications. Often the the problems are similar. And you end up working across these different areas. And and you find unexpected similarities, or that you can bring across different techniques or methods.

Stephanie J Dickinson: Great. Thank you. What is your future wish for the results of this work?

Jim E Griffin: Well, we've developed various packages that people can use to do the analysis of their data. The hope is that these become a standard part of the the toolkit for ecologists to analyze their data. I'm not sure we got that. But people are interested. So that's good. It's a good start.
And yeah, that that will be. That would be the kind of best possible outcome.

Stephanie J Dickinson: Great. Thank you. How could you apply what you have done to other areas?

Jim E Griffin: One obvious similar area is the study of the human microbiome. So that there, rather than in in ecology. You have species, and in the microbiome you have bacteria. For example, people are interested in the microbiome in the gut, and and how bacteria interact there? So, there's differences because in the microbiome you know whether that is connected to people having particular conditions, or, you know, particular health outcomes. Maybe if you can adjust. You see this lot. People are interested in the moment that if you can adjust your microbiome, then you can somehow have better, you can become healthier or stronger, or sleep better or something, right? And so they're interested in all these physical outcomes. In some ways the analysis in the lab is is in many is similar, very similar, and and a lot of the statistical challenges are quite similar as well. That's one area. That's kind of separate, but it but is similar.

Stephanie J Dickinson: That sounds great.

Jim E Griffin: Yeah, yeah, yeah. And and the other is, the other thing that I'm involved in is anti doping. And there, you have a similar thing because you have mostly have false negatives. Because people are very worried about false positives. For obvious reasons. But there you have the same kind of problem. You have false negatives. And yeah, you have a lot of kind of noise in the way that things are happening. So you're very interested in these types of errors. So there's a similarity there as well.

Stephanie J Dickinson: Great. Let's go to question 7. I heard someone once saying that they want to solve the world's problems with math. Do you think environmental statistics could be an example of something similar?

Jim E Griffin: Well, I think, as we were talking about earlier, these ideas of remote sensing are potentially very powerful for monitoring environments and those are always going to come with challenges that have to be dealt with through mathematical modeling or statistical modeling, or AI type of approaches. Or probably in reality, some a combination of all of those different ideas and different ways of thinking and and so I think, that will become increasingly important. 

Just to try and quantify what's going on right? Because it's just very difficult to understand particularly. Yeah, you think about insects and things like that. Then it's just incredibly difficult to monitor them in any other way. Then, through these types of approaches, you know, even now, people are discovering new species of insect. Right? Yeah, we just know very little about about what's going on. 

And when you get into the soil, what's going on in the soil? Different types of animals there, then, these things we really know very little. So it's going to really throw up a lot of a lot of interesting work, it seems to me. And I think statistics has an important role to to play there. 

And, as we were saying earlier also in terms of if people want to intervene in some way either through trying to to regroup in things or trying to build or try. And you know with this the different ways they can intervene in the environment. Then then you can be hopefully, you can begin to understand what the effects of those things are. I think that's incredibly challenging but also potentially incredibly important, and maths, in its general mathematical sciences in its general sense, will play a key role in that.

Stephanie J Dickinson: That does sound like there could be a lot of future work for people wanting to get into this type of statistics.

Jim E Griffin: Yeah.

Stephanie J Dickinson: We try and solve our climate mess.

Jim E Griffin: Yeah. And I think even different approaches. Right? So I think there's now an approach towards a marketized version way of trying to address these problems through green finance and ideas of building a market. And somehow, then you need to quantify what's going on in the environment in order for people to be able to actually run a market right? They have to have some idea.

If you say I'm going to improve in some area then, in a way, unless you quantify what you mean by that, does it mean that you have more of a particular type of species, or somehow you have more species in general, or, then it becomes, qualification becomes very important. At the moment, that's a direction, or that's one direction that people are moving in, particularly, I think, in terms of government and people.

Also another on top of the science side of things, and the things that do academic. There's also that side of things as well.

Stephanie J Dickinson: Question 8. Imagine a future where your work has become a standard method used all around the world for many different things. What would that world look like, and what would be different there to what it is like now?

Jim E Griffin: Well, hopefully, as I said, a lot of our work is really about trying to adjust these errors and biases in the data. And so hopefully, then from that, you can get a much better understanding of what's going on. And a much more realistic understanding of what's going on, and the effects of various factors in terms of deciding the communities within the environment. And so having a better understanding of that will lead to people making better decisions. It will lead to better science in the future, hopefully. I guess that's the outcome.

I'm not sure it will be radically different, but hopefully, a bit better.

Stephanie J Dickinson: Hopefully, people will know what's going on a bit more, and what they're doing. What they can do that won't cause harm.

Jim E Griffin: Yeah. So of course, I think that that's the next stage, right? Once people are able to get better science. As we improve the science and and we have better techniques for measuring things and for understanding things, then, yeah, of course, that can play into how people make decisions.

Stephanie J Dickinson: Question 9. Do you think there is a difference between the way you think about environmental statistics and other kinds of statistics?

Jim E Griffin: Don't know. I think for me. I covered it very much from the point of view of a statistical methodologist. I think I always see these these problems as data that comes from some system that needs to be modeled in in a stochastic way. And so my interest is always in in building these stochastic models and building methods for making inferencing these stochastic models. 

Whether that's in the environment or in anti doping or in econometrics. In many ways, the toolkit that you use. It's different for each of those different areas. But there's lots of things that are in common and a lot of common kind of techniques that you can bring across. So that's my view on it.

Stephanie J Dickinson: Thank you, and question 10. What unintended consequences of environmental statistics could potentially occur generally not relating to your own specific project?

Jim E Griffin: Yeah. So I think, as we as we've talked, I've emphasized the importance for adjusting for the kind of biases and errors that come in data. And I think environmental data often has a lot of that, because. Firstly, it's kind of an issue of design. So you kind of have biases that we would usually think about in statistics. 

You know that that you're going to decide to sample in a particular location and in a particular way. And maybe that potentially introduce biases. But also as people start using with kind of remote sensing and more complicated types of data. Maybe that also needs needs that needs some some kind of in modeling. And if you don't do that, you may start getting erroneous erroneous results.

I think also in terms of, we talked about measurement and regulation earlier. So it seems to me that always in in regulation there's a danger that the the measurement replaces the thing you're trying to measure. So you may be interested in biodiversity. But at some point you build a measure of it, and somehow the measure becomes the theme rather than the actual but which has to necessarily be a simplification of the actual thing you're interested in. And so always with quantification. That that's kind of the danger. That the people then just become obsessed by the actual measurements rather than the actual thing you're interested in. In the first place. But I think those are the possibilities. There's huge possibilities, I think, for environmental statistics, really. And huge advances. Recently.

Stephanie J Dickinson: That's really interesting. And I do think that we need to definitely think about unintended consequences, because from what I can see. That with other forms of research in the past there have been lots of unintended consequences. When people thought that they were doing the right thing. And I think that environmental statistics is another one of those where we are caught up in the idea that we're doing this because it's got purpose, and you know, doing the right thing. So we need to be careful that we don't get swept up and do the wrong thing while trying to do the right thing.

Jim E Griffin: Yeah, I completely agree. And I think, statistics has some idea of objectivity that isn't always true, right? Because at the end somebody is making a lot of decisions. And they may be bringing a lot of opinions that maybe are not quite. You know. They're not made explicit. And are kind of somehow implicit. Yeah, I agree. I think statisticians are aware, often a lot of these problems. It's often when it gets out, I think of the statistical community, and then those ideas get picked up, and people may be have less interest or less awareness of the nuances of method.

Stephanie J Dickinson: Hmm! Right? I think this is the last question. I think. Question 10, was the last question. Let me just check. Oh, no, I've got another question, maybe more than one. I can't remember. Right. Question 11.

It's become pretty evident that we all need to become more environmental and more sustainable. Do you think we should all try to improve our knowledge of statistics and get more people involved in similar work to do this?

Jim E Griffin: So I think always it's good for people to improve their knowledge of statistics. I think we live in a world where people throw a lot of statistics at us. And I think that can be good and bad, and as we've as we've already said, statistics are not neutral. And people often want to make a particular point, particularly around, you know, critical point right? And their  job is not really to present a fair picture of what's going on. It's to present a picture that backs up their point right?

So it's useful for people to be statistically literate. That's important. But I think also in environmental, particularly in  thinking about ecological things, then, statistics have been  quite useful. I think so. You often see these results, of surveys, people, monitoring birds and bees, and various types of insects and things. And and you know people are interested, right? That comes up in in the national media. So that goes beyond the academic or specialist kind of the audience.

And so there that seems to be quite useful for other people to understand what's going on and what's changing. And yeah,  this kind of statistical work has been very, I think, important for raising awareness.

Stephanie J Dickinson: So what you just said is sparked off a couple of more questions. I hope you don't mind me asking these. Firstly, what level do you think it would be useful? People try to aim to improve their knowledge of statistics in order to understand the statistics that are constantly thrown at us from various channels. And then, as a follow on, actually, no, let's just focus on that first.

Jim E Griffin: Well, I think it's good these days in in GCSE maths and and that kind of kind of level people increasingly have parts on statistical literacy. And so I think that's very good for young people to to be exposed to different types of data and and issues that are around that. So I'm not sure people need to have an amazingly technical knowledge.

It's just to be aware of some of the issues and the biases and some of the ways that people try and slant results so often. You often hear that something could be true, or this could happen, often. That means that it may still be incredibly unlikely to happen. Right? It's just that's the most extreme thing that could could happen. That could be true. But it's probably not true.

And often, often people present things in terms of this most extreme thing kind of end up concentrating on that. But being reality, that's probably not going to happen. And I think it's issues a bit like that. There's, you see, common things come up in newspapers right? Because I guess the people, the journalists also understand a little bit how to make a more exciting story. So I think that's under the level.

Stephanie J Dickinson: Okay, so you're saying, GCSE level would be a good level for people to aim to.

Jim E Griffin: Current Qcd. And things like that. But I think often there's courses around statistical literacy out there. I think that would be a really good thing for for a lot of people. And it's becoming increasingly important. Data is increasingly part of our lives, right? And that means that it's increasingly something that people useful to have a kind of basic understanding around.

Stephanie J Dickinson: And as a follow on from that, if the general population did start to improve their statistical knowledge, and to think more about statistics, how would that change what happens in the universities?

I'm sorry if that's a difficult question.

Jim E Griffin: I think it's quite separate, unless you know, some universities get involved in this idea of trying to teach people about statistical literacy. So before, when I was at Kent, we had kind of courses that, people within, outside, not academics, but professional services and people would would come to courses to build up their this kind of basic statistical literacy. So I think in a way, the people we work within the university are often and outside university are often highly, statistically literate, particularly in the areas they're working in. 

So thinking about this kind of environmental work. I was working with an outside company. And their people have, you know, a really good understanding of what the statistical issues that they're facing are so, I think, in terms of the university. We often are working with people who are very, at least I am working with. People are very aware of what's those kind of issues. But it would be good for society in general, it seems to me people to be more aware.

Stephanie J Dickinson: Okay, great. Thank you very much. I agree with that as well. I think that it would be fantastic if more people, including myself, could improve our knowledge of statistics, because, you know, I work in the department. But I'm not a statistician. I'm in professional services and I do the green champion role for the department, and I see what I'm doing here is trying to draw out information from you and other academics. To try and place that information for everyone to learn from on our website. So thank you very much. I really appreciate your answers to the questions. I'm pretty sure that is the last question. Yes, it is so. Thank you, Professor Jim Griffin. And do you have any last comments that you want to make?

Jim E Griffin: No, no, I think I've I've said everything that I was to say. So. Thank you for your time, Stephanie.
 

Stephanie J Dickinson: Hi! This is a sustainability interview with Domna Ladopoulou, Department of Statistical Science, UCL. Questions are by Statistical Science Green Champion, Stephanie Dickinson, myself. Hello, Domna, would you like to say something? 

Domna Ladopoulou: Hello, Stephanie, thank you very much for inviting me. 

Stephanie J Dickinson: This is part of a series of interviews that I'm doing for the Department of Statistical Science at UCL, to try to connect the subject of Statistical Science more with the discipline of Sustainability. That involves interviewing people who are doing research in statistics who have a relationship to sustainability within that research. I will ask some questions, and Domna will answer them. Let's get started with question one. What is the problem you are trying to solve? 

Domna Ladopoulou: I'm trying to tackle critical challenges on statistical modeling in the Wind Energy sector. I'm working with Professor Dellaportas, who is a member of the Statistical Science Department at UCL. 

My main research aim is to make a meaningful impact on the efficiency, reliability of wind energy production and more generally, we could say that we help in contributing to the global need for sustainable energy resources, and especially we would aim to contribute in some reliability issues in wind farms.  

During my PhD, I'm trying to develop statistical and machine learning modeling approaches that are appropriate for wind energy. One of the problems I'm trying to work on is to create a condition monitoring system for wind farms. 

Stephanie J Dickinson: Great. Thank you. How will your research benefit the solving of the problem? 

Domna Ladopoulou: I'm using data collected in wind farms which is called scatter data. We are trying to develop robust new modeling techniques. Currently, we are working on developing a probabilistic condition monitoring system for wind farms which should discover faults and failures happening in the wind generators early on. 

This system must be appropriate for the number of data collected which are millions. And ideally, it should be able to include an important number of different variables as inputs, so we could benefit from the amount of information available.  

We hope that if this system is used for condition monitoring of the wind generators, it will help us with the overall sustainability of the wind farms. 

Stephanie J Dickinson: Fantastic. That sounds really helpful. What area of statistics is your research based in? 

Domna Ladopoulou: I'm working on non-parametric probabilistic methods, such as Gaussian processes. 

And I'm also using some other machine learning models such as probabilistic neural networks that they offer good predictive capabilities and an advantage over Gaussian processes in terms of computational cost needed to train them. 

The advantages of both methods are that they offered flexibility. And I did adaptability in modeling complex relationship within the data. 

And this is something very useful for us working in the wind energy sector as the wind energy systems have a dynamic nature, and what I mean by that is that the variability of the wind speed creates the dynamic nature of the wind, power, output. 

Stephanie J Dickinson: Well, could you explain your methodology so that someone without a statistics background might understand what are the bones of it? And how does it work. 

Domna Ladopoulou: As I mentioned earlier, the scope is by using with our data to spot problems such as failures in wind farms early. So what we really do is that we use our models, which are appropriately designed for wind farms.  

To predict the wind power output, and then we compare the expected power which is the result of our model. With the observed power output. Using this information, we can build a condition monitoring system that can draw conclusions on whether the wind generators operate in a healthy way. 

So in a way, we are developing new techniques to improve wind energy, reliability and sustainability. 

Stephanie J Dickinson: Great. I can see why that would be very important. And yeah, I think improvements in efficiency. And reliability is very welcome. I think if more people who are interested in sustainability understood about methodology and applied it to their own actions, they have the potential to become more effective. 

What do you think methodologies do for sustainability? 

Domna Ladopoulou: I think that understanding what a specific methodology could potentially offer in solving a particular problem can empower people to act more effectively. 

So nowadays people can make more informed decision in comparison to the past due to the power of data analysis offered in our everyday lives. 

And by informed decision making. I mean that we can make choices based on a deep understanding of all the relevant information given for a specific problem, while at the same time we know the potential consequences of our action. 

Of course we can evaluate the risk of each choice, which is a very important benefit when we speak in terms of science. 

So for all, I think that making informed decision decisions can give us the privilege of implementing impactful strategies for sustainability practices. 

Stephanie J Dickinson: That sounds very positive. 

Domna Ladopoulou: Yeah. 

Stephanie J Dickinson: Do you think it would be possible to roll out your methodology on a mass scale? What would that look like. 

Domna Ladopoulou: I think it is feasible with the right infrastructure and collaboration. I mean that from a technical perspective, it would involve implementing a standardized data collection process across Wind farm. So the development of an automated anomaly detection system could be also feasible. 

Another component of probably higher importance this is more difficult to achieve is that it might be necessary to educate, or maybe even convince, the relevant stakeholders and policymakers about the benefits of such methods in wind energy, production and sustainability. 

And I think this is the most essential part. If we aim for widespread adoption. 

Stephanie J Dickinson: Are you collaborating with other departments across UCL or in different universities? 

Are there any external partners you are working with in industry? 

Domna Ladopoulou: Yes. So my research is multidisciplinary, since it requires some knowledge of engineering, and it also requires a statistical background. So to fill in my engineering gap. I have some collaborations with industry partners in the fields of wind energy, and this helped me to stay informed about real world challenges which I aim to solve with my research. 

But as for now, we don't have any academic collaborations, but we are more than open to explore further collaboration both in academia and industry. 

Stephanie J Dickinson: Great. 

What can you bring to the table with a potential collaboration with other academics? 

Domna Ladopoulou: I think we could bring a blend of statistical expertise and some domain knowledge in the wind energy sector. 

And we could work on open research questions in the field of wind energy. And probably we could work also in other energy related fields that require machine learning models suitable for almost the same type of complexities as these observed in wind power systems. 

Stephanie J Dickinson: Are there any particular fields of academia that you think would benefit from a collaboration with your research field? 

Domna Ladopoulou: Yes, I think so. Collaboration with scientists working on renewable energy engineering, environmental scientists, wind energy policy and could potentially bring some benefits in the field.  

In the case of engineering and environmental science, I think, using the appropriate models, we could increase renewable energy efficiency, and we could get a better understanding of the environmental impacts which I find quite important. 

And what I mean is that if we have a proper assessment study on the economic and then ecological footprint of wind farms, we could optimize their sustainability. 

And from a policy perspective, I think we could make informed policy decision making for wind energy production issues which are quite important. 

Stephanie J Dickinson: Fantastic 

When you collaborate with others, particularly thinking of your current industry collaborations. What is your main function in these projects? And how does your research change things? 

Domna Ladopoulou: The projects in which we work with the industry now is to improve the efficiency and reliability of wind energy production by developing an anomaly detection system. 

So, as I mentioned earlier, the main scope of this system is to identify anomalies in the operation of wind generators.  

And with the system we built, we detected, operational anomalies about a day before the existing conditioning monitoring system that is installed now in the wind farms. And this result could change something.  

One of them is the number of is the number and the cost of the interventions that must be done in a wind generator to return back to a healthy operation. 

So if we manage to detect earlier a potential fault, then we help in the overall reliability and sustainability of the farm by reducing a lot the maintenance cost of the wind farms, which is a very significant amount spent yearly for wind farms. 

Stephanie J Dickinson: Really helpful? Yeah. 

Where in the world will this particular piece of research be of the most use? 

Domna Ladopoulou: Yeah. This I think this work will be most beneficial in regions with a high reliance on wind power. And, of course, a significant presence of wind farms. 

Also regions with a growing or established wind energy industry might find this research useful. 

 And I'm referring to these regions, since they offer a higher concentration of wind farms, and as a result they face such challenges in the optimization of energy production. 

 The research could be also viable in markets where the development of wind farms is of growing interest, just because they will face this kind of issues on the later stage. 

Stephanie J Dickinson: Great. 

If you could think back to your formal education in statistics, what topics covered in the standard university curriculum had the most influence on your current research and your interactions with industry? 

Domna Ladopoulou: I studied in the Athens University of Economics and Business in the Department of Statistics for my bachelor degree, and I find it very hard to pick a specific topic covered. But the modules related to data, analysis and statistics for environment had some influence in my work. 

But I believe that whatever I learned in this department. Apart from helping me start, the Phd, still helps me today, since it provided me the foundation I need to grasp new concept in my new concepts in my field. 

But since you asked me this question, it's a great opportunity to thank my professors in Athens for what they taught me, and in general for what they offer to the students of this department, despite facing challenges with limited resources. 

And now, concerning my interactions with industry. I started working on wind energy, related topics during my masters dissertation, where it was my first interaction with industry experts in the field. 

And they introduced me to the wind energy related problems along with my supervisor. 

They offered me some data which I use for my thesis, and that was the starting point of my work in wind energy. 

Stephanie J Dickinson: Finally, the final question. 

Let's save the world. If you could pick anywhere or anything to apply methodology to that would have the biggest impact. What would that be? And what would the methodology change there? 

Domna Ladopoulou: So that's a great and very hard answer to answer question. But to keep the conversation in the environment, which is important. I think I would try to work on the most important to me of problems in the world, which is climate change.  

And this is an incredibly hard and complex problem to solve. But we could use AI in helping climate change. And there are currently a lot of brilliant scientists working on embedding AI to help climate change. 

I think applying statistical and machine learning models in such context would increase our awareness and understanding about climate change, which I think is an important component of tackling with this issue, or at least a good starting point. 

At the same time, AI can help with predicting natural disasters, they can offer plenty of other applications. 

For example, I watched on the news a few days ago about a system that can dive deeper than a human and using AI rather than normal GPS systems to navigate itself. It can capture very high quality video and track the corals and the sea life which is now due to the climate change moving to a deeper level of the sea in order to survive from the increased temperature of the sea. 

 So again, as I mentioned earlier. In our discussion, an important feature of statistics is that it helped us to make informed decision-making, and I believe it is a necessary step to help us act towards a more sustainable future. 

Stephanie J Dickinson: Great. I'm really pleased with that final answer to that question. 

I very much agree with you on those points. And I think that you're doing some really good work. And it sounds like there is a lot of hope for statistics to help solve these issues that that we're facing right now in the entire world. 

With climate change. So thank you very much for putting all this work into it as a professional working in the field of of statistics, on sustainability issues. 

And I think a lot of people would be very appreciative of that work. And, I think there's a lot of opportunities for people to work with statistics in the future to help solve these issues? 

So hopefully, we can all use our minds and get ourselves out of this problem. But yeah, thank you. Thank you very much, Domna. 

Domna Ladopoulou: Thank you very much Stephanie. 

Stephanie J Dickinson: Yeah, for this brilliant interview. 

Domna Ladopoulou: I'm very glad we you liked our work. 

Stephanie J Dickinson: Yes, right. If anyone wants to contact you about your research, how can they contact you? 

Domna Ladopoulou: So I have a web page, the UCL profile webpage under my name, which is Domna Ladopoulou. Also they can contact me on my email which is available. And I also have a Linkedin profile. So I'm happy to discuss further. 

Stephanie J Dickinson: Great right. Thank you very much. And hope you have a lovely rest of the day. 

Domna Ladopoulou: Thank you very much. 

Environmental Statistics

Data Science Methodology for Weather and Climate

• Recycling our old electronic equipment
• Creating bespoke initiatives to lower our carbon emissions caused by computer usage
• Setting temperature guidelines for heating
• Encouraging the use of reuseable containers for lunch
• Banning the use of disposables from our catered events, and only serving vegetarian food
• Using boiled water from an instant tap
• Only printing what we need, and digitising everything else
• Selling our vintage textbook collection with proceeds going to charity
• ...more to be announced, check back for updates on future sustainability projects!