Bob O’Hara: The Rise of the Ecological Modeller

Image Credit: Internet Archive Book Image, Public Domain, Image Cropped

Non-scientists still often think of ecologists as field workers in cargo shorts, running around a grassland with a notebook and a tape measure. Whilst I’d be remiss to say this wasn’t a percentage of us, the last two decades has seen the rise of ecological modelling, which has resulted in a new breed of ecologist. One who is capable of working almost exclusively with data, producing species distribution maps and population fluctuation graphs without leaving the office.

At the forefront of this group is Bob O’Hara, who has long claimed he plans to retire the moment he figures out whether he’s a biologist or statistician. Bob currently works at the Norwegian University of Science and Technology, spending his time with the Centre for Biodiversity Dynamics and the Departments of Mathematical Sciences. I spoke to Bob about the history of ecological modelling, its integration into the wider field, and problems with modern ecological modelling.

Sam Perrin (SP): Ecological modelling seems to be a relatively recent phenomena, but it’s been around for much longer, correct?

Professor Bob O’Hara, Department of Mathematical Sciences (BO): It’s gained traction recently, true, but we’ve been using modelling for hundreds of years. Fibonacci is a famous example, but even going back before then, there were Islamic scholars doing basic ecological modelling. There have been more recent landmarks of course. Macarthur and Wilson’s Theory of Island Biogeography was a landmark. People read it and realised that they can create models with are quite simple models but tell us quite a lot about ecological processes. It changed ecological perspective from just collection to a more theoretical field.

SP: Modelling these days is a facet of mainstream science. Yet 40-50 years ago you didn’t need any modelling to be an ecologist. Do you there was was a point at which modelling tipped over into the mainstream?

BO: It’s a difficult question. MacArthur and Wilson are seen as being when it all changed, but whether that is actually true I don’t know. There are important models which go back further. Lotka-Volterra modelled relationships between prey and predators back in the early 1900s. Before then, Preston and Fisher did some ecological modelling. But there have been big changes along the way, a lot of that is in line with changes in methods or technology. That’s what Macarthur and Wilson was about. And then with computers nowadays, models got a lot more complicated.

SP: The advent of new technology, and big data, has it been well integrated?

BO: I have this theory of how technology is used in science. Not just in ecology, also other things, molecular biology being another example. Someone finds this new amazing method, and everybody leap in and starts to use it. Then after a couple of years, people people put their hand up and say, “hang on, maybe we’re not using this new tech the right way”.

So MRI is one example. People were looking at brains, and seeing which areas lit up during certain activities. It was very exciting, but then people started to see cracks in the method, there were theoretical papers about false probabilities. Finally someone put a dead salmon in an MRI machine and it lit up. So then people realised they were doing something wrong with this new technology and they sat down and thought about how they should be using it. I think with any new technology this is what happens. People are in a rush to use it and at some point, someone goes “hang on a moment, this isn’t right”. And then you get the second phase where people start to think more seriously about what they should be doing.

SP: What are some of the main problems with ecological modelling today?

BO: My rather rude response is that it’s done by people who aren’t competent mathematically. I really feel that you should have some understanding of the mathematics. You don’t need to be professors in mathematics, but people should be able to get a reasonable grasp of the mathematics behind the models. A lot of it is the use of computers. They make it very easy to do very complicated things. And you get to the point where you don’t really understand how your model works. And what’s the point of having a model if you don’t understand what’s going on in it.

That’s the problem. It’s just become too easy for people to do things badly. Take species distribution models for example. The problem is that you just grab the data from GBIF, you put it into a modelling package like MaxEnt, you get some pretty graphs. But there’s lots of problems with GBIF data. You need to do some careful thinking through about what the data is, which of it to use, how to use it. And MaxEnt itself requires an understanding of the ecological processes behind your data. These aren’t just black boxes.

SP: You’ve just written an enormous review of statistical modelling with a number of prominent ecological modellers, among them people like Carsten Dormann, Miguel Araujo, Carsten Rahbek. What was the aim of the paper?

BO: That project was started by a group of people having similar frustrations as I’ve just expressed. That second stage of having new technology, where people sit back and think more deeply about how to use it, this review will hopefully kick this off. I was pulled in after the process had started, and we essentially had to put together a list of things to think about when doing ecological modelling. So that when you start a project, you can sit down with this list and make sure you’re considering your models correctly.

There was a paper in the journal Surgery, and the researchers showed that using checklists was saving lives. Just with lists of things you had to remember, cleaning your scalpel, that sort of thing. Just by going through that checklist, the number of errors people were making was being reduced. So producing a checklist for modelling was the same idea. By going through this checklist, maybe human lives won’t be saved, but other species will be.

Whilst new technology can make it easy to rush off and construct population models for species like these House sparrows, Bob urges caution, and a greater understanding of the mathematics behind the models

Whilst new technology can make it easy to rush off and construct population models for species like these House sparrows, Bob urges caution, and a greater understanding of the mathematics behind the models (Image Credit: hedera.baltica, CC BY-SA 2.0)

SP: You’ve mentioned that people often lack an understanding of the mathematics behind the models. Yet some of these models give very practical and useful outputs. Does everyone putting these models to use need a mathematical background?

BO: I think it’s making sure that maybe not the people using those maps, but that the people who are producing those maps understand what’s going on. If that’s decision makers then yes, they should understand the process. If it’s someone else then they should be able to explain the basic concepts to the decision makers.

SP: Is there a large lag between the advent of new technology and our rush to use it?

BO: Well mathematicians came up with the foundations of modelling ideas in the 30s and 40s. It was basically just logistic regression. The main thing is happening today is that rather than just a straight line they fit a complicated straight line. The technology to make the line you’re fitting a bit more wiggly is the new thing, but the basic underline is just a generalised linear model. So that’s a big lag. But I think it’s a product of the questions ecologists have been asking. And also the technological shift was in being able to collect the data.

BO: Do you think this reliance on plugging into models has detracted from the need to understand biology and the need to understand the biological processes behind a model?

BO: I suspect so. But I think that the rise of interdisciplinarity has filled that gap. When you’re doing macroecology and taking such a broad view of things, you can’t think too deeply about each individual species. But there are others out there who can. In my department we has some really good natural historians. So if you need to discuss sparrows, you know who to go to to discuss sparrows. That knowledge doesn’t need to be crammed into one head.

During my PhD I was working on plant pathology. One of the field technicians in the group, he was the guy to go to to discuss field trials, he just knew everything the sort of things you don’t learn if you do modelling. And this is what’s great about interdisciplinarity. In a sense, lacking one knowledge base doesn’t matter as long as you know who to talk to. So when I did my field trials, I would talk to our field technician. If I was going to do and molecular work I’d talk to the lab technician because she knew all that sort of stuff. So you do lose a bit of natural history if you do modelling. What’s important is to learn to be aware of what you don’t know.

You can read a full list of Bob O’Hara’s publications here.


  • I personally think that the rise of the modeller with little mathematical background has largely been at the expense of people with good taxononomic and organismal knowledge, just at the time when we are starting to realise that a lot of modelling in the service of biodiversity conservation is actually not very good, largely for the reasons that Bob outlines or implies. I spend an awful lot of my time constructively criticising modelling exercises from the point of view of data selection and processing, not only because modellers may have little taxonomic background (note that this is not just about identifying species in the field, but also about understanding the taxonomic process, and the related, but subtly different, process of databasing occurrence records), but also because they do not understand the observation processes that have created the datasets that they would like to use. For example, in the UK, even a database such as that of the Botanical Society of Britain and Ireland, with around 45 million observations, has a lot of historic quirks that need to be understood when using the data for any particular application at any particular scale. See, e.g.,


  • Pingback: In Silico Science: Ecology Without the Nature | Ecology for the Masses

  • Pingback: Ecological Modelling, the Coronavirus, and Why They’re Not A Perfect Match | Ecology for the Masses

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s