Throughout March, we are featuring articles shortlisted for the 2025 Robert May Prize. The Robert May Prize is awarded by the British Ecological Society each year for the best paper in Methods in Ecology and Evolution written by an early career author. Robin Boyd’s article ‘Using causal diagrams and superpopulation models to correct geographic biases in biodiversity monitoring data‘ is one of those shortlisted for the award.

Read Robs previous blog on the paper here – It is only by understanding what causes sampling bias that we can correct it – Methods Blog

About the paper

What is your shortlisted paper about, and what are you seeking to answer with your research?  

Biodiversity monitoring relies on data collected at a sample of sites to draw conclusions about what is happening to species across the wider landscape. Extrapolating beyond the sample requires assumptions. A common assumption is that, among sites with certain characteristics, species are faring similarly at sampled to non-sampled locations. Our paper describes a method for identifying the site characteristics that make this assumption plausible. It is essentially a principled approach to variable selection that can be used with any statistical model, and, unlike traditional approaches, it prioritises a reduction in bias over in-sample predictive performance.

Were you surprised by anything when working on it? Did you have any challenges to overcome? 

One thing that surprised me was just how much our bias-adjusted results differed from previous ones. For one of our two examples, the small pearl-bordered fritillary, the estimated trend in abundance switched in sign. That really brought home how strongly our inferences depend on the assumptions we make.

What is the next step in this field going to be?  

As policymakers increasingly demand information on how species are faring at local scales, analysts are becoming less interested in estimating parameters describing large-scale change and more interested in making site-level predictions. People often assume that prediction is all about minimising in-sample error. But if we want to make predictions for unsampled sites, we need to make the kind of assumption described in my answer to question one. Our method is therefore just as applicable to local monitoring and prediction as it is to estimating parameters at larger scales.

What are the broader impacts or implications of your research for policy or practice?  

Governments are beginning to set legally binding targets for halting and reversing declines in species’ abundances. Whether these targets have been met is ultimately an empirical question that can only be answered using a combination of the available data and assumptions. If our assumptions are plausible, then we will arrive at estimates that faithfully reflect progress towards targets. Otherwise, we won’t. Our method was designed specifically to help analysts of biodiversity data identify the variables that, once included in their models, make their assumptions plausible (or at least more plausible than alternatives).

About the author

How did you get involved in ecology?

I grew up by the sea, which got me into marine biology. From there it was a short journey to ecology more generally. But if I’m honest, what really fascinates me is how things work from a theoretical perspective.

What is your current position? 

I’m a quantitative ecologist working at the UK Centre for Ecology and Hydrology. I spend my time developing theory and methods to improve biodiversity monitoring. That’s the idea, at least!

Have you continued the research your paper is about? 

We have. I’m most excited about a theoretical paper, which we’ve just submitted. The paper makes the case that causal diagrams are a universal framework for displaying the assumptions needed to answer causal, descriptive and predictive questions. We’re also exploring how Large Language Models might assist domain experts in constructing causal diagrams. This would make it much easier to roll out our method to more species and datasets—something that we’re also planning to do.  

What one piece of advice would you give to someone in your field?  

Try not to think exclusively in terms of statistical models. Instead, define the theoretical quantity that you wish to learn about then consider the assumptions needed to recover its value from the available data. The statistical model is merely a device for invoking those assumptions to arrive at an answer.