← Latest papers
🌿 ecology

Reduced-Rank Generalised Dissimilarity Modelling (RRGDM): Model-Based Ordination of Dissimilarity Data

This paper introduces Reduced-Rank Generalised Dissimilarity Modelling (RRGDM), a novel framework that reframes dissimilarity modelling as a model-based ordination problem to resolve the mismatch between site-level predictors and pairwise responses, thereby enabling robust inference on environmental drivers of beta diversity while disentangling their effects on community composition versus dispersion.

Original authors: Hernandez-Carrasco, D., Koerich, G., Lai, H. R., Woolley, S. N. C., Tonkin, J. D., Hui, F. K. C.

Published 2026-10-28
📖 5 min read🧠 Deep dive

Original authors: Hernandez-Carrasco, D., Koerich, G., Lai, H. R., Woolley, S. N. C., Tonkin, J. D., Hui, F. K. C.

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ⚕️ This is an AI-generated explanation of a preprint that has not been peer-reviewed. It is not medical advice. Do not make health decisions based on this content. Read full disclaimer

In the natural world, life rarely exists in isolation. A patch of forest, a pond, or a stretch of coral reef is a complex assembly of countless species interacting with one another and their surroundings. Ecologists have long sought to understand how these communities change from place to place, a concept known as beta diversity. To measure this, scientists often calculate how different two communities are from each other, creating a "dissimilarity" score. If two sites share many of the same plants or animals, they are similar; if they share few, they are dissimilar. For decades, researchers have used these scores to map out the invisible forces shaping life on Earth, such as temperature, rainfall, or soil type. However, a fundamental problem has persisted: these scores are calculated between pairs of locations, yet the environmental factors driving the change act on individual sites. Trying to link a relationship between two points to the conditions at a single point has been like trying to describe the weather by only looking at the distance between two cities, rather than the conditions inside each city itself. This mismatch has made it difficult to untangle whether a change in community is due to a shift in the average types of species present, or simply because the variety of species has become more chaotic and unpredictable.

A team of researchers has now proposed a new way to solve this puzzle, offering a clearer lens through which to view the shifting tapestry of life. They developed a method called Reduced-Rank Generalised Dissimilarity Modelling, which reframes the problem entirely. Instead of forcing a connection between pairs of sites, the researchers imagine a hidden, low-dimensional space where every community exists as a single point. In this invisible landscape, the distance between two points represents how different the communities are. The researchers then map the environmental conditions of each site directly onto this space, allowing them to see how factors like snow cover or human disturbance push communities apart or pull them together. Crucially, this approach separates two distinct processes that often get confused: the directional shift in the average makeup of a community, and the random variability in how much individual samples differ from that average.

To test their idea, the team first ran extensive computer simulations. They created artificial communities with known rules for how they should change and how much they should vary. When they applied their new method to this simulated data, it successfully recovered the hidden rules they had set, accurately identifying which environmental factors were driving the changes and how strongly they acted. The simulations also showed that as the number of samples increased, the method became increasingly precise, correctly distinguishing between a true shift in the community's identity and a mere increase in random noise. This proved that the mathematical framework was sound and capable of separating the signal from the noise in a way previous methods struggled to do.

The researchers then put their method to work on real-world data, starting with a study of alpine plant communities in the French Alps. They examined how the length of the snow-cover season and different types of physical disturbance affected plant life. The new model revealed a dominant gradient driven by the duration of snow cover, showing a near-complete turnover of plant species as the snow melted later in the year. It also detected a significant effect from disturbance caused by marmots, which acted in a direction almost perpendicular to the snow gradient. Perhaps most revealing was the model's ability to show that even when the average plant community looked the same under identical snow conditions, the actual plants found in different spots could vary wildly. The method quantified this random variability, showing that under identical environmental conditions, the difference between two random samples could be substantial, a nuance that older methods often missed or conflated with the average change.

In a second application, the team revisited data from a pond experiment designed to test how productivity affects the diversity of invertebrate communities. Previous analyses of this data had relied on separate statistical tests to look at average changes and variability, a process that often struggled to keep the two distinct. The new method, however, handled both simultaneously within a single framework. It confirmed that ponds with high productivity had much greater variability in their communities than those with low productivity, while also detecting a clear difference in the average types of species present. By visualizing these results in the same hidden space, the researchers could see exactly how the high-productivity ponds were not just different in their average composition, but also much more scattered and unpredictable in their individual samples.

This approach offers a significant step forward for ecological science because it treats the data in a way that matches how nature actually works: environmental factors act on individual sites, and the resulting differences are observed between them. By placing samples in a shared, interpretable space, the method allows scientists to see not just how communities change, but how the very nature of that change varies. It provides a way to visualize the uncertainty and the randomness inherent in biological systems, rather than smoothing them over. While the method requires careful handling of large datasets, it opens the door to more accurate predictions about how biodiversity will respond to a changing world, distinguishing between a community that is steadily transforming and one that is becoming increasingly unstable. The result is a tool that does not just measure the distance between places, but explains the journey of life that connects them.

Drowning in papers in your field?

Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.

Try Digest →