Modelling habitats: the importance of accuracy in investigating suitable areas for species
Luca Chiaverini, a DPhil candidate in the Department of Biology’s Wildlife Conservation Research Unit, describes his work on habitat modelling to help understand how we can best protect endangered species. This work demonstrates the importance of accurate models when working with conservation policies and management plans.
When looking at how we can help endangered species, a key consideration is which environmental factors of their habitat make it suitable and where those habitats can be found. These factors could be natural such as the presence of other particular species or density of forests, or anthropic human-influenced factors, such as levels of forest loss or pollution in an area.
One of the ways we can investigate how these factors make a habitat suitable, and thus influence the success of a species, is through species distribution models – but there are multiple to choose from.
I have been investigating the distribution of small cat species, such as clouded leopards (Neofelis nebulosa and N. diardi) and the marbled cat (Pardofelis marmorata), across Southeast Asia using camera trap data, looking at whether species were present or absent in images to understand the areas they were inhabiting. I am particularly interested in looking at how anthropic disturbance influences habitat suitability for these species.
While modelling this distribution, I was testing a few of the available frameworks and was struck by the differences displayed by distinct models. While the models should clearly determine the relationship between species and their habitats, I found that there can be big or even contradictory differences in conclusions, which adds further complication to picking which one to use. Which of these models accurately describes the real-world relationship between these species and their relationship with their habitat?
This is worrying because species distribution models are widely used in conservation, to help predict things such as species resilience and survival based on their habitat. This in turn helps to inform conservation priorities and management policies, such as what elements of the landscape to preserve in order to help a species. However it seems we should be much more uncertain about the results than we thought. In this work I have tested only a few of the existing modelling techniques, so in the future I plan to expand the suite of models that I test.
To make sure we really understand these complex ecological relationships between species and their habitat, we need to compare outcomes from multiple models – particularly for different species and taxonomic groups – as we cannot use one model for all species. Different species interact with their habitat in very different ways, and certain models might be a better fit for some than others. I hope that through this work, we will have a better understanding of different models and work towards a system where we can pick the model that is most appropriate, accurate, and effective for a particular ecosystem and species to best protect these animals in the wild.