Buy a hand-crafted ceramic daffodil from the Ten Thousand Daffodils installation

Where’s Wally for animals: when species are hard to find

June 27th, 2024

Protected areas are crucial to offer animals a safe haven from threats like overfishing or hunting, but they only work if the target animals are there in the first place. This can be surprisingly difficult to determine, especially if a species is rare or if we don’t know where it lives. However, a recent study has built a reliable tool to map their whereabouts. The team of scientists who created it tested the model on 11 fish species and found that most marine protected areas are not where these animals are actually found.

Since we cannot inform animals of where a new protected area will be established, our best bet is to create them where we know the endangered species already live. As brilliant as this idea sounds, it’s hard to put it in practice. When it comes to protecting animals, in fact, most people assume that ‘how?’ is the most important question, but that is because we often take the ‘where?’ for granted. In truth, knowing where your target animals are found is a tricky business, and for some species you might just have old or patchy data to guide you [1]. Where then should you place a protected area? A team of scientists has recently developed a tool to answer this question. This tool is a ‘Combined Model for Accurate Prediction’; an innovative model that works the same way as a jigsaw puzzle, where multiple pieces each showing a little part of the scene on their own are combined to form a fuller picture. In this case each piece is another model, and together they form a single map of where the animal is most likely to be.

If you took two puzzle pieces and looked at them, you could see what looks like a bee in one and some snow in the other, leaving you confused about what scene you are trying to recreate. Similarly, different models can produce results that seem to conflict with each other [2]. What the scientists did in this scenario was to select a handful of informative puzzle pieces (models) for each of 11 species of fish and use them to produce two maps: one showing where a fish is abundant, and the other showing where it cannot live. The areas that these maps have in common tells us where the fish might be, and combining all three gives us a complete overview of how likely a fish is to be found in a particular place. Pretty impressive, isn’t it? The species used to test this model were all diadromous fish, like salmon and trout. ‘Diadromous’ is a fancy name given to all fish that migrate between rivers and seas [3]. As these fish migrate between different areas, they are harder to protect, and since we eat them, we need to make sure not to catch too many [4]. Multiple marine protected areas (MPAs) have been set up with the intent to protect these species, but to know whether they are working we need to know if diadromous fish cross them. These fish were also chosen because their importance to us mean that their whereabouts are now well monitored, so that they can be used to confirm if the final map made by the model is correct. This knowledge wasn’t available when MPAs were first created and, sadly, it shows.

As the model found out, most MPAs that were designed to protect diadromous fish are outside the areas where these species live. Additionally, even when a species is really abundant in an MPA, only half of that area is dedicated to their protection. The message is clear: marine protected areas are not where fish are, so they cannot shelter their populations from threats like overfishing. As distressing as this may sound, having a tool that can confidently tell you where species that we have little data on might be can help us reshape MPAs to become more effective year after year. And not only that! Although this model was tested on marine creatures, it can do the same for every animal. Think of the elusive harvest mouse and the rare spotted crake in the UK, the wolves and lynx quietly making their way back throughout mainland Europe [5]… This model has the potential to become a key helper for planning protected areas in the future, and it can already aid us in adjusting their position now.

How to find Another Way

Learning about the impressive advancements of science always fills me with a desire to take action and do my part. Luckily, there is always a way to help. Models like the ones used to produce the final map in this study rely on data. This is often gathered when professionals survey a particular species, but surveys are expensive and cannot be done too extensively or very often. If you are taking a walk, try and write down a few species you see along the way and where. You can then upload your sightings on apps such as iRecord, iNaturalist, and PlantNet. Who knows? Maybe they will help future projects decide the best course of action to protect the nature all around you.

Article by Maria Giulia Checchi


  1. Bean, W.T., Stafford, R., and Brashares, J.S. (2012). The effects of small sample size and sample bias on threshold selection and accuracy assessment of species distribution models. Ecography 35, 250–258.
  2. Elith, J., H. Graham, C., P. Anderson, R., Dudík, M., Ferrier, S., Guisan, A., J. Hijmans, R., Huettmann, F., R. Leathwick, J., Lehmann, A., et al. (2006). Novel methods improve prediction of species’ distributions from occurrence data. Ecography 29, 129–151.
  3. Lene Klubben Sortland, Aarestrup, K., and Birnie‐Gauvin, K. (2024). Comparing the migration behavior and survival of Atlantic salmon (Salmo salar) and brown trout (Salmo trutta) smolts. Journal of fish biology.
  4. Marie-Line Merg, Olivier Dézerald, Kreutzenberger, K., Demski, S., Yorick Reyjol, Philippe Usseglio-Polatera, and Belliard, J. (2020). Modeling diadromous fish loss from historical data: Identification of anthropogenic drivers and testing of mitigation scenarios. PloS one 15, e0236575–e0236575.
  5. Woodward I, Aebischer N, Burnell D, Eaton M, Frost T, Hall C, Stroud DA, Noble D. Population estimates of birds in Great Britain and the United Kingdom. Br. Birds. 2020; 113:69-104.