In the field of ecology, there is a growing interest in machine (including deep) learning for processing and automatizing repetitive analyses on large amounts of images collected from camera traps, drones and smartphones, among others. These analyses include species or individual recognition and classification, counting or tracking individuals, detecting and classifying behavior. By saving countless times of manual work and tapping into massive amounts of data that keep accumulating with technological advances, machine learning is becoming an essential tool for ecologists. We refer to recent papers for more details on machine learning for ecology and evolution (Besson et al. 2022, Borowiec et al. 2022, Christin et al. 2019, Goodwin et al. 2022, Lamba et al. 2019, Nazir & Kaleem 2021, Perry et al. 2022, Picher & Hartig 2023, Tuia et al. 2022, Wäldchen & Mäder 2018).

In their paper, Nakagawa et al. (2023) conducted a systematic review of the literature on machine learning for wildlife imagery. Interestingly, the authors used a method unfamiliar to ecologists but well-established in medicine called rapid review, which has the advantage of being quickly completed compared to a fully comprehensive systematic review while being representative (Lagisz et al., 2022). Through a rigorous examination of more than 200 articles, the authors identified trends and gaps, and provided suggestions for future work. Listing all their findings would be counterproductive (you’d better read the paper), and I will focus on a few results that I have found striking, fully assuming a biased reading of the paper. First, Nakagawa et al. (2023) found that most articles used neural networks to analyze images, in general through collaboration with computer scientists. A challenge here is probably to think of teaching computer vision to the generations of ecologists to come (Cole et al. 2023). Second, the images were dominantly collected from camera traps, with an increase in the use of aerial images from drones/aircrafts that raise specific challenges. Third, the species concerned were mostly mammals and birds, suggesting that future applications should aim to mitigate this taxonomic bias, by including, e.g., invertebrate species. Fourth, most papers were written by authors affiliated with three countries (Australia, China, and the USA) while India and African countries provided lots of images, likely an example of scientific colonialism which should be tackled by e.g., capacity building and the involvement of local collaborators. Last, few studies shared their code and data, which obviously impedes reproducibility. Hopefully, with the journals’ policy of mandatory sharing of codes and data, this trend will be reversed. 

REFERENCES

Besson M, Alison J, Bjerge K, Gorochowski TE, Høye TT, Jucker T, Mann HMR, Clements CF (2022) Towards the fully automated monitoring of ecological communities. Ecology Letters, 25, 2753–2775. https://doi.org/10.1111/ele.14123

Borowiec ML, Dikow RB, Frandsen PB, McKeeken A, Valentini G, White AE (2022) Deep learning as a tool for ecology and evolution. Methods in Ecology and Evolution, 13, 1640–1660. https://doi.org/10.1111/2041-210X.13901

Christin S, Hervet É, Lecomte N (2019) Applications for deep learning in ecology. Methods in Ecology and Evolution, 10, 1632–1644. https://doi.org/10.1111/2041-210X.13256

Cole E, Stathatos S, Lütjens B, Sharma T, Kay J, Parham J, Kellenberger B, Beery S (2023) Teaching Computer Vision for Ecology. https://doi.org/10.48550/arXiv.2301.02211

Goodwin M, Halvorsen KT, Jiao L, Knausgård KM, Martin AH, Moyano M, Oomen RA, Rasmussen JH, Sørdalen TK, Thorbjørnsen SH (2022) Unlocking the potential of deep learning for marine ecology: overview, applications, and outlook†. ICES Journal of Marine Science, 79, 319–336. https://doi.org/10.1093/icesjms/fsab255

Lagisz M, Vasilakopoulou K, Bridge C, Santamouris M, Nakagawa S (2022) Rapid systematic reviews for synthesizing research on built environment. Environmental Development, 43, 100730. https://doi.org/10.1016/j.envdev.2022.100730

Lamba A, Cassey P, Segaran RR, Koh LP (2019) Deep learning for environmental conservation. Current Biology, 29, R977–R982. https://doi.org/10.1016/j.cub.2019.08.016

Nakagawa S, Lagisz M, Francis R, Tam J, Li X, Elphinstone A, Jordan N, O’Brien J, Pitcher B, Sluys MV, Sowmya A, Kingsford R (2023) Rapid literature mapping on the recent use of machine learning for wildlife imagery. EcoEvoRxiv, ver. 4 peer-reviewed and recommended by Peer Community in Ecology.  https://doi.org/10.32942/X2H59D

Nazir S, Kaleem M (2021) Advances in image acquisition and processing technologies transforming animal ecological studies. Ecological Informatics, 61, 101212. https://doi.org/10.1016/j.ecoinf.2021.101212

Perry GLW, Seidl R, Bellvé AM, Rammer W (2022) An Outlook for Deep Learning in Ecosystem Science. Ecosystems, 25, 1700–1718. https://doi.org/10.1007/s10021-022-00789-y

Pichler M, Hartig F Machine learning and deep learning—A review for ecologists. Methods in Ecology and Evolution, n/a. https://doi.org/10.1111/2041-210X.14061

Tuia D, Kellenberger B, Beery S, Costelloe BR, Zuffi S, Risse B, Mathis A, Mathis MW, van Langevelde F, Burghardt T, Kays R, Klinck H, Wikelski M, Couzin ID, van Horn G, Crofoot MC, Stewart CV, Berger-Wolf T (2022) Perspectives in machine learning for wildlife conservation. Nature Communications, 13, 792. https://doi.org/10.1038/s41467-022-27980-y

Wäldchen J, Mäder P (2018) Machine learning for image-based species identification. Methods in Ecology and Evolution, 9, 2216–2225. https://doi.org/10.1111/2041-210X.13075

Capture-Mark-Recapture (CMR) data are commonly used to estimate ecological variables such as abundance, survival probability, or transition rates from one state to another (e.g. from juvenile to adult, or migration from one site to another). Many studies have shown how estimations can be affected by neglecting one aspect of the population under study (e.g. the heterogeneity in survival between individuals) or one limit of the methodology itself (e.g. the fact that observers might not detect an individual although it is still alive). Strikingly, very few studies have yet assessed the robustness of one fundamental assumption of all CMR-based inferences: marks are supposed definitive and immutable. If they are not, how are estimations affected? Addressing this issue is the main goal of the paper by Touzalin et al. (2023), and they did a very nice work. But, because the answer is not that simple, it also calls for further investigations.

When and why would mark loss bias estimation? In at least two situations. First, when estimating survival rates: if an individual loses its mark, it will be considered as dead, hence death rates will be overestimated. Second, more subtly, when estimating transition rates: if one individual loses its mark at the specific moment where its state changes, then a transition will be missed in data. The history of the marked individual would then be split into two independent CMR sequences as if there were two different individuals, including one which died.

Touzalin et al. (2023) thoroughly studied these two situations by estimating ecological parameters on 1) well-thought simulated datasets, that cover a large range of possible situations inspired from a nice compilation of hundreds of estimations from fish and bats studies, and 2) on their own bats dataset, for which they had various sources of information about mark losses, i.e. different mark types on the same individuals, including mark based on genotypes, and marks found on the soil in the place where bats lived. Their main findings from the simulated datasets are that there is a general trend for underestimation of survival and transition rates if mark loss is not accounting for in the model, as it would be intuitively expected. However, they also showed from the bats dataset that biases do not show any obvious general trend, suggesting complex interactions between different ecological processes and/or with the estimation procedure itself.

The results by Touzalin et al. (2023) strongly suggest that mark loss should systematically be included in models estimating parameters from CMR data. In addition to adapt the inferential models, the authors also recommend considering either a double marking, or even a single but ‘permanent’ mark such as one based on the genotypes. However, the potential gain of a double marking or of the use of genotypes is still to be evaluated both in theory and practice, and it seems to be not that obvious at first sight. First because double marking can be costly for experimenters but also for the marked animals, especially as several studies showed that marks can significantly affect survival or recapture rates. Second because multiple sources of errors can affect genotyping, which would result in wrong individual assignations especially in populations with low genetic diversity or high inbreeding, or no individual assignation at all, which would increase the occurrence of missing data in CMR datasets. Touzalin et al. (2023) supposed in their paper that there were no genotyping errors, but one can doubt it to be true in most situations. They have now important and interesting other issues to address.

References

Frédéric Touzalin, Eric J. Petit, Emmanuelle Cam, Claire Stagier, Emma C. Teeling, Sébastien J. Puechmaille (2023) Mark loss can strongly bias demographic rates in multi-state models: a case study with simulated and empirical datasets. BioRxiv, ver. 3 peer-reviewed and recommended by Peer Community in Ecology. https://doi.org/10.1101/2022.03.25.485763

Eyespots—round or oval spots, usually accompanied by one or more concentric rings, that together imitate vertebrate eyes—are found in insects of at least three orders and in some tropical fishes (Stevens 2005). They are particularly frequent in Lepidoptera, where they occur on wings of adults in many species (Monteiro et al. 2006), and in caterpillars of many others (Janzen et al. 2010). The resemblance of eyespots to vertebrate eyes often extends to details, such as fake « pupils » (round or slit-like) and « eye sparkle » (Blut et al. 2012). Larvae of one hawkmoth species even have fake eyes that appear to blink (Hossie et al. 2013). Eyespots have interested evolutionary biologists for well over a century. While they appear to play a role in mate choice in some adult Lepidoptera, their adaptive significance in adult Lepidoptera, as in caterpillars, is mainly as an anti-predator defense (Monteiro 2015). However, there are two competing hypotheses about the mechanism by which eyespots confer defense against predators. The « intimidation » hypothesis postulates that eyespots intimidate potential predators, startling them and reducing the probability of attack. The « deflection » hypothesis holds that eyespots deflect attacks to parts of the body where attack has relatively little effect on the animal’s functioning and survival. In caterpillars, there is little scope for the deflection hypothesis, because attack on any part of a caterpillar’s body is likely to be lethal. Much observational and some experimental evidence supports the intimidation hypothesis in caterpillars (Hossie & Sherratt 2012). In adult Lepidoptera, however, both mechanisms are plausible, and both have found support (Stevens 2005). The most spectacular examples of intimidation are in butterflies in which eyespots located centrally in hindwings and hidden in the natural resting position are suddenly exposed, startling the potential predator (e.g., Vallin et al. 2005). The most spectacular examples of deflection are seen in butterflies in which eyespots near the hindwing margin combined with other traits give the appearance of a false head (e.g., Chotard et al. 2022; Kodandaramaiah 2011). 

Most studies have attempted to test for only one or the other of these mechanisms—usually the one that seems a priori more likely for the butterfly species being studied. But for many species, particularly those that have neither spectacular startle displays nor spectacular false heads, evidence for or against the two hypotheses is contradictory.  

Iserhard et al. (2024) attempted to simultaneously test both hypotheses, using the neotropical nymphalid butterfly Caligo martia. This species has a large ventral hindwing eyespot, exposed in the insect’s natural resting position, while the rest of the ventral hindwing surface is cryptically coloured. In a previous study of this species, De Bona et al. (2015) presented models with intact and disfigured eyespots on a computer monitor to a European bird species, the great tit (Parus major). The results favoured the intimidation hypothesis. Iserhard et al. (2024) devised experiments presenting more natural conditions, using fairly realistic dummy butterflies, with eyespots manipulated or unmanipulated, exposed to a diverse assemblage of insectivorous birds in nature, in a tropical forest. Using color-printed paper facsimiles of wings, with eyespots present, UV-enhanced, or absent, they compared the frequency of beakmarks on modeling clay applied to wing margins (frequent attacks would support the deflection hypothesis) and (in one of two experiments) on dummies with a modeling-clay body (eyespots should lead to reduced frequency of attack, to wings and body, if birds are intimidated). Their experiments also included dummies without eyespots whose wings were either cryptically coloured (as in unmanipulated butterflies) or not. Their results, although complex, indicate support for the deflection hypothesis: dummies with eyespots were mostly attacked on these less vital parts. Dummies lacking eyespots were less frequently attacked, especially when they were camouflaged. Camouflaged dummies without eyespots were in fact the least frequently attacked of all the models. However, when dummies lacking eyespots were attacked, attacks were usually directed to vital body parts. These results show some of the complexity of estimating costs and benefits of protective conspicuous signals vs. camouflage (Stevens et al. 2008).

Two complementary experiments were conducted. The first used facsimiles with « wings » in a natural resting position (folded, ventral surfaces exposed), but without a modeling-clay « body ». In the second experiment, facsimiles had a modeling-clay « body », placed between the two unfolded wings to make it as accessible to birds as the wings. However, these dummies displayed the ventral surfaces of unfolded wings, an unnatural resting position. The study was thus not able to compare bird attacks to the body vs. wings in a natural resting position. One can understand the reason for this methodological choice, but it is a limitation of the study.

The naturalness of the conditions under which these field experiments were conducted is a strong argument for the biological significance of their results. However, the uncontrolled conditions naturally result in many questions being left open. The butterfly dummies were exposed to at least nine insectivorous bird species. Do bird species differ in their behavioral response to eyespots? Do responses depend on the distance at which a bird first detects the butterfly? Do eyespots and camouflage markings present on the same animal both function, but at different distances (Tullberg et al. 2005)? Do bird responses vary depending on the particular light environment in the places and at the times when they encounter the butterfly (Kodandaramaiah 2011)? Answering these questions under natural, uncontrolled conditions will be challenging, requiring onerous methods, (e.g., video recording in multiple locations over time). The study indicates the interest of pursuing these questions.

References

Blut, C., Wilbrandt, J., Fels, D., Girgel, E.I., & Lunau, K. (2012). The ‘sparkle’ in fake eyes–the protective effect of mimic eyespots in Lepidoptera. Entomologia Experimentalis et Applicata, 143, 231-244. https://doi.org/10.1111/j.1570-7458.2012.01260.x

Chotard, A., Ledamoisel, J., Decamps, T., Herrel, A., Chaine, A.S., Llaurens, V., & Debat, V. (2022). Evidence of attack deflection suggests adaptive evolution of wing tails in butterflies. Proceedings of the Royal Society B, 289, 20220562. https://doi.org/10.1098/rspb.2022.0562

De Bona, S., Valkonen, J.K., López-Sepulcre, A., & Mappes, J. (2015). Predator mimicry, not conspicuousness, explains the efficacy of butterfly eyespots. Proceedings of the Royal Society B, 282, 1806. https://doi.org/10.1098/RSPB.2015.0202

Hossie, T.J., & Sherratt, T.N. (2012). Eyespots interact with body colour to protect caterpillar-like prey from avian predators. Animal Behaviour, 84, 167-173. https://doi.org/10.1016/j.anbehav.2012.04.027

Hossie, T.J., Sherratt, T.N., Janzen, D.H., & Hallwachs, W. (2013). An eyespot that “blinks”: an open and shut case of eye mimicry in Eumorpha caterpillars (Lepidoptera: Sphingidae). Journal of Natural History, 47, 2915-2926. https://doi.org/10.1080/00222933.2013.791935

Iserhard, C.A., Malta, S.T., Penz, C.M., Brenda Barbon Fraga; Camila Abel da Costa; Taiane Schwantz; & Kauane Maiara Bordin (2024). The large and central Caligo martia eyespot may reduce fatal attacks by birds : a case study supports the deflection hypothesis in nature. Zenodo, ver. 1 peer-reviewed and recommended by Peer Community in Ecology. https://doi.org/10.5281/zenodo.10980357

Janzen, D.H., Hallwachs, W., & Burns, J.M. (2010). A tropical horde of counterfeit predator eyes. Proceedings of the National Academy of Sciences, USA, 107, 11659-11665. https://doi.org/10.1073/pnas.0912122107

Kodandaramaiah, U. (2011). The evolutionary significance of butterfly eyespots. Behavioral Ecology, 22, 1264-1271. https://doi.org/10.1093/beheco/arr123

Monteiro, A. (2015). Origin, development, and evolution of butterfly eyespots. Annual Review of Entomology, 60, 253-271. https://doi.org/10.1146/annurev-ento-010814-020942

Monteiro, A., Glaser, G., Stockslager, S., Glansdorp, N., & Ramos, D. (2006). Comparative insights into questions of lepidopteran wing pattern homology. BMC Developmental Biology, 6, 1-13. https://doi.org/10.1186/1471-213X-6-52

Stevens, M. (2005). The role of eyespots as anti-predator mechanisms, principally demonstrated in the Lepidoptera. Biological Reviews, 80, 573–588. https://doi.org/10.1017/S1464793105006810

Stevens, M., Stubbins, C.L., & Hardman C.J. (2008). The anti-predator function of ‘eyespots’ on camouflaged and conspicuous prey. Behavioral Ecology and Sociobiology, 62, 1787-1793. https://doi.org/10.1007/s00265-008-0607-3

Tullberg, B.S., Merilaita, S., & Wiklund, C. (2005). Aposematism and crypsis combined as a result of distance dependence: functional versatility of the colour pattern in the swallowtail butterfly larva. Proceedings of the Royal Society B, 272, 1315-1321. https://doi.org/10.1098/rspb.2005.3079

Vallin, A., Jakobsson, S., Lind, J., & Wiklund, C. (2005). Prey survival by predator intimidation: an experimental study of peacock butterfly defence against blue tits. Proceedings of the Royal Society B, 272, 1203-1207. https://doi.org/10.1098/rspb.2004.3034

Along with the generalization of open-access practices, large, heterogeneous datasets are becoming increasingly available to ecologists (Farley et al. 2018). While such data offer exciting opportunities for unveiling original patterns and trends, they also raise new challenges regarding how to extract relevant information and actually improve our knowledge of complex ecological systems, beyond purely descriptive correlations (Dietze 2017, Farley et al. 2018).

In this work, Caumette et al. (2024) develop an original ecoinformatics approach to relate multi-scale environmental factors to the temporal dynamics of a major pest in mango orchards. Their method relies on the recent tree-boosting method GPBoost (Sigrist 2022) to hierarchize the influence of environmental factors of heterogeneous nature (e.g., orchard composition and management; landscape structure; climate) on the emergence date of the oriental fruit fly, Bactrocera dorsalis. As boosting methods allows the analysis of high-dimensional data, they are particularly adapted to the exploration of such datasets, to uncover unexpected, potentially complex dependencies between ecological dynamics and multiple environmental factors (Farley et al. 2018). In this article, Caumette et al. (2024) make a special effort to guide the reader step by step through their complex analysis pipeline to make it broadly understandable to the average ecologist, which is no small feat. I particularly welcome this commitment, as making new, cutting-edge analytical methods accessible to a large community of science practitioners with varying degrees of statistical or programming expertise is a major challenge for the future of quantitative ecology. 

The main result of Caumette et al. (2024) is that temperature and humidity conditions both at the local and regional scales are the main predictors of B. dorsalis emergence date, while orchard management practices seem to have relatively little influence. This suggests that favourable climatic conditions may allow the persistence of small populations of B. dorsalis over the dry season, which may then act as a propagule source for early re-infestations. However, as the authors explain, the resulting regression model is not designed for predictive purposes and should not at this stage be used for decision-making in pest management. Its main interest rather resides in identifying potential key factors favoring early infestations of B. dorsalis, and help focusing future experimental field studies on the most relevant levers for integrated pest management in mango orchards.

In a wider perspective, this work also provides a convincing proof-of-concept for the use of boosting methods to identify the most influential factors in large, multivariate datasets in a variety of ecological systems. It is also crucial to keep in mind that the current exponential growth in high-throughput environmental data (Lucivero 2020) could quickly come into conflict with the need to reduce the environmental footprint of research (Mariette et al. 2022). In this context, robust and accessible methods for extracting and exploiting all the information available in already existing datasets might prove essential to a sustainable pursuit of science.

References
 
Caumette C, Diatta P, Piry S, Chapuis M-P, Faye E, Sigrist F, Martin O, Papaïx J, Brévault T, Berthier K. 2024. Hierarchizing multi-scale environmental effects on agricultural pest population dynamics: a case study on the annual onset of Bactrocera dorsalis population growth in Senegalese orchards. bioRxiv 2023.11.10.566583, ver. 3 peer-reviewed and recommended by Peer Community in Ecology.  https://doi.org/10.1101/2023.11.10.566583

Dietze MC. 2017. Ecological Forecasting. Princeton University Press
 
Farley SS, Dawson A, Goring SJ, Williams JW. 2018. Situating Ecology as a Big-Data Science: Current Advances, Challenges, and Solutions. BioScience, 68, 563–576, https://doi.org/10.1093/biosci/biy068
 
Lucivero F. 2020. Big Data, Big Waste? A Reflection on the Environmental Sustainability of Big Data Initiatives. Science and Engineering Ethics 26, 1009–1030. https://doi.org/10.1007/s11948-019-00171-7

Mariette J, Blanchard O, Berné O, Aumont O, Carrey J, Ligozat A-L, Lellouch E, Roche P-E, Guennebaud G, Thanwerdas J, Bardou P, Salin G, Maigne E, Servan S, Ben-Ari T 2022. An open-source tool to assess the carbon footprint of research. Environmental Research: Infrastructure and Sustainability, 2022. https://dx.doi.org/10.1088/2634-4505/ac84a4
 
Sigrist F. 2022. Gaussian process boosting. The Journal of Machine Learning Research, 23, 10565-10610. https://jmlr.org/papers/v23/20-322.html
 

Alarm calls are an important acoustic signal that can decide the life or death of an individual. Many birds are able to vary their alarm calls to provide more accurate information on e.g. urgency or even the type of a threatening predator. According to the acoustic adaptation hypothesis, the habitat plays an important role too in how acoustic patterns get transmitted. This is of particular interest for range-expanding species that will face new environmental conditions along the leading edge. One could hypothesize that the alarm call repertoire of a species could increase in newly founded ranges to incorporate new habitats and threats individuals might face. Hence selection for a larger active vocabulary might be beneficial for new colonizers. Using the Great-Tailed Grackle (Quiscalus mexicanus) as a model species, Samantha Bowser from Arizona State University and Maggie MacPherson from Louisiana State University want to find out exactly that. 

The Great-Tailed Grackle is an appropriate species given its high vocal diversity. Also, the species consists of different subspecies that show range expansions along the northern range edge yet to a varying degree. Using vocal experiments and field recordings the researchers have a high potential to understand more about the acoustic adaptation hypothesis within a range dynamic process. 

Over the course of this assessment, the authors incorporated the comments made by two reviewers into a strong revision of their research plans. With that being said, the few additional comments made by one of the initial reviewers round up the current stage this interesting research project is in. 

To this end, I can only fully recommend the revised research plan and am much looking forward to the outcomes from the author’s experiments, modeling, and field data. With the suggestions being made at such an early stage I firmly believe that the final outcome will be highly interesting not only to an ornithological readership but to every ecologist and biogeographer interested in drivers of range dynamic processes.

References

Bowser, S., MacPherson, M. (2021). A study on the role of social information sharing leading to range expansion in songbirds with large vocal repertoires: Enhancing our understanding of the Great-Tailed Grackle (Quiscalus mexicanus) alarm call. In principle recommendation by PCI Ecology. https://doi.org/10.17605/OSF.IO/2UFJ5. Version 3