Occupancy, density and the ecology of terrestrial British mammals

Mammals include species of ecological, economic and cultural importance. Determining the factors that drive their abundance and distribution, and developing effective management, rely on sustained and widespread monitoring. Britain boasts a rich history of detailed and informative studies of specific mammal populations. Over the same period, however, effective monitoring of the wide range of terrestrial mammal species has been lacking, exacerbated by the fact that many species occur at low densities, and are nocturnal or otherwise elusive. Consequently, calls for improved monitoring to underpin conservation and management have a long history.

Recent years have brought some notable declines in British mammals, including hedgehogs, weasels and wildcats. At the same time, invasive non-native species, including mammals such as muntjac and grey squirrel, continue to expand, causing problems for native species. Constant vigilance is required to avoid further invasions; the presence of the invasive greater white-toothed shrew, recently identified in north-east England, highlights the importance of improved monitoring systems.

In light of these concerns, the MammalWeb project was set up to encourage citizen scientists to engage in mammal monitoring by contributing camera trap images and associated metadata to a growing national database, and by helping with the task of classifying the resultant images. In this way, MammalWeb has assembled a database of millions of photos, videos and image classifications, from thousands of camera trap placements around Britain. This is a powerful resource but inference from these data is hampered by three issues: (i) long lags between data submission and data classification; (ii) known and unknown biases in camera placements; and (iii) the complex nature of the data, which demand elaborate analyses for robust inference. This project will tackle all three of these problems to produce timely and robust inferences with which to underpin insights for ecology, conservation and management.

Aims: (1) to use an array of camera traps to collect data on the occupancy and density of British mammals, calibrating, validating and improving inferences from citizen-led data collection; (2) to utilise recently-developed and high-performance AI models for image classification to develop efficient and prioritised workflows for robust image classification; (3) to analyse classified data from the MammalWeb dataset to answer questions regarding the occupancy and activity of British mammals, as well as their natural and anthropogenic drivers; and (4) to work with our end-user partners to showcase the use of ecological inferences to underpin management and policy.

Click on an image to expand

Image Captions

Figure 1: Imperilled species captured by MammalWeb contributors include the weasel (A) and the wildcat (B). Invasive non-natives include the grey squirrel (C) and the muntjac (D).,Figure 2: One challenge is combining classifications submitted via the MammalWeb portal (upper image) with those derived from machine learning models (lower image).


Methods will include deploying and calibrating camera traps, using cutting-edge analytical techniques to infer occupancy and density of selected mammal species; working with partners to classify image contents using bespoke machine learning (convolutional neural networks -CNNs) for British and European mammals, improving training for underrepresented species, and modifying classification algorithms on MammalWeb in light of the outputs; and hierarchical Bayesian modelling to derive insights into the drivers of occupancy and activity, including multi-species occupancy modelling. Training will be provided by all supervisors and, in addition, we anticipate a placement at the National Wildlife Management Centre (APHA), liaising directly with end-users of the data.

Project Timeline

Year 1

Learning techniques for camera trap use and analysis, including methods for occupancy and density analyses. Setting up camera trap studies. Working on inferences from existing AI classification approaches.

Year 2

Learning methods for hierarchical Bayesian modelling and applying it to existing datasets. Analysing the outcomes of the field data from year 1. Additional training of AI models for underrepresented species.

Year 3

Applying hierarchical Bayesian models to existing and newly-collected datasets. Finalising work-flows for classification algorithms that combine human and AI classifications. Placement with end-user organisation and exploration of management and policy uses of data. Commence writing thesis and drafting initial publications.

Year 3.5

Finalising thesis and submission of initial publications to journals.

& Skills

Full training will be given on camera trapping, viewshed methods for analysis, occupancy modelling, harnessing the outputs of AI models with statistical integration of human and machine-learned classifications, and hierarchical Bayesian modelling.

References & further reading

The MammalWeb platform: https://www.mammalweb.org/index.php/en/

Bowler, D.E., Nilsen, E.B., Bischof, R., O’Hara, R.B., Yu, T. T., Myint Aung, Y. & Linnell, J.D.C. (2019) Integrating data from different survey types for population monitoring of an endangered species: the Eld’s deer. Scientific Reports. 9, 7766. https://doi.org/10.1038/s41598-019-44075-9

Green, S.E., Stephens, P.A., Whittingham, M.J & Hill, R.A. (2022) Camera trapping with photos and videos: implications for ecology and citizen science. Remote Sensing in Ecology and Conservation. https://doi.org/10.1002/rse2.309

Hsing, P.-Y., Bradley, S.P., Kent, V.T., Hill, R.A., Smith, G.C., Whittingham, M.J., Cokill, J., Crawley, D., MammalWeb Volunteers, Stephens, P.A. (2019) Economical crowdsourcing for camera trap image classification. Remote Sensing in Ecology and Conservation. 4, 361-374. https://doi.org/10.1002/rse2.84

Hsing, P.-Y., Hill, R.A., Smith, G.C., Bradley, S.P., Green, S.E., Kent, V.T., Mason, S.S., Rees, J.P., Whittingham, M.J., Cokill, J. & Stephens, P.A. (2022) Large-scale mammal monitoring: the potential of a citizen science camera-trapping project in the UK. Ecological Solutions and Evidence, https://doi.org/10.1002/2688-8319.12180

Mason, S.S., Hill, R.A., Whittingham, M.J., Cokill, M.J., Smith, G.C. & Stephens, P. (2022) Camera trap distance sampling for mammal population monitoring: lessons learnt from a UK case study. Remote Sensing in Ecology and Conservation 8: 717–730. https://doi.org/10.1002/rse2.272

Rota, C.T., Ferreira, M.A.R., Kays, R.W, Forrester, T.D., Kalies, E.L., McShea, W.J., Parsons, A.W. & Millspaugh, J. J. (2016). A multispecies occupancy model for two or more interacting species. Methods in Ecology & Evolution, 7, 1164-1173. https://doi.org/10.1111/2041-210X.12587

Twining, J.P. et al. (2022) A comparison of density estimation methods for monitoring marked and unmarked animal populations. Ecosphere https://doi.org/10.1002/ecs2.4165

Apply Now