MammalWeb's findings

Many people are interested in the wildlife around them and many use camera traps for a variety of purposes. Ultimately, MammalWeb is intended to harness the efforts of those individuals and to ensure that, regardless of the purpose for which camera traps are deployed, the data don't just remain on individual SD cards and hard drives, but contribute towards a wider effort to document our wild mammals (and any other species of mammal and bird that are caught in images and videos).

Collating data in this way is not ideally suited to answering one or two specific science questions but, over time, it helps us to build up a picture of mammal distributions and what affects those, as well as mammal activity and its drivers. Moreover, with careful analysis, many specific questions can be answered along the way. These include methodological questions about citizen science, engagement and survey techniques, ecological questions about species associations, activity and occurrence, and management questions about the occurrence and whereabouts of non-native species, pest species or species of conservation concern. Moreover, you never know when routine monitoring might reveal a pattern that you weren't expecting to find, hinting at some previously undetected process.

The initial focus must be on how to build engagement with the project and how to interpret the data. The outputs and links below give some idea of the progress we are making, and the findings being generated along the way. All of the links should be accessible but please let us know (at This email address is being protected from spambots. You need JavaScript enabled to view it.) if you cannot access an article of interest.

Research articles from the MammalWeb team

This piece, led by Pen during his PhD research, describes the process and outcomes of outreach work focused around using the MammalWeb platform in a secondary school. It features school pupils and MammalWeb citizen scientists as co-authors and establishes some of the wider benefits that can arise from taking MammalWeb into schools. The journal's style prevented group authorship being attributed to volunteers, specifically, but accepted that "the MammalWeb citizen science project" could be credited.

This is a review that focuses on how citizen science can contribute to camera trapping image data processing, and how - given the constraints on that processing - deep learning (or "artificial intelligence") might be integrated with citizen scientists' efforts to make data processing more efficient. The research did not draw on MammalWeb data.

This piece, also led by Pen during his PhD, was a preliminary exploration of how classifications of image data submitted by citizen scientists can be interpreted, and how they affect the confidence we can have in what species feature in an image sequence. Pen showed that contributors have generally high accuracy but that, nonetheless, it can take from 7 to 9 classifications per sequence to have high confidence in what is featured. On the face of it, this finding is bleak. However, species-specific assessments show that most species can be determined with confidence much more rapidly (typically, with only a few classifications). Difficulties arise in classifying sequences featuring small rodents, as these are often mistaken for 'empty' sequences. This is less of a worry, as small rodents are not really the focus of camera trapping. Nevertheless, with better and more efficient approaches to image classification, we will be able to make more of the data, more rapidly.