12-09-2018

I would say, "We've all done it!" - but it's probably mostly just me.  Yes, once again I'm howling in anguish because I forgot that my camera trap uses a US time format (MM-DD-YYYY), rather than the more familiar (to me at least) DD-MM-YYYY format.  That means that photos from a camera trap I put out in early July, are all proudly stamped with the claim that they were taken at various times after the 7th of May!  If I was cataloguing and analysing my own camera trap photos, I'd have work-arounds for this error.  However, because MammalWeb takes information directly from the image EXIF data (Exchangeable image file format data, for those of you who care), I can't upload the images until I've sorted out the dodgy "Photo taken" dates and times.  This problem has been hanging over me for a while but, today, I decided to look into a solution.  Happily, I found one!

Owners of Macs (myself included) and users of Linux might not like me for this, but there's currently a bit of freeware available for PCs, called "Exif Date Changer".  There's probably something similar for Macs - but I have yet to find it.  If you do find one, please let us know (This email address is being protected from spambots. You need JavaScript enabled to view it.).  In the meantime, I'm very happy to report that the free version of Exif Date Changer is spectacularly straightforward and efficient - so much so, that it's quicker for me to go and find a PC to use it on, than to spend my time looking for alternatives that I can use on my own desktop.

The software allows you to choose the source folder containing the files you want to correct, as well as the output folder in which you want to store the corrected files.  That's particularly useful if - like me - it takes you a while to work out the difference between two dates (so you don't want to overwrite the originals until you've checked that you're correct).  There are a variety of options for changing the dates and times - but the most straightforward is just to set the numbers of days, hours, minutes and seconds by which you were out when you set the camera.  One thing to check is the "options" tab, which might suggest renaming the images.  You can easily change that, if you'd prefer that it left the names unchanged.  Of course, if you have a camera trap that puts the date and time on each photo, that will still be in error - but at least the dates and times recorded in the database will be correct!

Well, I'm pleased to have found this simple solution - and I hope that some of you find it useful too.

28-08-2018

Many of you will have been pleased to see the news, a couple of weeks ago, that pine marten appear to be spreading into Northumberland (e.g., see this article in The Guardian). The recovery of pine marten in England is an exciting development for nature enthusiasts, and could help to alleviate pressure on our native red squirrels.

We're really excited to announce a new partnership with NatureSpy, a non-profit organisation that - like MammalWeb - aims to engage people with the wildlife around them, and to research and conserve wildlife and habitats through study and monitoring. NatureSpy run a variety of projects and, at present, they are organising an intensive, 3-year project to study pine marten in the North York Moors. With the aid of volunteers, they are currently rotating camera traps around a variety of sites in the North York Moors, and a proportion of those traps are set to record still images. The traps yield large numbers of images and those will help to inform NatureSpy about the ecology of the area, aiding the development of a long-term conservation plan for pine marten in the area. If we're lucky, we might even see a picture or two of a pine marten!

If you are registered as a Spotter, the NatureSpy images will be available to classify by selecting "Classify All" from your Spotter home screen. To focus your efforts on helping the NatureSpy team, you can select "NatureSpy - North York Moors" and choose "Classify Selected Project Only".

You can also read a post about this project on the NatureSpy blog, here.

Much of the monitoring conducted by NatureSpy (as well as many other individuals and organisations) uses video, rather than still images. We know that a capacity to handle video (rather than just image sequences) is an aspect of the platform that many of you would like to see us develop. Over coming months, we will be looking for ways to resource this (both the development and the subsequent hosting of large video files); we'll keep you posted regarding progress!

19-07-2018

We (the project team, plus all of our participating volunteers) have published our first paper based on MammalWeb!  The paper is in the journal Remote Sensing in Ecology and Conservation, and can be found here.  When we think of Remote Sensing, we often think of satellite observations (a source of data that can be remarkably handy for ecologists, as well as many land managers, plus other scientists and social scientists).  However, Remote Sensing can cover many types of data collection for which the presence of people is not necessary at the time of data acquisition.  Camera trapping is, thus, a good example of a remote sensing approach.

A pre-requisite for making use of the data collected by all of MammalWeb's participants is knowing what's in the images.  By comparing user-submitted classifications to a set of over 10,000 image sequences that have been looked at by "experts" to determine their subjects, we can learn about how confident we can be about those classifications.  Happily, it turns out that participants are very good at identifying what's in an image sequence.  We need lots of examples to get a good idea of how accurate participants usually are, so we focused on 16 of the most commonly-occurring species (or species designations).  When those species occurred in a sequence, we looked at all the submitted classifications and asked how many of them correctly said that the species was pictured.  For almost all of the commonly-occurring species, that figure was 80% or more.  For several species, it was closer to 95%!  Given the risk, with all trail cameras, of getting blurry, dark, partial, or otherwise indistinct images, we think this is a real testament to the skill and dedication of those using the website!

Fig. 1. In general, spotters show high levels of accuracy.

Two "species" for which accuracy is notably lower are small rodents and brown hares.  Further analyses looking at the inaccurate classifications in those cases show that they arise for different reasons.  In particular, small rodents are often overlooked altogether, because they are often visible only from their eye-shine.  We hope that our site developments (especially the new approach to being able to move backwards and forwards through a sequence, which can help to identify any sort of movement) will reduce this problem.  By contrast, brown hares show lower accuracy than most other species because they are often misidentified as rabbits.  This is understandable, given their many similarities.  However, we would encourage anyone who is uncertain to take a look at web tutorials, such as this page or this one.

Fig. 2. Reasons for incorrect classifications by species or species group.  Where blue predominates, the species is more often missed than misclassified; green indicates the proportion of errors that arise from misclassification.

Clearly, based on these numbers alone, we can be much more confident of classifications regarding some species than others.  This might mean that we could remove many image sequences, allowing users to focus on the less distinct, the more unusual, or those that prove harder to reach consensus about.  However, that might result in a far less rewarding time for Spotters.  At present, we are keeping a close eye on techniques for automatic image analysis.  These might, at the very least, allow us to remove sequences very likely to contain no wildlife, which - based on comments received - would probably be very popular with many of our contributors!

Toby Atkinson-Coyle recently graduated from the biology department at Durham University: congratulations Toby! For his final year project, Toby worked with the MammalWeb team on a camera trapping project looking at the impact of urbanisation on mammals in Durham. Here, he writes about some of the findings from his project:

For my masters’ project, I used distance sampling with camera traps to assess the impact of urbanisation on mammals in Durham, UK. As discussed in previous posts, this method involves measuring the distance of animals from the camera which can then be used to calculate their density. I deployed cameras in both rural and urban habitats across a 3x4km area of Durham from 12th February through to 9th March 2018.

Over the study period, a total of 2,646 images of wild mammals were obtained, 1,285 in rural and 1,361 in urban habitats (see Figure 1). Six species were regularly detected in both habitats, including European rabbit, European badger, red fox, grey squirrel, roe deer and small rodents. European rabbit and grey squirrel were the most frequently photographed species, so I report on them in greater detail. To assess the overall impact of urbanisation on mammals, I was also able to calculate broader community-level metrics such as species richness.

Figure 1 A subset of mammals photographed in Durham, UK.

Species varied in density between habitats. Four out of the six detected mammal species decreased from rural to urban habitats whilst two increased. The largest differences in density were in rabbit and grey squirrel populations. Rabbit density in urban habitats was twice that in rural ones (see Figure 2A). Conversely, squirrel density was much greater in rural habitats (see Figure 2B).

Figure 2 Estimated density per km² of European rabbit and grey squirrel populations in rural and urban habitats.

I calculated species richness per deployment by counting the number of different species detected by each camera. Species richness was higher in rural habitats with an average of 1.60 (0.29 SE) species detected per deployment whilst in urban only 1.12 (0.23 SE) species were detected (see Figure 3).

Figure 3 Species richness per deployment of mammals in rural and urban habitats.

Urbanisation has a range of impacts on mammals at both the species and community levels. Broader ecological metrics indicate an overall decline of mammals with urbanisation but, at the species level, some appeared to benefit. Therefore, it is important to consider a range of species and community-level metrics when assessing the response of wildlife to disturbance.

09-07-2018

Many of you have asked us to increase the range of species that you can choose from when spotting. We will soon be implementing changes to give this much greater flexibility, without forcing you to wade through long species lists each time you classify what’s in an image sequence. Specifically, when you are spotting, you should soon see something like the screenshot below. At the top will be 3 or more “filters”, which will allow you to navigate rapidly between lists of common species, and full lists of either mammals or birds. In some cases, there might also be a list of species that are most likely to turn up among images associated with the project you are Spotting. For the vast majority of image sequences you look at, the pictured species will be in the “Common” list (which will usually be the default display). If you are lucky, however, sometimes you might see an unusual species. In those cases, you can click on the “Mammals” or “Birds” filters to find what you are looking for. Those lists are arranged alphabetically and you can use the page selectors to jump to a different page of the list. Bear in mind that it might take a little while before you find those rarer species in the lists. For example, if you are looking for a mink under “M”, you might need to check whether we have discriminated between mink species (in which case, the American mink will be under “A”). For some species, we have also retained less certain designations for when you know what sort of animal it is but you can’t identify it to species-level. For example, you can find “Unidentified bird” under “U” in the birds list, and "Small rodent (unknown species)" under “S” in the mammals list. If possible, however, do identify to species level.

We are particularly grateful to the Economic and Social Research Council’s Impact Acceleration Account for funding this (and some of our other recent developments) and to Helen for making our vague notions a useful reality! We’d welcome your feedback on this new approach once it’s available. Also, do let us know if we have overlooked any species, or if you think there are others that would be a higher priority for the “Common” species filters. As always, you can send suggestions to This email address is being protected from spambots. You need JavaScript enabled to view it..