Mammal Web 1st year Project Update and Future Plans

Vivien Kent Mammal Web, News, Uncategorized


The Spring Equinox on 20th March heralded the official start of Spring and activity on the cameras is definitely increasing, with hedgehogs reappearing and lots of animals chasing each other around. The clocks went forward to British Summer Time at the end of March too. However, can I remind all Trappers to please leave the clock on your camera(s) set to GMT. We have decided to do this to reduce the potential for confusion. This does mean that when you are uploading your images to the website you will need to enter your deployment times in GMT as well!

This month you can read a comprehensive project update from the whole MammalWeb team including future plans from Durham University below.

The First Year of MammalWeb

It is just over a year since we initiated this phase of the MammalWeb project. It seemed a good time to post the first of what we intend will become regular blog posts to update you on what you have achieved, how the project is developing and what our plans are for future developments.

The first thing to say is a massive THANK YOU to everyone who has deployed cameras, uploaded images, or classified images.  We have been really excited by the dedication of so many of you, which has already yielded useful ecological information and promises so much more for the future.

What have you achieved?

Since 28 March 2015, 54 trappers have deployed cameras at 129 sites for a total of 6,638 days. Sites now range from North Yorkshire to Northumberland, and include locations in the North York Moors National Park and the North Pennines Area of Outstanding Natural Beauty (Fig. 1).

Fig. 1: Map showing the locations of trapping sites.

Fig. 1: Map showing the locations of trapping sites.

 

Most deployments (periods of time between setting the camera and uploading) have been short term (less than 10 days) but some cameras have been deployed for 2 to 4 months, as shown (Fig. 2).

Fig. 2: Periods of time for which cameras have been deployed between uploads. The height of the bars indicates the number of camera deployments of the duration given on the horizontal axis (1-10 days, 11-20 days, etc.).

Fig. 2: Periods of time for which cameras have been deployed between uploads. The height of the bars indicates the number of camera deployments of the duration given on the horizontal axis (1-10 days, 11-20 days, etc.).

Trappers have uploaded 46,250 photos and, so far, 129 spotters have contributed to classifying the content of those images. To date, this has led to 90,642 classifications of 32,782 images.

The accumulation of photos has been fairly steady since August but spotting effort is rather more sporadic (Fig. 3). The impact of our site upgrade in early November can easily be seen.

Fig. 3: Number of photos captured by camera traps since the first deployments in April 2015 (grey). The black area shows the accumulation of unique photos that have been classified by spotters. Obviously, many photos have been classified multiple times, which is why there are many more classifications (90,642) than photos classified (32,782).

Fig. 3: Number of photos captured by camera traps since the first deployments in April 2015 (grey). The black area shows the accumulation of unique photos that have been classified by spotters. Obviously, many photos have been classified multiple times, which is why there are many more classifications (90,642) than photos classified (32,782).

 

Spotting effort has been quite varied among individuals, as shown (Fig. 4). Although most spotters have classified between 1 and 2,000 photos, a few heroic individuals have classified many more (17,185 in one case!). 19 spotters have classified the contents of at least 1,000 photos.

 

Fig. 4: Spotting effort by numbers of individuals. The height of the bars indicates the number of spotters who have classified the number of photos given on the horizontal axis (1-2,000 photos, 2,001-4,000 photos, etc.).

Fig. 4: Spotting effort by numbers of individuals. The height of the bars indicates the number of spotters who have classified the number of photos given on the horizontal axis (1-2,000 photos, 2,001-4,000 photos, etc.).

 

Importantly, because we have multiple classifications for many images (up to 20 in a few cases), we can start to look at how much agreement we need before we can be sure about what’s in a photo.  We are working on that at present.  Once we can treat the content of photos with confidence, we’ll be able to start reporting back on what they are showing.

One notable outcome already is that a raccoon was identified in Sunderland (Fig. 5).  Raccoons are charming and intelligent – but as invasive species, they can cause real problems.  They were introduced to Europe as domestic pets but are now quite widespread.  They carry diseases harmful to humans and other wildlife and are also extremely adaptable foragers.  They can cause problems for birds, as they have quite an appetite for eggs.  As charming as they are, therefore, it is better that they do not invade the UK.  The individual identified via MammalWeb has since been live-trapped and removed to Cleethorpes Zoo.

Fig. 5: One of several photos captured of a raccoon living in the Sunderland area.

Fig. 5: One of several photos captured of a raccoon living in the Sunderland area.

 

What do we have planned?

Now that we have a year’s worth of data from which to start to pull out patterns in trapping rates over time, we are pushing the project forward on several fronts.  Four particular developments that we hope will improve your experience of interacting with MammalWeb are as follows.

  1. We hope to include a ‘back to start of sequence’ button. Sometimes, to work out what species features in a sequence of images, it is necessary to go forward through the sequence quite a long way until a good, clear image is found.  At that point, it is possible to go back and classify earlier images as containing that species – but it can be annoying to go back one image at a time.  We hope, soon, to add a button that will return users directly to the start of the sequence.
  2. We hope to streamline the classification of blank images. Cameras vary in their sensitivity but even the most expensive camera traps are sometimes triggered when no animal is in shot.  When classifying images, it can become boring if a large proportion of those images have no animal in them.  When we look at the association between numbers of classifications and our confidence in what is pictured, we will consider these ‘blank’ images as a special case.  Ideally, we would be able to set a lower threshold at which those images can be removed from the pool of photos awaiting classifications.  That would reduce the number of blanks encountered by any spotter.  We are also investigating methods to weed out blanks automatically – but that is a tougher project!
  3. We hope to put better data validation procedures in place for Trappers. Some images are unusable to us, because we don’t have accurate data for either the site or the upload.  Sites should be associated with a grid reference (usually 2 letters followed by 2 groups of 3 digits, or 2 groups of 4 digits).  Initially, most grid references were recorded as 2 letters followed by 2groups of 2 digits.  This is ok, but if you can check your site details and, if possible, re-enter them (by typing, or by clicking on the map) to give 2 letters followed by 2 groups of 3 digits, that would be enormously helpful.  Likewise, if you can check that the site has an informative name (ideally not “[New site]”) and appropriate characteristics (like the habitat type), that would be very helpful.  Finally, if you can double-check the deployment and collection times before you upload, that would be great.  As we develop the site, we will try to put checks in place to ensure that all of that information is valid before you proceed.
  4. We hope to improve user feedback radically. We have held preliminary discussions with a web development firm who specialize in data visualization.  That firm is interested in helping us to develop the MammalWeb site to make it easy for trappers and spotters to see the patterns emerging from the underlying data, and even to ask your own questions of the data.  We will keep you updated on developments.

Some favourite photos so far

Here are just a few of the photos that users have flagged with the ‘Like’ button.  We hope to include access to these via the interactive part of the website in due course.

M2E38L128-128R399B307

0331:082915:07C:FLASS VALE:4

138a729458c4ff8b78352e3f4de50be3

b72eb6e62493654b1f8fe1e54183f923

f67adf918f99b3e692d1df3977000c3b

If you don’t currently contribute to MammalWeb (either as a trapper or a spotter) but are interested in doing so, please see http://www.mammalweb.org, or contact vkent@durhamwt.co.uk.