Thermal camera 5-20 times more sensitive for rats

Authors: David Blake and Grant Ryan

David Blake is a contributor to the Cacophony Project and has been running an experiment to test the relative sensitivity of Cacophony Project thermal cameras to off-the-shelf trail cameras for detection of rats. It is well documented that trail cameras perform much better than other detection tools like tracking tunnels and chew cards. In this experiment we are interested in comparing the performance Cacophony Project's thermal cameras to what is currently recognised as the best available tool.

The test setup is pictured below. Circled in red is a large rat trap attached to a tree that rats are known to use. This trap is as open as possible to try to increase the likelihood of rat interactions but we know this is not a viable approach at scale because it makes unwanted bycatch more likely. Immediately below the rat trap is a standard trail camera that is about 1.5 meters away from the rat trap. About 5 meters back is the thermal camera (gray box) and another trail camera beside that on the wooden post. There is also a bait station that can be seen just above the wooden post.

Rat detection experimental setup

This trial was conducted in Auckland in early May 2019. This is a time of year when we typically see some of the highest interactions with food based lures. It is also notable that the rats in the area are at the large end of the spectrum. A recently caught rat measured 440mm head to tail….. This creates a best case scenario for monitoring rats as they are large and probably very attracted to food based lures.

For one week, the different animal visits on each device were measured. Unfortunately the far away trail camera didn’t have the correct date and time set so its recordings couldn’t be line up to the recordings from the other cameras. The close trial camera was set for videos with the far trail camera configured for a mixture of pictures and videos. Below is a table showing the results for the three cameras.

Date Time Predator Heat Camera Close trail camera - saw predator Close trail camera - triggered but no predator Far trail camera Rats caught
28/04/2019 00:55:34 rat 1
28/04/2019 01:04:26 rat 1
28/04/2019 01:17:42 rat 1 1
29/04/2019 05:20:11 rat 1
01/05/2019 02:17:20 rat 1
01/05/2019 02:37:42 rat 1 1
01/05/2019 03:26:24 rat 1 1
01/05/2019 03:39:04 rat 1
01/05/2019 03:44:56 rat 1
02/05/2019 00:41:29 rat 1 1
04/05/2019 03:22:00 rat 1
04/05/2019 14:53:34 rat 1
04/05/2019 22:20:05 rat 1
05/05/2019 01:25:42 rat 1
05/05/2019 01:54:10 rat 1
05/05/2019 02:30:13 rat 1
05/05/2019 03:04:29 rat 1
05/05/2019 03:20:25 rat 1
30/04/2019 11:08:32 rat 1
Total 19 1 3 4 0

A visit was determined by looking through all the videos and if there were multiple videos around the same time it counted as only one visit. For example, there may have been 5 videos all around the same time so that is counted as one visit. The thermal camera also saw 11 cats and 9 hedgehogs but it is rats that are of primary interest for this study. The far camera recorded most of the cats but none of the hedgehogs so the data from the far camera is a little questionable and we will try to repeat the test with the far camera set just to videos.

The close trail camera triggered 4 times but only during one visit was a rat caught on the video. These cameras have a delay in start up to conserve power and it shows they are often too slow to wake up from low power mode.

Below is a video that shows all of the interactions on different cameras. While the trail camera videos have higher definition it clearly shows how the thermal camera is more sensitive at detection. Most of the heat videos are shown 10 times real speed to give an idea of just how much activity there was that was missed by the trail camera.

When rats did go in front of the trail camera they often missed because they either didn’t trigger or they were too slow at “waking up”. It is noticeable how hard it can be to see the rats in the 4 photos taken from the far trail camera - there were no trail camera videos taken of the rats from the far video apart from the one in a cats mouth at the end of the video.

The larger number of cases where the rat was visible in the general area but did not go to the lure on the tree so the close trail camera missed it. The trail camera view area for the close one is probably about four square meters but the thermal camera covers around 36 square meters. Previous tests show that the heat camera triggers at about 3 times the distance so with wide view has the potential to cover about 10 times the area (3 x 3).

This is obviously a small sample but the results seem consistent with what we would expect for cameras designed for large mammals. Previous testing showed the thermal camera is 3.5 times more sensitive for possums. For smaller animals we would expect it to work less well and this particular set up is probably a best case for rat detection given the size of them and the time of the year.

The table below summarises how much better the performance of the thermal camera is compared to the trail cameras. Between 5-20 times better at detection and vastly better at recording all animal behaviour. For the far camera, which was not configured to take videos, we assume that a video of 1 second was generated for each recording so that a comparison could be made.

Thermal Camera Trail Camera Times better
Detections - close camera 19 1 19
Detections - far camera 19 4 5
Video length close camera (seconds) 1524 6 254
Video length camera (seconds) 1524 1 1524

It could be said that a trail camera left out long enough will eventually detect the rat or if you have more trail cameras in the same area it will also eventually make a detection. Detection devices and trap effectiveness should be measured in terms of performance based on an area and time. What is the detection rate per hectare month for example? More sensitive detection devices can then be left out for shorter periods of time or in areas with lower predator densities to get the same detection rate. Therefore a higher priced, high sensitivity device can also end up being lower cost overall.

For applications where detection is essential like reinvasion, virtual fences or understanding animal behaviour then the high sensitivity is a key requirement. You can’t understand what is going on if your eyes are essentially closed 80% of the time. If you are interested in what is actually happening with predators in the bush then the Cacophony Project thermal camera gives much more information. More importantly we can link this sensitive detection device to intelligent catching/luring devices to make trapping dramatically more effective.