Insect Inspection

If you want to know something about the health of an ecosystem, the first thing you should do is take a look at who lives there. Variations in animal populations over time can indicate that environmental conditions of a region are changing in important ways. Whether the population of a species is dwindling or exploding, or animals that are not normally seen in a particular area start showing up, researchers can glean many insights to help drive a well-informed conservation effort.

Most people would probably assume that the most informative species to monitor would be perhaps birds, rodents, deer, or an apex predator. But insects make up half of all known species on Earth, and they can give us a lot of information. Because they can be observed by the thousands, their presence and activities can be easily correlated with other environmental data. Furthermore, their short lifespans and ranges offer many hyper-localized insights that are otherwise very difficult to obtain.

For these reasons, Digital Naturalism Laboratories developed a tool called Mothbox that can be deployed deep into the wilderness to autonomously monitor the populations and activities of insects. The energy-efficient and low-cost device was designed with easy-to-use, off-the-shelf components so that nearly anyone can build a Mothbox and so that they can be deployed far and wide.

The system is built around a Raspberry Pi 5 computer (a Raspberry Pi 4 is also acceptable). A high-resolution, 64-megapixel camera is used to capture images of insects, along with photography lights for nighttime illumination. There are also ultraviolet lights included in the design to attract insects to the Mothbox. The entire system is powered by a large, rechargeable battery, and the components are fitted into a case to protect them from the elements.

Software for the build is provided as a disk image that can be flashed to an SD card via the Raspberry Pi Imager. Included in the pipeline is a YOLO v8 object detection algorithm that was trained to recognize insects in captured images and crop those segments out. Another computer vision algorithm, BioCLIP, then identifies the insects and groups them according to their taxonomic classification. Mothbox runs in an ultra-low power state during the day to conserve the batteries, then wakes up at night to collect data when the insects come out in full force.

In total, the materials required to build a Mothbox cost about $375. A typical monitoring solution of this sort can cost up to $15,000, so that is quite a bargain. That does not factor in the time it will take to build the device, however, which appears to be quite substantial and requires some specialized equipment. The write-up is very extensive, and walks through everything step-by-step, so it should be possible to get one up and running successfully even if it takes a considerable amount of time.