Snippets in Clean Technology and Data Science: Wildlife and Ecosystems

 A fun and exciting area to use data science in clean tech is in monitoring wildlife! This sector uses a combination of computer vision, remote sensing, artificial intelligence among other tools to help us track wildlife across the world.

Here are a few examples of the kinds of problems and technologies that are in play these days!

First, An interesting study came out of California recently, where scientists from the University of Delaware, University of California at Davis and the US Geological Survey partnered to track the movement of waterfowl in the region.

My first question when I read the study was – why would people, other than wildlife biologists and conservationists, care about what happened to waterfowl?

And it turns out that the answer is really important from the perspective of the agricultural industry as well as from a human health perspective. Looking at how close or how far away waterfowl are from poultry farms can help track the spread of avian influenza or “bird flu”. Every time there’s an outbreak of bird flu, thousands of birds are killed and there’s usually a health scare about how it could spread to people and result in increased illness.

What the scientists did was use weather radar to track the birds’ characteristic take-off and landing patterns and then correlate those with the locations from which the patterns were observed. Data are collected at 5-10 minute intervals every day, which gives the researchers a large dataset in both space and time on how the birds are moving and where they are most likely to be concentrated.

And this kind of information can be combined with other environmental datasets, like the presence of water bodies, to help identify the type of environment as well as conservation measures that would be most useful in increasing the populations of endangered species.

Second, facial recognition for monkeys!

Most people, when thinking about facial recognition algorithms, associate them with airports, security cameras, movies like Minority Report… They would probably not think about monkeys and clean technology!

But that’s just where facial recognition has found one of the most interesting applications in wildlife monitoring and data science today. Monitoring animals in the wild is a challenging problem – certain animals are tagged with sensors and followed over months and years to understand animal behavior as well as the problems they are facing in the environment today. Now physically tagging the animals takes a lot of effort and isn’t always successful – not to mention, you might lose the animal that you have tagged to predation and death.

This is where a collaboration between computer scientists and wildlife biologists came up with something really interesting. They built a facial recognition algorithm called LemurFaceID to try and identify the lemurs of Madagascar, one of the species of monkeys found there. The algorithm is based on a patch-wise Multiscale Local Binary Pattern features analysis and could identify individual lemurs 98% of the time.

The algorithm was based off insights from human facial recognition – that the eyes are the most important feature for primates in general and adopted a number of techniques to smooth out and identify features like extra hair or patterns in the fur of the lemurs.

Now facial recognition is a subset of the field of image processing, which is one of the key fields that data scientists are employed in. Image processing is how photos in Pinterest or Facebook get tagged. It’s a field that’s been used in clean technology when looking at satellite images, but it’s only recently that all these other applications have started to come out.

The algorithm built here depended on having expertise in both data science and clean technology. Knowing the features of lemurs and what would be useful for an algorithm was key but at the same time knowledge of what image processing algorithms could be used and their strengths and weaknesses was essential.

What our community are reading

Moonshots, Models, IoT and Machine Learning in Agriculture

Our online community space is now live!

How much water should an email consume? Data centers and water use