00:00Samuel Guchu is looking for birds.
00:05The ornithologist for Kenya's National Museums heads to the field with his usual tools,
00:11binoculars, a finely tuned ear, and lots of patience.
00:16There's a white-browed robin, Chad, that I'm seeing.
00:30You can hear it calling.
00:33But Samuel is not just here for a walk in the woods.
00:37He's helping to collect the data for the forest census.
00:41He looks for species that can help identify threats to the ecosystem.
00:46Today, he hopes to find a bird called Hatlaup's turaco.
00:51So birds like those are forest dependent because they do rely on indigenous forest systems
00:58for their food and breeding and foraging areas.
01:02So when we lose species like the Hatlaup's turaco,
01:06we know that the forest's health is in decline.
01:09And maybe we'll indicate that we need to do something about that
01:14in terms of maybe reduce interference or plant more indigenous fruiting trees in the area.
01:22But this data collection method takes a long time.
01:26Spectacled weaver.
01:28And that's calling.
01:31And when it comes to deciphering bird calls, even professionals make mistakes.
01:37Luckily, a local university has developed a device that can help researchers like Samuel.
01:43So this is an acoustic sensor that we developed at the Center for Data Science and Artificial Intelligence
01:49for audio data collections with the goal of monitoring ecosystems using sounds.
01:55We have a solar panel to charge the battery.
01:59Samuel is excited about this invention.
02:02He believes it can help him in his work.
02:06Even being out here for a couple of hours now, almost four hours,
02:10and not being able to see the bird itself
02:14can show you how important acoustic monitoring such as these devices,
02:22how important they are to ecological studies.
02:25Because it's not every single time you go to the field that you get a chance to see your study subject.
02:32Actually having acoustic monitoring programs like these that remove the human aspect
02:40can maybe give us a better picture of what the ecosystem looks like without human interference.
02:49Machine learning powers the technology as well as most artificial intelligence.
02:55Recorded audio is converted into visualized data,
02:58which is then analyzed and used to teach the system to recognize bird calls and other acoustic markers.
03:05The recorded and processed audio data from the field helps researchers and ecologists draw conclusions about a forest's health.
03:16This data can be very useful when it comes to environmental decision-making.
03:21When it comes to ecosystems such as forests, birds are a good indicator of forest health.
03:29When we are getting lists submitted from those areas that are seeing decline in reporting rate of this forest-specific species,
03:40we are able to see that although maybe the forest might not be completely destroyed,
03:45we are seeing a decline in that ecosystem health of that area that could probably lead to further destruction.
03:54This can help to launch reforestation projects in areas damaged by human activity.
04:01Like here in Kiyambu County's Kiyeni district, more than 30,000 indigenous trees were planted in a habitat degraded by charcoal production and agricultural encroachment.
04:15The trees are monitored and tracked over time.
04:20But manual monitoring is time-consuming, so this scientist uses a special tool, an AI-powered stereoscope.
04:31And we capture two images of trees, and it is possible from two images to sort of acquire the depth of the points on the scene,
04:39you know, just know how far the points are, how tall they are, how off the ground they are, and such things.
04:45So it gives us sort of a three-dimensional understanding of the scene in front of the camera.
04:49Cedric's machine-learning-based software, TreeVision, automatically measures tree height, width, and other features.
04:57The data collected can then be used to project the amount of carbon that can be stored by a growing patch or forest.
05:04Just like the designs for the bio-acoustic microphone, the code for Cedric's software is open-source,
05:10so other researchers across Africa and beyond can freely adapt it to help keep ecosystems intact.
Comments