Latest / INSIGHT TODAY / This satellite can see inside forests

Transcript

This satellite can see inside forests · INSIGHT TODAY

• 0:00
  Speaker 1: What if we could measure every leaf on every branch
• 0:03
  Speaker 1: of every tree in the world's densest jungles. Until recently
• 0:07
  Speaker 1: that seemed impossible, But now scientists are using advanced technologies
• 0:11
  Speaker 1: to build these hyper detailed three D maps of the
• 0:14
  Speaker 1: world's forests. This is how they're measuring trees in extraordinary detail.
• 0:21
  Speaker 1: Scientists can measure how much carbon a tree holds by
• 0:24
  Speaker 1: measuring its biomass, and that's made up of different parts
• 0:28
  Speaker 1: like its bark and its leaves. And in the past,
• 0:30
  Speaker 1: the instruments they used to do this weren't that much
• 0:32
  Speaker 1: more sophisticated than this tape measure. But new technologies are
• 0:38
  Speaker 1: helping the world's ecologists measure forests in much greater detail.
• 0:42
  Speaker 2: The rainforest is a vast store of carbon. These are
• 0:46
  Speaker 2: hyper diverse ecosystems, different tree types, different histories. So we
• 0:52
  Speaker 2: have a problem in figuring out how much carbon there
• 0:55
  Speaker 2: is in the world's forests.
• 1:00
  Speaker 1: And that's where the Biomass satellite comes in. Launched in
• 1:04
  Speaker 1: twenty twenty five by the European Space Agency, it emits
• 1:07
  Speaker 1: the longest wavelength of radar ever sent from space. The
• 1:11
  Speaker 1: longer the wavelength, the bigger the elements that are picked up,
• 1:15
  Speaker 1: including tree stems and branches.
• 1:17
  Speaker 2: When we look at the Amazon from space with an
• 1:20
  Speaker 2: optical satellite, so imagine, just like human eyesight, you see
• 1:24
  Speaker 2: this huge green blob. The greenness doesn't give us a
• 1:28
  Speaker 2: lot of rich detail. Radar data takes us into a
• 1:32
  Speaker 2: new wavelength of viewing this landscape, and it looks through
• 1:37
  Speaker 2: the leaves, through the small branches and gives you information
• 1:40
  Speaker 2: about the big stems and about the soil and the
• 1:43
  Speaker 2: form of the land underneath the trees. These block shapes
• 1:53
  Speaker 2: are where people are chainsawed into the forest and cleared
• 1:56
  Speaker 2: for pasture. That's important because it's going to help us
• 2:00
  Speaker 2: how deforestation more effectively. But you can also see a
• 2:03
  Speaker 2: more red signal. So what this indicates is the presence
• 2:07
  Speaker 2: of water underneath the canopy. These flooded forests are important
• 2:11
  Speaker 2: because they will have different ecology and we currently don't
• 2:15
  Speaker 2: really understand these they are hard to access.
• 2:17
  Speaker 1: The satellite data also has to be validated. This means
• 2:23
  Speaker 1: cross checking it against other types of tree measurements taken
• 2:27
  Speaker 1: on the ground to see if they match up.
• 2:30
  Speaker 2: Every tree will be measured in a certain area, probably
• 2:35
  Speaker 2: around a square kilometer larger areas will be measured by
• 2:40
  Speaker 2: aircraft laser scanning. Bringing all that information together is absolutely critical.
• 2:52
  Speaker 1: Now this isn't the tropics, but scientists here at the
• 2:55
  Speaker 1: University of Cambridge are also using technology to revolutionize how
• 2:59
  Speaker 1: to reason measured.
• 3:01
  Speaker 3: It is really being pulled by the ways.
• 3:03
  Speaker 1: Doctor Emily Lines and her colleagues are using lidar to
• 3:07
  Speaker 1: scan forests across Europe.
• 3:09
  Speaker 4: Sensors like the one on this drone here send out
• 3:11
  Speaker 4: beams of light and they measure the amount of time
• 3:14
  Speaker 4: it takes without being to return back to the sensor,
• 3:16
  Speaker 4: So that gives us this three dimensional picture of forests
• 3:19
  Speaker 4: that we can see the height of the trees and
• 3:21
  Speaker 4: the size of their crowns, and from there we can
• 3:23
  Speaker 4: infer the amount of carbon.
• 3:24
  Speaker 3: The forests taking in What can lidar give you that
• 3:27
  Speaker 3: satellite data cannot give you?
• 3:29
  Speaker 4: Lyder obviously gives us much higher resolution. We're looking at
• 3:32
  Speaker 4: millimeter precision. So by collecting lighter data on the ground,
• 3:36
  Speaker 4: we can provide ground validation for the satellite data to
• 3:40
  Speaker 4: be checked against.
• 3:41
  Speaker 3: And lidor has been around for quite some time.
• 3:44
  Speaker 4: It's been around for about fifteen years. In my discipline.
• 3:46
  Speaker 4: What has really improved in recent years is processing tools,
• 3:50
  Speaker 4: particularly deep learning and AI, that take these from being
• 3:53
  Speaker 4: scientific instruments into more practical applicable use for forestry and conservation.
• 4:00
  Speaker 4: We will plan a flight route that flies the drone
• 4:02
  Speaker 4: low and slow over the forest in.
• 4:05
  Speaker 3: A grid pattern.
• 4:06
  Speaker 4: But we want to see individual leaves and small branches.
• 4:09
  Speaker 4: We need to be using the ground based instruments.
• 4:12
  Speaker 3: Its challenging to get through here.
• 4:15
  Speaker 4: So here we have another led our system, so similar
• 4:17
  Speaker 4: to what was on the drone, but because it's on
• 4:20
  Speaker 4: the ground and underneath the trees, we can get much
• 4:23
  Speaker 4: higher resolution imagery.
• 4:24
  Speaker 3: So we can pick up.
• 4:26
  Speaker 4: The individual twigs, We can see leaves, we can see
• 4:28
  Speaker 4: the texture of the bark.
• 4:30
  Speaker 3: Would you like to press start? Oh, I'd love to.
• 4:32
  Speaker 1: Okay, Time for a quick light our selfie in a
• 4:38
  Speaker 1: cool shapes. The three hundred and sixty degree image capture
• 4:44
  Speaker 1: means I get two goes at being in the shot.
• 4:48
  Speaker 3: All right, being in two places at once. It's really amazing.
• 4:58
  Speaker 4: So this is the most cutting edge of the technologies
• 5:01
  Speaker 4: that we use in my groups.
• 5:02
  Speaker 3: It's much more maneuverable.
• 5:03
  Speaker 4: This is a much lower cost option, so this is
• 5:06
  Speaker 4: a commercially available off the shelf lider, and it's mounted
• 5:10
  Speaker 4: on an open hardware three D printed system that anyone
• 5:14
  Speaker 4: with a three D printer could make for themselves. Once
• 5:16
  Speaker 4: it's powered up and these lights start going with pressco
• 5:19
  Speaker 4: and you can walk around this beautiful woodland.
• 5:21
  Speaker 3: And scan some trees for us.
• 5:23
  Speaker 1: Okay, now I'm not actually collecting any data here, but
• 5:27
  Speaker 1: it's easy to see how this small handheld device could
• 5:30
  Speaker 1: make lidar much more accessible. Back at Emily's office, she's
• 5:36
  Speaker 1: showing me how her team is using AI to process
• 5:39
  Speaker 1: the lidar data.
• 5:41
  Speaker 4: So we've used deep learning to label each point with
• 5:43
  Speaker 4: the probability that it might be a woody point, because
• 5:46
  Speaker 4: if we're monitoring bio mass, we want to be able
• 5:47
  Speaker 4: to monitor the woody structure and.
• 5:49
  Speaker 3: Remove the leaves away.
• 5:51
  Speaker 4: The redder it is, the more the model thinks it
• 5:53
  Speaker 4: is likely to be wood The deep learning model is
• 5:57
  Speaker 4: able to work out what species we're looking at as well,
• 6:00
  Speaker 4: and what that allows us to do is to better
• 6:02
  Speaker 4: understand the forests, so more diverse forests are more likely
• 6:06
  Speaker 4: to support different types of biodiversity in or wider type
• 6:08
  Speaker 4: of biodiversity.
• 6:10
  Speaker 1: We've been hearing so much about how destructive AI can
• 6:13
  Speaker 1: lead for the environment, So do you feel like the
• 6:15
  Speaker 1: benefits kind of outweigh the negative elements.
• 6:19
  Speaker 4: The tools that we are building here, we are building
• 6:21
  Speaker 4: them generally on our own desktop computers, so we are
• 6:25
  Speaker 4: not using the big data centers. We're trying to build
• 6:28
  Speaker 4: lightweight tools, and we want them to be able to
• 6:30
  Speaker 4: be usable by people who have smaller computers, maybe even
• 6:35
  Speaker 4: run them on laptop. So we try to choose deep
• 6:37
  Speaker 4: learning architectures and methods of learning that minimize the amount
• 6:41
  Speaker 4: of energy.
• 6:43
  Speaker 1: The hope is that both of these emerging technologies will
• 6:46
  Speaker 1: provide a richer understanding of how the world's forests is
• 6:49
  Speaker 1: transforming in a rapidly changing climate.