Hurricane Maria and unequal recovery

1:01:36uh artificial light driven by human activity and removing everything that is not that. >> Exactly. Yeah. And so in the next paper published in 2019, we use the black marble algorithm to talk about what happened after hurricane Maria. >> Mhm. >> This is Roman and Stokes. They shared first author credit. All five of the people that are being awarded the golden goose are also on the author's list here. The paper asks how nighttime lights can be used to estimate the geography and duration of electricity loss after hurricane Maria. And the first step is you just want to establish a baseline, right? You want to say this is what Puerto Rico was like before the

1:02:17hurricane came about. >> Mhm. >> And then you can ask, okay, what is it looking like afterwards? Now, there is a subtle question here that has to do with resolution, which I thought was very cool. So, look at the top left. >> The top left is actually the resolution of the satellite. >> Okay. And so, for audio listeners, this looks like uh a 8bit pixel video game or whatever the correct number is, but highly pixelated. >> Highly pixelated. And there's a reason for that, right? There's not a lot of light >> Yeah. coming at night. >> Yeah. >> Okay. >> Which is a a challenge in and of itself

1:02:58to capture in the first place. >> Exactly. Like in the daytime there's so much light coming in there that you can really resolve things very very finely. At night time there's not a lot of light coming in and so naturally your resolution is going to be not that great. How do you get over that? Well, you can combine it with LANCSAT and Sentinel 2 images which are separate >> which are separate satellites that tell you about the distribution of houses and during the daytime. During the daytime they have all of this resolution. If you combine these two then you can ask things like okay um you know >> there's a pixel there's a giant pixel

1:03:38let's say >> y >> from the nighttime a 500 meter x 500 meter pixel that's covering a town. So, this is half a kilometer wide. Um, but that town, that same pixel has a town. It's got a wooded hillside and it's got part of a reservoir. The satellite might report the pixel that's lost most of its light, but that measurement alone is not going to pinpoint where exactly the light came from. If you combine it with the LANCET images and the Sentinel 2 images, then you can the finer daytime resolution is going to give these researchers a guide and you build that in into your algorithm. So, >> right. So, in practice, what they're doing is interpolating in some sense the

1:04:18nighttime data >> onto a 30 m grid from a 500 meter grid to now 30 m, which is, you know, the size of communities, big houses >> kind of thing. M um it it's it's an incredible thing that they've done. So let's see how they how it worked. Right. The the paper actually reports an approximately 80% reduction in the island's total observed lights immediately after Maria. >> So crazy, >> right? Because there was just so much devastation caused by this hurricane. And it was on the news that the recovery from that hurricane, especially in terms of the power needs of the people of Puerto Rico, it was not

1:04:59that great. Right. What this actually shows is a resolution depth of which communities fared better than others. >> Oh, that's quite that's quite nice. >> Right. So here we're looking at a color-coded graph where the color represents the number of days without electricity post Hurricane Maria. >> Okay. >> Okay. The brighter yellowower spots means that almost immediately within 30 days, within a month, the power was back. The really dark spots, that's 150 days. That's like 6 months plus. >> Unbelievable, >> right? And you can see different parts of Puerto Rico have different amounts of

1:05:40power loss on the order of months we're talking. And even if you look at the aerosibo, which is the first row, third column, you'll find parts that the power returned immediately and then other parts where it didn't. You can then go into the data and say what is the difference between the parts that returned early and the parts that returned late. And almost always it has to do do with poorer rural communities. The urban communities got it first. The poor rural communities got it second. And so the poorer residents often bore long interruptions. >> I mean it's a story as old as time. And

1:06:20the what's interesting is you're you're getting this empirical data related to disaster recovery that uh makes it very black and white because it's also happening on over time. >> Yeah. >> And so you get this level of granularity where like you talked about earlier where it's like 150 days versus almost immediately. And I can just imagine in my mind, you know, if I'm thinking of, you know, something like in the U, like in FEMA in the US in terms of responses, like that data, if I was someone who was trying to plan to do make disaster recovery more improved in the future, >> uh, just something as simple as that

1:07:02context, >> um, you know, cuz then I could then go as someone who's in the disaster recovery space and say, okay, well, in these areas, are our processes different. Did we access them later? Do we have enough coverage from like a resources and human capital perspective? Um, and in theory, what you would want to be able to do, I'm just extrapolating over time, is look at because obviously these weather events continue to happen. >> Yeah. >> Um, >> they might be happening more frequently because of climate change. >> Correct. M >> you could start to at least look at this as a way to make progressive progress

1:07:42because you can see how it happened in X year in the past, look at Year, and then say, okay, have we increased have have some of these communities decreased in response time? And then you can then have real data to say we've done better. >> Yeah. >> Um just as an example, that's where my mind goes. >> Totally. I mean, it's it it's I I think it's a really cool way to look at empirical data from satellites and then like answer these societal questions about like do we do we is this okay? Right. >> Yeah. Yeah. >> Um so we move on to our paper number