Calibrating Remote Sensing for Climate-Risk Zones
Researchers are tackling the data gap in rapidly growing Sub-Saharan African cities by combining remote-sensing measurements, low-cost satellite imagery, and smarter machine-learning models. Their work focuses on validating and improving these tools—so models trained on U.S. suburbs can reliably map risks like flooding, heat, and transportation stress in places like Kigali. For the test & measurement community, it’s a story of making high-tech sensing and calibration methods robust where data is scarce—so climate-risk planning can work where it’s needed most.
Transcript
00:00:07 Emily: I am working on improving global access to data for climate change adaptation. We’re specifically focused on cities in Sub-Saharan Africa that maybe don’t have the same levels of data that we’re used to in the United States or Europe around cities, how they’re developing where people are living, and the sort of risks they face from climate change. We look at harnessing remote sensing data and satellite imagery that provides access to really low-cost data sources that you don’t get from any of these ground surveys or really long-term data collection efforts. And in order to interpret those and be able to provide information to local stakeholders and policymakers, try to interpret them with machine learning models. But those machine learning models aren’t super globally applicable. They’re often trained on data in the United States where a suburb in Texas looks very different from Kigali in Rwanda. And so, learning how we can use
00:00:55 different statistical methods in order to improve the performance of these models so that they’re reliable over time in these cities where we don’t have the data but really desperately need that information for climate change planning. So ultimately, we’re kind of hoping to develop methods to improve, like, all of these really cool technological developments that we’ve had as far as access to data and planning for climate change, and make sure that you can use those everywhere that they’ll have the greatest impact. As we start to prove these methods, we’re looking to be connecting with stakeholders in Kigali to start to be kind of sharing our results. There’s a lot of new and exciting data sources like satellite imagery, remote sensing, ways to interpret big data through machine learning, but paying special focus to making sure that those are
00:01:34 reliable methods in these data-scarce areas. So through statistical methods, through validating that and seeing how that progresses through an entire risk assessment is really important for making sure that we kind of have global impacts from all of these new technological developments. David: So, Emily’s work is really looking at the forefront of how we can respond to climate change. Really looking at from the lens of data access, because in a lot of these regions that are developing, especially Sub-Saharan Africa is a great example, they don’t have access to these data sets, and remote sensing is a very powerful way to be able to get access to these data sets. But in order to properly use remote sensing data, you have to be able to validate it. And so, Emily’s doing a lot of work to be able to understand how good are the data sets from
00:02:23 remote sensing, and then thinking about how they can be used to better understand urban flooding, transportation networks, urban heat island effects. There’s lots of different ways that these data sets can be used to support adaptation planning moving forward.