Eyes on Earth Episode 11 - EROS Fall Poster Session
Each fall, EROS invites its staff scientists and area graduate students to visit for a noontime poster session. The poster sessions offer a change for those researchers to present their results to their peers and get feedback from their fellow scientists. For this episode, we’ll hear about research into biofuels, cloud-friendly Landsat data, shrubland mapping and satellite-based fire monitoring.
Hult: Hello everyone. Welcome to Eyes on Earth, a podcast of the US Geological Surveyís Earth Resources Observation and Science Center. Iím your host, John Hult. On todayís podcast extra, weíll hear from a handful of the EROS staff members and graduate students who took part in our Fall Poster Session. A poster session is essentially a way for scientists to share their work with their colleagues in a public forum. The posters in the atrium for our session, covered a variety of topics on remote sensing, satellite technology and geography. I pulled aside a few of the authors during their hour-long site visit to hear more about their work. First up, weíll hear from a South Dakota State University graduate student who used land cover maps built with satellite data to investigate switchgrass as a more sustainable alternative to corn in ethanol production. We are here with Logan Megard. Heís got a poster that says, ìThe Potential Area of Switchgrass for Bio-fuel Production in South Dakota.î So, what problem are you looking at here, Logan?
Megard: I am looking at renewable energy source that can replace corn. What I am looking at specifically is switchgrass which can be grown on marginal land and has less inputs than corn currently has.
Hult: ìLess inputsî you meanÖ.cheaper to grow?
Megard: Less fertilizers, less irrigation. So, it can be more sustainable.
Hult: What did you do to sort of figure out how much area is available. What are we looking at here?
Megard: I used two data sources. The GAP LANDFIRE cover data from 2011 and the national land cover from 2016.
Hult: And since people arenít going to be able to see this, we are looking at a geospatial data set, we are essentially looking at a map that tells us what parts of the land are forests, developed, cropland, etcÖSo, thatís what you are talking about. You are trying to find areas that are not currently used for crops that could be suitable for this switchgrass.
Megard: Yes. We are finding areas that arenít currently used for cropland, or developed, or water, cut-out forests as well because we donít want to cut those down.
Hult: What did you find?
Megard: Between the two data sources, I found 3 million hectors and 2.1 million hectors that are in this useable area.
Hult: In those two data sets, you found 3 million in one and 2.1 million in the other?
Megard: Yes. For those 2-3 million acres I found that you could produce 5.37 to 7.67 billion liters of biofuel a year.
Hult: So, there is a significant possibility there, thereís a significant opportunity?
Megard: Yes. The only problem with it is that itís more renewable but itís also not profitable which is why it has taken off yet.
Hult: The National Land Cover database that Logan used for his study is produced at EROS. Logan used land cover data but the database has several components. One of them defines how much grass and shrub land covers each 30 x 30-meter plot of ground in the western United States. Letís visit with one of the EROS contractors behind that project. We are here with Matthew Rigge. He has a poster that says, ìValidating a Landsat Time-Series of Fractional Component Cover Across Western US Rangelands.î What on earth does that mean?
Rigge: It means that we are just trying to figure out a way to validate our fractional maps. So, 0-100% cover mapsÖ.
Hult: So, fractional is 1-100%?
Rigge: Exactly, yes. Across a big time and space. These maps span from ë84 to 2018. Since we donít have a time machine, weíre trying to figure out a way to use existing data sets to validate previous years where we canít go out and collect field data anymore.
Hult: And by validate, you mean make sure that you are right essentially? Check your work?
Rigge: Correct. But not even to make sure that we are right. To quantify accuracy. Letís put it that way. We are looking in Nevada, Wyoming and Montana. We had a series of high-resolution satellite images. So, like 2-meter satellite imagery. And field data collected at the same time and same place.
Hult: Gotcha, we are talking about field observations. You have someone on the ground saying, ìYes, this is sagebrush, this is cheatgrass, whatever it is you are looking for.
Rigge: Correct. And not only that but what is the cover of each of those? From 0 to 100%, what is the cover of those?
Hult: So, we know. Because we are on the ground we know this is 80% cover and then you get your 30 meter data and it says whatever it says.
Rigge: Right, and then compare those two.
Hult: And youíre talking about an accuracy assessment for a mapping product, right? Is this an NLCD product?
Rigge: Well, itís like a subset of an NLCD.
Hult: A subset of the national land cover database?
Hult: I see. And, Why does this matter? Why would you go to all this trouble?
Rigge: Itís important because we want users who would be like land managers for instance to have confidence in our products to know that they can reliably use these data to answer the science questions that they have.
Hult: The Landsat data used for land cover mapping will soon be available in the cloud. Letís hear from one of the EROS contractors working to make cloud-friendly Landsat a reality. We are here with Renee Pieschke. And your poster is Landsat in the Cloud. Can you tell us what you are looking at here, what you are studying, what this is about?
Pieschke: The Landsat Product of Improvements project has been working on putting Landsat into the cloud for the past two years. We are getting closer. Weíve got 6.5 petabytes in the cloud right now. Collection 2 will be processed in the cloud. So this poster is focused on how our end users are going to access the data and a little bit about the new cloud format, the cloud-optimized GeoTIFF and this new metadata format called the Spatio-Temporal Asset Catalog which is pointing directly to cloud assets as they are sitting in their cloud locations.
Hult: Spatio-Temporal Asset Catalog, this is STAC?
Hult: This is a tool? Is that right?
Pieschke: This is just a format of meta data that will allow you to access the data programmatically. So if you are in a Jupiter notebook, you can query Landsat data where it lives, you donít have to download any of the data at all.
Hult: And you say how soon is it that people will be able to use this?
Pieschke: We are looking at collection to release in the spring.
Hult: Remote sensing scientists spend a lot of their time working to fine tune satellite data, to make it more useful to those who need it on the ground. Letís hear from an EROS contractor who recently published a paper that explains how to improve the resolution of daily satellite readings over active fires. We are here with Sanath Kumar and his poster is titled, ìAn Algorithm to Downscale Motis Satellite One Kilometer Daytime Active Fire Detections.î Iím guessing some of our listeners might not understand what that means. Can you tell us what you are doing here? What you are looking at? What problems you are trying to solve?
Kumar: We are looking at fires in space. There is an inherent problem. The thermal sensors we have, they are about a 1 kilometer resolution. So basically what it means is If you say a fire happens somewhere on location on earth, we donít know where within the first one kilometer radius. Itís there in, right? If we can reduce the uncertainty that would be good. It would be good for fire managers to know where the fire is. So, this work reduces uncertainty. And hopefully going forward we can possibly do it to the nearest 30 meters to 10 meters.
Hult: Well, great! How did you do that?
Kumar: Physics. The Planckís function. It all comes to fundamental science. How radiation is, what wavelengths, depends on the temperature in the area, and what spectral wavelengths are of interest. For example if you heat a rod of iron, it first goes to red, right? And then it goes blue hot or white hot. So by just looking at the color temperature you know how hot it is. So, by measuring the intensity of how much blue or how much red it is we also know the size. So, by using thermal sensors we can actually have an idea of what the size of a fire and whatís the temperature of the fire. So, the big takeaway is Yes, technology is improving and we can go find a higher temporal resolution. Itís like having more precise data. Previous we had approximate data now we are zoning on to the more finer details.
Hult: Who is going to use this or who is using this method now?
Kumar: As of now, itís a recent publication. Nobody is using right now, except me. But hopefully the fire managers. For example the California fire that happened, if we know the location of the fire, it would be useful for people, the first responders. It would be useful for the management group. These are the immediate people that might find it of interest. The second interest is for the people who are monitoring the science aspect of it. How the fire spreads and grows. In the 1-kilometer fire, we canít see growth of the fire unless we map it for several days. Whereas if you have a higher resolution product, we can see how the fire, the direction in which it is moving so we can relate it more to the ground. So we can have a better fire spread models, that means we can understand the effects of fire. How it grows in the landscape, and it can help in the future fires, how accurately we can predict, so it has good repercussions, yes.
Hult: Thanks for listening to Eyes on Earth. We hope you join us for the next episode, which you can find by visiting usgs.gov/eros. This podcast is a production of the US Geological Survey, Department of the Interior.