Eyes on Earth Episode 40 – Landsat Collection 2

JOHN HULT:

Hello everyone. Welcome to this episode of Eyes on Earth. We're a podcast that focuses on our ever-changing planet and on the people here at EROS and around the globe who use remote sensing to monitor and study the health of Earth.

My name is John Hult, and I will be your host for this episode.

If you are a regular listener, you have surely heard us talk about how the Landsat satellite data archive represents the longest continuously collected record of the Earth's surface in existence. You have also heard about how scientists monitor the health of the planet by looking back through that nearly 50-year record to track change. But how can data collected in 1972, by a satellite with 1972 technology, possibly align with data collected yesterday, by a satellite launched 40 years later? The answer, for the most part, is Collections. Landsat Collection 1 saw all that data calibrated to match up as closely as possible to match up as closely as possible across all 7 satellite systems. The work allowed scientists to track points on the surface of the Earth more easily and gave them more confidence in their conclusions.

The Landsat team at EROS has just released Collection 2. An upgrade that improves accuracy and expands access to higher level products like land surface temperature. Collection 2 also makes Landsat data available in a cloud friendly format.

Here with us to talk about Collection 2 is Dr. Chris Barnes a contractor at EROS who supports the Landsat International Cooperator Network. Dr. Barnes, thank you for joining us.

CHRIS BARNES:

Thank you very much. Great to be here.

HULT:

Also joining us is Dr. Christopher BARBER. A remote sensing scientist with the USGS Land Change Monitoring Assessment and Projecting Initiative also known as LCMAP. Dr. BARBER, thank you for joining us.

CHRIS BARBER:

Not a problem. Happy to contribute.

HULT:

Dr. Barnes and Dr. Barber, you both work in remote sensing. You are both names Chris. My guess is you have probably gone to the same conference once or twice. You guys must have had your luggage messed up at the airport at least once, right?

BARBER:

We have had frequent flyer miles mixed up.

HULT:

Oh really? Who was the beneficiary of that?

BARNES:

I'm pleased to say that it was me that took a trip to South America.

HULT:

Oh, nice. Let's get into Collection 2 here. Dr. Barnes, we are going to start with you. Why don't you tell us what the word "collections" means in relation to satellite data. How does a collection help scientists study the Earth?

BARNES:

Absolutely. That's a great question. Back in 2016, USGS released the first Collection, Landsat Collection 1, which was a major shift in the management of the USGS archive. Before that, the Landsat archive was processed based on the most current calibration parameters that were available at the time, or the best known updates. The users would have to spend some time and effort trying to determine where that data came from, what system was used to process it. Not all Landsat instruments were processed using the same product generation system. So, in recognizing these challenges, the USGS worked with the Landsat user community and also with the joint agency USGS-NASA, Landsat Science Team to determine how they could provide a consistent archive of known data quality. The bonus in that is, it would allow users more time to conduct their scientific research using Landsat data.

HULT:

If I can interject here really quickly ... I think what I heard was that in the past, before collections, the newest Landsat data had the best calibration, the best accuracy and all of that, and something from 20 years ago didn't have all of the newest processing and it didn't align with the rest of the data for the most part. There were issues, I guess.

BARNES:

Yes. There was a little disconnect in what is being acquired today as to what was acquired back in the 1970s and 1980s.

HULT:

So in Collection 1, you did that. You did align all of the data as well as possible, is that right?

BARNES:

Absolutely, yes that is right. So all the data going back to 1972 from the days of Landsat 1 all the way to Landsat 8, the most current Landsat that is in orbit. All of that data, over 9 million scenes, have all been processed to the same calibration and validation parameters. So that allows users to go back and forth through the entire Landsat archive and conduct their research knowing that the most up to date parameters have been used to calibrate that data.

HULT:

Now, I want to jump over to Dr. Barber here, because you worked with Landsat data before Collections, as I understand it, in some remote parts of the world. Tell us about that work, and tell us what it was like to work with Landsat data before Collection 1.

BARBER:

Not all the Landsat data that exists was collected by what we call the United States ground station. There's a collection that are called the foreign ground stations. They were scattered in a bunch of countries around the world and each of those started with the U.S. version of software for processing Landsat data, but then they customized to their particular needs and their particular tastes for their local user community. So when we start working in Southeast Asia or South America, and you start working with data from 2, 3, 4 different foreign ground stations, they have all been processed with different versions of the software, using different algorithms coming in different physical formats. You had to change the way you worked with each piece of data depending on when and where it came from.

HULT:

So, before the consolidation, the Landsat Global Consolidation, where all that international Landsat data was moved from those ground stations into the EROS archive and later processed into Collection 1, you were relying on data from ground stations that may have been and still could be processing data differently to serve their own local needs. It almost sounds as though you were working with black and white photography versus colored photography, maybe one image that is zoomed in and the other that's a little wider. You kind of had to cobble all that together. Does that sound right?

BARBER:

That's a very broad, rough analogy, yeah. Even countries right next to each other like Thailand and Indonesia, they aren't right next to each other, but they are close. They would process data differently.

HULT:

Interesting. Well, now they would be looking at, if they were looking at Collection 1, they would be looking at the same processing, the same standards and things would match up a lot easier. 

Dr. Barnes, I want to turn back to you on this. How do we do that? I mean, how do we get to a place where we can compare a satellite image from 40 years ago to one collected just yesterday? Talk to us a little bit about the steps involved in making all those pixels align.

BARNES:

Well, all those kudos go to a very intelligent team of calibration and validation engineers both part of the USGS and NASA. Where they are constantly monitoring and looking at the performance of the instruments on board the Landsat spacecraft. They then publish those findings in peer reviewed journal articles. Those then get feedback by people in the calibration/validation community. It all comes down to monitoring on a daily basis how the instrument is performing and what changes, if necessary, need to be applied. That's kind of what's happened in this collection management structure. Any changes needed or observed that need to be applied, in a new collection, they make that call of when those will be implemented. That is exactly what's happened and in part what has triggered a reprocessing event for Collection 2.

HULT:

Collection 1 sounds pretty great. Sounds like you have done all the calibration/validation, they do all this work. They really dig down to make sure they have the right changes, the right alterations, the right fixes. They apply them all the way back through the archive and it sounds like you turned a Betamax video cassette into a DVD. Now we are looking at Collection 2. How much better could it possibly be? Tell us what's new with Collection 2 and what kinds of improvements we're going to see. 

BARNES:

That's a great question. Some of the main improvements that the users will be very excited to learn about is the substantial improvement in the absolute geo-location accuracy which is used to do the ground control reference data set. Basically pinpoints, very accurately, the Landsat scene onto the Earth's surface. Not only being able to go back through the Landsat archive, but also improves interoperability with the European Space Agency's Sentinel 2 satellites, which are very similar.

HULT:

And if we can put a finer point on that, that seems like a pretty big deal. We are talking about a point on the Earth's surface. You are talking about each pixel. Each 30 meter by 30 meter pixel of Landsat from further back in the archive, matching up even closer to the front of the archive, because there were times where it was a little bit off even in Collection 1 as I understand it, right?

BARNES:

A little bit, yes. And improving this interoperability allows users to be able to get more frequent observations of the same place on the Earth's surface by pulling in the Sentinel 2 series of satellites alongside Landsat.

HULT:

It's not just lining up Landsat pixels but it's also bringing those pixels closer to alignment with a very similar system to get more observations. Interesting.

BARNES:

Absolutely. Some other major highlight that users will be pleased to hear there are some updated global digital elevation modeling sources that are used in the Collection 2 processing system. Also, for the first time, USGS will be producing global surface reflectance and surface temperature products from the early 1980s will also be distributed as part of Collection 2.

HULT:

And when you say distributed, just to make this clear to the people who are listening, for Collection 1 you could ask for surface reflectance and surface temperature for anywhere in the world, is that right? But now it is just going to be there. If you log into EarthExplorer and search Collection 2, you will just be able to get it, is that right?

BARNES:

That is right. Up until Collection 2 was publicly available, surface reflectance was only available to the user community on demand. The surface temperature products was only available through the U.S. Analysis Ready data, data set for the United States. This time, USGS will be processing and making that available for the entire world.

HULT:

Dr. Barber, tell us a little about LCMAP. My understanding is LCMAP uses Collection 1. Tell us just broadly what LCMAP does and do you see any improvements to LCMAP with Collection 2?

BARBER:

One of the things to think about is, you know, 20 plus years ago, monitoring the land surface with Landsat data was a bit of a challenge because data was expensive, computer storage was expensive. A half a dozen, a dozen or 20 images for your study period was about all you could handle for cost and storage. Some big projects, maybe 100 images, 200 images. Today, data is free, computer storage is inexpensive. So the idea of, "well let's look at all the images for the United States and look at how land cover is changing across the United States and the land surface is changing, using all the inputs" becomes an idea that is cost effective and storage effective so, why not? Before Collections, the problem was there was inconsistent data through that historical record. So if you want to do monitoring over time, especially with any kind of automated method it's really important that you are measuring the same thing and measuring with the same measurements all the time. So for example, if you want to track temperature in your backyard, you aren't going to mix up measurements of, some days you're going to take the temperature in your backyard, some days your front yard and some days in your neighbors driveway, and mix up Fahrenheit and Celsius. You're going to put one thermometer in the backyard with one measurement system to monitor that temperature. That's what Collection 1 allows us to do with LCMAP. To look at the conterminous United States and really look at all Landsat data available and track it through time and analyze land cover change. With Collection 2 coming up, we expect to see some improvements, especially in the geolocations so that thermometer is always in the same place even more precisely, and improvements in the calibration and things like that. I think the advantages of Collection 2 are going to be much more evident in other parts of the world, outside the United States. A lot of the data over the United States has already been, even in Collection 1, was well advanced. At some point in the future there's a potential of taking LCMAP global, and at that point Collection 2 or even Collection 3 or beyond is going to be really invaluable for taking that work forward.

HULT:

So what you do with LCMAP is to look at every pixel back through time, to create these products. That's only possible because of Collection 1. And Collection 2 is going to perhaps improve the results there because of the accuracy, the thermometer issue that you brought up. But it's also going to make better data available to a broader swath of the world, potentially taking this approach from LCMAP and making it possible to do in parts of the world where it maybe wasn't before. There is something else I wanted to ask about with LCMAP. Is there some interplay between Collection 2 and the possibility of including other data sets in the algorithm you have now?

BARBER:

The European Space Agency has program called Copernicus with a satellite called Sentinel-2, which produces some data that is similar to Landsat, and there is work that has been done on how do we take Landsat data and Sentinel data and do what is called harmonize it to make the measurements comparable easily. So we can start to compare them directly. We are in a really rich time for satellite observations compared to 20 years ago or even 10 years ago. So over the next 2, 5, 10 years there's going to be more and more Earth observation satellites up there. So learning how to incorporate different sensors and different space platforms together is the way forward.

HULT:

Right. So you at LCMAP, you are thinking about this stuff-the idea of harmonizing these datasets and incorporating more observations into your work. I know NASA is working on a harmonized Landsat/Sentinel product as well. So broadly speaking, that is where the future is and that is something that this particular improvement in Collection 2 will make easier.

BARBER:

That's it exactly. So going from before Collections to Collection 1 to Collection 2, we are taking data from up to today, 8 different Landsats and making that data all work together so it is easy for researchers to use it all together. The moving forward is, how do we take those lessons and expand it to more satellite systems and different sensors and make that data all usable directly to researchers without having to worry about the engineering stuff in the background.

HULT:

Dr. Barnes, it sounds like Dr. Barber is pretty pleased with the direction we're heading. So congratulations there. Good work for your team. But, I want to talk about something else with Collection 2, which you briefly mentioned. The idea of land surface temperature and surface reflectance, those higher-level products being available right there. What do you think that particular advancement might mean to the world of remote sensing? What kind of research is that going to aid?

BARNES:

Absolutely. The first advantage to the user community is that preprocessing has already been taken care of by the USGS. Hopefully with these being globally available products, people are going to be able to do more extensive research. For example, surface reflectance account for aerosols, water vapor and ozone in the atmosphere and therefore it helps to accurately map changes in the Earth's surface. So applications from all around the world will be able to use looking at the Earth's surface for change and impacts to how the Earth's surface is changing. They will be able to use that product. For the land surface temperature product, because that will also be globally available, people will now start to incorporate that into global energy balance studies. Looking at how the Earth's global energy is changing over time. Looking at hydrological modeling, looking to sort of take crop monitoring and also trying to get an indicator of vegetation health and also it can be used for looking at extreme heat events such as natural disasters, volcanic eruptions, wild fires and also how urban heat islands are propagating through time as global population rises and urban centers continue to gentrify out.

HULT:

We're looking at the possibility of, because it's available right there, being able to sort of automate, if you're the kind of person who does this research, you are going to be able to, if you want to, automate these processes to do some of these analyses without having to make extra requests for one thing, and you're also expanding it globally. We are talking about being able to see whether a particular city in India or Pakistan is hotter now than it was in 1994 and being able to quantify that much more easily. Is that maybe one example?

BARNES:

Absolutely. That would be one of the example applications of the surface temperature product. And to go one step further the fact that USGS moved to making Landsat Collection 2 available in a commercial cloud environment definitely lends its hand to those users who do want to do global-scale or even continental scale analyses using Collection 2. They will be able to bring their algorithms to the data, whereas in the past and what has historically been done, which is what Dr. Barber was referring to earlier, people had to download large volumes of data. That took a lot of time, cost a lot of money. You had to store that data and then run your algorithm on that data locally if you had that capability or transfer it to a place where you will be able to do that. So being accessible in a cloud environment really opens up a plethora of options for the user community to do all kinds of research. And we are really excited to see what this engenders.

HULT:

The idea of putting the data in the cloud and being able to work in the cloud environment, with Landsat data, doesn't it sort of level the playing field for folks who maybe otherwise wouldn't have access to the computing power it would take if they had to download all this data. Is that a fair characterization of one possibility?

BARNES:

Yes. I think that is a very good way of looking at this. USGS has processed the archive to the highest possible standard in the history of the Landsat program and taken it one step further by putting it in this cloud environment to allow what you are eluding to a more even playing field of accessibility to retrieving that data.

HULT:

Right. Because as I understand it, some of the work that LCMAP might do and even further with some of the global monitoring things that are taking place, you need some pretty serious horsepower to do that work. Don't you Dr. Barber?

BARBER:

Indeed. You know it is getting less and less as we move forward. One of the things that is important to remember is that science operates on a budget. And some of that money goes to computational resources and computer storage and time for human resources to do analysis. Pre-Collections, a lot of that time was taken up in just preparing data for analysis. If we can get to the point with Collection 2 where we can get real measurements of surface reflectance and land surface temperature, ready to go for science, that leaves a lot more in our kind of human resources and computational budget to use for actual analysis rather than data prep.

HULT:

That's a good point and gets to something we should probably address here and make clear. The data is being made available in this cloud friendly format but it is not as though the USGS is providing cloud storage, right? Like a person would have to pay for a cloud storage plan, just as they would have had to pay for a computer and an internet connection to download the data before. The data is there and is available in that format. Is that right, Dr Barnes?

BARNES:

That is exactly right. The USGS is making the Collection 2 Landsat archive freely available. There is no change to the 2008 Open Data Policy but users will have to work with the respective commercial cloud providers if they want to do running algorithms on the archive in the cloud, and also exporting the results from running that algorithm in that cloud environment.

HULT:

Just to wrap up here ... is there anything else you would like the world to know? How does it feel to have this job done?

BARNES:

Yes, I think obviously this is the second major reprocessing event USGS has done with the Landsat archive in 4 or 5 years. The major accomplishment with this is the amount of enhancements that have gone into this version of the archive. Not only just improving the quality of the Landsat archive but also new data access and distribution capabilities. And of course the new products of surface reflectance and surface temperature. Another major leap that the USGS took with this is migrating to the cloud environment. That not only means data access and distribution but also the processing of the Landsat archive in the cloud. So it really goes to show you how far the USGS have come in these last 5 years between Collection 1 and Collection 2 of really being able to turn around and produce the most high quality Landsat archive known to date.

HULT:

We've been talking to Dr. Chris Barnes and Dr. Chris Barber about Collection 2 and improvements to the Landsat archive. Doctors, thank you for joining us.

BARNES:

Thanks, John

BARBER:

Thanks, John. Exciting times ahead. 

This podcast is a product of the U.S. Geological Survey, Department of Interior.