Eyes on Earth Episode 29 – Tracking Landscape Change with LCMAP

HULT:

Hello, everyone, and welcome to this episode of Eyes on Earth, a podcast focusing on our ever-changing planet and on the people here at EROS and across the globe who use remote sensing to monitor the Earth.

The U.S. Geological Survey took a bold step toward documenting change across the landscape with the launch of the first Landsat satellite in 1972. Since then, it's collected nearly five decades of imagery.

But it takes more than just imagery to understand change. It takes time, effort-and serious computing horsepower. The most powerful computers in the 1970s and 80s could scarcely begin to run the kinds of sophisticated algorithms needed to turn all those images into meaningful information.

Today, we're talking about the USGS Land Change Monitoring, Assessment, and Projection initiative, which we call LCMAP for short. LCMAP aims to use the full Landsat record to track change. That could mean abrupt changes like new development, loss of tree cover from wildfires or forest harvest, or the conversion of land from grasslands to crops, or changes that play out over several years. The LCMAP team looks at every high-quality Landsat pixel, going back to 1985. 

The collection not only tells us what's on the land year after year, but where and how that land might have changed during each year. 

The LCMAP approach can help us track land cover dynamics and tie change to events like hurricanes or fires, to public policies like the U.S. farm bill, to climate change and weather patterns, to invasive species, and much, much more.

Here with us to talk about LCMAP are Research Geographers Jess Brown who leads the initiative and Jen Rover who leads LCMAP's Applied Sciences. Jess and Jen welcome to Eyes On Earth.

BROWN:

Hi John. This is Jess. Good to be here.

ROVER:

Hi John. Thanks for inviting me. This is Jen.

HULT:

Alright, let's get into this now. Let's start with the basics. What is LCMAP? Where did this idea come from and why did EROS decide to pursue it?

BROWN:

So John, I will tackle that. LCMAP is all about characterizing, monitoring and understanding change on the Earth's surface. This idea really goes back a long time and we could go back to the beginning of Landsat actually. We are harnessing the power of the Landsat archive as a record. Folks that work at EROS primarily Dr. Tom Loveland and his colleagues that he has worked with over the years. Dr. Curtis Woodcock and Dr. Zhe Zhu and many others have really helped spawn this idea that we could use this incredible archive that we have and use it to really target change events and characterize the cover of the lands surface.

HULT:

Well, let's talk about that how it does that. How does LCMAP use Landsat data to find where and when the land surface is changing? What kinds of changes are we looking at? Are we just talking about leaves changing color or snow being on the ground where it wasn't there before? What kinds of changes are we talking about?

BROWN:

We're trying to target changes in the land cover state or condition. What I mean by land cover state is the change from something like a cropland land cover to a developed cover. Or forested to cropland or grassland. These characteristics of the land surface are different than just simply tracking phenology differences or vegetation growth differences. When you talked about leaves changing, that is not what we are doing and I think your other example was snow melting.

HULT:

Right

BROWN:

That's not what LCMAP does.

HULT:

How does LCMAP use Landsat data to find change?

BROWN:

We have reconditioned the archive into something called "Analysis Ready Data" or ARD and this is essentially a super large data cube of observations through time. After we recondition the archive, we can use an approach called CCDC or "Continuous Change Detection and Classification" to identify or detect change events and we can use the information to classify land cover.

HULT:

So if I can break in here real quick....when you are talking about stacking pixels, like each pixel would be a year, right? So (in) 1985 this pixel was like this, 1986 like, is that what you are talking about?

BROWN:

So a pixel is a picture element and I consider that attached to a specific location. 

HULT:

Okay

BROWN:

If you can see my hands, which you can't...you can see me with our X and Y dimensions of space and then our Z dimension, which is time so, really we are thinking about the data as stacked up like a big layer cake and we are running a vertical marker through that time stack.

HULT:

So some of the products are about land cover and other products are about change, but you are not just looking at big changes that might change whether land is grassland or developed. You are also seeing some of those smaller changes as you move though the model and some of the products identify those changes, is that right?

BROWN:

So CCDC identifies change on the land surface and the change that it finds could be a land cover change or it could be a different kind of change. These other changes could be just where drought has influenced the grass surface and it has dried substantially or where a forest thinning event has occurred. So you have tree cover, the cover hasn't changed, but perhaps you have fewer trees or perhaps the forest got some kind of insect infestation. So those are changes that don't initiate a land cover change. CCDC finds all of those. So the first step in our process is just to find change and then after the change has been found then we run the land cover and then we can identify which of those changes were actually a land cover shift.    

HULT:

Gotcha. So you start with change. You don't start by classifying the land you start by looking for the change and work forward from that.

BROWN:

Bingo. Yes

HULT:

OK. Great. Why is it so important to understand how the land surface changes through space and time? I mean why can't you just make a map and then come back and revisit and see if something has changed 5 or 6 years from now?

BROWN:

So land cover and these other kinds of changes that we're finding is a lot more dynamic then we ever imagined and if we are not monitoring continuously we are not going to be able to find that. What we can do with this approach, which we have never done before is actually identify the date, within a year, that we observed that change.

HULT:

That's interesting. So you could, potentially tie these changes based on time of year to things like hurricanes, that are close to, if you see a change that is close to the date that a hurricane happened or a drought you can actually tie it to events that we know about.

BROWN:

That's right. We have done that. This is still a work in progress but we do see really strong evidence especially with large change events such as those caused by hurricanes that we can, I wouldn't say pin point the exact date but it is better than we've had before.

HULT:

Right, right. Not exact but somewhere in the neighborhood of an event.

BROWN:

That's right.

HULT:

I think I want to bring Jen in here. Jen you have worked with this data a little bit and you looked at North Dakota as I understand it, some wetlands in North Dakota. What did you find using the data in North Dakota?

ROVER:

Well, just a little background on that geographic region. It's an area that has a really high density of wetlands on the landscape and so it is an interesting area to look at. Wetlands are very ephemeral. They change within a season. Using that information from the change day I can actually go in and map which pixels within a wetland extent, that wetland area, when they change, what time of year they change. So is it a wetland that has a lot of change in the spring or is it summer or is it fall or is it a combination of that through these 33 years that we have data?

HULT:

Why would that information be useful? Why would you want to know how a wetland is changing? At what times of the year, what seasons? How would that be useful potentially?

ROVER:

Well, the US Government has done some mapping and inventorying of wetlands and that started many decades ago. Those maps were originally created with aerial photography and so it was labor intensive. I am actually looking into how LCMAP can be used to identify areas that have had a lot of change and if we know there's been a lot of change then maybe those areas are prioritized for updates rather than areas that haven't had a lot of changes.

HULT:

When you're updating those areas you're hoping to maybe manage the area for wildlife or manage the area for water use is that the purpose of making these maps?

ROVER:

It's to understand from a wetlands perspective, what types of ecosystem services are available. So, yes that's part of it, is just understanding our resources but wetlands are protected by the Clean Water Act and so it is important to know where they are at so they are not disturbed accidently or somehow drained when they are expected to be a resource on our environment.

HULT:

Essentially the better you understand them and the more frequently you can observe them and understand the change, the better job you can do at managing these protected areas.

ROVER:

Correct

HULT:

You Jen, you had access to this stuff early on, because you are on the team. But now it's available to everybody, right? How do people get this stuff I mean, just a regular person finding this information? Is it online somewhere? Do you have to pay for it?

ROVER:

No, the data are available for anyone to use. I would say, if you just want to peruse the data the best place to go is our LCMAP Viewer. You can zoom in and zoom out on all the continental and contiguous US and you can also download data there. For our geek out users what they probably, maybe want to do is go into EarthExplorer and grab data there. That's where the Landsat data are also available.

HULT:

And the Viewer, just so we are clear on this, the Viewer is like a map of the United States and you can zoom into your house or you can zoom into North Dakota or you can zoom into wherever you want and you can get all 10 of these products, you can download it or you can just look at it. Either way it's kind of just a way to take a look at these things if you don't want to download huge chunks if it and just want to see what's happening. Is that right?

ROVER:

That's right and there's a capability there to open multiple windows so you can look at different products over the same area and sort of understand, ok we do have a change here. Is it a land cover change? Or is that a condition change? When did that happen? What day of year did it happen and what was the magnitude of that change?

HULT:

And again, you can also download it. We have a website as well, right? There's a USGS.gov/LCMAP and you can learn about this stuff as well. One of the things you might find on that website is some information about updates. Like, are you done? It goes from 85 to 2018, is that right? 2017, 2018?

ROVER:

2017 is the last year for the annual products. Right now the team is looking at how to update that for 2018 & 2019. The thing about the methodology that we are currently using, we need the entire calendar years worth of Landsat data before we can run the model it needs all the data and so we'll lag a little behind each calendar year for that product development. Now in the future the team is certainly looking at doing different methodologies that we can use to actually monitor in a shorter timeframe. So let's say for the upcoming season knowing what we know about a specific pixel and it's history, if we know we are going to expect to see drought conditions, we will be able to estimate what might happen on the landscape. Having all of these, I'm going to call it an increase in temporal and spatial frequency and a consistent product, is something that is really new and for the non-science user what that means is we have more data across geographically, across space but also through time. We are only just starting to understand how tracking those changes and integrating other datasets such as climate or other natural resources datasets, fire, modeling those types of things, what we can learn about our country.

HULT:

One thing that strikes me about LCMAP and again talking about integrating it with other products that come out of the building at EROS is that there are a lot of datasets out there. There is a lot of information out there that comes out every day or every year. There's stream gage data, there is climate data. There is a lot of stuff that is sort of instantly available but this seems like a way to bring land cover which is so time intensive and so difficult to put together with Landsat information. It seems like you are bringing that into the fold as a more timely product that you can use to understand what's going on.

BROWN:

Oh, yes. Absolutely. I mean there's a lot of possibilities for the future for LCMAP. You mentioned management earlier. I mean providing this information, this type of information on an annual basis for land managers is one of the things that we think is a big growth area.

HULT:

Can we touch on why LCMAP only goes back to 1985?

BROWN:

This is because of the frequency of the source Landsat observations that we are relying on for our process. Back in the early 1980's Landsat 4 was launched. Although Landsat 4 has similar characteristics to the later Landsats. During those first few years that it was functioning, there was not a lot of data that was saved. So when we use this CCDC process to detect change it works a lot more consistently to find change when we have a good observation record and we just don't have the observations back before 1985. We have run CCDC on data we have from 1982 forward but the early years are just not a very high quality, so we choose not to release them to the public.

HULT:

So the reason of using data from 1985 forward is really a question of quality. You want to make sure you have the highest quality and most consistent, reliable information that you are putting out there.

BROWN:

That's the goal. Yes

HULT:

What can we look for from LCMAP in the future? Are you going to look at doing Alaska and Hawaii for example?

BROWN:

Annual updating is our first priority. But after that we do plan to expand to Hawaii and Alaska next. Hawaii is a little easier to tackle just because it's a lot smaller. Alaska is a huge state. Those landscapes come with some challenges related to data quality, quantity and cloudiness and cloudiness does affect our ability to find change. I would say one of the other areas besides our applications that Jen leads is our assessments capability. This is where we really get to drive assessments and delve into understanding the products and what they are telling us about land surface change across our country so, I like the expression "every pixel tells a story" but also, it is more than that.

HULT:

We've been talking to Jess Brown and Jen Rover about LCMAP. It's been a fascinating conversation. Jess, Jen thank you for joining us.

BROWN:

John, thanks for having us.

ROVER:

Thanks John. I appreciate you visiting with us about LCMAP.

HULT:

This product is a podcast of the US Geological Survey, Department of the Interior