Eyes on Earth Episode 46 – Deforestation and Forest Degradation

ERIC BULLOCK:

I was just amazed that I could be handed a map of where there was a fire yesterday. And that I could go out and see the damage of that fire in real life. With my own eyes! That just blew my mind......

JOHN HULT:

Hello everyone and welcome to another episode of Eyes on Earth. We're a podcast that focuses on our ever-changing planet and on the people here at EROS and across the globe who use remote sensing to monitor and study the health of Earth. I'm your host, John Hult.

Deforestation and forest degradation are often most pronounced in the rainforests of the Amazon. They have far reaching impacts for biodiversity and the global climate. Landsat satellites are among the data sources used to track deforestation in remote locations., and researchers are constantly looking to refine and improve upon that work.

Our guest today is Dr. Eric Bullock of the U.S. Forest Service Rocky Mountain Research Station. As a post-doctoral researcher at Boston University, Dr. Bullock developed an open source algorithm for monitoring tropical deforestation using Landsat data. Dr. Bullock is now piloting this methodology for use as part of the United Nations Reducing Emissions from Deforestation and Degradation in Developing Countries Program., which offers incentives for reduction in emissions and sustainable practices.

Dr. Bullock, welcome to Eyes on Earth.

BULLOCK:

Thanks for having me on.

HULT:

It's great that we're able to connect and talk about this. First off, let's talk a little bit about your background. How did you come to remote sensing and how did you land on this particular area of research?

BULLOCK:

Like anyone, my road to where I am today has been rather long and windy. I was working as a forest restoration technician at the Santa Monica Mountains. That's a really beautiful park right outside of L.A. I was doing a lot of field work for restoration of ecosystems after fires, and what became really clear to me was the importance of data. Especially spatial data. What we really needed was real-time information on fire damage, specie distributions, ecosystem change. We just needed all the information that we could get our hands on while we were out in the field. I was trained as a sociologist and in GIS, so in very basic spatial analysis. There was a remote sensing scientist at the park who was providing me with data. So, all I knew was that was the data that I needed and that it was absolutely essential for doing the work that I was doing.

HULT:

So how did you get from there to working with satellite data and writing algorithms and looking into deforestation?

BULLOCK:

So that really interested me. Enough to want to pursue my master's in remote sensing. There's not a lot of programs to do that. One that really interested me was a GIS and remote sensing of environment master's at Boston University. I went there to try and learn more about how I could start to develop remotely sensed information. I just loved it. I got to be involved in a little bit of research through my master's, but it really wasn't until I started to pursue me PhD that I really started to get more involved in global initiatives to monitor forests.

HULT:

What's the difference between deforestation and degradation? And why is it important to monitor those things? How does satellite data help us do that?

BULLOCK:

Often when you see on the news or in movies, somewhere in Borneo or in the Amazon being cleared for something like agriculture or pastureland. That's really large-scale deforestation. So that's the conversion of a forest to another land cover, such as croplands. And that usually occurs over large areas, not always, but the most common of what you see in say the Amazon.

HULT:

So deforestation ... We're talking about clearing the forest and using the land for something else altogether. Often cropland. That's deforestation-one thing to another thing completely.

BULLOCK:

Right, right

HULT:

How about degradation? What are we talking about there?

BULLOCK:

Degradation is a process where forests are disturbed. Or they're affected by people through causes such as selective logging-just removing a branch here, some timber there. But it remains a forest. It's not a conversion to another land cover. It's a forest that's being impacted by people. Most of the time in a negative way or what we think of as a negative way, so removing trees or causing some sort of ecological damage. So if you look at it from above, if you look at it from the ground there's still going to be trees there.

HULT:

You still see something that looks like a forest, but it's not doing the job we would expect it to do, or performing quite as well as we would hope.

BULLOCK:

Exactly. And this is important because with the degradation of that, like selective logging. There's still a lot of the ecological consequences there, carbon being emitted into the atmosphere. The forest will still be reducing its ability to perform ecological functions.

HULT:

Why is it important to look at those two things from the sky? From above? How does satellite data help us to watch this stuff?

BULLOCK:

Satellites are really the only way we can monitor deforestation or degradation over really large areas. It's not practical to go into the entire Amazon and get accurate information on how that forest is changing. Just from a practical perspective, it is a lot easier to obtain information on where deforestation or degradation is occurring from satellites than it would be from the ground.

HULT:

And I suspect that even if you did have the thousands or hundreds of thousands of volunteers that you needed might not be super easy to get into the depths of a rainforest. 

BULLOCK:

Even the United States, non-tropical areas, where we might think that forests would be a bit more accessible and the weather is less extreme, there are still very remote areas. It's just a logistical nightmare to try and get to or it's impossible.

HULT:

So logistically on many, many levels, it makes the most sense to look at it from above. From the sky. Let's talk about how you do that. The approach that you've taken. Tell us about the Continuous Degradation Detection algorithm. Which is CODED. Is that how you say that?

BULLOCK:

Yep

HULT:

How did you develop this? What does it look for? What kinds of data go into this algorithm and how has it been applied?

BULLOCK:

CODED comes out of what has really been a paradigm shift in the remote sensing community towards the use of time series analysis. The basic idea is that instead of just using one image or one bit of data from one year and comparing it to five or ten years later, you're using all of the data that you have available to you. What CODED does is characterizes what the normal state of a stable forest is through time, and then automatically detects the subtle deviations from that trend, then labels them either degradation or deforestation.

HULT:

So, you're looking for very specific things. You're not looking to characterize land cover or maybe the time of change, you're looking at some pretty specific parameters to determine whether there is degradation or deforestation?

BULLOCK:

It's looking at what you just mentioned, as well. It's also trying to determine the time of change. If it's a deforestation event, what is that land cover after the event? Does it convert to a pastureland or croplands? What it was really designed to do was identify these very subtle changes from selective logging or from understory fire that can be challenging to identify if you're just looking at one or two images.

HULT:

What did you learn by applying CODED to the Amazon?

BULLOCK:

I've now used CODED in over 20 countries. What we're finding has been pretty consistent everywhere. The more you look in detail for these subtle degradation events, the more change that we're finding. So in the Amazon for example, we found that there was between 44 and 60 percent more area of disturbed Amazonian forest. Previous analysis had mostly ignored these degradation events, either for practical purposes or because it was not the focus of that study. This is similar to what we're finding elsewhere. When you account for degradation, a lot more of the forest is disturbed then what we previously thought.

HULT:

Interesting. So, a big part of the reason that you found more is because you were looking harder?

BULLOCK:

Exactly

HULT:

It's my understanding that you previously worked on a deforestation algorithm using NASA's MODIS sensor. Can you tell us about that? How did, or how does that work factor into what you did with Landsat? Or what you're doing with Landsat?

BULLOCK:

So what we were just talking about with CODED has primarily been done using historical data. Between 30 and five years ago, for example, just trying to understand how much of the forest has been disturbed. Landsat data is recorded between every eight and sixteen days. Whereas, MODIS data you get a new image of almost everywhere in the world every single day. This is ideal for near real-time monitoring. We developed an algorithm that used daily MODIS data to create alerts of forest disturbance. If you are trying to mitigate illegal deforestation, if you had an alert of disturbance as it was occurring, you could theoretically go out and try and stop that event. We tested this in the Amazon and it worked well, really well actually. What we are currently trying to do is create a similar system, but instead of using this daily Modus data that is rather course and therefore you can't find real small-scale clearings, trying to use a combination of different sensors that are higher resolution to try and replicate that daily repeat cycle of MODIS, but at the spatial resolution of something like Landsat. 

HULT:

So, to summarize at least, so that I can wrap my head around it. With the Landsat work, you were trying to figure out what has happened historically and how much change there has been. And then with the MODIS work you were really trying to find out what kind of change is happening right now. You put those two concepts together and at this point your looking at, I'm guessing, maybe Landsat and Sentinel. European Space Agency Sentinel 2 satellites. You're looking at putting those together to do what you were doing with MODIS, but with finer resolution so you can see more change. Is that kind of where you're at now? Is that right?

BULLOCK:

That's a good summary of it. Yep

HULT:

What data sources or data products are on the way that you see as particularly useful for monitoring deforestation?

BULLOCK:

I'm really excited about Harmonized Landsat/Sentinel. Sentinel 2 is a couple sensors from the European Space Agency that records very similar data to Landsat. HLS is a product that as a user you'll get an image that looks like a Landsat image or a Sentinel 2 image. You won't need to know which sensor it is coming from. It's comparable to the Landsat data we've already been working with, so we can input it directly into our analysis and just get much, much more data.

HULT:

So, the collaboration between USGS and NASA on the Landsat side and ESA - European Space Agency on the Sentinel side is really benefiting everyone.

BULLOCK:

Exactly. And that will be really useful for many applications. For one would be near real-time monitoring in which we want really as much data as possible and as fast as possible. 

HULT:

Hey, guys. This is John. Just jumping in to let you know that if you want to know more about Harmonized Landsat Sentinel data, you can check out episode 45 of Eyes on Earth, which is available on the show page at usgs.gov/eros, or through Apple Podcasts or Google Podcasts. Now, back to the show.

BULLOCK:

There is a lot, lot, lot of cool sensors and technology that are coming out now. There is one specific sensor that I am really excited about called GEDI. GEDI is onboard the International Space Station. And it is a lidar sensor. What lidar does it uses laser light to basically measures distance. So it can be used to study different attributes of a forest that you can't get with an image like from Landsat. This will be really useful, not so much for trying to identify deforestation or change of any sort, but for learning about the different properties of that forest. GEDI's going to be used to create a pretty high resolution, high quality forest biomass data set. Which will be really useful in trying to better understand carbon implications from cutting down the forest.

HULT:

So you're not necessarily going to be finding out whether some logging is happening from GEDI. But you're going to be able to use that data set to know how much biomass is there. Sort of the composition of the forest, and maybe make some determinations about the impact that the change that you're seeing might have.

BULLOCK:

That's a great summary. Now the other general type of new data that I'm excited about is radar data. We've thus far been talking about optical imagery, which is basically like taking a photo from space and doing an analysis on that. That's what Landsat is. That's what MODIS is, Sentinel-2 It's really good for identifying change on the landscape when there is not clouds. Because if there is a cloud, that image you are taking from space is really taking an image of the cloud. In the tropics, it is very frequent that there is cloud cover. Right now, I am working on a project to incorporate not just different sources of optical imagery but incorporate radar data through a sensor called Sentinel-1. That is a C-band radar data. A radar beam can penetrate through a cloud. Reflect off of a forest or off the canopy or something like that, and return to the satellite without being horribly affected by that cloud.

HULT:

You're looking to sort of fill in the gaps that the clouds create with data that sees through the clouds.

BULLOCK: 

Exactly.

HULT:

Dr. Bullock, you said that you have now applied the CODED algorithm to 20 countries around the world. Tell us about the UN's Reducing Emissions From Deforestation and Degradation in Developing Countries Program. What role do you play there? And where have you taken it so far?

BULLOCK: 

The general idea is that if a country could prove that they're reducing their forest-related emissions, they will receive monetary compensation through the United Nations. In order to do that, they need to be able to monitor their emissions from deforestation and degradation, to prove that a, this is what their emissions used to be, and b, now they have taken actions to reduce those emissions, and this is what they are now. We've piloted CODED now in a couple of countries West Africa, the Pacific Islands. It is shown to be a pretty effective tool in getting this really necessary information on forest change, deforestation and degradation.

HULT:

Essentially, we have a program that will pay for doing that right thing and reducing emissions. But in order to figure out whether those emissions have been reduced, we have to know if deforestation and degradation has occurred. What you are doing with this publicly available data is to create an algorithm that has that measuring stick, So they can say 'here is what we're doing and here is the difference it has made,' to make those determinations.

BULLOCK:

Measuring stick is a good way to put it. Yeah.

HULT:

So, that's what you're doing for this U.N. program. But you work at US Forest Service which is kind of interesting. How does that translate into something that could potentially be useful for the Forest Service or useful state-side?

BULLOCK:

The Forest Service monitors for change in U.S. forests. The algorithms that I develop or that my colleagues develop, could then be applied in the U.S. to better monitor or change what is happening here.

HULT:

Where would you use this in the U.S.?

BULLOCK:

It kind of goes back to my reason for getting into remote sensing in the first place, right? I was a restoration technician that required real-time information on fire damage. That is a really big problem that is only getting worse in the United States. The algorithms that I help develop could be used back in the United States for looking at disturbance from fire. Pest damage is another one-beetle damage-or just for getting a better handle of the state of our forests, even the ones that are not changing. We talked about lidar and how that could be used for estimating biomass, and that's also incredibly relevant in the United States. We have a vast inventory of field plots, where forest service scientists are going out and measuring trees every couple years. If this could be done using satellites, it would save a whole lot of people a whole lot of time. And it could possibly enable us to have an even better understanding of our own forests.

HULT:

So ultimately what you are working on here, if it were applied to the U.S., would be some data that would be useful to the younger Eric. To someone who is out in the field and needs this information. And then, as you said, just the idea of using remote sensing helps you to fill in those gaps and you don't need to have as many boots on the ground. Or if you do, you can sort of use it to validate the data. I mean, it is useful all around.

Dr. Bullock, any closing thoughts? Anything you would like to make sure that we talk about on this podcast?

BULLOCK:

Yes, I do just want to say, and you touched upon this already, you know all of the research that I get to do wouldn't be possible without free and open data policies like at USGS , NASA and European Space Agency, who are both creating this data and making it very accessible to scientist like me. I do think it is incredibly important to give credit where that is due, because it's really, really remarkable that we get all this high-quality data to work with.

HULT:

We've been talking to Dr. Eric Bullock about deforestation and degradation and how remote sensing can help us track change in the world's forests. Dr. Bullock, thank you so much.

BULLOCK:

Thanks for having me on.