One of the more impressive earth observation achievements in recent years was the collaboration between Google and the University of Maryland to create a global forest change assessment and visualization. This effort was unprecedented in terms of the raw computing power applied to the substantial Landsat archive, and yielding results that opened a lot of eyes to the change occurring around the globe. Sensors & Systems (S&S) special correspondent Matteo Luccio recently interviewed professor Matthew C. Hansen of the Department of Geographical Sciences at the University of Maryland about this effort that he spearheaded as well as the state of earth observation, global monitoring, and the need for Big-Earth sustainability efforts.
S&S: What first inspired you to become a scientist? Why did you specialize in remote sensing?
Hansen: I was into service. I was a Peace Corp volunteer in Zaire and was trained as an engineer. I did fish farming. When I came back from the Peace Corp I decided to go into geography for the simple fact that I like the spatial domain; in particular, I like atlases. I didn’t even know what kind of job you could get in geography, but I knew that I had a natural inclination for it. Within geography, I wanted to do something applied, not basic science. I chose remote sensing, which is very applied: you can clearly see what’s happening at a large scale with landscapes, land use, the way cities grow, the natural hazards therein, the way croplands are healthy or not, the way forest resources are used—it’s just a whole great perspective. My analytical or scientific background, coupled with the idea that this could be a really nice applied information domain, is how I got into it, but the key thing is that it was spatial.
S&S: What is your definition of geography? What is remote sensing’s role in the study of geography?
Hansen: For me geography is the study of space to understand how we interact with the natural world. It’s coming from the human perspective, because we are the biggest change agents on the face of the Earth. When I was growing up, they said, “The only object you can see from space is the Great Wall of China.” Actually, all you can see from space is human activity. What’s not modified by humans is a more appropriate question, and there’s very little of that. So, for me it’s the study of the natural space and, more importantly, how humans interact with it. It’s analogous to history, except in space: history is to time what geography is to space. It’s too big a field, really.
Remote sensing gives us a really good, internally-consistent set of facts from which we can quantify these different changing things in geographical space. So, this is a beautiful, wall-to-wall kind of dataset that is not influenced by other strange things that you might find in a set of samples. It is clean. It is a physical, quantitative, wall-to-wall measurement of the Earth. It’s just awesome.
S&S: Compared to 25 or 50 years ago, have only the tools changed or have the tools—GPS, GIS, advances in remote sensing, etc.—dramatically changed geography as a field of study and as a profession?
Hansen: There is no question that we have more data, more information, more analytical and quantitative tools at our disposal. Some geographers think that if you do GIS or remote sensing you’re just a technician. No, these are information domains that allow us to more quantitatively approach geographical problems and answer them. We can ask more questions knowing that we have better data sources and better tools to examine those data. It has changed the way we ask and answer questions, but I think it’s also been a bit disruptive. Some of the more traditional people think quantitative tools and approaches are not legitimate geography, and I think that’s wrong. For sure, it’s made the discipline more valuable, the degree more valuable, because you can go out into the real world, in government, civil society, the commercial side, and solve real problems, study markets and connectivity and things like that. The new tools and data have made geography incredibly more relevant to the real world.
S&S: Operators of EO satellites, such as Landsat, like operators of other large data collection systems, face a constant dilemma: should they continue to use the same hardware, software, and methods so that the data can be meaningfully compared over long periods of time or should they adopt the latest scientific and technical advances that will improve the quality of the data? As a user of data from EO satellites, what advice would you give to satellite operators on this question?
Hansen: Landsat is important in terms of continuity. The idea that you have a very similar measurement over the long term is really important, because if you drastically change the scale of your measurement or the bandwidth you are using, you lose continuity, and then it’s hard to draw conclusions about how the Earth is changing. That’s really important. Of course you can augment that with other measurements, we can go down to finer spatial detail, and maybe even eventually come up with Landsat-like systems at different scales or even in different information domains. This is huge.
We need to move towards operational land monitoring in general. We don’t have that yet. When you turn on the TV at night, the weather image is always there. You always see it, meaning it’s operational. That’s for good reasons, because it is important for safety and commerce and all that. But the land is getting a lot of pressure too, and we don’t have, strictly speaking, similar land-monitoring capabilities. Moving in that direction is really important. Software and hardware can change, even drastically, but if you have an archive of consistent measurements you can always reprocess that archive with a new software tool, a new algorithm, and with new hardware. What we really need is the quality, continuity, and consistency of the data to monitor how our land is changing.
When Apollo 17 took that picture of Earth against the black background of space and it became an icon of the environmental movement— the Blue Marble—in 1971, that was when the first Landsat was launched. We were shipping Landsat images in the 70s primarily as glossy photos, because we couldn’t efficiently process them digitally. The biggest photo lab was the EROS archive out in South Dakota. With the 80s came TM and better computing. In 2008 they opened up the archive and now we have cloud computing.
All these things came together that allowed us to create this 30m map of global change 40 years after the first Landsat was launched. That archive is not perfectly consistent. It was only in 2000 that we started getting global acquisitions, so Landsat has changed amazingly over time. Starting in 2000, with Landsat 7’s global acquisition strategy, we’re going to have these great records and we’re going to keep following it. Before 2000, Landsat data are great, but kind of spotty. We took 40 years to put all this stuff together, but at least the data were there.
S&S: What are some of the specific ways in which remote sensing of large regions can help us understand climate change?
Hansen: You can look at that in a number of ways. One is disturbance of forests and particularly of tropical forests. Between 10 and 20 percent of the anthropogenic contribution to climate change comes from land use change and most of that is in the tropical forest—the clearing of old growth, high biomass forest. So, tracking that consistently with the same observational framework, like Landsat, allows us to build baselines, to quantify more precisely what the rates of change are, and to model those emissions. That’s a clear one.
With long-term records from Earth observation, we can see the way different natural systems are responding to climate change and maybe get hints of eventual outcomes. For example, whether we have high-latitude vegetation greening, more woody vegetation up there, whether we have more severe and more frequent droughts in the tropics that were largely absent in the recent past. You can see that with satellite data. So, there are many ways in which we can track outcomes of climate change effects on the land, but also the drivers of the climate change itself through perturbations of the landscape.
S&S: What has MODIS accomplished?
Hansen: Boy, MODIS is awesome! MODIS, first of all, was the first instrument of its type that had land bands. It had seven land bands, so it was a fantastic instrument for observing the land surface. It has oceans, land, and atmosphere teams. I’m on the land team. The idea in the land domain was this daily-acquired set of observations that were systematically processed, gridded, and turned into very user-friendly formats. That enabled a ton of people to get into the game of doing science with these data. They learned a bit from the Advanced Very High Resolution Radiometer (AVHRR). We had some AVHRR standard products. Global scale, time series standard products allowed people to investigate a ton of different questions. MODIS set a standard in terms of the organization of these science teams and the systematic processing and calibration and provision of uncertainty of both spectral inputs and derived products, different levels of products—levels one, two, three. It’s just a cool system.
What I’d say about MODIS and NASA datasets in general—and now Landsat, which is a joint program of NASA and the U.S. Geological Survey (USGS)—is that to make your data really useful for large area analysis, and certainly for global analysis, you have to have systematic global acquisitions like clockwork. You have to have the data freely available, which they are, and they have to be accessible, which they are, with really fast and efficient portals on the Internet. You can get all the data you need, the whole archive if you want. You also need some pre-processing, to take the burden of a lot of the corrections off of the user who has a science question to answer. They don’t want to do geometry, they don’t want to do radiometry. They don’t have to. No other systems around the world meet those criteria. It’s really NASA and the USGS that have prototyped this approach.
S&S: MODIS checks all those boxes?
Hansen: Definitely, it does. Now, since 2008, Landsat checks those boxes too.
S&S: What are you hoping to accomplish by taking the global processing model for MODIS and applying it to the Landsat archive?
Hansen: When we map forest loss disturbance with MODIS—and we do it conservatively, to try to avoid commission errors—we see about half of the change globally that we see with Landsat. With Landsat, we can see logging roads and small-holder agriculture dynamics in developing countries that we can’t see with MODIS. Many urbanization effects are very small-scale and not huge change dynamics; we see that with Landsat. Landsat, at 30-meter scale, blows away MODIS in terms of spatial detail. So, if we take the same processing we have with MODIS and apply it to Landsat, wow! Our application base also gets a lot more interesting. For example, in terms of modeling and habitat and things like that; biodiversity groups were never interested in our MODIS products, because they were too coarse. Now, however, they are one of our strongest advocates and users of our Landsat products.
S&S: Can the monitoring of cropland using MODIS, Landsat, and RapidEye data sets help individual growers make decisions?
Hansen: That’s a good question. One of our goals, for example with the Landsat forest products, is to have a globally-consistent, locally-relevant product. In other words, if you cut out any particular part of the globe, it will have meaning and utility to someone at that scale. With agricultural products we’re not there yet, where we do wall-to-wall mapping at Landsat scale, and much less at Rapid Eye scale. We’re working toward that, but if we got into that kind of scale, then the local relevance would jump out at us.
Right now we do sample-based characterizations of crop type because we’re kind of limited in the time domain to make wall-to-wall maps. There’s clearly some work to be done there. I don’t need high time temporal richness to map forests and how they’re changing. However, to map agricultural crops—for example, to differentiate cotton, corn, soybean, and what-have-you—I do need that and we don’t have that wall-to-wall at finer spatial resolutions, so that limits a little bit the local relevancy of it. The high-end farmers have precision ag technology, including yield monitors on their systems, so they know what the heck’s going on in their fields. I don’t know how much these datasets will help them individually.
These datasets will help us understand what’s happening on a national, regional, and global scale by helping us characterize a particular crop and how it’s doing in different regions at different times of the year. So, if we use these things smartly we can take some of the uncertainty out of agriculture production estimates.
S&S: How will the advent of micro satellites and cube satellites change remote sensing of large regions?
Hansen: If you don’t have a proper data policy, it’s not going to change it very much. In other words, if it’s really a high-cost model to get into the game, large regions are going to cost too much money for those people. The big question is how to take a MODIS-like Landsat and Landsat-like data policy and get it down to the fine-scale of these systems. Clearly, constellations at the fine spatial scale are going to be awesome by giving us high temporal or high spatial resolution information. That’s the Holy Grail, at least in the optical domain. If you could have daily 5 meter data, wow, that would be cool. The application space will really blow up. The question is, what’s the data policy attached to that? Is it going to enable our large-area analysis or not?
Also, are these satellites going to have the calibration you need? Landsat’s expensive, but man, it is an awesome instrument! Are you going to have the same quality observations that allow you to mass process data, to map large regions? If the calibration is not good, you’re not going to do it. So we’ll need to resolve some technical challenges as well as data policy and some data access issues.
S&S: What are the key scientific questions and technical challenges in remote sensing today?
Hansen: From my perspective, sustainability of the Earth system is a big one. When we have growing populations and populations that consume more per capita, how do we make for a sustainable planet? Earth observations are going to be really important to doing that—both tracking how successful or not we are, but also building up records that allow us to model different scenarios by having real data that can be used for cause and effect. Doing that requires global data—whether it’s to model climate change or global biodiversity, or to analyze population changes. Big Earth-system sustainability is where I think it’s at.