Terry Sohl is fairly unique among U.S. Geological Survey (USGS) research physical scientists at the Earth Resources Observation and Science (EROS) Center in that he spends less time looking at the past and present, and more time peering into the future.
As the Land Resources Mission Area representative for the USGS’ Earth Monitoring, Analysis, and Projections (EarthMAP) initiative, the projections lead for of the Land Cover Monitoring, Assessment, and Projection (LCMAP) initiative at EROS, and the head of the FOREcasting SCEnarios of Land-Use Change (FORE-SCE) project at the Center, Sohl is all about future possibilities in water availability, land use and ecosystem change, mineral and energy stores, and more.
He sat down recently to talk about that work.
Note: Hear Terry talk about this topic by clicking here for his episode of the EROS podcast "Eyes on Earth"
You helped develop the FORE-SCE model. What is that, and what does it do?
“This is work that started about 12 years ago. The idea was, EROS has this wonderful archive of remote sensing data that can look at current and past land cover. What we wanted to know is, is there a way to take these data and project what may happen in the future or even go back in time before the dates where we have remote sensing data? The FORE-SCE model was developed specifically to take advantage of USGS land cover data, use that to parameterize the model, and then kind of stretch out these land cover databases that we produce for the current and the past.”
To project how land cover/land use may change going forward, you have to understand the past?
“We base our model very much on the historical land cover databases that are produced at EROS. So, we have these national scale projects, such as the National Land Cover Database (NLCD), that measures land cover change over time. Those form a primary basis for the model. We also try to model back in time to get a better understanding of what’s happened in the past, because that helps us to predict what happens in the future. When we model back in time, we can really go back as far as we want. It’s just the uncertainty increases because the data quality gets less as you go back in time. But we can use things like Agricultural Census data from the USDA (U.S. Department of Agriculture). We can look at Census data from the past. We can look at old satellite imagery, or old aerial photography, even going all the way back to the 1930s to help us parameterize the model.”
How do you assess your accuracy?
“What we try to do is compare what we’ve modeled with a past period. If we model out into the future, what we do is we typically start the model about 15 to 20 years in the past. The reason we do that is, we can look at that 15- or 20-year period and compare the model performance to actual performance of land use on the ground as measured from remote sensing data. So, we kind of calibrate our model that way.”
Can you go back to 1938, 1945, model what it looked like then, and be comfortable that it’s accurate?
“The uncertainties definitely do increase, and your ability to validate your model also decreases as you go back in time. But you can still do some basic comparisons. If you have something like Agricultural Census data, that’s a county level product. And what it does is, it tells you how much corn or how much soybeans or how much other agricultural commodities were in a given location at the time. What we can do is, we can compare the overall quantity that we modeled with that to see if we’re matching the actual overall quantity. We can also look at spatial patterns by looking at things like aerial photography from the past.”
How important are remote sensing systems like Landsat in all this?
“They are vitally important. One thing that makes our model unique is our heavy reliance on remote sensing data. That’s why it’s a perfect location here at USGS EROS to do this work. The model is so heavily reliant on looking at Landsat imagery and other satellite imagery that we get in the building to look at spatial patterns, for example. So, historical spatial patterns of change we can look at from a historical perspective and use that to parameterize our model for how things will look in the future. Or rates of change. Or overall sectors of change. How much corn is coming from grassland? How much corn is coming from forest? Those types of transitions that we can measure in the past, directly inform the model for the future.”
How do you get to the reasons why land cover may change in the future?
“The driving forces of change are really what are the most important factors with land cover modeling. What we can do is, we can take two dates in the past and extrapolate the past out into the future and call that a projected land use model. But those aren’t really what are valuable. What’s really valuable is understanding the economics, understanding the demographics. How are people moving on the landscape, or how are migrations changing populations in a given area? Or climate. How is climate impacting the landscape? And so, those driving forces of change and understanding them are really a key to the model, just as much as the historical remote sensing data is.”
Who uses this information, and how?
“We have partners from across the Federal agencies. The EPA (Environmental Protection Agency) has used our work. We have groups like Audubon from NGOs (non-governmental organizations) that use our work. We have quite a few partners in academia right now. I’ll give an example of one partnership, a project that was funded by the National Science Foundation with the University of South Dakota, the University of Wyoming, and Montana State. It’s looking at biofuels in the Northern Great Plains, how biofuels are tied into economics, and how these scenarios of biofuels going out into the future could impact things like bird populations, climate, weather, and hydrology.”
When you project into the future, how many scenarios do you provide?
“It’s a little bit different than being a weather forecaster. We’re not trying to predict the future. What we’re trying to do is say that there’s a lot of uncertainty ... things like the economics or demographics. Energy, that’s another one. Where’s your energy going to come from in the future? Is it going to be renewables, or is it going to be fossil fuels? All these things have a huge impact. They all interrelate, and they all make it very difficult to predict what’s going to happen in the future. So, what we do is, we produce these scenarios. We’ll have multiple scenarios that try to capture that uncertainty so that if I’m a farmer in eastern South Dakota, I can look at a climate change scenario, for example, that’s quite severe, and try to figure out how I might adapt to that. Or, I might have a less severe climate change scenario. So, having multiple scenarios helps people to adapt to potential change, and hopefully to mitigate any negative consequences before they actually happen.”