April 27, 2018
Spread across a desk, a sequence of satellite imagery photos lie in neat rows across a desk. Mutlu Özdoğan carefully takes a photo and brings it to eye level for a closer inspection. His eyes quickly sweep the image, left to right, row by row, searching for any distinguishable land features. This used to be a lot of his work, though he doesn’t do so much of that anymore because he trains computers to find the patterns for him.
Özdoğan is a professor in the Nelson Institute for Environmental Studies and Department of Wildlife Ecology. He is a geologist by training, but he’s since adapted his expertise to land use and land conversion and how that impacts water and energy usage. Now he uses satellites and computer algorithms to understand how human beings are affecting and managing agricultural and forested landscapes.
Though computers do much of the work now, Özdoğan still needs to train his eyes to recognize landscape features, part of a technique in his field known as pattern recognition. While he declares humans the better pattern finders, he looks towards computers as the future of Earth observation data interpreters.
“Humans are the best image interpreters ever, but we’re really slow. So we use computers to analyze large volumes of data,” Özdoğan said. “My expertise is to write computer algorithms and come up with ideas to extract environmentally relevant information from satellite imagery.”
For human beings, our ability to discern shapes and trends among data is referred to as pattern recognition. But for a computer, categories like water and land do not exist unless it is programmed to understand those items. With the proper input, Özdoğan says humans can teach computers to understand patterns in images by providing a set of initial training data to use as an example—a process known as machine learning. He will give the computer numerous examples of land, water, vegetation and any other land cover types to act as parameters for his categories, then he’ll use a computer algorithm to generate a sort of map from those inputs. “The beauty of it is that instead of mapping just one field, we can do it for millions of fields,” Özdoğan said. “And that’s how we mapped the global croplands.”
The data Özdoğan uses for his cropland research is provided by Earth-observing satellites in space, like Landsat 8 and MODIS. Both of these satellites follow near polar sun-synchronous orbits, meaning the satellites will observe the same spot on the Earth at the same local time, which is very advantageous for his research. This allows him to access a historical archive of data for any location and track how a site has changed over time as far back as up to about 40 years ago.
Satellites have another unique property that deals with how they observe. Satellite sensors work similarly to the human eye by sensing reflected solar light from objects, but satellites can see a little better than humans can. In addition to the visible light spectrum from blue to red, satellites can record in the infrared portion of the light spectrum as well. Each object reflects a unique spectral signature and Özdoğan can use that detailed profile to determine what’s on the ground, even distinguishing between what kind of crop is growing in a field.
By enhancing his research with satellite images, Özdoğan has been able to see the world at a global scale and work on an international collaboration to map the world’s croplands. The map resulted from NASA grant funded work titled Global Food Security-Support Analysis Data @ 30-m (GFSAD30) project.
The focus for this NASA project is global food security. By knowing where our global food production is centered, Özdoğan said we can better understand how to manage it. However, he also noted that these agricultural uses of the land take tolls on the environmental health of the landscape. He recalls a saying by Al Gore that refers to agricultural land use as the other “inconvenient truth” for global change, which he says is an accurate description of what’s happening to the Earth.
Overall, Özdoğan says mapping these agro-intensive land changes, especially over time, can show just how much food production alters the land. He hopes to change other people’s perspectives and show how individual action can have a greater impact on the environment.
“One of the most important ways humans have changed this planet is through agriculture,” Özdoğan said. “We have a growing population that needs to be fed and now we’re busting at the seams in terms of the good agricultural land. People need to recognize that their choices can majorly transform the landscape.”