BUENOS AIRES–There may be nothing more depressing than watching a deforestation map in real time, knowing that each time a green pixel turns red, the corresponding square of earth has been denuded of trees.
That must make the folks at Terra-i some of the biggest sadists (or masochists) in the world, as they programmed a system that lets you do just that.
Launched at the Rio+20 United Nations Conference on Sustainable Development in June, the Terra-i system uses NASA satellite data to create image frames that, when sequenced together in video, show real-time deforestation in Latin America, from Mexico to Argentina. Developed over more than three years, Terra-i is run out of the International Center for Tropical Agriculture (CIAT) near Cali, in southwest Colombia, with help from the HEIG-VD in Switzerland, and the Nature Conservancy, which funds it.
The way satellite deforestation monitoring systems like Terra-i work is fairly straightforward, although they do require an immense amount of computing power. Satellite deforestation monitoring is, at base, an exercise in color-coding. Using an instrument called a MODIS (or Moderate Resolution Imaging Spectroradiometer), NASA’s Terra and Aqua satellites take daily images of the earth’s surface. By measuring the greenness of the vegetation in each pixel of those images–technically, the ”greenness” is measured on the Normalized Difference Vegetation Index (NVDI)–and then following how that changes over time, scientists can see which areas are being deforested.
“All vegetation changes in greenness on a normal cycle with rainfall and drought. So what deforestation monitoring systems have to do is look for changes that aren’t the normal pattern of greenness change for the season,” said Mark Mulligan, a reader in the department of geography at King’s College London who invented an early monitor, Conservation Eye, which used the NASA data to track deforestation on an annual basis.
What Terra-i did was take systems like Conservation Eye to another order of complexity. Because clouds often make parts of the images unusable, NASA combines the four daily images provided by the Terra and Aqua satellites into one cloud-free mosaic image every 16 days. Using those images from 2001 to 2004, the Terra-i team–led by Louis Reymondin, Terra-i’s Swiss main developer and a PhD student working under Mulligan at King’s College–crunched the NVDI numbers to determine the pattern of changing greenness over the course of a normal climatic year for each pixel. The Terra-i team then took that knowledge to calibrate a neural network that then used the data from 2005 onward to create images of which areas were being deforested (they appear as red in the maps).
“Whenever we see, every 16 days, a change it greenness, we know if it’s inside or outside of normal range of greenness,” Mulligan said. “If it’s outside the normal range, then we flag it as a ‘pixel of concern’. And then if after another 16 days it’s still outside the normal range we can be pretty sure it’s probably a deforestation effect.”
In the videos produced by the Terra-i process, pixels flicker from green to red, a color that then usually spreads like an inkblot (each pixel represents a 250m x 250m patch of land). These videos of destruction are beautiful in a cold, horrible way, and seeing one is like watching a silent aerial film of Dresden being carpet-bombed (albeit with Paraguay playing Dresden and the carpet bombs replaced by soybean plantations).
The amount of deforestation Terra-i shows since 2004 is staggering. The greatest amount of deforestation has been, not surprisingly, in the Brazilian Amazon, and there has also been a rapid increase in deforestation in Caquetá, Colombia. But the most significant increase has been in the Paraguayan part of the Gran Chaco, an oppresively hot lowland area that crosses the borders of Paraguay, Argentina, Bolivia and Brazil. In the Paraguayan Chaco, about 5 million acres–or about 14%–have been deforested since 2004, says Reymondin. About 3.5 million acres, or 7%, of the Argentine Chaco has been deforested over the same period, largely to make way for cash crops.
“The models are not able to tell us what’s now on the fields. We know that there was forest and now there is not,” Reymondin said. “But from literature and discussions with other organizations, it’s almost all about soy.”
Besides deforestation itself, Terra-i clearly shows the effects of policy and political will. In Paraguay, the Defensores del Chaco National Park is surrounded by deforested areas but there is almost no clearing in the protected zone. “This is something positive, even if in a few years the protected area will be the only part remaining,” Reymondin said.
And in Brazil, the border between the states of Acre and Rondônia show vast differences in the importance state governments have given to environmental protection. “Acre put in new road but implemented sustainable policies,” Reymondin said. “There’s almost no impact in Acre in the protected areas near the road, but in Rondônia, where the situation is different, you can see that some of the protected area is almost completely destroyed.”
In the end, Terra-i’s participants see their mission as opening eyes. In the past, governments and other organizations relied on reports from the Food and Agriculture Organization of the United Nations, which come out every five years and offered national scale statistics, not real time maps of what was being cut and where.
“Clearly deforestation is a complicated problem where a lot of it occurs in order to provide agricultural land to feed people. So it’s a difficult issue, but we hope that by having information really easily accessible, people will be able to see it and understand it close to real time,” Mulligan said. “What you really need is to know where it’s occurring and then why it’s occurring. Is it individual farmers who have no other options, in which case it’s the poverty issue you’ve got to address. Or is it large scale industrial deforestation that’s unregulated, in which case you’ve got to look at regulation.”
Going forward, Terra-i hopes to expand to tropical areas around the globe, and it has already done some preliminary work in Central Africa and Borneo. To make an expansion possible, it will have to increase its computing power and workforce. Right now, it follows data from between 800 million and 1 billion image pixels in Latin America, Reymondin says, and has an annual budget of about $200,000.
Photos courtesy of Terra-i.