When Alfredo Huete saw Scott Saleska’s poster presentation at a meeting of the American Geophysical Union in 2002, he felt like he had been vindicated. Several years before, Huete had been sponsored by NASA to develop techniques for mapping global vegetation using data from a new sensor planned for two of the space agency’s upcoming Earth-observing satellite missions. For several years after Terra, the first satellite, launched in 1999, the University of Arizona remote-sensing ecologist had been worrying over the data processing and mapping technique he and his team had proposed.
For nearly two decades, scientists had been mapping global vegetation patterns using a vegetation scale, or index, based on data from a series of satellite sensors operated by the National Oceanic and Atmospheric Administration. The NASA sensors built on and even surpassed the capabilities of the previous sensors, but still, Huete had to deal with a new kind of satellite sensor, a new method for producing the vegetation maps—and the awareness that he was making a product that would go out into a global research community with NASA’s name on it. “I felt a lot of pressure,” Huete says.
As they were developing and testing their technique, Huete and his team frequently checked that the maps matched real-world seasonal changes in vegetation in different ecosystems, from African savannas to eastern North American forests. Although their maps captured the expected seasonal changes in most areas, one area bothered Huete: the Amazon. As data from Terra began accumulating, he noticed something peculiar: the Amazon rainforest looked greener to the satellite in the forest’s dry season than it did during the rainy season. Huete knew that parts of the forest go several months with little or no rain. How could the forest be thriving during those times of seasonal drought?
Before the 2002 conference Huete had spent several years repeatedly tinkering with the data and the mapping technique. “When you see something you are not expecting, you have to ask yourself, ‘What are all the possibilities for a remote-sensing product going wrong?’” Among the possibilities are things in the atmosphere that keep the satellite from having a clear view of the surface. “We checked for aerosols [particles in the air, such as smoke from biomass burning] and clouds, which can potentially reduce the vegetation signal obtained by satellites. Someone suggested that maybe there was flooding on the forest floor during the wet season, so we looked at that. We looked how the vegetation maps changed if the light [hitting a particular patch of vegetation] was direct or diffuse. We just kept re-doing and re-doing the data products,” he says. Each time they made a change, they wondered if the dry-season green-up would disappear. But with each refinement, it stayed. His confidence grew, but Huete still wasn’t sure. Was this for real? Or was it just a sign he was still doing something wrong?