With tributaries extending from the vast savannas to itsnorth and south, the Amazon River runs almost 4,000miles (1 mile equals 1.6 kilometers) across northernSouth America from the highland biomes in the foothillsof the Andes Mountains to the Atlantic Ocean. It carriestwenty percent of all river water discharged into Earth’soceans—ten times the volume of the Mississippi River. Ifthe Amazon River Basin were draped over thecontinental United States, it would cover more thanthree fourths of the country.

Introduction: The Large-scaleBiosphere-AtmosphereExperiment in Amazonia

Part 1: Escape from theAmazon

Part 2: From Forest to Field

Part 3: Stealing Rain from theRainforest

Part 4: Defying Dry: AmazonGreener in Dry Season thanWet

From December to May each year, torrential rains andsnow melt from the Andes increase the main riverchannel’s depth 30-45 feet (9 to 14 meters), and waterbacks up in tributaries and inundates forest severalmiles from the main channel. In the central AmazonBasin alone, the flood waters can cover an area up to97,000 square miles. The river and the flooded foreststhen come together as a giant, slow-moving swamp.Surrounding these waters are over 2.7 million squaremiles (7 million square kilometers) of lush forestexploding with life. In fact, perhaps as much as one halfof all life forms on the planet live in the Amazon RiverBasin.

The Amazon is more than a habitat, however; it is also aclimate regulator. Located near the equator, where thesun’s daily rays are most intense, the uninterruptedexpanse of lush vegetation absorbs incoming radiationand keeps things cool. The forest also absorbs and storesmoisture. The Amazon forest canopy is so dense and sobiologically productive that scientists have alsorecognized the region as a key component of the globalcarbon cycle. The continent-spanning tracts of forestinhale tons of carbon dioxide during photosynthesis andexhale oxygen. With respect to carbon, however, theseforests aren’t all take. Through deforestation,decomposition, respiration, and export of organic andinorganic matter to the oceans, they also give.

The Amazon River runs almost 4,000 milesfrom the Andes Mountains in the west tothe Atlantic Ocean in the east. The riverbasin plays a key role in heat, moisture,and carbon cycles both regionally andglobally. The region is also the mostbiologically diverse location on Earth,supporting perhaps half of all species onthe planet.

The dense vegetation of the rainforestcanopy is a sink for atmospheric carbondioxide.

In this era of heightened concern about the relationshipbetween the build up of atmospheric carbon dioxide andclimate change, scientists are working to itemize all theways carbon moves between the atmosphere and theelements of Earth’s surface, including life, water andsoil. Forests are of particular interest in large partbecause many nations now manage the forests withintheir borders, deciding where and when to harvest treesand when to leave the forest alone. Now those decisionsare influenced by the role forests play in the globalcarbon cycle. Forests’ ability to take in and sequestercarbon during photosynthesis has ceased to besomething we accept without thought; the biologicalservices they provide have instead become a productwith a market value to be traded between nations likeradio parts or soybeans. Just as humans have turned toforests for fuel, food, and shelter for hundreds ofthousands of years, we now look to them to help uscompensate for the atmospheric excesses of ourcombustion-engine civilization.

Whether or not forests will respond as we hope isunclear. Factors other than carbon dioxide availabilityinfluence rates of photosynthesis—factors such as wateravailability and heat stress. In addition, the carbon cycleof a forest involves more than just carbon dioxideuptake because forests burn, decompose, and respire,re-releasing some of their carbon stash back into theatmosphere. We must consider the contribution of manyprocesses to the overall cycle before we can say whatfuture role forests will play in the global carbon cycle orhow much we can rely on them to absorb steadilyincreasing atmospheric carbon dioxide.

Perhaps nowhere on Earth do questions about the roleof forests in the carbon cycle need answers more than inthe Amazon Rainforest. The largest expanse of tropicalforest on Earth, the Amazon covers just 5% of theEarth’s land surface (neglecting Antarctica), and yet is

The Amazon rainforest is avast area of dense jungle.The vegetation “breathes”carbon dioxide andconverts the carbon intobiomass—tree-trunks,branches, and leaves.Therefore, the forest actsas a huge reservoir ofstored carbon. But otherprocesses, such asrespiration and the decayof dead plants, releasecarbon back into theatmosphere. Scientists arecurrently studying theforest to learn whether, inthe end, the Amazon is asource or a sink of carbondioxide. (Photographcopyright Digital Vision)

Escaping carbon

At first glance, the simplest explanation might appear tobe deforestation. When forests are cut down or burned,the carbon stored in the forest biomass is released intothe atmosphere. Combined with the other processes thatcarry carbon out of the rainforest ecosystem—decomposition, respiration, soil and sediment run offinto the Atlantic—deforestation might be the big sourceof carbon scientists are seeking. But calculations suggestotherwise. In Brazil alone, deforestation is proceeding ata rate of about 20,000 square kilometers per year as theAmazon is cleared for farming and ranching (Houghton,et al., 2002), but these losses still do not appear to belarge enough to offset the large carbon intake measuredby the flux towers.

If deforestation wasn’t the culprit, then how couldscientists account for the apparent discrepancy betweenhow much carbon the flux towers indicated was cominginto the forest and the lesser amount of carbon actuallycontained in the biological material? Researchers had nolack of alternative explanations. Maybe the globalmodels were wrong. Maybe estimates of the rates ofdeforestation were too low. Maybe there was somethingwrong with how scientists were collecting the flux towerdata. A few scientists, though, did not discount thepossibility that the Amazon could be hiding a large,yet-to-be-discovered source of carbon emissions. Richeythought he knew where.

“We had been working in the Amazon for almost 20years, collecting all kinds of river samples, includingmeasurements of the carbon dioxide dissolved in thewater. So as far back as 20 years ago, we werepublishing papers saying that the amount of carbon inthe waters of the Amazon was greater than that in theair. For years I had been listening to the carbonmodelers complaining about the discrepancies in thetropics, and I said to myself, ‘I know that carbon dioxideis moving out of the water into the atmosphere.’ But atthat time the scientists doing the carbon modeling didn’ttalk to the people doing the flux tower measurements,and they didn’t talk to those of us who were down on thewater.”

The state of Acre in Brazil,shown in the satelliteimage above, isundergoing rapiddeforestation. Move yourmouse cursor over theimage to compare thelandscape in 2000 withhow it appeared in 2002.The tan areas are recentlycleared patches. The 2002image is also shrouded inhaze from nearby fires setto clear land. (Images byClare Averill, NASA JPLMISR Science Team)

LBA brings the right scientists togetherBut then in 1998, the Brazilian science community,joined by an international team of scientists, launchedthe Large-scale Biosphere-Atmosphere Experiment inAmazonia (LBA). Their aim was to study how Amazoniafunctions as a regional entity within the larger Earthsystem and how changes in land use and climate willaffect the biological, physical, and chemical functioningof the region’s ecosystem. With the Amazon as theirlaboratory, scientists have been studying climate,atmospheric chemistry, the carbon cycle, nutrientcycling, land surface hydrology and water chemistry,land use and land cover, and the interaction of humanswith the landscape.

Richey credits the LBA project for bringing a diversegroup of scientists together and encouraging them tospeak a common language. It was on a return flight froman LBA conference that Richey began a dialogue with acarbon cycle modeler. He says,“On the plane we startedcomparing notes. I realized that we had always talked interms of pressures of carbon dioxide, and they spoke interms of mass, so many tons of carbon in and out of theecosystem each year. I realized we would need to put ourresults into that common language.”

Although directmeasurements of the air inthe Amazon [from fluxtowers (left)] showed theforest removing largeamounts of carbon dioxidefrom the atmosphere,mathematical models ofthe global atmosphereshowed the Amazon as asource of CO2. JeffreyRichey and a team ofscientists had beenstudying the region’s riversand streams for 20 years,and knew that highconcentrations of CO2were dissolved in the water(right). Perhaps the excessCO2 was coming from theAmazon River and itstributaries. [Photographscourtesy Michael Keller,USDA Forest ServiceInstitute of Tropical Forestry(right), and Jeffrey Richey,University of Washington(left).]

Richey knew that what they needed was a grand total:how much total carbon was emitted from water surfaces(a process called evasion) across the Amazon every year.To get a grand total, they required two pieces ofinformation; as many measurements as they could get ofthe amount of carbon dioxide released by numerousareas within the basin and an estimate of the totalsurface area covered by water in the Amazon. To comeup with these numbers, Richey and his colleagues madeuse of data sources that ranged from low tech— morethan a decade’s worth of air and water samples collectedfrom the bows of small fishing boats—to a sophisticated,satellite-based radar.

Small teams of scientistsworking in remote reachesof the Amazon discoveredthat large amounts ofcarbon dioxide wereescaping from the surfaceof Amazon flood waters.Carbon dioxide wastrapped by bowls turnedupside down over thewater. (Photographcourtesy Jeffrey Richey. )

Using the Right Tools

Richey already had a lot of the river water samples heneeded. Between 1982 and 1992, he and his colleagueshad periodically gone out on six-week river cruises on a60-foot, double-decker research boat. In describingthose thousand-kilometer expeditions, Richey says, “TheAmazon is almost beyond anything you can imagine.There’s this vast life and energy surrounding you. Thesky is moving. The river is swirling and churning. There

are birds everywhere. Then you get off the big boat andinto outboards to go into the narrower floodplains, andyou are overwhelmed by the smell of all the vegetation.And all day, there’s the pressure of the sun.”

In addition to the standard, canned, camp fare you’dexpect on a month-long research venture into the depthsof the Amazon, Richey says the crew ate delicious localfood, especially the fish they bought from localfisherman. The trips were not always idyllic, however.The researchers had one of their scarier moments afterbeing confronted by a local tribe who mistakenlythought the researchers had arrived to take them awayand claim a bounty on the tribe offered by drugtraffickers. Richey and his colleagues beat a hastyretreat, more than willing to sacrifice a few data pointsto preserve the peace.

Richey and his colleagues collected more than 1800river water and air samples within the central AmazonRiver Basin. In some cases, they used huge winches tohaul up samples from deep in the river. In other cases,they captured gas emissions from the water surfaceusing what Richey called “floating dishpans,” anddescribed as inverted bowls placed over the water.

The second piece of information Richey needed was agood estimate of just how big an area was covered bywater during the year. The Amazon may be perpetuallywet, but it is wetter at some times than others. FromDecember to May each year, torrential rains and snowmelt from the Andes increase the main river channel’sdepth 30 to 45 feet, and water backs up in tributariesand inundates forest miles from the main channel. Theriver and the flooded forests, called várzea inPortuguese, become a giant, slow-moving swamp.Richey needed to know how big.

Richey’s team measuredthe carbon dioxidedissolved in the AmazonBasin’s rivers by samplingthe water directly. Adecade of river cruisesgave the researchersextensive knowledge ofthe region. (Photographcourtesy Jefferey Richey,University of Washington)

Given the immense area under study, an afternoon trekthrough the jungle with a camera in hand was out of thequestion. Satellite mapping was the only real possibility;satellites such as NASA’s Landsat series had beenmapping the Amazon basin for years in true- andfalse-color imagery. Optical sensors like those onLandsat, which work like digital cameras, have a seriouslimitation, however. If there is one thing that you cancount on in the Amazon during the wet season, it’s rain.At precisely the time of year when Richey neededimagery to reveal the extent of the flooding, the rainclouds hid the forests from a satellite’s view. To map theflooded Amazon forests, Richey needed a remote-sensing device that could see through clouds. He turnedto radar.

The area covered bywater in the Amazon isn’tconstant; it varies wildlywith the change ofseasons. The várzea—flooded forest—isinundated every May byrain and Andeansnowmelt. (Photographscourtesy Max-Planck-Institute for Limnology)

Seeing Through Clouds

Unlike traditional optical sensors, radar is consideredactive as opposed to passive remote sensing. Instead ofpassively recording how much energy is being reflectedby or emitted from the Earth as the spacecraft travelsoverhead, radar works by sending out a pulse of radiowaves toward a target and then recording the strengthand return time of the signal as it bounces back. Thatinformation tells the scientists both how far away thetarget is and what the surface looks like, since differentsurfaces will absorb and reflect the pulse in differentways.

Although LBA is a Brazil-led study, it is an internationalaffair. The National Space Development Agency(NASDA) of Japan mapped the Amazon floodplain aspart of their Global Rainforest Mapping Project, usingradar data collected by the Japanese Earth ResourcesSatellite (JERS-1). As the satellite mapped tropicalrainforests around the globe, different groups aroundthe world became responsible for processing the dataand making them available to the scientific communityin an easy-to-use format.

Bruce Chapman is a senior engineer at NASA’s JetPropulsion Laboratory (JPL) in California, which is theorganization selected by NASDA to handle the datacoming in from South America. Chapman was aprincipal investigator on the project. “With an opticalsensor,” he says, “it can take years to create a cloud-freeimage of the Amazon. Even the supposedly ‘cloud-free’image still has some clouds because there are places inthe Amazon where the clouds just never go away. Radarwavelengths penetrate the clouds and provide a detailedimage of the forests below. The radio waves can evenpenetrate the forest canopy and reveal the layers ofstructure within the forest right down to the ground.”

Radar maps of theAmazon Basin reveal theseasonally flooded forest.In the pair of imagesabove, black representspermanent waterways,dark grey represents forest,and light grey representsflooded areas. (Imagesbased on data providedby the Global RainforestMapping Project)

It’s this ability to see the underlying structure thatenabled them to map the extent of the flooding. Thewater underlying the forest canopy provides a kind ofamplification of the returned radar signal. ExplainsChapman, “The water underneath the canopy providessomething we call a ‘double bounce reflection.’ Thisdouble bounce occurs when the radar waves bounce offtwo perpendicular structures: the very reflective surfaceof the water and the tree trunks. This double bouncemakes the return signal very bright. When we see thatreally bright signal in the Amazon, there is a goodchance there are partially submerged trees.”

Making the maps

Flooded areas appearbright to radar becausethe radar waves arereflected directly back atthe sensor. The firstbounce, off the watersurface, is away from the

The mapping of the Amazon took place in two phases:one data collection for the dry season and a second onefor the wet. The first strip of radar data was obtained onSeptember 27, 1995, over the east coast of SouthAmerica. The satellite mapping progressed about 75kilometers westward each day for the next 62 days, withthe last strip collected over the west coast inmid-November. Beginning May 4, 1996, the satellitemapped the Amazon in flood. The picture was completeby July 3. Chapman and his team at JPL made the finalmaps available to the scientific community in March2001.

Even with the radar data, though, there werelimitations. The radar could only see rivers and streamsat least 100 meters wide, but hundreds, possiblythousands of small streams branch across the Amazon.“To get those streams,” explains Richey, “we had to drilldown even further, using Geographic InformationSystems (GIS) data sets that had been collected over theyears.” For the smallest streams they had computermodels predict the volume and area based ontopographic and geologic features.

sensor, but the secondbounce, off the tree trunksand canopy, redirects thebeam back towards thesource. (Photographcourtesy Jeffrey Richey)

Putting Together Maps and Measurements

Despite the aid of satellite data and years ofobservations, Richey and his colleagues couldn’t hope tostudy the whole Amazon. Instead, they focused theirefforts on a large area in the central Amazon basin. Theycategorized the waters of the 1.77-million-square-kilometer study area into four geographic regions basedon the hydrological characteristics: the main Amazonchannel, the main channel floodplain, tributaries greaterthan 100 meters wide, and tributaries less than 100meters wide. The region was further subdivided intoup-, mid- and downriver regions. Based on the carbondioxide detected in the river samples from each of thesecategories, they came up with an estimate for the entirestudy area.

Richey said they had suspected for years that theamount of carbon dioxide evasion could be large, butuntil they could combine their ground-basedmeasurements with the satellite maps of the totalflooded area, they had no hard evidence, no “smokinggun.” When the amount of carbon dioxide emitted fromthe sampled water surfaces was extrapolated to theentire flooded area within the study site, it totaled 120million grams (264,550 pounds) of carbon per squarekilometer per year. A rough estimate for the amount ofcarbon given off by the entire Amazon River basin washalf a gigatonnne of carbon every year—a mass ofcarbon equivalent to more than 90 million adultelephants!

The area covered by theAmazon River and itstributaries more than triplesover the course of a year.In an average dry season110,000 square km of landare water-covered, whilein the wet season, theflooded area of theAmazon Basin rises to350,000 square km. (Graphadapted from Richey,Melack, Aufdenkampe,Ballester, and Hess)

Says Richey, “When we put our measurements togetherwith the satellite-based flood maps, we got an estimateof carbon dioxide emissions that was greater than 10times the amount of carbon that washes out to sea in theriver outflow. Hydrologists had long thought that themost important role of river systems in the globalcarbon cycle was in the carbon that flowed out to sea asdissolved organic and inorganic compounds. And nowwe had an estimate that the carbon dioxide flowing intothe atmosphere directly from the river surface wasalmost 13 times larger than that amount.” For the firsttime, there was solid evidence of a large carbon sourcewithin the forest sink.

Identifying the Source of the SourceThe carbon in the rivers comes from a number of places.Richey and his colleagues’ believe that most of thecarbon originates in the non-flooded, upland forests.Accounting for 35 percent of the total, they believe, isforest litter that washes down from highland forests.The litter decomposes, giving off carbon dioxide.Another 25 percent of the carbon comes into the systemdirectly as carbon dioxide when plant and tree roots giveoff carbon dioxide during respiration. The carbondioxide becomes dissolved in groundwater that flowsinto streams and rivers. Another 15 percent comes fromcarbon-containing compounds that leach out of soil, leaflitter, and other biological matter. Those dissolvedorganic carbon compounds get metabolized by river life,ultimately returning to the atmosphere as carbondioxide.

The rivers in the AmazonBasin carry a large amountof dissolved carbondioxide gas, created byrotting leaves and othersources in the forestupstream. As the riversystem floods each year,some of this carbondioxide is released into theatmosphere, peaking at alevel of about 35teragrams of carbon permonth. (Graph adaptedfrom Richey, Melack,Aufdenkampe, Ballester,and Hess)

Richey estimates that only about 25 percent of thecarbon given off by the Amazon River and its tributariesactually originates within the river itself, mostly in theform of aquatic vegetation that first fixes carbon dioxideduring photosynthesis and then respires some of it backinto the water. He admits those numbers are onlyestimates at this time. Despite the surprising discoveryof this large source of carbon emissions, he says, so farthe scientific community doesn’t seem bothered by themagnitude of his estimate. “There is definitely a sense of‘here is a missing piece’ of the tropical carbon budgetpuzzle.”

Answers Produce More Questions

By identifying the carbondioxide being transferredfrom the rivers of theAmazon Basin to theatmosphere, scientists areenhancing theirunderstanding of the rolethe Amazon plays in theglobal carbon cycle. Thisunderstanding will helpclarify how natural andhuman-caused changes inthe Amazon could changethe world. (Photographcourtesy Jeffrey Richey)

Where that carbon is coming from is more hotlydebated. If most of the carbon dioxide released from theAmazon waters comes from carbon originally absorbedby the upland forests and washed down into rivers andstreams, as Richey believes, then it would represent areal carbon loss from the ecosystem. But if it turns outthe carbon dioxide is produced by vegetation in the riverand in the adjacent flooded forests and lakes, ratherthan the upland forests, then the large emissions onlycounterbalance a large carbon intake by the aquaticvegetation. The source of the carbon dioxide seeping outof the Amazon waters is the subject of several ongoingstudies.

Richey’s enthusiasm for the project and his excitementabout the results don’t seem to have dimmed since thepaper was published in the journal Nature in April2002. “This study was a terrific assemblage of waterchemistry data, GIS, theory, remote sensing, and towerdynamics. That’s why this was so fun—theintegration—all these disciplines coming together towork on a problem.” The implication is that the couplingbetween the land and the atmosphere, and also betweenthe terrestrial Amazon and the aquatic Amazon, istighter than scientists previously thought.

Those who say that for every question science answers,it generates a dozen more can find evidence in Richey’swork. Richey himself is already thinking ahead. Hewonders about the effect on this source of carbon fromglobal warming and land-use change. He’s alsobeginning to think globally, and has also begun planninga similar study of the rivers and rainforests near theMekong River in southeast Asia. And he’s not done withthe Amazon yet either. Says Richey, “Not all the data weused in this study was gathered specifically to answerthis question. Now we have to go back and get better,more detailed measurements, specifically targeted toanswering our questions.”

References:

1. Richey, J. E., Melack, J.M., Aufdenkampe, A.K.,Ballester, V.M., and Hess, L.L. (2002) Outgassing fromAmazonian rivers and wetlands as a large source oftropical CO2, Nature, 416: 617-6202. Houghton, R.A, Skole, D. L., Nobre, C.A., Hackler,J.L., Lawrence, K.T., and Chomentowski, W.H. (2002)Annual fluxes of carbon from deforestation andregrowth in the Brazilian Amazon, Nature, 403:301-304

Resources:

1. Amazon Facts from the Smithsonian National Zoo.2. LBA-ECO Website

Before widespread human settlement began to encroachon the borders of South America’s Amazon forests, therewas no such thing as an Amazon fire season. Now, firemay pose the biggest threat to the survival of theAmazon ecosystem.

Slash-and-burn agriculture converts forest to farm land,but that obvious destruction is only the beginning.Intentional fires get out of control and burn through theunderstory of nearby forests, killing, but not completelyburning small trees, vines and shrubs. The dead anddying trees collapse, spilling firewood and kindling tothe ground and ripping a great tear in the tent of theforest overhead. Logging has a similar effect. Theintense tropical sun, previously deflected by the greencanopy, heats the forest floor, pushing fire danger evenhigher. Smoke hangs over the forest and suppressesrainfall. In this damaged, fragmented landscape, theonset of the natural dry season becomes ominous. TheEl Niño-driven droughts that typically arrive a couple oftimes per decade become devastating.

A fire sweeps through aportion of the AmazonRainforest, reducing thevegetation to ash. Oncerarely touched by fire, theportions of the Amazon nearareas of human developmentare becoming increasinglysusceptible to damaging fires.(Photograph copyrightWoods Hole Research Center)

Ecologist Dan Nepstad of the Woods Hole ResearchCenter is engaged in an activity that might seem crazyfor someone who cares about forests as much as he does.For the past two years, this veteran of tropical forestresearch has been stealing the rain over two and halfacres of forest in the eastern Amazon.

Strangely, no one seems to mind. None of his colleagues,including fellow Amazon researcher and remote-sensingexpert Greg Asner of the Carnegie Institution andStanford University—whose career in tropical forestresearch began with the environmental group TheNature Conservancy—has tried to stop him. In fact, ifyou ask Asner, he’ll say the whole thing is a great idea.

Nepstad’s ‘grand theft water’ isn’t supporting anexclusive tropical resort or even a hydroelectric project.In fact, he has no need for the water at all. He justdoesn’t want the forest to have it. Nepstad and Asnerwant to know how much drought the forest can takebefore it begins to show signs of stress, what those signsare, and whether any of them can be detected fromspace.

Dan Nepstad and hiscolleagues constructed aunique structure, similar to agreenhouse, to deprive asmall plot of the AmazonRainforest of rain. Thisseemingly bizarre activity waspart of a study on the effectsof drought on the forest.[video (4.5 MB WindowsMedia)] (Photographcourtesy Dan Nepstad, videocourtesy Paul Lefebvre)

“We started thinking about simulated droughtexperiments back in 1994, when the Amazon wascoming out of a major drought caused by a severe ElNiño, and the forest almost completely ran out ofwater,” Nepstad says. The fact that the Amazonexperiences drought often comes as a surprise to people.It’s the rainforest, after all; doesn’t it rain all the time?

“That’s one of the most fascinating things about theAmazon,” explains Nepstad. “The east and southeasternparts of the forest actually go months each year withlittle or no rain. The trees survive by tapping soilmoisture as far down as 20 meters.” During strong ElNiño events, wet-season rainfall decreases, making thedry season even drier. Under those conditions, even thedeep-rooted trees begin to suffer.

Nepstad is concerned that longer and more severedroughts hover on the Amazon’s horizon. Somescientists are predicting that El Niño events will becomemore frequent and severe as Earth’s climate warms.Large-scale deforestation and smoke from biomassburning interfere with local cloud formation andrainfall. Identifying the precise signals of a drought-stressed forest would benefit the region’s farmers,timber operators, fire planners, and conservationists.Being able to detect those changes from a satellite wouldbe a huge advantage; the Amazon is enormous and inmany places still remote and difficult or impossible tosurvey on foot.

In the always-green AmazonRainforest, signs of droughtare rarely obvious. To simulatethe severe droughts theAmazon can experienceduring El Niño events,scientists diverted rainfall froman experimental forest in Braziland then observed theimpacts from ground-leveland from space. (Photographcourtesy 3rd LBA ScientificConference)

Nepstad’s and Asner’s shared interests in forestecosystems led both of their careers to Brazil’sLarge-scale Biosphere-Atmosphere Experiment inAmazonia (LBA, for short), the largest cooperativeinternational scientific project ever to study theinteraction between the Amazon Forest and theatmosphere and ultimately, the climate. NASA’sLBA-ECO program is one of numerous participants inthe effort. The scientists funded under the LBA-ECOprogram concentrate on the processes and effects ofland use change, often using NASA satellite data to adda wider view to what they observe on the ground.

“Through LBA,” Asner says, “Dan and I realized we hadcommon research interests. I was working on how to useremote sensing to describe the structure and function offorest canopies, and he was working on the impact ofdrought, fire, and logging. We realized that the droughtexperiment he was planning was the perfect opportunityto try to find some field-based and remote-sensingindicators of drought stress.”

It’s a Jungle out There

Although the AmazonRainforest is evergreen, manyplaces in the east andsoutheast go months eachyear with little or no rain.Vegetation survives thesenatural droughts by tappingmoisture deep in the soil. Thedrought simulation occurredat an experimental site in theTapajós National Forest inBrazil, near where the TapajósRiver joins the Amazon. (Mapby Robert Simmon)

It’s a Jungle out There

This perfect opportunity wasn’t going to just fall out ofthe sky, though. It would require tremendous creativity,ingenuity, and sweat from Nepstad and the localcarpenters, self-educated engineers, and laborers hehired to help figure out how to exclude rainfall from a100-meter-by-100-meter (1 hectare) plot of rainforest.

It might seem easier to just go out and find a locationthat was experiencing a naturally occurring drought andthen compare it to a location that wasn’t. The troublewith that idea is that the plots need to be similar to eacheither other in as many ways as possible—from thenumber and kinds of trees, to topography and altitude,to soil type—so that the scientists could be sure anydifferences between them were due only to droughtstress. In a place as diverse as the Amazon, that kind ofsimilarity is rare. Once you found two such places, youmight have to observe the two for years—perhaps alifetime—in the hope that at some point, one wouldexperience a drought and the other wouldn’t.

“If you look at all the trees in a 1-hectare plot with aminimum diameter of say, 10 centimeters, at least 2/3 ofthose trees will be a single individual of a species. That’show diverse the place is. Just to find two 1-hectare plotsthat had several species in common, we had to survey 22

Hand-crafted woodenwalkways gave scientistsaccess to the upper levels ofthe forest where they couldmeasure the effects ofdrought on the rainforestcanopy. (Photographcourtesy Paul Lefebvre)

hectares,” explains Nepstad.

Then the sites faced another test. “The first big challengeat any potential site,” he continues, “was whether wewould be able to dig a 10-meter pit in the ground so thatwe could measure soil moisture at various depths. I hadto carry along an auger to test-drill at potentiallocations.”

There was no guarantee that the ground would besuitable for digging. Almost the entire Amazon Basinwas once covered with a vast lake, and the region’s soilsare dense clay formed from sediments that settled to thebottom over hundreds of thousands of years. Asnerlaughs as he says he is glad his part of the project didn’tinvolve much digging. “In the wet season, the soil turnsto a thick clinging mud that sticks to shovels and bootsand everything. In the dry season, it turns to brick.”

Asner faced challenges of his own, though. The biggestproblem with remote sensing in the Amazon is theclouds. “Even if imagery is mostly clear,” says Asner, “itseems like there will always be a cloud in the area youwant [to see]. At some locations you might get only oneor two cloud-free observations per year, which isn’tmuch, but it turns out to be sufficient for land usechange and selective logging studies. It works pretty wellfor drought stress, too, because the one to twocloud-free overpasses are in mid to late dry season,which is the most drought-stressed time of the year forthe forests.”

A researcher descends into a10-meter-deep pit to measuresoil moisture. Normally,rainforest trees survive monthsof seasonal drought eachyear by drawing on waterstored deep in the soil. Duringsevere droughts, however, thesoil can become so dry thateven deeply rooted treessuffer. (Photograph courtesyDan Nepstad)

In with the New

The satellite observations for the study came fromNASA’s first satellite-based hyperspectral remotesensor. The difference between a hyperspectral sensorand a multispectral sensor is that a multispectral sensordetects electromagnetic energy in a sampling of broadslices (groups of wavelengths) of the spectrum while ahyperspectral sensor detects hundreds of very narrowslices of the spectrum that are contiguous, meaning thatone slice touches the next, leaving no gaps. CalledHyperion, the hyperspectral sensor is flying on the EarthObserving-1 satellite. Hyperion is one of several newsensors that NASA is testing in an effort to producesmaller, less expensive devices with more capabilitiesthan its current generation of sensors.

The current Landsat satellites, for example, whoseobservations have been the centerpiece ofhigh-resolution land cover and land cover changemapping for years, only collect observations of 7different spectral bands (the broad slices of wavelengthsof electromagnetic energy described above) reflected or

About 50 kilometers south ofwhere the Tapajós River joinsthe Amazon River, the TapajósNational Forest lies betweenthe Tapajós River in the westand the Santarém-CuiabáHighway (BR-163) in the east.Large, rectangular clearingspunctuate the forest beyondthe highway. The TapajósNational Forest is a 6,000-square-kilometer protectedarea where internationalteams of scientists carry outresearch into how theAmazon responds to humanand natural disturbances.(NASA image by RobertSimmon, based on dataLandsat 7 data provided bythe Global Land CoverFacility)

emitted from Earth. Hyperion detects 220. If amultispectral sensor can be compared to you standing ata paint counter in a home improvement store asking forwhite paint, a hyperspectral sensor would be the clerkasking, “Did you want antique white, or colonial white,or off-white, or eggshell white, or ...?”

With all those wavelengths to choose from, Asner had amuch better chance of detecting the changes in pigmentactivity, leaf area, and carbon balance that he suspectedwould change when the forest got stressed by drought.“The problem with remote sensing of vegetationconditions in the Amazon—well, aside from theclouds—is how lush everything is,” says Asner.

Most satellite-based indicators of vegetation describevegetation “greenness,” which is a general characteristicof vegetation that results from leaf area, canopy cover,and architecture. Greenness indicators are based on therelative amounts of visible light and infrared light beingreflected from the forest. Chlorophyll and otherpigments in the plant leaves absorb visible wavelengths(except green), while the chemicals that make up theleaves’ cell structure reflect near-infrared light. Thetrouble is that because it is so green in the tropicalrainforest, the signal can get “saturated,” which meansthat above a certain level of greenness, it all looks thesame to the satellite.

Asner decided to test the traditional vegetationindicators against some that used the new informationprovided by Hyperion to see which ones did the best jobof detecting the changes in vegetation brought about byNepstad’s simulated drought. In addition to twotraditional greenness indicators, he tested three newindicators that could only be made from hyperspectralobservations: one that was sensitive to the amount of achlorophyll-helper pigment called xanthophyll, one thatwas sensitive to a pigment called anthocyanin, and onethat was sensitive to the water content in the leaves ofthe forest canopy.

A hyperspectral sensor likeHyperion observes energy in aseries of adjacent, narrowslices of the electromagneticspectrum. Plotted on thisgraph, Hyperion’sobservations appear as acontinuous black line.Multispectral sensors likeLandsat detect a single,average value (red dot) for ablock of wavelengths (pinkrectangles). For wavelengthsnear 800 nanometers, forexample, Hyperion detectsseveral peaks and valleys inthe energy reflected from asurface, while Landsat recordsa single measurement.(Graph by Robert Simmon,based on data provided bythe EO-1 science team)

Stealing the Rain

While Asner was getting familiar with the new kinds ofdata coming from Hyperion, and designing and refiningcomputer programs to process the observations andcalculate the different vegetation indicators, Nepstadhad located a suitable pair of sites in the TapajósNational Forest, a managed forest south of Santarem,Brazil, where the Tapajós River joins the Amazon. Hewas busy figuring out how to temporarily thwart MotherNature.

“I just love that kind of thing,” Nepstad says of thechallenge of devising a plan to prevent rain fromreaching the rainforest. “Challenging” doesn’t seem likea strong enough word to describe the situation. “Wecame up with panels—roofing plates that you can get atlocal hardware stores—suspended on a woodenstructure [about one and a half to two meters off theground] and tilted to run off into gutters. Each panelwas half a meter by 3 meters, and there were 5,600 ofthem, made out of essentially clear, greenhouse plastic.Then there were 1,700 meters of gutters--a mile ofgutter! All that flows into a trench around plot. Thetrench is lined with plastic, and the water flows off siteinto a gully about 300 meters away.”

Scientists prevented rain fromreaching the ground in onearea (drought plot) andallowed rain to fall normally inanother (control plot).Although the forest looksuniform from above, speciesdiversity is so high thatscientists had to survey 54acres of tropical forest inorder to find two, 2.5-acreplots with enough treespecies in common for agood comparison. (Image byRobert Simmon, based onIKONOS data copyrightSpace Imaging)

The LBA team built a system ofpanels, gutters, and trenchesto steal the rain from theAmazon Rainforest. Cleargreenhouse plastic wrappedaround wooden framesdeflected rain as it fell ...

... into a network of gutters ...

... that drained into a narrowtrench, which carried thewater off the site.(Photographs courtesy DanNepstad)

To minimize damage to the forest during construction,the 1- to 2-meter-deep trenches around the plot weredug by hand—all 1,500 meters of them. “Of course, aswe dug them out, they became great congregating placesfor every kind of snake you can imagine, boaconstrictors, everything,” says Nepstad with a shortlaugh.

Snakes weren’t the only animal visitors. “With thevolume of wood and other materials we were bringing inwith trucks, we quickly created lakes in the road, andpretty soon we started seeing eyes in them.” The eyesbelonged to caiman, a kind of crocodile. “A number oftimes the workers would be hauling wheelbarrows ofdirt from the trenches out to a pile, and find jaguarssitting on the dirt pile, staring at them.”

“In all we probably brought in a volume of wood equal toabout thirty percent of what was there in the forestitself. I would say about 100 tons of wood was broughtin from local, legal,” he stresses, “saw mills. The peoplewho were bringing in all this wood thought we werecrazy.”

Aside from the trenches they dug to divert water fromthe site, they also had to dig out 11-meter pits so thatthey could measure soil moisture at various depthsthroughout the course of the experiment. “It took threepeople six weeks to dig each pit, and there were ten ofthem. At the height of the construction, we had about 45people employed at the site, and it took about a year andhalf to set it all up.”

The team spent about a year establishing a set ofbaseline measurements for 12 characteristics in eachplot: rainfall, soil moisture, pre-dawn and mid-day leafwater potential (indicates water stress), litterfall (leaves,twigs, flowers, etc, dropped from the trees to theground), litter decomposition, leaf area and canopyopenness, photosynthesis, flower and fruit production,stem growth, stem respiration (the release of carbondioxide back into the atmosphere as the tree consumesthe sugars and starches it creates duringphotosynthesis), gases emitted from the soil, andsolution chemistry (the different kinds of chemicals thatleach into rainwater as it drips down through forest andonto the ground).

Prolific native wildlifefrequently crawled in and outof the study area. This caimanmay look threatening, but it’sonly about 30 centimeters (12inches) long. (Photographcourtesy Paul Lefebvre)

Artificial El Niño Gets Underway

By January 2001, the experiment was officiallyunderway. Between January 7 and May 31 about 1,368millimeters of rain fell over the two sites. Over thecourse of the wet season, rainfall averaged 9.5millimeters per day, and the structure Nepstad and hiscrew had constructed diverted about 50 percent of thatfrom the site, bringing the average down to 4.7millimeters of rain a day, simulating El Niño droughtconditions.

After the wet season ended, the panels were removed. Asthe dry season progressed, the effects of the rainfalldiversion became obvious. By the end of the dry seasonin November 2001, the amount of soil water available toplants at depths between 0 and 11 meters in the“drought” plot was several hundred millimeters lessthan the control site, where Nepstad didn’t interferewith the normal rainfall. Leaf area was 17 percent lowerthan it was at the control site, and mid-day leaf waterpotential was 30 percent lower.

Not surprisingly, all this stress affected tree growth. Netprimary production (the total amount of carbon thatwinds up in trees as a result of what they take in duringphotosynthesis minus what they give off duringrespiration) was almost thirty percent less in thedrought plot than in the normal plot; the mass of carbonin the control plot increased by 2.6 megagrams, while inthe drought plot, it increased by only 1.9 megagrams.

At the field experiment site,rainfall (top graph) began inJanuary, peaked in May, anddropped off dramatically inJuly. Plant available water(bottom graph) increasedand remained high at thecontrol site (top line)throughout the rainy season,while at the drought site(bottom line), available waterwas already declining bymid-February, bottoming outat 400 millimeters by the endof the dry season inDecember. Hyperioncollected data over the plotsin July and November (greylines). (Graph courtesy Asneret. al.)

The fact that drought interfered with the tree growthisn’t surprising. What is surprising, says Nepstad, iswhere in the tree this slow-down occurs. “We thoughtthat early drought stress would show up first inleaves—that leaf area would decrease significantly andthat litterfall would increase as leaves died and droppedoff the trees,” said Nepstad. “Instead, we found onlysmall decreases in leaf area, and litterfall actuallydecreased. It turns out that wood production is the mostsensitive to drought stress. Trees just stop growing indiameter, which has important consequences for timberproduction.”

The second of Nepstad’s two big surprises was whichtrees were most likely to die as a result of the droughtstress. It seemed logical that a smaller tree would have aharder time in a drought than a large tree, since thesmaller tree’s root system couldn’t reach as deeply intothe soil for water as a larger tree’s could. Instead, saysNepstad, “the first trees to die are the big ones, probablybecause they are in the sun high in the canopy.”

Scientists were surprised todiscover that their simulateddrought decreased woodproduction more than it didleaf area. Scientists numberedsome individual leaves (topimage) to monitor leaf areaand mortality. Scientists alsomeasured the girth of treetrunks (bottom image) overthe course of the experiment.(Photographs courtesy PauloBrando)

Asner’s analysis of the Hyperion data confirmed that thecommonly used indicators of vegetation greenness andleaf area just weren’t sensitive enough to detect thesmall differences against a background of such lushvegetation. When he calculated net primary production(net carbon intake) based on the traditional greennessobservations collected by Hyperion, the resultssuggested that the carbon content in the drought plotand the control plot were the same.

Before the experiment, thescientists thought that biggertrees, with their larger rootsystems, would survive thedrought better than smallertrees. Instead, theydiscovered that the big treesthat extend above the forestcanopy (pictured at left)were the first to die.(Photograph courtesy 3rd LBAScientific Conference)

NPP Drought:Control RatioJulyNov.2001Greenness 0.990.980.98Greenness and Pigments 1.0 0.830.85Canopy Water 1.0 0.490.69Canopy Water and Pigments1.0 0.420.67Field Measurements — — 0.73

The indicators that made use of the new hyperspectralinformation from Hyperion were much more successfulat detecting the changes in carbon content broughtabout by drought stress. When Asner factored inobservations of xanthophyll pigment activity (increasedxanthophyll activity is a sign that a tree is using lightefficiently), anthocyanin pigment activity (reddishanthocyanin pigments are most visible in newly formedleaves and buds), and canopy water content, thesatellite-based calculations of net carbon intake camevery close to matching the growth that had beenmeasured on the ground.

Based on field measurements,net primary production (NPP)at the drought site was 73percent less than the controlsite in 2001 (bottom row oftable). Rows 2-4 show theratio of NPP at the two sitesestimated from different typesof satellite observations inJuly, November, and for theentire year. Estimates basedon hyperspectral observationsof chlorophyll-helperpigments and canopy watercontent matched theground-based measurementsmore closely than did theestimate based on greennessalone. (Table courtesy Asneret al.)

When Alfredo Huete saw Scott Saleska’s posterpresentation at a meeting of the AmericanGeophysical Union in 2002, he felt like he hadbeen vindicated. Several years before, Huete hadbeen sponsored by NASA to develop techniquesfor mapping global vegetation using data from anew sensor planned for two of the space agency’supcoming Earth-observing satellite missions. Forseveral years after Terra, the first satellite,launched in 1999, the University of Arizonaremote-sensing ecologist had been worrying overthe data processing and mapping technique heand his team had proposed.

For nearly two decades, scientists had beenmapping global vegetation patterns using avegetation scale, or index, based on data from aseries of satellite sensors operated by theNational Oceanic and AtmosphericAdministration. The NASA sensors built on andeven surpassed the capabilities of the previous sensors,but still, Huete had to deal with a new kind of satellitesensor, a new method for producing the vegetationmaps—and the awareness that he was making a productthat would go out into a global research community withNASA’s name on it. “I felt a lot of pressure,” Huete says.

Layer upon layer ofvegetation stretches from theforest floor to the top of theAmazon Rainforest canopy.Surprisingly, parts of this lushforest go several months eachyear with little or no rain. Theforest actually greens up inthe dry season, due to sunnydays and water sucked upfrom deep in the soil by longroots. (Photograph copyrightNaadir Jeewa.)

Before the 2002 conference Huete had spent severalyears repeatedly tinkering with the data and themapping technique. “When you see something you arenot expecting, you have to ask yourself, ‘What are all thepossibilities for a remote-sensing product goingwrong?’” Among the possibilities are things in theatmosphere that keep the satellite from having a clearview of the surface. “We checked for aerosols [particlesin the air, such as smoke from biomass burning] andclouds, which can potentially reduce the vegetationsignal obtained by satellites. Someone suggested thatmaybe there was flooding on the forest floor during thewet season, so we looked at that. We looked how thevegetation maps changed if the light [hitting a particularpatch of vegetation] was direct or diffuse. We just keptre-doing and re-doing the data products,” he says. Eachtime they made a change, they wondered if thedry-season green-up would disappear. But with eachrefinement, it stayed. His confidence grew, but Huetestill wasn’t sure. Was this for real? Or was it just a signhe was still doing something wrong?

Satellite vegetation mapsshould match well-knownseasonal changes inecosystems. Satellitemeasurements collected overground-based research sitesin Massachusetts (top) andBrazil (lower) are shown hereas graphs and as a filmstrip ofpictures. At Harvard Forest,the seasonal (48-day) satelliteobservations matched thescientists’ expectations: thenumbers were highest andthe pictures were greenest insummer, lowest and brownestin winter. But over theAmazon, numbers on thegraph and the greenness inthe pictures went up duringthe dry season—whenscientists expected the forestto be under stress. (Map byRobert Simmon and JesseAllen, based on data from theOak Ridge National LaboraryDAAC.)

When satellite maps ofAmazon vegetation showedunexpected seasonalpatterns, Huete and hiscolleagues began to checkout ways that remote-sensingobservation can go wrong.This NASA satellite imageacquired October 7, 2005,shows thin clouds and smokepartially obscuring the landsurface. Huete’s teamchecked their data for errorscaused by clouds, smoke,

and changes in lighting toensure their measurementswere accurate. (Image byRobert Simmon, based ondata from the ModerateResolution ImagingSpectroradiometer.)

Abandoning His Doubts

In late 2002, Huete got just the sign he needed toput his doubts behind him. At a meeting of theAmerican Geophysical Union, he saw a poster byecologist Scott Saleska, then part of a researchgroup led by Steven Wofsy of Harvard University,showing results of field studies at a location inthe Tapajos National Forest in Brazil. The sitewas home to a research tower holding scientificinstruments that Saleska and his colleagues haddesigned and operated as part of NASA’scontribution to a Brazilian-led internationalresearch project called “LBA,” short for the“Large-Scale Biosphere-Atmosphere Experimentin Amazonia.”

Among the most important measurementscollected on the 60-meter tower was the uptakeand release of carbon dioxide by the forest.Vegetation takes in carbon duringphotosynthesis, but also releases it duringrespiration. Decomposition of dead trees and vegetationalso releases carbon dioxide back into the atmosphere.Over the course of the project, they collectedobservations of the flux of carbon dioxide from theforest and compared those observations with the growthand death of trees surrounding the tower. Theobservations would reveal whether, overall, the area wasa sink or a source of carbon.

After they had about two years of data, Saleska realizedthey were seeing something surprising. Althoughecosystem models suggested that plants should betaking up less carbon dioxide in the dry season, whentrees were expected to be water-stressed, what they sawwas the opposite. Photosynthesis was greater during thedry season than during the wet season. “When I firstplotted the comparison between our data and themodels,” said Saleska, “I thought for a minute that I hadmade a mistake by plotting the model predictions upsidedown.”

At the Tapajós National Forestin Brazil, scientific instrumentson a tall tower measure therate at which the forest andthe atmosphere around thetower exchange gases, suchas carbon dioxide. In theimage ecologist Scott Saleskascales the tower to check theequipment. (Photographcourtesy Alfredo Huete.)

But results from a second flux tower nearby, operated bySaleska’s colleagues, Michael Goulden from theUniversity of California-Irvine, and Humberto da Rocha,from Brazil’s University of Sao Paulo, showed exactlythe same “backwards” pattern, boosting Saleska’sconfidence that the measured pattern at both towers wasright. Saleska and his colleagues combined the resultsfrom the towers and presented them on a poster for themeeting.

“For several years, we had been seeing this dry-seasongreen up [in the satellite data] and wondering whether itwas real or not,” says Huete. “But when I saw Scott’sresults, that the ground data at the tower sites showedthe same thing as the satellite data, it really changed

When Saleska and his teamfirst graphed themeasurements of carbondioxide exchange betweenthe forest and theatmosphere at research sitesin the Amazon, he thought hehad done it backwards.Measurements (green line)were the opposite ofpredictions from ecosystemmodels (yellow). Modelspredicted that the forestwould release carbon to theatmosphere (positiveexchange) in the dry season(July-December) becausewater stress would slowgrowth. Instead, the treesconsumed (negativeexchange) several hundredkilograms of carbon perhectare per month during thedry season. (Graph by RobertSimmon, adapted fromSaleska 2003.)

everything for me. From that moment, I realized wecould stop focusing on ‘what’s wrong’ and instead focuson how to demonstrate that what we were seeing wasreal.”

Getting further encouragement

As he began pulling the satellite data together anddeciding the best way to demonstrate to other scientiststhat the dry-season green-up wasn’t a mistake, Huetegot some further encouragement that he was on theright track. In 2003, ecologist Rama Nemani of NASAAmes Research Center (a long-time colleague ofHuete’s) and other researchers published research thatlinked global changes in vegetation productivity between1982 and 1999 to the three environmental conditionsthat most affect plant growth: precipitation, sunlight,and temperature. Using 20 years of climate datacombined with satellite-based vegetation maps, theteam developed a model that predicted which of thethree factors most influenced the vegetation in differentplaces on Earth.

Among the most interesting findings, says Nemani, wasthat the Amazon had experienced a large increase inproductivity, apparently because of decreased cloudcover and increased sunlight. “Our analysis indicatedthat growth in the Amazon was light-limited, rather thanwater-limited,” explains Nemani. Over the long-term,the Amazon had become more productive in response togreater sunlight. Huete and Nemani thought that thesame phenomenon could be operating on the seasonalscale.

Detecting the Amazon’s Seasonal Signal

Trying to describe the big picture of seasonal dynamicsacross the entire Amazon Basin puts scientists in acatch-22. The forest is so big that satellites are the onlyway to make observations of the entire forest. Butmeasurements collected from hundreds of kilometersabove can sometimes be hard to tie to specific biologicalprocesses on the ground. On the other hand, a handfulof ground stations scattered throughout 7.5 millionsquare kilometers of forest can’t tell the forest’s entirestory, either. To make a convincing case for anAmazon-wide, dry-season green-up, Huete knew that hewould need both perspectives: space-based and ground-based.

For the satellite-view, Huete and his research teamcompiled 5 years of satellite vegetation data from theModerate Resolution Imaging Spectroradiometer sensoron NASA’s Terra satellite. The maps are based on therelative amounts of red and near-infrared light that thesensor detects over a location on Earth. Chlorophyll in

Flux tower sites (red dots)provide on-the-groundevidence of forest processes,but are too widely scatteredto describe seasonal changesacross the whole Amazon.Satellite data cover the wholearea, but can be hard to linkto specific forest biology, likecarbon uptake. Huete’s teamused both types of data todescribe basin-wide seasonalchanges in Amazonvegetation. (Map by RobertSimmon.)

vegetation absorbs red light, while “scaffolding” (like cellwalls) in the plants’ leaves reflect near-infrared. An areathat reflects very little red light but a lot of near-infraredlight back to space is likely covered in vegetation;scientists call this signal “greenness.” Greenness is anoptical (light-based) way to measure forest productivity.

Like most tropical locations, the Amazon is very cloudy,especially in the rainy season. In the dry season, it canbe pretty smoky from slash and burn deforestation andagricultural fires. The high humidity (water vapor) inthe atmosphere can also interfere with measurementsfrom satellites. To make sure their vegetation maps werefree of clouds and other data contamination, Huete andhis colleagues selected only the best-quality data from2000-2005 and averaged them into a single “typical”Amazon year.

Once they had their example year, the team pored overthe maps, looking at seasonal changes in greenness atthree different scales: the whole Amazon, regional slicescalled transects, and small areas surrounding researchtowers like the one where Saleska worked. Theysubtracted wet season and dry season greenness valuesto identify seasonal patterns. At tower locations inundisturbed forests and sites that had been converted topasture, Huete and Saleska made meticulousweek-by-week comparisons of satellite greenness andground-based measurements of carbon dioxide uptake.

Leaves interact with differentwavelengths of light indifferent ways. Chlorophylland other light-harvestingpigments absorb red light verystrongly, while a layer ofspongy “scaffolding” tissue inleaves reflects near-infraredlight. Land covered withvegetation will absorb redlight and reflect near-infraredlight. Remote-sensing scientistscall this light signaturegreenness. (Photographcopyright M. J. Davidson.)

At each scale, they saw the same pattern: undisturbedrainforests became “greener” and increased theirphotosynthesis throughout the dry season. In theregional transects, the scientists discovered that thelonger the area’s dry season was, the greater thegreening effect was. Even though it might seem like

After combining five years ofsatellite vegetation data,Huete and his teamsubtracted wet seasonmeasurements from dryseason measurements. Thismap shows the results:vegetation index values were

months with little or no rain ought to slow down theforests’ ability to photosynthesize, Huete says thereverse appears true. “The dry season, with less cloudsand higher sunlight, is actually the ‘good’ season.”

Forecasting the Future of the Amazon Abandoning His Doubts

higher (greener) during thedry season than the wetseason in places where theAmazon forests wereundisturbed. At pasture sites,such as in the southeasternAmazon, the vegetation wasless productive (browner)during the dry season.Apparently, lack of sunlight inthe wet season limits forestgrowth more than lack ofwater during the dry season.(Map by Robert Simmon,based on data from theUniversity of Arizona TerrestrialBiophysics and RemoteSensing Lab.)

Forecasting the Future of the Amazon

As counterintuitive as a “good” dry season might seem,that response is perfectly in tune with research about thesoil-water-tapping potential of mature rainforest trees.Since the early 1990s, field studies and soil-moisturemodeling research have been accumulating evidencethat in the undisturbed rainforest, roots extend as far as20 meters (more than 60 feet) into the soil, where thewet-season rains are stored. Rather than being a time ofstress, the normal dry season may be the forests’ mostproductive time of year because the rain clouds clear up,and more sunlight reaches the forest.

But the dry-season green-up only happens inundisturbed forests, stresses Huete. At locations wherethe forest has been converted to pasture or farmland,the dry season has the more intuitive effect: thevegetation “browns down” in response to decreased soilmoisture. Once the deep roots of the mature trees arelost, the access to the water stored deep in the soil is lostas well.

Pastures cleared fromrainforests show the oppositeseasonal pattern fromundisturbed rainforest. Thesegraphs show the relationshipbetween vegetationgreenness (top), rainfall(middle), and sunlight(bottom). Greenness in forestareas (green line) increaseswhen rainfall is low andsunlight is high (July-December). Greenness inpastures (gold line) increasesin the rainy season (January-June). (Graphs by RobertSimmon, adapted from Huete2006.)

“Some current ecological models of the Amazon actuallyhave the Amazon getting browner during the dryseason,” explains Huete. If the seasonal cycle ofgreen-up and photosynthesis in a model is wrong, theability to predict uptake and release of carbon dioxide,water availability, and fire risk would probably be off aswell.

Among the key implications, explains Saleska, is the fateof the Amazon itself. In most models that linksimulations of global climate to vegetation dynamics,the Amazon has a big influence on carbon because evensmall changes in the great stores of carbon in the vastAmazon forests can have significant impacts at theglobal scale. “But what we are learning is that Amazon ishighly sensitive to hydrological change,” he says. “Whenthe modelers couple their climate models to ecosystemmodels of the Amazon and run them out over the nextcentury, some models predict that Earth’s warmingclimate will cause the whole Amazon ecosystem tocollapse, to become a savanna.”

Instead of a dense forest richly layered with trees,shrubs, plants, and vines, the whole area could insteadbe covered by grass and scattered trees. In addition tothe radical transformation of the ecosystem and loss ofbiodiversity, “there would be significant global impactsof that collapse,” explains Saleska. After taking intoaccount the current rates of deforestation, scientiststhink the Amazon region may be neutral with respect tocarbon losses and gains. But if savanna replaced therainforest, says Saleska, “the whole area would switchfrom being close to neutral on average to being a bigsource of carbon.”

Because of the Amazon’s sizeand productivity, smallchanges in the carbon storedthere can produce a globaleffect. Many ecologicalmodels are based on theidea that the dry season is atime of stress. Some modelspredict that future warmingand drying of Earth’s climatemay cause the rainforestecosystem to collapse,turning the area from lushforest (top) into tropicalsavanna (below). Such atransformation woulddramatically change theAmazon’s role in the globalcarbon cycle. The discoveryof a dry-season green-upindicates that many questionsremain about how theAmazon will respond toclimate change. (Rainforestphotograph copyright BrantOlson, pasture photographcopyright Lady Drid.)

“But if these models are getting the seasonality wrong,then the impacts [of climate change] may not be whatwe expect,” he continues. Predictions of ecosystemcollapse are based on the idea that the dry season is atime of stress and declining greenness. If that isn’t true,then perhaps the Amazon will be more resilient than themodels predict. On the other hand, a typical dry seasonisn’t the same as a lengthy El Niño-induced drought.Previous studies, including a drought-simulationexperiment conducted during LBA, indicate that themore severe, extended declines in rainfall that canhappen during strong El Niño events do produce stressin the forest, especially fragmented or damaged areas.With forest disturbance on the rise and predictions bysome climate models that El Niño events may increaseas climate warms, the fate of the Amazon is unclear.

In the near future, Saleska and Huete will be workingtogether to do a synthesis of all the available LBA towerobservations with the satellite data to come up with abasin-wide estimate for the flux of carbon in theAmazon. In the meantime, says Saleska, everyone iskeeping their eyes out for the next strong El Ninobecause observations collected during the event couldprovide the next key piece of the puzzle of how theAmazon responds to large-scale climate variation andchange. Knowing the Amazon’s baseline seasonalresponse should help scientists judge when and howfuture climate events may disturb the balance of such animportant and sensitive ecosystem.

References:

Huete, A.R., Didan, K., Shimabukuro, Y.E., Ratana, P., Saleska,S. R., Hutyra, L.R., Yang, W., Nemani, R.R., and Myneni,R.(2006) Amazon rainforests green-up with sunlight in dryseason. Geophysical Research Letters, 33, L06405,doi:10.1029/2005GL025583.

Saleska, S.R., Miller, S.D., Matross, D.M., Goulden, M.L. Wofsy,S.C., da Rocha, H.R., de Camargo, P.B. Crill, P., Daube,B.C., de Freitas, H.C., Hutyra, L., Keller, M., Kirchhoff, V.,Menton, M. Munger, J.W., Hammond Pyle, E. Rice, A.H.,Silva, H. (2003) Carbon in Amazon Forests: UnexpectedSeasonal Fluxes and Disturbance-Induced Losses. Science302, 1554–1557.

Links:

Introduction to the Large-scale Bisophere-AtmosphereExperiment in Amazonia

LBA-ECO Project Website.

Alfredo Huete’s Webpage. Department of Soil, Water, andEnvironmental Science at the University of Arizona.

by Holli Riebeek· design by Robert Simmon· March 13,2006

When Alfredo Huete saw Scott Saleska’s posterpresentation at a meeting of the AmericanGeophysical Union in 2002, he felt like he hadbeen unleashed. Several years before, Huete hadbeen sponsored by NASA to develop techniquesfor mapping global vegetation using data from anew sensor planned for two of the space agency’supcoming Earth-observing satellite missions. Forseveral years after Terra, the first satellite,launched in 1999, the University of Arizonaremote-sensing ecologist had been worrying overthe data processing and mapping technique heand his team had proposed.

For nearly two decades, scientists had beenmapping global vegetation patterns using avegetation scale, or index, based on data from aseries of satellite sensors operated by theNational Oceanic and AtmosphericAdministration. Although the new NASA sensorsdrew on and even surpassed the capabilities of theprevious sensors, they weren’t identical to what hadcome before. Huete had to deal with a new kind ofsatellite sensor, a new method for producing thevegetation maps—and the awareness that he wasmaking a product that would go out into a globalresearch community with NASA’s name on it. “I felt a lotof pressure,” Huete says.

caption (Photographcopyright Naadir Jeewa)

As they were developing and testing their technique,Huete and his team frequently checked that the mapsmatched real-world seasonal changes in vegetation indifferent ecosystems, from African savannas to easternNorth American forests. Although their maps capturedthe expected seasonal changes in most areas, one areabothered Huete: the Amazon. As data from Terra beganaccumulating, he noticed something peculiar: theAmazon rainforest looked greener to the satellite in theforest’s dry season than it did during the rainy season.Huete knew that parts of the forest go several monthswith little or no rain. How could the forest be thrivingduring those times of seasonal drought?

Amazonia covers most oftropical South America eastof the Andes Mountains.Primarily rainforest, the regionincludes some of the densestvegetation and greatestbiodiversity on Earth. This mapshows enhanced vegetationindex, a measure of plantvigor, in South America.Satellite measurements of EVIrevealed that the Amazon isgreener in the dry seasonthan during the rainy season,contradicting conventionalwisdom. (Map by RobertSimmon)

graph of EVI over time

Huete had spent several years repeatedly tinkering withthe data and the mapping technique. “When you seesomething you are not expecting, you have to askyourself, ‘What are all the possibilities for a remote-sensing product going wrong?’” Among the possibilitiesare things in the atmosphere that keep the satellite fromhaving a clear view of the surface. “We checked foraerosols [particles in the air, such as smoke frombiomass burning] and clouds, which can potentiallyreduce the vegetation signal obtained by satellites.Someone suggested that maybe there was flooding onthe forest floor during the wet season, so we looked at

caption (Graph by RobertSimmon and Jesse Allen)

that. We looked how the vegetation maps changed if thelight [hitting a particular patch of vegetation] was director diffuse. We just kept re-doing and re-doing the dataproducts,” he says. Each time they made a change, theywondered if the dry-season green-up would disappear.But with each refinement, it stayed. His confidencegrew, but Huete still wasn’t sure. Was this for real? Orwas it just a sign he was still doing something wrong?