This content has been downloaded from IOPscience. Please scroll down to see the full text.

Download details:

IP Address: 128.83.166.162

This content was downloaded on 03/12/2015 at 16:35

Please note that terms and conditions apply.

Sunlight mediated seasonality in canopy structure and photosynthetic activity of Amazonian

rainforests

View the table of contents for this issue, or go to the journal homepage for more

2015 Environ. Res. Lett. 10 064014

(http://iopscience.iop.org/1748-9326/10/6/064014)

Home Search Collections Journals About Contact us My IOPscience

Environ. Res. Lett. 10 (2015) 064014 doi:10.1088/1748-9326/10/6/064014

LETTER

Sunlight mediated seasonality in canopy structure andphotosynthetic activity of Amazonian rainforests

JianBi1,20, Yuri Knyazikhin1,20, SunghoChoi1, Taejin Park1, JonathanBarichivich2, PhilippeCiais3, Rong Fu4,SangramGanguly5, ForrestHall6, ThomasHilker7, AlfredoHuete8,Matthew Jones9, JohnKimball9,Alexei I Lyapustin10,MattiMõttus11, Ramakrishna RNemani12, Shilong Piao13,14, Benjamin Poulter15,Scott R Saleska16, Sassan S Saatchi17,18, LiangXu17, LimingZhou19 andRangaBMyneni1

1 Department of Earth and Environment, BostonUniversity, Boston,MA02215,USA2 Climatic ResearchUnit, School of Environmental Sciences, University of East Anglia, NorwichNR4 7TJ, UK3 Laboratoire des Sciences duClimat et de l’Environnement, IPSL-LSCE, CEA-CNRS-UVSQ, 91191Gif sur Yvette Cedex, France4 Department ofGeological Sciences, TheUniversity of Texas at Austin, Austin, TX 78712,USA5 BayArea Environmental Research Institute, NASAAmesResearchCenter,Moffett Field, CA 94035,USA6 Biospheric Sciences Laboratory, NASAGoddard Space Flight Center, Greenbelt,MD20771,USA7 College of Forestry, Oregon StateUniversity, Corvallis, OR 97331,USA8 Plant Functional Biology andClimate ChangeCluster, University of Technology Sydney, New SouthWales 2007, Australia9 Numerical Terradynamic SimulationGroup, TheUniversity ofMontana,Missoula,MT59812,USA10 Climate andRadiation Laboratory, NASAGoddard Space Flight Center, Greenbelt,MD20771,USA11 Department ofGeosciences andGeography, University ofHelsinki, FI-00014,Helsinki, Finland12 NASAAdvanced SupercomputingDivision, AmesResearchCenter,Moffett Field, California 94035,USA13 Department of Ecology, PekingUniversity, Beijing 100871, People’s Republic of China14 Institute of Tibetan PlateauResearch, Chinese Academy of Sciences, Beijing 100085, People’s Republic of China15 Department of Ecology,Montana StateUniversity, Bozeman,MT59717,USA16 Department of Ecology and Evolutionary Biology, University of Arizona, Tucson, AZ 85721,USA17 Institute of the Environment and Sustainability, University of California, Los Angeles, CA 90095,USA18 Radar Science and Engineering section, Jet Propulsion Laboratory, California Institute of Technology, Pasadena, CA 91109,USA19 Department ofAtmospheric andEnvironmental Sciences,University atAlbany, StateUniversity ofNewYork,Albany,NewYork12222,USA20 These authors contributed equally to thiswork.

E-mail: jknjazi@bu.edu

Keywords:Amazonian rainforests, seasonality, remote sensing,MISR,MODIS

Supplementarymaterial for this article is available online

AbstractResolving the debate surrounding the nature and controls of seasonal variation in the structure andmetabolism of Amazonian rainforests is critical to understanding their response to climate change.In situ studies have observed higher photosynthetic and evapotranspiration rates, increased litterfalland leafflushing during the Sunlight-rich dry season. Satellite data also indicated higher greennesslevel, a proven surrogate of photosynthetic carbonfixation, and leaf area during the dry season relativeto thewet season. Some recent reports suggest that rainforests display no seasonal variations and theprevious results were satellitemeasurement artefacts. Therefore, here we re-examine several years ofdata from three sensors on two satellites under a range of sun positions and satellitemeasurementgeometries and document robust evidence for a seasonal cycle in structure and greenness of wetequatorial Amazonian rainforests. This seasonal cycle is concordant with independent observations ofsolar radiation.We attribute alternative conclusions to an incomplete study of the seasonal cycle, i.e.the dry season only, and to prognostications based on a biased radiative transfermodel. Consequently,evidence of dry season greening in geometry corrected satellite datawas ignored and the absence ofevidence for seasonal variation in lidar data due to noisy and saturated signals wasmisinterpreted asevidence of the absence of changes during the dry season.Our results, grounded in the physics ofradiative transfer, buttress previous reports of dry season increases in leaf flushing, litterfall,photosynthesis and evapotranspiration inwell-hydrated Amazonian rainforests.

OPEN ACCESS

RECEIVED

26 February 2015

REVISED

28May 2015

ACCEPTED FOR PUBLICATION

1 June 2015

PUBLISHED

16 June 2015

Content from this workmay be used under theterms of theCreativeCommonsAttribution 3.0licence.

Any further distribution ofthis workmustmaintainattribution to theauthor(s) and the title ofthework, journal citationandDOI.

© 2015 IOPPublishing Ltd

1. Introduction

Understanding the seasonal variation in functioningof rainforests and its controls are requisite for under-standing how rainforests will respond to climatechange. In situ studies report counter-intuitive seaso-nal variation in wet equatorial Amazonian rainforests—higher photosynthetic and evapotranspiration ratesand increased litterfall and leaf flushing during theSunlight-rich dry season (Saleska et al 2003, da Rochaet al 2004, Goulden et al 2004, Rice et al 2004, Haslerand Avissar 2007, Hutyra et al 2007, Negrón Juárezet al 2009, Costa et al 2010, Jones et al 2014). Waterlimitation during the dry season is alleviated in theseforests through deep roots and hydraulic redistribu-tion (Nepstad et al 1994, Oliveira et al 2005). Satellitedata, which cover a large area and span a long timeperiod, support findings of in situ studies—higherradiometric greenness level and green leaf area duringthe dry season compared to the wet season (Xiaoet al 2005,Huete et al 2006,Myneni et al 2007, Samantaet al 2012, Hilker et al 2014, Jones et al 2014, Maedaet al 2014). This convergent view of seasonality, parsedfrom several studies, shows how sunlight interactswith adaptive mechanisms to result in higher rates ofleaf flushing, litterfall, photosynthesis and evapotran-spiration in tropical forests if water limitation isabsent (Wright and Van Schaik 1994, Restrepo-Coupeet al 2013, Borchert et al 2015, Guan et al 2015).

This community-consensual view was questionedin recent studies (Galvão et al 2011, Mortonet al 2014). The studies claim that the dry seasongreening inferred from passive remote sensing dataresulted from an artificial increase in forest canopyreflectance at near-infrared (NIR) wavelengths causedby variations in sun-satellite sensor geometry. Theiranalyses of satellite-borne lidar data suggested thatthese forests exhibited no seasonal variations incanopy structure or leaf area. Relying on model simu-lations to guide and imbue a physical meaning to thesatellite data analysis, the studies conclude that Ama-zon rainforests maintain consistent structure andgreenness during the dry season.

These contradictory results justify a re-examina-tion of the same satellite data with the goal of assessingseasonality in wet equatorial Amazonian rainforests.In addition to data fromNASA’s Moderate ResolutionImaging Spectroradiometer (MODIS) on the Terraplatform and the Geoscience Laser Altimeter System(GLAS) instrument onboard the Ice, Cloud and landElevation Satellite (ICESat) used in (Mortonet al 2014), we also include data from the MODISinstrument on Aqua andMultiangle Imaging Spectro-radiometer (MISR) on the Terra satellite. The MISRsensor views the Earth’s surface with nine camerassimultaneously, as opposed to the two MODIS sen-sors, which are capable of only one view each. This fea-ture enables the rigorous use of the theory of radiativetransfer in vegetation canopies—the fundamental

theory that explains from first principles the mechan-isms underlying the signals generated by the canopyand measured by a remote sensor (Knyazikhinet al 2005).

This study is focused on terra firme rainforests incentral Amazonia that are relatively undisturbed byhuman activities (supplementary data and methodssection 1, figure S1). The period June toMay is treatedas one seasonal cycle as per convention (Hueteet al 2006, Morton et al 2014). It consists of a short dryseason, June to October, and a long wet season there-after (supplementary data and methods section 1).The following analysis of satellite borne sensor dataaddresses the question at the center of current debate—did previous studies (Xiao et al 2005, Hueteet al 2006, Myneni et al 2007, Brando et al 2010,Samanta et al 2012) misinterpret changes in near-infrared (NIR) reflectance caused by seasonal changesin sun-satellite sensor geometry (figures S2 and S3) asseasonal variations in rainforest canopy structure andgreenness (Galvão et al 2011,Morton et al 2014)?

2.Data andmethods

A detailed description of methods and data used isgiven in the supplementary information available atstacks.iop.org/ERL/10/064014/mmedia. A brief sum-mary is provided here. The study region and thevarious data analysed in this study are detailed in thesupplementary data and methods section 1−2. Thesun-sensor geometry relevant to the discussion in thisarticle is presented in the supplementary data andmethods section 3. The theory of remote measure-ments and evaluation of NIR reflectance angularsignatures (figure 3) and their interpretation isdescribed in the supplementary data and methodssection 4. A critical look at Morton et al 2014 analysesof MODIS and GLAS data is presented in thesupplementary discussion. Abbreviations and symbolsare listed in supplementary table S5.

3. Results and discussioin

3.1. Leaf area index seasonalityThe seasonal cycle of green leaf area inferred fromsatellite data (figure 1(a)) exhibits rising values duringthe dry season (June to October), high values duringthe early part of the wet season (November toFebruary) and decreasing values thereafter (March toMay). This seasonal variation of about 20% is imposedon a base value of Leaf Area Index (LAI, one-sidedgreen leaf area per unit ground area) of about 5.75, isgreater than the uncertainty of the LAI product (0.66LAI, Yang et al 2006) and is observed in nearly 70% ofthe rainforests in the study domain (figure S4(a)); therest lacked valid data. Is this seasonal variation real or amisinterpretation of changes in satellite-sensor mea-surements caused by seasonal changes in sun position

2

Environ. Res. Lett. 10 (2015) 064014 J Bi et al

in the sky and the manner in which the sensormeasures reflected radiation (‘sun-sensor geometry’)?The answer requires an understanding of how thisgeometry changes during the seasonal cycle, which isdescribed in the supplementary data and methodssection 3.

The seasonal cycle of leaf area in figure 1(a) cannotbe an artefact of seasonal changes in sun-sensor geo-metry because the algorithm with which leaf area isderived explicitly accounts for geometry changes, i.e.the algorithm is capable of differentiating betweenchanges in measurements caused by leaf area changesand those caused by geometry changes (Knyazikhinet al 1999, Knyazikhin et al 1998). This is also evidentfrom the fact that the seasonal cycle of leaf area doesnot track the seasonal course of either the Sun positionin the sky (figure 1(b)) or theMODIS sensor sampling(figures 1(c) and (d)). Instead, it tracks independentlyobtained observations of seasonal variation in sunlight(figure 1(a)). This behavior is consistent with the ideathat sunlight acts as a proximate cue for leaf produc-tion in moist tropical forests if water limitation isabsent (Wright and Van Schaik 1994, Borchertet al 2015, Guan et al 2015). Thus, relatively high sun-light levels from absence of clouds during the dry sea-son cause leaf area to increase, which in turn generateshigher rates of photosynthesis (Saleska et al 2003, DaRocha et al 2004, Restrepo-Coupe et al 2013, Gattiet al 2014). But, photosynthesis becomes decoupledfrom sunlight during the early to middle part of the

wet season. This results in increasing rates of photo-synthesis, which are possibly sustained by still suffi-ciently high levels of light and increasing leafproduction (Restrepo-Coupe et al 2013). All threedecrease rapidly thereafter. A bimodal seasonal cycleof LAI reported in one instance could be site-specific(figure 2 in Doughty and Goulden (2008)) as alternatein situ evidence does not exist (Restrepo-Coupeet al 2013, Xiao et al 2005, Asner et al 2000, Carswellet al 2002, Chave et al 2010, Malhado et al 2009,Negrón Juárez et al 2009).

3.2. Evidence for seasonality after sun-sensorgeometry correctionThe Enhanced Vegetation Index (EVI) is a provenproxy for the potential photosynthetic carbon fixationby vegetation (Xiao et al 2005,Huete et al 2006, Brandoet al 2010). It is calculated from satellite-sensormeasurements of reflected solar radiation at threedifferent wavelength bands. These measurementsdepend on sun-sensor geometry, but this dependencycan be eliminated by expressing themeasurements in afixed geometry (Morton et al 2014, Lyapustinet al 2012). The EVI calculated from MODIS sensormeasurements in a fixed geometry, i.e. nadir viewingdirection and 45° solar zenith angle, shows a distinctwet season decrease (figure 2(a)) and dry seasonincrease (figure 2(b)). These changes are greater than ahighly conservative estimate of the precision in 43% ofthe pixels during the wet season and 31% of the pixels

Figure 1. Seasonal variations in green leaf area of central Amazonian rainforests. (a) Seasonal cycles of TerraMODIS leaf area index(LAI), at-surface photosynthetically active radiation (PAR) fromCERES, andTRMMprecipitation. The PARpolynomial regressioncurve excludes the circled data point. The seasonal profiles represent average values over pixels that exhibited dry season greening in atleast 4 out of 7 seasonal cycles analyzed (63%of all forest pixels). (b)–(d) Seasonal cycle of LAI, as in panel (a), contrasted againstseasonal variations in (b) solar zenith angle, (c) sensor view relative azimuth angle and (d) view zenith angle.

3

Environ. Res. Lett. 10 (2015) 064014 J Bi et al

in the dry season. Here, the precision is estimated asthe spatial standard deviation of the EVI data in thestudy domain. Analogous to EVI, pixel level estimatesof green leaf area show a strong decrease in the wetseason and increase during the dry season. The wetseason decrease (figure 2(a)) suggests net leaf abscis-sion, i.e. more older leaves dropped than those newlyflushed, and the dry season increase indicates net leafflushing (figure 2(b)), resulting in a sunlight mediatedphenological behavior (Myneni et al 2007). The factthat both EVI and LAI show congruent changes duringthe seasonal cycle even though the Sun-sensor geome-try effect is removed from measurements in differentways (Knyazikhin et al 1999, Knyazikhin et al 1998,Lyapustin et al 2012, Hilker et al 2014, Maedaet al 2014) is particularly noteworthy.

3.3. Evidence for seasonality frommultiple sensorsand geometriesNow we turn to satellite-sensor measurements ofreflected solar radiation at the NIR wavelength band,which are at the heart of the controversy. Thesemeasurements are usually expressed as normalizedquantities called reflectances (supplementary data andmethods section 4.1−4.2). The geometric structureand radiation scattering properties of the rainforestcanopy determine the magnitude and angular distri-bution of reflected radiation. The angular signatures ofreflectance are therefore unique and rich sources ofdiagnostic information about rainforest canopies(Diner et al 1999). We first examine NIR angularsignatures from the late dry season (October 15 to 30)and the middle part of the wet season (March 5 to 20).The Solar Zenith Angle (SZA) at the time when Terra(10:30 am) and Aqua (1:30 pm) satellites view thecentral Amazonian forests in March and October isbetween 20° and 30°. This variationminimally impactsthe shape of angular signatures (supplementary data

and methods section 4.4). MODIS and MISR sensorssample the rainforests very differently (figures S2(c)–(f); also see figure S1(c)). However, all the sensorsrecord a distinct decrease in reflected NIR radiation inall view directions between October and March withno change in the overall shape of the angular signatures(figures 3(a) and (b)). Such a simple change inmagnitude can only result from a change in canopyproperties—this conclusion is based on the physics ofhow solar radiation interacts with foliage in vegetationcanopies (supplementary data and methods section4.3, figures S5(a) and (b)). The EVI, although evalu-ated from reflectances at NIR, red and bluewavelengthbands, is tightly linked to NIR reflectance (Samantaet al 2012). Thus, the decrease in sun-sensor geometrycorrected EVI (figure 2(a)) is in agreement withdirectly observed decreases in NIR angular signaturesfromOctober toMarch (figures 3(a) and (b)).

The wet season reduction in greenness is incon-sistent with the hypothesis of invariant dry seasongreenness. Indeed the net loss of leaf area, without acorresponding net gain elsewhere during the seasonalcycle, will result in rainforests without leaves in a fewyears. If wet Amazonian forests somehow maintainconsistent canopy structure and greenness during thedry season, then they must be either aseasonal or theentire seasonal cycle must be confined to the wet sea-son, but this argument lacks empirical support. Thequestion then arises whether variations in angular sig-natures of forest reflectance during the dry season sup-port this inference?

Therefore, let us now consider NIR reflectancesfrom early (25 June to 10 July) and the late dry season(15 October to 30 October) when both sun position inthe sky and sensor sampling vary significantly (figuresS2(a)−(d); also see figure S1(c)). MODIS and MISRmeasurements are made at significantly higher SZA inJune (∼35°–40°) compared to October (∼20°–30°).

Figure 2.Wet and dry season changes in sun-sensor geometry corrected estimates of leaf area and greenness. Per-pixel changes inMODIS leaf area index (LAI) andMODISMAIAC enhanced vegetation index (EVI) from (a)October toMarch and (b) June toOctober. LAI values are normalized by 10. The changes are calculated as the difference between the values inMarch andOctober, andOctober and June, respectively.

4

Environ. Res. Lett. 10 (2015) 064014 J Bi et al

The magnitude and shape of angular signatures areimpacted when both canopy properties and SZA vary.However, a higher or equal reflectance at lower SZArelative to reflectance at higher SZA always indicatesan increase in leaf area and foliage scattering proper-ties according to the physics of radiation interaction invegetation (supplementary data and methods section4.4−4.5, figures S5(c)−(f)). This is observed clearly inMISR data (figure 3(d)) because this sensor views theEarth’s surface with nine cameras simultaneously, asopposed to the two MODIS sensors (figure 3(c)),which are capable of only one view each (figure S3).Further, the juxtaposition of the two angular sig-natures in figure 3(d) is significantly different thanthat predicted by theory for the case of identical cano-pies (supplementary data and methods section 4.6).Thus, the NIR angular signatures in figure 3(d) indi-cate a change in vegetation structure (LAI) and green-ness (EVI) during the dry season.

4. Conclusions

Satellite data indicate a distinct sunlight-mediatedseasonality in leaf area and photosynthetic carbonfixation over unstressed rainforests in central Amazo-nia. This seasonal cycle is not an artefact of seasonalchanges in sun position in the sky or how the satellite-sensor measures the reflected radiation field. Thespatially expansive remote sensing data agree withavailable in situ data. A better understanding of how

the rainforests will respond to climate change dependson future ground campaigns as satellite data cancomplement, but not substitute,field data.

Acknowledgments

This study was supported by NASA Earth ScienceDivision.

Author contributionsJ B, Y K, S C and T P performed the analyses. RBM

wrote the initial draft. All authors contributed withideas to the analyses and with writing the article. Theauthors declare no conflict of interest.

References

AsnerGP, TownsendAR andBraswell BH2000 Satelliteobservation of ElNino effects onAmazon forest phenologyand productivityGeophys. Res. Lett. 27 981–4

Brando PM,Goetz S J, Baccini A,NepstadDC, Beck P S andChristmanMC2010 Seasonal and interannual variability ofclimate and vegetation indices across the Amazon Proc. NatlAcad. Sci. USA 107 14685–90

Borchert R, Calle Z, Strahler AH, Baertschi A,Magill R E,Broadhead J S, Kamau J, Njoroge J andMuthuri C 2015Insolation and photoperiodic control of tree developmentnear the equatorNewPhytol. 205 7–13

Carswell F et al 2002 Seasonality inCO2 andH2O flux at an easternAmazonian rain forest J. Geophys. Res. Atmos. 107 LBA43–1

Chave J et al 2010Regional and seasonal patterns of litterfall intropical SouthAmericaBiogeosciences 7 43–55

CostaMH et al 2010Atmospheric versus vegetation controls ofAmazonian tropical rain forest evapotranspiration: are the

Figure 3. Seasonal changes in angular signatures of near-infrared (NIR) reflectance from three satellite borne sensors. Angularsignatures of NIR reflectance inMarch (5Mar to 20Mar), June (25 Jun to 10 Jul) andOctober (15Oct to 30Oct). TheAquaMODISsignature forOctober is shown in panel (c) for clarity. The phase angle is the angle between the directions to the Sun and sensor (figureS1(c)).

5

Environ. Res. Lett. 10 (2015) 064014 J Bi et al

wet and seasonally dry rain forests any different? J. Geophys.Res. Biogeosci. 115G04021

DaRochaHR,GouldenML,Miller SD,MentonMC, Pinto LD,De FreitasHC and Silva Figueira AME2004 Seasonality ofwater and heatfluxes over a tropical forest in easternAmazoniaEcol. Appl. 14 S22–32

DinerD J, AsnerGP,Davies R, Knyazikhin Y,Muller J P,Nolin AW, Pinty B, Schaaf CB and Stroeve J 1999Newdirections in earth observing: scientific applications ofmultiangle remote sensingBull. Am.Meteorol. Soc. 802209–28

DoughtyCE andGouldenML 2008 Seasonal patterns of tropicalforest leaf area index andCO2 exchange J. Geophys. Res.Biogeosci. 113G00B06

Galvão L S,Dos Santos J R, Roberts DA, Breunig FM,ToomeyMandDeMoura YM2011On intra-annual EVIvariability in the dry season of tropical forest: a case studywithMODIS and hyperspectral dataRemote Sens. Environ.115 2350–9

Gatti L et al 2014Drought sensitivity of Amazonian carbon balancerevealed by atmosphericmeasurementsNature 506 76–80

GouldenML,Miller SD,DaRochaHR,MentonMC,De FreitasHC, Silva Figueira AME andDe SousaCD2004Diel and seasonal patterns of tropical forest CO2 exchangeEcol. Appl. 14 S42–54

GuanK et al 2015 Photosynthetic seasonality of global tropicalforests constrained by hydroclimateNatureGeosci. 8 284–9

HaslerN andAvissar R 2007What controls evapotranspiration intheAmazon basin? J. Hydrometeorol. 8 380–95

Hilker T et al 2014Vegetation dynamics and rainfall sensitivity ofthe Amazon Proc. Natl Acad. Sci. USA 111 16041–6

Huete AR et al 2006Amazon rainforests green‐upwith sunlight indry seasonGeophys. Res. Lett. 33 L06405

Hutyra L R et al 2007 Seasonal controls on the exchange of carbonandwater in anAmazonian rain forest J. Geophys. Res.Biogeosci. 112G03008

JonesMO,Kimball J S andNemani RR 2014AsynchronousAmazon forest canopy phenology indicates adaptation tobothwater and light availability Environ. Res. Lett. 9 124021

Knyazikhin Y et al 1999MODIS leaf area index (LAI) and fraction ofphotosynthetically active radiation absorbed by vegetation(FPAR) product (MOD15)AlgorithmTheoretical BasisDocument (https://lpdaac.usgs.gov/products/modis_products_table/mcd15a2)

Knyazikhin Y,Martonchik J,Myneni R,DinerD andRunning S1998 Synergistic algorithm for estimating vegetation canopyleaf area index and fraction of absorbed photosyntheticallyactive radiation fromMODIS andMISR data J. Geophys. Res.Atmos. 103 32257–75

Knyazikhin Y,MarshakA andMyneni RB 2005Three-dimensionalradiative transfer in vegetation canopies and cloud-vegetationinteractionThreeDimensional Radiative Transfer in theCloudyAtmosphere edAMarshak andABDavis (Berlin,Heidelberg: Springer) pp 617–51

Lyapustin A I et al 2012Multi-angle implementation of atmosphericcorrection forMODIS (MAIAC): 3. Atmospheric correctionRemote Sens. Environ. 127 385–93

Maeda E E,Heiskanen J, Aragão L EOC andRinne J 2014CanMODIS EVImonitor ecosystem productivity in theAmazonrainforest?Geophys. Res. Lett. 41 7176–83

MalhadoA,CostaMH, de Lima F Z, PortilhoKCandFigueiredoDN2009 Seasonal leaf dynamics in anAmazonian tropical forest For. Ecol.Manage. 258 1161–5

MortonDC et al 2014Amazon forestsmaintain consistent canopystructure and greenness during the dry seasonNature 506221–4

Myneni RB et al 2007 Large seasonal swings in leaf area of Amazonrainforests Proc. Natl Acad. Sci. USA 104 4820–3

Negrón Juárez R I, da RochaHR,GouldenML andMiller SD 2009An improved estimate of leaf area index based on thehistogram analysis of hemispherical photographsAgric. For.Meteorol. 149 920–8

NepstadDC et al 1994The role of deep roots in the hydrological andcarbon cycles of Amazonian forests and pasturesNature 372666–9

Oliveira R S,DawsonTE, Burgess S S andNepstadDC2005Hydraulic redistribution in three Amazonian treesOecologia145 354–63

Restrepo-CoupeN et al 2013What drives the seasonality ofphotosynthesis across the Amazon basin? A cross-site analysisof eddy flux towermeasurements from the BrasilfluxnetworkAgric. For.Meteorol. 182 128–44

Rice AH et al 2004Carbon balance and vegetation dynamics in anold-growthAmazonian forest Ecol. Appl. 14 55–71

Saleska SR et al 2003Carbon inAmazon forests: unexpectedseasonal fluxes and disturbance-induced losses Science 3021554–7

Samanta A et al 2012 Seasonal changes in leaf area of Amazon forestsfrom leafflushing and abscission J. Geophys. Res. Biogeosci.117G01015

Wright S J andVan SchaikCP 1994 Light and the phenology oftropical treesAm.Nat. 143 192–9

XiaoX et al 2005 Satellite-basedmodeling of gross primaryproduction in a seasonallymoist tropical evergreen forestRemote Sens. Environ. 94 105–22

YangW et al 2006MODIS leaf area index products: from validationto algorithm improvement IEEETrans. Geosci. Remote Sens.44 1885–98

6

Environ. Res. Lett. 10 (2015) 064014 J Bi et al