Leaf area index of a tropical semi-deciduous forest of the southern ...

Int J Biometeorol (2011) 55:109–118DOI 10.1007/s00484-010-0317-1ORIGINAL PAPERLeaf area index of a tropical semi-deciduous forestof the southern Amazon BasinOsvaldo Borges Pinto-Júnior & Luciana Sanches &Francisco de Almeida Lobo &Adilson Amorim Brandão & José de Souza NogueiraReceived: 25 July 2009 /Revised: 7 April 2010 /Accepted: 7 April 2010 /Published online: 9 June 2010# ISB 2010Abstract Leaf area index (LAI) is an important ecophysiologicalvariable because leaves are the organs responsiblefor gas exchange between plants and the atmosphere. Thisvariable can be calculated from primary values of leaf areaassessed by destructive or non-destructive methods, whichis relatively easy when crop species are investigated, but isnot the case when the focus is on natural wood plantscommunities. In this paper, we analyze the seasonality ofLAI estimated by three different methods in the Amazoniasavannahtransitional forest, located 50 km north-east ofSinop city, Mato Grosso, Brazil. In the first method, wecombine Monsi and Saekis' original method [Monsi M,Saeki T (1953) Jpn J Bot 14:22–52], which measures LAIusing the Beer-Lambert extinction law, and the propositionO. B. Pinto-Júnior (*) : F. de Almeida Lobo : A. A. BrandãoDepartment of Soil and Rural Engineering,Federal University of Mato Grosso,Av. Fernando Correa da Costa s/n,Cuiabá, MT 78.060-900, Brazile-mail: osvaldo.borges@gmail.comF. de Almeida Loboe-mail: f_a_lobo@ufmt.brA. A. Brandãoe-mail: aabrandaojr@gmail.comL. SanchesDepartment of Sanitary and Environmental Engineering,Federal University of Mato Grosso,Av. Fernando Correa da Costa s/n,Cuiabá, MT 78.060-900, Brazile-mail: luciana.sanches@pq.cnpq.brJ. de Souza NogueiraDepartment of Physics, Federal University of Mato Grosso,Av. Fernando Correa da Costa s/n,Cuiabá, MT 78.060-900, Brazile-mail: nogueira@ufmt.brof Goudriaan [Goudriaan J (1988) Agric For Meteorol43:155–169] to estimate the extinction coefficient fromsolar height. The second method differed from the first onlyin the way in which the daily fraction of interceptedphotosynthetic active radiation (FPAR) was calculated, asproposed by Charles-Edwards and Lawn (Charles-EdwardsDA, Lawn RJ (1984) Plant Cell Environ 7:247–251]. In thethird method, we used a remote sensing technique[MOD15_BU-collection 4, produced and distributed byEROS Data Center Distributed Active Archive Center(EDC DAAC)]. We found that the first and the secondmethods revealed the expected LAI dynamics, whichincreased during the dry–wet transition and wet season,and decreased during the wet–dry transition and dry season.From 20 randomly distributed sets in a 1.0 ha area, only 3showed significant differences in LAI estimated from thefirst two methods; conversely, LAI was overestimated bythe third method.Keywords Radiation interception . Photosyntheticallyactive radiation . Extinction coefficientIntroductionOne of the most important characteristics of canopystructure is leaf area index (LAI), the total one-sidedfoliage area per unit soil surface area. LAI is critical forthe estimation of a number of important ecosystemprocesses, including CO 2 flux, evapotranspiration, rainfallinterception, and litter deposition (Chason et al. 1991).There are two main types of procedure used to estimateLAI: direct and indirect methods. Direct methods consist ofdestructive or non-destructive field measurements assessingthe leaves themselves, while indirect methods derive LAI

110 Int J Biometeorol (2011) 55:109–118from models based on measured variables other than actualleaves. Both methods have their specific problems andlimitations. The decision of which method to use dependson many factors such as: the required accuracy, the timeavailable for performing measurements, the research scale,the available budget, etc. (Jonckheere et al. 2004), andother, site-specific restrictions, e.g., some methods are notsuitable for LAI estimates in tropical forests (Wasseige etal. 2003).The use of micrometeorological towers with sensorsinstalled to measure radiation at different times and pointsis an effective method of estimating global or interceptedphotosynthetically active radiation (PAR) captured by thecanopy, which, according to Monsi and Saeki (1953), canbe used to estimate LAI in natural ecosystems, especiallyforests. Based on the Beer-Lambert law of light extinction,these sensors must be installed in representative locationswithin the environment, especially in dense forests withheterogeneous vegetation.Sometimes, if a micrometeorological tower is notavailable, and where it is not possible to gather continuousradiation data, alternative methods must be used. Charles-Edwards and Lawn (1984) proposed an alternative methodthat made possible to estimate the daily fraction ofintercepted radiation by the canopy using only an instantaneousmeasurement at midday. This method not onlyallowed the use of portable sensors and reduced the timerequired for taking measurements, but also allowed morethan a single point in the field to be used for LAIestimation, enhancing the accuracy of this estimate.Remote sensing techniques have been developed toestimate biophysical or biochemical attributes at thecanopy or leaf color scale (Kodani et al. 2002). Moderateresolution imaging spectroradiometer (MODIS) is a sensoraboard the National Aeronautics and Space Administration(NASA) Terra satellite, which provides estimates ofsoil cover, and these estimates can be used to calculateLAI, daily fraction of intercepted PAR (FPAR), grossprimary productivity (GPP) and net primary productivity(NPP), which makes it extremely convenient for use inconjunction with simulation models of ecosystems. Presently,the emphasis within the MODIS program is onvalidation of the algorithms and their products (Privette etal. 2002). The MODIS LAI algorithm is based on sixcanopy architectural types: grasses and cereal crops;shrubs; broadleaf crops; savannas; broadleaf forests; andneedle leaf forests (Myneni et al. 2002).In this paper, we compare different methods of estimatingLAI in a transitional tropical forest in the AmazonBasin, testing three of the methods most accessible to manyresearches working in natural ecosystems. The methodsevaluated were that of Monsi and Saeki (1953) with FPARobtained from continuous measurements of radiation,Monsi and Saeki (1953) with FPAR obtained by themethod proposed by Charles-Edwards and Lawn (1984),and the remote sensing technique MOD15_BU-collection 4produced and distributed by EROS Data Center DistributedActive Archive Center (EDC DAAC).Material and methodsSite descriptionThe experiment was conducted in a transitional forestlocated 50 km NE of Sinop (11°24.75′S:55°19.50′W—coordinated from a micrometeorological tower in thetransitional forest, Fig. 1), Mato Grosso, Brazil.The vegetation is composed of semi deciduous treespecies, such as Brosimum lactescens, Qualea paraensisand Tovomita schomburkii, with a maximum height of28 m (Sanches et al. 2008). The 30-year mean annualtemperature is 24°C with little seasonal variation, andrainfall is 2 m year −1 with a dry season between May andSeptember, and a wet season between October and April.The sites are located on the Araguaia formation, with thesame soil texture (Oxisol soil; Marcelino et al. 2005). Thesoil near the forest tower site is a quartzarenic neosolcharacterized by sandy texture (84% sand, 4% silt, and12% clay in the upper 50 cm of soil; Priante-Filho et al.2004). The soils are quite sandy, poor in nutrients, theyhave high porosity and drain rapidly following rainfallevents (Vourlitis et al. 2002).Meteorological and ancillary measurementsA 40 m micrometeorological tower was installed in thestudy area equipped with different kind of sensorscontrolled by a datalogger (model CR 10X; CampbellScientific, Logan, UT) and recorded on a memory module(model SM16M; Campbell Scientific).PAR sensors (model LI-190; LI-COR Bioscience, Lincoln,NE) were installed at heights of 1, 20, and 40 m above theground in the micrometeorological tower. These sensorsprovided all data employed to calculate LAI using the methodproposed by Monsi and Saeki (1953) withFPARobtainedfrom continuous measurements of radiation.To estimate LAI by the method proposed by Monsi andSaeki (1953) with FPAR obtained by the proposition ofCharles-Edwards and Lawn (1984), 20 random sets werechosen in a 1.0 ha area (Fig. 2). At those set points,measurement of LAI below the canopy was performedusing a PAR sensor (model LI-190; LI-COR Bioscience)coupled to a datalogger (model CR10; Campbell Scientific)programmed to measure and record instantaneous PAR at arate of 1.0 s. For each set point, the average value of ten

Int J Biometeorol (2011) 55:109–118 111Fig. 1 Location of the areastudied, approximately50 km N of Sinop city, MatoGrosso state, Brazil. The micrometeorologicaltower (graycircle) was located in a transitionalforest at 11°24.75′S and55°19.50′Wmeasurements was taken as the instantaneous data. Thetime required to measure all 20 set points was from 11:00 a.m. to 1:00 p.m. (GMT−4). PAR measured above the canopywas taken from the micrometeorological tower.Air temperatures and relative humidities were taken40 m above the ground at the micrometeorologicaltower (12–14 m above the forest canopy) using ashielded relative humidity sensor (model HMP-35;Vaisala, Helsinki, Finland).Precipitation was measured every 30 min at the top ofthe tower using a tipping-bucket rainfall gauge (model TE-525; Texas Electronics, Dallas, TX); however, gaps in datacollection precluded use of the rainfall data measured onsite, and data obtained from a manual rain gauge that wasread daily at the Fazenda Continental located 5 km E of thestudy site was used instead. Rainfall data collected by themanual rainfall gauge was highly correlated to datacollected on-site (Vourlitis et al. 2008).The atmospheric vapor pressure deficit (VPD) wascalculated as the difference between saturation vaporpressure and actual vapor pressure from temperature andhumidity measurements made at the top of the tower.Estimate of LAIEstimates of LAI using the Monsi and Saeki (1953)method with FPAR calculated from continuous PARrecord, or FPAR estimated by the proposition of Charles-Edwards and Lawn (1984) were calculated from August2007 to June 2008.Equation 1 was used to calculate FPAR from total dailycumulated PAR above and below the canopy, and Eq. 2was used when instantaneous data of PAR at noon wasrequired.ðFPAR ¼ PAR 40 PAR 40r Þ PAR 1ð1ÞðPAR 40 PAR 40r ÞWhere FPAR is the fraction of daily interceptedradiation, PAR 40 is the incident daily PAR on the top ofthe canopy (mol m −2 day −1 ); PAR 40r is the daily PARreflected by the canopy (mol m −2 day −1 ); PAR 1 is the dailyPAR transmitted through the canopy (mol m −2 day −1 ).ð PAR 40rn Þ PAR 1nPAR 40rn Þð Þ PAR 1nðPAR 40n PAR 40rn ÞFPAR ¼ 2 PAR 40nðPAR 40n1 þ PAR 40n PAR 40rnð2ÞWhere FPAR is the fraction of daily interceptedradiation, PAR 40n is the incident instantaneous PAR atnoon on the top of the canopy (μmol m −2 s −1 ); PAR 40rn isthe instantaneous PAR at noon reflected by the canopy(μmol m −2 s −1 ); PAR 1n is the instantaneous PAR at noontransmitted through the canopy (μmol m −2 s −1 ).The LAI was calculated by the fraction of dailyintercepted PAR, according to Eq. 3.LAI ¼ lnð1FPARÞð3ÞKWhere LAI is the leaf area index (m 2 m −2 ), FPAR is thefraction of daily intercepted radiation; K is the extinctioncoefficient of the canopy.

112 Int J Biometeorol (2011) 55:109–118Fig. 2 Location of 20 samplingpoints distributed randomly overa 1-ha area in the transitionalforestThe extinction coefficient (k) estimated is an estimateproposed by Goudriaan (1988), and considers the randomand circular distribution of the leaves in the canopy. Theinterception and extinction of light through the canopy canbe derived from the ratio between the average projection ofthe leaves in the direction of the solar beam (O) and theelevation sun angle based on the horizon (Β) according toEq. 4.k ¼Oð4ÞsenBFor a distribution consisting of leaf angle, the averageprojection was determined based on the integration over theangle distribution of leaves (F) between zero and π/2,simplified according to Eq. 5.OðBÞ ¼F 1 O 1 ðBÞþF 2 O 2 ðBÞþF 3 O 3 ðBÞð5ÞWhere, F 1 , F 2 and F 3 are the relative distributionfrequencies of leaf inclination in classes of 15°, 45° and75°, respectively, each covering 30°, and of the sphericaldistribution, F 1 , F 2 and F 3 with values of 0.134, 0.366 and0.5, respectively. According to Goudriaan (1988) O 1 andO 2 are dependent on solar elevation angle; this approximatesto a combination of a sine wave and a constantlower limit, according to Eqs. 6 and 7.O 1 ¼ maxð0:26; 0:93senðBÞÞ ð6ÞO 2 ¼ maxð0:47; 0:68senðBÞÞ ð7ÞThe O 3 value is a reference projection for a sphericaldistribution, which regardless of the solar elevation, shouldbe equal to 0.5. Based on these frequencies, the equation ofO 3 was estimated according to Eq. 8.O 3 ¼ 1 0:268O 1 0:732O 2 ð8ÞThe solar elevation angle defined as the angle betweenthe direction of the sun and the horizon (β) was determinateaccording to Eq. 9.B ¼ 90 Z ð9Þwhere, Z is the zenith angle, and Z depends on day time,place latitude and year time and was estimated according toEq. 10.cosZ ¼ senB ¼ sen8 sen dþ cos8 cosdcos½15ðt t 0 ÞŠ ð10Þwhere, 8 is the local latitude, δ is the solar declination, t isthe time and t o is the sunset.The solar declination is calculated according to Eq. 11.sin d ¼ 0:39785 sin½278:97 þ 0:9856J þ 1:9165 sinð356:6 þ 0:9856JÞŠwhere, J is the Julian day.Estimate of LAI by remote sensingð11ÞLAI products from MODIS are produced with a spatialresolution of 1 km on a Sinusoidal grid. Values areprovided as a daily product (MOD15A1) and as a collatedmaximum value over a period of 8 days (MOD15A2), andas a monthly average product (MOD15_BU). The MODISLAI algorithm is based on three-dimensional radiativetransfer theory and developed by inversion using a look-

Int J Biometeorol (2011) 55:109–118 113Fig. 3 Daily cycle for eachmonth of a incoming solar radiation(W m −2 ), b air temperature(°C), c relative humidity (%)and d vapor pressure deficit(VPD; kPa) from August 2007to July 2008

114 Int J Biometeorol (2011) 55:109–118Table 1 Average (±standard deviation) monthly global radiation, air temperature, relative humidity and vapor pressure deficit (VPD) from August2007 to July 2008Month/year Global radiation (W m −2 ) Air temperature (°C) Relative humidity (%) VPD (kPa)August/2007 570.0±386.4 24.0±6.0 59.1±21.6 1.4±1.3September/2007 460.9±315.9 25.3±5.4 64.5±21.1 1.3±1.3October/2007 479.6±329.1 25.3±3.8 40.8±2.9 0.37±0.32November/2007 487.4±298.0 24.6±3.1 79.7±12.6 0.62±0.62December/2007 453.7±259.5 24.2±2.5 82.8±10.6 0.47±0.47January/2008 529.4±272.5 23.5±2.3 85.6±8.8 0.42±0.41February/2008 475.5±283.2 23.6±2.4 85.4±9.3 0.32±0.3March/2008 439.6±265.1 23.8±2.6 85.3±10.11 0.11±0.26April/2008 425.2±286.2 24.3±2.8 84.2±11.5 0.4±0.5May/2008 419.3±283.5 23.9±3.6 79.4±14.4 0.6±0.4June/2008 499.1±348.5 23.6±5.0 71.4±20.7 1.0±1.1July/2008 535.1±374.0 24.1±5.6 60.4±20.7 1.4±1.3up table approach. The product used in this study isMOD15_BU-collection 4, produced and distributed to thepublic by the EROS Data Center Distributed ActiveArchive Center (EDC DAAC, available at http://daac.ornl.gov/cgi-bin/MODIS/GR_col5_1/mod_viz.html). The studyarea is located at coordinates −11.41S and −55.32W, whichincludes the micrometeorological tower. For each cell, theLAI value was analyzed and the quality assessed (QA),which determined the type of algorithm used or theinexistence of data. The monthly LAI was obtained by thearithmetic mean of four grid points, when data existedbased on five levels of the QA value for each cell. Toreduce noise and obtain a higher percentage of clear-skydata, MODIS products are distributed as maximum valuecomposites of 8-day ranges. Nevertheless, in the humidtropical regions, availability of high quality consistent timeseries remains low; in the case of our study site area (SinopMato Grosso, Brazil) only a part (∼50%) of the compositesof the evaluated 1-year period have the highest QA level of0 and 1.Analysis methodMeteorology (global radiation, air temperature, humidityand VPD) were summarized as diurnal averages calculatedby averaging each 30-min datum for a particular time (e.g.,0900–0930 hours LT) over monthly intervals to obtain acomposite average trend in meteorology. This averagingprocess was conducted to allow the utilization of all datacollected and to provide an indication of how the LAI andmeteorology variables varied, on average, during thedifferent hydrological periods.Assuming the dry season is the number of months withrainfall

Int J Biometeorol (2011) 55:109–118 115Table 3 Linear correlation matrix for average monthly incoming radiation, average monthly air temperature, average monthly relative humidity,average monthly VPD and leaf area index (LAI). Correlation coefficients were calculated over monthly intervals (n=12 months)Global radiation Air temperature Relative humidity VPD LAIGlobal Radiation 1Air temperature −0.182 1Relative humidity −0.408 −0.617* 1VPD 0.589* 0.176 −0.502 1LAI −0.097 0.364 0.090 −0.468 1*P

116 Int J Biometeorol (2011) 55:109–118Fig. 6 LAI estimated by theLambert-Beer method (blackcircles) and by product MODIS(white circles) ina 2007 and b2008between air temperature and relative humidity, and betweenair temperature and VPD.Seasonality of LAIThe LAI estimated by the micrometeorological data fromAugust 2007 to June 2008 resulted in a maximum value inthe wet season (4.25 m 2 m −2 ) and a minimum value in thedry season (3.32 m 2 m −2 ; Fig. 4). LAI was higher in the wetand dry–wet seasons due to an increase in forest vegetationgrowth as a consequence of increased water availability.Furthermore, during the wet season there was a higherglobal radiation (Table 1). Light is essential for photosynthesisand other plant processes, and consequently plays animportant role in driving growth (King et al. 2005).The behavior pattern of the vertical profile of the LAIdemonstrated reduced values in the lower stratum. Onaverage, in the upper stratum, the LAI was 20% higher (atheight 40 m; Fig. 5) due to the foliage density at a canopyheight of 28–40 m.Another reflectance index that is commonly related toforest LAI is the LAI product MODIS. Satellite-derivedestimates of MODIS have been used extensively to observethe green-up and brown-up of terrestrial ecosystems and tomap LAI in time and space. Figure 6 shows the relationshipbetween estimated LAI and LAI product MODIS. Figure 7shows the estimated LAI and LAI product MODIS in a boxplot figure.Average estimated LAI using the product MODIS was3.4 m 2 m −2 from January to May. The months with the mostreliable data were June and July—a period with lesscloudiness and also high radiation. The months with lessreliable data were January, February, March and April—months with higher rainfall rates, when estimates wereinfluenced by cloudiness. There was no significant correlationbetween estimates using different methods. Using theproduct MODIS, the mean LAI values identified in forestareas were low, which could be due to clouds, which arenot filtered by the algorithm, generating low LAI valuesdue to high reflectance, consequently reducing the meanLAI. The lower values compared with LAI estimated by theLambert-Beer method were likely due to influences such asnoise effects (atmospheric path radiances), the exhibition ofasymptotic (saturated) signals over high biomass condi-Fig. 7 LAI estimated by theLambert-Beer method and productMODIS displayed as a boxplot figure in a 2007 and b2008. Dotted line Mean, solidline median

Int J Biometeorol (2011) 55:109–118 117Table 4 Paired t-test of average of LAI obtained by the Monsi andSaeki method (at the point of the tower) and Charles-Edwards andLawn method (on 20 sampling points). A t-test was used to analyzethe relationship between the tower and the spatial variation on 20sampling points (t-test, t=2.30). P is the interval confidenceLAI average at thetower (Monsi andSaeki method)SamplingpointsAverage LAI over 20sampling points(Charles-Edwards andLawn method)PFig. 8 Correlation between LAI calculated from radiation measures atthe 20 sampling points (Charles-Edwards and Lawn method) withthose calculated from the radiation data taken from the micrometeorologicaltower (Monsi and Saeki method). Dotted line 1:1tions, and the sensitivity of this method to canopybackground variations (Aragão et al. 2005). The interfacebetween ecosystems and the atmosphere is very difficult toquantify due to spatial variability (vertical and horizontal)and temporal variability: annual cycles and inter-yearvariation interact with structural patterns, stratification andhomogeneity (Bréda 2003). Factors such as the differencesin absorption should be considered. While the MODISalgorithm determines LAI from the absorption of radiationby leaves only, LAI estimated from micrometeorologicaldata takes into account PAR absorbed by leaves (Senna andCosta 2005).The higher spatial resolution of the MODIS data mayenable more accurate derivation of surface longwaveradiation because it provides more detailed informationabout the atmosphere and land surface than sensors with acoarser spatial resolution (Wang and Liang 2008). Due tothe high resolution (down to 1 km), land surface modelsusing MODIS data have been applied to natural ecosystems;however, the spatial resolution that specifies the pixelsize of satellite images covering the Earth’s surface for theLAI data MODIS algorithm determines LAI at 1 km spatialresolution, and in this paper LAI estimated by fieldobservation was determined at a single point of observation.Thus, the monthly LAI was obtained by the arithmeticmean of four grid points in the transitional forest to reducethe possible influence of spatial resolution of MODISproducts.The LAI estimated by the Lambert-Beer method resultedin an asymmetric distribution skewed towards the right(Fig. 8). The distance from the third quartile to the medianis more than the distance from the medium to the firstquartile, suggesting that 75% of the LAI estimates liebetween the third quartile and the upper limit. The LAIestimated by MODIS product resulted in an asymmetric4.2 Point 1 4.5 0.704.2 Point 2 4.4 0.844.2 Point 3 4.0 0.794.2 Point 4 4.6 0.414.2 Point 5 4.9 0.244.2 Point 6 4.5 0.574.2 Point 7 4.3 0.884.2 Point 8 4.6 0.354.2 Point 9 4.5 0.594.2 Point 10 5.4 0.04*4.2 Point 11 5.0 0.05*4.2 Point 12 5.2 0.05*4.2 Point 13 5.1 0.02*4.2 Point 14 4.9 0.03*4.2 Point 15 4.9 0.174.2 Point 16 4.4 0.674.2 Point 17 4.1 0.904.2 Point 18 4.7 0.084.2 Point 19 3.8 0.324.2 Point 20 4.2 0.93*P0.05).Spatial variation of LAIA t-test (P>0.05) was used to analyze the relationshipbetween the tower point and spatial variation based on 20sampling points distributed randomly in a 1-ha area(Table 4). During December 2007, there was a higherspatial LAI variability X I:C: 95% ¼ 5:1 0:6m 2 m 2probably due to marked horizontal variations in radiationavailability—often only 1–2% of full sunlight penetrated tothe understory.

118 Int J Biometeorol (2011) 55:109–118There were significant differences between LAI estimatedby the Monsi and Saeki method taken from the towerpoint and LAI taken from the 20 sampling points,especially at points 10, 11, 13 and 14.Uncertainty related to the LAI estimate at the canopylevel must be noted. LAI could be underestimated if solarradiation passed through the canopy and focused directlyon the sensor installed below the canopy. It would thenprovide an estimate fraction with minimal values or zeroand this would lead to smaller LAI estimates. Alternatively,LAI could be overestimated if there was a physicalimpediment other than leaves. This would lead to overestimatesof the fraction intercepted.ConclusionsLAI estimated from randomly distributed sampling pointsin a 1-ha area was significantly correlated in 17/20 pointswith LAI estimated at the tower point.The above correlation notwithstanding, the LAI productfrom MODIS is important for monitoring and modelingglobal change and terrestrial dynamics at many scales.However, cloud contamination effects on MODIS LAIproduct need to be taken into account carefully beforeanalyzing LAI patterns. The effect of cloudiness can lead toerroneous interpretation of the results. In this study, LAIestimated using the MODIS product underestimated thevalues. The correlation between LAI estimated using theMODIS product and the Lambert-Beer method resulted insignificant differences.Acknowledgments This research was supported in part by UniversidadeFederal de Mato Grosso (UFMT), Conselho Nacional deDesenvolvimento Científico e Tecnológico (CNPq), and the Coordenaçãode Aperfeiçoamento de Pessoal de Nivel Superior (CAPES).ReferencesAragão LEOC, Shimabukuro YE, Espírito-Santo FDB, Williams M(2005) Spatial validation of the collection 4 MODIS LAI productin Eastern Amazonia. IEEE Trans Geosci Remote Sens 43(11):2526–2534Bréda NJJ (2003) Ground-based measurements of leaf area index: areview of methods, instruments and current controversies. J ExpBot 54:2403–2417Charles-Edwards DA, Lawn RJ (1984) Light interception by grainlegume row crops. Plant Cell Environ 7:247–251Chason JW, Baldocchi DD, Huston MA (1991) A comparison ofdirect and indirect methods for estimating forest canopy leaf area.Agric For Meteorol 57:107–128Goudriaan J (1988) The bare bones of leaf-angle distribution inradiation models for canopy photosynthesis and energy exchange.Agric For Meteorol 43:155–169Jonckheere I, Fleck S, Nackaerts K, Muysa B, Coppin P, Weiss M,Baret F (2004) Review of methods for in situ leaf area indexdetermination part I. Theories, sensors and hemisphericalphotography. Agric For Meteorol 121:19–35. doi:10.1016/j.agrformet.2003.08.027King DA, Davies SJ, Supardi MNN, Tan S (2005) Tree growth isrelated to light interception and wood density in two mixeddipterocarp forests of Malaysia. Funct Ecol 19:445–453Kodani E, Awaya Y, Tanaka K, Matsumura N (2002) Seasonalpatterns of canopy structure, biochemistry and spectral reflectancein a broad-leaved deciduous Fagus crenata canopy. ForEcol Manage 167:233–249Marcelino T, Shirawa S, Borges WR, Priante-Filho N, Raiter F (2005)GPR para verificação do nível d’água subterrânea em transiçãoFloresta Amazônica e Cerrado. Acta Amazonica 35:369–376Monsi M, Saeki T (1953) Über den Lichtfakor in den Pflanzengesellschaften,seine Bedeutung für die Stoffproduckion. Jpn J Bot14:22–52Myneni RB, Hoffman S, Knyazikhin Y, Privette JL, Glassy J, TianY,WangY,SongS,ZhangY,SmithGR,LotschA,FrieldM,Morrisette JT, Votava P, Nemani RR, Running SW (2002)Global products of vegetation leaf area and fraction absorbedPAR from year one of MODIS data. Remote Sens Environ83:214–231Priante-Filho N, Vourtilis GL, Hayashi MMS, Nogueira JS, Campelo-Junior JH, Nunes PC, Sanches L, Couto EG, Hoeger W, Raiter F,Trienweiler JL, Miranda EJ, Priante PC, Fritzen CL, Lacerda M,Pereira LC, Biudes MS, Suli GS, Shiraiwa S, Paulo SR, SilveiraM (2004) Comparison of the mass and energy exchange of apasture and a mature transitional tropical forest of the southernAmazon Basin during a seasonal transition. Glob Chang Biol10:863–876Privette JL, Myneni RB, Knyazikhin Y, Mukufute M, Roberts G, TianY, Wang Y, Leblanc SG (2002) Early spatial and temporalvalidation of MODIS LAI product in Africa. Remote SensEnviron 83:232–243Sanches L, Valentini CMA, Pinto-Júnior OB, Nogueira JS, VourlitisGL, Biudes MS, Silva CJ, Bambi P, Lobo FA (2008) Seasonaland interannual litter dynamics of a tropical semideciduous forestof the southern Amazon Basin, Brazil. J Geophys Res 113(2008):G04007. doi:10.1029/2007JG000593Senna MCA, Costa MH (2005) Fraction of photosynthetically activeradiation absorbed by Amazon tropical forest: a comparison offield measurements, modeling, and remote sensing. J GeophysRes 110:G01008. doi:10.1029/2004JG000005Shuttleworth WJ, Gash JHC, Lloyd CR et al (1988) Eddy correlationmeasurements of energy partitioning for Amazonian forest. Q J RMeteorol Soc 110:1143–1162Vourtilis GL, Priante-Filho N, Hayashi MMS, Nogueira JS, Caseiro F,Campelo-Júnior JH (2002) Seasonal variations in the evapotranspirationof a transitional tropical Forest of Mato Grosso, Brazil.Water Resour Res 38. doi:10.1029/2000WR000122Vourlitis GL, Nogueira JS, Lobo FA, Sendall KM, Faria JLB, DiasCAA, Andrade NLR (2008) Energy balance and canopyconductance of a tropical semi-deciduous forest of the southernAmazon Basin. Water Resour Res 44:W03412. doi:10.1029/2006WR005526Wang W, Liang S (2008) Estimation of high-spatial resolution clearskylongwave downward and net radiation over land surfacesfrom MODIS data. Remote Sens Environ 113(4):745–754.doi:10.1016/j.rse.2008.12.004Wasseige C, Bastin D, Defourny P (2003) Seasonal variation oftropical forest LAI based on field measurements in CentralAfrican Republic. Agric For Meteorol 119:181–194