1
BIOLOGICAL SCIENCES: Ecology, Environmental Sciences
Large Seasonal Swings in Leaf Area
of Amazon Rainforests
Ranga B. Myneni*, Wenze Yang*, Ramakrishna R. Nemani‡, Alfredo R. Huete§, Robert
E. Dickinson¶†, Yuri Knyazikhin*, Kamel Didan§, Rong Fu¶, Robinson I. Negrón Juárez¶,
Sasan S. Saatchi║, Hirofumi Hashimoto**, Kazuhito Ichii††, Nikolay V. Shabanov*, Bin
Tan*, Piyachat Ratana§, Jeffrey L. Privette‡‡, Jeffrey T. Morisette§§, Eric F. Vermote¶¶,‡‡,
David P. Roy║║, Robert E. Wolfe***, Mark A. Friedl*, Steve W. Running†††, Petr
Votava**, Nazmi El-Saleous‡‡‡, Sadashiva Devadiga‡‡‡, Yin Su*, Vincent V.
Salomonson§§§
*
Department of Geography and Environment, Boston University, 675 Commonwealth
Avenue, Boston, MA 02215, USA.
‡
Ecosystem Science and Technology Branch, NASA Ames Research Center, Mail Stop
242-4, Moffett Field, CA 94035, USA.
§
Department of Soil, Water and Environmental Science, University of Arizona, Tucson,
AZ 85721, USA.
¶
School of Earth and Atmospheric Sciences, Georgia Institute of Technology, 311 Ferst
Drive, Atlanta, GA 30332, USA.
║
Jet Propulsion Laboratory, California Institute of Technology, 4800 Oak Grove Drive,
Pasadena, CA 91109, USA.
**
California State University at Monterey Bay and Ecosystem Science and Technology
Branch, NASA Ames Research Center, Mail Stop 242-4, Moffett Field, CA 94035, USA.
††
San Jose State University and Ecosystem Science and Technology Branch, NASA
Ames Research Center, Mail Stop 242-4, Moffett Field, CA 94035, USA.
‡‡
Biospheric Sciences Branch, NASA Goddard Space Flight Center, 8600 Greenbelt
Road, Mail Code 614.4, Greenbelt, MD 20771, USA.
§§
Terrestrial Information Systems Branch, NASA Goddard Space Flight Center, 8600
Greenbelt Road, Mail Code 614.5, Greenbelt, MD 20771, USA.
2
¶¶
Department of Geography, University of Maryland, College Park, MD 20742, USA.
║║
Geographic Information Science Center of Excellence, South Dakota State University,
Wecota Hall, Box 506B, Brookings, SD 57007, USA.
***
Raytheon TSC at NASA Goddard Space Flight Center, 8600 Greenbelt Road, Mail
Code 614.5, Greenbelt, MD 20771, USA.
†††
School of Forestry, University of Montana, Missoula, MT 59812, USA.
‡‡‡
Science Systems and Applications Inc., NASA Goddard Space Flight Center, 8600
Greenbelt Road, Mail Code 614.5, Greenbelt, MD 20771, USA.
§§§
Senior Scientist (Emeritus) of NASA Goddard Space Flight Center and Research
Professor, Department of Geography and Meteorology, University of Utah, Salt Lake
City, Utah 84112-0110.
†
To whom correspondence should be addressed. Robert E. Dickinson, School of Earth
and Atmospheric Sciences, Georgia Institute of Technology, 311 Ferst Drive, Atlanta,
GA 30332, USA; Telephone: 404-385-1509; Fax: 404-385-1510; E-mail:
robted@eas.gatech.edu.
This manuscript has 12 pages and 3 figures. The abstract contains 183 words, and the
manuscript contains 14,365 characters (without space, and not including supporting
information).
Supporting information has 18 pages, 1 table, and 4 figures. The supporting text contains
25,691 characters (without space).
Abbreviations: MODIS, moderate resolution imaging spectroradiometer; LAI, leaf area
index; CERES, Clouds and the Earth's Radiant Energy System; GOES, Geostationary
Operational Environmental Satellite; TRMM, Tropical Rainfall Measuring Mission.
.
3
Despite early speculation to the contrary, all tropical forests studied to date display
seasonal variations in the presence of new leaves, flowers and fruits. These past
studies were focused on the timing of phenological events and their cues, but not on
the accompanying changes in leaf area which regulate vegetation-atmosphere
exchanges of energy, momentum and mass. Here we report, from analysis of five
years of recent satellite data, seasonal swings in green leaf area of about 25% in a
majority of the Amazon rainforests. That is, leaf area equivalent to nearly 28% the
size of South America appears and disappears each year in the Amazon. This
seasonal cycle is timed to the seasonality of solar radiation in a manner that is
suggestive of anticipatory and opportunistic patterns of net leaf flushing during the
light rich dry season and net leaf abscission during the cloudy wet season. These
heretofore unknown seasonal swings in leaf area are critical to initiation of the
transition from dry to wet season, seasonal carbon balance between photosynthetic
gains and respiratory losses, and litterfall nutrient cycling in moist tropical forests.
4
The trees of tropical rainforests are known to exhibit a range of phenological behavior,
from episodes of ephemeral leaf bursts followed by long quiescent periods to continuous
leafing, and from complete intraspecific synchrony to complete asynchrony (1). Several
agents - herbivory, water stress, day length, light intensity, mineral nutrition, flood pulse,
etc. - have been identified as proximate cues for leafing and abscission in these
communities (1-8) These studies were focused on the timing of phenological events but
not on the accompanying changes in leaf area. Leaves selectively absorb solar radiation,
emit longwave radiation and volatile organic compounds, and facilitate growth by
regulating carbon dioxide influx and water vapor efflux from stomates. Therefore, leaf
area dynamics are relevant to studies of climate, hydrological and biogeochemical cycles.
The sheer size and diversity of rainforests preclude a synoptic view of leaf area
changes from ground sampling. We therefore used data on green leaf area of the Amazon
basin (approximately 7.2  106 km2) derived from measurements of the Moderate
Resolution Imaging Spectroradiometer (MODIS) onboard the National Aeronautics and
Space Administration’s Terra satellite (9, 10). These data were expressed as one-sided
green leaf area per unit ground area (leaf area index, LAI).
Results
Seasonality in Leaf Area Index Time Series. Leaf area data of the Amazon rainforests
exhibit notable seasonality, with an amplitude (peak to trough difference) that is 25% of
the average annual LAI of 4.7 (Fig. 1A). This average amplitude of 1.2 LAI is about
twice the error of a single estimate of MODIS LAI and thus is not an artifact of remote
5
observation or data processing (Supporting Text, which is published as supporting
information on the PNAS web site). The aggregate phenological cycle appears timed to
the seasonality of solar radiation in a manner that is suggestive of anticipatory and
opportunistic patterns of leaf flushing and abscission. These patterns result in leaf area
leading solar radiation during the entire seasonal cycle, higher leaf area during the shorter
dry season when solar radiation loads are high, and lower leaf area during the longer wet
season when radiation loads decline significantly. This seasonality is roughly consistent
with the hypothesis that in moist tropical forests, where rainfall is abundant and herbivore
pressures are modest, seasonal increase in solar radiation during the dry season might act
as a proximate cue for leaf production (1, 2, 4).
In a community dominated by leaf-exchanging (11) evergreen trees, leaf area can
increase if some of the older leaves that are photosynthetically less efficient because of
epiphylls and poor stomatal control are exchanged for more numerous new leaves. Leaf
area can decrease if the new leaves are less numerous than the older ones that are
dropped. If such exchanges are staggered in time amongst the individuals over a large
area, for example due to asynchrony (7), they can result in a gradually increasing spatially
averaged leaf area over a period of several months during the ascending phase of the
seasonal cycle, and a gradually decreasing leaf area during the descending phase, while
maintaining the evergreen character of the rainforest (Fig. 1A). These patterns of net leaf
flushing and abscission also generate higher leaf litterfall in the dry season relative to the
wet season, as reported (12-14). Such a leaf strategy will enhance photosynthetic gain
6
during the light rich dry season (15-20), provided the trees are well hydrated (2), and
reduces respiratory burden during the cloudy wet season.
Leaf area changes in the adjacent grasslands and savannas in Brazil are
concordant with rainfall data (Fig. 1B) - higher leaf area in the wet season and lower leaf
area in the dry season. This expected behavior imbues confidence in the opposing
seasonality of deep rooted and generally well hydrated (2), but light limited (2, 4, 18, 19),
rainforests inferred from the same LAI data set.
Geographic Details of Leaf Area Changes. The satellite data provide geographic details
of leaf area changes in the Amazon (Fig. 2A). The region with a distinct seasonality of
leaf area spans a broad contiguous swath of land that is anchored to the Amazon river,
from its mouth in the east to its western-most reaches in Peru, in the heart of the basin.
This pattern is notable for at least two reasons. First, for its homogeneity – a higher dry
season leaf area relative to the wet season is observed in about 58% of all rainforest
occupied pixels, while only 3% show the opposite change (Fig. 2B). Second, the
homogeneous region roughly overlies the precipitation gradient (21) in the basin
(Supporting Text and Fig. 4C, which are published as supporting information on the
PNAS web site), suggesting that the amplitude is, to a first approximation, independent of
the duration and intensity of the dry season. For example, an amplitude of about one LAI
unit is observed in areas with two to five dry months in a year. Ostensibly, these forests
maintain high leaf area (20, 22) and remain well hydrated during the dry season in nondrought years (Supporting Text and Fig. 6, which are published as supporting information
7
on the PNAS web site) via their deep root systems (2, 23) and/or through hydraulic
redistribution (24, 25). Similar changes are not seen in about 40% of the rainforest pixels,
some of which are transitional and drier rainforests to the south and east.
Correlation among Changes in Leaf Area, Solar Radiation and Precipitation. To
associate quantitatively the changes in leaf area, solar radiation and precipitation, we
correlated successive monthly differences of these variables, first using the spatially
averaged data shown in Fig. 1A, and second, using pixel level data. Changes in LAI are
both positively correlated with changes in solar radiation (p<0.0001) and negatively
correlated with precipitation changes (p<0.0001), but the correlations between leaf area
and radiation changes are larger and at the pixel level more numerous (Fig. 3 and Fig. 7,
which is published as supporting information on the PNAS web site). The negative
correlations between LAI and precipitation are likely an indirect effect of the changes in
cloudiness and radiation associated with precipitation changes (18). These results,
together with the past phenological studies, support the idea of an evolved pattern of
endogenously controlled vegetative phenology that is timed to the seasonality of solar
radiation (2, 11).
Discussion
The consistency between leaf area, solar radiation and precipitation data from different
satellite instruments is especially noteworthy. However, the strong seasonality in cloud
cover and tropospheric aerosol loading may introduce seasonally opposing artifacts in
MODIS leaf area. To minimize the impact of significant wet season cloud cover in the
8
Amazon, we used a coarse resolution – eight kilometer and monthly - data set that was
derived by averaging the best quality LAI values from the standard one kilometer, eight
day MODIS data set (Supporting Text, which is published as supporting information on
the PNAS web site). Although some of the coarse resolution LAI values were based on
fewer high quality estimates in the wet season, this did not bias the inferred seasonal LAI
amplitudes.
The high aerosol content in the dry season, from biomass burning, natural
biogenic emissions, and soil dust re-suspension (26) can result in artificially low LAI
values, unless the reflectance data are corrected for aerosol effects. The MODIS
processing system was found to correct well for such effects (Supporting Text, which is
published as supporting information on the PNAS web site). The LAI values may have
been underestimated by about 5% from any residual aerosol effects. This effect is small
and of opposite timing relative to the observed seasonality. Other possible sources of
bias, such as reflectance saturation at high leaf area and changes in light scattering and
absorption properties of leaves due to aging and epiphylls (27), were found to be small,
and with the wrong timing, to significantly alter our estimates of the amplitude of LAI
seasonality (Supporting Text, which is published as supporting information on the PNAS
web site).
The average seasonal range of 1.2 LAI recorded over 4.2  106 km2 (58%) of
Amazon rainforests implies a seasonal fluctuation in one-sided green leaf area of about
5.0  106 km2. That is, leaf area equivalent to nearly 28% the size of South America
9
appears and disappears each year in the Amazon. The greener dry season can enhance the
buoyancy of surface air, and thus convection, through higher latent heat fluxes, and
initiate transition to the wet season (28, see also Supporting Text, which is published as
supporting information on the PNAS web site). Also, the seasonal dynamics and interplay
between canopy photosynthesis and ecosystem respiration will be altered by this
unexpected seasonality in leaf area (12, 15-20, 29), with attendant consequences for
litterfall nutrient cycling (30). Therefore, it is important to investigate the significance of
these changes for climate, hydrological and biogeochemical cycles, and whether such
similar swings in leaf area also exist in the moist forests of Africa and Asia.
This work was supported by grants from the National Aeronautics and Space
Administration to the authors.
10
1. van Schaik, C. P., Terborgh, J. W. & Wright, S. J. (1993) Annu. Rev. Ecol Syst. 24,
353-377.
2. Wright, S. J. (1996) in Tropical Forest Plant Ecophysiology, Eds. Stephen, S. M.,
Robin, L. C. & Alan, P. S. (Chapman & Hall, New York), pp. 440-460.
3. Morellato, P. C. (2003) in Phenology: An Integrative Environmental Science, Ed.
Schwartz, M. (Kluwer Academic Publishers, The Netherlands), pp. 75-92.
4. Wright, S. J. & van Schaik, C. P. (1994) Am. Nat. 143, 192-199.
5. Borchert, R. (1994) Ecology 75, 1437-1449.
6. Schöngart, J., Piedade, M. T. F., Ludwigshausen, S., Horna, V. & Worbes, M. (2002)
J. Trop. Ecol. 18, 581-597.
7. Singh, K. P. & Kushwaha, C. P. (2005) Curr. Sci. 89, 964-975.
8. Borchert, R., Rivera, G. & Hagnauer, W. (2002) Biotropica 34, 27-39.
9. Salomonson, V. V., Barnes, W. L., Maymon, P. W., Montgomery, H. E., Ostrow, H.
(1989) IEEE Trans. Geosci. Remote Sens. 27, 145-153.
10. Supporting information on materials and methods is available on PNAS web set.
11. Borchert, R. (2000) in Dormancy in Plants: From Whole Plant Behaviour to Cellular
Control Eds. Viemont, J. D. & Crabbe, J. (CAB International, Wallingford), pp. 87107.
12. Goulden, M. L., Miller, S. D., da Rocha, H. R., Menton, M. C., Freitas, H. C.,
Figueira, A. M. S. & de Sousa, C. A. D. (2004) Ecol. Appl. 14 supplement, S42-S54.
13. Luizao, F. J. (1989) GeoJournal 19, 407-417.
14. Rodrigues, W. A., Furch, K. & Klinge, H. (2001) Amazoniana 16, 441-462.
15. Carswell, F. E. (2002) J. Geophys. Res. 107, doi:10.1029/2000JD000284.
11
16. Würth, M. K. R., Peláez-Riedl, S., Wright, S. J. & Körner, C. (2005) Oecologia 143,
11-24.
17. Graham, E. A., Mulkey, S. S., Kitajima, K., Phillips, N. G. & Wright, S. J. (2003)
Proc. Nat. Acad. Sci. 100, 572-576.
18. Schuur, E. A. G. (2003) Ecology 84, 1165-1170.
19. Nemani, R. R., Keeling, C. D., Hashimoto, H., Jolly, W. M., Piper, S. C., Tucker, C.
J., Myneni, R. B. & Running, S. W. (2003) Science 300, 1560-1563.
20. Huete, A. R. Didan, K., Shimabukuro, Y. E., Ratana, P., Saleska, S., Yang, W.,
Nemani, R. R., Myneni, R. B., Hutyra, L. & Fitzjarrald, D. (2006) Geophys. Res.
Lett., 33, L06405, doi:10.1029/2005GL025583.
21. Sombroek, W. (2001) Ambio 30, 388-396.
22. Asner, G. P., Nepstad, D. C., Cardinot, G. & Ray, D. (2004) Proc. Nat. Acad. Sci.
101, 6039-6044.
23. Nepstad, D. C., de Carvalho, C. R., Davidson, E. A., Jipp, P. H., Lefebvre, P. A.,
Negreiros, G. H., da Silva, E. D., Stone, T. A., Trumbore, S. E. & Vieira, S. (1994)
Nature 372, 666-669.
24. da Rocha, H. R., Goulden, M. L., Miller, S. D., Menton, M. C., Pinto, L. D. V. O., de
Freitas, H. C. & Silva Figueira, A. M. E. (2004) Ecol. Appl. 14 supplement, S22-S32.
25. Oliveira, R. S., Dawson, T. E., Burgess, S. S. O. & Nepstad, D. C. (2005) Oecologia
doi 10.1007/s00442-005-0108-2.
26. Echalar, F., Artaxo, P., Martins, J. V., Yamasoe, M., Gerab, F., Maenhaut, W. &
Holben, B. (1998) J. Geophys. Res. 103(D24), 31,849-31,864.
27. Roberts, D. A., Nelson, B. W., Adams, J. B. & Palmer, F. (1998) Trees 12, 315-325.
12
28. Fu, R. & Li, W. (2004) Theor. Appl. Climatol. 78, 98-110.
29. Seleska, 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. et al. (2003)
Science 302, 1554-1557.
30. Vitousek, P. M. & Sanford, R. L. Jr. (1986) Ann. Rev. Col. Syst. 17, 137-167.
Fig. 1. (A) Time series of monthly leaf area index from the Terra MODIS instrument (in
green), monthly maximum of hourly average surface solar radiation from the Terra
CERES and GOES-8 instruments (in red), and monthly merged precipitation from the
Tropical Rainfall Measuring Mission (TRMM) and other sources (in blue). The time
13
series are data averaged over all Amazon rainforest pixels, as identified in the MODIS
landcover map (Fig. 4B, which is published as supporting information on the PNAS web
site), south of the equator. The start of the data record is March 2000 and the end points
are September 2005 (leaf area index), May 2005 (solar radiation) and August 2005
(precipitation). The shaded areas
denote dry seasons, defined as months with
precipitation less than 100 mm or less than one-third the precipitation range
[0.33*(maximum-minimum) + minimum]. The solar radiation data are for all sky
conditions and include direct and diffuse components. (B) Same as (A) except that the
data are from savanna and grassland pixels adjacent to the Amazon basin in Brazil and
south of the equator (Fig. 4B, which is published as supporting information on the PNAS
web site). The shaded areas denote dry seasons, defined as months with precipitation less
than 50 mm. Information on the data can be found in Supporting Text, which is published
as supporting information on the PNAS web site (10).
Fig. 2. (A) Color coded map of leaf area index (LAI) amplitudes greater than 0.66 or less
than -0.66 – this threshold (|0.66|) is the smallest LAI difference discernable with the
MODIS LAI data set (Supporting Text, which is published as supporting information on
the PNAS web site). The amplitude, in regions with dry seasons longer than three
months, is calculated as the difference between the maximum four-month average LAI in
the dry season minus the minimum four-month average LAI in the wet season. Where the
dry season is three or fewer months, the amplitude is calculated as the difference between
the dry season average LAI and the minimum four-month average LAI in the wet season.
The dry and wet seasons are defined based on the precipitation data set at quarter degree
14
spatial resolution (Supporting Text, which is published as supporting information on the
PNAS web site). Thus, the seasons vary spatially and inter-annually. (B) The distribution
of LAI amplitude for all Amazon rainforest pixels. The color scheme is similar to that in
panel (A).
Fig. 3. (A) Correlation between first differences of leaf area index (LAI) and solar
radiation. The first differences of LAI [LAI(t)] are calculated as LAI(t+1)-LAI(t), where
t is months in the timeline March 2000 to May 2005. The number of data points is 62 for
each pixel. Correlation coefficients greater than 0.25, or less -0.25, are shown (p<0.05).
The analysis was performed for rainforest pixels with LAI amplitudes greater than 0.66
or less than -0.66 - this threshold (|0.66|) is the smallest LAI difference discernable with
the MODIS LAI data set (Supporting Text, which is published as supporting information
on the PNAS web site). (B) Correlation between first differences of LAI and
precipitation.
15
A
Leaf Area Index
950
5.0
900
850
4.5
800
4.0 300
Precipitation 750
(mm/mo)
200
Solar Radiation (W/m²)
1000
5.5
100
2000
2001
2002
2003
2004
2005
2006
3.0
1000
2.5
900
2.0
800
1.5
300
Precipitation
(mm/mo)
200
100
2000
Fig. 1.
2001
2002
2003
2004
2005
2006
700
Solar Radiation (W/m²)
Leaf Area Index
B
16
A
-1.2 -1.1 -1.0 -0.9 -0.8 -0.66 0 0.66 0.8 0.9 1.0
Seasonal Amplitude of Leaf Area Index
1.2 1.4
B
7
% Forest Area
6
5
4
3
2
1
0
-4.0
-3.0
-2.0
-1.0
0.0
1.0
2.0
3.0
Seasonal Amplitude of Leaf Area Index
Fig. 2.
4.0
17
A
-0.8
-0.6 -0.325 -0.25
0
0.25 0.325 0.6
Correlation Coefficient for Radiation and LAI
0.8
-0.8
-0.6 -0.325 -0.25
0
0.25 0.325 0.6
0.8
Correlation Coefficient for Precipitation and LAI
B
Fig. 3.
18
Supporting Text
Leaf Area Index Data. We used a continuous record of data on green leaf area from the
Moderate Resolution Imaging Spectroradiometer (MODIS) onboard the National
Aeronautics and Space Administration’s (NASA) Terra satellite to track leaf area
changes over the Amazon basin, from March 2000 to September 2005. The MODIS
instrument measures, amongst others, solar radiation reflected by the Earth’s surface and
the atmosphere at different narrow wavelength bands (1). The instrument has an on-board
calibrator and a dedicated state-of-the-art processing system that geo-locates the data (2),
does cloud filtering (3) and corrects for atmospheric contamination of surface reflectance
(4). The surface reflectances are then input to a suite of algorithms for generation of
several geophysical data sets, also called products, on an operational basis in near-real
time, with explicit pixel level quality flags (5, 6).
Green leaf area index (LAI) is defined as the one-sided green leaf area per unit
ground area in broadleaf canopies and as half the total needle surface area per unit ground
area in coniferous canopies. It is one of the standard operational data sets generated at
one kilometer spatial resolution, every day, for nearly the entire land surface of the Earth
(7). Not all pixels, however, have values of sufficient quality because of sub-pixel cloud
cover, residual atmospheric corruption, algorithm performance, etc. Therefore, the best
quality daily retrievals are averaged over an eight day period and distributed from the
Earth
Resources
Observation
and
(http://edc.usgs.gov/products/satellite/modis.html).
Science
The
MODIS
Data
data
Center
have
been
19
reprocessed several times, each time applying the latest available version of the science
algorithms to the MODIS instrument data and using the best available calibration and
geolocation information. The data sets generated during each reprocessing belong to a
collection. In this study, the latest collection 4 LAI data set was used.
Over five years of Terra MODIS data were acquired, processed and analyzed to
obtain the results presented in this paper. During these five years, the MODIS instrument
acquired over 125 terabytes of data globally. Almost one petabyte (about 3 million files)
of geophysical data, including land surface reflectance, were generated to support
production of global LAI products, 4% of which covers the Amazon region.
Our research on MODIS LAI spans a 10 year period, starting with the design of
the algorithm in the mid-1990s, to its implementation, product quality assessment and
validation. The physics of the algorithm and subsequent refinements are described in
several articles (8-13). Algorithm performance and quality assessment have also been
reported in refereed literature (7, 14-17). The LAI data have been validated through
comparisons with field measurements from several sites, representative of major
vegetation types, around the world (Table 1).
LAI estimates from MODIS data are obtained from either the main or back-up
algorithm. The main algorithm is based on a three-dimensional description of radiation
transfer and interactions in vegetation canopies, and solves the inverse problem of LAI
estimation using Look-Up-Tables (8). When this algorithm fails, a back-up algorithm
based on biome-specific NDVI-LAI relations is used. The main algorithm fails in cases
20
of cloud or aerosol contaminated reflectances and thus LAI estimates from back-up
algorithm are generally incorrect (17). The five year average success rate of the main
algorithm is about 67% globally, but is only 30% over the Amazon basin because of
persistent cloud cover (17). We therefore created an eight kilometer monthly data set by
averaging the cloud free main algorithm LAI estimates available in the standard one
kilometer eight day data set. For each eight kilometer monthly pixel, a quality index was
derived as the percentage of cloud free best quality retrievals. For example, if 35% of the
one kilometer eight day pixels met these criteria, these were averaged to generate a LAI
value with a 35% quality index. About 70% of the eight kilometer monthly pixels had
quality indices greater than 50%. This improved, but coarser spatial and temporal
resolution, data set was used in this study. The annual average green leaf area index
derived from this data set is shown in Fig. 4A.
Recently, three research teams collected field measurements suitable for
validation of MODIS LAI near the Tapajos FLUXNET site in the Amazon (13). The
BigFoot team (18) performed measurements on February 26 in year 2002 over a 7x7 km
area near the Tapajos FLUXNET site (center lat=-2.8697450, lon=-54.943550). Huete et
al. (19) performed similar measurements in July of year 2000 at several 250 m transects
near the KM67 tower at the same site (lat=-2.8566670, lon =-54.9588890). The mean
MODIS LAI over the Tapajos site is 5.6  0.9 (late February) and 5.2  1.4 (early July).
The corresponding ranges of field measurements are 5.4 to 7.0 and 5.8 to 6.7,
respectively. This suggests satisfactory agreement between MODIS LAI and ground
measurements at the sampling locations.
21
A third group (20) performed measurements along transects in the Tapajos
National Forest (13000 km2, upper-left lat=-2.750 & lon=-55.500, lower-right=-4.000 &
lon=-54.900) which included the Tapajos FLUXNET site. They found MODIS LAI
values to be large overestimates in relation to their measurements. Some of this
discrepancy is due to the comparison of dry season MODIS LAI values with wet season
field measurements. The transect measurements (mean LAI of 3.8) also disagreed
significantly with field measurements by the other two research teams (mean LAI of 6.2).
These disagreements could also be due to differences in the scale of observations,
experimental design (plots versus transects), and potential errors in scaling point ground
measurements to one km2 area averages.
Landcover Data. The landcover map describes global distribution of vegetation classes
using the International Geosphere-Biosphere Programme’s (IGBP) scheme of landcovers.
The map is generated from one year of Terra MODIS data at one km resolution (21). This
data set is available from the Earth Resources Observation and Science Data Center
(http://edc.usgs.gov/products/satellite/modis.html). We aggregated the IGBP vegetation
types to the following four classes - forests, savannas, other vegetation and nonvegetated. The resulting map is shown in Fig. 4B. This data set was used to identify
rainforest and savanna pixels in our analysis. All identified classes had a confidence flag
(21) value of at least 50%.
22
Precipitation Data. Monthly precipitation data at quarter degree spatial resolution for the
period January 1998 to August 2005 were obtained from the NASA Goddard Space
Flight Center (http://lake.nascom.nasa.gov/data/dataset/TRMM/). This data set is a
merged quality product generated from instruments of the Tropical Rainfall Measuring
Mission (TRMM) and other sources (22). The spatial variability of precipitation in the
Amazon can be represented (23) as the distribution of areas under different number of dry
months (Fig. 4C). Months with precipitation less than 100 mm may be considered dry (23,
24). We used a slightly different definition – months with precipitation less than 100 mm
or less than one-third the precipitation range [0.33*(maximum-minimum) + minimum].
This ensured a dry season of at least one month in the more humid regions of western
Amazon.
Solar Radiation Data. Monthly solar radiation data at one degree spatial resolution for
the period March 2000 to May 2005 were obtained from the Langley Atmospheric
Sciences Data Center (http://eosweb.larc.nasa.gov/PRODOCS/ceres/table_ceres.php).
These Level 3 data (25) are monthly means of hourly downward at-surface solar radiation
fluxes. They were generated from the Terra satellite’s Clouds and the Earth's Radiant
Energy System (CERES) data and Geostationary Operational Environmental Satellite-8
(GOES) derived broad band fluxes (CERES Product SFC R4V3). The GOES derived
fluxes have been normalized to the CERES fluxes to reduce discrepancies. The radiation
data used here were derived from observations corresponding to all-sky conditions. When
compared to data from 36 global stations, they have a bias of 1.5% and a regional root
mean square scatter of 9.6% (Personal Communication by Dr. Doeling of NASA Langley
23
Research Center). The maximum values of the monthly means of hourly solar radiation
data were used in this study, given the difficulties associated with precisely estimating
surface solar radiation from top-of-the-atmosphere measurements under cloudy
conditions in the Amazon. This selection of maximum values resulted in utilization of the
best quality data and resolved the seasonal cycle well as ascertained by comparisons to
similar data from the International Satellite Cloud Climatology Project (26).
Uncertainty Analysis. The MODIS LAI data have been validated against ground
measurements of leaf area in a host of vegetation types across the globe (12, 16, 27-32).
Table 1 summarizes these validation activities. The following regression is obtained
when field measurements from 29 different sites, representative of major vegetation types
across different continents, are regressed against the corresponding MODIS LAI values
(33),
MODIS LAI = 1.12*Field LAI + 0.11 ,
(1)
R2 = 0.87, RMSE = 0.66 LAI, N = 29.
(2)
In words, the MODIS LAI values may overestimate field measurements by about 12% in
dense vegetation. The MODIS values can explain 87% of the variability of LAI between
sites, and on average MODIS values will be in error in their estimation by 0.66 LAI.
The seasonal leaf area amplitudes reported in this paper are differences between
MODIS LAI estimates (dry season maximum LAI and wet season minimum LAI). At
24
any given pixel, we would expect from Eqs. (1) and (2) that on average the MODIS
estimation would be in error by 0.66 LAI. If a single independent dry season and wet
season estimation are made, their difference would change this error by 21/2 or about 0.84
LAI. However, normally more than one time sample will be available and we are
examining many points in space. The overall error given this multiple sampling is
difficult to estimate but likely to be around 0.1 to 0.3 LAI However, to be conservative,
we use the simple RMS estimate of 0.66 LAI and consider any difference less than this
threshold to be within the estimation error of MODIS LAI. Thus, two MODIS LAI values
that differ by less than |0.66| LAI are treated as about equal, but differences greater than
+0.66 LAI, or less than -0.66 LAI, between two MODIS values are assumed to
correspond to a true change in field measured LAI. The amplitude in Fig. 1A is about 1.2
LAI, which is nearly two times the threshold value of |0.66| LAI. Figures 2A and 3 depict
results for regions with amplitudes greater than this threshold only.
The factor of 1.12 in Eq. (1) is itself uncertain and contributes to the overall
uncertainty. Its uncertainty corresponds to 0.66 / N1/2 = 0.11. Thus, the MODIS LAI
could be giving little bias in its estimation of leaf area or as much as double that indicated
by Eq. (1).
Clouds, Aerosols, Reflectance Saturation and Leaf Spectral Effects.
A. Clouds: The MODIS data processing system has a well developed cloud screening
algorithm that uses 14 of the 36 MODIS spectral bands to maximize the reliability of
cloud detection (3). Information on pixel cloud state is contained in the quality flags
25
accompanying the data sets. Globally, about 65 to 75% of the vegetated pixels are cloud
free, depending on the time of the year, 10% are partially cloudy and the rest are cloud
covered (17). In the Amazon region, however, there is much more cloudiness, and this
varies greatly between the wet and dry seasons. This seasonality in cloud cover can bias
the results if cloud contaminated retrievals are not screened out from the analysis.
Atmospheric water vapor is not a problem as the MODIS channel data that are used for
land remote sensing completely avoid the water vapor absorption bands.
The percentage of cloud free, partially cloudy and cloud covered pixels in the
standard one kilometer eight day LAI data set is about 82, 11 and 7 in the dry season and
32, 17 and 51 in the wet season over the Amazon rainforests. All cloudy retrievals, and
some partially cloudy retrievals, are of suspect quality. Therefore, we created an eight
kilometer monthly data set by averaging the cloud free main algorithm LAI retrievals
available in the standard data set. For each eight kilometer monthly pixel, a quality index
was derived as the percentage of best quality cloud free retrievals. For example, if 35%
of the one kilometer eight day pixels met these criteria, these were averaged to generate a
LAI value with a 35% quality index.
In the dry season, about 95% of the pixels had quality indices greater than 50%.
In the wet season, only about 45% of the pixels had quality indices greater than 50%,
because of enhanced cloud cover. This suggests a higher proportion of pixels with lower
quality indices in the wet season. A lower quality index may correspond to a biased
coarse resolution LAI value because it is based on fewer higher resolution samples, but
26
not necessarily, given the relative homogeneity of LAI in the rainforests and the random
spatial occurrence of clouds. Nevertheless, it is instructive to see the difference in
average LAI between pixels of low and high quality indices, in order to ascertain if pixels
with lower quality indices do provide a biased estimate of spatially averaged LAI.
The average LAI difference between pixels with quality indices greater than 75%
(higher quality pixels) and less than 25% (lower quality pixels) is only 0.4 in the wet
season and even less (0.16) in the dry season. These differences are less than the smallest
true difference between any two MODIS LAI values (|0.66|, cf. described above), and
therefore, should be treated as not significantly different than zero. Thus, we conclude
that the results derived from the eight kilometer monthly LAI data are not an artifact of
cloud seasonality in the Amazon.
B. Aerosols: There is considerable variation in tropospheric aerosol loading between the
dry and wet seasons in the Amazon. The aerosol optical depth inferred from MODIS data
(34) indicates this seasonal difference to be as large as 0.5 (Fig. 5A). The high aerosol
loading in the dry season is due to biomass burning, natural biogenic emissions, and soil
dust re-suspension (35). A question arises as to how this seasonality in aerosol loading
impacts LAI retrievals, and if the inferred LAI seasonality is an artifact of aerosol
seasonality. The following analysis was performed to address this question.
We focus our investigation on a 1200 x 1200 km region in the Amazon, where
93% of the vegetation is classified as rainforests. The top-of-the-atmosphere MODIS
27
measurements are corrected for atmospheric effects, including aerosol scattering, to
derive the surface reflectances used in LAI estimation (4). The efficacy of this correction
depends on the amount of aerosol loading. The derived surface reflectances are quality
flagged as either residual-aerosol or aerosol-free, that is, reflectances contaminated with
residual aerosol effects or free of such effects after the atmospheric correction procedure
has been applied. Aerosol scattering tends to enhance reflectance over low reflective
surfaces, in both red and near-infrared channels – the two principal inputs to the LAI
algorithm (Fig. 5B). However, the difference between aerosol-free and residual-aerosol
reflectances does not show a seasonal pattern. Thus, the correction appears not to be
biased by the seasonal changes in aerosol optical depth level. The difference in LAI
values estimated from these two groups of reflectances, across the year, is quite small,
about -5%. Thus, residual aerosol effects may have resulted in an underestimation of
LAI, on an average, by 5% and this is likely to happen in the dry season because of high
aerosol optical depths (Fig. 5A). This effect is small, and importantly, opposite in
direction, when compared to the LAI seasonality reported in the manuscript. Thus, we
conclude that the leaf area amplitudes reported in the manuscript are not an artifact of
tropospheric aerosol seasonality.
C. Reflectance Saturation: In dense vegetation canopies, such as the Amazon
rainforests, the reflectances tend to an asymptotic value, that is, they saturate. The
MODIS algorithm is designed to identify such conditions and retrieve the threshold LAI
value at which reflectances saturate (8). The incidence of reflectance saturation in the
rainforest increased, on an average, by 11% in the dry season relative to the wet season.
28
This suggests higher leaf areas in the dry season compared to the wet season, in
agreement with the results presented in the manuscript. But because of reflectance
saturation, the reported amplitudes are likely underestimates, that is, the actual leaf area
difference between the dry and wet seasons could be somewhat larger.
D. Changes of Canopy Reflectance from Changing Leaf Optical Properties: The
MODIS LAI algorithm estimates LAI by use of red and near-infrared reflectances (7).
Leaf optical properties might change either because of changes of the leaf itself or of
other materials on top of the leaf. For example, observations by Roberts et al. (36)
demonstrated a change of leaf optical properties because of covering by epiphylls. Xiao
et al. (37) interpreted the observed dry season increases in canopy reflectance near the
Tapajos observational tower to result from the fall of old leaves and the emergence of
new leaves with a consequent reduction of such epiphyte cover. In addition, water drops
are more likely to remain on leaves during the wet season. The MODIS algorithm
assumes all leaves are the same but if actual leaves had properties that differed either
with age, water drops and epiphyll cover or location in canopy, the estimated LAI values
could be potentially biased.
The study by Roberts et al. (36) although in a rainforest area near Manaus was for
a forest growing on waterlogged and nutrient poor soil with low species diversity and low
biomass, and so was referred to as Caatinga, a form of vegetation adopted to semi-arid
conditions and with very long lived leaves. Because no other similar studies have been
done, its implications need to be addressed. They observed time history of leaf optical
29
properties over a period of 16 months. New leaves, because of their thinness, had
substantially lower reflectances in the near-infrared, initially by about 10% according to
their measurements. The reflectance peaked after several months and then began to
slowly decline from the gradual development of a coating of epiphytes (in this case
primarily large patches of dark fungi). After 2-3 months some of the leaves began to
become covered by epiphytes, after 6-9 months the majority of the leaves had a slight
covering of epiphytes, with a few classified as moderate or necrotic (near-dead), and by
the end of the sample period, few leaves were entirely uncoated. They show in their Fig.
2 that leaf reflectance can change from 10% for moderate to 50% for heavy colonization;
the latter condition was apparently uncommon enough not to be reflected in their overall
statistics, but after a year up to 20-40% of their leaves could be in the moderate or
necrotic category (their Fig. 7). We estimate that the overall average change of the
reflectances of the leaves they were sampling would have been about 3% in 6-months and
about 5% in a year. While the system they consider may not be entirely unusual, most of
the rainforest is different. They assert that “with the higher species diversity of terra firme
forest …considerable variation in phenology, leaf turnover and susceptibility to epiphylls,
it is unlikely that epiphylls would have the same impact at the canopy scale.” They
indeed mention that an examination of leaves at another site of secondary forest found no
significant impact of epiphytes.
The above study indicates that one year old leaves would have somewhat lower
reflectances than six month old leaves. Adding new darker leaves (most exposed to solar
irradiation) can only further reduce canopy reflectances. Maximum reflectances would be
30
expected several months later as the new leaves mature and before they are significantly
colonized. Adding water drops to a canopy would enhance its reflectance a small amount
because of the additional scattering surface exposed. In conclusion, the overall
seasonality of leaf reflectances is likely too small, and with the wrong timing, to
significantly affect this paper’s estimate of the amplitude of LAI seasonality, but if there
is significant seasonality in canopy reflectance, it would probably be such as to lead to an
underestimation of the seasonality of LAI. Even if we were to assume that the
colonization observed by Roberts et al. (36) after six months goes away by the dry
season, and that new leaves are the same as mature leaves, and that all the Amazon has
the properties of the observed caatinga forest, the consequent seasonality still appears to
be too small to account for the observed seasonality of the near-infrared reflectances from
the Amazon forest. Moreover, canopy reflectance is independent of leaf optical properties
in the limit leaf albedo is close to unity, and depends only on LAI. This further suggests
that the reported changes in near-infrared leaf optical properties are unlikely to be the
source of observed seasonal variations in surface reflectances, and therefore LAI.
Vegetation Control of Dry to Wet Season Transition. The onset of wet season controls
year-to-year changes in Amazon rainfall (38). What drives the onset of wet season, and in
particular, whether or not the land surface/vegetation plays an active role in the transition
from dry to wet season, is an important question given that Amazon rainfall contributes
approximately 20% to global rainfall and river discharge.
31
There is emerging evidence that the rainforest plays a critical role in initiating the
onset of wet season over the Amazon. Rainfall begins to increase prior to the reversal of
large-scale atmospheric wind (39). The latter brings moisture from the Atlantic Ocean
into the Amazon, which further increases the rainfall and leads to wet season onset. Thus,
such an increase of rainfall at the end of dry season is essential to the establishment of
wet season large-scale atmospheric circulation pattern (40).
An increase of surface evapotranspiration (ET) at the end of dry season appears to
be the primary cause for increasing the buoyancy of surface air, which consequently
increases the probability of atmospheric convection and rainfall (41). Most of the severe
delays in onset of wet seasons during the 1980s were traced to a weaker increase of ET at
the end of the dry season (42). In situ observations from different parts of the Amazon
have consistently shown a higher ET in the forested area during the dry season compared
to the wet season (43) (Fig. 6), possibly due to higher dry season leaf area and/or solar
radiation load. The 25% increase in LAI over nearly 60% of the Amazon rainforest
during the dry season reported in this paper therefore suggests a potentially important
vegetation control on the initiation of the wet season.
32
1. Salomonson, V. V., Barnes, W. L., Maymon, P. W., Montgomery, H. E. & Ostrow, H.
(1989) IEEE Trans. Geosci. Remote Sens. 27, 145-153.
2. Wolfe, R. E., Nishihama, M., Fleig, A. J., Kuyper, J. A., Roy, D. P., Storey J. C. &
Patt, F. S. (2002) Remote Sens. Environ. 83, 31-39.
3. Platnick, S., King, M. D., Ackerman, S. A., Menzel, W. P., Baum, B. A., Riedi, J. C. &
Frey, R. A. (2003) IEEE Trans. Geosci. Remote Sens. 41(2), 459-473.
4. Vermote, E. F., El-Saleous, N. & Justice, C. O. (2002) Remote Sens. Environ. 83, 97111.
5. Justice, C. O., Townshend, J. R. G., Vermote, E. F., Masuoka, E., Wolfe, R. E., ElSaleous, N. Roy, D. P. & Morisette, J. T. (2002) Remote Sens. Environ. 83, 3-15.
6. Roy, D. P., Borak, J. S., Devadiga, S., Wolfe, R. E., Zheng, M. & Descloitres, J.
(2002) Remote Sens. Environ. 83, 62-76.
7. Myneni, R. B., Hoffman, S., Knyazikhin, Y., Privette, J. L., Glassy, J., Tian, Y., Wang,
Y., Song, X., Zhang, Y. & Smith, G. R. et al. (2002) Remote Sens. Environ. 83, 214231.
8. Knyazikhin, Y., Martonchik, J. V., Myneni, R. B., Diner, D. J. & Running, S. W.
(1998) J. Geophys. Res. 103, 32,257-32,274.
9. Panferov, O., Knyazikhin, Y., Myneni, R.B., Szarzynski, J., Engwald, S., Schnitzler,
K.G. & Gravenhorst, G. (2001) IEEE Trans. Geosci. Remote Sens. 39, 241-253.
10. Wang, Y., Tian, Y., Zhang, Y., El-Saleous, N., Knyazikhin, Y., Vermote, E. &
Myneni, R. B. (2001) Remote Sens. Environ. 78, 296-311.
11. Tian, Y., Wang, Y., Zhang, Y., Knyazikhin, Y., Bogaert, J. & Myneni, R. B. (2002)
Remote Sens. Environ. 84, 143-159.
12. Wang, Y., Woodcock, C.E., Buermann, W., Stenberg, P., Voipio, P., Smolander, H.,
Hame, T., Tian, Y., Hu, J., Knyazikhin, Y. & Myneni, R.B. (2004) Remote Sens.
Environ. 91, 114-127.
13. Shabanov, N. V., Huang, D., Yang, W., Tan, B., Knyazikhin, Y., Myneni, R. B., Ahl,
D. E., Gower, S. T., Huete, A. & Aragao, L. E. O. C. et al. (2005) IEEE Trans.
Geosci. Remote Sens. 43, 1855-1865.
14. Tian, Y., Zhang, Y., Knyazikhin, Y., Myneni, R. B., Glassy, J. M., Dedieu, G. &
Running, S.W. (2000) IEEE Trans. Geosci. Remote Sens. 38, 2387-2401.
33
15. Tian, Y., Dickinson, R. E., Zhou, L., Zeng, X., Dai, Y., Myneni., R. B., Knyazikhin,
Y., Zhang, X., Friedl., M. A. & Yu, H. et al. (2004) J. Geophys. Res. 109, D01103,
doi:10.1029/2003JD003777,.
16. Tan, B., Huang, D., Hu, J., Yang, W., Zhang, P., Shanabov, N. V., Knyazikhin, Y.,
Nemani, R. R. & Myneni, R. B. (2006) Agric. For. Meteorol. 135, 124-134.
17. Yang, W., Huang, D., Tan, B., Stroeve, J. C., Shabanov, N. V., Knyazikhin, Y.,
Nemani, R. R. & Myneni, R. B. (2006) IEEE Trans. Geosci. Remote Sens., in press.
18. BigFoot project home, at http://www.fsl.orst.edu/larse/bigfoot/index.html.
19. Huete, A. R., Patana, P., Didan, K., Shimabukuro, Y., Barbosa, H., Ferreira, L. &
Miura, T. (2003) Anais Xi SBSR, 05–10, April (pp. 665– 672). Belo Horizonte,
Brasila Instituto Nacional de Pesquisas Espaciais.
20. Aragão, L. E. O. C., Shimabukuro, Y. E., Espirito-Santo, F. D. B. & Williams, M.
(2005) IEEE Trans. Geosci. Remote Sens. 43, 2526-2533.
21. Friedl, M. A., McIver, D. K., Hodges, J. C. F., Zhang, X. Y., Muchoney, D., Strahler,
A. H., Woodcock, C. E., Gopal, S., Schneider, A. & Cooper, A. (2002) Remote Sens.
Environ. 83, 287-302.
22. Huffman, G. J., Adler, R. F., Rudolph, B., Schneider, U. & Keehn, P. (1995) J. Clim.
8, 1284-1295.
23. Sombroek, W. (2001) Ambio 30(7), 388-396.
24. Zhang, X., Friedl, M. A., Schaaf, C. B., Strahler, A. H., Liu, Z. (2005) J. Geophys.
Res. 110, D12103, doi:10.1029/2004JD005263.
25. Wong, T., Young, D. F., Minnis, P., Cess, R. D., Ramanathan, V., Mitchum, M. V.,
Doelling, D. R., Gibson, G. G. & Sullivan, S. E. (1997) CERES ATBD Subsystem
10.0. at http://asd-www.larc.nasa.gov/ATBD/ATBD.html.
26. Zhang, Y-C., Rossow W. B., Lacis A. A., Oinas V. & Mishchenko M. I. (2004) J.
Geophys. Res. 109, D19105, doi:10.1029/2003JD004457.
27. Privette, J. L., Myneni, R. B., Knyazikhin, Y., Mukufute, M., Roberts, G., Tian, Y.,
Wang, Y. & Leblanc, S. G. (2002) Remote Sens. Environ. 83, 232-243.
28. Tian, Y., Woodcock, C. E., Wang, Y., Privette, J. L., Shabanov, N. V., Zhou, L.,
Zhang, Y., Buermann, W., Dong, J. & Veikkanen, B. et al. (2002) Remote Sens.
Environ. 83, 414-430.
34
29. Tian, Y., Woodcock, C. E., Wang, Y., Privette, J. L., Shabanov, N. V., Zhou, L.,
Zhang, Y., Buermann, W., Dong, J. & Veikkanen, B. et al. (2002) Remote Sens.
Environ. 83, 431-441.
30. Shabanov, N. V., Wang, Y., Buermann, W., Dong, J., Hoffman, S., Smith, G. R., Tian,
Y., Knyazikhin, Y. & Myneni, R. B. (2003) Remote Sens. Environ. 85, 410-423.
31. Tan, B., Hu, J., Zhang, P., Huang, D., Shabanov, N. V., Weiss, M., Knyazikhin, Y. &
Myneni, R. B. (2005) J. Geophys. Res. 110, D01107, doi:10.1029/2004JD004860.
32. Huemmrich, K. F., Privette, J. L., Mukelabai, M., Myneni, R. B. & Knyazikhin, Y.
(2005) Int. J. Remote Sens. 26, 4381-4398.
33. Yang, W., Tan, B., Huang, D., Rautiainen, M., Shabanov, N. V., Wang, Y., Privette, J.
L., Huemmrich, K. F., Fensholt, R. & Sandholt, I. et al. (2006) IEEE Trans. Geosci.
Remote Sens., in press.
34. Remer, L. A., Kaufman, Y. J., Tanrè, D., Mattoo, S., Chu, D. A., Matins, J. V., Li, RR., Ichoku, C., Levy, R. C. & Kleidman, R. G. et al. (2005) J. Atmos. Sci. 62, 947973.
35. Echalar, F., Artaxo, P., Martins, J. V., Yamasoe, M., Gerab, F., Maenhaut, W. &
Holben, B. (1998) J. Geophys. Res. 103(D24), 31,849-31,864.
36. Roberts, D. A., Nelson, B. W., Adams, J. B. & Palmer, F. (1998) Trees 12, 315-325.
37. Xiao, X., Zhang Q., Saleska S., Hutyra L., de Camargo P., Wofsy S., Frolking S.,
Boles S., Keller M. & Moore III, B. (2005) Remote Sens. Environ. 94, 105-122.
38. Marengo, J. A., Liebmann, B., Kousky, V. E., Filizola, N. P. & Wainer, I. C. (2001) J.
Climate 14, 833-852.
39. Horel, J. D., Hahmann, A. N. & Geisler, J. E. (1989) J. Climate 2, 1388-1403.
40. Lenters, J. D. & Cook, K. H. (1997) J. Atmos. Sci. 54, 656-678.
41. Li, W. & Fu, R. (2004) J. Climate 17, 2637-2651.
42. Fu, R. & Li, W. (2004) Theor. Appl. Climatol. 78, 97-110, doi: 10.1007/s00704-0040046-7.
43. Negrón Juárez, R. I., Hodnett, H. & Fu, R. (2006) J. Geophys. Res., in press.
44. Scholes, R. J., Frost, P. G. H. & Tian, Y. (2004) Global Change Biol. 10, 292-302.
35
45. Cohen, W. B., Maiersperger, T. K., Yang, Z., Gower, S. T., Turner, D. P., Ritts, W.
D., Berterretche, M. & Running, S. W. (2003) Remote Sens. Environ. 88, 233-255.
46. Fensholt, R., Sandholt, I. & Rasmussen, M. S. (2004) Remote Sens. Environ. 91, 490507.
47. Kang, S., Running, S.W., Lim, J.-H., Zhao, M., Park, C.-R. & Loehman, R. (2003)
Remote Sens. Environ. 86, 232-242.
48. Pandya, M. R., Chaudhari, K. N., Singh, R. P., Sehgal, V. K., Bairag, G. D., Sharma,
R. & Dadhwal, V. K. (2003) Curr. Sci. 85, 1777-1782.
49. Fernandes, R., Leblanc, S., Butson, C., Latifovic, R. & Pavlic, G. (2002) IEEE Int.
Geosci. Remote Sens. Sym. 4, 2097-2099.
50. Fang, H. & Liang, S. (2005) Remote Sens. Environ. 94, 405-424.
51. Baret, F., Weiss, M., Moreno, J., Chen, J., Pavageau, K., Beal, D., Berthelot, B. &
Gonzalez, M. C. (2005) Report on the validation of MERIS TOA_VEG land
products, INRA, Avignon, France.
36
Table 1. MODIS LAI validation activities.
Site
Vegetation Type
IGBP Kalahari Transect (22)
Semiarid woodlands and savannas
IGBP Kalahari transect, Mongu, Zambia Woodlands
(31)
IGBP Kalahari transect (44)
Six savanna sites on deep, sandy soils,
along a gradient of increasing aridity
Ruokolahti, Finland (12)
Needle leaf forest mixed with large and
small lakes
Alpilles, France (30)
Agricultural site
BOREAS Northern Study Area,
Cropland; Prairie grassland; Boreal needle
Canada; Harvard Forests, MA, USA;
leaf and Temperate mixed forests
Konza, KS, USA; Bondville, IL, USA
(45)
Bondville, IL, USA (16)
Agricultural site
Senegal, Western Sudano-Sahelian zone Grass-savanna
(46)
Korea (47)
Temperate mixed forests
Madhya Pradesh, India (48)
Agricultural areas
BOREAS (49)
Needle leaf forests
USDA Agrocultural Research Center,
Agricultural area, Boreal forests
MD, USA; BOREAS Southern Study
Area, Canada (50)
EOS Core Sites (32)
All biome types
VALERIE & AERONET Sites (51)
All biome types
37
A
B
0.0 0.1 0.3 0.5 0.8 1.2 1.6 2.1 2.8 3.4 4.2 4.6 5.0 5.4 7.0
Annual Average Leaf Area Index
Forests
Other
Savannas
Non-vegetated
C
1 - 2 3 - 4 5 6
7
8 9 - 10 11 - 12
Number of Dry Months in a Year
Fig. 4. (A) Annual average green leaf area index of the Amazon derived from the Terra
MODIS Collection 4 LAI data. (B) Simplified landcover map of the Amazon derived from
the Terra MODIS Collection 4 landcover data. (C) Color coded map of the number of dry
months in a year. Months with precipitation less than 100 mm or less than one-third the
precipitation range [0.33*(maximum-minimum) + minimum] are considered dry.
38
0.7
0.4
B
0.6
Surface Reflectance
Aerosol Optical Depth
A
0.5
0.4
0.3
0.2
0.1
0.0
2000
2001
2002
2003
Year
2004
2005
0.3
Near- infrared
Residual aerosol
0.2
Aerosol free
Red
0.1
0.0
0
2
4
6
8
10
12
Month in 2004
Fig. 5. Aerosol effects on remote sensing of leaf area with MODIS data. (A) Time series of aerosol
optical depth at 550 nanometers in the Amazon retrieved from observations of the Terra MODIS
instrument. (B) Atmosphere corrected surface reflectances, separated by their aerosol quality flags,
at red and near-infrared channels from the Terra MODIS instrument in tile h12v09 – this is a 1200 x
1200 km region in the Amazon (0-10oS, 49.9917-60.9256oW). The residual aerosol and aerosol free
quality flags indicate whether the reflectances are contaminated with residual aerosol effects or free
of such effects after the atmospheric correction procedure has been applied.
39
Fig. 6. Precipitation (Prec) and evapotranspiration (ET) from various flux tower
experimental sites in the Amazon Forest (AF). (A) Dry season average values of
precipitation and ET. (B) Daily rates of ET during the wet and dry season at various
sites. After Negrón Juárez et al. (43).
A
First Difference of Leaf Area Index
40
1.0
0.8
0.6
0.4
N = 62
R = 0.56
t = 5.28
P < 0.0001
0.2
0.0
-0.2
-0.4
-0.6
-0.8
-1.0
-150
-100
-50
0
50
100
150
First Difference of Solar Radiation
1.0
First Difference of Leaf Area Index
B
N = 62
R = -0.63
t = 6.23
P < 0.0001
0.8
0.6
0.4
0.2
0.0
-0.2
-0.4
-0.6
-0.8
-1.0
-150
-100
-50
0
50
100
150
First Difference of Precipitation
Fig. 7. Relationship between changes in green Leaf Area Index (LAI), solar radiation
and precipitation. (A) Correlation between first differences of LAI and first differences
of solar radiation. (B) Correlation between first differences of LAI and first differences
of precipitation. To associate quantitatively the degree of association between changes
in leaf area, solar radiation and precipitation, we correlated successive monthly
differences of these variables, using the spatially averaged data shown in Fig. 1A. For
example, successive time differences of LAI, LAI(t), were calculated as LAI(t+1)LAI(t), where t is months in the timeline March 2000 to May 2005. The number of data
points (N) in both cases is 62. R is the correlation coefficient. The Student’s t values for
the correlation coefficient are significant at probability p shown in the panels.