biological processes dominate seasonality of remotely
TRANSCRIPT
Biological processes dominate seasonality of remotely sensed canopy greenness in an Amazon evergreen forest
J. Wu
To be published in "New Phytologist"
November 2017
Environmental & Climate Sciences
Brookhaven National Laboratory
U.S. Department of EnergyUSDOE Office of Science (SC),
Biological and Environmental Research (BER) (SC-23)
Notice: This manuscript has been authored by employees of Brookhaven Science Associates, LLC under Contract No. DE-SC0012704 with the U.S. Department of Energy. The publisher by accepting the manuscript for publication acknowledges that the United States Government retains a non-exclusive, paid-up, irrevocable, world-wide license to publish or reproduce the published form of this manuscript, or allow others to do so, for United States Government purposes.
BNL-114821-2017-JA
DISCLAIMER
This report was prepared as an account of work sponsored by an agency of the United States Government. Neither the United States Government nor any agency thereof, nor any of their employees, nor any of their contractors, subcontractors, or their employees, makes any warranty, express or implied, or assumes any legal liability or responsibility for the accuracy, completeness, or any third party’s use or the results of such use of any information, apparatus, product, or process disclosed, or represents that its use would not infringe privately owned rights. Reference herein to any specific commercial product, process, or service by trade name, trademark, manufacturer, or otherwise, does not necessarily constitute or imply its endorsement, recommendation, or favoring by the United States Government or any agency thereof or its contractors or subcontractors. The views and opinions of authors expressed herein do not necessarily state or reflect those of the United States Government or any agency thereof.
Submitted To: New Phytologist 1
2
Title: Biological processes dominate seasonality of remotely sensed canopy greenness in 3
an Amazon evergreen forest 4
5
Brief Heading: Biotic controls of canopy reflectance seasonality in Amazon 6
7
Lab-associated twitter name: @testgroup_bnl8
9
Author List: Jin Wu1,2*
, Hideki Kobayashi3, Scott C. Stark
4, Ran Meng
2, Kaiyu Guan
5, 10
Ngoc Nguyen Tran6, Sicong Gao
6, Wei Yang
7, Natalia Restrepo-Coupe
1, Tomoaki 11
Miura8, Raimundo Cosme Oliviera
9, Alistair Rogers
2, Dennis G. Dye
10, Bruce W. 12
Nelson11
, Shawn P. Serbin2, Alfredo R. Huete
6, and Scott R. Saleska
113
14
Author Affiliation: 15
[1] Department of Ecology and Evolutionary Biology, University of Arizona, Tucson, AZ16
85721 17
[2] Environmental & Climate Sciences Department, Brookhaven National Laboratory,18
Upton, New York, NY 11973 19
[3] Department of Environmental Geochemical Cycle Research, Japan Agency for20
Marine-Earth Science and Technology, Yokohama, Kanagawa Prefecture 236-0001, 21
Japan 22
[4] Department of Forestry, Michigan State University, East Lansing, MI 4882423
[5] Department of Natural Resources and Environmental Sciences, and National Center24
for Supercomputing Applications, University of Illinois at Urbana Champaign, Urbana, 25
IL 61801 26
[6] Climate Change Cluster, University of Technology Sydney, Ultimo, NSW 2007,27
Australia 28
[7] Center for Environmental Remote Sensing, Chiba University, Chiba-shi, Chiba, 263-29
8522, Japan 30
[8] Department of Natural Resources and Environmental Management, University of31
Page 1 of 43 New Phytologist
Havaii, Honolulu, HI 96822 32
[9] Embrapa Amazônia Oriental, Santarém PA 68035-110, Brazil 33
[10] School of Earth Sciences and Environmental Sustainability, Northern Arizona 34
University, Flagstaff, AZ 86011 35
[11] Environmental Dynamics Department, Brazil's National Institute for Amazon 36
Research (INPA), Manaus, AM 69067-375, Brazil 37
38
* Corresponding author: 631-344 3361; [email protected] 39
40
Key Words: Leaf optics, canopy phenology, leaf age, LiDAR canopy structure, 41
PROSAIL, FLiES, WorldView-2, MODIS EVI 42
43
Type of Paper: Original Research 44
45
Words Record: 46
Summary: 199 words 47
Maintext Total: 6538 words 48
Introduction: 1047 49
Methods: 2413 50
Results: 1099 51
Discussion: 1739 52
Conclusion: 120 53
Acknowledgement: 120 54
55
Figures/Tables Record: 56
Figures: 1-6 57
Tables: 1 58
59
Supporting Information: 60
Figures S1-S16 61
Tables S1-S5 62
Methods S1-S6 63
Page 2 of 43New Phytologist
Summary 64
• Satellite observations of Amazon forests show seasonal and inter-annual 65
variations, but the underlying biological processes remain debated. 66
• Here we combined radiative transfer models (RTMs) with field observations of 67
Amazon forest leaf and canopy characteristics to test three hypotheses for 68
satellite-observed canopy reflectance seasonality: seasonal changes in leaf area 69
index, in canopy-surface leafless crown fraction, and/or in leaf demography. 70
• Canopy RTMs (PROSAIL and FLiES), driven by these three factors combined, 71
well-simulated satellite-observed seasonal patterns, explaining ~70% of 72
variability in a key reflectance-based vegetation index (MAIAC EVI which 73
removes artifacts that would otherwise arise from clouds/aerosols and sun-sensor 74
geometry). Leaf area index, leafless crown fraction and leaf demography 75
independently accounted for 1%, 33% and 66% of FLiES-simulated EVI 76
seasonality respectively. These factors also strongly influenced modeled near-77
infrared (NIR) reflectance, explaining why both modeled and observed EVI, 78
which is especially sensitive to NIR, captures canopy seasonal dynamics well. 79
• Our improved analysis of canopy-scale biophysics rules out satellite artifacts as 80
significant causes of satellite-observed seasonal patterns at this site, implying that 81
aggregated phenology explains the larger scale remotely-observed patterns. This 82
work significantly reconciles current controversies about satellite-detected 83
Amazon phenology, and improves our use of satellite observations to study 84
climate-phenology relationships in the tropics. 85
86
Page 3 of 43 New Phytologist
Introduction 87
A fundamental unanswered question for global change ecology is the degree to 88
which tropical forests are vulnerable to climate change. Increasingly, satellite remote 89
sensing is being used to tackle this question by investigating how forests respond to 90
climatic variations at multiple spatial and temporal scales (Saleska et al., 2007; Xu et al., 91
2011; Lee et al., 2013; Saatchi et al., 2013; Hilker et al., 2014; Zhou et al., 2014; Guan et 92
al., 2015). Many remote sensing products – such as the Moderate Resolution Imaging 93
Spectroradiometer (MODIS) Vegetation Indices (VIs), which are spectral transformations 94
of two or more reflectance bands – provide estimates of canopy greenness. These 95
products are composite indices of both leaf biochemistry (leaf cellular structure, 96
chlorophyll content and biochemical composition) and canopy structure (leaf area, crown 97
geometry, leaf demography) (Huete et al., 2002; Doughty & Goulden, 2008; Brando et 98
al., 2010; Lopes et al., 2016; Wu et al., 2016, 2017). If accurate, they can reveal 99
important mechanisms regulating the response of tropical forests to seasonal and inter-100
annual climatic variability, the same mechanisms which we rely on to validate and 101
improve earth system model simulations. 102
However, a number of key issues remain in the understanding and biophysical 103
interpretation of satellite remote sensing. Recent studies of satellite observations of 104
Amazon phenology show a significant seasonality in tropical evergreen forests (e.g. 105
Jones et al., 2014; Bi et al., 2015; Maeda et al., 2016; Saleska et al., 2016). However, 106
seasonal variation in leaf optical characteristics (Roberts et al., 1998; Toomey et al., 107
2009; Chavana-Bryant et al., 2017; Wu et al., 2017), canopy leaf area index (Brando et 108
al., 2010; Samanta et al., 2012a), canopy leaf demography (Doughty & Goulden, 2008; 109
Brando et al., 2010; Lopes et al., 2016), and canopy structure (Anderson et al., 2010; 110
Tang & Dubayah, 2017) may complicate the biophysical interpretation of true canopy 111
reflectance seasonality. Furthermore, some studies suggest that some standardized 112
MODIS products (e.g. Enhanced Vegetation Index, or EVI), which are processed to 113
provide a coherent data product across space and time, may still be sensitive to seasonally 114
varying artifacts – particularly atmospheric cloud/aerosol contamination (Samanta et al., 115
2010, 2012b), or sun-sensor geometry (Galvão et al., 2011; Morton et al., 2014), – that 116
may confound a sensor’s ability to accurately capture the real phenology and functional 117
Page 4 of 43New Phytologist
response of tropical forests to climate variability. To help untangle these confounding 118
factors we need a complimentary bottom-up approach that considers physically-based 119
canopy radiative transfer models (RTMs) and upscales observable leaf- and canopy-level 120
properties. 121
Canopy RTMs, such as PROSAIL (Baret et al., 1992; Jacquemound et al., 2009), 122
GEOSAIL (Huemmrich, 2001; Ustin et al., 2010), and Forest Light Environmental 123
Simulator (FLiES; Kobayashi & Iwabuchi, 2008; Kobayashi et al., 2012), which 124
encompass the interactive effects of surface optical elements (leaf, bark, litter and soil) 125
and canopy structural properties within a forest community, sun-sensor geometry and 126
atmospheric radiation condition, can potentially be used to differentiate the roles of 127
biophysical processes from potential artifacts in canopy-scale reflectance and greenness 128
seasonality. These modeling approaches, parameterized by field-observed leaf and 129
canopy properties, have been assessed previously across diverse ecosystems (e.g. boreal, 130
temperate and agricultural), showing good agreements between models and observations 131
(e.g. Verhoef & Bach, 2003; Koetz et al., 2007; Kobayashi et al., 2012; Schneider et al., 132
2014). However, these modeling approaches have rarely been applied and tested in 133
tropical evergreen forests over seasonal timescales. Prior studies have investigated the 134
sensitivity of remote sensing to certain biological effects (e.g. Toomey et al., 2009; 135
Morton et al., 2014, 2016), or focused on retrieval of important biophysical variables 136
(e.g. chlorophyll concentration, Hilker et al., 2017), but few studies have integrated field 137
observations with RTMs to comprehensively assess the contributions of competing 138
biological processes to the aggregated canopy-scale reflectance and greenness 139
seasonality. 140
Here we use canopy RTMs to connect field-observed leaf and canopy 141
characteristics with satellite remote sensing to mechanistically interpret canopy-scale 142
reflectance and greenness seasonality in Amazonian evergreen forests. Specifically, we 143
examine three phenological factors that could drive seasonality in satellite-observed 144
tropical forest canopy reflectance, including (i) the leaf area index (LAI) effect, i.e. the 145
change in reflectance due to the seasonality of LAI and light-scattering by multi-leaf-146
layer (Verhoef, 1984; Samanta et al., 2012a), (ii) the leafless crown fraction effect, i.e. 147
the change in reflectance due to seasonal change in whole-canopy optical properties, 148
Page 5 of 43 New Phytologist
specifically the fraction of leafless crowns, which is strongly anti-correlated with LAI 149
seasonality (e.g. Lopes et al., 2016), and (iii) the leaf demography effect, i.e. the change 150
in reflectance due to the seasonality of leaf age distributions, as leaf reflectance and 151
transmittance show strong dependence on leaf age (e.g. Roberts et al., 1998; Wu et al., 152
2017). 153
We used two canopy RTMs, PROSAIL and FLiES. Their main difference lies in 154
FLiES being a 3-D canopy RTM, allowing for more sophisticated and realistic 155
representation of canopy structure, relative to PROSAIL, a 1-D canopy RTM. 156
Comparison of the two models permits an assessment of the more realistic 3-D canopy 157
structure on canopy-scale reflectance and greenness seasonality. 158
To assess these phenology-related effects and to compare PROSAIL and FLiES, 159
we used field-observed leaf and canopy characteristics collected over the annual cycle in 160
a central-eastern Amazonian evergreen forest. These characteristics include monthly LAI 161
and litterfall of a 1-ha control plot (Brando et al., 2010); monthly leafless crown fraction 162
of the upper canopy derived from a tower-camera (Wu et al., 2016); field-observed leaf 163
reflectance spectra at different leaf ages (Wu et al., 2017); and an airborne LiDAR survey 164
of 3-D canopy structure (Stark et al., 2012, 2015; Hunter et al., 2015). PROSAIL and 165
FLIES were parameterized and driven by the above three phenological factors separately 166
and in combination, in order to assess the relative contribution of each factor responsible 167
for canopy-scale reflectance and greenness seasonality. Specifically, we asked two 168
questions: (1) When driven by all three phenological factors (i.e. LAI, leafless crown 169
fraction and leaf demography), can PROSAIL and FLiES capture tropical forest canopy 170
reflectance seasonality? (2) What is the relative contribution of each phenological factor 171
toward explaining canopy-scale reflectance seasonality? By answering these questions, 172
we provide a benchmark for scaling leaf- and canopy- characteristics to landscapes and 173
for broad application across multiple sites, and ultimately increase our understanding of 174
fundamental biophysical processes that regulate tropical canopy reflectance seasonality, 175
enabling more accurate usage of existing and future remote sensing platforms in the 176
tropics. 177
178
Materials and Methods 179
Page 6 of 43New Phytologist
Satellite observations 180
We used satellite observations targeted at the k67 tower site where ground 181
observations were made. The k67 site (54°58’W, 2°51’S) is located in the Tapajós 182
National Forest, near Santarém, Pará, Brazil. It is an evergreen tropical forest on a well-183
drained clay-soil plateau (Rice et al., 2004), with a mean upper canopy height of ~40 m 184
(Hutyra et al., 2007). Mean annual precipitation is ~2000 mm year-1
with a 5-month dry 185
season when evapotranspiration exceeds precipitation from approximately mid-July to 186
mid-December (Restrepo-Coupe et al., 2013). 187
Two kinds of satellite observations were used to evaluate RTMs performance, 188
including a WorldView-2 (WV-2) image and two versions of the collection 6 time-series 189
MODIS land surface reflectance and VIs products. 190
The WV-2 image was acquired on July 28th
, 2011 (Fig. S1). The image has 2m 191
spatial resolution, 2° off nadir view, 42° solar zenith angle, and 2.6% cloud cover. It 192
includes eight spectral bands (see Table S1 and Fig. S2 for spectral response functions). 193
For details on data availability, image pre-processing, atmospheric correction, and 194
reflectance calculation refer to Methods S1 and Meng et al (2017). Based on our 195
knowledge of the study site and vegetation spectroscopy, we used a true-color RGB 196
composite to visually identify three phenophases for fully illuminated upper-canopy 197
crowns within the 1x1 km2 footprint surrounding the k67 site. These were young (bright 198
green in color), old (dark green) and leafless (largely occupied by bare branches) (Fig. 199
S1d). Thirty crowns of each phenophase were selected, totaling at least 300 pixels per 200
phenophase, to derive landscape average reflectance. We subsequently used the derived 201
phenophase-specific canopy reflectance to validate the RTMs (below). 202
We primarily used the collection 6 MODIS Multi-Angle Implementation of 203
Atmospheric Correction product (MAIAC; Lyapustin et al., 2012) for years 2000 to 2014 204
to investigate remotely sensed vegetation seasonality. For details on data acquisition, 205
processing and quality control refer to Methods S2. MAIAC incorporates a new 206
Bidirectional Reflectance Distribution (BRDF; corrected to nadir view and 45° solar 207
zenith angle) and strict atmospheric corrections for clouds and aerosols (Lyapustin et al., 208
2012). It includes four MODIS bands (Blue, Green, Red and near-infrared, NIR; see 209
response functions in Fig. S2), and associated VIs (i.e. Normalized Difference Vegetation 210
Page 7 of 43 New Phytologist
Index, NDVI and EVI; see Table S2 for equations). We here focus on the MAIAC data 211
because the data have become commonly used for phenology monitoring in the tropics, 212
with several empirical validations from site-level phenology data (Lopes et al., 2016; 213
Wagner et al., 2016; Wu et al., 2016), eddy covariance data (Jones et al., 2014; Guan et 214
al., 2015; Saleska et al., 2016), and satellite data of other sources (Maeda et al., 2016) 215
and alternative approaches (Bi et al., 2015; Saleska et al., 2016). We expect that the 216
multi-year (2000-2014) average annual cycle of monthly BRDF-corrected MAIAC data 217
with robust removal of cloud/aerosols can be used as a benchmark to assess the integrated 218
phenology effects (i.e. the three phenological factors above) on modeled canopy 219
reflectance seasonality. We also tested the robustness of the MAIAC results through 220
retrievals of different target areas (3x3 km2 and 5x5 km
2) and through comparisons with 221
an alternative MODIS product, the BRDF and Albedo product (MCD43A1; Wang et al., 222
2015). We analyzed the MCD43A1 product (5x5 km2) for the same sun-sensor geometry 223
as MAIAC and two levels of quality control flags (QA3 and QA5; Methods S2). Our 224
results (Figs. S3 and S4) indicate that the relative seasonality of MAIAC data of 5x5 km2 225
is very robust, especially in the most critical products, NIR and EVI. We observed some 226
differences in the visible bands between the MAIAC and MCD43A1 products (Fig. S4), 227
presumably due to corresponding differences in associated atmospheric corrections. 228
These differences, however, did not affect the overall seasonality in EVI much, which 229
was dominated by NIR and is the main focus of the study. 230
231
Ground observations 232
To parameterize canopy RTMs and to assess the relative contribution of the three 233
phenological factors to canopy-scale reflectance seasonality, we leveraged a number of 234
field observations at the k67 site, including measurements of tissue optics (i.e. leaves, 235
barks and litter) and canopy characteristics (i.e. phenology and structure). 236
(1) Tissue optics. Reflectance spectra of leaves, barks and litter were measured using 237
a portable spectrometer (ASD FieldSpec Pro; Analytical Spectra Devices, ASD Inc., 238
Boulder, CO; spectral range: 350-2500 nm; spectral resolution: 3nm@350-1000nm and 239
8nm@1000-2500nm). Spectral data were interpolated to 1nm before analysis using the 240
default ASD output. For each tree, we obtained spectra of leaves at young (≤2 months), 241
Page 8 of 43New Phytologist
mature (3-5 months) and old (6-14 months) age classes (Wu et al., 2017). Seven tree 242
species with all three leaf age classes were used (see Tables S3 and S4 for species 243
identification and number of leaf replicate): four species were from the upper canopy 244
(accounting for 22.2% local basal area); the other three species were from 20-30m mid-245
canopy stratum. Leaves were sampled to represent three incident canopy light conditions: 246
upper canopy sunlit, upper canopy shaded, and mid-canopy shaded. For details on this 247
dataset (Fig. 1a) and the protocols used for spectral measurements and leaf age 248
classification see Wu et al (2017). Bark reflectance spectra were measured in 2002 by Dr. 249
T. Miura. Bark samples were harvested from 13 canopy trees at ~1.3 meters above the 250
ground (Table S5 for species identification). These samples were kept in sealed plastic 251
bags in a cooler box, and reflectance measurements (Fig. 1b) were made within 24 hours 252
after sampling. The litter spectra (Fig. 1b) were measured in March 2014, by randomly 253
collecting 40 litter foliage over various locations across the forest floor. Reflectance 254
measurements were made within 1 hour after sampling. 255
Together with reflectance, leaf transmittance regulates the leaf single scattering 256
albedo and is an important component of canopy RTMs and process models (Sellers et 257
al., 1997; Pinty et al., 2004). However, acquiring accurate and reproducible 258
measurements of leaf transmittance is challenging, primarily due to limitations inherent 259
of integrating sphere instrumentation (Shiklomanov et al., 2016). Instead, we estimated 260
leaf transmittance (Fig. 1c) and subsequent absorptance (Fig. 1d) by inverting the leaf 261
reflectance model PROSPECT (Jacquemoud & Baret, 1990; Feret et al., 2008) after it 262
was optimized to closely match the field-observed leaf reflectance. Shiklomanov et al 263
(2016) showed that PROSPECT-inverted leaf transmittance is highly consistent with 264
values obtained with an integrating sphere for fresh leaves. As a check, we compared our 265
estimated transmittance against a set of independent measurements made in 2002 by Dr. 266
T. Miura (see Methods S3 for more details). Similar to Shiklomanov et al (2016), we 267
found strong agreements between measured and modeled transmittance, in terms of both 268
mean values and standard deviation (Fig. S5a). Furthermore, similar trends across leaves 269
of multiple species were found for field-observed leaf NIR reflectance and transmittance 270
and for field-observed NIR reflectance and PROSPECT-inverted transmittance (Fig. 271
S5b), except that modeled transmittance was biased low in Manikara huberi (M.huberi) 272
Page 9 of 43 New Phytologist
leaves with high NIR reflectance (>0.6). This might be because M.huberi has thick, waxy 273
leaves which are not currently well represented/constrained in PROSPECT, but a more 274
in-depth understanding is still needed. The age-dependent leaf reflectance, transmittance 275
and absorptance for each of all 11 tree canopy light conditions (i.e. four canopy sunlit, 276
four canopy shaded and three mid-canopy shaded) are shown in Fig. S6. 277
(2) Canopy characteristics (phenology and structure). Three components of canopy 278
phenology were available at the k67 site. These were, firstly, the mean annual cycle of 279
monthly field-derived LAI (or LAI_field) obtained with the LAI-2000 instrument at 280
ground level (01/2000-12/2005). See Brando et al (2010), their Fig. 4 and our Fig. 2a for 281
more details. Secondly, we used the mean annual cycle of monthly leafless crown 282
fraction (1 minus the green crown fraction; Fig. 2a) from tower-mounted camera image 283
timeseries (01/2010-12/2011). The timeseries of the green crown fraction was derived 284
using a camera-based tree inventory approach, and see Wu et al (2016) and their Fig. S8 285
for more details. Thirdly, we used leaf age fractions in three age classes (Fig. 2b): young 286
(≤2 months), mature (3-5 months) and old (≥6 months). These were derived from a leaf 287
age demography model described in Wu et al (2016). 288
Canopy structure was derived from an August 2012 airborne LiDAR survey at 289
k67 (Hunter et al., 2015). For details about the LiDAR sensor and airborne survey refer 290
to Stark et al (2012, 2015). We here estimated the canopy area density (CAD) in 3-D 291
canopy voxels with a 2x2x2 m3 grain, following approaches developed in Stark et al 292
(2012, 2015). For details on how we derived 3-D voxel data from the LiDAR survey refer 293
to Methods S4. Additionally, a constant was adjusted to set the vertically-integrated 294
landscape-average LiDAR canopy area to match empirical estimates from LAI_field 295
(Fig. 2a) and to extend one-time LiDAR-derived 3-D canopy structure into the seasonal 296
scale (see Methods S5). The LiDAR-derived 3-D canopy structure is shown in Fig. S7, 297
and the landscape average height-CAD relationship is shown in Fig. 2c. 298
299
Canopy Radiative Transfer Models 300
The two canopy RTMs used are PROSAIL and FLiES. The former has the 301
advantage of being simple and less computationally demanding, while FLiES uses more 302
realistic 3-D canopy structure, in this case derived from airborne LiDAR. Here, we 303
Page 10 of 43New Phytologist
parameterized the two models (see Methods S5), aiming to (1) explore whether the 304
relative contributions of the three phenological factors to canopy reflectance seasonality 305
are consistent between the two, and (2) assess the impact of using a more realistic 3-D 306
canopy structure on modeled canopy reflectance by cross-model comparisons. 307
308
PROSAIL 309
PROSAIL (Jacquemoud et al., 2009) is a combination of the PROSPECT leaf 310
optical properties model (Jacquemoud & Baret, 1990) and the SAIL canopy bidirectional 311
reflectance model (Verhoef, 1984, 1985). Details of PROSPECT and SAIL are shown in 312
Methods S6. Coupling of PROSPECT and SAIL as implemented in PROSAIL allows 313
simulation of the joint effect of leaf biochemistry (morphological and chemical 314
parameters), canopy characteristics (LAI and crown geometry), sun-sensor geometry (sun 315
angle and sensor angle) and clear/diffuse sky on canopy-scale reflectance. We used the 316
MATLAB version, PROSAIL_5B_Matlab, available at 317
http://teledetection.ipgp.jussieu.fr/prosail/. 318
319
FLiES 320
FLiES consists of a 1-D atmospheric RTM and a 3-D canopy RTM, based on the 321
Monte Carlo ray tracing method (Kobayashi et al., 2012). The original FLiES model (e.g. 322
Kobayashi & Iwabuchi, 2008) employed geometric objects such as cone, cylinder and 323
spheroid to delineate individual tree structure. Here we extended FLiES for LiDAR-based 324
voxel representation of 3-D canopy structure, which enables a more realistic depiction of 325
forest canopies. Each voxel contains leaf and woody elements, parameterized by using 326
the LiDAR-derived 3-D canopy area, with 89% assigned to leaf elements and 11% 327
assigned to woody elements, following a recent field survey in a Costa Rican tropical 328
evergreen forest (Olivas et al., 2013). Voxel grains of 2x2x2 m3, representing a 3-D 329
forest landscape of 600x600 m2 surrounding the k67 site (Fig. S7), were used for FLiES 330
simulations. Additionally, shoot-scale clumping, a metric quantifying the foliage 331
clumping within a shoot (Chen et al., 1997), is an important parameter in FLiES 332
(Kobayashi et al., 2012). It increases with increasing clumping, and was set equal to 1 (or 333
no shoot-scale clumping) for the default model simulations. The FLiES code in fortran 334
Page 11 of 43 New Phytologist
and the voxelized data for the k67 site are available from the authors upon request. 335
336
Model experiments 337
We used PROSAIL and FLiES as our main tools (see Method S5 for more 338
details), which were set to the same sun-sensor geometry as MAIAC, and performed a 339
suite of model experiments to assess the relative contributions of the three phenological 340
factors (P) on canopy-scale reflectance seasonality. Details are shown as below: 341
(P1): Assess the LAI effect due to light-scattering by multi-leaf-layers. The 342
models were run under different monthly LAI_field (Fig. 2a) for PROSAIL and the 343
interpolated LiDAR 3-D canopy structure for FLiES (Method S5), with fixed tissue 344
optics (i.e. annual mean leaf demography-weighted leaf reflectance and transmittance, 345
and multi-species average bark/litter reflectances). 346
(P2): Assess the leafless crown fraction effect due to the distinct canopy 347
reflectance characteristics of leafless and green canopy phenophases. This is described as: 348
Rcanopy,t = (1− fleafless,t )×Rgreen + fleafless,t ×Rleafless (1) 349
Where Rcanopy,t is canopy-scale reflectance at given month t, Rgreen
and Rleafless are 350
modeled canopy-scale reflectance for green and leafless phenophases respectively, 351
fleafless,t is the tower camera-derived leafless crown fraction at month t (Fig. 2a). Rgreen 352
was modeled under fixed LAI_field (i.e. annual mean LAI_field) for PROSAIL and the 353
corresponding LiDAR canopy structure for FLiES, and fixed tissue optics (annual mean 354
leaf demography-weighted leaf reflectance and transmittance, and multi-species average 355
bark/litter reflectances). Rleaflesswas modeled under the same fixed LAI and multi-species 356
average bark/litter reflectances, while leaf optics were set equal to bark optics (field-357
observed bark reflectance and bark transmittance=0). 358
(P3): Assess the leaf age effect through seasonally varying leaf demographics by 359
scaling age-dependence of leaf optics to the canopy. We described this as: 360
Rcanopy,t = fY ,t ×RY + fM ,t ×RM + fO,t × RO (2) 361
Where RY, RM
and RO are modeled canopy-scale reflectance of young, mature and old 362
phenophases respectively, fY ,t , fM ,t and fO,t are their relative abundances at month t (Fig. 363
Page 12 of 43New Phytologist
2b). RY, RM
and ROwere again modeled under fixed annual average LAI_field for 364
PROSAIL and the corresponding LiDAR canopy structure for FLiES, multi-species 365
average bark/litter reflectances and age-dependence of leaf reflectance and transmittance. 366
(P1+P2+P3): Assess the joint effects of the three phenological factors (i.e. P1 to 367
P3) on canopy reflectance seasonality. We described this as: 368
Rcanopy,t = (1− fleafless,t )×Rgreen,t + fleafless,t ×Rleafless,t (3) 369
Rgreen,t = fY ,t ×RY ,t + fM ,t × RM ,t + fO,t ×RO,t (4) 370
Where Rgreen,t and Rleafless,t
are modeled canopy reflectance at month t for green and 371
leafless phenophases, respectively. RY ,t , RM ,t and RO,t were modeled under seasonally 372
varying LAI_field for PROSAIL and the interpolated LiDAR 3-D canopy structure for 373
FLiES, age-specific leaf reflectance and transmittance, and multi-species average 374
bark/litter reflectances. 375
We performed the above model experiments parametrized by the leaf optics from 376
each of the 11 tree canopy light conditions (Fig. S6), and the mean and standard deviation 377
of the total 11 modeling runs were calculated for the final analysis (see Method S5). To 378
assess the relative importance of three phenological factors (i.e. P1 to P3) on the 379
‘comprehensive’ model (i.e. P1+P2+P3) results, we used a relative importance of 380
regressors in R (package relaimpo, using the method called ‘betasq’), following the 381
approach of Grömping (2015). 382
It is worth noting that the interaction term, or the scattering between green and 383
leafless crowns, or among crowns dominated by different canopy phenophases (i.e. 384
young, mature, old and leafless), was not considered in this study. At nadir view angle 385
(i.e. WV-2 and MAIAC data), this interaction term might likely be a second order effect 386
in controlling canopy-scale reflectance seasonality, compared with the three phenological 387
factors already explored here; future analysis is still needed to fully understand such 388
interaction term, which is beyond the scope of the current paper. 389
390
Results 391
Age-dependence of leaf optics 392
Despite marked variation in leaf optical properties across all 11 tree-canopy light 393
Page 13 of 43 New Phytologist
conditions, within each tree-canopy light condition, field-measured leaf reflectance and 394
PROSPECT-inverted leaf transmittance and absorptance show strong dependence on leaf 395
ages (Figs. 1, S6 and S8). The mean visible (400-700 nm) reflectance shows continuous 396
decline with leaf ages, from 0.08 in young leaves to 0.05 and 0.04 in mature and old 397
leaves respectively (Fig. 1a). The mean NIR (700-1100 nm) and SWIR (1100-2500 nm) 398
reflectance initially increase with leaf ages from 0.44 (NIR) and 0.19 (SWIR) in young 399
leaves, and then reach a peak at 0.46 and 0.22 after full leaf expansion (Fig. 1a). By 400
contrast, the mean values of visible, NIR and SWIR transmittance all decline with leaf 401
ages (Fig. 1c), from 0.10 (visible), 0.47 (NIR) and 0.26 (SWIR) in young leaves, to 0.03, 402
0.39 and 0.22 in mature leaves, and 0.01, 0.36 and 0.21 in old leaves. Our derived leaf 403
absorptance (=1-reflectance-transmittance) also shows strong age-dependence, and the 404
mean values of leaf absorptance continuously increase with leaf ages throughout the full 405
spectral range (Fig. 1d). 406
407
Phenophase effects on canopy reflectance 408
We used PROSAIL and FLiES, together with observations from the WV-2 image, 409
to explore the phenophase effects (i.e. young, mature, old and leafless) on canopy 410
reflectance. Our results show that canopy reflectance varies in concert with canopy 411
phenophases with a fixed LAI (i.e. LAI_field=5 m2 m
-2; Figs. 3) and strongly depends on 412
the combined variation in LAI and phenophases (Figs. S9 and S10). Specifically, canopy 413
visible reflectance, though typically low magnitude (<0.1), shows strong dependence on 414
canopy phenophases, with the young phenophase displaying much higher value than that 415
of either mature or old phenophase. The leafless phenophase has a similar mean visible 416
reflectance as the young phenophase. Canopy NIR reflectance shows a much broader 417
magnitude of variation (>0.2) across all phenophases, and displays continuous decline 418
from young to old, with the minimum value at the leafless stage. Such phenophase effects 419
on canopy reflectance are consistent across the two models (Fig. 3a,b for PROSAIL and 420
Fig. 3c for FLiES), though PROSAIL results show consistently higher intra-phenophase 421
variability than FLiES. Additionally, we performed a sensitivity analysis (with and 422
without the thick-leaved species M.huberi which has bias in modeled leaf transmittance), 423
and our results show that the observed phenophase effects on canopy reflectance are also 424
Page 14 of 43New Phytologist
consistent (Fig. S11). Finally, both PROSAIL and FLiES show similar phenophase 425
effects on canopy reflectance compared with the WV-2 observations (Fig. 3d), suggesting 426
that the two RTMs can reproduce similar phenophase effects as observations. 427
428
Phenology effects on canopy reflectance seasonality 429
We used PROSAIL and FLiES to explore how including model representation of 430
three phenological factors influences modeled canopy-scale reflectance seasonality, as 431
evaluated against the MAIAC data. The seasonal pattern of MAIAC NIR and EVI shows 432
an initial decline in the late wet season and then an increase in the dry season (Fig. 4). 433
Our results show that the models driven by all three phenological factors (P1+P2+P3; or 434
the ‘comprehensive’ model; Fig. 4) are best able to capture the MAIAC data, respectively 435
explaining 87% and 75% of seasonal variation in MAIAC NIR and EVI using PROSAIL, 436
and 64% and 69% using FLiES, though PROSAIL has larger model bias than FLiES. 437
We further ran the models separately parameterized by each of the three 438
phenological factors (P1 to P3), aiming to quantify their relative contributions. The 439
results are shown in Figs. 5 (all 11 tree-canopy light conditions) and S12 (all but 440
excluding M.huberi), and Table 1. Our results show that both PROSAIL and FLiES 441
attributed very comparable phenological contribution in NIR and Green reflectances, but 442
with some variations in Blue and Red reflectances, and VIs. Additionally, both models 443
show that canopy-surface leafless crown fraction (P2) dominated the magnitude of the 444
‘comprehensive’ modeled NIR and EVI seasonality; canopy-surface leafless crown 445
fraction (P2) and leaf demographics (P3) jointly determined the relative seasonality of the 446
‘comprehensive’ modeled NIR and EVI; canopy LAI phenology (P1) barely contributed 447
to the ‘comprehensive’ modeled NIR and EVI seasonality. Furthermore, when the models 448
were run with and without M.huberi the simulated relative seasonalities were almost 449
identical to each other (Fig. S12), suggesting that our results are very robust, despite 450
some transmittance bias associated with M.huberi (e.g. Fig. S5b). 451
We also analyzed both modeled and observed canopy reflectance seasonality for 452
Blue, Green and Red bands and reflectance-derived NDVI. The results are summarized in 453
Figs. S13 and S14. Overall, the modeled canopy visible reflectances were small (<0.05) 454
and had similar magnitudes comparable to the observations, though their relative 455
Page 15 of 43 New Phytologist
seasonalities were different. We also compared the canopy RTMs results with the 456
MCD43A1 product (Fig. S15), and the results are practically the same as with MAIAC. 457
Comparing PROSAIL with FLiES, we found that the relative seasonalities of 458
observed EVI and NIR reflectance were more closely simulated by PROSAIL than 459
FLiES, although FLiES, with less of an offset, more closely matched with the absolute 460
value (Fig. 4). This was somewhat unexpected: if 3-D structure of the forest matters for 461
reflectance dynamics, we would expect in principle that a model (like FLiES) 462
representing forest structure in 3-D should be able to do better than a 1-D model (like 463
PROSAIL). This suggests that some aspects of the three dimensional dynamics (which 464
are less constrained than the 1-D version) are mis-parameterized. For example, FLiES 3-465
D structure is more capable of representing woody elements than is PROSAIL, but if the 466
woody fraction (which is not well constrained by observation but has less seasonality) is 467
over-represented, this would artifactually reduce modeled canopy NIR reflectance 468
seasonality. A fuller understanding of the role of 3-D versus 1-D structure in driving 469
temporal variation in canopy reflectance is needed, but for the purposes of the present 470
seasonality study, it is sufficient to note that both models appear to capture well the 471
dynamics of vegetation indices over seasons. 472
473
Model sensitivity to the clumping effect 474
We used the ‘comprehensive’ model (P1+P2+P3) to explore the extent to which 475
the clumping effect within a voxel in FLiES (approximated by shoot-scale clumping; 476
Kobayashi et al., 2010) affected modeled canopy-scale reflectance seasonality. We found 477
that although the absolute values varied greatly, the relative seasonalities of modeled 478
reflectance remained consistent across a wide range of shoot-scale clumping (from 1.0 to 479
1.39) (Fig. S16). Since the value of shoot-scale clumping shows strong biome 480
dependence (He et al., 2012), and is difficult to measure in the field, our model 481
sensitivity results suggest that the shoot-scale clumping effect can be an important source 482
of uncertainty that cause the absolute magnitude difference between models and 483
observations as shown in Fig. 4. 484
485
Discussion 486
Page 16 of 43New Phytologist
Our work consists of two main results: first, that when field-observed leaf and 487
canopy characteristics are used to drive canopy RTMs, they simulate canopy-scale 488
reflectance seasonality patterns that closely match satellite observations, robustly vetted 489
to remove known artifacts, of the same region where the ground observations were made 490
(Fig. 4). And second, that simulated reflectance seasonality is shown to arise in the model 491
directly from two main phenological factors: changes in canopy-surface leafless crown 492
fraction and changes in leaf demography, with only a very minor contribution from 493
changes in LAI (Figs. 5, S12 and S13; Table 1). We here discuss three broad implications 494
of this work: 495
496
Assessing differential phenological effects on canopy reflectance seasonality should help 497
to correctly interpret climate-phenology relationships in the Amazon 498
We start by discussing the implications of our second main finding that seasonal 499
variations in canopy-surface leafless crown fraction and leaf demography are the two 500
main phenological factors driving modeled reflectance seasonality (Table 1). This is 501
because canopy reflectance shows strong dependence on canopy phenophases (i.e. young, 502
mature, old, and leafless) (Fig. 3) and their seasonally changing fractions (Figs. 2a,b). 503
Though LAI_field and the canopy-surface leafless crown fraction were highly negatively 504
correlated at our site (Fig. 2a), our partitioning analysis attributes very little of the 505
reflectance seasonality to changes in LAI (see Table 1): canopy-scale NIR reflectance 506
increases with LAI through light scattering by multi-leaf-layer (Verhoef, 1984; Samanta 507
et al., 2012a), but LAI-induced NIR increase saturates when LAI exceeds 4 m2 m
-2 (Figs. 508
S9-S10). This implies that small seasonal changes detected in LAI (~1 m2 m
-2; Fig. 2a) in 509
a dense tropical forest canopy, where LAI exceeds 5 m2 m
-2 all year, will have little direct 510
impact on canopy reflectance seasonality (Table 1). This also implies that the previously 511
reported correlation between LAI and EVI at several sites in the Amazon (e.g. Brando et 512
al., 2010; Wu et al., 2016) is likely driven by the associated correlated decline in canopy-513
surface leafless crown fraction (Fig. 2a), which significantly regulates canopy-scale 514
reflectance seasonality (Fig. 5 and Table 1). 515
Interestingly, although leaf age strongly regulates leaf reflectance (Fig. 1a; see 516
also Chavana-Bryant et al., 2017; Wu et al., 2017), NIR reflectance and EVI had 517
Page 17 of 43 New Phytologist
opposite trends with age at the canopy vs. leaf scales, with young canopy phenophase 518
having highest NIR reflectance, whereas young leaves themselves having lowest NIR 519
reflectance (Fig. 1a vs Fig 3a; Fig. 6). Canopy-scale reflectance is a joint consequence of 520
overall canopy structure (LAI and crown geometry) in addition to leaf optical properties 521
(reflectance and transmittance) (Roberts et al., 1998; Ollinger, 2011). Young leaves have 522
much higher NIR transmittance (due to their low absorptance and reflectance; Fig. 1) 523
leading to more multiple-scattering of NIR light within a very dense canopy (e.g. LAI>3 524
m2 m
-2) dominated by young leaves. The effect of more multiple-scattering events in a 525
canopy dominates that of lower NIR reflectance in individual leaves, leading to higher 526
NIR reflectance at canopy-scale (Figs. 3, S9 and S10). This highlights the importance of 527
connecting leaf optics of single scattering albedo (reflectance+transmittance) to canopy 528
structure to correctly represent or interpret canopy-scale reflectance signatures. 529
By identifying the two phenological factors (leafless crown fraction and leaf 530
demography) that explain satellite-detected seasonality (Figs. 4, 5 and S15), our work has 531
direct implications for characterizing both spatial and temporal variation in satellite-532
detected tropical phenology (e.g. using MAIAC EVI), as these two factors represent 533
different ecophysiological strategies for tropical tree’s responses to seasonal/inter-annual 534
resource availability: (1) leafless crowns are a manifestation of deciduous or near-535
deciduous habit, related to the hydrological sensitivity of tropical trees. Higher water 536
stress in dry seasons can lead to higher abundance of deciduous trees over space (e.g. 537
Bohlman, 2010; Guan et al., 2015; Xu et al., 2016) or increased drought-induced leaf 538
shedding and mortality over time (e.g. Nepstad et al., 2007; Xu et al., 2016), resulting in 539
an increased canopy-surface leafless crown fraction (and thus ‘brown-down’); (2) leaf 540
age demographics are more related with the light-limited tropical trees which show 541
sensitivity to dry-season increased sunlight, and associated light-induced new leaf 542
flushing ( ‘green-up’) in response to seasonal and/or inter-annual drought (e.g. Wright & 543
van Schaik, 1994; Brando et al., 2010; Doughty et al., 2015; Lopes et al., 2016; Wu et 544
al., 2016). Leaf demography, especially dry-season leaf turnover, might also be an 545
evolved strategy of herbivory avoidance for tropical trees (e.g. Aide 1988, 1992). 546
Promisingly, these two phenological factors exert different effects on seasonal variation 547
in visible (especially in green band; Fig. S13 and Table 1) and SWIR (Fig. 3a) 548
Page 18 of 43New Phytologist
reflectances. Therefore it would be technically feasible and scientifically important to use 549
remote sensing products from multi-spectral or hyperspectral sensors that include visible 550
and NIR (, and ideally also SWIR) reflectances to differentiate these two processes in 551
future. 552
553
Canopy RTMs modeled seasonality helps reconcile the Amazon phenology debate 554
Our findings also have important implications for recent debates about the 555
mechanisms of satellite-observed ‘green-up’ of Amazon forest canopies (Huete et al., 556
2006; Brando et al., 2010; Galvao et al., 2011, 2013; Morton et al., 2014; Bi et al., 2015; 557
Saleska et al., 2016). Several studies (Galvao et al., 2011; Morton et al., 2014) have 558
suggested that satellite-observed seasonality in vegetation greenness (as captured by EVI) 559
are artifacts of sun-sensor geometry and not representative of biophysical factors in 560
forests measured on the ground. But our modeling of biophysical factors from the 561
bottom-up, by simulating EVI seasonality (with wet-season declines followed by dry 562
season green-up, Fig. 4b) that is consistent with top-down satellite observations that have 563
been corrected for artifacts of clouds/aerosols and of sun-sensor geometry (e.g. Brando et 564
al., 2010; Bi et al., 2015; Maeda et al., 2016; Saleska et al., 2016), suggests that these 565
biophysical factors are in fact driving the observed vegetation seasonality, at least at this 566
site. 567
Though derived from one site, these findings reveal general mechanisms for 568
phenology that effectively rule out the interpretation that remotely sensed patterns are 569
artifacts of changing sun-sensor geometry, as suggested by Galvão et al (2011, 2013) and 570
Morton et al (2014). Like this work, Morton et al (2014), in particular, sought to use 571
sophisticated RTM simulations to identify causal mechanisms for observed seasonality, 572
concluding that forest biophysical factors (e.g. phenology) were dominated by sun-sensor 573
geometry artifacts in the satellite observations. What accounts for the different 574
conclusions between the RTM-based study of Morton et al (2014) and the RTM-based 575
study presented here? 576
First, the BRDF correction applied by Morton et al (2014) to the satellite 577
observations appears to overestimate the size of the artifact that needs correction (Bi et 578
al., 2015) and in any case does not in fact eliminate detectable seasonality in greenness 579
Page 19 of 43 New Phytologist
(Saleska et al., 2016), as determined both by comparison to the BRDF-corrected EVI 580
used in Morton et al (2014) (Saleska et al., 2016), and to the more rigorously validated 581
BRDF-corrected MAIAC product (Lyapustin et al., 2012). 582
Second, in investigating the potential effects of vegetation phenology, though 583
Morton et al (2014) anticipated that individual leaf reflectances may change with age, 584
they did not account (i) for the fact that leaf transmittance also changes with age, or (ii) 585
for the possibility of seasonally changing canopy structure caused by the changing 586
fraction of leafless crowns. As discussed in the section above, the lower NIR absorptance 587
of young leaves (and consequent higher multiple-leaf-scattering) is what causes canopies 588
composed of young leaves to have higher overall canopy-scale NIR reflectance (Fig. 3), 589
and hence, higher EVI. Further, leafless crown fraction is observed, in tower-based 590
camera images, to change dramatically (by a factor of four, Fig 2b), with substantial 591
effect on the modeled reflectance seasonality. 592
This work thus clarifies relationships between tropical canopy phenology and 593
satellite-observed seasonality, helping to resolve debates regarding satellite-detected 594
Amazon phenology (e.g. Morton et al., 2014 vs. Bi et al., 2015 and Saleska et al., 2016). 595
In particular, we note that not accounting for higher NIR transmittance in young leaves, 596
or leafless crown fraction dynamics, will substantially diminish the simulated effect of 597
vegetation phenology on reflectance seasonality. 598
599
Phenological impacts on canopy reflectance, with implications for hyperspectral 600
retrievals of biophysical traits 601
By assessing the effects of leaf phenology on canopy reflectance spectra (Figs. 3-602
5), our results help extend leaf-to-canopy scaling with RTMs – widely applied to scaling 603
across space (e.g. Wessman et al., 1988; Weiss et al., 2000; Clark et al., 2005; Baret & 604
Buis, 2008; Asner & Martin, 2011; Asner et al., 2011, 2016; Singh et al., 2015) – to the 605
temporal domain. This extension is enabled by recent studies that demonstrate strong 606
relationships of leaf traits and spectra with leaf age across diverse growth environments 607
and species in tropical forests (Chavana-Bryant et al., 2017; Wu et al., 2017) and other 608
biomes (e.g. Yang et al., 2014, 2016; Meerdink et al., 2016). Our work builds on these 609
leaf-scale studies to provide support for the importance of remotely detectable temporal 610
Page 20 of 43New Phytologist
convergent relationships in traits and spectra at the canopy-scale as well, suggesting the 611
potential for deriving temporal variability in plant traits using canopy reflectance 612
spectroscopy techniques. 613
Importantly, our work also provides a practical guide for effective trait retrieval 614
using canopy reflectance spectroscopy techniques, in the context of both model inversion 615
(Weiss et al., 2000; Baret & Buis, 2008) and empirical statistical approaches, such as 616
partial least squares regression (PLSR; Asner et al., 2011; Singh et al., 2015). These are: 617
(i) Model inversion: In our work, though RTM-simulated canopy reflectance 618
captured the seasonal dynamics of satellite observations (which was our main focus), 619
there was a significant offset between the two, especially in the PROSAIL simulation of 620
NIR and EVI (Figs. 4 and S14). These differences are likely associated with uncertainty 621
in the foliage clumping effect (e.g. Fig. S16), which causes uncertainty in the absolute 622
magnitude of canopy reflectance. More fine-scale measurements and modeling 623
experiments are needed to fully understand such model-observation magnitude 624
differences to better constrain the RTMs and reduce potential bias in model-inverted key 625
plant traits. 626
(ii) Empirical statistical approaches (e.g. PLSR): since leaf age affects both leaf 627
traits and spectra (Wu et al., 2017), as well as canopy-scale reflectance spectra (Fig. 3), 628
our work highlights the importance of sampling leaves of different representative leaf 629
ages and of accounting for leaf demography within canopies, which will allow 630
development of more reliable canopy-scale trait-spectra relationships in the tropics. 631
Beyond the tropical forests studied here, this consideration is also likely important for 632
any biome that has seasonally changing traits and spectra (e.g. Yang et al., 2016; 633
Meerdink et al., 2016). 634
635
Conclusion 636
This study demonstrates that different components of leaf phenology (primarily 637
leafless crown fraction and leaf demography) contribute significantly to satellite-638
observed seasonal variations in canopy greenness (i.e. MAIAC EVI). These findings 639
effectively reconcile current controversies about satellite-detected vegetation seasonality 640
in the Amazon, and provide a robust basis for using satellite remote sensing (after 641
Page 21 of 43 New Phytologist
properly processed, like MAIAC EVI used here) to monitor phenology and study 642
climate-phenology relationships in the tropics. This work thus lays the foundation for the 643
study of how these phenological factors, which represent different ecophysiological 644
strategies by which tropical trees respond to varying resource availability, structure the 645
large scale satellite observations of forest dynamics across space and over time, thus 646
offering new insights to study tropical climate-phenology relationships. 647
648
Acknowledgements 649
This work was supported by NASA (#NNX11AH24G, #NNX17AF56G, and NESSF to 650
J.W.), NSF (PIRE #0730305), and the Next-Generation Ecosystem Experiments–Tropics 651
project supported by the U.S. DOE, Office of Science, Office of Biological and 652
Environmental Research and through contract #DE-SC0012704 to Brookhaven National 653
Laboratory. K.H. was supported by JAXA GCOM-C (RA6 #111) and JSPS KAKENHI 654
(#16H02948). S.C.S. was supported by NSF (EF-1340604). A.R.H., N.N.T. and S.G. 655
were supported by an Australian Research Council - Discovery Project (ARC-656
DP140102698, CI Huete). DigitalGlobe data were provided by NASA's NGA 657
Commercial Archive Data (cad4nasa.gsfc.nasa.gov) under the National Geospatial 658
Intelligence Agency's NextView license agreement. We also thank the three anonymous 659
reviewers for their constructive comments to improve the scientific rigor and clarity of 660
the manuscript. 661
662
Author Contributions 663
J.W., H.K., S.C.S., S.R.S., and A.R.H. designed the project. J.W. and T.M. performed 664
field-based tissue optical measurements, with the help from R.C.O. H.K. and Y.W. 665
developed the FLiES code. S.C.S. processed and analyzed the airborne LiDAR data. 666
S.P.S. and R.M. provided and processed the WorldView-2 image. A.R.H., N.N.T and 667
S.G. downloaded and processed the MODIS data. J.W., H.K. and S.C.S. performed the 668
synthetic data analysis. J.W. drafted the manuscript, and H.K., S.C.S., S.R.S., A.R.H., 669
S.P.S., B.W.N., K.G., D.G.D., A.R., T.M., N.R., R.M., N.N.T., and S.G. contributed to 670
the final version. 671
672
Page 22 of 43New Phytologist
673
References 674
Aide TM. 1988. Herbivory as a selective agent on the timing of leaf production in a 675
tropical understory community. Nature 336: 574–575. 676
Aide TM. 1992. Dry season leaf production: an escape from herbivory. Biotropica 24: 677
532–537. 678
Anderson LO, Malhi Y, Aragão LE, Ladle R, Arai E, Barbier N, Phillips O. 2010. 679
Remote sensing detection of droughts in Amazonian forest canopies. New 680
Phytologist 187:733-750. 681
Asner GP, Martin RE, Knapp DE, Tupayachi R, Anderson C, Carranza L, 682
Martinez P, Houcheime M, Sinca F, Weiss P. 2011. Spectroscopy of canopy 683
chemicals in humid tropical forests. Remote Sensing of Environment 115:3587-98. 684
Asner GP, Martin RE. 2011. Canopy phylogenetic, chemical and spectral assembly in a 685
lowland Amazonian forest. New Phytologist 189:999-1012. 686
Asner GP, Knapp DE, Anderson CB, Martin RE, Vaughn N. 2016. Large-scale 687
climatic and geophysical controls on the leaf economics spectrum. Proceedings of 688
the National Academy of Sciences 113, E4043-E4051, 689
doi:10.1073/pnas.1604863113. 690
Baret F, Jacquemoud S, Guyot G, Leprieur C. 1992. Modeled analysis of the 691
biophysical nature of spectral shifts and comparison with information content of 692
broad bands. Remote Sensing of Environment 41:133-42. 693
Baret F, Buis S. 2008. Estimating canopy characteristics from remote sensing 694
observations: Review of methods and associated problems. In: Liang S, eds. 695
Advances in land remote Sensing. Springer, Dordrecht, 173-201. 696
Bi J, Knyazikhin Y, Choi S, Park T, Barichivich J, Ciais P, Fu R, Ganguly S, Hall 697
F, Hilker T et al. 2015. Sunlight mediated seasonality in canopy structure and 698
photosynthetic activity of Amazonian rainforests. Environmental Research 699
Letters 10: 064014. 700
Bohlman SA. 2010. Landscape patterns and environmental controls of deciduousness in 701
forests of central Panama. Global Ecology and Biogeography 19:376-85. 702
Brando PM, Goetz SJ, Baccini A, Nepstad DC, Beck PS, Christman MC. 2010. 703
Seasonal and interannual variability of climate and vegetation indices across the 704
Amazon. Proceedings of the National Academy of Sciences 107: 14685-14690. 705
Page 23 of 43 New Phytologist
Chavana-Bryant C, Malhi Y, Wu J, Asner GP, Anatasiou A, Enquist BJ, Saleska 706
SR, Doughty C, Gerard F. 2017. Leaf aging of Amazonian canopy trees 707
revealed by spectral and physiochemical measurements. New Phytologist 708
214:1049-1063. 709
Chen JM, Plummer PS, Rich M, Gower ST, Norman JM. 1997. Leaf area index 710
measurements. Journal of Geophysical Research 102:429-443. 711
Clark ML, Roberts DA, Clark DB. 2005. Hyperspectral discrimination of tropical rain 712
forest tree species at leaf to crown scales. Remote Sensing of Environment 96:375-713
98. 714
Doughty CE, Goulden ML. 2008. Seasonal patterns of tropical forest leaf area index 715
and CO2 exchange. Journal of Geophysical Research: Biogeosciences (2005–716
2012) 113, G00B06, doi:10.1029/2007JG000590. 717
Doughty CE, Metcalfe DB, Girardin CA, Amézquita FF, Cabrera DG, Huasco WH, 718
Silva-Espejo JE, Araujo-Murakami A, Da Costa MC, Rocha W et al. 2015. 719
Drought impact on forest carbon dynamics and fluxes in Amazonia. Nature 720
519:78-82. 721
Feret JB, François C, Asner GP, Gitelson AA, Martin RE, Bidel LP, Ustin SL, Le 722
Maire G, Jacquemoud S. 2008. PROSPECT-4 and 5: Advances in the leaf 723
optical properties model separating photosynthetic pigments. Remote Sensing of 724
Environment 112:3030-43. 725
Galvao LS, dos Santos JR, Roberts DA, Breunig FM, Toomey M, de Moura YM. 726
2011. On intra-annual EVI variability in the dry season of tropical forest: A case 727
study with MODIS and hyperspectral data. Remote Sensing of Environment 728
115:2350-2359. 729
Galvão LS, Breunig FM, dos Santos JR, de Moura YM. 2013. View-illumination 730
effects on hyperspectral vegetation indices in the Amazonian tropical forest. 731
International Journal of Applied Earth Observation and Geoinformation 21:291-732
300. 733
Grömping U. 2015. Variable importance in regression models. WIREs Comput Stat 7: 734
137-152. 735
Guan K, Pan M, Li H, Wolf A, Wu J, Medvigy D, Caylor KK, Sheffield J, Wood EF, 736
Malhi Y et al. 2015. Photosynthetic seasonality of global tropical forests 737
constrained by hydroclimate. Nature Geoscience 8:284-9. 738
He L, Chen JM, Pisek J, Schaaf CB, Strahler AH. 2012. Global clumping index map 739
Page 24 of 43New Phytologist
derived from the MODIS BRDF product. Remote Sensing of Environment 740
119:118-30. 741
Hilker T, Lyapustin AI, Tucker CJ, Hall FG, Myneni RB, Wang Y, Bi J, de Moura 742
YM, Sellers PJ. 2014. Vegetation dynamics and rainfall sensitivity of the 743
Amazon. Proceedings of the National Academy of Sciences 111:16041-16046. 744
Hilker T, Galvão LS, Aragão LE, de Moura YM, do Amaral CH, Lyapustin AI, Wu 745
J, Albert LP, Ferreira MJ, Anderson LO et al. 2017. Vegetation chlorophyll 746
estimates in the Amazon from multi-angle MODIS observations and canopy 747
reflectance model. International Journal of Applied Earth Observation and 748
Geoinformation 58:278-87. 749
Huete A, Didan K, Miura T, Rodriguez EP, Gao X, Ferreira LG. 2002. Overview of 750
the radiometric and biophysical performance of the MODIS vegetation indices. 751
Remote Sensing of Environment 83:195-213. 752
Huete AR, Didan K, Shimabukuro YE, Ratana P, Saleska SR, Hutyra LR, Yang W, 753
Nemani RR, Myneni R. 2006. Amazon rainforests green‐up with sunlight in dry 754
season. Geophysical Research Letters 33, L06405, doi:10.1029/2005GL025583. 755
Hunter MO, Keller M, Morton D, Cook B, Lefsky M, Ducey M, Saleska S, de 756
Oliveira Jr R.C., Schietti J. 2015. Structural dynamics of tropical moist forest 757
gaps. PloS One 10: p.e0132144. doi:10.1371/journal.pone.0132144. 758
Huemmrich KF. 2001. The GeoSail model: a simple addition to the SAIL model to 759
describe discontinuous canopy reflectance. Remote Sensing of 760
Environment 75:423-431. 761
Hutyra LR, Munger JW, Saleska SR, Gottlieb E, Daube BC, Dunn AL, Amaral DF, 762
de Camargo PB, Wofsy SC. 2007. Seasonal controls on the exchange of carbon 763
and water in an Amazonian rain forest. Journal of Geophysical Research: 764
Biogeosciences (2005–2012) 112, G03008, doi:10.1029/2006JG000365. 765
Jacquemoud S, Baret F. 1990. PROSPECT: A model of leaf optical properties spectra. 766
Remote Sensing of Environment 34:75-91. 767
Jacquemoud S, Verhoef W, Baret F, Bacour C, Zarco-Tejada PJ, Asner GP, 768
François C, Ustin SL. 2009. PROSPECT+ SAIL models: A review of use for 769
vegetation characterization. Remote Sensing of Environment 113:56-66. 770
Jones MO, Kimball JS, Nemani RR. 2014. Asynchronous Amazon forest canopy 771
phenology indicates adaptation to both water and light availability. Environmental 772
Research Letters 9:124021, doi: 10.1088/1748-9326/9/12/124021. 773
Page 25 of 43 New Phytologist
Kobayashi H, Iwabuchi H. 2008. A coupled 1-D atmosphere and 3-D canopy radiative 774
transfer model for canopy reflectance, light environment, and photosynthesis 775
simulation in a heterogeneous landscape. Remote Sensing of Environment 776
112:173-85. 777
Kobayashi, H., N. Delbart, R. Suzuki, K. Kushida. 2010. A satellite-based method for 778
monitoring seasonality in the overstory leaf area index of Siberian larch forest, 779
Journal of Geophysical Research 115, G01002, doi:10.1029/2009JG000939. 780
Kobayashi H, Baldocchi DD, Ryu Y, Chen Q, Ma S, Osuna JL, Ustin SL. 2012. 781
Modeling energy and carbon fluxes in a heterogeneous oak woodland: a three-782
dimensional approach. Agricultural and Forest Meteorology 152:83-100. 783
Koetz B, Sun G, Morsdorf F, Ranson KJ, Kneubühler M, Itten K, Allgöwer B. 2007. 784
Fusion of imaging spectrometer and LIDAR data over combined radiative transfer 785
models for forest canopy characterization. Remote Sensing of Environment 786
106:449-59. 787
Lee JE, Frankenberg C, van der Tol C, Berry JA, Guanter L, Boyce CK, Fisher JB, 788
Morrow E, Worden JR, Asefi S et al. 2013. Forest productivity and water stress 789
in Amazonia: observations from GOSAT chlorophyll fluorescence. Proceedings 790
of the Royal Society of London B: Biological Sciences 280: 20130171, 791
http://dx.doi.org/10.1098/rspb.2013.0171. 792
Lopes AP, Nelson BW, Wu J, de Alencastro Graça PM, Tavares JV, Prohaska N, 793
Martins GA, Saleska SR. 2016. Leaf flush drives dry season green-up of the 794
Central Amazon. Remote Sensing of Environment 182:90-98. 795
Lyapustin AI, Wang Y, Laszlo I, Hilker T, Hall FG, Sellers PJ, Tucker CJ, Korkin 796
SV. 2012. Multi-angle implementation of atmospheric correction for MODIS 797
(MAIAC): 3. Atmospheric correction. Remote Sensing of Environment 127:385-798
393. 799
Maeda EE, Moura YM, Wagner F, Hilker T, Lyapustin AI, Wang Y, Chave J, 800
Mõttus M, Aragão LE, Shimabukuro Y. 2016. Consistency of vegetation index 801
seasonality across the Amazon rainforest. International Journal of Applied Earth 802
Observation and Geoinformation 52:42-53. 803
Meerdink SK, Roberts DA, King JY, Roth KL, Dennison PE, Amaral CH, Hook SJ. 804
2016. Linking seasonal foliar traits to VSWIR-TIR spectroscopy across California 805
ecosystems. Remote Sensing of Environment 186:322-38. 806
Meng R, Wu J, Schwager KL, Zhao F, Dennison PE, Cook BD, Brewster K, Green 807
TM, Serbin SP. 2017. Using high spatial resolution satellite imagery to map 808
Page 26 of 43New Phytologist
forest burn severity across spatial scales in a Pine Barrens ecosystem. Remote 809
Sensing of Environment 191:95-109. 810
Morton DC, Nagol J, Carabajal CC, Rosette J, Palace M, Cook BD, Vermote EF, 811
Harding JD, North PR. 2014. Amazon forests maintain consistent canopy 812
structure and greenness during the dry season. Nature 506:221-224. 813
Morton DC, Rubio J, Cook BD, Gastellu-Etchegorry JP, Longo M, Choi H, Hunter 814
MO, Keller M. 2016. Amazon forest structure generates diurnal and seasonal 815
variability in light utilization. Biogeosciences 13:2195. 816
Nepstad DC, Tohver IM, Ray D, Moutinho P, Cardinot G. 2007. Mortality of large 817
trees and lianas following experimental drought in an Amazon forest. Ecology, 818
88:2259-69. 819
Olivas PC, Oberbauer SF, Clark DB, Clark DA, Ryan MG, O’Brien JJ, Ordonez H. 820
2013. Comparison of direct and indirect methods for assessing leaf area index 821
across a tropical rain forest landscape. Agricultural and Forest Meteorology 822
177:110-6. 823
Ollinger SV. 2011. Sources of variability in canopy reflectance and the convergent 824
properties of plants. New Phytologist 189:375-94. 825
Pinty B, Widlowski JL, Taberner M, Gobron N, Verstraete MM, Disney M, Gascon 826
F, Gastellu JP, Jiang L, Kuusk A et al. 2004. Radiation Transfer Model 827
Intercomparison (RAMI) exercise: Results from the second phase. Journal of 828
Geophysical Research: Atmospheres 109: D06210, doi:10.1029/ 2003JD004252. 829
Restrepo-Coupe N, da Rocha HR, Hutyra LR, da Araujo AC, Borma LS, 830
Christoffersen B, Cabral OMR, de Camargo PB, Cardoso FL, da Costa ACL 831
et al. 2013. What drives the seasonality of photosynthesis across the Amazon 832
basin? A cross-site analysis of eddy flux tower measurements from the Brasil flux 833
network. Agricultural and Forest Meteorology 182: 128-144. 834
Rice AH, Pyle EH, Saleska SR, Hutyra L, Palace M, Keller M, de Camargo PB, 835
Portilho K, Marques DF, Wofsy SC. 2004. Carbon balance and vegetation 836
dynamics in an old-growth Amazonian forest. Ecological Applications 14: 55-71. 837
Roberts DA, Nelson BW, Adams JB, Palmer F. 1998. Spectral changes with leaf aging 838
in Amazon caatinga. Trees 12: 315-325. 839
Saatchi S, Asefi-Najafabady S, Malhi Y, Aragão LE, Anderson LO, Myneni RB, 840
Nemani R. 2013. Persistent effects of a severe drought on Amazonian forest 841
canopy. Proceedings of the National Academy of Sciences 110:565-70. 842
Page 27 of 43 New Phytologist
Saleska SR, Didan K, Huete AR, Da Rocha HR. 2007. Amazon forests green-up 843
during 2005 drought. Science 318:612-612. 844
Saleska, S.R., J. Wu, K. Guan, A.C. Araujo, A. Huete, A.D. Nobre, N. Restrepo‐‐‐‐845
Coupe. 2016. Brief Communications Arising: Dry season greening of Amazon 846
forests. Nature 531: E4-E5. 847
Samanta A, Ganguly S, Hashimoto H, Devadiga S, Vermote E, Knyazikhin Y, 848
Nemani RR, Myneni RB. 2010. Amazon forests did not green‐up during the 849
2005 drought. Geophysical Research Letters 37: L05401, 850
doi:10.1029/2009GL042154. 851
Samanta A, Knyazikhin Y, Xu L, Dickinson RE, Fu R, Costa MH, Saatchi SS, 852
Nemani RR, Myneni RB. 2012a. Seasonal changes in leaf area of Amazon 853
forests from leaf flushing and abscission. Journal of Geophysical Research: 854
Biogeosciences 117: G01015, doi:10.1029/2011JG001818 855
Samanta A, Ganguly S, Vermote E, Nemani RR, Myneni RB. 2012b. Why is remote 856
sensing of amazon forest greenness so challenging?. Earth Interactions 16:1-4. 857
Schneider FD, Leiterer R, Morsdorf F, Gastellu-Etchegorry JP, Lauret N, Pfeifer N, 858
Schaepman ME. 2014. Simulating imaging spectrometer data: 3D forest 859
modeling based on LiDAR and in situ data. Remote Sensing of Environment 860
152:235-50. 861
Sellers PJ, Dickinson RE, Randall DA, Betts AK, Hall FG, Berry JA, Collatz GJ, 862
Denning AS, Mooney HA, Nobre CA et al. 1997. Modeling the exchanges of 863
energy, water, and carbon between continents and the atmosphere. Science 864
275:502-9. 865
Shiklomanov AN, Dietze MC, Viskari T, Townsend PA, Serbin SP. 2016. 866
Quantifying the influences of spectral resolution on uncertainty in leaf trait 867
estimates through a Bayesian approach to RTM inversion. Remote Sensing of 868
Environment 183:226-38. 869
Singh A, Serbin SP, McNeil BE, Kingdon CC, Townsend PA. 2015. Imaging 870
spectroscopy algorithms for mapping canopy foliar chemical and morphological 871
traits and their uncertainties. Ecological Applications 25:2180-97. 872
Stark SC, Leitold V, Wu JL, Hunter MO, de Castilho CV, Costa FR, McMahon SM, 873
Parker GG, Shimabukuro MT, Lefsky MA et al. 2012. Amazon forest carbon 874
dynamics predicted by profiles of canopy leaf area and light environment. 875
Ecology Letters 15:1406-1414. 876
Page 28 of 43New Phytologist
Stark SC, Enquist BJ, Saleska SR, Leitold V, Schietti J, Longo M, Alves LF, 877
Camargo PB, Oliveira RC. 2015. Linking canopy leaf area and light 878
environments with tree size distributions to explain Amazon forest demography. 879
Ecology Letters 18:636-45. 880
Tang H, Dubayah R. 2017. Light-driven growth in Amazon evergreen forests explained 881
by seasonal variations of vertical canopy structure. Proceedings of the National 882
Academy of Sciences 114:2640-2644. 883
Toomey M, Roberts D, Nelson B. 2009. The influence of epiphylls on remote sensing of 884
humid forests. Remote Sensing of Environment 113: 1787-1798. 885
Ustin SL, Riaño D, Hunt ER. 2012. Estimating canopy water content from 886
spectroscopy. Israel Journal of Plant Sciences 60:9-23. 887
Verhoef W. 1984. Light scattering by leaf layers with application to canopy reflectance 888
modeling: the SAIL model. Remote Sensing of Environment 16:125-141. 889
Verhoef W. 1985. Earth observation modeling based on layer scattering matrices. 890
Remote Sensing of Environment 17:165-178. 891
Verhoef W, Bach H. 2003. Simulation of hyperspectral and directional radiance images 892
using coupled biophysical and atmospheric radiative transfer models. Remote 893
Sensing of Environment 87:23-41. 894
Wang D, Liang S, He T, Yu Y, Schaaf C, Wang Z. 2015. Estimating daily mean land 895
surface albedo from MODIS data. Journal of Geophysical Research: Atmospheres 896
120:4825-41. 897
Wagner FH, Anderson LO, Baker TR, Bowman DM, Cardoso FC, Chidumayo EN, 898
Clark DA, Drew DM, Griffiths AD, Maria VR et al. 2016. Climate seasonality 899
limits leaf carbon assimilation and wood productivity in tropical forests. 900
Biogeosciences 13:2537-2562. 901
Weiss M, Baret F, Myneni R, Pragnère A, Knyazikhin Y. 2000. Investigation of a 902
model inversion technique to estimate canopy biophysical variables from spectral 903
and directional reflectance data. Agronomie 20:3-22. 904
Wessman CA, Aber JD, Peterson DL, Melillo JM. 1988. Remote sensing of canopy 905
chemistry and nitrogen cycling in temperate forest ecosystems. Nature 335:154-906
156. 907
Wright SJ, Van Schaik CP. 1994. Light and the phenology of tropical trees. The 908
American Naturalist 143:192-199. 909
Page 29 of 43 New Phytologist
Wu J, Albert LP, Lopes AP, Restrepo-Coupe N, Hayek M, Wiedemann K, Guan K, 910
Stark SC, Christoffersen B, Prohaska N et al. 2016. Leaf development and 911
demography explain photosynthetic seasonality in Amazonian evergreen forests. 912
Science 351: 972-976. 913
Wu J, Chavana‐‐‐‐Bryant C, Prohaska N, Serbin SP, Guan K, Albert LP, Yang X, 914
Leeuwen WJ, Garnello AJ, Martins G et al. 2017. Convergence in relationships 915
between leaf traits, spectra and age across diverse canopy environments and two 916
contrasting tropical forests. New phytologist 214:1033-1048. 917
Xu L, Samanta A, Costa MH, Ganguly S, Nemani RR, Myneni RB. 2011. 918
Widespread decline in greenness of Amazonian vegetation due to the 2010 919
drought. Geophysical Research Letters 38: L07402, doi:10.1029/2011GL046824. 920
Xu X, Medvigy D, Powers JS, Becknell JM, Guan K. 2016. Diversity in plant 921
hydraulic traits explains seasonal and inter‐annual variations of vegetation 922
dynamics in seasonally dry tropical forests. New Phytologist 212:80-95. 923
Yang X, Tang J, Mustard JF. 2014. Beyond leaf color: Comparing camera‐based 924
phenological metrics with leaf biochemical, biophysical, and spectral properties 925
throughout the growing season of a temperate deciduous forest. Journal of 926
Geophysical Research: Biogeosciences 119:181-91. 927
Yang X, Tang J, Mustard JF, Wu J, Zhao K, Serbin S, Lee JE. 2016. Seasonal 928
variability of multiple leaf traits captured by leaf spectroscopy at two temperate 929
deciduous forests. Remote Sensing of Environment 179:1-12. 930
Zhou L, Tian Y, Myneni RB, Ciais P, Saatchi S, Liu YY, Piao S, Chen H, Vermote 931
EF, Song C et al. 2014. Widespread decline of Congo rainforest greenness in the 932
past decade. Nature 509:86-90. 933
934
935
Page 30 of 43New Phytologist
Figure Legends 936
937
Fig 1. Plant tissue optics at the k67 site: (a) field observed leaf reflectance of three age 938
classes, young (n=11 tree-canopy light conditions, see Table S3; ≤2 months, in blue), 939
mature (n=11; 3-5 months, in green), and old (n=11; 6-14 months, in red); (b) field 940
observed bark reflectance at 1.3m above the ground (n=13 tree species, see Table S5; in 941
black), and litter reflectance (n=40 tree species, see Methods; in orange); (c) PROSPECT 942
model inversion of leaf transmittance of three age classes (n=11 tree-canopy light 943
conditions; see Methods); (d) modeled leaf absorptance of three age classes 944
(absorptance=1-reflectance-transmittance; n=11 tree-canopy light conditions). Color lines 945
for the mean; shading for one standard deviation. 946
947
Fig 2. Field derived canopy-scale phenological and structural indices at the k67 site: (a) 948
ground measurements of mean annual cycle of leaf area index (LAI; monthly 949
measurements from January 2000 to December 2005; in green) and tower-camera derived 950
mean seasonality of leafless crown fraction (daily measurements from January 2010 to 951
December 2011, adapted from Wu et al., 2016, and see Methods; in black); (b) field-952
estimated seasonality of leaf age fraction for three leaf age classes (adapted from Fig. 3A 953
in Wu et al., 2016), using a leaf demography model as Wu et al (2016), constrained to 954
sum to total ground-observed LAI (green squares in panel a), with young (1-2 months, in 955
blue), mature (3-5 months, in green), and old (>=6 months, in red); (c) a 2012 airborne 956
LiDAR estimated and gap-filled mean canopy area density (i.e. the sum of leaf and 957
woody area density, which were assumed a constant fraction, 0.89, for leaves; m2
m-3
) 958
along the entire vertical canopy profile. Error bars in (a) indicate one standard deviation; 959
Shadings in (a) and (b) indicate the dry season; Shading in (c) indicates 95% confidence 960
interval on the mean of gap-filled data. 961
962
Fig 3. Canopy radiative transfer models (RTMs) simulated and WorldView-2 (WV2) 963
observed canopy-scale reflectances at each of different canopy phenophases at the k67 964
site. (a) PROSAIL simulated canopy-scale reflectance spectra for young (leaf age ≤2 965
months; in blue), mature (leaf age: 3-5 months; in green), old (leaf age: 6-14 months; in 966
red), and leafless (in black) phenophases; (b) PROSAIL simulated canopy-scale 967
reflectance convolved to WorldView-2 (WV-2) spectral bands (see Table S1); (c) FLiES 968
simulated canopy-scale reflectance convolved to WV-2 spectral bands (see Table S1); (d) 969
the WV-2 image (acquired on 28 July 2011 with near-nadir view; see Methods) observed 970
canopy-scale reflectances for each of three phenophases. All RTMs results shown here 971
are based on the scenario when canopy LAI=5m2 m
-2, tree-environment specific (n=11 972
tree-canopy light conditions), age-dependent leaf optics for young, mature, and old 973
phenophases respectively, multi-species (n=13 tree species) average bark optics for 974
leafless phenophase, SZA=45° and nadir view. Shading for one standard deviation. See 975
Fig. S2 for WV-2 band spectral response functions; see Figs. S9 and S10 for model 976
simulations of other canopy LAI. 977
978
Fig 4. Comparisons between canopy radiative transfer models (RTMs) simulated and 979
MAIAC version of MODIS observed canopy-scale seasonality of (a) NIR reflectance, 980
and of (b) EVI. The models were parameterized by ‘comprehensive’ monthly 981
Page 31 of 43 New Phytologist
phenological components (leaf area index—LAI, leafless crown fraction and leaf 982
demographics, i.e. P1+P2+P3 in Methods), with PROSAIL in grey and FLiES in black. 983
RTMs results shown here are the mean of their respective 11 RTM simulations (driven by 984
tree-environment specific leaf optics; see Methods S5). MAIAC data (in red) represents 985
monthly means for 2000-2014, spatially averaged over a 5x5 km2 window, and fully 986
accounts for sun-sensor geometry and cloud/aerosol contamination (see Methods). r2, 987
coefficient of determination, is based on the linear regression between model and 988
observations with intercept. Error bar indicates one standard deviation; Shading indicates 989
the dry season. 990
991
Fig 5. Relative roles of different phenological components in accounting for 992
‘comprehensive’ models simulated canopy reflectance seasonality in (a, c) NIR 993
reflectance, and (b, d) EVI, convolved to MODIS spectral bands (see Fig. S2). Top panel 994
for the PROSAIL model and bottom panel for the FLiES model. Canopy radiative 995
transfer models (RTMs) results shown here are the mean of their respective 11 RTM 996
simulations (driven by tree-environment specific leaf optics; see Methods S5). The 997
models that include only the direct LAI phenology effect (P1) are shown in blue, the 998
models that include only canopy-surface leafless crown fraction seasonality (P2) are 999
shown in red, and the models that include only the leaf demography seasonality effect 1000
(P3) are shown in green, and the ‘comprehensive’ models which include all three 1001
phenological components (P1+P2+P3) are shown in black. Shading indicates the dry 1002
season. 1003
1004
Fig 6. Different leaf age effects on leaf- and canopy-scale (a) NIR reflectance and (b) 1005
EVI, convolved to MODIS spectral bands (see Fig. S2). Leaf-scale patterns (red lines) are 1006
based on the field observations shown in Fig. 1a and canopy-scale patterns (black lines) 1007
are based on the PROSAIL model results shown in Fig. 3. Note the reversal of the age 1008
ranks for both NIR reflectance and EVI when going from leaf scale to canopy scale. 1009
Since the FLiES model simulates an age-dependency of canopy reflectance spectra 1010
similar to that of PROSAIL, we expect that FLIES has a similar change in the ranking. 1011
Error bars are one standard deviation among 11 tree-canopy light conditions (leaf-scale) 1012
and associated PROSAIL simulations (canopy-scale). 1013
1014
Page 32 of 43New Phytologist
Table 1. Relative contributions of the three phenological factors in accounting for the 1051
‘comprehensive’ canopy radiative transfer models (PROSAIL and FLiES) simulated 1052
seasonalities of canopy reflectance and vegetation indices (see Figs. 5 and S13). 1053
Spectra\Relative contribution P1 P2 P3
PROSAIL Blue 0.00 0.95*** 0.05**
Green 0.00 0.03 0.97***
Red 0.01 0.63** 0.37**
NIR 0.02*** 0.31*** 0.67***
EVI 0.03*** 0.68*** 0.29***
NDVI 0.00 0.83*** 0.17**
FLiES Blue 0.00 0.48*** 0.52***
Green 0.00 0.04*** 0.96***
Red 0.00 0.52*** 0.48***
NIR 0.00 0.30*** 0.70***
EVI 0.01 0.33*** 0.66***
NDVI 0.00* 0.46*** 0.54***
where PROSAIL and FLiES simulated canopy reflectance and vegetation indices were 1054
convolved to MODIS spectral bands (see Fig. S2); the three phenological factors include: 1055
P1—the leaf area index (LAI) effect, P2–the canopy-surface leafless crown fraction 1056
effect, and P3—the leaf age effect; the relative contributions were assessed by using a 1057
relative importance of regressors in R (package relaimpo; see Methods); asterisks 1058
indicate levels of significance (*, P=0.05; **, P=0.01; ***, P=0.001). 1059
1060
1061
Page 33 of 43 New Phytologist
Supporting Information 1062
Additional Supporting Information may be found online in the Supporting Information 1063
tab for this article: 1064
1065
Fig. S1 A WorldView-2 true-color RGB image of the k67 site acquired on 28 July 2011, 1066
near-nadir view. 1067
Fig. S2 Spectral response functions for the eight bands of WorldView-2 (WV-2) and for 1068
the four MODIS bands used in this study. 1069
Fig. S3 Effects of spatially averaged window sizes, centered at the k67 site, on MAIAC 1070
version of MODIS land surface reflectances and vegetation indices seasonality. 1071
Fig. S4 Consistency among different versions of MODIS BRDF-corrected products of 1072
land surface reflectances and vegetation indices seasonality using spatial averaging in a 1073
5x5 km2 window centered at the k67 site. 1074
Fig. S5 PROSPECT-inverted leaf transmittance and the associated validation. 1075
Fig. S6 Age-dependence of leaf optical properties (i.e. reflectance, transmittance, and 1076
absorptance) for 11 tree-canopy light conditions at the k67 site. 1077
Fig. S7 Airborne LiDAR derived two-dimensional canopy surface height aboveground of 1078
a 600x600 m2 forest landscape at the k67 site. 1079
Fig. S8 Cross comparisons of multi-species average age-dependent leaf optical properties 1080
between all 11-tree canopy light conditions and all but excluding M.Huberi which has 1081
significant bias in PROSPECT-inverted leaf transmittance. 1082
Fig. S9 PROSAIL simulated canopy-scale reflectances over the full spectral range (400-1083
2500 nm), for canopies containing only one of four different canopy phenophases, under 1084
varying canopy LAI. 1085
Fig. S10 FLiES simulated canopy-scale reflectances of eight WorldView-2 (WV-2) 1086
bands, for canopies containing only one of four different canopy phenophases, under 1087
varying canopy LAI. 1088
Fig.S11 Cross comparisons of multi-species average age-dependent canopy-scale 1089
reflectance between all 11-tree canopy light conditions and all but excluding M.Huberi 1090
which has significant bias in PROSPECT-inverted leaf transmittance. 1091
Fig. S12 Cross comparisons of relative roles of different phenological components in 1092
accounting for ‘comprehensive’ models simulated canopy reflectance seasonalities 1093
between all 11-tree canopy light conditions and all but excluding M.Huberi which has 1094
significant bias in PROSPECT-inverted leaf transmittance. 1095
Page 34 of 43New Phytologist
Fig. S13 Relative contributions of different phenological factors in accounting for the 1096
‘comprehensive’ models simulated canopy reflectance seasonalities in Blue, Green, and 1097
Red, and NDVI seasonality. 1098
Fig. S14 Comparisons between canopy radiative transfer models (RTMs) simulated and 1099
MAIAC version of MODIS observed canopy-scale reflectances and vegetation indices 1100
seasonality. 1101
Fig. S15 Comparisons between canopy radiative transfer models (RTMs) simulated and 1102
MCD43A1 version of MODIS observed canopy-scale reflectances and vegetation indices 1103
seasonality. 1104
Fig. S16 Sensitivity analysis of FLiES simulation results on shoot-scale clumping (an 1105
important parameter in FLiES), including FLiES simulated canopy-scale reflectances and 1106
vegetation indices. 1107
1108
Table S1 Specifications of the WorldView-2 (WV-2) image. 1109
Table S2 Equations for NDVI and EVI, based on MODIS reflectance bands in NIR, Red 1110
(R) and Blue (B). 1111
Table S3 Tree-canopy light conditions and associated canopy environments for leaf 1112
spectral measurements at the k67 site. 1113
Table S4 Number of leaves with spectral measurements for each of 11 tree-canopy light 1114
conditions at the k67 site. 1115
1116
Table S5 Species identification for trees with leaf and/or bark spectral measurements at 1117
three sites of Tapajos National Forests. 1118
1119
Methods S1 Image processing of the WorldView-2 image 1120
Methods S2 Data retrieval and processing of collection 6 MODIS data 1121
Methods S3 Leaf transmittance measurements at the Tapajos National Forests by Dr. T. 1122
Miura 1123
Methods S4 Retrieval of canopy area density from a 2012 airborne LiDAR survey at the 1124
k67 site 1125
Methods S5 Parameterization of the two canopy radiative transfer models (PROSAL and 1126
FLiES) 1127
Page 35 of 43 New Phytologist
Methods S6 The PROSAIL Model 1128
Page 36 of 43New Phytologist
Fig 1. Plant tissue optics at the k67 site: (a) field observed leaf reflectance of three age classes, young (n=11 tree-canopy light conditions, see Table S3; ≤2 months, in blue), mature (n=11; 3-5 months, in
green), and old (n=11; 6-14 months, in red); (b) field observed bark reflectance at 1.3m above the ground (n=13 tree species, see Table S5; in black), and litter reflectance (n=40 tree species, see Methods; in
orange); (c) PROSPECT model inversion of leaf transmittance of three age classes (n=11 tree-canopy light conditions; see Methods); (d) modeled leaf absorptance of three age classes (absorptance=1-reflectance-
transmittance; n=11 tree-canopy light conditions). Color lines for the mean; shading for one standard deviation.
218x151mm (300 x 300 DPI)
Page 37 of 43 New Phytologist
Fig 2. Field derived canopy-scale phenological and structural indices at the k67 site: (a) ground measurements of mean annual cycle of leaf area index (LAI; monthly measurements from January 2000 to December 2005; in green) and tower-camera derived mean seasonality of leafless crown fraction (daily
measurements from January 2010 to December 2011, adapted from Wu et al., 2016, and see Methods; in black); (b) field-estimated seasonality of leaf age fraction for three leaf age classes (adapted from Fig. 3A in Wu et al., 2016), using a leaf demography model as Wu et al (2016), constrained to sum to total ground-observed LAI (green squares in panel a), with young (1-2 months, in blue), mature (3-5 months, in green),
and old (>=6 months, in red); (c) a 2012 airborne LiDAR estimated and gap-filled mean canopy area density (i.e. the sum of leaf and woody area density, which were assumed a constant fraction, 0.89, for
leaves; m2 m-3) along the entire vertical canopy profile. Error bars in (a) indicate one standard deviation; Shadings in (a) and (b) indicate the dry season; Shading in (c) indicates 95% confidence interval on the
mean of gap-filled data.
Page 38 of 43New Phytologist
114x220mm (300 x 300 DPI)
Page 39 of 43 New Phytologist
Canopy radiative transfer models (RTMs) simulated and WorldView-2 (WV2) observed canopy-scale reflectances at each of different canopy phenophases at the k67 site. (a) PROSAIL simulated canopy-scale reflectance spectra for young (leaf age ≤2 months; in blue), mature (leaf age: 3-5 months; in green), old
(leaf age: 6-14 months; in red), and leafless (in black) phenophases; (b) PROSAIL simulated canopy-scale reflectance convolved to WorldView-2 (WV-2) spectral bands (see Table S1); (c) FLiES simulated canopy-scale reflectance convolved to WV-2 spectral bands (see Table S1); (d) the WV-2 image (acquired on 28
July 2011 with near-nadir view; see Methods) observed canopy-scale reflectances for each of three phenophases. All RTMs results shown here are based on the scenario when canopy LAI=5m2 m-2, tree-
environment specific (n=11 tree-canopy light conditions), age-dependent leaf optics for young, mature, and old phenophases respectively, multi-species (n=13 tree species) average bark optics for leafless
phenophase, SZA=45° and nadir view. Shading for one standard deviation. See Fig. S2 for WV-2 band spectral response functions; see Figs. S9 and S10 for model simulations of other canopy LAI.
252x155mm (300 x 300 DPI)
Page 40 of 43New Phytologist
Fig 4. Comparisons between canopy radiative transfer models (RTMs) simulated and MAIAC version of MODIS observed canopy-scale seasonality of (a) NIR reflectance, and of (b) EVI. The models were
parameterized by ‘comprehensive’ monthly phenological components (leaf area index—LAI, leafless crown
fraction and leaf demographics, i.e. P1+P2+P3 in Methods), with PROSAIL in grey and FLiES in black. RTMs results shown here are the mean of their respective 11 RTM simulations (driven by tree-environment specific
leaf optics; see Methods S5). MAIAC data (in red) represents monthly means for 2000-2014, spatially averaged over a 5x5 km2 window, and fully accounts for sun-sensor geometry and cloud/aerosol
contamination (see Methods). r2, coefficient of determination, is based on the linear regression between model and observations with intercept. Error bar indicates one standard deviation; Shading indicates the dry
season.
220x65mm (300 x 300 DPI)
Page 41 of 43 New Phytologist
Fig 5. Relative roles of different phenological components in accounting for ‘comprehensive’ models simulated canopy reflectance seasonality in (a, c) NIR reflectance, and (b, d) EVI, convolved to MODIS spectral bands (see Fig. S2). Top panel for the PROSAIL model and bottom panel for the FLiES model. Canopy radiative transfer models (RTMs) results shown here are the mean of their respective 11 RTM
simulations (driven by tree-environment specific leaf optics; see Methods S5). The models that include only the direct LAI phenology effect (P1) are shown in blue, the models that include only canopy-surface leafless crown fraction seasonality (P2) are shown in red, and the models that include only the leaf demography
seasonality effect (P3) are shown in green, and the ‘comprehensive’ models which include all three
phenological components (P1+P2+P3) are shown in black. Shading indicates the dry season.
223x129mm (300 x 300 DPI)
Page 42 of 43New Phytologist
Fig 6. Different leaf age effects on leaf- and canopy-scale (a) NIR reflectance and (b) EVI, convolved to MODIS spectral bands (see Fig. S2). Leaf-scale patterns (red lines) are based on the field observations
shown in Fig. 1a and canopy-scale patterns (black lines) are based on the PROSAIL model results shown in Fig. 3. Note the reversal of the age ranks for both NIR reflectance and EVI when going from leaf scale to canopy scale. Since the FLiES model simulates an age-dependency of canopy reflectance spectra similar to that of PROSAIL, we expect that FLIES has a similar change in the ranking. Error bars are one standard deviation among 11 tree-canopy light conditions (leaf-scale) and associated PROSAIL simulations (canopy-
scale).
191x69mm (300 x 300 DPI)
Page 43 of 43 New Phytologist