Decomposing reﬂectance spectra to track gross primary … · 2020. 9. 15. · 4524 R. Cheng et al.: Decomposing reﬂectance spectra to track gross primary production lighting all

Biogeosciences 17 4523ndash4544 2020httpsdoiorg105194bg-17-4523-2020copy Author(s) 2020 This work is distributed underthe Creative Commons Attribution 40 License

Decomposing reflectance spectra to track gross primary productionin a subalpine evergreen forestRui Cheng1 Troy S Magney111 Debsunder Dutta212 David R Bowling3 Barry A Logan4 Sean P Burns56Peter D Blanken5 Katja Grossmann78 Sophia Lopez4 Andrew D Richardson9 Jochen Stutz810 andChristian Frankenberg12

1Division of Geological and Planetary Sciences California Institute of Technology Pasadena CA USA2NASA Jet Propulsion Laboratory California Institute of Technology Pasadena CA USA3School of Biological Sciences University of Utah Salt Lake City UT USA4Department of Biology Bowdoin College Brunswick ME USA5Department of Geography University of Colorado Boulder CO USA6National Center for Atmospheric Research Boulder CO USA7Institute of Environmental Physics University of Heidelberg Heidelberg Germany8Joint Institute for Regional Earth System Science and Engineering University of California Los AngelesLos Angeles CA USA9Center for Ecosystem Science and Society and School of Informatics Computing and Cyber SystemsNorthern Arizona University Flagstaff AZ USA10Department of Atmospheric and Oceanic Sciences University of California Los Angeles Los Angeles CA USA11Department of Plant Sciences University of California Davis CA USA12Department of Civil Engineering Indian Institute of Science Bengaluru India

Correspondence Rui Cheng (ruichengcaltechedu) and Christian Frankenberg (cfrankencaltechedu)

Received 6 February 2020 ndash Discussion started 17 February 2020Revised 11 June 2020 ndash Accepted 2 August 2020 ndash Published 15 September 2020

Abstract Photosynthesis by terrestrial plants represents themajority of CO2 uptake on Earth yet it is difficult to measuredirectly from space Estimation of gross primary produc-tion (GPP) from remote sensing indices represents a primarysource of uncertainty in particular for observing seasonalvariations in evergreen forests Recent vegetation remotesensing techniques have highlighted spectral regions sensi-tive to dynamic changes in leafneedle carotenoid composi-tion showing promise for tracking seasonal changes in pho-tosynthesis of evergreen forests However these have mostlybeen investigated with intermittent field campaigns or withnarrow-band spectrometers in these ecosystems To investi-gate this potential we continuously measured vegetation re-flectance (400ndash900 nm) using a canopy spectrometer systemPhotoSpec mounted on top of an eddy-covariance flux towerin a subalpine evergreen forest at Niwot Ridge ColoradoUSA We analyzed driving spectral components in the mea-sured canopy reflectance using both statistical and process-

based approaches The decomposed spectral components co-varied with carotenoid content and GPP supporting the inter-pretation of the photochemical reflectance index (PRI) andthe chlorophyllcarotenoid index (CCI) Although the entire400ndash900 nm range showed additional spectral changes nearthe red edge it did not provide significant improvements inGPP predictions We found little seasonal variation in bothnormalized difference vegetation index (NDVI) and the near-infrared vegetation index (NIRv) in this ecosystem In addi-tion we quantitatively determined needle-scale chlorophyll-to-carotenoid ratios as well as anthocyanin contents usingfull-spectrum inversions both of which were tightly corre-lated with seasonal GPP changes Reconstructing GPP fromvegetation reflectance using partial least-squares regression(PLSR) explained approximately 87 of the variability inobserved GPP Our results linked the seasonal variation in re-flectance to the pool size of photoprotective pigments high-

Published by Copernicus Publications on behalf of the European Geosciences Union

4524 R Cheng et al Decomposing reflectance spectra to track gross primary production

lighting all spectral locations within 400ndash900 nm associatedwith GPP seasonality in evergreen forests

1 Introduction

Terrestrial gross primary production (GPP) the gross CO2uptake through photosynthesis is the largest uptake of at-mospheric CO2 (Ciais et al 2013) yet the uncertain-ties are large hampering our ability to monitor and pre-dict the response of the terrestrial biosphere to climatechange (Ahlstroumlm et al 2012) Hence accurately mappingGPP globally is critical In contrast to unevenly distributedground-level measurements such as Fluxnet (Baldocchi et al2001) satellites can infer GPP globally and uniformly Re-mote sensing techniques are based on the optical responseof vegetation to incoming sunlight which can track photo-synthesis via the absorption features of photosynthetic andphotoprotective pigments (Rouse et al 1974 Liu and Huete1995 Gamon et al 1992 2016) Progress is particularly im-portant for evergreen forests which can have large seasonaldynamics in photosynthesis but low variability in canopystructure and color However these promising techniquesstill lack a comprehensive evaluationvalidation using bothcontinuous in situ measurements and process-based simula-tions

GPP can be expressed as a function of photosyntheticallyactive radiation (PAR) the fraction of PAR absorbed by thecanopy (fPAR) and light-use efficiency (LUE)

GPP= PAR middot fPAR middotLUE (1)

with LUE representing the efficiency of plants to fix carbonusing absorbed light (Monteith 1972 Monteith and Moss1977) The accuracy of remote-sensing-derived GPP is lim-ited by the estimation of LUE which is more dynamic anddifficult to measure remotely than PAR and fPAR particu-larly in evergreen ecosystems There have been many stud-ies inferring the light absorbed by canopies (ie fPAR) fromvegetation indices (VIs) that estimate the ldquogreennessrdquo ofcanopies (Running et al 2004 Zhao et al 2005 Robinsonet al 2018 Glenn et al 2008) such as the normalized dif-ference vegetation index (NDVI Rouse et al 1974 Tucker1979) the enhanced vegetation index (EVI Liu and Huete1995 Huete et al 1997) and the near-infrared vegetation in-dex (NIRv Badgley et al 2017) Current GPP data productsderived from Eq (1) rely on the modulation of abiotic con-ditions to estimate LUE (Xiao et al 2004) LUE is derivedempirically by defining a general timing of dormancy for allevergreen forests with the same plant functional type (egKrinner et al 2005) or the same meteorological thresholds(eg Running et al 2004) However within the same cli-mate region or plant functional type forests are not identicalndash leading to uncertainties in estimated LUE (Stylinski et al

2002 Gamon et al 2016 Zuromski et al 2018) whichpropagate to the estimation of GPP

Because evergreen trees retain most of their needles andchlorophyll throughout the entire year (Bowling et al 2018)LUE in evergreens is regulated by needle biochemistry AsLUE falls with the onset of winter due to unfavorable envi-ronmental conditions and seasonal downregulation of pho-tosynthetic capacity evergreen needles quench excess ab-sorbed light via thermal energy dissipation that involves thexanthophyll cycle and other pigments (Adams and Demmig-Adams 1994 Demmig-Adams and Adams 1996 Verho-even et al 1996 Zarter et al 2006) Thermal energy dis-sipation is a primary de-excitation pathway measured bypulse-amplitude fluorescence as non-photochemical quench-ing (NPQ Schreiber et al 1986) At the same time a smallamount of radiation solar-induced fluorescence (SIF) viathe de-excitation of absorbed photons is emitted by photo-system II (Genty et al 1989 Krause and Weis 1991)

Some vegetation indices are sensitive to photoprotectivepigments (eg carotenoids) and can characterize the sea-sonality of evergreen LUE with some success For instancethe photochemical reflectance index (PRI Gamon et al1992 1997) and chlorophyllcarotenoid index (CCI Ga-mon et al 2016) both use wavelength regions that repre-sent carotenoid absorption features around 531 nm at the leaflevel (Wong et al 2019 Wong and Gamon 2015a b) andshow great promise for estimating photosynthetic seasonality(Hall et al 2008 Hilker et al 2011a) Due to the relativelyinvariant canopy structure in evergreen forests CCI and PRIhave been applied at the canopy level as well (Gamon et al2016 Garbulsky et al 2011 Middleton et al 2016) In addi-tion the green chromatic coordinate (GCC Richardson et al2009 2018 Sonnentag et al 2012) an index derived fromthe brightness levels of RGB canopy images is also capableof tracking the seasonality of evergreen GPP (Bowling et al2018) However the full potential of spectrally resolved re-flectance measurements to explore the photosynthetic phe-nology of evergreens has not been comprehensively exploredat the canopy scale The evaluation of pigment-driven spec-tral changes in evergreen forests over the course of a season isnecessary to determine where when and why certain wave-length regions could advance our mechanistic understand-ing of canopy photosynthetic and photoprotective pigmentsHowever this has not been done with both empirical andprocess-based methods using continuously measured canopyhyperspectral reflectance and in situ pigment samples

Here we used continuous measurements in both spec-tral space (full spectrum between 400 and 900 nm) and time(daily over an entire year) to evaluate the potential of hy-perspectral canopy reflectance for better understanding thesensitivity of VIs to pigment changes that regulate GPP inevergreen forests Continuous measurements of spectrally re-solved reflectance at the canopy scale have so far been sparseat evergreen forest sites (Gamon et al 2006 Hilker et al2011b Porcar-Castell et al 2015 Rautiainen et al 2018

Biogeosciences 17 4523ndash4544 2020 httpsdoiorg105194bg-17-4523-2020

R Cheng et al Decomposing reflectance spectra to track gross primary production 4525

Wong et al 2020) There are only a few empirical studies onhyperspectral canopy reflectance in evergreen forests (Smithet al 2002 Singh et al 2015) Yet empirically decom-posed canopy spectral reflectance has been used as a predic-tor of maximum photosynthetic capacity (Serbin et al 2012Barnes et al 2017 Dechant et al 2017 Silva-Perez et al2018 Meacham-Hensold et al 2019) GPP (Matthes et al2015 Huemmrich et al 2017 DuBois et al 2018 Huemm-rich et al 2019 Dechant et al 2019) and other physiolog-ical properties (Ustin et al 2004 2009 Asner et al 2011Serbin et al 2014)

In contrast to empirical methods process-based ap-proaches such as canopy radiative transfer models (RTMs)can help to quantitatively link canopy photosynthesis withleaf-level contents of photosyntheticphotoprotective pig-ments (Feret et al 2008 Jacquemoud et al 2009) WithRTMs we can use spectrally resolved reflectance to directlyderive leaf pigment contents (Feacuteret et al 2017 Jacquemoudet al 1995) and plant traits (Feacuteret et al 2019)

In addition to seasonal changes in pigment concentrationscanopy SIF was found to correlate significantly with the sea-sonality of photoprotective pigment content in a subalpineconiferous forest (Magney et al 2019a) Steady-state SIF isregulated by NPQ and photochemistry (Porcar-Castell et al2014) and it provides complementary information on canopyGPP Yang and van der Tol (2018) justified that the relativeSIF SIF normalized by the reflected near-infrared radiationis more representative of the physiological variations in SIFas it is comparable to a SIF yield (Guanter et al 2014 Gentyet al 1989) Our continuous optical measurements makeit possible to differentiate mechanisms undergoing seasonalchanges by comparing the decomposed reflectance spectrumagainst relative far-red SIF Additionally using relative SIFcan effectively correct for incoming irradiance and accountfor the sunlit and shade fractions within the observation fieldof view (FOV) of PhotoSpec (Magney et al 2019a)

In the present study we analyzed continuous canopy re-flectance data from PhotoSpec at a subalpine evergreen forestat the Niwot Ridge AmeriFlux site (US-NR1) in ColoradoUS and sought to understand the mechanisms controlling theseasonality of photosynthesis using continuous hyperspec-tral remote sensing We first explored empirical techniquesto study all seasonal variations in reflectance spectra identi-fied specific spectral regions that best explained the seasonalchanges in GPP and then linked these spectral features topigment absorption features that impacted both biochemicaland biophysical traits We also used full-spectral inversionsusing a canopy RTM to infer quantitative estimates of leafpigment pool sizes Finally we compared the spring onset ofphotosynthesis captured by different methods VIs and rel-ative SIF to determine the underlying mechanisms that con-tributed to photosynthetic phenology

2 Material and methods

21 Study site

The high-altitude (3050 m above sea level) subalpine ev-ergreen forest near Niwot Ridge Colorado US is anactive AmeriFlux site (US-NR1 lat 400329 N long1055464W tower height 26 m Monson et al 2002Burns et al 2015 2016 Blanken et al 2019) Three speciesdominate subalpine fir (Abies lasiocarpa var bifolia) En-gelmann spruce (Picea engelmannii) and lodgepole pine (Pi-nus contorta) with an average height of 115 m a leaf area in-dex of 42 (Burns et al 2016) and minimal understory Theannual mean precipitation and air temperature are 800 mmand 15 C respectively (Monson et al 2002) The highelevation creates an environment with cold winters (withsnow present more than half the year) while the relativelylow latitude (40 N) allows for year-round high solar irra-diation (Monson et al 2002) Thus trees have to dissipatea considerable amount of excess sunlight during winter dor-mancy which makes this forest an ideal site for studying sea-sonal variation in NPQ including the sustained componentof it during dormancy (Bowling et al 2018 Magney et al2019a)

22 Continuous tower-based measurements of canopyreflectance

PhotoSpec (Grossmann et al 2018) is a 2D scanning tele-scope spectrometer unit originally designed to measure SIFIt also features a broadband Flame-S spectrometer (OceanOptics Inc Florida USA) used to measure reflectancefrom 400 to 900 nm at a moderate (full width at half maxi-mum= 12 nm) spectral resolution with a FOV of 07 (moredetails in Grossmann et al 2018 Magney et al 2019a) Inthe summer of 2017 we installed a PhotoSpec system on thetop of the US-NR1 eddy-covariance tower from where wecan scan the canopy by changing both viewing azimuth an-gle and zenith angles On every other summer day and ev-ery winter day PhotoSpec scans the canopy by changing theview zenith angle with small increments at fixed view az-imuth angles ie elevation scans Only one azimuth positionis kept after 18 October 2017 to protect the mechanism frompotentially damaging winter conditions at the site Spectrallyresolved reflectance was calculated using direct solar irradi-ance measurements via a cosine diffuser mounted in the up-ward nadir direction (Grossmann et al 2018) as well as re-flected radiance from the canopy The reflectance data usedin this study are from 16 June 2017 to 15 June 2018

Here we integrated all elevation scans to daily-averagedreflectance (every other day before 18 October 2017) by us-ing all scanning viewing directions with vegetation in thefield of view over the course of a day filtering for both low-light conditions and thick clouds by requiring PAR to beboth at least 100 micromol mminus2 sminus1 and 60 of theoretical clear-

httpsdoiorg105194bg-17-4523-2020 Biogeosciences 17 4523ndash4544 2020

sky PAR A detailed description of data processing can befound in Appendix B To further test whether bidirectionalreflectance effects impacted our daily averages we com-pared the NDVI and NIRv at various canopy positions givena range of solar zenith and azimuth angles (Figs A1ndashA3)Neither of the daily averaged VIs was substantially impactedby the solar geometry supporting the robustness of daily av-eraged canopy reflectance An additional analysis (Fig A4)has also shown the variation in phase angle at a daily timestep is not a critical factor for the change in reflectance

About 49 winter days exhibited significantly higher re-flectances attributable to snow within the field of viewwhich we corroborated with canopy RGB imagery from thetower After removing data strongly affected by snow andexcluding the days of instrument outages 211 valid sampledays remained among which 96 valid sample days were be-tween DOY 100 and 300 The daily-averaged reflectance wascomputed as the median reflectance from all selected scansfor a single day which was then smoothed by a 10-point(37 nm) box-car filter over the spectral dimension (400ndash900 nm) to remove the noise in the spectra Figure 1a showsthe seasonally averaged and spectrally resolved canopy re-flectances measured by PhotoSpec

To further emphasize the change in reflectance as a re-sult of changes in pigment contents we transformed the re-flectance (shown as Rλ) using the negative logarithm (Eq 1)as light intensity diminishes exponentially with pigment con-tents (Horler et al 1983)

Rλ prop exp(minusC middot σ(λ)) (2a)C middot σ(λ)propminus log(Rλ) (2b)

Here σ is the absorption cross section of pigmentsTherefore the log-transformed reflectance (Fig 1b)

should correlate more linearly with pigment contents (shownas C) We also considered a variety of typical VIs using thereflectance data from PhotoSpec

NDVI=R800minusR670

R800+R670(Rouse et al 1974) (3a)

NIRv= NDVI middotR800 (Badgley et al 2017) (3b)

PRI=R531minusR570

R531+R570(Gamon et al 1992) (3c)

CCI=R526ndash536minusR620ndash670

R526ndash536+R620ndash670(Gamon et al 2016) (3d)

GCC=RGreen

RRed+RGreen+RBlue

(Richardson et al 2009) (3e)

In order to calculate GCC we convolved the reflectance us-ing the instrumental spectral response function (Fig S1 inthe Supplement Wingate et al 2015) of the StarDot Net-Cam SC 5 MP IR (StarDot Technologies Buena Park CAUSA) which is the standard camera model used by the Phe-noCam Network protocol (Sonnentag et al 2012)

Figure 1 (a) Seasonally averaged canopy reflectance in winterdormancy (red) and the growing season (black) from PhotoSpec(b) Seasonally averaged negative logarithm transformation of re-flectance (400ndash900 nm) For comparison we normalized the re-flectance by the value at 800 nm on each day Here we referred to13 Novemberndash18 April as dormancy and 2 Junendash21 August as themain growing season The seasonal averaged canopy reflectance iscomposed of 39 daily-average reflectance in the growing season and113 daily-averaged reflectance in the dormancy

In addition to the reflectance measurements we also in-cluded relative SIF far-red SIF normalized by the reflectednear-infrared radiance at 755 nm The far-red SIF (745ndash758 nm Grossmann et al 2018) was measured simultane-ously with reflectance with a QEPro spectrometer (OceanOptics Inc Florida USA) The daily relative SIF was pro-cessed in the same fashion as the reflectance

23 Eddy covariance measurements and LUE

Observations of net ecosystem exchange (net flux of CO2NEE) PAR and meteorological variables made at the US-NR1 tower are part of the official AmeriFlux Network data(Burns et al 2016) GPP was estimated in half-hourly inter-vals (Reichstein et al 2005) using the REddyProc package

(Wutzler et al 2018) allowing us to compute LUE (Gouldenet al 1996 Gamon et al 2016) at half-hourly intervals

According to the light response curves GPP is a non-linear function of PAR (Fig 2 Harbinson 2012) Magneyet al (2019a) showed that fPAR does not significantly varywith seasons We started to observe a photosynthetic satura-tion between 500 and 1000 micromol mminus2 sminus1 of PAR (Fig 2)when the carboxylation rate driven by maximum carboxyla-tion rate (Vcmax) became the limiting factor (Farquhar et al1980) Thus we defined the light-saturated GPP (GPPmax)as the mean half-hourly GPP at PAR levels between 1000and 1500 micromol mminus2 sminus1 a range which was covered through-out the year (Fig 2) even in winter Therefore GPPmax wasless susceptible to short-term changes in PAR Yet due to thelower light intensity during storms GPPmax was not alwaysavailable As suggested by the low PAR value at which lightsaturation happened plants remained in a light-saturatedcondition for most of the daytime A higher GPPmax indi-cates a greater Vcmax and maximum electron transport rate(Jmax) when the variation in GPPmax is independent of stom-atal conductance and intercellular CO2 concentration (Leun-ing 1995) Therefore GPPmax was closely correlated withdaily LUE driven by physiology (see Sect S24 in the Sup-plement)

We refrained from normalizing GPPmax by absorbed pho-tosynthetically active radiation (APAR) due to some of theAPAR measurements (see Sect S21 in the Supplement) notavailable in the beginning of growing season GPPmax wassignificantly linearly correlated with normalized GPPmax byAPAR (Fig S2c)

We also included air temperature (Tair) and vapor pressuredeficit (VPD) provided from the AmeriFlux network dataDaytime daily mean Tair and VPD were computed from av-eraging the half-hourly Tair and VPD when PAR was greaterthan 100 micromol mminus2 sminus1

24 Pigment measurements

To link canopy reflectance with variations in pigment con-tents we used pigment data (Bowling et al 2018 Bowl-ing and Logan 2019 Magney et al 2019a) at monthlyintervals over the course of the sampling period Herewe focused on the xanthophyll cycle pool size (vio-laxanthin+ antheraxanthin+ zeaxanthin V+A+Z) totalcarotenoid content (car) and total chlorophyll content (chl)measured on Pinus contorta and Picea engelmannii nee-dles with units of moles per unit fresh mass Car includesV+A+Z lutein neoxanthin and beta-carotene We alsocomputed the ratio of chlorophyll to carotenoid contents(chl car) because CCI derived from the Moderate Resolu-tion Imaging Spectroradiometer (MODIS) can track chl car(Gamon et al 2016) Overall we can match 10 individ-ual leaf-level sampling days for both pine and spruce sam-ples with reflectance measured withinplusmn2 d Among these 10

Figure 2 Half-hourly GPP as a function of PAR during the mea-surement period Points were colored by month Bold points werethe median GPP when PAR was binned every 100 micromol mminus2 sminus1

approximately The solid lines represent the canopy light responsecurve

valid sample days 6 sample days are between DOY 100 and300

25 Data-driven spectral decomposition

We assumed that the spectrally resolved reflectance is a re-sult of mixed absorption processes by different pigmentsThis allowed us to apply an independent component analy-sis (ICA Hyvaumlrinen and Oja 2000) to decompose the log-transformed reflectance matrix (day of the year in rows andspectral dimension in columns) into its independent com-ponents An advantage of the ICA is that it can separatea multivariate signal into additive subcomponents that aremaximally independent without the condition of orthogonal-ity (Comon 1994) We extracted three independent compo-nents which explained more than 9999 of the varianceusing the ICA algorithm (FastICA Python package scikit-learn v0210 Sect S4 in the Supplement) such as

minus log(RλDOY)=sum

(spectral componentiλ

middot temporal loadingiDOY) (4)

where i is the ith component in spectral spaceThe decomposed spectral components revealed character-

istic features that explain most of the variance in the re-flectance matrix which dictated the time-independent spec-tral shapes of pigment absorption features based on Eq (1)The corresponding temporal loadings showed temporal vari-ations in these spectral features ie the variations in pigmentcontents We will introduce the method of extracting pigmentabsorption features in a quantitative model-driven approachin Sect 26

In addition to analyzing the transformed reflectance alonewe empirically correlated the reflectance with GPPmax us-ing partial least-squares regression (PLSR Python packagescikit-learn v0210) PLSR is a predictive regression modelwhich solves for a coefficient that can maximally explainthe linear covariance of the predictor with multiple variables(Wold et al 1984 Geladi and Kowalski 1986) PLSR hasbeen used to successfully predict photosynthetic propertiesusing reflectance matrices in previous studies from the leaf tocanopy scales (eg Serbin et al 2012 2015 Barnes et al2017 Silva-Perez et al 2018 Woodgate et al 2019) Ap-plying the PLSR to the hyperspectral canopy reflectance andGPPmax resulted in a time-independent coefficient that em-phasizes the key wavelength regions which contribute to thecovariation of reflectance and GPPmax such as

GPPmaxDOY =minus log(RλDOY)timesPLSR coefficientGPPmaxλ (5)

We implemented another set of PLSR analyses on the re-flectance with individual pigment measurement as the tar-get variable such as the mean values of V+A+Z car andchl car such as

pigment measurement=minus log(RλDOY)

timesPLSR coefficientpigment measurementλ (6)

We did not include chl as one of the target variables in thisPLSR analysis since Bowling et al (2018) and Magney et al(2019a) have already shown chl did not vary seasonally inour study site Fitting the minimal variance in chl will lead tooverfitting the PLSR model

Comparing the PLSR coefficient of pigment measure-ments at the leaf level with the PLSR coefficient of GPPmaxconnected the changes in GPPmax to the pool size of photo-protective pigments because the reflectance is regulated bythe absorption of pigments

26 Process-based methods

PROSPECT+SAIL (PROSAIL Jacquemoud et al 2009)is a process-based 1-D canopy RTM that models canopyreflectance given canopy structure information (SAIL) aswell as leaf pigment contents (PROSPECT) (Jacquemoudand Baret 1990 Vilfan et al 2018)

We used PROSAIL (with PROSPECT-D Feacuteret et al2017) to compute the derivative of the daily-averagednegative logarithm transformed reflectance with respectto individual pigment contents namely chlorophyll con-tent (chlorophyll Jacobian partminuslog(R)

partCchl) and carotenoid content

(carotenoid Jacobian partminuslog(R)partCcar

) (Dutta et al 2019) Thishelped explain the decomposed spectral components fromthe empirical analysis

We also used PROSAIL to infer pigment contents (ieCchl Ccar Cant) by optimizing the agreement betweenPROSAIL-modeled reflectance and measured canopy daily-

mean reflectance from PhotoSpec We fixed canopy struc-tural parameters (eg the leaf area index (LAI) to 42 asreported in Burns et al 2015) and fitted leaf pigment compo-sitions as well as a low-order polynomial for soil reflectance(Appendix C) similar to Vilfan et al (2018) and Feacuteret et al(2017) The cost function J in Eq (7) represents a least-squares approach where R is the modeled reflectance

800 nmsumλ=450 nm

(Rλminus Rλ)2 (7)

We used the spectral range between 450 and 800 nm whichencompasses most pigment absorption features

3 Results and discussion

31 Seasonal cycle of GPPmax and environmentalconditions

As can be seen in Fig 3 the subalpine evergreen forest atNiwot Ridge exhibits strong seasonal variation in GPP TairVPD GPPmax and PAR GPP and GPPmax dropped to zerowhile sufficient PAR required for photosynthesis was stillavailable in the dormancy which suggests that the abiotic en-vironmental factors impact photosynthesis seasonality non-linearly and jointly

Abiotic factors played a strong role in regulating GPPmaxin this subalpine evergreen forest over the course of theseason For instance there was a strong dependence ofGPPmax with Tair However photosynthesis completely shutdown during dormancy even when the Tair exceeded 5 C(Fig 3) During the onset and cessation periods of photosyn-thesis GPPmax rapidly increased with temperature (Fig S3aleft panel) potentially because needle temperature co-variedwith Tair and needle temperature controls the activity ofphotosynthetic enzymes which affect Vcmax Spring warm-ing approaches the optimal temperature for photosyntheticenzymes leading to activation of photosynthesis while cool-ing in the early winter inhibits these enzymes (Rook 1969)Warming in spring melted frozen boles and made them avail-able for water uptake (Bowling et al 2018) and thus causedthe recovery of GPPmax (Monson et al 2005) Once the tem-perature was around the optimum (in the growing season)Tair was no longer the determining factor for photosynthesisHigher VPD caused by rising Tair can stress the plants suchthat stomata closed intercellular CO2 reduced and photo-synthesis decreased (Fig S3a right panel) When intercellu-lar CO2 concentration was not a limiting factor GPPmax wasmore representative of Vcmax and did not vary T significantly

32 Seasonal cycle of reflectance

In Fig 4 the Jacobians show the maximum sensitivity of thereflectance spectral shape to carotenoid content at 524 nmand near 566 and 700 nm for chlorophyll The first peak of

Figure 3 Daily-averaged time series of air temperature (Tair) vapor pressure deficit (VPD) photosynthetically active radiation (PAR) grossprimary production (GPP) from half-hourly data when PAR was greater than 100 micromol mminus2 sminus1 and time series of GPPmax DOY 166(2017) was the first day of observation The vertical dashed line divides the observations from day of year (DOY) for the years 2017 and2018

the chlorophyll Jacobian covers a wide spectral range in thevisible range while the second peak around the red edge isnarrower

It can be seen that the first spectral ICA component has asimilar shape as the chlorophyll Jacobian The correspondingtemporal loading has a range between minus02 and 02 withoutany obvious seasonal variation consistent with a negligibleseasonal cycle in chlorophyll content as shown in the pig-ment analysis However there is a gradual increase beforeDOY 50 in the first temporal loading which appears to beanti-correlated with the temporal loading of the second ICAstructure

Two major features in the second spectral component canbe observed One is a negative peak centered around 530 nmwhich aligns with the carotenoid Jacobian At the negativelogarithm scale the negative values resulting from the neg-ative ICA spectral peak multiplied by the positive ICA tem-poral loadings (growing season in Fig 4 middle plots) indi-cate there were fewer carotenoids during the growing season(Eqs 1 and 4) Conversely positive values resulting froma negative spectral peak multiplied by the negative tempo-ral loadings (dormancy in Fig 4 middle plots) indicate therewere more carotenoids during dormancy (ie sustained pho-toprotection via the xanthophyll pigments Bowling et al2018) Another feature is the valley-trough shape which is

co-located with the chlorophyll Jacobian center at the longerwavelength in the red-edge region The center of this fea-ture occurs at the shorter-wavelength edge of the chloro-phyll Jacobian but does not easily explain changes in to-tal chlorophyll content which should show equal changesaround 600 nm The corresponding temporal loading appar-ently varied seasonally with GPPmax

The second temporal loading transitioned more graduallyfrom dormancy to the peak growing season than GPPmaxUnfortunately we were missing data to evaluate the relativetiming of GPPmax cessation

The third spectral component is similar to the mean shapeof reflectance spectra Its temporal loading remained aroundzero throughout the year

Overall the second ICA spectral component is more rep-resentative of the seasonal variation in the magnitude of totalcanopy reflectance than the other spectral components Thespectral changes around the red edge in the second compo-nent are interesting and might be related to structural nee-dle changes in chlorophyll-a and chlorophyll-b contributions(de Tomaacutes Mariacuten et al 2016 Rautiainen et al 2018) whichare not separated in PROSPECT

CCI and PRI (Fig 5andashb) followed the seasonal cycle ofGPPmax closely CCI and PRI use reflectance near the centerof the 530 nm valley feature (Eqs 3cndash3d) the spectral range

Figure 4 A set of three spectral components (top colored) and corresponding temporal loadings (bottom colored) from ICA decompositionThe first spectral component is overlaid with the chlorophyll Jacobian ( partminuslog(R)

partCchl dashedndashdotted) and the second spectral component is

overlaid with the carotenoid Jacobian ( partminuslog(R)partCcar

dotted) The third spectral component is overlaid on the annual mean shape of transformedreflectance spectra Temporal loadings are overlaid with GPPmax (grey line) The axis of Jacobians is not shown because its magnitude isarbitrary here The vertical dashed line divides the observations from DOY for the years 2017 and 2018

that is most sensitive to the change of carotenoid contentso that they matched changes in GPPmax very well PRI wasthe smoothest throughout the year without any significantfluctuations within the growing season as opposite to whatwas observed in GPPmax which co-varied with Tair and VPD(Fig S3a and b) This performance is intriguing given thatPRI was originally developed to track short-term variationsin LUE (Gamon et al 1992) such as day-to-day and sub-seasonal scales

GCC (Fig 5c) also correlated well with GPPmax but lessthan CCI and PRI As can be seen in Fig S1 the peak ofthe green channel used for GCC is close to the carotenoidJacobian peak while the red channel feature covers a partof the chlorophyll Jacobian feature This explained the sensi-tivity of the GCC to changes in both carotenoid content andchlorophyll The bands used in GCC are broader than theones used by PRI and CCI however it still captured thesevariations and can be computed using RGB imagery Gen-tine and Alemohammad (2018) found that the green bandhelps to reconstruct variations in SIF using reflectances fromMODIS While they speculated that most variations in SIFare related to variations in PAR middot fPAR (Gentine and Alemo-hammad 2018) we suggest here that the green band indeedcaptures variations in LUE as well

NDVI (Fig 5e) and NIRv (Fig 5f) did not show an obvi-ous seasonal variability

Similar to the ICA components all VIs were quite noisyduring dormancy especially prior to DOY 50 This noisemay be due to snow because we only removed the reflectancewhen the canopy was snow covered Scattered photons pos-sibly still reached the telescope when there was snow on theground which is true for our study site as snowpack exists inwinter (Bowling et al 2018)

33 PLSR coefficients of reflectance with GPPmax andpigment measurements

The spectral shape of the PLSR coefficient with GPPmaxhighlighted a peak (centering at 532 nm) near that of thecarotenoid Jacobian with the same valley-trough feature ob-served near the second peak of the chlorophyll Jacobian(Fig 6a)

The reconstructed GPPmax captured the onset and cessa-tion of growth while the day-to-day noise in reflectance dur-ing dormancy propagated to the reconstructed GPPmax (minus2to 5 micromol mminus2 sminus1) During the growing season the day-to-day variations in GPPmax were not captured by any of themethods using pigment absorption features (Figs 5andashc and

Figure 5 Magenta points are time series of VIs (a) CCI (b) PRI (c) GCC (d) relative SIF (e) NDVI (f) NIRv The grey points in thebackground show GPPmax The Pearson r2 values of regressing VIs and GPPmax are noted in each plot The p values of all correlations inthis figure are less than 0005 The vertical dashed line divides the observations from DOY for the years 2017 and 2018

6b) which indicates those variations were not related to pig-ment content but rather changes in environmental conditionsthat lead to day-to-day changes in photosynthesis (Fig S3a)Overall the observed GPPmax was significantly correlatedwith the PLSR reconstruction (Pearson r2

= 087) but verysimilar compared to CCI and PRI

A similar PLSR model of reflectance but with pigmentmeasurements (Fig 7) showed a direct link between pig-ment contents and reflectance It can be seen that the PLSRcoefficients of reflectance are very similar irrespective ofthe target variable They feature a valley near the peak ofthe carotenoid Jacobian and a valley-trough feature nearthe peak at the longer wavelength of the chlorophyll Jaco-bian This spectral shape is also very similar to the secondICA spectral component and PLSR coefficients of GPPmaxV+A+Z chl car and car were all nicely reconstructed byusing the PLSR coefficients and reflectance (Fig 7b) Thereconstructed V+A+Z car and chlcar are correlated withthe measured ones with Pearson r2 values of 084 071 and093 respectively

The second ICA component and PLSR empiricallyshowed the seasonality of reflectance using two differentempirical frameworks ICA only used the reflectance whilethe PLSR model accounts for variations in both reflectanceand GPPmax or pigment content Yet both ICA and PLSRagreed on similar spectral features that co-varied seasonallywith GPPmax This indicates that the resulting spectral fea-tures were primarily responsible for representing this sea-sonal cycle The overlap of these features with the chloro-phyllcarotenoid absorption features showed that the season-ality of GPPmax was related to variation in pigment contentat the canopy scale which was directly validated with a sim-ilar PLSR coefficient of reflectance and pigment contentsThese results are consistent with leaf-level measurements ofa higher ratio of chlorophyll to carotenoid content during thegrowing season in this forest (Fig 7)

The highlighted spectral feature around 530 nm from ICAand PLSR closely overlaps with one of the bands used inCCI PRI and GCC (Eqs 3andash3e) which provides a justifica-tion that these VIs can remarkably capture the LUE season-ality The comparable Pearson r2 values of PLSR CCI and

Figure 6 (a) The PLSR coefficient of reflectance with GPPmax is the blue line The overlaid dashedndashdotted and dotted lines are chlorophylland carotenoid Jacobians respectively The overlaid orange solid line is the second ICA spectral component which was scaled to fit to theplot without a y axis (b) The reconstructed GPPmax (blue) by PLSR is overlaid with the observed GPPmax (red) The vertical dashed linedivides the observations from DOY for the years 2017 and 2018

PRI with GPPmax suggest the pigment-driven seasonal cycleof GPPmax is sufficiently represented by CCI and PRI Thespectral feature around the red edge does not make PLSRsignificantly more correlated with GPPmax than CCI or PRIwhich implies the feature is not driven by total chlorophyllor carotenoid contents

34 Process-based estimation of pigment content

PROSAIL inversion results further supported the link be-tween canopy reflectance pigment contents and GPPmaxFigure 8 shows a continuous time series of Cchl Ccar antho-cyanin content (Cant) and Cchl

Ccarderived from the PROSAIL

canopy RTM inversion model Examples of simulated andmeasured reflectance spectra shown are in Fig C1 Antho-cyanins are another type of photoprotective pigment (Pietriniet al 2002 Lee and Gould 2002 Gould 2004) that protectsthe plants from high light intensity (Hughes 2011) The pig-ment inversions closely matched the seasonality of GPPmaxCchlCcar

showed the greatest sensitivity in capturing the seasonalcycle with the strongest correlation to leaf level measure-ments (Fig 8c) The inverted Cchl had the weakest empiri-cal relationship with the measured one (Fig 8a right panel)Apparently some of the inversion errors of individual Ccarand Cchl contents canceled out in the ratio making the ratiomore stable Cant performed similarly to Ccar since they bothare photoprotective and the anthocyanins absorb at 550 nm

(Sims and Gamon 2002) which is close to the center ofcarotenoid absorption feature Even though we lacked fieldmeasurements of anthocyanins to validate anthocyanin re-trievals the inversions showed that more than just carotenoidcontent can be obtained from full-spectral inversions

Strictly speaking the complex canopy structure of ever-greens makes the application of 1D canopy RTMs such asPROSAIL difficult (Jacquemoud et al 2009 Zarco-Tejadaet al 2019) Yet Moorthy et al (2008) Ali et al (2016)and Zarco-Tejada et al (2019) reasonably discussed the pig-ment retrieval in conifer forests with careful applications Inour study the reflectance was collected from needles with avery small FOV and our study site has a very stable canopystructure throughout the year (Burns et al 2016) Thus theinversion results are meaningful for discussing the seasonal-ity of pigment contents In the future radiative transfer mod-els that properly describe conifer forests such as LIBERTY(Dawson et al 1998) could be used

35 Comparison across methods

Although decomposing the hyperspectral canopy reflectanceand using relative SIF (Fig 5d) both successfully trackedthe seasonal cycle of evergreen LUE they underlie differentde-excitation processes During the growing season environ-mental conditions primarily drove the day-to-day variationsin GPPmax Relative SIF responded to such environmental

Figure 7 (a) PLSR coefficients of reflectance and three pigment measurements The overlaid dashedndashdotted and dotted lines are chlorophylland carotenoid Jacobians respectively The overlaid solid grey line is the second ICA spectral component which is scaled to fit to the plotwithout a y axis (b) The reconstructed pigment measurements (blue) by PLSR are overlaid with the measured mean pigment measurements(red) The error bar is 1 standard deviation of the measurements The vertical dashed line divides the observations from DOY for the years2017 and 2018

stresses (van der Tol et al 2014) so that it appeared to tracksub-seasonal variations better than reflectance particularlyduring the growing season (Fig S5f) Yet reflectance de-compositions and VIs were less sensitive to such day-to-dayvariations (Figs 6 S3b)

There was also some variability between reflectance-basedmethods and relative SIF during the transition periods be-tween the growing season and dormancy We focused onthe growing season onset since the reflectance measurementswere not available during the cessation period The onset(DOY 60 to 166) described by all the methods mentionedabove as well as the relative SIF are compared in Fig 9 us-ing a sigmoid fit to available data (Fig D1) The observedGPPmax had the most rapid yet latest growing onset Themethods and VIs derived from or related to the pigment con-tents increased earlier than GPPmax ndash such as the ICA com-ponent PLSR coefficient PROSAIL Cchl

Ccar and CCI How-

ever they built up slowly to reach the maximum which sug-

gests that reduction of the carotenoid content is a slowerprocess than the recovery of LUE Reflectance-based VIs(Fig 5) and decomposing methods (Figs 4 and 8b c) hada slower growing season onset than GPPmax as found inBowling et al (2018) as well On the other hand relativeSIF started the onset at almost the same time as the GPPmaxand it quickly reached the maximum Therefore using bothSIF and reflectance to constrain the LUE prediction (van derTol et al 2014) can further improve the prediction accuracy

4 Conclusion and future work

In this study we analyzed seasonal co-variation in GPP andthe spectrally resolved visible and near-infrared reflectancesignal as well as several commonly used VIs The mainspectral feature centered around 530 nm is most importantfor inferring the seasonal cycle of reflectance (400ndash900 nm)

Figure 8 The left panels are the estimations of (a) Cchl (b) Ccar Cant and (c) CchlCcar

from the PROSAIL overlaid with the GPPmax Wenormalized two metrics because they report the pigment contents in different units The vertical dashed line divides the observations fromDOY for the years 2017 and 2018 The plots on the right compare the pigment contents from leaf-level measurements and using PROSAIL(a) chl vs Cchl (b) car vs Ccar and (c) chl car vs Cchl

Ccar The correlations are statistically significant except Cchl

Figure 9 Temporal evolution of the growing season onset usingsigmoid fits (scaled) of PLSR ICA CCI chlorophyll-to-carotenoidratio and relative SIF

and LUE which corresponds to changes in carotenoid con-tent This explains why CCI PRI and GCC track GPP sea-sonality so well as most variations are driven by carotenoidpool changes Our analysis included RTM simulation andin situ pigment measurements throughout the season con-firming the link between reflectanceVIs and pigment con-

tents The comparison of reflectanceVIs and relative SIF re-veals differences in the timing of the growing season onsetpigment changes and SIF indicating the potential of usingboth reflectance and SIF to track the seasonality of photo-synthesis However the close correspondence between bothSIF and reflectance suggests that hyperspectral reflectancealone provides mechanistic evidence for a robust approachto track photosynthetic phenology of evergreen systems Be-cause seasonal variation in pigment concentration plays astrong role in regulating the seasonality of photosynthesisin evergreen systems our work will help to inform futurestudies using hyperspectral reflectance to achieve accuratemonitoring of these ecosystems While indices like PRI andCCI are performing sufficiently as our methods which use thefull-spectrum analysis at the canopy scale the application ofthe full spectrum might be more robust for space-based mea-surements In addition we found seasonal changes of canopyreflectance near the red-edge region which could be relatedto leaf structural changes or changes in chlorophyll a and bOur PLSR coefficients are good references for customizingVIs to infer the photosynthetic seasonality in evergreen forestwhen there are restrictions to use the specific bands from cur-rently existing VIs (such as PRI and CCI) While our currentstudy is limited to a subalpine evergreen forest and canopy-scale measurements applications to other regions vegetationtypes and observational platforms will be a focus for futureresearch

Appendix A Bidirectional reflectance effect

A1 NDVI and NIRv

The impact of geometry and small FOV is relatively negligi-ble First our method only used the scans when FOV is onthe needles by setting a NDVI threshold Second we plottedthe NDVI and NIRv against the solar geometry at each in-dividual tree target throughout a year NDVI and NIRv arequite homogeneous regardless of various solar geometries asshown in the following figures

Figure A1 NDVI and NIRv of all scans targeting a pine at different solar azimuth angles and solar zenith angles throughout a year

Figure A2 NDVI and NIRv of all scans targeting a fir at different solar azimuth angles and solar zenith angles throughout a year

Figure A3 NDVI and NIRv of all scans targeting a spruce at different solar azimuth angles and solar zenith angles throughout a year

A2 PLSR on phase angle and reflectance

We did a PLSR analysis on individual measurements ofphase angle and reflectance for 3 summer days (1 to3 July 2017) The results are the same from other sampledays Indeed the reflectance has different sensitivities to thephase angle However the poor correlation of PLSR recon-structed phase angle and the measured one suggests the vari-ations in phase angle should not be the critical factor for thechange in reflectance In our study we primarily removedthe bidirectional impact by averaging all the individual re-flectance that was measured at different solar geometry andviewing geometry

Figure A4 PLSR analysis on phase angle and reflectance

Appendix B Detailed processes on integratingdaily-averaged canopy reflectance

First we chose scans targeting vegetation only by requiringan NDVI greater than 06 Second it is important to ensurethat the solar irradiation did not change between the acquisi-tion of the solar irradiance and the reflected radiance mea-surement To achieve this we matched the timestamps ofa PAR sensor (LI-COR LI-190SA LI-COR EnvironmentalLincoln Nebraska US) to the timestamps of PhotoSpec andwe compared the PAR value from the PAR sensor during thePhotoSpec irradiance acquisition with PAR during the actualtarget scan of the reflected radiance from vegetation We onlyused the scans when the ratio of the two was 10plusmn 01 en-suring stable PAR conditions Third in order to avoid un-stable PAR because of clouds (Dye 2004) we also removedcloudy scenes by requiring PAR to be at least 60 of a the-oretical maximum driven by solar geometry (Fig B1) Fur-ther only data when PAR was greater than 100 micromol mminus2 sminus1

were considered to eliminate the impact of low solar angleson reflectance data The VIs shown in Fig 5 were extractedin the same fashion as above

Figure B1 The distribution of the ratio of the measured PAR to the PAR at theoretical maximum from all individual scans

Appendix C PROSAIL fits

We used the following range constraints for variables in-cluded in the state vector of the PROSAIL inversion

ndash Leaf mesophyll structure (N ) 09ndash11

ndash Chlorophyll content (Cchl) 0ndash120 micromol cmminus2

ndash Carotenoid content (Ccar) 0ndash70 micromol cmminus2

ndash Anthocyanin content (Cant) 0ndash10 micromol cmminus2

ndash Brown pigments (Cbrown) 0ndash06

ndash Water content (Cw) 0ndash02 cm

ndash Dry matter content (Cm) 0ndash02 g cmminus2

ndash Xanthophyll cycle status (Cx) 0ndash1

ndash Leaf area index (LAI) fixed to 42

Figure C1 The observed and fitted reflectance spectra at low (a) and high (b) CchlCcar

Appendix D Sigmoid fit

The sigmoid equation is

y = b+aminus b

1+ exp(dminusxc

) In this form a and b represent the maximum and minimumvalues of the sigmoid fit And d is the half maximum of thefit We obtained the optimal values of these parameters

ProofIf xrarr+infin exp

(dminusxc

)rarr 0 So

limxrarr+infin

If xrarrminusinfin exp(dminusxc

)rarr+infin So

limxrarrminusinfin

Figure D1 Individual sigmoid fits of the onset of growth from different methods and more VIs The fitted curve has been expressed asthe derivation as above The Pearson r2 and p values listed in each subplot were calculated from the correlation of observed and fittedvariables The residual was calculated as the average L2 norm of the difference between observed (y) and fitted variables (y) normalizedby the observation ie 1

sumi(yminusyy )2 The fittings are overall good Because the ICA loading lacks a clear sigmoid shape ICA has a larger

residual

The first derivative of y is

aminus b

(1+ exp(dminusxc

exp(d minus x

At the half maximum point (x = xhalf) y = a+b2 There-

fore we need to solve

aminus b

1+ exp(dminusxhalfc

) Hence xhalf = d

Code and data availability Our data presented in this paper areprovided at httpsdoiorg1022002D11597 (Cheng et al 2020)and httpsdoiorg1022002d11231 (Magney et al 2019b) ThePROSAIL model in the Julia programming language used inour study can be obtained from httpsgithubcomCliMALand(CliMA 2020)

Supplement The supplement related to this article is available on-line at httpsdoiorg105194bg-17-4523-2020-supplement

Author contributions RC TSM DD and CF designed researchRC TSM DD DRB BAL SPB PDB KG SL ADR JS and CFperformed data analyses RC TSM DD DRB BAL SPB PDBKG SL ADR JS and CF wrote the paper

Competing interests The authors declare that they have no conflictof interest

Acknowledgements We thank the sponsors from the Caltech Grad-uate First-year Fellowship for Rui Cheng and NASA Carbon Moni-toring Systems program for David R Bowling The US-NR1 site isa part of the AmeriFlux Management Project (AMP) The NationalCenter for Atmospheric Research (NCAR) is sponsored by the NSF

Financial support This research has been supported by the NASACarbon Monitoring Systems program (grant no NNX16AP33G)and the DOE Office of Science through the AmeriFlux Manage-ment Project (AMP) at Lawrence Berkeley National Laboratory(grant no 7094866)

Review statement This paper was edited by Soumlnke Zaehle and re-viewed by Shari Van Wittenberghe and one anonymous referee

References

Adams W W and Demmig-Adams B Carotenoid composi-tion and down regulation of photosystem II in three coniferspecies during the winter Physiol Plantarum 92 451ndash458httpsdoiorg101111j1399-30541994tb08835x 1994

Ahlstroumlm A Schurgers G Arneth A and Smith B Robust-ness and uncertainty in terrestrial ecosystem carbon responseto CMIP5 climate change projections Environ Res Lett 7044008 httpsdoiorg1010881748-932674044008 2012

Ali A M Darvishzadeh R Skidmore A K van Duren I Hei-den U and Heurich M Estimating leaf functional traits by in-version of PROSPECT Assessing leaf dry matter content andspecific leaf area in mixed mountainous forest Int J Appl EarthObs 45 66ndash76 2016

Asner G P Martin R E Knapp D E Tupayachi R An-derson C Carranza L Martinez P Houcheime M SincaF and Weiss P Spectroscopy of canopy chemicals in hu-

mid tropical forests Remote Sens Environ 115 3587ndash3598httpsdoiorg101016JRSE201108020 2011

Badgley G Field C B and Berry J A Canopy near-infraredreflectance and terrestrial photosynthesis Sci Adv 3 e1602244httpsdoiorg101126sciadv1602244 2017

Baldocchi D Falge E Gu L Olson R Hollinger D RunningS Anthoni P Bernhofer C Davis K Evans R Fuentes JGoldstein A Katul G Law B Lee X Malhi Y Meyers TMunger W Oechel W Paw U K T Pilegaard K Schmid HP Valentini R Verma S Vesala T Wilson K and Wofsy SFLUXNET A new tool to study the temporal and spatial variabil-ity of ecosystem-scale carbon dioxide water vapor and energyflux densities B Am Meteorol Soc 82 2415ndash2434 2001

Barnes M L Breshears D D Law D J van LeeuwenW J D Monson R K Fojtik A C Barron-GaffordG A and Moore D J P Beyond greenness Detect-ing temporal changes in photosynthetic capacity with hy-perspectral reflectance data PLOS ONE 12 e0189539httpsdoiorg101371journalpone0189539 2017

Blanken P D Monson R K Burns S P Bowling D R andTurnipseed A A Data and information for the AmeriFlux US-NR1 Niwot Ridge Subalpine Forest (LTER NWT1) Site Amer-iFlux Management Project Berkeley CA Lawrence BerkeleyNational Laboratory httpsdoiorg1017190AMF12460882019

Bowling D and Logan B Carbon Monitoring Sys-tem (CMS)Conifer Needle Pigment Composition Ni-wot Ridge Colorado USA 2017ndash2018 0021358 MBhttpsdoiorg103334ORNLDAAC1723 2019

Bowling D R Logan B A Hufkens K AubrechtD M Richardson A D Burns S P Anderegg W RBlanken P D and Eiriksson D P Limitations to win-ter and spring photosynthesis of a Rocky Mountainsubalpine forest Agr Forest Meteorol 252 241ndash255httpsdoiorg101016JAGRFORMET201801025 2018

Burns S P Blanken P D Turnipseed A A Hu J and Mon-son R K The influence of warm-season precipitation on thediel cycle of the surface energy balance and carbon dioxide at aColorado subalpine forest site Biogeosciences 12 7349ndash7377httpsdoiorg105194bg-12-7349-2015 2015

Burns S P Maclean G D Blanken P D Oncley S P Sem-mer S R and Monson R K The Niwot Ridge Subalpine For-est US-NR1 AmeriFlux site ndash Part 1 Data acquisition and siterecord-keeping Geosci Instrum Method Data Syst 5 451ndash471 httpsdoiorg105194gi-5-451-2016 2016

Cheng R Frankenberg C Magney T Grossmann K Bowl-ing D Burns S Stutz J and Blanken P Hyperspectral re-flectance at Niwot Ridge Colorado (Version 10) [Data set] Cal-techDATA httpsdoiorg1022002D11597 2020

Ciais P Sabine C Bala G Bopp L Brovkin V Canadell JChhabra R DeFries R Galloway J Heimann M Jones CLe Queacutereacute C Myneni R Piao S and Thornton P Carbonand Other Biogeochemical Cycles in Climate Change 2013The Physical Science Basis Contribution of Working Group Ito the Fifth Assessment Report of the Intergovernmental Panelon Climate Change edited by Stock T Qin D Plattner G-K Tignor M Allen S Boschung J Nauels A Xia Y BexV and Midgley P chap 6 465ndash570 Cambridge University

Press Cambridge United Kingdom and New York NY USAhttpsdoiorg101017CBO9781107415324015 2013

CliMA Land GitHub available at httpsgithubcomCliMALand last access 6 February 2020

Comon P Independent component analysis a new concept Sig-nal Processing 36 287ndash314 1994

Dawson T P Curran P J and Plummer S E LIBERTYndashModeling the effects of leaf biochemical concentration on re-flectance spectra Remote Sens Environ 65 50ndash60 1998

Dechant B Cuntz M Vohland M Schulz E and Doktor DEstimation of photosynthesis traits from leaf reflectance spectracorrelation to nitrogen content as the dominant mechanism Re-mote Sens Environ 196 279ndash292 2017

Dechant B Ryu Y and Kang M Making full use of hyperspec-tral data for gross primary productivity estimation with multi-variate regression Mechanistic insights from observations andprocess-based simulations Remote Sens Environ 234 111435httpsdoiorg101016jrse2019111435 2019

Demmig-Adams B and Adams W W The role of xanthophyllcycle carotenoids in the protection of photosynthesis Trend PlantSci 1 21ndash26 httpsdoiorg101016S1360-1385(96)80019-71996

de Tomaacutes Mariacuten S Novaacutek M Klancnik K and Gaberšcik ASpectral signatures of conifer needles mainly depend on theirphysical traits Pol J Ecol 64 1ndash14 2016

DuBois S Desai A R Singh A Serbin S P Goulden M LBaldocchi D D Ma S Oechel W C Wharton S KrugerE L and Townsend P A Using imaging spectroscopy to de-tect variation in terrestrial ecosystem productivity across a water-stressed landscape Ecol Appl 28 1313ndash1324 2018

Dutta D Schimel D S Sun Y van der Tol C and FrankenbergC Optimal inverse estimation of ecosystem parameters from ob-servations of carbon and energy fluxes Biogeosciences 16 77ndash103 httpsdoiorg105194bg-16-77-2019 2019

Dye D G Spectral composition and quanta-to-energy ra-tio of diffuse photosynthetically active radiation under di-verse cloud conditions J Geophys Res 109 D10203httpsdoiorg1010292003JD004251 2004

Farquhar G D von Caemmerer S and Berry J AA biochemical model of photosynthetic CO2 assim-ilation in leaves of C3 species Planta 149 78ndash90httpsdoiorg101007BF00386231 1980

Feret J-B Franccedilois C Asner G P Gitelson A AMartin R E Bidel L P Ustin S L le MaireG and Jacquemoud S PROSPECT-4 and 5 Advancesin the leaf optical properties model separating photosyn-thetic pigments Remote Sens Environ 112 3030ndash3043httpsdoiorg101016JRSE200802012 2008

Feacuteret J B Gitelson A A Noble S D and JacquemoudS PROSPECT-D Towards modeling leaf optical propertiesthrough a complete lifecycle Remote Sens Environ 193 204ndash215 httpsdoiorg101016jrse201703004 2017

Feacuteret J-B Le Maire G Jay S Berveiller D Bendoula RHmimina G Cheraiet A Oliveira J Ponzoni F SolankiT de Boissieu F Chave J Nouvellon Y Porcar-Castell AProisy C Soudani K Gastellu-Etchegorry J-P and Lefegravevre-Fonollosa M-J Estimating leaf mass per area and equivalentwater thickness based on leaf optical properties Potential and

limitations of physical modeling and machine learning RemoteSens Environ 231 110959 2019

Gamon J Pentildeuelas J and Field C A narrow-waveband spec-tral index that tracks diurnal changes in photosynthetic efficiencyRemote Sens Environ 41 35ndash44 httpsdoiorg1010160034-4257(92)90059-S 1992

Gamon J Rahman A Dungan J Schildhauer M and Huemm-rich K Spectral Network (SpecNet) ndash What is it andwhy do we need it Remote Sens Environ 103 227ndash235httpsdoiorg101016JRSE200604003 2006

Gamon J A Serrano L and Surfus J S The pho-tochemical reflectance index an optical indicator of pho-tosynthetic radiation use efficiency across species func-tional types and nutrient levels Oecologia 112 492ndash501httpsdoiorg101007s004420050337 1997

Gamon J A Huemmrich K F Wong C Y S Ensminger IGarrity S Hollinger D Y Noormets A and Pentildeuelas J Aremotely sensed pigment index reveals photosynthetic phenol-ogy in evergreen conifers P Natl Acad Sci USA 113 13087ndash13092 httpsdoiorg101073pnas1606162113 2016

Garbulsky M F Pentildeuelas J Gamon J Inoue Y and Filella IThe photochemical reflectance index (PRI) and the remote sens-ing of leaf canopy and ecosystem radiation use efficiencies Areview and meta-analysis Remote Sens Environ 115 281ndash2972011

Geladi P and Kowalski B R Partial least-squares re-gression a tutorial Anal Chim Acta 185 1ndash17httpsdoiorg1010160003-2670(86)80028-9 1986

Gentine P and Alemohammad S H Reconstructed Solar-InducedFluorescence A Machine Learning Vegetation Product Basedon MODIS Surface Reflectance to Reproduce GOME-2 Solar-Induced Fluorescence Geophys Res Lett 45 3136ndash3146httpsdoiorg1010022017GL076294 2018

Genty B Briantais J-M and Baker N R The relationship be-tween the quantum yield of photosynthetic electron transport andquenching of chlorophyll fluorescence Biochim Biophys Acta990 87ndash92 1989

Glenn E Huete A Nagler P Nelson S Glenn E P HueteA R Nagler P L and Nelson S G Relationship Be-tween Remotely-sensed Vegetation Indices Canopy Attributesand Plant Physiological Processes What Vegetation Indices Canand Cannot Tell Us About the Landscape Sensors 8 2136ndash2160 httpsdoiorg103390s8042136 2008

Gould K S Naturersquos Swiss army knife the diverse protective rolesof anthocyanins in leaves BioMed Research International 2004314ndash320 2004

Goulden M L Munger J W Fan S-M Daube BC and Wofsy S C Measurements of carbon sequestra-tion by long-term eddy covariance methods and a criti-cal evaluation of accuracy Glob Change Biol 2 169ndash182httpsdoiorg101111j1365-24861996tb00070x 1996

Grossmann K Frankenberg C Magney T S Hurlock S CSeibt U and Stutz J PhotoSpec A new instrument to mea-sure spatially distributed red and far-red Solar-Induced Chloro-phyll Fluorescence Remote Sens Environ 216 311ndash327httpsdoiorg101016JRSE201807002 2018

Guanter L Zhang Y Jung M Joiner J Voigt M BerryJ A Frankenberg C Huete A R Zarco-Tejada P Lee J-E Moran M S Ponce-Campos G Beer C Camps-Valls G

Buchmann N Gianelle D Klumpp K Cescatti A BakerJ M and Griffis T J Global and time-resolved monitoring ofcrop photosynthesis with chlorophyll fluorescence P Natl AcadSci USA 111 E1327ndashE1333 2014

Hall F G Hilker T Coops N C Lyapustin A Huemm-rich K F Middleton E Margolis H Drolet G andBlack T A Multi-angle remote sensing of forest lightuse efficiency by observing PRI variation with canopyshadow fraction Remote Sens Environ 112 3201ndash3211httpsdoiorg101016JRSE200803015 2008

Harbinson J Modeling the protection of photosyn-thesis P Natl Acad Sci USA 109 15533ndash15534httpsdoiorg101073pnas1213195109 2012

Hilker T Coops N C Hall F G Nichol C J Lya-pustin A Black T A Wulder M A Leuning R BarrA Hollinger D Y Munger B and Tucker C J Infer-ring terrestrial photosynthetic light use efficiency of temper-ate ecosystems from space J Geophys Res 116 G03014httpsdoiorg1010292011JG001692 2011a

Hilker T Gitelson A Coops N C Hall F G and BlackT A Tracking plant physiological properties from multi-angular tower-based remote sensing Oecologia 165 865ndash876httpsdoiorg101007s00442-010-1901-0 2011b

Horler D N H Dockray M and Barber J The red edgeof plant leaf reflectance Int J Remote Sens 4 273ndash288httpsdoiorg10108001431168308948546 1983

Huemmrich K F Campbell P K E Gao B-C Flana-gan L B and Goulden M ISS as a Platform for Opti-cal Remote Sensing of Ecosystem Carbon Fluxes A CaseStudy Using HICO IEEE J Sel Top Appl 10 4360ndash4375httpsdoiorg101109JSTARS20172725825 2017

Huemmrich K F Campbell P Landis D and Middleton EDeveloping a common globally applicable method for opticalremote sensing of ecosystem light use efficiency Remote SensEnviron 230 111190 2019

Huete A Liu H Batchily K and Van Leeuwen W A compar-ison of vegetation indices over a global set of TM images forEOS-MODIS Remote Sens Environ 59 440ndash451 1997

Hughes N M Winter leaf reddening in ldquoevergreenrdquo species NewPhytol 190 573ndash581 2011

Hyvaumlrinen A and Oja E Independent component analysisalgorithms and applications Neural Networks 13 411ndash430httpsdoiorg101016S0893-6080(00)00026-5 2000

Jacquemoud S and Baret F PROSPECT A model of leaf opticalproperties spectra Remote Sens Environ 34 75ndash91 1990

Jacquemoud S Baret F Andrieu B Danson F and JaggardK Extraction of vegetation biophysical parameters by inver-sion of the PROSPECT+SAIL models on sugar beet canopyreflectance data Application to TM and AVIRIS sensors Re-mote Sens Environ 52 163ndash172 httpsdoiorg1010160034-4257(95)00018-V 1995

Jacquemoud S Verhoef W Baret F Bacour C Zarco-TejadaP J Asner G P Franccedilois C and Ustin S L PROSPECT+SAIL models A review of use for vegetation characterizationRemote Sens Environ 113 S56ndashS66 2009

Krause G H and Weis E CHLOROPHYLL FLU-ORESCENCE AND PHOTOSYNTHESIS TheBasics Annu Rev Plant Phys 42 313ndash349httpsdoiorg101146annurevpp42060191001525 1991

Krinner G Viovy N de Noblet-Ducoudreacute N Ogeacutee J PolcherJ Friedlingstein P Ciais P Sitch S and Prentice I CA dynamic global vegetation model for studies of the cou-pled atmosphere-biosphere system Global Biogeochem Cy 19GB1015 httpsdoiorg1010292003GB002199 2005

Lee D W and Gould K S Why leaves turn red pigments calledanthocyanins probably protect leaves from light damage by directshielding and by scavenging free radicals Am Sci 90 524ndash5312002

Leuning R A critical appraisal of a combined stomatal-photosynthesis model for C3 plants Plant Cell Environ 18 339ndash355 1995

Liu H Q and Huete A A feedback based modification of theNDVI to minimize canopy background and atmospheric noiseIEEE T Geosci Remote Sens 33 457ndash465 1995

Magney T S Bowling D R Logan B A GrossmanK Stutz J Blanken P Burns S P Cheng R Gar-cia M A Koumlhler P Lopez S Parazoo N Raczka BSchimel D and Frankenberg C Mechanistic evidence fortracking the seasonality of photosynthesis with solar inducedfluorescence P Natl Acad Sci USA 116 11640ndash11645httpsdoiorg101073pnas1900278116 2019a

Magney T Frankenberg C Grossmann K Bowling D LoganB Burns S and Stutz J Canopy and needle scale fluorescencedata from Niwot Ridge Colorado 2017ndash2018 (Version 11) [Dataset] CaltechDATA httpsdoiorg1022002d11231 2019b

Matthes J H Knox S H Sturtevant C Sonnentag O Ver-faillie J and Baldocchi D Predicting landscape-scale CO2flux at a pasture and rice paddy with long-term hyperspectralcanopy reflectance measurements Biogeosciences 12 4577ndash4594 httpsdoiorg105194bg-12-4577-2015 2015

Meacham-Hensold K Montes C M Wu J Guan K FuP Ainsworth E A Pederson T Moore C E BrownK L Raines C and Bernacchi C J High-throughput fieldphenotyping using hyperspectral reflectance and partial leastsquares regression (PLSR) reveals genetic modifications tophotosynthetic capacity Remote Sens Environ 231 111176httpsdoiorg101016JRSE201904029 2019

Middleton E Huemmrich K Landis D Black T Barr Aand McCaughey J Photosynthetic efficiency of northern forestecosystems using a MODIS-derived Photochemical ReflectanceIndex (PRI) Remote Sens Environ 187 345ndash366 2016

Monson R K Turnipseed A A Sparks J P Harley P C Scott-Denton L E Sparks K and Huxman T E Carbon sequestra-tion in a high-elevation subalpine forest Glob Change Biol8 459ndash478 httpsdoiorg101046j1365-2486200200480x2002

Monson R K Sparks J P Rosenstiel T N Scott-Denton L EHuxman T E Harley P C Turnipseed A A Burns S PBacklund B and Hu J Climatic influences on net ecosys-tem CO2 exchange during the transition from wintertime carbonsource to springtime carbon sink in a high-elevation subalpineforest Oecologia 146 130ndash147 2005

Monteith J L Solar Radiation and Productivity in TropicalEcosystems J Appl Ecol 9 747ndash766 1972

Monteith J L and Moss C J Climate and the Efficiency of CropProduction in Britain and Discussion Philos T R Soc B 281277ndash294 1977

Moorthy I Miller J R and Noland T L Estimating chlorophyllconcentration in conifer needles with hyperspectral data An as-sessment at the needle and canopy level Remote Sens Environ112 2824ndash2838 2008

Pietrini F Iannelli M and Massacci A Anthocyanin accumula-tion in the illuminated surface of maize leaves enhances protec-tion from photo-inhibitory risks at low temperature without fur-ther limitation to photosynthesis Plant Cell Environ 25 1251ndash1259 2002

Porcar-Castell A Tyystjaumlrvi E Atherton J Van der Tol CFlexas J Pfuumlndel E E Moreno J Frankenberg C and BerryJ A Linking chlorophyll a fluorescence to photosynthesis forremote sensing applications mechanisms and challenges J ExpBot 65 4065ndash4095 2014

Porcar-Castell A Mac Arthur A Rossini M Eklundh LPacheco-Labrador J Anderson K Balzarolo M Martiacuten MP Jin H Tomelleri E Cerasoli S Sakowska K HueniA Julitta T Nichol C J and Vescovo L EUROSPECat the interface between remote-sensing and ecosystem CO2flux measurements in Europe Biogeosciences 12 6103ndash6124httpsdoiorg105194bg-12-6103-2015 2015

Rautiainen M Lukeš P Homolova L Hovi A Pisek J andMottus M Spectral properties of coniferous forests A review ofin situ and laboratory measurements Remote Sensing 10 207httpsdoiorg103390rs10020207 2018

Reichstein M Falge E Baldocchi D Papale D AubinetM Berbigier P Bernhofer C Buchmann N GilmanovT Granier A Grunwald T Havrankova K Ilvesniemi HJanous D Knohl A Laurila T Lohila A Loustau D Mat-teucci G Meyers T Miglietta F Ourcival J-M PumpanenJ Rambal S Rotenberg E Sanz M Tenhunen J SeufertG Vaccari F Vesala T Yakir D and Valentini R Onthe separation of net ecosystem exchange into assimilation andecosystem respiration review and improved algorithm GlobChange Biol 11 1424ndash1439 httpsdoiorg101111j1365-24862005001002x 2005

Richardson A D Braswell B H Hollinger D Y Jenkins J Pand Ollinger S V Near-surface remote sensing of spatial andtemporal variation in canopy phenology Ecol Appl 19 1417ndash1428 2009

Richardson A D Hufkens K Milliman T Aubrecht D MChen M Gray J M Johnston M R Keenan T F Kloster-man S T Kosmala M Melaas E K Friedl M A and Frol-king S Tracking vegetation phenology across diverse NorthAmerican biomes using PhenoCam imagery Scientific Data 5180028 httpsdoiorg101038sdata201828 2018

Robinson N P Allred B W Smith W K Jones M OMoreno A Erickson T A Naugle D E and RunningS W Terrestrial primary production for the conterminousUnited States derived from Landsat 30 m and MODIS 250 mRemote Sensing in Ecology and Conservation 4 264ndash280httpsdoiorg101002rse274 2018

Rook D A The influence of growing temperature on photosynthe-sis and respiration of Pinus radiata seedlings New Zeal J Bot7 43ndash55 httpsdoiorg1010800028825X1969104291011969

Rouse Jr J Haas R Schell J and Deering D Paper A 20in Third Earth Resources Technology Satellite-1 SymposiumThe Proceedings of a Symposium Held by Goddard Space Flight

Center Washington DC 10ndash14 December 1973 Goddard SpaceFlight Center Vol 351 p 309 Scientific and Technical Informa-tion Office National Aeronautics and Space 1974

Running S W Nemani R R Heinsch F A Zhao MReeves M and Hashimoto H A Continuous Satellite-Derived Measure of Global Terrestrial Primary Produc-tion BioScience 54 547ndash560 httpsdoiorg1016410006-3568(2004)054[0547acsmog]20co2 2004

Schreiber U Schliwa U and Bilger W Continuous recording ofphotochemical and non-photochemical chlorophyll fluorescencequenching with a new type of modulation fluorometer Photo-synth Res 10 51ndash62 1986

Serbin S P Dillaway D N Kruger E L and TownsendP A Leaf optical properties reflect variation in photosyntheticmetabolism and its sensitivity to temperature J Exp Bot 63489ndash502 httpsdoiorg101093jxberr294 2012

Serbin S P Singh A McNeil B E Kingdon C C andTownsend P A Spectroscopic determination of leaf morpho-logical and biochemical traits for northern temperate and borealtree species Ecol Appl 24 1651ndash1669 2014

Serbin S P Singh A Desai A R Dubois S G JablonskiA D Kingdon C C Kruger E L and Townsend P A Re-motely estimating photosynthetic capacity and its response totemperature in vegetation canopies using imaging spectroscopyRemote Sens Environ 167 78ndash87 2015

Silva-Perez V Molero G Serbin S P Condon A GReynolds M P Furbank R T and Evans J R Hyper-spectral reflectance as a tool to measure biochemical andphysiological traits in wheat J Exp Bot 69 483ndash496httpsdoiorg101093jxberx421 2018

Sims D A and Gamon J A Relationships between leaf pigmentcontent and spectral reflectance across a wide range of speciesleaf structures and developmental stages Remote Sens Environ81 337ndash354 2002

Singh A Serbin S P McNeil B E Kingdon C C andTownsend P A Imaging spectroscopy algorithms for mappingcanopy foliar chemical and morphological traits and their uncer-tainties Ecol Appl 25 2180ndash2197 2015

Smith M-L Ollinger S V Martin M E Aber J D HallettR A and Goodale C L Direct estimation of aboveground for-est productivity through hyperspectral remote sensing of canopynitrogen Ecol Appl 12 1286ndash1302 2002

Sonnentag O Hufkens K Teshera-Sterne C Young A MFriedl M Braswell B H Milliman T OrsquoKeefe J andRichardson A D Digital repeat photography for phenologi-cal research in forest ecosystems Agr Forest Meteorol 152159ndash177 httpsdoiorg101016JAGRFORMET2011090092012

Stylinski C Gamon J and Oechel W Seasonal patternsof reflectance indices carotenoid pigments and photosynthe-sis of evergreen chaparral species Oecologia 131 366ndash374httpsdoiorg101007s00442-002-0905-9 2002

Tucker C J Red and photographic infrared linear combinationsfor monitoring vegetation Remote Sens Environ 8 127ndash150httpsdoiorg1010160034-4257(79)90013-0 1979

Ustin S L Roberts D A Gamon J A Asner G P and GreenR O Using imaging spectroscopy to study ecosystem processesand properties BioScience 54 523ndash534 2004

Ustin S L Gitelson A A Jacquemoud S Schaepman M As-ner G P Gamon J A and Zarco-Tejada P Retrieval of foliarinformation about plant pigment systems from high resolutionspectroscopy Remote Sens Environ 113 S67ndashS77 2009

van der Tol C Berry J A Campbell P K E andRascher U Models of fluorescence and photosynthesisfor interpreting measurements of solar-induced chlorophyllfluorescence J Geophys Res-Biogeo 119 2312ndash2327httpsdoiorg1010022014JG002713 2014

Verhoeven A S Adams W W and Demmig-Adams B Closerelationship between the state of the xanthophyll cycle pigmentsand photosystem II efficiency during recovery from winter stressPhysiol Plantarum 96 567ndash576 httpsdoiorg101111j1399-30541996tb00228x 1996

Vilfan N Van der Tol C Yang P Wyber R Malenovskyacute ZRobinson S A and Verhoef W Extending Fluspect to simulatexanthophyll driven leaf reflectance dynamics Remote Sens Env-iron 211 345ndash356 httpsdoiorg101016JRSE2018040122018

Wingate L Ogeacutee J Cremonese E Filippa G MizunumaT Migliavacca M Moisy C Wilkinson M Moureaux CWohlfahrt G Hammerle A Houmlrtnagl L Gimeno C Porcar-Castell A Galvagno M Nakaji T Morison J Kolle OKnohl A Kutsch W Kolari P Nikinmaa E Ibrom A Gie-len B Eugster W Balzarolo M Papale D Klumpp KKoumlstner B Gruumlnwald T Joffre R Ourcival J-M HellstromM Lindroth A George C Longdoz B Genty B LevulaJ Heinesch B Sprintsin M Yakir D Manise T GuyonD Ahrends H Plaza-Aguilar A Guan J H and Grace JInterpreting canopy development and physiology using a Euro-pean phenology camera network at flux sites Biogeosciences12 5995ndash6015 httpsdoiorg105194bg-12-5995-2015 2015

Wold S Ruhe A Wold H and Dunn III W The collinearityproblem in linear regression The partial least squares (PLS) ap-proach to generalized inverses SIAM J Sci Stat Comp 5 735ndash743 1984

Wong C Y and Gamon J A The photochemical reflectance indexprovides an optical indicator of spring photosynthetic activationin evergreen conifers New Phytol 206 196ndash208 2015a

Wong C Y S and Gamon J A Three causes of variation inthe photochemical reflectance index (PRI) in evergreen conifersNew Phytol 206 187ndash195 httpsdoiorg101111nph131592015b

Wong C Y DrsquoOdorico P Bhathena Y Arain M A and En-sminger I Carotenoid based vegetation indices for accuratemonitoring of the phenology of photosynthesis at the leaf-scalein deciduous and evergreen trees Remote Sens Environ 233111407 httpsdoiorg101016jrse2019111407 2019

Wong C Y DrsquoOdorico P Arain M A and Ensminger I Track-ing the phenology of photosynthesis using carotenoid-sensitiveand near-infrared reflectance vegetation indices in a temperateevergreen and mixed deciduous forest New Phytol 226 1682ndash1695 httpsdoiorg101111nph16479 2020

Woodgate W Suarez L van Gorsel E Cernusak LDempsey R Devilla R Held A Hill M and Nor-ton A tri-PRI A three band reflectance index track-ing dynamic photoprotective mechanisms in a mature eu-calypt forest Agr Forest Meteorol 272ndash273 187ndash201httpsdoiorg101016JAGRFORMET201903020 2019

Wutzler T Lucas-Moffat A Migliavacca M Knauer JSickel K Šigut L Menzer O and Reichstein MBasic and extensible post-processing of eddy covarianceflux data with REddyProc Biogeosciences 15 5015ndash5030httpsdoiorg105194bg-15-5015-2018 2018

Xiao X Hollinger D Aber J Goltz M Davidson E A ZhangQ and Moore B Satellite-based modeling of gross primaryproduction in an evergreen needleleaf forest Remote Sens En-viron 89 519ndash534 httpsdoiorg101016jrse2003110082004

Yang P and van der Tol C Linking canopy scattering of far-redsun-induced chlorophyll fluorescence with reflectance RemoteSens Environ 209 456ndash467 2018

Zarco-Tejada P Hornero A Beck P Kattenborn T Kempe-neers P and Hernaacutendez-Clemente R Chlorophyll content es-timation in an open-canopy conifer forest with Sentinel-2A andhyperspectral imagery in the context of forest decline RemoteSens Environ 223 320ndash335 2019

Zarter C R Adams W W Ebbert V Cuthbertson D JAdamska I and Demmig-Adams B Winter down-regulation of intrinsic photosynthetic capacity coupled withup-regulation of Elip-like proteins and persistent energy dis-sipation in a subalpine forest New Phytol 172 272ndash282httpsdoiorg101111j1469-8137200601815x 2006

Zhao M Heinsch F A Nemani R R and Running S W Im-provements of the MODIS terrestrial gross and net primary pro-duction global data set Remote Sens Environ 95 164ndash176httpsdoiorg101016JRSE200412011 2005

Zuromski L M Bowling D R Koumlhler P FrankenbergC Goulden M L Blanken P D and Lin J C Solar-Induced Fluorescence Detects Interannual Variation in GrossPrimary Production of Coniferous Forests in the West-ern United States Geophys Res Lett 45 7184ndash7193httpsdoiorg1010292018GL077906 2018

Abstract

Introduction

Material and methods

Study site

Continuous tower-based measurements of canopy reflectance

Eddy covariance measurements and LUE

Pigment measurements

Data-driven spectral decomposition

Process-based methods

Results and discussion

Seasonal cycle of GPPmax and environmental conditions

Seasonal cycle of reflectance

PLSR coefficients of reflectance with GPPmax and pigment measurements

Process-based estimation of pigment content

Comparison across methods

Conclusion and future work


Appendix A1 NDVI and NIRv

Appendix A2 PLSR on phase angle and reflectance

Appendix B Detailed processes on integrating daily-averaged canopy reflectance



Code and data availability

Supplement

Author contributions

Competing interests

Acknowledgements

Financial support

Review statement

References

lighting all spectral locations within 400ndash900 nm associatedwith GPP seasonality in evergreen forests

1 Introduction

Terrestrial gross primary production (GPP) the gross CO2uptake through photosynthesis is the largest uptake of at-mospheric CO2 (Ciais et al 2013) yet the uncertain-ties are large hampering our ability to monitor and pre-dict the response of the terrestrial biosphere to climatechange (Ahlstroumlm et al 2012) Hence accurately mappingGPP globally is critical In contrast to unevenly distributedground-level measurements such as Fluxnet (Baldocchi et al2001) satellites can infer GPP globally and uniformly Re-mote sensing techniques are based on the optical responseof vegetation to incoming sunlight which can track photo-synthesis via the absorption features of photosynthetic andphotoprotective pigments (Rouse et al 1974 Liu and Huete1995 Gamon et al 1992 2016) Progress is particularly im-portant for evergreen forests which can have large seasonaldynamics in photosynthesis but low variability in canopystructure and color However these promising techniquesstill lack a comprehensive evaluationvalidation using bothcontinuous in situ measurements and process-based simula-tions

GPP can be expressed as a function of photosyntheticallyactive radiation (PAR) the fraction of PAR absorbed by thecanopy (fPAR) and light-use efficiency (LUE)

GPP= PAR middot fPAR middotLUE (1)

with LUE representing the efficiency of plants to fix carbonusing absorbed light (Monteith 1972 Monteith and Moss1977) The accuracy of remote-sensing-derived GPP is lim-ited by the estimation of LUE which is more dynamic anddifficult to measure remotely than PAR and fPAR particu-larly in evergreen ecosystems There have been many stud-ies inferring the light absorbed by canopies (ie fPAR) fromvegetation indices (VIs) that estimate the ldquogreennessrdquo ofcanopies (Running et al 2004 Zhao et al 2005 Robinsonet al 2018 Glenn et al 2008) such as the normalized dif-ference vegetation index (NDVI Rouse et al 1974 Tucker1979) the enhanced vegetation index (EVI Liu and Huete1995 Huete et al 1997) and the near-infrared vegetation in-dex (NIRv Badgley et al 2017) Current GPP data productsderived from Eq (1) rely on the modulation of abiotic con-ditions to estimate LUE (Xiao et al 2004) LUE is derivedempirically by defining a general timing of dormancy for allevergreen forests with the same plant functional type (egKrinner et al 2005) or the same meteorological thresholds(eg Running et al 2004) However within the same cli-mate region or plant functional type forests are not identicalndash leading to uncertainties in estimated LUE (Stylinski et al

2002 Gamon et al 2016 Zuromski et al 2018) whichpropagate to the estimation of GPP

Because evergreen trees retain most of their needles andchlorophyll throughout the entire year (Bowling et al 2018)LUE in evergreens is regulated by needle biochemistry AsLUE falls with the onset of winter due to unfavorable envi-ronmental conditions and seasonal downregulation of pho-tosynthetic capacity evergreen needles quench excess ab-sorbed light via thermal energy dissipation that involves thexanthophyll cycle and other pigments (Adams and Demmig-Adams 1994 Demmig-Adams and Adams 1996 Verho-even et al 1996 Zarter et al 2006) Thermal energy dis-sipation is a primary de-excitation pathway measured bypulse-amplitude fluorescence as non-photochemical quench-ing (NPQ Schreiber et al 1986) At the same time a smallamount of radiation solar-induced fluorescence (SIF) viathe de-excitation of absorbed photons is emitted by photo-system II (Genty et al 1989 Krause and Weis 1991)

Some vegetation indices are sensitive to photoprotectivepigments (eg carotenoids) and can characterize the sea-sonality of evergreen LUE with some success For instancethe photochemical reflectance index (PRI Gamon et al1992 1997) and chlorophyllcarotenoid index (CCI Ga-mon et al 2016) both use wavelength regions that repre-sent carotenoid absorption features around 531 nm at the leaflevel (Wong et al 2019 Wong and Gamon 2015a b) andshow great promise for estimating photosynthetic seasonality(Hall et al 2008 Hilker et al 2011a) Due to the relativelyinvariant canopy structure in evergreen forests CCI and PRIhave been applied at the canopy level as well (Gamon et al2016 Garbulsky et al 2011 Middleton et al 2016) In addi-tion the green chromatic coordinate (GCC Richardson et al2009 2018 Sonnentag et al 2012) an index derived fromthe brightness levels of RGB canopy images is also capableof tracking the seasonality of evergreen GPP (Bowling et al2018) However the full potential of spectrally resolved re-flectance measurements to explore the photosynthetic phe-nology of evergreens has not been comprehensively exploredat the canopy scale The evaluation of pigment-driven spec-tral changes in evergreen forests over the course of a season isnecessary to determine where when and why certain wave-length regions could advance our mechanistic understand-ing of canopy photosynthetic and photoprotective pigmentsHowever this has not been done with both empirical andprocess-based methods using continuously measured canopyhyperspectral reflectance and in situ pigment samples

Here we used continuous measurements in both spec-tral space (full spectrum between 400 and 900 nm) and time(daily over an entire year) to evaluate the potential of hy-perspectral canopy reflectance for better understanding thesensitivity of VIs to pigment changes that regulate GPP inevergreen forests Continuous measurements of spectrally re-solved reflectance at the canopy scale have so far been sparseat evergreen forest sites (Gamon et al 2006 Hilker et al2011b Porcar-Castell et al 2015 Rautiainen et al 2018

21 Study site

NDVI=R800minusR670

R800+R670(Rouse et al 1974) (3a)

PRI=R531minusR570

R531+R570(Gamon et al 1992) (3c)

GCC=RGreen

RRed+RGreen+RBlue

800 nmsumλ=450 nm

(Rλminus Rλ)2 (7)

Ccar and CCI How-

A1 NDVI and NIRv

y = b+aminus b

1+ exp(dminusxc

(dminusxc

)rarr 0 So

limxrarr+infin

)rarr+infin So

limxrarrminusinfin

residual

aminus b

(1+ exp(dminusxc

exp(d minus x

aminus b

1+ exp(dminusxhalfc

) Hence xhalf = d

References

Abstract

Introduction

Study site




















Supplement


Competing interests

Acknowledgements

Financial support

Review statement

References

21 Study site

NDVI=R800minusR670

R800+R670(Rouse et al 1974) (3a)

PRI=R531minusR570

R531+R570(Gamon et al 1992) (3c)

GCC=RGreen

RRed+RGreen+RBlue

800 nmsumλ=450 nm

(Rλminus Rλ)2 (7)

Ccar and CCI How-

A1 NDVI and NIRv

y = b+aminus b

1+ exp(dminusxc

(dminusxc

)rarr 0 So

limxrarr+infin

)rarr+infin So

limxrarrminusinfin

residual

aminus b

(1+ exp(dminusxc

exp(d minus x

aminus b

1+ exp(dminusxhalfc

) Hence xhalf = d

References

Abstract

Introduction

Study site




















Supplement


Competing interests

Acknowledgements

Financial support

Review statement

References

NDVI=R800minusR670

R800+R670(Rouse et al 1974) (3a)

PRI=R531minusR570

R531+R570(Gamon et al 1992) (3c)

GCC=RGreen

RRed+RGreen+RBlue

800 nmsumλ=450 nm

(Rλminus Rλ)2 (7)

Ccar and CCI How-

A1 NDVI and NIRv

y = b+aminus b

1+ exp(dminusxc

(dminusxc

)rarr 0 So

limxrarr+infin

)rarr+infin So

limxrarrminusinfin

residual

aminus b

(1+ exp(dminusxc

exp(d minus x

aminus b

1+ exp(dminusxhalfc

) Hence xhalf = d

References

Abstract

Introduction

Study site




















Supplement


Competing interests

Acknowledgements

Financial support

Review statement

References

800 nmsumλ=450 nm

(Rλminus Rλ)2 (7)

Ccar and CCI How-

A1 NDVI and NIRv

y = b+aminus b

1+ exp(dminusxc

(dminusxc

)rarr 0 So

limxrarr+infin

)rarr+infin So

limxrarrminusinfin

residual

aminus b

(1+ exp(dminusxc

exp(d minus x

aminus b

1+ exp(dminusxhalfc

) Hence xhalf = d

References

Abstract

Introduction

Study site




















Supplement


Competing interests

Acknowledgements

Financial support

Review statement

References

800 nmsumλ=450 nm

(Rλminus Rλ)2 (7)

Ccar and CCI How-

A1 NDVI and NIRv

y = b+aminus b

1+ exp(dminusxc

(dminusxc

)rarr 0 So

limxrarr+infin

)rarr+infin So

limxrarrminusinfin

residual

aminus b

(1+ exp(dminusxc

exp(d minus x

aminus b

1+ exp(dminusxhalfc

) Hence xhalf = d

References

Abstract

Introduction

Study site




















Supplement


Competing interests

Acknowledgements

Financial support

Review statement

References

Ccar and CCI How-

A1 NDVI and NIRv

y = b+aminus b

1+ exp(dminusxc

(dminusxc

)rarr 0 So

limxrarr+infin

)rarr+infin So

limxrarrminusinfin

residual

aminus b

(1+ exp(dminusxc

exp(d minus x

aminus b

1+ exp(dminusxhalfc

) Hence xhalf = d

References

Abstract

Introduction

Study site




















Supplement


Competing interests

Acknowledgements

Financial support

Review statement

References

Ccar and CCI How-

A1 NDVI and NIRv

y = b+aminus b

1+ exp(dminusxc

(dminusxc

)rarr 0 So

limxrarr+infin

)rarr+infin So

limxrarrminusinfin

residual

aminus b

(1+ exp(dminusxc

exp(d minus x

aminus b

1+ exp(dminusxhalfc

) Hence xhalf = d

References

Abstract

Introduction

Study site




















Supplement


Competing interests

Acknowledgements

Financial support

Review statement

References

Ccar and CCI How-

A1 NDVI and NIRv

y = b+aminus b

1+ exp(dminusxc

(dminusxc

)rarr 0 So

limxrarr+infin

)rarr+infin So

limxrarrminusinfin

residual

aminus b

(1+ exp(dminusxc

exp(d minus x

aminus b

1+ exp(dminusxhalfc

) Hence xhalf = d

References

Abstract

Introduction

Study site




















Supplement


Competing interests

Acknowledgements

Financial support

Review statement

References

Ccar and CCI How-

A1 NDVI and NIRv

y = b+aminus b

1+ exp(dminusxc

(dminusxc

)rarr 0 So

limxrarr+infin

)rarr+infin So

limxrarrminusinfin

residual

aminus b

(1+ exp(dminusxc

exp(d minus x

aminus b

1+ exp(dminusxhalfc

) Hence xhalf = d

References

Abstract

Introduction

Study site




















Supplement


Competing interests

Acknowledgements

Financial support

Review statement

References

Ccar and CCI How-

A1 NDVI and NIRv

y = b+aminus b

1+ exp(dminusxc

(dminusxc

)rarr 0 So

limxrarr+infin

)rarr+infin So

limxrarrminusinfin

residual

aminus b

(1+ exp(dminusxc

exp(d minus x

aminus b

1+ exp(dminusxhalfc

) Hence xhalf = d

References

Abstract

Introduction

Study site




















Supplement


Competing interests

Acknowledgements

Financial support

Review statement

References

A1 NDVI and NIRv

y = b+aminus b

1+ exp(dminusxc

(dminusxc

)rarr 0 So

limxrarr+infin

)rarr+infin So

limxrarrminusinfin

residual

aminus b

(1+ exp(dminusxc

exp(d minus x

aminus b

1+ exp(dminusxhalfc

) Hence xhalf = d

References

Abstract

Introduction

Study site




















Supplement


Competing interests

Acknowledgements

Financial support

Review statement

References

A1 NDVI and NIRv

y = b+aminus b

1+ exp(dminusxc

(dminusxc

)rarr 0 So

limxrarr+infin

)rarr+infin So

limxrarrminusinfin

residual

aminus b

(1+ exp(dminusxc

exp(d minus x

aminus b

1+ exp(dminusxhalfc

) Hence xhalf = d

References

Abstract

Introduction

Study site




















Supplement


Competing interests

Acknowledgements

Financial support

Review statement

References

y = b+aminus b

1+ exp(dminusxc

(dminusxc

)rarr 0 So

limxrarr+infin

)rarr+infin So

limxrarrminusinfin

residual

aminus b

(1+ exp(dminusxc

exp(d minus x

aminus b

1+ exp(dminusxhalfc

) Hence xhalf = d

References

Abstract

Introduction

Study site




















Supplement


Competing interests

Acknowledgements

Financial support

Review statement

References

y = b+aminus b

1+ exp(dminusxc

(dminusxc

)rarr 0 So

limxrarr+infin

)rarr+infin So

limxrarrminusinfin

residual

aminus b

(1+ exp(dminusxc

exp(d minus x

aminus b

1+ exp(dminusxhalfc

) Hence xhalf = d

References

Abstract

Introduction

Study site




















Supplement


Competing interests

Acknowledgements

Financial support

Review statement

References

y = b+aminus b

1+ exp(dminusxc

(dminusxc

)rarr 0 So

limxrarr+infin

)rarr+infin So

limxrarrminusinfin

residual

aminus b

(1+ exp(dminusxc

exp(d minus x

aminus b

1+ exp(dminusxhalfc

) Hence xhalf = d

References

Abstract

Introduction

Study site




















Supplement


Competing interests

Acknowledgements

Financial support

Review statement

References

y = b+aminus b

1+ exp(dminusxc

(dminusxc

)rarr 0 So

limxrarr+infin

)rarr+infin So

limxrarrminusinfin

residual

aminus b

(1+ exp(dminusxc

exp(d minus x

aminus b

1+ exp(dminusxhalfc

) Hence xhalf = d

References

Abstract

Introduction

Study site




















Supplement


Competing interests

Acknowledgements

Financial support

Review statement

References

References

Abstract

Introduction

Study site




















Supplement


Competing interests

Acknowledgements

Financial support

Review statement

References

Abstract

Introduction

Study site




















Supplement


Competing interests

Acknowledgements

Financial support

Review statement

References

Abstract

Introduction

Study site




















Supplement


Competing interests

Acknowledgements

Financial support

Review statement

References

Abstract

Introduction

Study site




















Supplement


Competing interests

Acknowledgements

Financial support

Review statement

References

Abstract

Introduction

Study site




















Supplement


Competing interests

Acknowledgements

Financial support

Review statement

References

Documents

Decomposing reﬂectance spectra to track gross primary … · 2020. 9. 15. · 4524 R. Cheng et al.: Decomposing reﬂectance spectra to track gross primary production lighting all