Embed Size (px)
Citation preview
SHOHEI WATANABE
LA COULEUR DE L'EAU ET LA TRANSMISSION DE LA LUMIÈRE DANS LES ÉCOSYSTÈMES
LACUSTRES
Thèse présentée à la Faculté des études supérieures de l'Université Laval
dans le cadre du programme de doctorat en biologie pour l'obtention du grade de Philosophiae Doctor (Ph.D.)
DEPARTEMENT DE BIOLOGIE FACULTÉ DES SCIENCES ET DE GÉNIE
UNIVERSITÉ LAVAL QUÉBEC
2011
©Shohei Watanabe, 2011
Ill
Résumé L'objectif général était de faire progresser notre compréhension des caractéristiques
du rayonnement solaire dans une gamme optiquement diversifiée des eaux continentales et
en particulier pour le rayonnement photosynthétiquement actif (PAR, 400-700 nm). Les
mesures sur le terrain et en laboratoire des propriétés optiques apparentes et inhérentes (les
AOP et les IOP, respectivement) ont été réalisées en utilisant des instruments d'optique de
pointe. Ces instruments et méthodes avaient jusqu'à ce jour été utilisés surtout aux eaux
claires de l'océan au large des zones côtières. J'ai tout d'abord examiné les mécanismes
contrôlant les différences de couleur entre les mares thermokarstiques dans le Québec
subarctique. Les résultats ont montré que les variations quantitatives du carbone organique
dissous et des particules en suspension non algales étaient les facteurs principaux.
L'analyse d'une image satellite suggère que ces composantes peuvent être estimées à
distance. Les résultats indiquent que la télédétection par satellite pourrait permettre la mise
à l'échelle de certains processus biogéochimiques. J'ai par la suite quantifié les processus
d'atténuation du PAR dans un réservoir d'eau potable. Les résultats d'analyse du budget de
photons ont également révélé les changements spectraux du PAR en fonction de la
profondeur affectant l'estimation des coefficients d'absorption moyen du PAR in situ pour
la zone euphotique; par exemple, le coefficient de l'eau pure était dix fois plus élevé que
celui mesuré dans l'océan. Dans la dernière partie de cette étude, j'ai entrepris une analyse
de la «fermeture optique» afin d'évaluer la relation entre les AOP et les IOP pour un lac
hypereutrophe urbain. Une similarité a été observée entre la luminance spectrale mesurée
par télédétection avec un système de profilage dans l'eau et les valeurs obtenues grâce à un
modèle de transfert radiatif (HydroLight). En conclusion, ces études réalisées à trois sites
contrastés ont permis d'étendre la gamme de mesures et de modélisation optiques pour les
écosystèmes lacustres, et de faire progresser notre compréhension des processus qui
influencent l'éclairement spectral dans les eaux douces. Ces résultats soulignent la valeur
mais aussi les sources d'incertitude associées à l'application de l'optique hydrologique dans
les eaux continentales optiquement complexes.
Abstract The overarching objective of this doctoral thesis was to advance our understanding
of the behavior of solar radiation in an optically diverse range of inland waters, specifically
for the photosynthetically active radiation waveband (PAR, 400-700 nm). Field and
laboratory measurements of apparent and inherent optical properties (AOPs and IOPs,
respectively) were made by applying advanced optical instruments and techniques, whose
applications to date have been especially in clear waters of the offshore ocean. Firstly, I
examined the mechanisms underlying the striking color differences among turbid
permafrost thaw ponds in subarctic Québec. The results showed that quantitative
differences in dissolved organic carbon and non-algal suspended particulate matter
produced these large differences in optical conditions. Analysis of a satellite image
indicated that these two components can be estimated by satellite remote sensing. This
result may allow scaling up of biogeochemical processes from local to regional scale in
future research. Secondly, I quantified the underwater PAR attenuation processes in a
drinking water reservoir. Colored dissolved organic matter played the dominant role, with
phytoplankton having a much lesser effect on PAR attenuation. The results of photon
budget analysis revealed that the large changes in component-specific PAR attenuation
depended on variations in the spectral distribution of PAR. This affected the average in situ
absorption coefficients for the euphotic zone; for example, the coefficient for pure-water
was ten-fold higher than the value that has been measured in the ocean. Thirdly, I
undertook an optical closure analysis to evaluate the relationship between AOPs and IOPs
in an urban hypereutrophic lake. There was good agreement between spectral remote
sensing reflectance measured in the field and the values estimated from measured IOPs
with a radiative transfer model (HydroLight), supporting the validity of this modeling
approach for inland waters. However, there was strong sensitivity of the absorption
measurements to the method used for correcting the scattering error. In sum, these inter
related studies at contrasting sites have extended the previous range of optical
measurements and modeling for lacustrine systems, and have underscored the value but
also identified sources of uncertainty in applying hydrologie optics to optically complex
inland waters.
Vil
Avant-propos Cette thèse est l'apogée de ma recherche doctorale sous la direction du professeur
Warwick F. Vincent et la co-direction de M. Michio Kumagai. Elle est composée de cinq
chapitres, dont trois sous forme d'articles. Je suis l'auteur principal de tous ces articles et le
responsable de la planification et la réalisation des travaux sur le terrain, la plupart des
analyses en laboratoire, l'interprétation de données et la rédaction des articles.
L'organisation de cette thèse est la suivante :
Chapitre 1 - Introduction générale
Chapitre 2 - Un chapitre sur la diversité optique des mares thermokarstiques. Ce chapitre
est déjà publié comme Watanabe, S., I. Laurion, K. Chokmani, R. Pienitz et W. F. Vincent.
2011. «Optical diversity of thaw ponds in discontinuous permafrost: a model system for
water color analysis», Journal of Geophysical Research-Biogeosciences 116: G02003, doi:
10.1029/2010JG001380.
Chapitre 3 - Un chapitre sur l'atténuation de la lumière dans un réservoir d'eau potable. Ce
chapitre sera soumis sous peu à Limnology and Oceanography comme Watanabe, S. et al.
2011. «Light attenuation in a drinking water reservoir: photon budget analysis and
importance of abiotic factors».
Chapitre 4 - Un chapitre sur la modélisation de la lumière sous-marine dans un lac
hypertrophe. Ce chapitre sera soumis sous peu à Journal of Remote Sensing comme
Watanabe, S. et al. 2011. «Optical closure in a shallow urban lake: steps toward modeling
of remote sensing reflectance in optically complex inland waterbodies».
Chapitre 5 - Conclusion générale
IX
Remerciements Je voudrais tout d'abord remercier mon directeur de thèse, M. Warwick F. Vincent
du Département de biologie de l'Université Laval, pour ses conseils et son appui
indéfectible tout au long de mes études. J'ai tout spécialement apprécié l'incroyable
diversité d'opportunités qu'il m'a présentées durant ces cinq années à l'Université Laval.
De ma vie, je n'aurais jamais pensé visiter la région du permafrost Subarctique et l'océan
Arctique; ceci s'est avéré une très belle expérience aussi bien pour ma carrière scientifique
que pour ma vie personnelle. De plus, son énergie positive à mener d'importants et
passionnants projets scientifiques autour du monde a grandement stimulé mon inspiration.
Je suis convaincu que cette immersion dans ces passionnantes activités scientifiques
contribuera de façon significative au futur développement de ma carrière. Je voudrais aussi
remercier mon co-directeur M. Michio Kumagai de Lake Biwa Environmental Research
Institute (LBERI-Japon). Connaître ce limnologue distingué a été une chance, j 'ai beaucoup
appris lors de nos rencontres aux congrès scientifiques et lors de mes visites au LBERI. J'ai
tout spécialement apprécié ses efforts à me présenter de nombreux scientifiques japonais en
écologie aquatique; ceci ne serait jamais arrivé si je ne l'avais pas eu comme co-directeur
puisque mes études graduées ont principalement eu lieu en Amérique du Nord.
Je voudrais également remercier les membres de mon comité d'encadrement Mme
Isabelle Laurion (INRS-ETE, Québec) et M. Reinhard Pienitz (Département de géographie,
Université Laval) pour leurs précieux conseils tout au long de mes recherches. Mon étude a
été grandement stimulée par la qualité et l'intelligence de leurs travaux lors de notre
échantillonnage de terrain et l'écriture d'article. Je voudrais aussi remercier M. Emanuel
Boss (Maine In-situ Sound & Color Laboratory, School of Marine Sciences, University of
Maine, Orono, Maine, USA), d'avoir accepté d'être l'examinateur externe de cette thèse. Je
suis enfin très reconnaissant aux chercheurs suivants: Karem Chokmani (INRS-ETE),
Marcel Babin (Directeur du laboratoire Takuvik, CNRS-ULaval, Québec), Pierre Francus
(INRS-ETE), Dariusz Stramski, Rick A. Reynolds et Kuba Tatarkiewicz (Marine Physical
Laboratory, Scripps Institution of Oceanography, University of California San Diego,
Californie, USA). J'aimerai tout spécialement remercier Mme Kanako Ishikawa (LBERI)
pour ses conseils et son support qui m'ont permis de surmonter les difficultés rencontrées.
Dans le laboratoire de M. Vincent, j 'ai grandement apprécié l'aide technique de
Tommy Harding, Sébastien Bourget et Delphine Rolland. J'ai aussi apprécié l'aide de
Roxane Tremblay, Gabriel Sarasin, Marc-Antoine Couillard, Frédéric Bouchard, Yuya
Matsukawa, et Keita Suzuki lors de travaux sur le terrain. Mes remerciements tous spéciaux
vont à Marie-Josée Martineau pour son aide dans de nombreux aspects de la vie du
laboratoire. Merci aussi à Leira Rétamai, Marie Lionard, et Aaron Vincent, ainsi qu'aux
autres membres des laboratoires Vincent-Lovejoy et du laboratoire Laurion. Enfin, j 'ai
beaucoup apprécié l'aide des personnes de l'INRS pour les analyses chimiques lors de mes
travaux en laboratoire là bas.
Merci à mes chers amis Ayako Konomi, Yoshihiko Murao, Hidenori Natsui,
Takafumi Oishi, et Mariko Shimizu pour leurs encouragements. J'ai toujours été inspiré par
ces amis ayant une fantastique intelligence et personnalité. J'attends avec impatience nos
futurs échanges avec les nouveaux membres de leurs familles.
Pour finir, j'aimerai remercier ma famille de tout mon cœur, Yasuo, Masayo, et Nao
Watanabe pour leur amour et leurs encouragements indéfectibles. Sans leur compréhension
pour mes objectifs de carrières et leur support dans mon choix de vie dans un autre pays, je
n'aurai jamais pu aller au bout de cette aventure.
XI
Table des matières Résumé iii Abstract v Avant-propos vii Remerciements ix
Table des matières xi Liste des tableaux xv liste des figures xvii
Chapitre 1 Introduction générale 1 1.1 Introduction 1 1.2 Concepts fondamentaux en optique hydrologique 3
1.2.1 Le rayonnement solaire et rayonnement photosynthétiquement actif (PAR).... 3 1.2.2 Les propriétés optiques apparentes et inhérentes (AOP et IOP) 3 1.2.3 Les composantes optiquement actives 4
1.3 Sujets spécifiques d'optique hydrologique 4 1.3.1 La transmission de la lumière sous l'eau 4 1.3.2 La couleur de l'eau (le rayonnement qui sort de l'eau) 5 1.3.3 Les modèles de transfert radiatif. 6
1.4 Description des sites d'étude 6 1.4.1 Les mares thermokarstiques 7 1.4.2 Le lac Saint-Charles 8 1.4.3 Le lac Saint-Augustin 8
1.5 Organisation de la thèse 9
Chapitre 2 Optical diversity of thaw ponds in discontinuous permafrost: a model system for water color analysis 15
2.1 Résumé 15 2.2 Abstract 16 2.3 Introduction 17 2.4 Materials and methods 19
2.4.1 Study site and field sampling 19 2.4.2 Limnological variables 19 2.4.3 Apparent optical properties (AOPs) 20
2.4.3.1 Within water AOPs 20 2.4.3.2 Above-water AOPs 20 2.4.3.3 CIE L*a*b* color coordinates 21
2.4.4 Inherent optical properties (IOPs) 23 2.4.4.1 Absorption coefficient of colored dissolved organic matter 23 2.4.4.2 Absorption coefficients of suspended particulate matter 24 2.4.4.3 AC-S measurements 25 2.4.4.4 Absorption coefficient of fine particles 27 2.4.4.5 Validation of AC-S related measurements 27
2.4.5 Remote sensing analysis 28
xn
2.4.6 Statistical analyses 29 2.5 Results 30
2.5.1 Ambient conditions 30 2.5.2 AOPs 30 2.5.3 IOPs 32
2.5.3.1 Absorption coefficient of colored dissolved organic matter 32 2.5.3.2 Absorption coefficient of non-algal particles 32 2.5.3.3 Absorption coefficient of algal particles 32 2.5.3.4 Absorption coefficient of fine particles 33 2.5.3.5 Total absorption 33 2.5.3.6 Scattering 33
2.5.4 Remote sensing analysis 34 2.6 Discussion 35
2.6.1 Limnological conditions 35 2.6.2 Above-water AOPs and color coordinates 35 2.6.3 IOPs 36
2.6.3.1 Absorption coefficient of colored dissolved organic matter 36 2.6.3.2 Absorption coefficient of non-algal particles 37 2.6.3.3 Absorption coefficient of algal particles 39 2.6.3.4 Absorption coefficient of fine particles 39 2.6.3.5 Total absorption 40 2.6.3.6 Scattering 41
2.6.4 Remote sensing analysis 42 2.7 Conclusions 43 2.8 Acknowledgements 43
Chapitre 3 Light attenuation in a drinking water reservoir: photon budget estimation and the importance of abiotic factors 63
3.1 Résumé 63 3.2 Abstract 64 3.3 Introduction 65 3.4 Methods 66
3.4.1 Study site and limnological conditions 66 3.4.2 Apparent optical properties (AOPs) 66 3.4.3 Laboratory analyses of limnological variables 67 3.4.4 Inherent optical properties (IOPs) 68
3.4.4.1 Absorption coefficient of colored dissolved organic matter 68 3.4.4.2 Absorption coefficients of suspended particulate matter 68 3.4.4.3 AC-S measurements 69
3.4.5 Photon budget calculation 70 3.5 Results 72
3.5.1 Limnological variables 72 3.5.2 AOPs 72 3.5.3 IOPs 73
3.5.3.1 Absorption coefficients of colored dissolved organic matter 73 3.5.3.2 Absorption coefficients of suspended particulate matter 73 3.5.3.3 Scattering coefficients 73
Xlll
3.5.4 Photon budget calculations 74 3.6 Discussion 75
3.6.1 IOPs and specific characteristics of optically active components 75 3.6.1.1 Colored dissolved organic matter 75 3.6.1.2 Non-algal particles 76 3.6.1.3 Algal particles 76 3.6.1.4 Fine particles 77 3.6.1.5 Scattering coefficients 78
3.6.2 Photon budget calculation 78 3.6.2.1 Vertical attenuation of PAR 78 3.6.2.2 Contributions to PAR attenuation 79 3.6.2.3 Seasonal dynamics of PAR attenuation 79 3.6.2.4 Attenuation by water 80 3.6.2.5 Application of the photon budget approach 80 3.6.2.6 Estimation of average cosine and back scattering ratio 81
3.7 Conclusions 83 3.8 Summary of the notation used in this article 84
Chapitre 4 Optical closure in a shallow urban lake: steps toward modeling of remote sensing reflectance in optically complex inland waters 97
4.1 Résumé 97 4.2 Abstract 98 4.3 Introduction 99 4.4 Materials and Methods 102
4.4.1 Study site and field sampling 102 4.4.2 Apparent optical properties (AOPs) 102 4.4.3 Limnological variables 103 4.4.4 Inherent optical properties (IOPs) 104
4.4.4.1 Absorption coefficient of colored dissolved organic matter 104 4.4.4.2 Absorption coefficients of suspended particulate matter 105 4.4.4.3 AC-S measurements 106
4.4.5 Radiative transfer modeling (RTM) 107 4.5 Results 107
4.5.1 Ambient conditions and IOPs 107 4.5.2 Optical closure 108
4.6 Discussion 109 4.6.1 Limnological conditions 109 4.6.2 IOPs for optically active components 109 4.6.3 Optical closure 110 4.6.4 Uncertainties in IOP measurements 111
4.6.4.1 Near-infrared absorption 111 4.6.4.2 Null point correction of quantitative filter technique (QFT) 111 4.6.4.3 The path length amplification factor 112 4.6.4.4 The absorption coefficient of fine particles 112 4.6.4.5 Future improvement opportunities of particulate absorption easurements
113 4.6.5 Importance of fluorescence estimations on the RTM 113
XIV
4.7 Conclusions 114
Chapitre 5 Conclusion générale 125
Références bibliographiques 131
Annexe 1 Error analyses for optical measurements 143 Al.l Introduction 143 A1.2 Methods 143
Al.2.1 Varian Cary 100 bench-top spectrophotometer 143 Al.2.2 WET Labs AC-S in situ spectrophotometer 143 Al.2.3 Satlantic radiometer system 144
Al.3 Results and discussion 144 AI.3.1 Varian Cary 100 bench-top spectrophotometer 144 Al.3.2 WET Labs AC-S in situ spectrophotometer 144 Al.3.3 Satlantic radiometer system 145
Annexe 2 Particle size distribution analysis of thaw ponds 151 A2.1 Introduction 151 A2.2 Methods 151 A2.3 Results and discussion 152 A2.4 References 152
Annexe 3 The pathlength amplification factor (beta factor) for the quantitative filter technique 155
A3.1 Introduction 155 A3.2 Methods 155
A3.2.1 The quantitative filter technique 155 A3.2.2 WET Labs AC-S analysis 156
A3.3 Results and discussion 156 A3.4 References 157
XV
Liste des tableaux Table 2.1 Limnological variables for the 14 permafrost thaw ponds 44
Table 2.2 Summary of the absorption coefficients (a), specific absorption coefficients (a*) and exponential slope parameters (S) for optically active components in 14 thaw ponds 45
Table 2.3 The relative contribution of each absorbing component to total absorption at 440 nm (in %) for all sampled thaw ponds 46
Supplement Table 2.1 Standardized coefficients to calculate canonical scores from satellite data for the estimation of total suspended particulate matter (SPMT) and dissolved organic carbon (DOC) 60
Table 3.1 Limnological conditions of Lake St. Charles in summer of 2008 86
Table 3.2 Summary of inherent optical properties measured by a laboratory bench top spectrophotometer and a WET Labs AC-S spectrophotometer 87
Table 3.3 Summary of the photon budget estimation 88
xvn
Liste des figures Figure 1.1 Le rayonnement solaire à la surface de la Terre 12
Figure 1.2 Diagramme schématique de modélisation directe et de fermeture optique dans les eaux optiquement complexes 13
Figure 2.1 Map of study area 47
Figure 2.2 Aerial photograph of the study site, showing some of the sampled thaw ponds... 48
Figure 2.3 Dimensionless ratio of above surface upwelling radiance to 27. corrected
downwelling irradiance (R(Xj) of studied ponds 49
Figure 2.4 Color of studied ponds expressed on CIE L*a*b* color coordinate 50
Figure 2.5 The absorption coefficient spectra from 400 to 700 nm 51 Figure 2.6 Relationship between absorption coefficient by colored dissolved organic matter
at 440 nm (OCDOM(440)) and dissolved organic carbon (DOC) 52
Figure 2.7 Relationship between absorption coefficient of non-algal particles at 440 nm (ctNAp(440)) and total suspended particulate matter (SPMT) 53
Figure 2.8 Relationship between chlorophyll a concentration (Chi a) and absorption
coefficient of algal particles at 440 nm (0^(440)) 54
Figure 2.9 Examples of component absorption spectra for representative pond types 55
Figure 2.10 The particulate scattering coefficient spectra (bp(X)) from 400 to 700 nm 56
Figure 2.11 The single scattering albedo spectra (cop(X)) for all 14 sampled ponds 57
Figure 2.12 Reflectance spectra of the 34 ponds obtained from the QuickBird satellite
image in July 2006 58
Figure 2.13 Comparison of measured to estimated values 59
Supplement Figure 2.1 Validation of the AC-S measurements 61
Figure 3.1 Map showing the location and bathymetry of Lake St. Charles 89 Figure 3.2 The profile of natural log-transformed spectral downward photon flux density
(Eq:d(z;X)) on 8 July 2008 90
XVlll
Figure 3.3 The spectral diffuse attenuation coefficient of downward photon flux density (Kd(X)) 91
Figure 3.4 Spectral profiles of inherent optical properties 92
Figure 3.5 Vertical profiles of A. the estimated diffuse attenuation coefficient for PAR (Kd'(z;PAR)) for all 8 sampling dates, and B. the estimated (solid line) and measured (broken line) diffuse attenuation coefficient for PAR (Kd'(z;PAR) and Kd(z;PAR), respectively) on 8 July 2008 93
Figure 3.6 Vertical profiles of component-specific diffuse absorption/scattering coefficients for 8 July 2008 94
Figure 3.7 Seasonal changes in the contribution of each absorption component to the diffuse attenuation coefficient for PAR (Kd'(z;PAR)) 95
Figure 3.8 Spectra of the component-specific, per volume absorption/scattering of photon flux (ad;i(z;X)xEq;d(z;X)) at different depths down the water column 96
Figure 4.1 . Map showing the location and bathymetry of Lake St. Augustin 115
Figure 4.2 Vertical profiles of physical and biological variables in Lake St. Augustin (5 October 2010) measured by the YSI 6600 sonde 116
Figure 4.3 The inherent optical properties of the surface water of Lake St. Augustin measured by the bench-top spectrophotometer 117
Figure 4.4 The inherent optical properties of Lake St. Augustin surface water measured by the AC-S 118
Figure 4.5 Normalized remote sensing reflectance spectra (Rrs(X), sr"1) for the Lake St. Augustin measured by the in-water radiometers 119
Figure 4.6 Results of the optical closure analysis 120
Figure 4.7 Normalized remote sensing reflectance spectra (Rrs(X), sr"1) 121
Figure 4.8 Results of the sensitivity test on forward modeling 122
Figure 4.9 Differences in inherent optical properties measured by the AC-S as a result of the different correction methods 123
Figure 4.10 Raw absorption spectra in the near-infrared radiation (NIR) region 124
Figure Al.l Observed noise on pure water (blank) readings for a Varian Cary 100 spectrophotometer 146
XIX
Figure A 1.2 Observed difference between duplicate readings for a Varian Cary 100 spectrophotometer 147
Figure A1.3. Measurement error for the WET Labs AC-S spectrophotometer 148
Figure A 1.4 Observed noise on the dark count for a Satlantic spectroradiometer system 149
Figure Al .5 Measurement error for the Satlantic radiometer positioned above the water 150
Figure A2.1 Particle size distribution of six thaw ponds 153
Figure A3.1 Comparison of the absorption coefficients of suspended particulate matter (ap(X)) obtained by different methods 159
Chapitre 1
Introduction générale
1.1 Introduction Le régime de lumière dans l'eau des lacs, des rivières et des mers a un effet majeur
sur leur structure et fonction écologiques. La lumière est la principale source d'énergie à la
base des écosystèmes aquatiques, et la distribution spectrale verticale du rayonnement
solaire dans la colonne d'eau exerce une influence fondamentale sur la répartition et
l'ampleur de la production primaire (Falkowski et Raven 2007; Kirk 1994). Le profil
d'absorption de la radiation solaire contrôle le budget de chaleur et la stratification des eaux
naturelles, et il affecte donc leur structure verticale et leur dynamique physique (Lewis
1983; Morel et Antoine 1994; Patterson et Hamblin 1988; Zaneveld et al. 1981). La plus
haute énergie du rayonnement solaire (rayonnement de courte longueur d'onde, dans les
gammes d'ondes UV et visible) a de nombreux effets sur la photochimie et la photobiologie
qui influencent le fonctionnement des écosystèmes aquatiques (Frenette et Vincent 2003).
En plus de leurs multiples rôles dans les processus aquatiques, les conditions optiques des
eaux naturelles ont des implications importantes pour les pratiques de gestion des
ressources hydriques. Les propriétés optiques peuvent être un indicateur sensible aux
changements dans le couplage des milieux terrestres et aquatiques (Pienitz et Vincent 2000),
à la croissance du phytoplancton et aux changements de la qualité de l'eau (International
Ocean Color Coordinating Group; IOCCG 2008). Par ailleurs, la couleur perçue ainsi que
la transparence de l'eau affectent sa valeur esthétique, et donc l'importance économique de
certains plans d'eau.
L'étude de la lumière dans les écosystèmes aquatiques (optique hydrologique) a
progressée rapidement au cours des dernières décennies en réponse à une demande
croissante des connaissances ainsi qu'à la suite de diverses avancées technologiques.
Indépendamment de leurs échelles spatiales et temporelles, les études des écosystèmes
aquatiques nécessitent des informations détaillées sur la distribution spectrale et verticale
de la lumière à travers la colonne d'eau et de son influence déterminante sur la physique, la
biogéochimie et la biologie (par ex., Fujii et al. 2007). De plus, la surveillance de l'état
biologique des écosystèmes aquatiques par des techniques de télédétection nécessite des
connaissances détaillées sur le comportement de la lumière depuis la surface de l'eau vers
l'atmosphère (IOCCG 2000). Pour répondre à ces besoins, les instruments et les techniques
d'analyse des propriétés optiques des écosystèmes aquatiques ont rapidement progressé et
sont devenus plus accessibles (voir Mueller 2002; Mueller et al. 2002; Pegau et al. 2003).
En conséquence, des séries de données optiques de plus en plus détaillées et disponibles ont
permis des avancées majeures dans notre compréhension de la lumière dans les
écosystèmes aquatiques (par ex., la base de données Worldwide Ocean Optics, Office of
Naval Research, États-Unis; http://wood. jhuapl.edu/wood/).
La disponibilité commerciale des radiomètres et des photomètres hyperspectraux
(instruments qui mesurent des bandes de fréquences étroites sur une large gamme spectrale)
a grandement accéléré l'acquisition de données et la compréhension de l'optique
hydrologique (Roesler and Boss 2008). Cependant, à ce jour, ces instruments ont surtout
été exploités dans des systèmes où la complexité optique est réduite, comme les eaux
bleues océaniques et les lacs ultra-oligotrophes. Il existe beaucoup moins de connaissances
sur les caractéristiques des eaux optiquement complexes, telles que celles des lacs plus
riches en nutriments, des rivières et des zones côtières (IOCCG 2000; mais voir Bukata et
al. 1995, Arst 2003, Davies-Colley et al. 2003, et Miller et al. 2005). Il est donc opportun et
important d'appliquer et d'adapter les connaissances et les techniques développées en eaux
océaniques aux eaux continentales afin de mieux comprendre leurs processus optiques et
d'étendre l'optique hydrologique à un plus large éventail de conditions. Ces efforts
contribueront aussi à une meilleure compréhension des mécanismes écologiques et
biogéochimiques influencés par la lumière, ainsi qu'au développement des pratiques de
gestion des précieuses ressources mondiales en eaux douces (Boss et al. 2007).
Les objectifs principaux de mon étude doctorale ont été de mesurer et de modéliser
les propriétés optiques des écosystèmes lacustres en appliquant des approches
océanographiques et d'élargir ainsi la portée de l'optique hydrologique. Dans ce chapitre
d'introduction, je présenterai tout d'abord un aperçu succinct des concepts fondamentaux en
optique hydrologique, suivi par une présentation plus détaillée des thèmes spécifiques
pertinents à ma thèse. J'introduirai ensuite les sites d'étude, puis je décrirai l'organisation de
la thèse et les objectifs spécifiques à chaque chapitre.
1.2. Concepts fondamentaux en optique hydrologique 1.2.1 Le rayonnement solaire et le rayonnement photosynthétiquement actif (PAR)
Le rayonnement solaire est la première source d'énergie des écosystèmes aquatiques
(Falkowski et Raven 2007; Kirk 1994); les propriétés de la lumière dans l'eau sont donc
essentielles à une bonne compréhension de leur structure et de leur fonction. Le
rayonnement solaire à la surface de la Terre s'étend sur une large gamme spectrale (Fig.
1.1). Le pic de rayonnement solaire se produit dans la zone allant de 400 à 700 nm et
constitue 45 % de l'éclairement total atteignant la surface de la Terre (Falkowski et Raven
2007; Kirk 1994; Mobley 1994). Cette zone, correspondant à la gamme des longueurs
d'onde perçues par l'œil humain, a des implications particulièrement importantes pour
l'optique hydrologique. En effet, les organismes photosynthétiques utilisent cette portion de
longueurs d'onde pour la production primaire, c'est pourquoi elle est appelée le
rayonnement photosynthétiquement actif (PAR). Cette gamme a également été utilisée par
de nombreux systèmes de télédétection satellitaire (IOCCG 1998). Ainsi, la majorité des
recherches en optique hydrologique se concentrent sur la distribution temporelle, verticale
et spectrale du PAR dans les écosystèmes aquatiques.
1.2.2 Les propriétés optiques apparentes et inhérentes (AOP et IOP)
Deux grandes classes de variables optiques sont utilisées pour caractériser la
lumière dans l'eau: les propriétés optiques apparentes (AOP) et les propriétés optiques
inhérentes (IOP). Les AOP sont dépendantes à la fois des caractéristiques du milieu et des
conditions de lumière ambiantes (par ex. de la distribution angulaire des photons incidents,
ce qui est le cas de la lumière solaire). Il s'agit notamment de variables telles que la
réflectance spectrale détectée à distance (Rrs(X)). En revanche, les IOP sont ainsi nommées
puisqu'elles sont inhérentes au milieu et indépendantes de la lumière ambiante. Elles sont
mesurées pour un faisceau de photons parallèles généré instrumentalement (collimated
beam). Elles comprennent les coefficients d'absorption spectrale (a(X)), les coefficients de
diffusion spectrale (b(X)), et les coefficients d'atténuation spectrale d'un faisceau (c(X)).
Définir les variations verticales et spectrales de ces paramètres est nécessaire pour la
modélisation du transfert radiatif et pour caractériser les composantes optiquement actives
dans la colonne d'eau (Kirk 1994; Mobley 1994).
1.2.3 Les composantes optiquement actives
Les IOP des eaux naturelles sont contrôlées par des substances «optiquement
actives» dans la colonne d'eau. Les molécules d'eau, la matière organique dissoute colorée
(CDOM), le phytoplancton, et d'autres particules organiques et inorganiques en suspension
(particules non-algales, NAP) en sont les constituants majeurs, et leur quantité et
distribution déterminent les propriétés optiques de l'eau. Ces substances optiquement
actives sont à leur tour influencées par les diverses caractéristiques du système, y compris
la biogéochimie interne et les processus hydrodynamiques, ainsi que les effets du paysage
incluant la géologie, la géomorphologie, la végétation, le climat et les activités
anthropiques. En conséquence, les propriétés optiques des écosystèmes aquatiques peuvent
être exploitées pour estimer les composantes biogéochimiques de la colonne d'eau (par ex.,
Boss et al. 2009), et pour fournir des mesures quantitatives des changements
environnementaux. La croissance du phytoplancton, les flux de carbone, les changements
dans la qualité de l'eau, les impacts climatiques et le couplage avec le bassin versant ont
tous le potentiel d'être estimés à partir de variables optiques (Kirk 1994, Mobley 1994).
1.3 Sujets spécifiques d'optique hydrologique 1.3.1 La transmission de la lumière sous l'eau
La transparence de l'eau et ses facteurs de contrôle sont une partie importante des
études en eaux douces depuis les débuts de la limnologie (Kalff 2002; Wetzel 2001). Les
progrès technologiques récents ont permis de mesurer l'éclairement de façon plus détaillée
en fonction de la longueur d'onde et de la profondeur de la colonne d'eau (Lewis 2008;
Roesler and Boss 2008). L'éclairement descendant (Ed(X)) est la variable la plus
couramment mesurée et son extinction verticale est quantifiée comme le coefficient
d'atténuation diffuse (Kd(X); Lewis 2008). Cette valeur pour le PAR (Kd(PAR)) est
particulièrement importante dans les études d'écologie aquatique, car le coefficient
décrivant la lumière disponible pour la photosynthèse à chaque profondeur de la colonne
d'eau est utilisé pour définir la limite verticale de la production primaire (Kalff 2002; Kirk
1994). Il peut également être utilisé pour calculer le budget calorifique des plans d'eau
(Caplanne et Laurion 2008; Morel et Antoine 1994; Patterson et Hamblin 1988). Le
coefficient Kd(PAR) peut être subdivisé en composantes optiquement significatives dans la
colonne d'eau (Kirk 1994; Mobley 1994), afin de connaître la contribution proportionnelle
de chacune des composantes et déterminer les principaux facteurs contrôlant l'atténuation
spectrale.
1.3.2 La couleur de l'eau (le rayonnement qui sort de l'eau)
Une fraction du rayonnement entrant dans la colonne d'eau est réfléchie vers
l'atmosphère. Cette portion de la lumière visible (PAR) réfléchie est perçue comme étant la
couleur de l'eau par l'œil humain, et elle peut être quantifiée par des capteurs optiques
comme la réflectance détectée à distance (Rrs(X)). Cette dernière est de plus en plus utilisée
pour estimer les caractéristiques biologiques et biogéochimiques des eaux naturelles
(IOCCG 1998). Dans les systèmes océaniques, la relation entre Rrs(X) et les caractéristiques
biologiques de la colonne d'eau a été largement étudiée, et des modèles robustes ont été
développés (IOCCG 1998). Cependant, ces modèles ont surtout été utilisés pour estimer les
concentrations en Chi a dans les systèmes océaniques, et ils ont été appliqués pour des
estimations biogéochimiques à l'échelle mondiale (par ex., Arrigo et al. 2008; Morel 1991).
Dans les eaux continentales optiquement complexes et les plans d'eau côtiers, ces relations
sont encore mal comprises, et le développement d'un modèle robuste universellement
applicable est difficile en raison de la diversité des conditions optiques (Bukata et al. 1995;
IOCCG 2000). Les modèles développés et adaptés localement se révèlent cependant être
des outils de plus en plus utiles pour les zones côtières et pour la gestion des ressources en
eaux continentales, où ils ont été appliqués avec succès pour estimer les concentrations en
Chi a (Oyama et al. 2009; Tyler et al. 2006), la phycocyanine (Simis et al. 2005), et
l'abondance en matières particulates en suspension (Li et al. 2003). Ces études sont à leurs
débuts, et l'extension à un plus large éventail de conditions est encore nécessaire.
1.3.3 Les modèles de transfert radiatif
Le terme «fermeture optique» se réfère à la tentative de relier de façon théorique les
AOP et les IOP dans un modèle balancé (Zaneveld 1994). Les incertitudes et sensibilités
des mesures optiques et de la modélisation peuvent aussi être examinées à travers l'analyse
de la fermeture optique (Fig. 1.2; Gallegos et al. 2008; Zaneveld 1994). La prédiction
d'AOP à partir d'IOP, est appelée modélisation directe (forward modeling) et peut être
effectuée en utilisant l'application d'un modèle de transfert radiatif (RTM; Mobley 1994),
que l'on peut exprimer sous la forme simplifiée suivante:
c o s 6 Ë Ù Ê J Ê . = _c(_)L(z, 6,cp) + [ p{z, 0,ç;&,ç.')z(<9',tp')dU eq. 1.1 dz *n
où L(6,<p) est la radiance à l'angle zénithal 6 et l'angle nadir tp, c(z) le coefficient
d'atténuation du faisceau tp à la profondeur z, L(z,d,tp) la radiance à z, ,3(z,9,(p;d',(p') la
fonction de diffusion de volume, et ___ l'angle du rayonnement entrant dans le système
(Mobley 1994). La modélisation RTM de la réflectance détectée à distance (Rrs(X)) à partir
des IOP est particulièrement nécessaire pour le développement d'algorithmes robustes pour
la télédétection (IOCCG 2000). Malgré son rôle central dans l'optique hydrologique, la
modélisation directe a surtout été appliquée dans les milieux océaniques, avec seulement
quelques exceptions notables dans les plans d'eau optiquement complexes (Belzile et al.
2004; Gallegos et al. 2008; O'donnell et al. 2010; Tzortziou et al. 2006). La validation des
modèles par des procédés tels que la démonstration de fermeture optique est nécessaire
pour le développement de modèles optiques et la surveillance par télédétection des lacs, des
réservoirs et des eaux côtières.
1.4 Description des sites d'étude Ce travail de thèse vise à étendre la portée de l'observation et de la modélisation
hydrologique dans le domaine de l'optique grâce à des mesures détaillées provenant d'eaux
continentales optiquement contrastées. Trois sites d'étude ont été choisis afin de couvrir une
large gamme de conditions optiques: les mares thermokarstiques formées par la fonte du
pergélisol dans la toundra forestière subarctique du Québec, qui sont connues pour leur
remarquable éventail de couleurs (Breton et al. 2009; Laurion et al. 2010; Vincent et
Laybourn-Parry 2008); le lac Saint-Charles, réservoir d'eau potable de la ville de Québec
qui a récemment connu des problèmes de qualité de l'eau; et le lac Saint-Augustin, un lac
urbain hypereutrophe contenant des densités élevées de phytoplancton (Bouchard-Valentine
et al. 2004; Pienitz et al. 2006). La section suivante présente chacun de ces sites d'étude.
1.4.1 Les mares thermokarstiques
Dans les régions arctiques et subarctiques, des millions de mares thermokarstiques
(ou mares de fonte) sont issues de la fonte irrégulière et de l'érosion des sols gelés (Pouliot
and Bhiry 2005; Payette et al. 2004; Pienitz et al. 2008). Cette composante importante des
écosystèmes aquatiques de haute latitude attire une attention scientifique croissante en
raison de la capacité des sols nordiques de la planète à stocker d'importantes quantités de
carbone organique (Ping et al. 2008; Zimov et al. 2006). Les mares thermokarstiques
peuvent agir comme des bioréacteurs en libérant dans l'atmosphère des gaz à effet de serre
dont le carbone provient de la fonte et de l'érosion des sols (Laurion et al. 2010; Walter et
al. 2006). Ces systèmes contiennent des concentrations élevées en dioxyde de carbone
(CO2) et en méthane (Laurion et al. 2010; Walter et al. 2006), mais ils demeurent trop peu
étudiés en raison des difficultés d'accès aux régions nordiques éloignées. Dans la région de
la toundra forestière près de Whapmagoostui-Kuujjuarapik, Québec, Canada (55°N, 77°W),
les mares thermokarstiques englobent une diversité frappante de couleurs et de
transparences, même entre des mares très près l'une de l'autre (Breton et al. 2009). La
gamme de couleurs peut ainsi aller du vert au brun, en passant par le noir. Cette différence
marquée dans la couleur implique des variations dans les conditions optiques de l'eau,
reflétant la grande variabilité limnologique entre ces mares (Breton et al. 2009). Ce type de
système fournis ainsi un ensemble de conditions idéales pour développer et valider les
modèles, et donc pour réaliser des avancées dans le domaine de l'optique hydrologique. De
plus, les propriétés optiques de ces eaux ont été corrélées avec leurs taux d'émission de gaz
à effet de serre (Laurion et al. 2010). Par conséquent, des informations détaillées sur les
conditions optiques des mares thermokarstiques peuvent fournir des indications sur leur
limnologie ainsi que sur la biogéochimie régionale de ce biome de haute latitude.
8
1.4.2 Le lac Saint-Charles
Le lac Saint-Charles est situé à 20 km au nord de la ville de Québec. Son bassin
versant comprend un mélange de zones forestières et résidentielles (Pienitz et Vincent
2003). Ce lac est la source principale d'eau potable pour la ville de Québec. Toutefois, son
état trophique est défini comme étant mésotrophe avancé, avec des concentrations en
phosphore total allant jusqu'à 13 pg L"1 (Pienitz et Vincent 2003; Tremblay et al. 2001).
Les études paléolimnologiques conduites sur le lac dans les années 90 n'ont fournies aucune
preuve de changement rapide de l'état trophique du lac, alors que les concentrations en
métaux ont augmentées depuis la fin du I9ème siècle, suggérant l'entrée de polluants
atmosphériques (Pienitz et Vincent 2003; Tremblay et al. 2001). Des proliférations de
cyanobactéries dans le lac ont été observées pour la première fois en 2007, et de récentes
études limnologiques ont fourni des signes d'eutrophisation. L'appauvrissement en oxygène
dans le fond du lac est observé durant l'été et l'hiver (Pienitz et Vincent 2003; Tremblay et
al. 2001), ce qui peut contribuer au changement dans le cycle des nutriments internes et
ainsi accélérer l'eutrophisation. La surveillance attentive des conditions trophiques du lac et
le développement d'un plan de gestion solide pour la qualité de l'eau sont donc des enjeux
cruciaux pour la ville.
1.4.3 Le lac Saint-Augustin
Le lac Saint-Augustin est un lac urbain situé immédiatement à l'ouest de la ville de
Québec, et est entouré de résidences. L'état trophique du lac s'est fortement dégradé au
cours du siècle passé suite aux activités anthropiques dans son bassin versant (Pienitz et al.
2006). Ce lac est actuellement considéré comme hypereutrophe, avec des concentrations en
phosphore total de plus de 50 pg L"1 et avec une concentration en chlorophylle a pouvant
atteindre plus de 100 pg L"1 à la fin de l'été (Bouchard-Valentine et al. 2004). Le lac est
populaire pour les activités estivales de loisirs comme le canotage, la pêche et la natation.
Ces activités y sont cependant interdites par la municipalité à la fin de l'été en raison de la
prolifération des cyanobactéries à potentiel toxique (Bouchard-Valentine et al. 2004).
Ainsi, la surveillance de la qualité de l'eau est un enjeu majeur pour une utilisation
sécuritaire du lac et pour une gestion adéquate du lac et de son bassin versant.
1.5 Organisation de la thèse Le but ultime de cette thèse est de faire progresser notre compréhension quantitative
de l'optique hydrologique des eaux continentales en appliquant les connaissances et les
techniques développées dans les études océaniques. J'ai mesuré à la fois la transmission de
la lumière sous l'eau et la couleur de l'eau (le rayonnement spectral qui sort de l'eau), puis
j 'ai examiné leurs facteurs de contrôle afin de décrire les processus optiques des lacs sous
un large éventail de conditions.
Le chapitre 2 décrit l'application de ces approches océanographiques de
quantification des conditions optiques dans les mares thermokarstiques des régions
subarctiques, et la détermination des facteurs qui génèrent les différences marquées de la
couleur de l'eau. Le premier objectif spécifique était de quantifier la couleur de l'eau par le
biais de mesures de réflectance spectrale, puis de déterminer l'influence des conditions
limnologiques sur la couleur des mares. Le deuxième objectif était de mesurer les IOP de
ces mares afin de caractériser les composantes optiquement actives dans la colonne d'eau.
Enfin, je voulais faire une évaluation initiale de l'utilité de la télédétection satellitaire
comme approche pour estimer certaines caractéristiques limnologiques et biogéochimiques
clés de ces mares en vue d'appliquer cette démarche aux échelles locales et régionales du
paysage.
Dans ce travail sur les mares thermokarstiques, la couleur de l'eau a été quantifiée
selon un système de coordonnées de couleur, et la relation entre les valeurs des couleurs et
les variables limnologiques ont été examinées afin de comprendre la diversité des couleurs.
Pour décrire les IOP associées à chaque composante optiquement active, les
caractéristiques spécifiques d'absorption et de diffusion ont été analysées. La faisabilité de
l'estimation par télédétection de deux composantes principales, le carbone organique
dissous et les particules non-algales en suspension, a été évaluée en utilisant une image
prise par le satellite QuickBird, équipé d'un capteur multispectral de haute résolution
spatiale.
10
L'objectif global du chapitre 3 était d'évaluer les processus optiques et leurs facteurs
de contrôle dans le lac Saint-Charles, un réservoir d'eau potable. Le but était de déterminer
les IOP des composantes optiquement actives dans la colonne d'eau, et leurs relations avec
les variables limnologiques, afin de parvenir à une meilleure compréhension générale de
l'optique hydrologique des lacs tempérés, et d'aider à la modélisation écologique et à la
surveillance de cette source importante en eau potable. Les processus d'atténuation du PAR
dans la colonne d'eau ont été décrits par le couplage des IOP et des AOP afin d'estimer le
budget en photons, c'est-à-dire une analyse quantitative du devenir des photons dans la
colonne d'eau en fonction de la profondeur et des composés optiquement actifs. Ce chapitre
vise à évaluer le potentiel de cette approche «budgétaire» pour de futures études optiques
dans les systèmes optiquement complexes.
Les analyses des IOP pour le lac Saint-Charles portaient sur l'évaluation spectrale
de l'absorption et de la diffusion, ainsi que sur la dynamique saisonnière des composantes
optiquement actives dans le lac, c'est à dire le CDOM, les particules algales et non-algales,
les particules fines et les molécules d'eau. Ces observations des composantes optiquement
actives, leurs sources potentielles et leurs contributions relatives aux processus optiques de
la colonne d'eau ont été examinées en concert avec la surveillance optique du lac. Les
résultats de ces analyses détaillées des IOP ont été utilisés pour le calcul du budget en
photons en les couplant aux mesures de l'éclairement spectral du PAR.
L'objectif global du chapitre 4 a été de réaliser une première évaluation de la
fermeture optique pour un lac hypereutrophe. Il existe un besoin urgent de développer des
techniques de surveillance de la qualité de l'eau au lac Saint-Augustin, incluant des
méthodes optiques (Pienitz et Vincent 2003). De plus, ce lac représente un environnement
riche en phytoplancton, ainsi complétant la gamme limnologique et optique des sites
examinés dans les autres chapitres. J'ai comparé les mesures de réflectance détectée à
distance obtenues à l'aide d'un radiomètre hyperspectral in situ, avec les résultats
modélisés estimés par un modèle de transfert radiatif exploitant les IOP. Ceci a permis de
tester la validité du modèle dans une masse d'eau optiquement complexe. Un objectif
additionnel était d'évaluer les différentes méthodes de correction d'erreur de diffusion sur
11
les mesures d'IOP effectuées par spectrophotométrie in situ, afin de tester la validité et
l'applicabilité future de telles méthodes.
12
E a
T C §
I
E o "<5 u <•_. x 3
ce <
O co
3.0-
2.0
1.0
EXTRATERRESTRIAL SOLAR RADIATION DIRECT SOLAR RADIATION AT GROUND LEVEL
T
! UV / PAISIBLE I INFRARED __*£&.--.
i i T i i i r i i i 0.8 1.2 1.6 2.0 2.4 2.8 3.2
WAVELENGTH (pm)
Figure 1.1 Le rayonnement solaire à la surface de la Terre. La ligne continue montre le rayonnement à la surface terrestre et la ligne brisée montre le rayonnement au sommet de l'atmosphère. La figure provient de Wetzel (2001).
13
Closure experiment: I) measurements consister
li) assumptions in model
Measured IOPs ( . \ . b ,e . . . ) and [chl-a)
Measured radiation fields <_,«, L M , L^R,,...)
Use of measurements and mode^estimations ki : M Validation & interpretation
of satellite measured L.. R,, * Comparison with satellite derived products
(e.g. IchK). b* -COOM) 31 Investigation of applicaMity of satellite
bk>optical algorithms for specific waters SI Development, parameterization & refinement
ol regional specific bio-opocal models
Problems / uncertainties regarding:
_ model assumptions on water optical properties
3i accuracy of measurements. y consistency among Independently
measured quantities
Figure 1.2 Diagramme schématique de modélisation directe et de fermeture optique dans les eaux optiquement complexes. La figure provient de Tzortziou et al. (2006).
14
15
Chapitre 2
Optical diversity of thaw ponds in discontinuous
permafrost: a model system for water color analysis
2.1 Résumé Les mares formées par la fonte du pergélisol (lacs thermokarstiques ou mares de
fonte) sont issues de la fonte irrégulière et de l'érosion des sols gelés. Ils sont des sites
importants d'émissions de gaz à effet de serre dans le Nord circumpolaire. En région de
pergélisol discontinu au Nunavik, Canada, les mares de fonte montrent des différences
marquées en couleur, même parmi les mares avoisinantes, allant du blanc au vert, brun et
noir. Pour quantifier ces variations optiques et déterminer leurs mécanismes de contrôle,
nous avons étudié les propriétés optiques inhérentes et apparentes, ainsi que les
caractéristiques limnologiques des mares. Les couleurs étaient bien séparées sur un
diagramme de coordonnées de couleur, avec des valeurs déterminées par la réflectance
spectrale. L'analyse des propriétés optiques et de leurs relations empiriques avec les
substances optiquement actives ont montré que les différences de couleur peuvent être
entièrement attribuées à des variations dans la concentration de deux composantes dans la
colonne d'eau: le carbone organique dissous, dominant l'absorption spectrale, et les
particules non-algales en suspension, contribuant à la diffusion spectrale et à l'absorption.
Cette dernière composante était dominée par les particules de petite taille qui avaient des
propriétés d'absorption et de dispersion (par unité de masse) exceptionnellement élevées.
L'analyse d'une image satellite multi-spectrale de haute résolution de ces mares a montré
que ces deux composantes optiques pourraient être estimées par modélisation multi-variée.
Les résultats indiquent que de la télédétection peut fournir de précieuses informations
synoptiques sur les mares de fonte du pergélisol dans la vaste région subarctique, et
pourrait permettre l'extrapolation des mesures de flux de gaz à effet de serre de l'échelle
locale aux échelles régionale et circumpolaire.
16
2.2 Abstract Permafrost thaw ponds result from the irregular melting and erosion of frozen soils,
and they are active sites of greenhouse gas emissions to the atmosphere throughout the
circumpolar North. In the discontinuous permafrost region of Nunavik, Canada, thaw ponds
show pronounced differences in color even among nearby ponds, ranging from white to
green, brown and black. To quantify this optical variation and to determine its underlying
controlling mechanisms, we studied the apparent and inherent optical properties and
limnological characteristics of the ponds. The pond colors were well separated on a color
coordinate diagram, with axis values determined from above-water spectral reflectance
measurements. Our analyses of optical properties and their empirical relationships with
optically active substances showed that the differences in color could entirely be attributed
to variations in the concentration of two optically active substances: dissolved organic
carbon, which was a major contributor to spectral absorption, and non-algal suspended
particulate matter, which contributed to spectral scattering as well as absorption. The latter
component was dominated by small-sized particles that had unusually high mass-specific
absorption and scattering properties. Analysis of high spatial resolution, multi-spectral
satellite imagery of these ponds showed that these two optically important constituents
could be estimated by multivariate modeling. The results indicate that remote sensing
surveys will provide valuable synoptic observations of permafrost thaw ponds across the
vast subarctic region, and may allow scaling up of local greenhouse gas flux measurements
to regional and circumpolar scales.
17
2.3 Introduction Permafrost thaw lakes and ponds (also referred to as thermokarst systems) are
widespread throughout high latitude regions, and are formed by the localized melting and
erosion of permafrost terrain (Pienitz et al. 2008, and references therein). These aquatic
systems are currently attracting considerable scientific attention related to the impact of
climate change. Permafrost soils contain vast amounts of organic carbon (Ping et al. 2008;
Tarnocai 2009; Zimov et al. 2006) and thaw ponds can serve as biogeochemical hot spots
of microbial and photochemical degradation, converting permafrost carbon to greenhouse
gases (CO2 and CH4), which are then released to the atmosphere (Laurion et al. 2010;
Walter et al. 2006). Despite their wide distribution and importance in regional and global
biogeochemistry, the limnological characteristics of thaw ponds have been little explored.
The discontinuous permafrost region in Nunavik, Canada, contains numerous thaw
ponds (Fig. 2.1). One of their striking features is their variation in color, ranging from white
to green, brown, or black, even among adjacent ponds (Fig. 2.2). This pronounced
difference in color implies variation in their underwater optical conditions, reflecting the
large limnological variability among ponds (Breton et al. 2009). Additionally, it has been
reported that the optical properties of these waters correlate with their rates of greenhouse
gas emission (Laurion et al. 2010). Therefore, detailed information regarding the optical
conditions of thaw ponds may provide insights into their limnology, as well as the regional
biogeochemistry of this high latitude biome.
Color perceived by the human eye can be numerically described by transposing
radiometric measurements onto a colorimetric coordinate system. Standardized conversion
methods have been proposed by the Commission Internationale d'Éclairage (CIE), and
these have been widely applied (Schanda 2007). For natural waterbodies, measurements of
their above-water apparent optical properties (AOPs) can be converted into colorimetric
coordinate values, which may provide a quantitative means to test the relationship between
color and specific limnological variables. This approach offers a potentially useful tool for
resource management (e.g., for assessments of aesthetic value and water quality); however,
it has been applied only to a limited extent to date (Duntley 1963; Oyama and Shibahara
18
2009; Tyler 1964).
Optical conditions reflect the biogeochemical characteristics of the waterbody, and
can therefore be used to monitor changes in the ecosystem (IOCCG 2000). The
relationships among water column constituents, their inherent optical properties (IOPs), and
remotely monitored optical properties are well known for open ocean systems, where there
is a close correlation with chlorophyll a concentrations (Case 1 waters; Morel 1988), and
the resultant models have been applied to evaluate oceanic biogeochemical processes at the
global scale (Arrigo et al. 2008; IOCCG 2000). Such relationships are, however, still poorly
understood in coastal ocean and inland waterbodies, where optical components such as
non-algal suspended particulate matter, terrigenous dissolved organic matter, and
phytoplankton pigments independently vary (Case 2 waters), and development of a
universally applicable robust model is difficult due to the diversity and complexity of
optical conditions (IOCCG 2000). Locally adapted models are, however, proving to be
increasingly useful tools for coastal and inland water resource management, and have been
successfully applied to estimate chlorophyll a (Oyama et al. 2009; Tyler et al. 2006),
phycocyanin (Simis et al. 2005), and suspended particulate matter concentrations (Li et al.
2003). Such models to estimate limnological correlates of CO2 and CH4 concentrations in
thaw ponds (Laurion et al. 2010; Sobek et al. 2005) would provide an approach towards the
scaling up of greenhouse gas measurements in high latitude waters to regional and global
scales. Additionally, the IOP analysis of diverse thaw ponds provides opportunities to
extend existing models, and to expand our knowledge of optical processes in optically
complex Case 2 waters.
In the present study, our overall objectives were to quantify the range of optical
conditions in thaw ponds in the discontinuous permafrost region, and to determine the
underlying mechanisms causing their striking differences in color. Firstly, we quantified
pond colors from above-water spectral reflectance measurements, and analyzed the
influence of limnological conditions on water color. Secondly, we examined the IOPs of
these ponds and evaluated the optical characteristics of optically active substances within
the water column. Finally, we made an initial assessment of satellite imagery as an
19
approach towards estimating key limnological and biogeochemical characteristics of
permafrost thaw ponds.
2.4 Materials and methods
2.4.1 Study site and field sampling The study site was located in a discontinuous permafrost region near
Whapmagoostui-Kuujjuarapik, Nunavik, Canada (55°20' N, 77°30' W; Fig. 2.1), where
there are more than 40 ponds within an area of approximately 0.2 km2 (Fig. 2.2). The ponds
were numbered and referred to as KWK ponds. At the time of sampling, each pond was 10-
30 m in diameter with a maximum depth of 2-3 m. Detailed limnological information for
these waters is given in Breton et al. (2009) and Laurion et al. (2010).
Fourteen ponds representing the full range of limnological conditions (Breton et al.
2009) were sampled between 28 June and 7 July 2007 to characterize AOPs, IOPs, and
their controlling factors. Sampling was conducted between 9:00 and 15:00 local time (EST)
at the center of the ponds, which generally corresponded to the deepest point. Surface water
samples (50 cm below the surface) were taken with Nalgene amber polyethylene bottles
and kept refrigerated (4°C) in the dark until processing within 10 h of collection.
2.4.2 Limnological variables Total phosphorus (TP) was measured by spectrophotometry as described by
Stainton et al. (1977). Water samples were filtered onto Whatman GF/F filters and stored
frozen (-40 to -80°C) until laboratory analysis for chlorophyll a (Chi a). Pigments were
extracted in ethanol and Chi a concentration was determined by fluorometry (Varian Cary
Eclipse spectrofluorometer, Varian Inc., Canada) before and after acidification (Nusch
1980). Samples were also filtered onto pre-combusted, pre-weighed GF/F filters to quantify
suspended particulate matter. Filters were weighed after drying for 2 h at 60°C to determine
total suspended particulate matter (SPMT). Subsequently, filters were combusted at 500°C
for 2 h, and then inorganic suspended particulate matter (SPM_) was quantified. Organic
suspended particulate matter (SPMo) estimates were obtained as the difference between
SPMT and SPMi. Dissolved organic carbon (DOC) concentrations were measured with a
20
Shimadzu TOC-5000A carbon analyzer on filtrates passed through prerinsed cellulose
acetate filters (pore size 0.22 pm, Advantec Micro Filtration Systems, USA).
2.4.3 Apparent optical properties (AOPs) 2.4.3.1 Within water AOPs
9 1
Vertical profiles of spectral downwelling irradiance (Ed(X), pW cm" nm" ) were
obtained with a hyperspectral radiometer system (Satlantic Inc., Canada). The instrument
was manually lowered at approximately 1 cm s"1. Raw radiometer data were processed by
ProSoft (Version 7.7.9., Satlantic Inc., Canada) to obtain the spectral profile of Ed(X) from
352 to 797 nm at 1 nm intervals, and the depth profile at 1 cm intervals. The vertical profile
of photosynthetically active radiation (PAR) of downwelling irradiance (Ed(PAR)) was
calculated by integrating Ed(X) from 400 to 700 nm. Kd(PAR) values, the attenuation
coefficients for Ed(PAR), were then calculated as the slopes of the linear regression of
natural log-transformed Ed(PAR) against depth. This calculation was conducted for the
depth interval from the surface to the point where Ed(PAR) was reduced to 10 % of surface
irradiance.
2.4.3.2 Above-water AOPs
Ed(X) and the spectral upwelling radiance (LU(X), pW cm"2 nm"1 sr"1) were measured
above each waterbody with the hyperspectral radiometer immediately prior to the vertical
profiling. The instrument was held 10-15 cm above the surface and data were recorded for
20 s. The variation of measured downwelling irradiance (Ed(X)) was less than 1 % for most
observations (maximum 2 %), and that of upwelling radiance (LU(X)) was less than 5 %.
These above-surface measurements were then averaged over the sampling period. To
quantify the color of each waterbody, the proportion of surface leaving radiance to incident
irradiance (spectral reflectance, R(Xj) was calculated as:
R(A)= L " { X ) eq.2.1 Ed(X)l2n
21
The Ed(X) values were divided by 2K steradians giving units that were the same for the
denominator and numerator (pW cm"2 nm"1 sr"1), and thereby producing spectra of
reflectance in dimensionless units. This ratio differs from the remote sensing reflectance in
that the surface leaving radiance acquisition was not conducted at the recommended angle
for such a measure and a correction for sky reflection by the sky radiance was not applied
(Mobley 1999; Mueller et al. 2002). However, two lines of evidence suggest that sky
reflection did not have a major effect on the shape of the R(X) spectra. Firstly, in optically
shallower ponds where underwater profiles oîLu(X) were obtained, the LU(X) spectra for just
below the water surface closely matched the above-water LU(X) spectra. Secondly, our
measured R(X) values clearly quantified and separated the differences in pond water color
(which would not be the case, if sky reflection dominated the measurements), which
thereby allowed a quantitative assessment of the influence of limnological variables on
water color.
2.4.3.3 CIE L*a*b* color coordinates
We used the CIE L*a*b* color coordinate system, in which axis L* describes
lightness in the range from zero to 100 and axis a* expresses the green (negative values) to
red (positive values) transition. Similarly, the b* axis describes the blue (negative values) to
yellow (positive values) transition (Schanda 2007). We adopted this color coordinate
system because it had an advantage over other indices, such as the CIE xyY color space (see,
Duntley 1963; Oyama and Shibahara 2009; Tyler 1964), in that it includes a component for
lightness. This greatly differentiated the waterbodies of the present study.
To convert above-water AOP measurements to CIE L*a*b* coordinates, R(X) was
translated into X, Y, and Z coordinates that describe the quantitative response of human
eyes to red, green, and blue light, respectively, using the following equations:
780
X = kYJRWS(X)-x(À)M eq. 2.1 _=380
22
780
Y = kJ^R(Â)-S(Â)-y(Â)AA eq. 2.2 _=380
780
z=k Y , R W ■ s w ■ Z(À)^ eq. 2.3 1=380
where S(X) is the CIE standard illuminant D65, which represents the radiant power
distribution of incident solar radiation giving natural viewing conditions for the ponds
(CIE 1986); x(X), y(X), and z(X) are tristimulus constants, which represent wavelength
dependent responses of human eyes to red, green, and blue, respectively (CIE Cie 1931);
and k is a constant to set Y = 100 for an object whose R(X) is 1 (perfect reflective material),
defined as:
100 780
Z _=380
^S(A.)-y(A)AA eq. 2.3
These summations were calculated at 1 nm intervals. The resultant X, Y, and Z coordinates
were translated into L*a*b* color space as:
L* = 116l — VnJ
16 eq. 2.4.1
a* = 500 < X ^
K*nJ
f Y ^
v-n. eq. 2.4.2
b* = 200 f t r \
J n J
( 7 .
K Z * J eq. 2.4.3
23
where X„, Y„, and Z„ are X, Y, and Z values of a perfectly reflective material, and defined as:
780 X n = k ^ S ( A ) x ( A ) A A eq.2.5.1
_=380
780 Yn = k £ 5(A) • y(A)AA eq. 2.5.2
_=380
780
Z n = k ^ S ( A ) ■ z(A)AA eq. 2.5.3 _=380
2.4.4 Inherent optical properties (IOPs) 2.4.4.1 Absorption coefficient of colored dissolved organic matter
Water samples filtered through Whatman GF/F filters were then re-filtered through
prerinsed cellulose acetate filters (pore size 0.22 pm), and refrigerated in amber glass
bottles until colored dissolved organic matter (CDOM) analysis. The absorbance (A(XJ) of
the filtrate against pure-water prepared by a Barnstead EASY pure II UV/UF system was
measured from 250 to 800 nm at 1 nm intervals in a 1 cm quartz cuvette using a Varian
Cary 100 dual beam spectrophotometer (Varian Inc., Canada). The absorption coefficients
of CDOM (acDOMfX)) were calculated as:
.,. 2.303.4(A) acDOM W = ; e T 2 - 6
where L is the path length of the cuvette (Beer's law). Null point correction was conducted
by subtracting the average of acDOiu(X) from 750 to 760 nm from all spectra. Note that the
range used for the correction here is somewhat short and the spectral curves have not yet
flattened around this range for these highly colored water samples (Mitchell et al. 2000;
2002). The null point correction at this range was, however, adopted to maintain
consistency with other optical measurements, especially with the AC-S (see below). The
24
exponential slope parameters (SCDOM) of the spectra were calculated by non-linearly fitting
the measured value from 300 to 650 nm to the equation:
"CDOM ( A ) = a C D O M (A0 ) e ~ s ™ « i X - ^ eq. 2.7
where Ao is the reference wavelength (Bricaud et al. 1981; Stedmon et al. 2000). SCDOM can
be sensitive to the null point correction value (e.g., at 850 nm) and the fitted wavelength
range (e.g., 305-440 nm (Morris et al. 1995), 375-500 nm (Bricaud et al. 1981)). In our
results, differences in SCDOM values among different methods were small. For example, the
absorption spectra of KWK 7 showed SCDOM of 0.0146, 0.0144, and 0.0144 nm"1 for the
fitted range of 305-440 nm, 375-500 nm, and 300-650 nm, respectively, when the spectra
were corrected with the null point value as described above. For an alternative null point
value at 850 nm, the resultant SCDOM values were 0.0145, 0.0141, and 0.0142 nm'1. The
DOC-specific absorption of CDOM (acDOM*(X)) was calculated as OCDOM(X) divided by
DOC concentration.
2.4.4.2 Absorption coefficients of suspended particulate matter
Suspended particles were collected on Whatman GF/F filters and stored at -40
to -80 °C until optical analysis. The A(X) of suspended particulate matter was measured by
the wet filter technique in the same spectrophotometer as above equipped with an
integrating sphere (Labsphere Inc., USA), following Mitchell et al. (2002). The A(X) of
non-algal particles (NAP) was measured by extracting algal pigments with boiling
methanol (Kishino et al. 1985; Mitchell et al. 2002). The absorption coefficients of
suspended particulate matter and NAP (ap(X) and a^Ap(X), respectively) were calculated as:
2.303_4(A)-r B-V
ap (A) or aNAP(A) = — . ; / eq. 2.8
where T is the filtered area, V is filtered volume, and /. is the path length amplification
factor (Mitchell et al. 2002). B = 2 was used after Roesler (1998) (see, Annexe 3 for further
discussion regarding /?). Null point correction was applied in the same manner as above.
25
The absorption of algal particles (a^(X)) was obtained by subtracting aNAp(X) from ap(X).
Similar to OCDOM(X), the exponential slope of ONAPA), SNAP, was calculated by fitting to the
equation 2.7. Non-linear regression was conducted over the spectral range of 380-730 nm
excluding the ranges 400-480 and 620-710 nm to eliminate the residual pigment absorption
(Babin et al. 2003b; Belzile et al. 2004). The SPMT-specific absorption of NAP (aNAP*(X))
and Chi a-specific absorption of algal particles (a<p*(X)) were also calculated.
Chemical oxidation by the NaCIO method (Ferrari and Tassan 1999; Tassan and
Ferrari 1995) was also conducted on separately prepared filters. The results showed
noticeably lower ONAP(X) values compared to the results of methanol extraction, and we
suspect that NaCIO oxidation changed the optical characteristics of suspended detritus
particles. We therefore did not use this method, and suggest that it be used with caution for
waterbodies rich in strongly colored organic particles.
2.4.4.3 AC-S measurements
The absorption and beam attenuation coefficients of water samples excluding those
of pure-water (at.w(X) and ct.w(X), respectively) were measured using a WET Labs AC-S
in situ spectrophotometer (25 cm path length, WET Labs Inc., USA) on the laboratory
bench. All measurements were completed within 10 h of sampling, and the instrument
configuration and calibration were as described in Sullivan et al. (2006). Samples were
transferred very slowly to the instrument to avoid inclusion of bubbles inside the tubing
system, and once the tubes were filled, the spectra were recorded for 20 s. This
measurement process was repeated three times for each sample. Each reading was averaged
over recorded time, and then all triplicates were averaged to obtain raw readings for each
waterbody.
The AC-S readings were calibrated against pure-water, and were corrected for
temperature and scattering. The pure-water calibration values were obtained immediately
before the sampling period, and subtracted from the raw readings (Pegau et al. 2003;
Twardowski et al. 1999). The temperature correction was applied by using the spectral
correction coefficients obtained specifically for our instrument (Pegau et al. 2003; Sullivan
26
et al. 2006; Twardowski et al. 1999). The null point correction, subtracting the absorption
coefficient at the longest wave band measured by the instrument (751.7 nm), was adopted
as the scattering correction method for our study (Pegau et al. 2003). Although some
authors recommended the proportional correction, which takes into account the wavelength
dependency of measurement error of the reflective tube absorption meter (Kirk 1992; Pegau
et al. 2003; Zaneveld et al. 1994), we found that this often yielded lower values compared
to absorption measurements obtained by conventional spectrophotometry. The at.w(X)
values were always lower than the sum of ap(X) and acDOM(X), and often even lower than
acDOMfX), and so this method was not applied. Similarly, an approach suggested by
Gallegos and Naele (2002) to obtain scattering correction coefficients did not provide
reasonable estimates (the linear equation system gave negative values), and could not be
applied in our study. Finally, the absorption coefficients of pure-water (aw(X)) (Pope and
Fry 1997) were added to at.w(X) to obtain total absorption of the waterbodies (at(X)).
The total scattering coefficients by suspended particles (bp(X)) were obtained by
subtracting at.w(X) from c,.w(X). SPMT-specific scattering coefficients (bp*(X)) were obtained
by dividing bp(X) by SPMT. To characterize the shape of bp(X) spectra, the observations
were nonlinearly fitted to the hyperbolic model as:
bp(A) = bp(555)x * A V"
v555. eq. 2.9
where n is the hyperbolic exponent (Babin et al. 2003a; Belzile et al. 2004). The particle
single scattering albedo (cop(X)) was obtained as:
m 7 , W a>AX) = - E -— eq.2.10
where cp(X) is the beam attenuation coefficient by particles that was obtained by subtracting
aCDOM(X) from c,.w(X).
27
2.4.4.4 Absorption coefficient of fine particles
We found that filtrates of GF/F filters were often visibly turbid as a result of high
concentrations of fine particles, and these particles were removed by subsequent filtration
using the cellulose acetate filters. There were also significant differences between a,.w(X)
obtained with the AC-S and the sum ofap(X) plus OCDOM(X). This difference was defined as
the absorption coefficient of fine particles (afme(X)) in this study.
2.4.4.5 Validation of AC-S related measurements
The initial value of both the absorption and attenuation measurements made by the
AC-S often exceeded the instrumental upper detection limit (10 m"1). Therefore, samples
were diluted by measured volumes of pure-water until the values reached less than 10 m"1
(dilution factor of 2 to 7). Even though dilution was the only feasible solution to measure
samples from extremely colored and turbid systems by the instrument, the process might
affect the water optical characteristics, for example by the flocculation of particles and
precipitation of dissolved matter. The effect of dilution was, therefore, tested by two
different approaches.
Firstly, we compared measured values of diluted and undiluted samples at a longer
wavelength (at.w(604.1) and c,.w(604)), where observed raw values were always below the
detection limit for at.w(604.1) (four highly turbid samples showed ct-w(604) > 10 m"1 even at
this wavelength). The average of the absolute difference between two measurements was
0.13 m"1 (5.9 %) and 0.13 m"1 (3.3 %) for absorption (n = 14) and attenuation (n = 10),
respectively. Secondly, we compared absorption measurements obtained by the two
different photometric methods. Surface water was sampled in four ponds (KWK 6, 16, 23,
38) in July 2008. The at.w(X) was determined by the AC-S with dilutions of one to six, and
(*CDOM(X) and ap(X) were measured by the laboratory bench top spectrophotometer as
described above. In addition, filtrate from GF/F filters was re-filtered through cellulose
acetate filters and the absorption coefficient of materials collected on the filters was
determined using the same method as for normal particulate absorption analysis (i.e., ap(X)
quantification), as described above (Ferrari and Tassan 1996; Gallegos 2005). Then at.w(X)
and the sum of all bench top spectrophotometer measurements were compared. The
28
obtained spectra showed close correspondence with respect to their shape (Supplement Fig.
2.1a). The average of the absolute difference at 440 nm was 0.52 m"1 (6.7 %). These two
sets of tests indicated that the effect of dilution on the measurement by AC-S was small
within the present study.
The measurement of afine(X) was also validated with data obtained in four ponds
sampled in 2008. We compared aflne(X) determined as difference between at.w(X) and the
sum of ap(X) plus acDOM(X) with the absorption coefficients of materials collected on the
0.22 pm cellulose acetate filters measured as above. The absolute difference was 0.52 m"1
(41.6 %) on average, and the spectra showed differences in shape especially at shorter
wavelengths of samples with low to medium concentrations of fine particles (Supplement
Fig. 2.1b). This difference resulted from two intrinsically different spectrophotometer
designs. For the AC-S measurement, the error may be exacerbated by the spectral
dependency of scattering and its influence on reflective tube absorption measurements
(Kirk 1992; Zaneveld et al. 1994). Our measurements of afme(X) therefore cannot be
considered a precise estimate of spectral absorption by fine particles. However, differences
between at.w(X) and the sum of OCDOM(X) and ap(X) provided an index of site-to-site
variability in the contribution of fine particles to total absorption.
2.4.5 Remote sensing analysis A remote sensing algorithm was developed and evaluated for these thaw ponds
using limnological data and satellite imagery obtained in July 2006. Thirty-four ponds were
sampled, and SPMT and DOC were measured with the same methods as described above.
We used a high spatial resolution multispectral image acquired by QuickBird (DigitalGlobe
Co., USA) during clear sky conditions on July 7, 2006. The satellite provides four spectral
bands: blue (450-520 nm), green (520-600 nm), red (630-690 nm), and IR (760-900 nm).
An atmospheric correction was conducted with Geomatica® (PCI Geomatics, Canada),
using the ATCOR2 module for standard subarctic summer atmospheric conditions.
Reflectance of each pond was obtained as an average of pixels over the water and
excluding those overlapping the adjacent land. (14-105 pixels depending on pond size,
where each pixel was 2.4 x 2.4 m). It is possible that there was some adjacency effect;
29
however, visual inspection of the satellite image indicated that this effect was minor
because there was no obvious gradient in color of pixels towards the center of each pond.
Bottom effects can also be a problem for satellite remote sensing of shallow waterbodies;
however, this was negligible in the present study because all the ponds were optically deep,
with sufficiently high dissolved color and suspended particles to obscure the bottom. For
modeling and validation, sampled ponds were randomly separated into calibration (n = 24)
and validation (n = 10) data sets.
Color inversion for water quality estimation of Case 2 inland waters is relatively
difficult compared to that of oceanic Case 1 waters due to the complexity of optical
processes in the water column (IOCCG 2000). To overcome this difficulty, various
attempts have been made to apply multivariate statistical analyses instead of simple or
multiple linear regressions (e.g., Flink et al. 2001; Svab et al. 2005). We adopted the
approach of canonical correlation analysis (CCA), a multivariate statistical technique that
examines interrelationships between two sets of variables (multiple Ys against multiple Xs)
(Johnson and Wichern 2007). It also can be conducted on variable sets with one Y against
multiple Xs, providing coefficients for Xs to align with variation of Y. The latter method of
CCA has been successfully applied to multispectral satellite data for terrestrial systems
(e.g., Cohen et al. 2003; Heiskanen 2006). We applied this type of CCA to limnological
variables (Y = SPMT or DOC) against four reflectance bands of QuickBird data (Xs). The
relationships between the limnological variables and the obtained canonical scores were
linearly fitted. The developed models were then evaluated using the validation data set.
2.4.6 Statistical analyses All statistical analyses were conducted using the R language and environment for
statistical computing (R Development Core Team 2009). Statistical significance was tested
at the usual p = 0.05 level, and exact p values are given for probabilities above 0.01.
Pearson's product moment correlation coefficients (r) are given where the significance of
correlation between two variables was tested. In turn, r squared (r2) values are given for
linear regression models, along with the associated regression coefficients (slope and
30
intercept) to quantify the relationship between optical properties and optically active
variables of studied waterbodies.
2.5 Results 2.5.1 Ambient conditions
The thaw ponds in this region encompass a broad range of colors: green, black,
brown , and white (Fig. 2.2). Most were thermally stratified with a thermocline at around 1
m depth. The mean surface temperature was 13.7°C (range: 10.4-18.6°C, n = 13) at the
time of sampling. TP varied from 33.6 to 108.9 pg L"1 with a mean of 60.4 pg L"1 (Table
2.1). DOC ranged from 3.1 to 10.5 mg L"1 with a mean of 7.7 mg L" . The mean value of
Chi a was 6.2 pg L"1. One site (KWK 11) showed exceptionally high Chi a (23.8 pg L"1),
but the rest of the observations fell within the range 1.4 to 10.3 pg L" . There were large
variations in suspended particulate matter concentrations: SPMT ranged from 5.0 to
42.4 mg L"1 (mean: 16.8 mg L"1), SPMi from 2.1 to 36.7 mg L"1 (mean: 13.0 mg L"1), and
SPMo from 1.6 to 7.1 mg L"1 (mean: 3.8 mg L"1). The relationship between SPMi and
SPMo was significant (r = 0.55, p = 0.04). These variables showed much stronger
relationships when the outlier with exceptionally high Chi a (KWK 11) was removed (r =
0.87, p < 0.01). The mean of %SPMi contribution to SPMT was 71 % with a coefficient of
variation (CV) of 26 %. Chi a did not correlate with SPMT (r = 0.04, p = 0.89), but did
correlate with SPMo(r = 0.75, p < 0.01); however, the relationship was weaker without
KWK 11 (r = 0.60, p = 0.05).
2.5.2 AOPs Kd(PAR) ranged from 1.70 to 7.50 m"1 with a mean of 3.94 m"1. Among the
observed limnological variables, Kd(PAR) was strongly correlated with SPMT (r = 0.90, p <
0.01) and moderately with DOC (r = 0.62, p = 0.03), but not with Chi a (r = 0.18, p = 0.58).
Forward stepwise regression analysis of these data selected SPMT and DOC as explanatory
variables: Kd(PAR) = 0.11 SPMT + 0.25 DOC (r2 = 0.89; both SPMT and DOC were
significant at p = 0.05, and the intercept was not significantly different from zero).
31
The R(X) spectra showed large variations in shape that clustered according to the
visual appearance of the ponds (Figs. 2.3a-d). Green (KWK 6; see, Fig. 2.2) or white
(KWK 38 and 39) ponds showed relatively high R(X) that peaked around 550 nm (Fig. 2.3a).
Black ponds (KWK 2 and 11) showed low R(X) curves with flat spectra (Fig. 2.3b). Brown
ponds showed intermediate levels of R(X) peaking around the orange to red part of the
spectrum (Fig. 2.3c). Light brown ponds had high levels of R(X) peaking around 650-700
nm (Fig. 2.3d).
The CIE L*a*b* color coordinate values, converted from R(X) spectra, well
described and separated the pond colors (Fig. 2.4). Three ponds that had green to white
color were plotted in the green-yellow quadrant (negative a* and positive b*) with
relatively high b* values, and the rest of the ponds clustered in the red-yellow quadrant
(positive both in a* and b*). Among the latter ponds, two black ones showed low b* values.
Pond KWK 11 had especially low values in both a* and b*. The majority of brown ponds,
except for KWK 33, showed intermediate b* values, while the light brown ponds had high
b* values.
The a* axis values were strongly correlated with DOC (r = 0.73, p < 0.01), but not
with SPMT (r = 0.48, p = 0.08), Chi a (r = 0.21, p = 0.51), or the DOC:SPMT ratio (r
= -0.11, p = 0.73). In turn, b* showed a significant negative correlation with the
DOC:SPMT ratio (r = -0.70, p = 0.01), but not with DOC (r = -0.35, p = 0.27) or SPMT (r =
0.46, p = 0.10). The correlation between b * and Chi a was statistically significant (r = -0.61,
p = 0.04); however, this result was largely driven by an outlier that showed exceptionally
high Chi a and low b* (KWK 11). Without the outlier, the relationship was not significant
(r = -0.03, p = 0.92). The lightness index (L*) ranged from 16.8 to 55.0 with a mean of 36.0,
and was significantly correlated with b* (r = 0.84, p < 0.01). This index was negatively
correlated with the DOC:SPMT ratio (r = -0.57, p = 0.05), but not with DOC (r = -0.14, p =
0.66), SPMT (r = 0.49, p = 0.07) or Chi a (r = -0.54, p = 0.07).
32
2.5.3 IOPs 2.5.3.1 Absorption coefficient of colored dissolved organic matter
OCDOM(X) spectra showed wide variations among ponds (Fig. 2.5a). The absorption
coefficient at 440 nm (acDOM(440)) ranged from 1.654 to 9.333 m"1 (mean: 5.461 m ; Table
2.2). In contrast, SCDOM varied over a narrow range, from 0.0144 to 0.0159 nm"1 (mean:
0.0150 nm"1). The mean value of aCDOM*(440) was 0.743 m2 g"1 (0.491-0.962 m2 g"1). The
CICDOM(440) values and DOC were strongly correlated (r = 0.82, p < 0.01; Fig. 2.6, circles).
The relationship was similar to that reported for the data collected in 2006 by Breton et al.
(2009) (n = 36, r = 0.85, p < 0.01; Fig. 2.6, triangles). Regression analyses were conducted
on the pooled data (n = 48). Both linear and power models were highly significant and
indicators of goodness-of-fit were similar between the two models. The linear model was
CICDOM(440) = 0.742 DOC (r2 = 0.65, p < 0.01, Akaike's information criteria (AIC) = 153.7,
regression mean square error (RMSE) = 1.15, intercept not significantly different from
zero), and the power model was aCDoM(440) = 0.495 DOC1 129 (p < 0.01, AIC = 154.0,
RMSE =1.16).
2.5.3.2 Absorption coefficient of non-algal particles
CINAP(X) spectra also showed a wide variation among the studied ponds (Fig. 2.5b
and Table 2.2). aNAP(440) values ranged from 0.464 to 2.240 m"1 with a mean of 1.049 m"1.
In contrast, the SNAP varied little (range: 0.0096-0.0136 nm"1) and most ponds (n = 10) had
values ± 0.0006 nm"1 from the mean (0.0111 nm"1). aNAP*(440) varied three-fold (Table
2.2), with the majority of values in the range 0.060 to 0.093 m2 g"1. The ONAP(440) values
showed a strong linear relationship with SPMT: ONAP(440) = 0.042 SPMT + 0.342 (r2 = 0.90,
p < 0.01; Fig. 2.7).
2.5.3.3 Absorption coefficient of algal particles
Spectral a<p(X) consistently showed a bimodal distribution with some variation in
shape (Fig. 2.5c). There was a six-fold variation in a<p(440) with a mean of 0.415 m"1 (CV:
61 %; Table 2.2), and 10 of the 14 observations were below 0.5 m". The range oîaa,*(440)
was 0.044 to 0.164 m2 mg"1 (mean: 0.086 m2 mg"1, CV: 47 %). The a0(44O) values and
Chi a were strongly related (Fig. 2.8). Both linear and power models were significant and
33
the goodness-of-fit was similar for the two models: a<p(440) = 0.038 Chi a + 0.167 (r2 =
0.87, p < 0.01, AIC = -18.39, RMSE = 0.096); a*(440) = 0.136 Chi a0636 (p < 0.01, AIC
= -18.40, RMSE = 0.096).
2.5.3.4 Absorption coefficient of fine particles
ctfme(X) showed the greatest variation of any of the component spectra (Fig. 2.5d).
The value of afine(440) ranged from zero to 8.865 m"1 (mean: 2.180 m"1, CV: 127 %; Table
2.2). The correlation analysis showed that this variable had a strong significant correlation
with SPMT (r = 0.97, p < 0.01) and aNAp(440) (r = 0.93, p < 0.01), although these variables
were obtained on GF/F filters which did not retain the fine particles.
2.5.3.5 Total absorption
There was a more than five-fold variation in at(440) (Fig. 2.9 and Table 2.2), and
the contribution of each absorbing component to the total absorption was also variable
among ponds (Table 2.3). On average, CDOM accounted for more than half of the
absorption at 440 nm, but varied from 30 to 85 %. The mean NAP contribution to total
absorption at 440 nm was 12 % with little variation (CV: 25 %). Algal particles were
always a minor constituent, ranging from 2 to 14 % with a mean of 5 %. The contribution
of fine particles varied between 0 and 53 % with a mean of 19 %. For half the ponds, fine
particles accounted for more than 20 % of total absorption. Multiple linear regression
analysis was conducted to determine the relationships between total absorption and
limnological variables, and forward stepwise regression analysis selected SPMT and DOC
as explanatory variables, giving the model: a,(440) = 0.288 SPMT + 0.650 DOC (r2 = 0.90;
both SPMT and DOC were significant at p = 0.05 level, intercept not significantly different
from zero).
2.5.3.6 Scattering
The value of bp(555) ranged from 1.957 to 21.818 m"1 (mean: 7.345 m"1). At 440 nm,
the range was 2.595-32.685 m"1 with a mean of 10.832 m"1. The bp(440) values were
significantly correlated with variables related to SPMT (r = 0.97, p < 0.01), ONAP(440) (r =
0.93, p < 0.01), and afme(440) (r = 0.99, p < 0.01), but not with algal particles: Chi a
34
(r= -0.03, p = 0.90) and a0(44O) (r = 0.33, p = 0.25). bp*(555) ranged from 0.300 to
0.688 m2 g"1 with a mean of 0.441 m2 g'1. Most of the scattering spectra showed a strong
wavelength dependency represented by a power model (Fig. 2.10a). The spectrum of
KWK 11 was the sole exception showing a departure from the hyperbolic curve in the
range coinciding with photosynthetic pigment absorption (Fig. 2.10b). The range of the
hyperbolic exponent (n) was from 0.831-1.861, with a mean of 1.545 excluding KWK 11.
The exponent significantly correlated with variables associated with suspended particulate
matter concentrations: SPMT (r = 0.57, p = 0.02) and SPMi (r = 0.64, p = 0.01), and not
with those representing dissolved organic carbon: OCDOM(440) (r = 0.25, p = 0.39) or DOC
(r = 0.16, p = 0.63). The single scattering albedo at 555 nm ranged from 0.814 to 0.921
(mean: 0.856), and the value decreased toward shorter wavelengths (Fig. 2.11).
2.5.4 Remote sensing analysis The thirty-four ponds sampled in 2006 showed the same wide variety of conditions
as those sampled in 2007, with DOC ranging from 3.8 to 12.9 mg L"1 and a mean of 8.6
mg L"1 (CV: 23 %), and SPMT from 1.8 to 55.2 mg L"1 with a mean of 19.8 mg L"1 (CV:
82 %). There was no significant correlation between these two limnological variables (r =
0.09, p = 0.61). The reflectance spectra showed relatively high values in the IR band (Fig.
2.12). This may suggest that there was some influence from sun and sky glint or from
adjacent land effects that are difficult to remove due to the limitations in current
methodology. However, the shape of the reflectance spectra showed large differences
among the observed ponds, and contained sufficient information for analysis of their wide
variation in limnological conditions despite the uncertainties. The CCA provided the
standardized coefficients to calculate canonical scores as shown in the supplement Table
2.1. Linear regression models between obtained canonical scores and limnological variables
were significant for both SPMT (r2 = 0.78, p < 0.01) and DOC (r2 = 0.35, p < 0.01). For the
validation data set, the estimated and measured values from the regressions agreed well for
both SPMT (r2 = 0.88, p < 0.01, RMSE = 6.27 mg L"1, Slope = 1.02, intercept not
significantly different from zero; Fig. 2.13a) and DOC (r2 = 0.74, p < 0.01, RMSE =
1.20 mg L , Slope = 1.05, intercept not significantly different from zero; Fig. 2.13b).
35
2.6 Discussion 2.6.1 Limnological conditions
The striking variability in color among permafrost thaw ponds (Fig. 2.2) implies
large variations in their limnological properties, and our analyses provided clear evidence
of such variability. DOC concentration, which would mostly originate from surrounding
soils and vegetation affected by permafrost (Breton et al. 2009), varied by a factor of three.
Suspended particulate matter showed eight-fold variation. It was dominated by non-algal
organic and inorganic suspended particulate matter (marine clays) in similar proportion
among ponds, and the contribution of algal particles was minor. This non-algal suspended
particulate matter also likely originates from the surrounding and underlying soils, which
are rich in clays and silts (Calmels et al. 2008b). Chi a concentrations differed by a factor
of 17; however, they were generally low, particularly by comparison with TP
concentrations. The controlling mechanisms for this low Chi a to TP ratio are at present
unknown, and future research is needed to address this observation. The wide variation in
limnological conditions of nearby ponds may reflect differences in local topography,
hydrology, soil characteristics, and vegetation causing differential erosion, together with
different developmental stages of the thermokarst ponds (Bouchard et al. 2011; Calmels et
al. 2008a; Pienitz et al. 2008). To better understand mechanisms underlying this variability
in nutrients, suspended particles, and DOC inputs, more detailed studies regarding
hydrology and geomorphology.
2.6.2 Above-water AOPs and color coordinates Our above-water spectroradiometric measurements captured the variation in visual
appearance of the ponds (Fig. 2.3), and their converted colorimetric values provided an
appropriate analytical tool to numerically evaluate this variation in perceived color (Fig.
2.4). Further, the relationship between color coordinate values and limnological variables
helped to elucidate the underlying mechanisms controlling this observed variation. These
results support the use of above-water AOPs to describe the color of Case 2 waters. Such an
approach continues to offer much potential in water color evaluation and prediction for
water resource management.
36
Our analysis of these optically diverse thaw ponds showed that color is determined
by the specific mixture of DOC and SPMT. More precisely, the green to red color transition
(a* axis) was controlled by DOC concentrations that represent CDOM absorption
(discussed below). Due to the absorption spectral shape (Fig. 2.5a), blue light was strongly
absorbed even at low DOC concentrations, while green light was still perceivable. Along
with the increase in DOC, green light was also strongly absorbed, and water color shifted
toward red. The magnitude of transition on this axis was small relative to that of other axes;
however, the observed values were distributed in both the green (negative a*) and red
(positive a*) quadrants. Thus, this transition considerably influenced the pond color as
perceived by the human eye.
The blue to yellow transition (_>*) and lightness (L*) were controlled by the
DOC:SPMT ratio, which may represent the ratio of dissolved absorption and scattering (see
below). Light scattering contributed to the upward flux of light back to surface and
atmosphere, while absorption did the opposite. The balance of these two functions dictates
the amount of surface leaving radiation, i.e., the perceived lightness of the waterbodies.
This function was pronounced at the yellow to red end of the spectrum where absorption
was low relative to the blue-green waveband; thus, yellowness as well as lightness
increased as a function of decreasing DOC:SPMT.
2.6.3 IOPs 2.6.3.1 Absorption coefficient of colored dissolved organic matter
The observed range in CDOM absorption was surprisingly wide, especially for such
closely located ponds. Detailed analyses on optical properties suggest that the wide
variation in CDOM would be due mostly to quantitative differences in the DOC pool, not
to qualitative differences. SCDOM, a variable characterizing the shape of CDOM absorption
curves that is often considered as an indicator of DOC composition in natural waters
(Stedmon et al. 2000; Zepp et al. 2008), showed relatively low values with narrower range
compared to the values in other natural waters (Markager and Vincent 2000). Also, the
range in OCDOM*(440), another variable indicating DOC composition (Morris et al. 1995),
was much narrower than that reported in other inland waterbodies (Bukaveckas and
37
Robbins-Forbes 2000; Morris et al. 1995; Watanabe et al. 2009). The observed low
variability in these variables despite wide variation in absolute absorption indicates a strong
similarity in the chemical composition of the DOC pool among ponds. This result is
consistent with DOC fluorescence analysis by Breton et al. (2009) suggesting that the DOC
in these ponds was of similar terrigenous origin.
The compositional similarity and optical resemblance of the DOC pool in these
permafrost thaw ponds increases the prospect of inferring DOC concentrations from optical
information. Significant correlations between DOC and CDOM have been previously
documented (e.g., Bukaveckas and Robbins-Forbes 2000; Cole et al. 2002) and such
relationships can be used to estimate CDOM (e.g., Pienitz and Vincent 2000) and to predict
the attenuation of ultraviolet radiation (UVR) (e.g., Morris et al. 1995; Scully and Lean
1994) from DOC concentration, or vice versa. These applications, however, would be valid
only when the composition of the DOC pool is similar among waterbodies (Watanabe et al.
2009). In our study sites, a strong DOC-CDOM relationship (Fig. 2.6) was observed,
although the lack of observations at low concentrations precludes the application of a
power model as in Bukaveckas and Robbins-Forbes (2000). The clear relationship within
the observed range, as well as compositional similarity of the DOC pool noted above
suggest that optical measurements, including spectrophotometry and remote sensing, would
be a useful approach to estimate DOC and UVR attenuation. Algorithms from this region,
however, should be applied with caution elsewhere, given the observed differences in DOC
composition in other thaw pond regions (Breton et al. 2009).
2.6.3.2 Absorption coefficient of non-algal particles
Most thaw ponds contained high concentrations of suspended particulate matter,
and NAP absorption values lay at or above the range that has been previously observed in
coastal marine waters (Babin et al. 2003b) but lower than that of the upper bound for
eutrophic lakes (Sun et al. 2009; Zhang et al. 2010). In addition, the specific absorption
characteristics of the particles found in thaw ponds may contribute to their high ONAP(440)
values. The slope of aNAp(440)-SPaW.j model (0.042) was more than 35 % higher than that
reported elsewhere (0.031 (Babin et al. 2003b), 0.0235 (Bowers et al. 1996), 0.0214
38
(Binding et al. 2008)). Also, the mean aNAp*(440) (0.073 m2 g"1; Table 2.2) was higher than
found elsewhere (0.05 (Belzile et al. 2004), 0.041 (Babin et al. 2003b), 0.040 (Binding et al.
2008)). These results suggest that NAP in thaw ponds have unusually high mass-specific
absorption properties. Three possible explanations may account for this. Firstly, particle
size is an important determinant of absorption characteristics of suspended particles in
natural waters (Stramski et al. 2007). The surrounding permafrost soils, likely to be the
main source of suspended particles for these ponds, contain mostly fine clay particles
(Calmels et al. 2008b), and the particle size distribution analysis showed that the ponds
may contain considerable amounts of sub-micrometre particles (see, Annexe 2).. Because
of their small size, these particles may cause higher absorption due to the particle package
effect (Stramski et al. 2007). Secondly, the composition of NAP is also an important
determinant of ONAP(X) characteristics (Babin et al. 2003b; Binding et al. 2008). Absorption
by the organic component of NAP, including both algal and terrigenous detritus, is
incorporated into aNAp(X). In our study sites, organic component quantified as SPMo largely
accounted for SPMT (30 % on average), and absorption characteristics of this fraction may
contribute to the high specific absorption of NAP in thaw ponds. Finally, CDOM adsorbed
onto NAP might cause higher absorptions. SNAP values were negatively correlated with
OCDOM(440) (r = -0.59, p = 0.03), indicating that the concentration of CDOM might have an
influence on ONAP(X) absorption spectra.
Although NAP showed unusually high mass-specific absorption and large variation
in absorption among ponds, its mass-specific absorption (aNAP*(440)) and shape of
absorption spectra (SNAP) showed little variation among ponds. Also, there was a strong
relationship between aNAp(440) and SPMT for these waterbodies. These results suggest
compositional similarity and optical resemblance of NAP of the ponds in the studied area.
Similar to OCDOM(X) discussed above, the variation in ONAP(X) would be due to the
quantitative difference of particles in the water column, suggesting that optical estimation
of SPMT would be feasible, as for DOC. Furthermore, there was a strong correlation
between SPMT and TP in the ponds (r = 0.84, p < 0.01, n = 33, 2006 data), implying that an
optical estimation of nutrient content (TP) may also be possible (Kutser et al. 1995).
39
2.6.3.3 Absorption coefficient of algal particles
The absorption by algal particles showed a wide range of values among ponds (Fig.
2.8), and the increase of a<p(440) per unit Chi a was higher (i.e., the relationship had a
steeper slope) relative to those previously reported in freshwater systems (Perkins et al.
2009; Sun et al. 2009; Zhang et al. 2010) and oceanic systems (Bricaud et al. 1995;
Matsuoka et al. 2007). Furthermore, the mean value of a<p*(440) was above previously
reported values (Bricaud et al. 1995; Dekker et al. 2002; Zhang et al. 2010). These
observations imply that the pigment composition in thaw pond phytoplanktonic
communities differs greatly from that of other natural waterbodies. However, it has been
reported that particulate humic color extracted by methanol may induce overestimation of
a<p(X) in waterbodies with high concentrations of NAP (Gallegos 2005). Additionally, the
observed range of a0(44O) and Chi a in the present study may be too narrow to
appropriately evaluate the model and a&*(440), and more detailed analysis with larger data
sets is required.
2.6.3.4 Absorption coefficient of fine particles
In the present study, we defined fine particles as those passing through GF/F filters
and then caught on cellulose acetate filters. The size range of these particles is therefore
assumed to be 0.2-0.7 pm. Suspended particles in this size range may have important
implications for aquatic ecosystems by causing light limitation of algal growth in highly
turbid waterbodies (Knowlton and Jones 2000). The optical property of fine particles
suspended in natural waters, however, has been little examined (Gallegos 2005; Stramski et
al. 2007; Stramski and Wozniak 2005). In our study sites, fine particle absorption was
always observed except in ponds with the lowest suspended particulate matter
concentrations. Most of these fine particles were probably inorganic, and derived from the
surrounding soils rich in clays (Calmels et al. 2008b). Phytoplankton particles would have
had little contribution to the measured absorption values because no Chi a peaks at 440 or
680 nm were apparent in the spectra. The mass-specific absorption of fine particles has
important implications for future parameterization of optical properties of these ponds, and
may be higher than that of larger particles due to size-dependent effects (Stramski et al.
2007). This could not, however, be appropriately quantified from the current data set
40
because of uncertainty in precision of af,m(X) estimation due to the methodological
limitations discussed above, and the absence of dry weight measurements of the fine
particle fraction. Careful evaluation of analytical methods for measuring afme(X) spectra
(Ferrari and Tassan 1996; Gallegos 2005) is required in future studies of such fine particle
containing systems.
The influence of fine particles on estimation of SPMT by optical measurements in
the study sites can be regarded as a minor factor, although the detailed optical
characteristics of this component remain unknown. The strong linear relationships between
ctfme(440) and measures of suspended particles caught on GF/F filters, SPMT and ONAP(440),
showed that the concentration of fine particles increased in tandem with larger particle
concentrations and absorption. This result suggests that the influence of fine particles can
be accounted for by measuring larger particles with conventional methods.
2.6.3.5 Total absorption
As a result of the large variation in each absorbing component, at(X) showed a
five-fold variation (Table 2.2), and the proportional contribution of each component to at(X)
also varied greatly (Fig. 2.9 and Table 2.3). Overall, CDOM was the dominant absorption
factor in most ponds, and was only a lesser component in those ponds where fine particles
dominated. The absorption by suspended particulate matter caught on glass fiber filters,
measured as ONAP(X) and aq,(X), played a lesser role in the spectra of all ponds. In particular,
the contribution of algal particle was minor (less than 15 % for all ponds). In contrast to
absorption by those larger particles, fine particles played an important role in ponds with
high suspended particulate matter concentrations (Fig. 2.9d).
Estimation of Chi a by remotely sensed data is well developed and commonly
applied to oceanic Case 1 waters, where it is a major light-absorbing component as well as
highly correlated with other absorbing components (Morel 1988). This approach, however,
will not be successful in our study sites, as algal particles exhibited a negligible role in the
absorption spectra and varied independently of other optically active components. Instead,
the two dominant absorbing factors, CDOM and fine particles, were strongly related to
41
DOC and SPMT concentrations, respectively. Therefore, the estimation of these two
variables from optical data should be feasible for this system.
2.6.3.6 Scattering
Few data are available for spectral scattering coefficients in freshwater ecosystems
and the present study increases the availability of such information. There was large
variation in these values among ponds, and within the range previously observed by similar
instruments (AC-9, WET Labs Inc., USA) in coastal (Babin et al. 2003a; Doxaran et al.
2009) and riverine systems (Gallegos 2005). The characteristics of bp(X) spectra and their
relationship with limnological variables well portrayed the characteristics of scattering
components at our study sites, which is consistent with the results of the particulate
absorption analyses, as discussed above. Firstly, the correlation analyses and the shapes of
scattering spectra, well represented by the hyperbolic model described by the equation 2.9,
indicated that non-algal suspended particulate matter dominated scattering, and that algal
particles played a minor role, as in coastal marine waters (Babin et al. 2003a). The only
exception was a pond with high Chi a concentrations (KWK 11) where the influence of
algal pigment absorption on the scattering spectra was observed. Secondly, the obtained
hyperbolic exponents (mean: 1.545) were higher relative to previously reported values in
coastal waters (Doxaran et al. 2009). A theoretical study suggested that this exponent
represents the size distribution of scattering particles when the distribution is assumed to
follow the Junge type power law, and that the exponent increases as a function of the
proportion of small size particles in the total scattering particle population (see, Babin et al.
2003a, and references therein). In the present study, the values did indeed show a strong
positive correlation with the proportion of afme(440) in the total particulate absorption
(at.w(440) - OCDOM(440); r = 0.82, p < 0.01). This result is consistent with the theoretically
simulated particle size (see, Babin et al. 2003 a, Fig. 4), and implies the domination of small
size particles in thaw pond scattering. Finally, the spectral shape of cop(X) showed a strong
influence of absorption at the blue end of scattering spectra. In summary, scattering
processes in the studied systems were dominated by small non-algal particles having high
absorption characteristics.
42
The SPMT-specific scattering coefficients were similar to those for coastal waters
and to theoretically predicted values for inorganic particles (Babin et al. 2003a; Doxaran et
al. 2009). However, these values must be interpreted with caution. The fraction of particles
accounted for by our two measurements differed; bp(X) accounted all particles in the water
column but SPMT was assumed to include only particles larger than 0.7 pm. Fine particles
likely make a substantial contribution to the total mass of suspended particulate matter.
Inclusion of dry weight of this size fraction into the bp*(X) calculation would result in lower
mass-specific values especially for highly turbid systems, which may affect subsequent
optical modeling. Turbid systems therefore require a much closer analysis of particle size
distribution, as well as attention to the optical characteristics of fine particles.
2.6.4 Remote sensing analysis Our results suggest that high spatial resolution multispectral satellite imagery may
be a useful tool for estimating limnological properties of permafrost thaw ponds, which are
abundant and widely distributed over remote and poorly accessible areas. Remote sensing
techniques have already been applied to inland waterbodies where small-scale spatial
heterogeneity has important implications for water resource management (e.g., Oyama et
al. 2009; Tyler et al. 2006). The present results also show an application to small-scale
systems, specifically inland waterbodies with high inter-system variation in limnological
conditions, as small as 10 m in diameter. The goodness-of-fit of the CCA modeling
indicates that two optically important constituents, DOC and SPMT, can be derived from
satellite imagery using multivariate statistical modeling. These two variables would provide
useful information not only on optical processes but also on the biogeochemical cycles of
the systems. DOC is of special interest since it has important implications for microbial
processes, including greenhouse gas emissions from these waterbodies (Laurion et al. 2010;
Sobek et al. 2005). Furthermore, both optically active constituents can regulate
photosynthetic production by strongly attenuating light in the water column (Karlsson et al.
2009; Knowlton and Jones 2000; Vincent et al. 1996). Therefore, remote sensing
techniques have the potential to estimate the significance of biological activity in thaw
ponds within the global carbon cycle (P/R balance, greenhouse gas production, and
underwater UVR exposure). For future applications of such models, validation is required
43
based on larger data sets covering disparate geographic areas with potentially different
limnological and optical characteristics.
2.7 Conclusions Thaw ponds in the discontinuous permafrost region of Nunavik, Canada showed a
wide variety of optical conditions that were strikingly apparent as differences in water color.
This variation is largely the result of variable combinations of dissolved organic carbon and
non-algal suspended particulate matter concentrations in the water column. Analyses of
IOPs of the waterbodies showed that compositions of these optically important components
were similar among ponds; i.e., the optical diversity was derived from their differences in
concentration but not composition. Analysis of high spatial resolution multispectral satellite
imagery of these ponds showed that these two optically important constituents could be
remotely estimated by multivariate modeling. Given the importance of these two
constituents in aquatic biogeochemical cycles, remote sensing surveys will provide
valuable synoptic observations of permafrost thaw ponds across the vast subarctic regions,
and may allow scaling up of the local greenhouse gas flux measurements to regional and
circumpolar scales.
2.8 Acknowledgments We thank T. Harding, F. Bouchard, J. Breton, C. Dupont, L. Rétamai, and A.
Rouillard for their assistance with field work and laboratory analyses, and R. Fillion for the
satellite image treatment and extraction of spectral signature. We also thank M. Babin
(CNRS, France) for valuable comments on the manuscript. This research was funded by Le
Fonds Québécois de la recherche sur la nature et les technologies (FQRNT), the Natural
Sciences and Engineering Research Council of Canada (NSERC), the Networks of Centres
of Excellence program ArcticNet, and the Canada Research Chair program.
44
Table 2.1 Limnological variables for the 14 permafrost thaw ponds: chlorophyll a (Chi a), dissolved organic carbon (DOC), total suspended particulate matter (SPMT), inorganic suspended particulate matter (SPMi), organic suspended particulate matter (SPMo), and total phosphorus (TP).
Variables units range mean CV (%)
ChlaT UgL"1 1.4-23.8 6.2 100
DOC1 mgL"1 3.1-10.5 7.7 27
SPMT mgL"1 5.0 - 42.4 16.8 74
SPMI mgL"1 2.1-36.7 13.0 88
SPMO mgL"1 1.6-7.1 3.8 47
T p t .gL"1 33.6-108.9 60.4 41
f Data from KWK 16 and 39 were missing (n=12).
45
Table 2.2 Summary of the absorption coefficients (a), specific absorption coefficients (a*) and exponential slope parameters (S) for optically active components in 14 thaw ponds: colored dissolved organic matter (CDOM), non-algal (NAP), algal, and fine particles.
Optical variable units range mean CV
CDOM OCDOM(320) tu
OCDOM(440) rn1
OCDOM*(440) t mV
SCDOM nm"1
NAP aNAp(440) rn"1
aNAP*(440) mV
SNAP nm"1
Algal particles a,p(440) m
a<p*(440) T m2 mg"1
Fine particles afine(440) _.-! m
Total at(440) m
9.882 - 56.047 34.78 36
1.654-9.333 5.461 37
0.491 - 0.962 0.743 21
0.0144 - 0.0159 0.0150 3
0.464 - 2.240 1.049 53
0.045-0.141 0.073 34
0.0096 - 0.0136 0.0111 10
0.162-1.037 0.415 61
0.044 - 0.164 0.086 47
0 - 8.865 2.18 127
2.922 - 16.773 9.080 49
t Data from KWK 16 and 39 were missing (n=12).
46
Table 2.3 The relative contribution of each absorbing component to total absorption at 440 nm (in %) for all sampled thaw ponds.
Variables range mean CV
CICDOM(440) 29-85 65 26
aNAp(440) 8-21 12 25
a0(44O) 2-14 5 60
afine(440) 0-53 19 84
aw(440) < 0.2 % for all ponds
47
Figure 2.1 Map of study area. The study sites were located in a discontinuous permafrost region near Whapmagoostui-Kuujjuarapik, Nunavik, Canada (55°20' N, 77°30' W).
48
Figure 2.2 Aerial photograph of the study site, showing some of the sampled thaw ponds numbered as in the text (Photo by I. Laurion).
49
a. Green/White
•x $ c. Brown
0.30
0.25
0.20
0.15
0.10
0.05
0.00 400 450 500 550 600 650 700 400 450 500 550 600 650 700
Wavelength (nm)
Figure 2.3 Dimensionless ratio of above surface upwelling radiance to 2JT corrected downwelling irradiance (R(Xj) of studied ponds. Ponds were grouped by visual appearance of color determined on sight and by aerial photographs: a. green to white, b. black, c. brown, and d. light brown.
KWK1 KWK 7
— KWK 33 — KWK 35
KWK 36
~ ' ~ z
,.9„ *■"•"."%
~ ' ~ z *""*" ._— ""*"* *"*" '"̂ '̂
50
b*{+): Yello 3fr
a*(-): Green 30 20 '% °
6 o » 2o> .6
o 23 . o33 O X 38
o
2o> .6 o
23 . o33 O X
A
3 o 5 ° 7
2 o
- o 11
■ >
<_-Green ° Red
■ >
-2 0 a*
Figure 2.4 Color of studied ponds expressed on CIE L*a*b* color coordinate. The a* axis shows green to red transition, with negative values indicating greenness and positive values indicating redness. Similarly, the b* axis shows blue to yellow transition, with negative values indicating blueness, and positive values indicating yellowness. The L* axis, indicating lightness, is not shown on the plot (mean of 36.0 with range from 16.8 to 55.0).
51
a. CDOM b.NAP
a*
c. Algal particles d. Fine particles
400 450 500 550 600 650 700 Wavelength (nm)
400 450 500 550 600 650 700 Wavelength (nm)
Figure 2.5 The absorption coefficient spectra from 400 to 700 nm of: a. Colored dissolved organic matter (acDOMfXj), b. non-algal particles (ONAP(X)), c. algal particles (a<p(X)), and d. fine particles (afme(X)).
52
12
10
8
6
o 2007 A 2006 A
Linear model ° Power model ^ /
0 8 10 12 14 16 18 -i DOC (mg L"1)
Figure 2.6 Relationship between absorption coefficient by colored dissolved organic matter at 440 nm (acooM(440)) and dissolved organic carbon (DOC). Open circles represent data collected in 2007 (n= 12), and triangles in 2006 (n = 36). The solid line represents the linear model and the broken line the power model on pooled data (n = 48).
53
0,
2.5
2.0 -
1.5
1.0
0.5
0.0
o
0 . ^ a ' O
/o o
°J*^ o
oo
■ - - 1 I " 1
0 10 20 30 -i
SPMT (mg L"1) 40
Figure 2.7 Relationship between absorption coefficient of non-algal particles at 440 nm (QNAP(440)) and total suspended particulate matter (SPMT). The line represents the linear regression model.
54
1.2
1.0 Linear model Power model
0 10 15 20
Chi a (|ag L"1) 25 30
Figure 2.8 Relationship between chlorophyll a concentration (Chi a) and absorption coefficient of algal particles at 440 nm (a<p(440)). The solid line represents the linear model and the broken line is for the power model applied to the data (n = 12).
55
-*. - -a. Green/White (KWK 6)
— J 1
v 20 - ■ °NAP
■ a 0
15 ■ aClX)M
■ a .
10
5
0
b. Black (KWK 2)
c. Brown (KWK 7) d. Light Brown (KWK 3)
400 450 500 550 600 650 700 400 450 500 550 600 650 700
Wavelength (nm)
Figure 2.9 Examples of component absorption spectra for representative pond types: a. Low absorption pond with green or white color (KWK 6), b. High dissolved absorption pond with black color (KWK 2), c. High dissolved absorption pond with medium particulate absorption (KWK 7), and d. High dissolved and particulate absorption pond (KWK 3).
56
a. All fourteen ponds
40
_— 30 6 R 20
10
0
KWK I KWK 2 KWK 3 KWK 6 KWK 7 KWK II KWK 16
KWK 21 KWK 23 KWK 33 KWK 35 KWK 36 KWK 38 KWK 39
b. KWK 11
400 450 500 550 600 Wavelength (nm)
650 700
Figure 2.10 The particulate scattering coefficient spectra (bp(X)) from 400 to 700 nm for: a. All 14 sampled ponds, and b. Pond KWK 11 where bp(X) was relatively low compare to other ponds and the influence of absorption of by algal pigments was noticeable.
57
1.00
0.95
0.90
^ 0.85
0.80
0.75 ■
0.70 -
KWKl KWK 2 KWK 3 KWK 6 KWK 7 KWK 11 KWK 16
KWK 21 KWK 23 KWK 33 KWK 35 KWK 36 KWK 38 KWK 39
400 450 500 550 600 Wavelength (nm)
650 700
Figure 2.11 The single scattering albedo spectra (cop(X)) for all 14 sampled ponds.
58
Blue Green Red
Spectral Bands
IR
Figure 2.12 Reflectance spectra of the 34 ponds obtained from the QuickBird satellite image in July 2006.
59
_*
S
CL
-a a.
LU
60
50
40
30
20
10
0
a. SPM,
s s _ / - " ^ *■***" *~ 1 ; 1 O / V
n on model y \
y Kegress on model y \
y
y y / y
y y y y
~ y y o °yy o y y
- o y y o y y y /
y y j * y
o y y
y y y y
y y
y6 -O , - T 1 1 1
10 20 30 40 50 60
Measured SPMT (mg L'1)
b. DOC 12 .
I I / y
o />* y y
y y y y yy o
. V
10 Regression model
/ y o />* y y
y y y y yy o '-J 10 Regression model
/ y o />* y y
y y y y yy o
ai) O o 95 E 8 y y
y y
8 6 y>
y o G 6 -a > o OJ
s 4 y y o
J •*-< 2 -
n y y
y y
0 4 6 8
Measured DOC (mg L'1)
10 12
Figure 2.13 Comparison of measured to estimated values of: a. total suspended particulate matter (SPMT) and b. dissolved organic carbon (DOC) for the validation data set (n = 10) when applying the canonical model. Solid lines show the regression model and broken lines the one-to-one relationship.
60
Supplement Table 2.1 Standardized coefficients to calculate canonical scores from satellite data for the estimation of total suspended particulate matter (SPMT) and dissolved organic carbon (DOC).
Variables Spectral band
Blue Green Red IR
SPMT
DOC
0.0559
-3.0971
1.5810
0.0893
-2.4563
2.39726
-0.0550
0.5187
61
a. ahJA) estimation
3.0
2.5
- r 2.o E
S ,* 5
tf 1.0
0.5 0.0
Bench top spectrophotometer
b. aflm(A) estimation
\ \
s \
Bench top spectrophotometer
\ \
X ***>
** 1 1
%r*̂ *A_.
400 450 500 550 600 650 700 Wavelength (nm)
Supplement Figure 2.1 Validation of the AC-S measurements. Two photometric measurements were compared for a. the total absorption coefficient minus pure water (at.w(X)) and b. the absorption coefficient of fine particles (afine(X)). Solid lines are values estimated using the AC-S, and broken lines are from the bench top spectrophotometer (details in the methods section). Pond KWK 38 data are presented as an example.
62
63
Chapitre 3
Light attenuation in a drinking water reservoir: photon
budget estimation and the importance of abiotic factors
3.1 Résumé L'atténuation du rayonnement photosynthétiquement actif (PAR) et ses facteurs de
contrôle ont été caractérisés dans le lac Saint-Charles, un lac tempéré utilisé pour
l'approvisionnement en eau potable de la ville de Québec, Canada. L'absorption et la
diffusion spectrale ont montré une relation relativement stable avec les variables
limnologiques au cours de l'été, suggérant que les variations dans les propriétés optiques
apparentes et inhérentes (AOP et IOP) de ce lac étaient causées par des changements de la
quantité plutôt que de la qualité des substances optiquement actives de la colonne d'eau.
L'atténuation verticale du PAR a été analysée selon une approche budgétaire de photons en
couplant les mesures d'AOP et d'IOP. Les résultats ont montré le rôle prédominant de la
matière organique dissoute colorée (CDOM) par rapport aux autres composantes
optiquement actives, ce qui implique que l'utilisation des méthodes traditionnelles de
mesure de la qualité de l'eau, telles que la profondeur de Secchi et la transparence
radiométrique, ne permettrait pas de suivre l'état biologique de la colonne d'eau de façon
adéquate pour la range de concentrations de chlorophylle a observée. Le budget de photons
a également révélé de grands changements verticaux dans la contribution absolue et relative
des différentes composantes optiques sur l'atténuation du PAR dans la colonne d'eau. Ces
changements ont affecté la moyenne des coefficients d'absorption in situ pour la zone
euphotique. Par exemple, la mesure du coefficient d'absorption de l'eau pure était dix fois
plus élevée que la valeur mesurée dans l'océan, valeur qui a pourtant souvent été appliquée
aux autres types d'eau. Les résultats soulignent ainsi la pertinence de la méthode du budget
de photons et la nécessité de considérer toutes les composantes optiquement actives dans la
surveillance de la qualité des eaux continentales.
64
3.2Abstract Optical measurements were conducted to characterize the attenuation processes of
photosynthetically active radiation (PAR) and its controls in Lake St. Charles, a north
temperate lake used as the drinking water supply for Québec City, Canada. The spectral
absorption and scattering showed relatively stable relationships with limnological variables
during summer, suggesting that the observed variations in apparent and inherent optical
properties (AOPs and IOPs, respectively) of the lake were caused by changes in quantity
rather than quality of optically active components in the water column. The vertical
attenuation of PAR was analyzed by a photon budget approach, coupling measurements of
both AOPs and IOPs. The results showed the dominant role of CDOM relative to other
optically active components, implying that the application of traditional measures of water
quality, such as Secchi depth and radiometric transparency, would not provide a
meaningful estimate of the biological state of the water column in the observed range of
chlorophyll a concentration. The photon budget also revealed large changes with depth in
the absolute and relative contribution of individual optical components to PAR attenuation.
This vertical feature affected the average in situ absorption coefficients for the euphotic
zone; for example the absorption coefficient for pure-water was ten-fold higher than the
value previously measured in the clear open ocean and erroneously applied to lakes and
coastal waters. The results highlight the value of the photon budget method, and the need to
consider all optically active components in monitoring inland water resources.
65
3.3 Introduction Although the factors controlling underwater light have been well studied in Case 1
waters (Morel 1988), the optical controls on Case 2 waters of lakes, reservoirs and the
coastal ocean have received much less attention (see, Bukata et al. 1995). In the latter
systems, optically active components, such as colored dissolved organic matter (CDOM),
non-algal particles (NAP), and phytoplankton pigments vary independently (International
Ocean Color Coordinating Group (IOCCG) 2000), and their apparent optical properties
(AOPs) are a complex function of inherent optical properties (IOPs) as well as of the
ambient light field. An AOP of major ecological and geophysical significance is the
Kd(PAR), the diffuse attenuation coefficient for downwelling irradiance (EJ) in the
photosynthetically active waveband (PAR; 400-700 nm). This sets the depth extent of the
euphotic zone (Kalff 2002; Kirk 1994), as well as the availability of light for visual
predators and other photobiological processes, and is a key parameter that is also used in
heat budget calculations (Morel and Antoine 1994; Patterson and Hamblin 1988).
Kd(PAR) can be partitioned into components associated with each of the optically
active materials in the water column (Kirk 1994; Mobley 1994), and such an analysis
provides information on the controlling variables for underwater light transmission. These
component-specific attenuation coefficients are typically estimated by conducting multiple
linear regressions on data obtained under a wide range of optical conditions (e.g., Duarte et
al. 1998; Perkins et al. 2009; V.-Balogh et al. 2009). An alternative approach, termed the
photon budget calculation, is based on combined AOP and IOP measurements, and was
introduced by Smith et al. (1989) to estimate the effects of depth and optical components
on PAR attenuation in the open ocean. There have been few applications of this technique
to oceanic waters (Giles-Guzman and Alvarez-Borrego 2000), and none in coastal and
inland waters to our knowledge.
The overall objective of the present research was to describe underwater optical
processes and its controlling factors in north temperate lakes, by way of detailed
measurements in a drinking water reservoir, a category of aquatic resources for which there
is an urgent need for improvements in environmental monitoring protocols. We examined
66
the IOPs of optically active components in the water column and related them to
limnological variables to understand their characteristics and implications for future
ecological modeling and remote observations of the lake. We also applied the photon
budget approach to determine how the optically active components of lake water affect the
vertical distribution of underwater light, and how their individual contributions to spectral
attenuation vary as a function of depth. Our secondary aim was to evaluate the potential of
the photon budget method for future studies of optically complex systems.
3.4 Methods
3.4.1 Study site and limnological conditions Lake St. Charles (46°56' N, 71°23' W; Fig. 3.1) is a drinking water reservoir located
20 km north of Québec City, Québec, Canada. It has surface area of 3.6 km , and catchment •y
area of 166 km that is primarily covered by forests (53 %). The mean hydraulic flushing
rate is 23 days (Pienitz and Vincent 2003; Tremblay et al. 2001). The lake was sampled on
10 June 2008, and biweekly from 8 July to 30 September 2008 (n = 8) at the deepest point
(16.6 m) located in the north basin. Sampling was always conducted within 3 hours of local
solar noon. Surface water (50 cm below the surface) was taken with Nalgene amber
polyethylene bottles (Thermo Fisher Scientific Inc., USA) and kept refrigerated in the dark
until processing in the laboratory within 5 h of collection. Temperature and dissolved
oxygen profiles were taken with a YSI 6600 sonde (YSI Inc., USA).
3.4.2 Apparent optical properties (AOPs) The vertical profiles of spectral downward irradiance (Ed(z;X), W m"2 nm"1) were
obtained with a hyperspectral radiometer system (Satlantic Inc., Canada). The instrument
was manually lowered at approximately 5 cm s". Raw radiometer data were linearly
interpolated to obtain the spectral profiles of Ed(z;X) from 352 to 797 nm at 1 nm intervals,
and the depth profiles at 10 cm intervals. For subsequent data analyses, Ed(z;X) was
converted to the spectral downward photon flux density (Eq:d(z;X), pmol m" nm" s" ) as,
E r Ar ,^^y Oil h c - N A
67
where h is Planck's constant (6.626 *10"34 J s), c the speed of light (299,792,458 m s"1), and
NA Avogadro constant (6.022 x IO23 mol-1). The vertical profiles of downward photon flux
density for photosynthetically active radiation (Eq;d(z;PAR), pmol m"2 s"1) were then
obtained by integrating Eq:d(z;X) over the range from 400 to 700 nm.
Three different diffuse attenuation coefficients were calculated by linearly
regressing log-transformed photon flux density against depth. The spectral diffuse
attenuation coefficients of downward photon flux density (Kd(X), m"1) were obtained by
regressing Eq:d(z;X) over the depth range from just below the surface to the depth where
Eq:d(z;X) reached 1 % of surface value. Similarly, the diffuse attenuation coefficient for
PAR in the euphotic zone depth (Kd(PAR), m"1) was calculated for Eq:d(z;PAR) from the
surface to the limit of the euphotic zone, where Eq:d(z;PAR) reached 1 % of the surface
value. The diffuse attenuation coefficient for PAR at depth z (Kd(z;PAR), m"1) was
calculated by using seven data points of Eq:d(z;PAR) centered around z m (i.e., depth range
of 0.6 m). For example, the coefficient at 2 m depth (Kd(2;PAR), m"1) was calculated with
Eq:d(z;PAR) at 1.7, 1.8, 1.9, 2.0, 2.1, 2.2, and 2.3 m.
3.4.3 Laboratory analyses of limnological variables Water samples were filtered on GF/F glass fiber filters (Whatman Inc., USA) and
stored frozen (-80°C) until laboratory analysis for chlorophyll a concentrations (Chi a,
pg L"1). Pigments were extracted in ethanol and Chi a concentration was determined by
fluorometry (Cary Eclipse spectrofluorometer, Varian Inc., Canada) before and after
acidification (Nusch 1980). Water samples were also filtered onto pre-combusted, pre-
weighed GF/F filters to quantify suspended particulate matter. Filters were weighed after
drying for 2 h at 60°C to determine total suspended particulate matter (SPMT, rng L"1).
Subsequently, filters were combusted at 500°C for 2 h, and then reweighed to estimate
inorganic suspended particulate matter (SPMi, mg L"1). Organic suspended particulate
matter (SPMo, mg L"1) estimates were obtained as the difference between SPMT and SPMi.
Dissolved organic carbon (DOC, mg L" ) concentrations were measured with a TOC-
5000A carbon analyzer (Shimadzu Co., Japan) on filtrates passed through pre-rinsed
nitrocellulose membrane filters (pore size 0.22 pm, Millipore Co., USA).
68
3.4.4 Inherent optical properties (IOPs) 3.4.4.1 Absorption coefficient of colored dissolved organic matter
For analysis of colored dissolved organic matter (CDOM), water samples were
filtered through GF/F filters and subsequently through the membrane filters, and the
filtrates were refrigerated in amber glass bottles until analysis. The absorbance (A(X),
dimensionless) of the filtrate against pure-water prepared by a EASY pure II UV/UF
system (Barnstead International, USA) was measured from 200 to 850 nm at 1 nm intervals
(spectral slit width 2 nm) in a 1 cm quartz cuvette using a Cary 100 dual beam
spectrophotometer (Varian Inc., Canada). The absorption coefficients of CDOM (OCDOM(^),
m"1) were calculated as:
. . , 2.303.4(A) - . <*CDOM W = ; eq- 3.2
where L is the path length of the cuvette (Beer's law). Null point correction was conducted
by subtracting the average of acDOM(X) from 750 to 760 nm where the absorption curves
became flat (Mitchell et al. 2000; 2002). The exponential slope parameters of the spectra
(SCDOM) were calculated by non-linearly fitting the measured value from 300 to 600 nm to
the equation:
aaxjuW^cDOMW-e- 3™"-™ eq. 3.3
where Xo is the reference wavelength (Bricaud et al. 1981; Stedmon et al. 2000). The DOC-
specific absorption coefficient of CDOM (OCDOM*(X), m2 g"1) was calculated as OCDOM(X)
divided by DOC concentration.
3.4.4.2 Absorption coefficients of suspended particulate matter
Water samples were filtered onto GF/F filters and stored at -80°C until optical
analysis. The absorbance of particulate matter was measured by the wet filter technique
(quantitative filter technique) in the same spectrophotometer as above equipped with an
integrating sphere (Labsphere Inc., USA) following Mitchell et al. (2002). The absorbance
69
of non-algal particles was measured by extracting phytoplanktonic pigments with methanol
(Kishino et al. 1985; Mitchell et al. 2002). The absorption coefficients of total and non-
algal particulate matter (ap(X) and ONAP(X) in m"1, respectively) were calculated as:
. . . 2.3034(A)-T , „ ap(A) or aNAP(A) = y eq. 3.4
where T is the filtered area, V is filtered volume, and /? is the path length amplification
factor (Mitchell et al. 2002). ,6 = 2 was used after Roesler (1998) (see, Annexe 3 for further
discussion regarding fl). Null point correction was applied in the same manner as above.
The absorption coefficient of algal particles (a&(X), m"1) was obtained by subtracting
O-NAPOO from ap(X). Similar to OCDOM(X), the exponential slope parameter of the ONAP(X)
spectra (SNAP) was calculated by fitting equation 3.3 to the data. Non-linear regression was
conducted over the spectral range of 380-730 nm excluding the ranges 400-480 and
620-710 nm to eliminate the residual pigment absorption (Babin et al. 2003a; Belzile et al.
2004). The SPMT-specific absorption coefficient of non-algal particulate matter (aNAP*(X),
m2 g"1) and Chi ..-specific absorption coefficient of algal particles (a&*(X) , m2 mg"1) were
also calculated. The GF/F filtrate was filtered onto the membrane filters and the absorption
coefficient of materials collected on the filters (fine particles, afine(X), m"1) was determined
using the same method as for the ap(X) quantification described above (Ferrari and Tassan
1996; Gallegos 2005; Watanabe et al. 2011).
3.4.4.3 AC-S measurements
The absorption and beam attenuation coefficients excluding those of pure-water for
unfiltered water samples (at.w(X) and ct.w(X) in m"1, respectively) were measured using a
AC-S in situ spectrophotometer (25 cm path length, WET Labs Inc., USA) on the
laboratory bench (Watanabe et al. 2011). The pure-water calibration values obtained
immediately before the sampling period were subtracted from the raw readings (Pegau et al.
2003; Sullivan et al. 2006; Twardowski et al. 1999). The temperature correction was
applied by using the spectral correction coefficients obtained specifically for our instrument
(Pegau et al. 2003; Sullivan et al. 2006; Twardowski et al. 1999). A null point correction
70
was applied by subtracting the absorption coefficient at the longest wavelength measured
by the instrument (751.7 nm), as in Pegau et al. (2003).
The measured spectral ranges of our instrument were 403.0-751.7 nm and 401.6-
750.8 nm for at.w(X) and ct.w(X), respectively. The obtained spectra were linearly
interpolated for the range 403 to 700 nm at 1 nm intervals, and were also extrapolated to
400 nm by fitting the measured values from 403 to 410 nm to the exponential model as in
equation 3.3. The scattering coefficients of suspended particulate matter (bp(X), m"1) were
then obtained by subtracting a,.w(X) from ct.w(X). The SPMT-specific scattering coefficients
for suspended particulate matter (bp*(X), m2 g"1) were obtained by dividing bp(X) by SPMT.
Finally, the total absorption coefficient (at(X), m"1) was obtained by adding the absorption
coefficient of pure-water (aw(X), m"1) (Pope and Fry 1997) to the at.w(X), and the total
scattering coefficient (bt(X), m"1) was obtained by adding the scattering coefficient of pure-
water bw(X) (Buiteveld et al. 1994) to the bp(X).
3.4.5 Photon budget calculation The classic work by Preisendorfer (1961; 1976) showed that the diffuse attenuation
coefficient oîEd(z;X) can be calculated as:
Kd(z;A)=ad(z;A) + bb.d(z;A)-bb.M(z;A)R(z;A) eq. 3.5
where ad(z;X) is the diffuse absorption coefficient for downward irradiance, bt,;d(z;X) is the
diffuse backscattering coefficient for downward irradiance, bb,u(z;X) is the diffuse
backscattering coefficient for upward irradiance, and R(z;X) is the irradiance reflectance,
defined as the ratio of upward irradiance to downward irradiance. Theoretically, ad(z;X) can
be subdivided into those of optically active components due to the additive nature of the
absorption coefficient. The bb;u(z;X)R(z;X), representing downward scattering of upward
irradiance, should be negligible relative to ad(z;X) and bb,d(z;X) in waterbodies where the
ratio of scattering to absorption is low (b/a < 5, Kirk 1981; Smith et al. 1989). Thus, the
diffuse attenuation coefficient of downward irradiance can be estimated as:
71
Kd fe A)=X ad, & À) + Kd (*; A) eq. 3.6 1=1
where, ad:i(z;X) is the diffuse absorption coefficient of optically active component i. The
conceptual background to this Kd(z;X) approximation from IOPs have been discussed in
detail by Smith et al. (1989) and Kirk (1994).
In the present study, four absorbing components measured by laboratory
spectrophotometry (CDOM, NAP, algal particles, and fine particles) and pure-water (Pope
and Fry 1997) were considered as the most significant light absorbing components in the
water column, and their absorption spectra were applied for the subsequent estimation
process. The back scattering coefficient (bb(X), m"1) was estimated by multiplying bp(X) by
the backscattering ratio of 0.018, which is a widely accepted value for particles in natural
water (Petzold 1972). These absorption and back scattering spectra measured with
collimated light were converted to the diffuse absorption and backscattering coefficients of
downward radiation (ad:i(z;X) and bb,d(z;X), respectively) by dividing by the average cosine
of downward irradiance (pd(z;A), dimensionless), which is the ratio of the downward
irradiance to the downward scalar irradiance (Kirk 1994). The parameter pd(z;A) was
estimated by using the HydroLight radiative transfer model (Sequoia Scientific Inc., USA)
with the input of lake location, sample time, measured above-water spectral irradiance,
at.w(X) and ct.w(X) measured by AC-S, a backscattering ratio of 0.018, and sky conditions for
each sampling day. These diffuse absorption coefficients were subsequently incorporated
into the spectrally weighed diffuse absorption coefficient of component i for PAR
(ad:i(z;PAR)) as:
<¥00
a„(r,PAR)=*±-^ "- eq. 3.7
[J ,M^A
72
The spectrally averaged diffuse backscattering coefficient for PAR (bbd(z;PAR)) was
calculated in an analogous manner. Finally, the estimated diffuse attenuation coefficient for
PAR (Kd'(z;PARj) was calculated as:
Kd'(z;PAR) = ̂ a d l (z ;PAR) + bb;d(z;PAR) eq. 3.8 < = 1
3.5 Results 3.5.1 Limnological variables
Limnological data for Lake St. Charles over the sampling period are given in Table
3.1. SPMT varied more than two-fold, with no clear seasonal trend. The percent
contribution of SPMi to SPMT varied from 18 to 52 % (mean: 40 %). Chi a values were in
the mesotrophic range, and varied by a factor of 5, with higher values in late July and
August (9.5-15.4 pg L"1) and low to moderate values (3.0-7.4 pg L"1) in the rest of the
season. DOC concentrations varied to a much lesser extent than SPM and Chi a, increasing
from 10 June (2.47 mg L"1) to 8 July (4.2 mg L"1), and then gradually decreasing throughout
the summer to 3.09 mg L"1 on 30 Sep 2008.
3.5.2 AOPs As shown in an example profile of 8 July 2008 (Fig. 3.2), the Eq:d(z;X) attenuated
rapidly at shorter relative to longer wavelengths, and similar spectral characteristics were
observed throughout summer. The Kd(X) was, therefore, higher at the blue end of the
spectra and lower at the red end of the spectra (Fig. 3.3). The average Kd(440) was
3.565 m"1 with a range from 2.517 to 4.628 m"1 (CV: 20 %), and the average Kd(680) of
0.947 m"1 with a range of 0.789 to 1.068 m"1 (CV: 9 %). The Kd(PAR) showed the lowest
value (0.769 m"1) early in summer (10 June 2008); the value increased to a maximum of
1.289 m"1 on 8 July, and gradually decreased thereafter to 1.005 m"1 during the rest of the
summer. The estimated euphotic zone depth ranged from 3.57 to 5.98 m (mean: 4.41 m;
CV: 17 %).
73
3.5.3 IOPs 3.5.3.1 Absorption coefficients of colored dissolved organic matter
The aCDOM(440) showed a two-fold variation (Fig. 3.4A, Table 3.2), with a clear
trend of decreasing values during mid to late summer. The lowest value of OCDOM(440)
(1.383 m"1) was observed at the beginning of the summer (10 June 2008). The value
increased to a maximum of 2.651 m"1 one month later (8 July 2008), and gradually
decreased throughout the summer to 1.804 m"1. The summer average was 2.155 m"1.
Although the OCDOM*(440) showed a similar seasonal trend as OCDOM(440), the variation
was small (Table 3.2). SCDOM showed little variation (CV: 1 %) with a mean of 0.0157 nm"1.
There was a strong correlation between acDOM(440) and DOC (r = 0.98, p < 0.01).
3.5.3.2 Absorption coefficients of suspended particulate matter
Similar to acDOM(440), the ONAP(440) values also showed a two-fold variation (Fig.
3.4B, Table 3.2), but no clear seasonal trend. The specific values, ONAP*(440), ranged from
0.190 to 0.259 m2 g"1 with a mean of 0.219 m2 g"1. The mean value of SNAP was 0.0101 nm"1
with little variation (Table 3.3). The aNAp(440) was significantly correlated with both SPMT
(r = 0.94, p < 0.01) and SPMi (r = 0.81, p = 0.02). The a0(44O) showed two-fold variation
with a mean of 0.119 m"1 (Fig. 3.4C, Table 3.2). It was strongly correlated with Chi a (r =
0.92, p < 0.01). The range of a9*(440) was from 0.010 to 0.025 m2 mg"1 (CV: 32 %). The
absorption of fine particles showed three-fold variation with a mean of 0.257 m"1 (Table
3.2), with no clear seasonal trend.
3.5.3.3 Scattering coefficients
The scattering coefficient bp(440) showed two-fold variation (Table 3.2). It was
significantly correlated with SPMT (r = 0.99, p < 0.01). Specific values, bp*(440), showed
little variation (Table 3.2). The shape ofbp(X) spectra is often characterized by a hyperbolic
model as:
b(A)=b(555)x ( A V eq. 3.9 555 J
74
where n is the power model exponent (Babin et al. 2003a; Belzile et al. 2004). However,
the scattering spectra obtained for Lake St. Charles particles did not fit this model due to
the influence of absorption peaks of phytoplankton pigments. The ratio of the total
scattering to the total absorption coefficient (bJat(X), dimensionless) showed a two-fold
seasonal variation. The spectra had highest values between 580 to 590 nm except for 5
August 2008 when the highest value was at 640 nm. The average value for b/at(585) was
2.355, with a range from 1.567 to 2.884 (CV: 19 %).
3.5.4 Photon budget calculations On all dates of sampling, the estimated diffuse attenuation coefficient for PAR
(Kd'(z;PAR)) decreased with depth (Fig. 3.5A), dropping from 54 to 62 % (mean: 58 %) of
the surface value at the bottom of the euphotic zone. This estimated Kd'(z;PAR) profiles
were comparable to Kd(z;PAR) obtained by regression analysis of the Eq:d(z;PAR) profiles
(Fig. 3.5B).
Vertical profiles of the absorption and scattering coefficients for specific optically
active components are illustrated in the Figure 3.6A (data from 8 June 2008). The
ctd,cDOM(z;PAR.) values decreased from 1.185 m"1 at just below the surface to 0.386 m" at
the bottom of the euphotic zone (3.5 m). Other non-water absorbing components, NAP,
algal particles, and fine particles also showed similar vertical decreases. For example,
surface to euphotic zone depth change of the ad:fme(z;PAR) was from 0.096 to 0.033 m".
The ad:w(z;PAR) was the only component that increased with depth, rising from 0.193 at the
surface to 0.374 m"1 at the bottom of the euphotic zone. In contrast, bb,d(z;PAR) showed
little vertical variation. The proportional contribution of each optical component to total
attenuation at just below surface was 64, 15, 5, 4, 2, and 10 % for CDOM, NAP, fine
particles, back scattering, and water, respectively (Fig. 3.6B). The values at the euphotic
zone depth were 39, 12, 3, 5, 3, and 37 % (totaling slightly below 100 % due to rounding).
These vertical characteristics and proportional contributions were consistent throughout the
summer (Figs. 3.6B, D, and E).
75
Seasonal variation in the estimated diffuse attenuation coefficient for PAR averaged
over euphotic zone depth (Kd'(zave;PAR)), just below the surface (Kd'(0;PAR)), and at the
bottom of the euphotic zone (Kd'(zeu;PAR)) are shown in Figures 3.7 A, C, and E,
respectively. The seasonal trend for these variables was consistent with that of measured
Kd(PAR), with an increase from June to July followed by a gradual decrease throughout the
summer. For the average over euphotic zone depth, the range of ad:cDOM(zave',PAR) was
from 0.334 to 0.569 m"1, and a ^ z ^ P A R ) from 0.265 to 0.320 m"1 (Table 3.3). Other
components showed two- to four-fold variations, but their absolute changes were smaller
relative to those of CDOM and water (Table 3.3). There were close correlative relationships
between the optical and limnological variables: ad-cDOM(zave',PAR) versus DOC, r = 0.97 (p
< 0.01); ad;NAp(zave;PAR) versus SPMT, r = 0.92 (p < 0.01); ad:NAp(zave;PAR) versus SPMb
r = 0.88 (p= 0.01); ad:0(zave;PAR) versus Chi a, r = 0.95 (p < 0.01); and bb,d(zave;PAR)
versus SPMT, r = 0.87 (p < 0.01).
3.6 Discussion 3.6.1 IOPs and specific characteristics of optically active components 3.6.1.1 Colored dissolved organic matter
The IOP analyses provided a detailed description of optically active components in
Lake St. Charles. The observed high OCDOM*(X) and low SCDOM values imply that dissolved
organic matter in the lake is highly colored (Morris et al. 1995; Stedmon et al. 2000). Also,
little seasonal variations in these variables suggest that there was little qualitative variation
throughout the summer, and that seasonal changes in CDOM absorption were almost
entirely due to shifts in the concentration of dissolved organic matter. This carbon pool may
be dominated by terrigenous carbon delivered by the inflowing river from catchments
mostly covered by deciduous forests (Pienitz and Vincent 2003; Tremblay et al. 2001). The
influence of photochemical or biological degradation, which is known to cause seasonal
changes in the optical properties of CDOM (Vàhâtalo and Wetzel 2004), may be minimal
due to high hydraulic flushing rate (23 days; Tremblay et al. 2001). This qualitative
stability in carbon pool and strong linearity in the CDOM-DOC relationship implies that
CDOM absorption can serve as an index to estimate DOC concentrations in the water
column, and vice versa.
76
3.6.1.2 Non-algal particles 9 1
The ONAP*(440) in the lake (mean 0.219 m g" ) was considerably higher than
values previously reported for other aquatic systems (Babin et al. 2003b; Belzile et al.
2004; Binding et al. 2008). This high mass-specific absorption may be the result of the
NAP particle size distribution (Stramski et al. 2007) or composition (Babin et al. 2003b;
Binding et al. 2008). Adsorption of the highly colored CDOM in the water to these non-
algal particles could also contribute to a higher mass-specific absorption. This effect cannot
be evaluated from the present data set. ONAP(X) is assumed to describe the absorption by
both inorganic and organic component of non-algal particles, while SPMT incorporates the
dry weight of algal particles in addition to non-algal components. Also, SPMi is assumed to
be exclusively composed of the inorganic portion of non-algal particles. This difference
between the absorption and dry weight measurements may confound interpretation of the
mass-specific absorption estimates. Despite this uncertainty, however, the small variations
observed in ONAP*(440) and SNAP, and the strong £*#,(/>(<.'. 0,1-SPMT/SPMI relationship,
indicate that the optical characteristics of NAP were relatively stable over the sampling
period, and that variation in ONAP(440) was largely the result of changes in concentration of
suspended particles. This characteristic means that it would be feasible to estimate
suspended particle concentrations from optical data, and conversely to apply SPM
measurements in combination with the aNAP*(X) for the forward modeling of AOPs of the
lake.
3.6.1.3 Algal particles
The precise determination of algal absorption is essential for the parameterization
of optical processes in natural waterbodies, and for ecological and biogeochemical
modeling (Babin et al. 2003b; Bricaud et al. 1995). The two-fold variations in algal
absorption were well explained by the changes in Chi a concentration in the lake, and the
Chi a-specific absorption values were comparable to those reported elsewhere (Bricaud et
al. 1995; Dekker et al. 2002; Zhang et al. 2010). These characteristics suggest that
estimation of algal absorption spectra from Chi a concentrations and subsequent modeling
of primary production (Falkowski and Raven 2007; Morel 1991) would be feasible in this
lake. Remote observation of Chi a, however, may be difficult because the contribution of
77
algal absorption to overall PAR attenuation was relatively low compared to other optical
components (Fig. 3.4), and these other components (CDOM, NAP) were not statistically
correlated, unlike Case 1 waters.
3.6.1.4 Fine particles
Fine particles are defined as particulate matter passing through glass fiber filters
(Whatman GF/F in the present study) but caught on membrane filters with a pore size of
0.22 pm (Millipore nitrocellulose membrane filters in the present study). Particles in this
size fraction have been commonly observed in a variety of highly turbid systems such as
reservoirs in an agricultural area (Jones et al. 2008), thaw ponds in subarctic discontinuous
permafrost (Watanabe et al. 2011), and the estuary of a large river (Gallegos 2005). In
certain systems, they can be an important optical component (Gallegos 2005; Watanabe et
al. 2011). To our knowledge, however, they have not been previously considered in
relatively clear inland waters such as the lake sampled here, and their detailed optical
properties have been little explored. In the present study, the absorption by fine particles
was detected throughout the summer. Although their absorption was lower relative to
CDOM absorption, it was equivalent to NAP absorption at times (Fig. 3.4). This suggests
that fine particles may exert a significant influence on optical processes, not only in highly
turbid systems but also in mesotrophic waters. Therefore, their light absorbing and
scattering properties require much closer attention in optically complex waterbodies.
The fine particles in the lake had exponentially shaped absorption spectra with a
small peak at 680 nm (Fig. 3.4D) implying that they were mostly non-algal with only a
minor contribution of phytoplankton. The proportion of these particles may vary seasonally.
On two sampling dates, there were relatively higher fine particle absorption values (Fig.
3.4D). The early summer observation (afine(440) = 0.45 m"1, 10 June 2008) showed a clear
peak at 680 nm, indicating a significant contribution by phytoplankton pigments. On the
other hand, the mid-summer curve (af,ne(440) = 0.40 m"1, 3 September 2008) showed no
such peak, suggesting less algal contribution to absorption at that time. This difference
illustrates the compositional shifts in fine particles through time, and the need for ongoing
analysis.
78
3.6.1.5 Scattering coefficients
Observations of the spectral scattering coefficient for optically complex coastal
and inland waters have been limited to date (Babin et al. 2003a), and the present set of
measurements extends the range of optical conditions for such data. The shapes of the bp(X)
spectra were consistent throughout the summer, except for the one observation showing a
steeper curve (8 July 2008, Fig. 3.4F). Although differences in particle composition or size
distribution may produce such a deviation (Babin et al. 2003 a), characterization of the
spectra by a power model exponent (eq. 3.9, Babin et al. 2003a; Belzile et al. 2004) was not
possible for the present data set due to the influence of algal pigment absorption. Thus,
estimation of spectra for forward modeling (Mobley 1994) by applying the power model
would not be appropriate. However, the relatively consistent spectral shape and the strong
linear relationship between bp(X) and SPMT may allow the estimation of SPMT from bp(X),
and the application of SPMT for forward modeling of AOPs in a manner analogous to that
for NAP absorption.
3.6.2 Photon budget calculation 3.6.2.1 Vertical attenuation of PAR
As shown in Figure 3.5, Kd'(z;PAR) was highest at the surface and decreased with
depth, due mainly to changes in the spectral distribution oïEq:d(z;X) in the water column. In
the upper layer, higher absorption at shorter wavelength efficiently attenuated Eq:d(z;X) and
contributed to the high Kd'(z;PAR) value. As incident solar radiation penetrated deeper into
the water column, the irradiance at shorter wavelengths was rapidly depleted, and the
remaining longer wavelength PAR was attenuated less strongly, resulting in lower values of
Kd'(z;PAR) at depth. The vertical decrease of Kd'(z;PAR) is well known in oceanic systems
(Giles-Guzman and Alvarez-Borrego 2000; Smith et al. 1989), but the way in which
spectral distribution changes with depth in such Case 1 waters differs from the present
study. In the more transparent oceanic systems, light at red end of the spectrum attenuates
more rapidly relative to blue light, opposite to the pattern observed here because of the
strong terrigenous CDOM and NAP influence.
79
3.6.2.2 Contributions to PAR attenuation
The PAR attenuation process was well described by the budget analysis of
contributions by each optical component to photon loss down the water column. The
component-specific coefficient profiles showed vertical changes depending upon their
absorption and scattering characteristics (Fig. 3.6), and the mechanisms of variation were
apparent in the estimates of per volume spectral absorption as a function of depth
(ad:i(z;X)xEq;d(z;X); Fig. 3.8). There were large vertical decreases in ad:cDOM(z;PAR), with
more than a two-fold change from surface to the bottom of the euphotic zone. This change
was associated with depletion of light at shorter wavelengths where CDOM has highest
absorption. Similar vertical changes were observed for both NAP and fine particles,
components with similar exponential absorption curves as for CDOM. The absorption by
algal particles also showed a depth-dependent decrease; however, it was not as considerable
as those of abiotic factors due to the bimodal shape of chlorophyll absorption spectra, with
peaks in both the blue and red wavebands of the spectrum. Pure-water was the only
component that showed a strong vertical increase, due to its higher absorption at longer
wavelengths. Back scattering showed little change down the water column because of its
small spectral dependency.
3.6.2.3 Seasonal dynamics of PAR attenuation
The seasonal measurements of Kd'(z;PAR) and its component structure indicate that
variation in PAR attenuation of Lake St. Charles is mostly accounted for by changes in
CDOM absorption (Fig. 3.7), which in turn is controlled by DOC concentration. Other
components showed two- to four-fold seasonal variation in tandem with variations in their
respective limnological variables; however, the ranges of variation and proportional
contribution to total attenuation process were small relative to those of CDOM. Algal
particles, which are often considered to be the major controlling factor for underwater light
and an important indicator of lake water quality, made little contribution to PAR
attenuation and its seasonal variability. Therefore, optical observations (e.g., radiometric
determination of the diffuse attenuation coefficient for PAR and Secchi depth
measurements of water column transparency) would provide a misleading index of
phytoplankton dynamics and water quality. Similar caution must be applied to lake
80
monitoring practices throughout the north temperate region, where there are typically large
inputs of highly colored dissolved organic matter to lakes from their catchments (Prairie et
al. 2002).
3.6.2.4 Attenuation by water
The absorption coefficient of pure-water for PAR has often been considered as a
constant (Morel 1988) given its fixed absorption spectrum (Pope and Fry 1997). Our results,
however, show that the in situ absorption values for pure-water are subject to striking
vertical and seasonal variations (Fig. 3.6A, Table 3.3). In the present study and on all dates
of sampling, ad;w(z;PAR) increased with depth, in sharp contrast to absorption by the other
optically active materials. This is also contrary to the pattern reported for oceanic systems,
where in situ ad;w(z;PAR) decreases with depth because of the shift to bluer wavelengths
(Giles-Guzman and Alvarez-Borrego 2000; Smith et al. 1989). The average value for the
Lake St. Charles euphotic zone (0.304 m" ; Table 3.3) was an order of magnitude higher
than the classic values for oceanic Case 1 waters (0.027 m"1; Smith and Baker 1978; 0.0384
m"1, Lorenzen 1972), and which have often been applied in studies elsewhere, including for
Case 2 coastal and inland waters (e.g., Obrador and Pretus 2008; V.-Balogh et al. 2009, and
references therein). This disparity between offshore marine and inland waters is the result
of differences in spectral irradiance down their respective water columns. Although the
seasonal variation of the water absorption of PAR in Lake St. Charles was not as
considerable as that of CDOM, a^z^eiPAR), there were measurable changes in its value
among sampling dates (from 0.265 to 0.320 m"1). This result clearly shows that the
absorption (attenuation) coefficient of water for PAR cannot be considered a constant for
the euphotic zone, and varies as a function of the other absorption and scattering
components that change the spectral distribution of light with depth. Temperature
dependence of the absorption coefficient of water (Schiebener et al. 1990) might also affect
the estimation of the PAR absorption, and future evaluation is needed for this aspect.
3.6.2.5 Application of the photon budget approach
The present study of attenuation processes in Lake St. Charles combined field
measurements of spectral irradiance with laboratory measurements of component-specific
81
absorption and scattering spectra. Although similar AOP and IOP coupling has long been
applied to model solar radiation-dependent processes in aquatic ecosystems, such as
primary production (Falkowski and Raven 2007; Markager and Vincent 2001; Morel 1991)
and photodegradation of dissolved organic matter (Vàhâtalo and Wetzel 2004), there has
been little application of this spectral approach towards understanding the mechanisms of
light attenuation, and such studies have been mostly limited to oceanic systems (Giles-
Guzman and Alvarez-Borrego 2000; Smith et al. 1989). This coupled AOP and IOP
analysis has several advantages over the estimation of component-specific attenuation
coefficients by multiple linear regressions, as has been routinely applied to optically
complex waters (e.g., Duarte et al. 1998; Perkins et al. 2009; V.-Balogh et al. 2009). Firstly,
the photon budget approach provides detailed descriptions of the vertical gradient in
spectral attenuation, while the more common regression approach provides only a water
column averaged Kd(PAR) value. Secondly, the present estimation can be conducted with a
single set of measurements. In contrast, multiple linear regression requires a large number
of observations, yet produces only a single averaged estimate that does not capture the large
intra- and inter-system variation in space and time. Furthermore, the resultant estimate fails
to take into account the spectral changes in the ambient light field, which can result in order
of magnitude errors. The photon budget approach will serve as a useful tool to understand
the complex underwater optical processes of coastal and inland waters, and also to model
ecological and biogeochemical processes of such systems. This spectrally explicit approach
is also increasingly attractive given that data sets from hyperspectral radiometers and
advanced techniques for IOP analysis are becoming widely available.
3.6.2.6 Estimation of average cosine and back scattering ratio
There are ongoing technical issues to be addressed for future applications of the
methods developed here. Neither p.d(z;A) nor the back scattering ratio have been
measured frequently in natural waters; however, estimates of both variables were required
for the present calculations. The pd(z;A) value represents the average angular distribution
of downward photon flux at a point in the water column (Kirk 1994). It is used to convert
absorption and scattering coefficients obtained with collimated light to those for diffuse
light field. Theoretical studies have shown that there are spectral and vertical variations in
82
pd(z; A) depending on a number of factors, such as the scattering to absorption ratio,
scattering phase function, optical depth, and solar angle (Berwald et al. 1995; Kirk 1981).
Despite its importance for describing the underwater light field, this parameter is difficult
to obtain either by field measurement, for example by simultaneous deployment of
radiometers with both cosine and scalar collectors, or by numerical modeling, requiring
detailed information regarding underwater light field of the target waterbody as applied in
the present study. Smith et al. (1989) assumed pd(z;A) to be a constant of 0.8 for the
entire water column in their Kd(z;PAR) estimation. When this assumption was applied to
our data set, the difference between Kd'(z;PAR) estimated with modeled pd(z;A) and the
constant was only -5 to 6 % for the entire water column throughout the summer. This
implies that pd(z;A) can be reasonably approximated at a constant value of 0.8 for oligo-
to mesotrophic waters with relatively low b/a ratio, if measured or estimated values are not
available. The back scattering ratio expresses the loss of downward photon flux by upward
scattering. Although it was assumed to be the constant value of 0.018 (Petzold 1972) in the
present study, this value can vary widely and have a substantial influence on the underwater
light field in optically complex waters (Tzortziou et al. 2006). An estimation from field
radiometric measurements by a method based on Monte Carlo simulation (bb(X) =
a(X)-Lu(X)/0.082 Ed(X); Kirk 1994) showed that the value could vary considerably in Lake St.
Charles (range: 0.011-0.032, mean: 0.017). This possible variation in back scattering ratio,
however, had little influence on the total Kd(z;PAR) estimation. Because of the relatively
minor contribution of back scattering to the total Kd(z;PAR) of the lake (Fig. 3.7, Table 3.3),
a three-fold change in the back scattering ratio resulted in a maximum of only 3 % change
in Kd'(z;PAR). Thus, the application of the constant value appears to be appropriate for
Kd(z;PAR) estimation in Lake St. Charles and similar mesotrophic lakes. However, this
should carefully be examined in highly turbid systems where scattering makes a greater
contribution to PAR attenuation.
In the present study, IOPs were measured only in the surface water and were
assumed to be constant throughout the water column. This was sufficient for the lake where
vertical changes in the limnological conditions are minor; however, vertical variation
83
should be considered where necessary. The application of IOP profiles taken by an in situ
spectrophotometer (AC meters) or fluorometric measurement of CDOM and Chi a may
also be useful for such precise estimates of vertical structure of the attenuation process.
Furthermore, emission of light associated with CDOM and chlorophyll fluorescence, and
Raman scattering may be considered as significant optical processes.
3.7 Conclusions The suite of optical variables measured in the present study provided a detailed
description of spectral PAR attenuation and its controls in Lake St. Charles. The observed
stable characteristics of optically active components through the summer suggests potential
application of this information for future modeling and monitoring of limnological
conditions of the lake; however, the trophic state of the lake cannot be reliably estimated
from transparency. The photon budget calculation was shown to be a promising tool to
describe vertical PAR attenuation processes in optically complex waters, where multiple
components vary independently. Therefore, this spectrally explicit approach will provide
important information for future ecological and biogeochemical modeling, particularly as
advanced optical instruments and techniques become increasingly available.
84
3.8 Summary of the notation used in this article X wavelength, m
z depth, m
Chi a chlorophyll a, pg L"1
S P M T total suspended particulate matter, mg L"1
SPMi inorganic suspended particulate matter, mg L"1
SPMo organic suspended particulate matter, mg L"1
DOC dissolved organic carbon, mg L"1
Ed(z;X) spectral downward irradiance, W m" nm" "? 1 1
Eq:d(z;X) spectral downward photon flux density, pmol m" nm" s" Eqd(z;PAR) photon flux density for photosynthetically active radiation (PAR), pmol
m-V Kd(X) spectral diffuse attenuation coefficient of downward photon flux density, _.-' m Kd(PAR) diffuse attenuation coefficient for PAR in the euphotic zone depth, m"1
Kd(z;PAR) diffuse attenuation coefficient for PAR at depth z, m"1
<*CDOM(X) absorption coefficient of colored dissolved organic matter (CDOM), m"1
acDOM*(X) DOC-specific absorption coefficient of CDOM, m2 g"1
SCDOM exponential slope parameter of the CDOM absorption spectra
ap(X) absorption coefficient of total particulate matter, m"1
CINAP(X) absorption coefficient of non-algal particulate matter (NAP), m"1
<*NAP*(X) SPMT-specific absorption coefficient of NAP, m2 g"1
SNAP exponential slope parameter of the NAP absorption spectra
a<p(X) absorption coefficient of algal particles, m"1
a&*(X) Chi a-specific absorption coefficient of algal particles, m2 mg"1
ctflne(X) absorption coefficient of fine particles, m"1
at.w(X) absorption coefficient excluding those of pure-water for unfiltered water samples, m"1
Ct-w(X) beam attenuation coefficient excluding those of pure-water for unfiltered water samples, m"1
85
_-i bp(X) scattering coefficient of suspended particulate matter, m" bp*(X) SPMT-specific scattering coefficient of suspended particulate matter,
mV at(X) total absorption coefficient, m"' bt(X) total scattering coefficient, m"1
aw(X) absorption coefficient of pure-water, m"1 (Pope and Fry 1997)
bw(X) scattering coefficient of pure-water, m"1 (Buiteveld et al. 1994)
b/at(X) ratio of total scattering to total absorption coefficient, dimensionless
fid(z;X) average cosine for downward irradiance, dimensionless
ad;i(z;X) diffuse absorption coefficient of optically active component i for downward irradiance, m"1
adcDOM(z;PAR) diffuse absorption coefficient of CDOM for PAR, m"1
adNAp(z;PAR) diffuse absorption coefficient of NAP for PAR, m"1
ad<t>(z;PAR) diffuse absorption coefficient of algal particles for PAR, m"1
ad:rine(z;PAR) diffuse absorption coefficient of fine particles for PAR, m"1
adw(z;PAR) diffuse absorption coefficient of pure-water for PAR, m"1
bbd(z PAR) diffuse back scattering coefficient for PAR, m"1
Kd'(z;PAR) estimated diffuse attenuation coefficient for PAR at depth z, m"1
Kd'(zave;PAR) estimated diffuse attenuation coefficient for PAR averaged over euphotic zone depth, m"1
Kd'(0;PAR) estimated diffuse attenuation coefficient for PAR at just below the surface, m"1
Kd'(zeu;PAR) estimated diffuse attenuation coefficient for PAR at euphotic zone depth, m"1
86
Table 3.1 Limnological conditions of Lake St. Charles in summer of 2008: chlorophyll a (Chi a), dissolved organic carbon (DOC), total suspended particulate matter (SPMT), inorganic suspended particulate matter (SPMi), and organic suspended particulate matter (SPMo). CV = coefficient of variation (standard deviation as percentage of the mean).
Variables units Range mean CV (%)
Chi a ugL"1 3.0-15.4 8.1 53
DOC mgL"1 2.47 - 4.20 3.43 16
SPMT + mgL"1 1.1-2.9 2.1 29
SPMif mgL"1 0.2-1.2 0.9 33
SPMo1 mgL"1 0.8-1.7 1.2 30
f Data from 8 June 2008 were missing (n = 7).
87
Table 3.2 Summary of inherent optical properties measured by a laboratory bench top spectrophotometer and a WET Labs AC-S spectropthometer: the absorption coefficients (a), specific absorption coefficients (a*), and exponential slope parameters (S) for absorbing components: colored dissolved organic matter (CDOM), non-algal (NAP), algal, and fine particles, the scattering coefficients of suspended particulate matter (bp), specific scattering coefficients (bp*), and the ratio of total scattering to total absorption coefficient (bt/at).
Optical variable units range mean CV (%)
CDOM CICDOM(320) m 1 9.919-18.768 15.297 19
OCDOM(440) tu 1.383-2.651 2.155 20
aCDOM*(440) + rn2 g 1 0.560 - 0.690 0.624 6
SCDOM nm"1 0.0155-0.0158 0.0157 1
NAP aNAp(440) m"1 0.276 - 0.552 0.438 19
ONAP*(440) ! mV 0.190-0.259 0.219 13
SNAP nm"1 0.0096 - 0.0106 0.0101 4
Algal particles a0(44O) m 0.075-0.160 0.119 25
a0(66O) m 0.056-0.142 0.096 34
a0*(44O) m2 mg"1 0.010 - 0.025 0.017 31
a0*(66O) m2 mg"1 0.009 - 0.019 0.013 23
Fine particles afine(440) m 1 0.156-0.450 0.257 43
Scattering bp(440) m 1 0.946-1.978 1.532 22
bp(555) m 1 0.783 - 1.737 1.312 24
bp*(440) m2 g"1 0.635 - 0.870 0.761 13
bp*(555) m2 g"1 0.537 - 0.755 0.652 13
b/a ratio bM(440) NA 0.407 - 0.781 0.588 19
bt/a,(585) NA 1.567-2.884 2.355 19
f Data from 8 June were missing (n = 7).
88
Table 3.3 Summary of the photon budget estimation: the estimated diffuse attenuation coefficient for PAR (Kd'(z;PAR)), the diffuse absorption coefficients of colored dissolved organic matter (ad;cDOM(zave',PAR)), non-algal particulate matter (ad;NAp(zave',PAR)), algal particles (a^ZaveiPAR)), fine particles (ad:Fine(zave;PAR)), and water (ad:w(zave;PAR)), the diffuse back scattering coefficient (bbd(zave;PAR)), and their percent contribution to Kd'(z;PAR).
Factor range mean CV (% Kd'(Zave;PAR) 0.911 -1.180 1.020 10 Cld;CDOM(Zave;PAR) 0.334 - 0.569 0.446 17 ad;NAp(Zave',PAR) 0.074 ■0.163 0.128 20 a^Zave lPAR) 0.023 - 0.079 0.054 35 a^FinefZay^PAR) 0.031 ■0.133 0.061 57 ad;w(Zave;PAR) 0.265 - 0.320 0.304 6 bb;d(zave;PAR) 0.017 - 0.038 0.029 24
%ad:CDOM(Zave,PAR) 37--48 44 9 %ad:NAp(Zave;PAR) 8- 15 13 17 % ad;<p(zave;PAR) 3 -7 5 31 %ad:fine(zave;PAR) 3 - 15 6 66 %ad;w(zave;PAR) 26 -33 30 8 %bb,d(zaVe;PAR) 2--3 3 18
89
Figure 3.1 Map showing the location and bathymetry of Lake St. Charles. The figure is taken from Tremblay et al. (2001).
90
400 450 500 550 600 650 700 Wavelength (nm)
Figure 3.2 The profile of natural log-transformed spectral downward photon flux density (Eq:d(z;X)) on 8 July 2008. The spectral profiles are shown from just below the surface to 5.0 m at 0.5 m intervals.
91
J 0 400 450 500 550 600 650 700
Wavelength (nm)
Figure 3.3 The spectral diffuse attenuation coefficient of downward photon flux density (Kd(X)). All observations (n = 8) are shown.
92
A CDOM B. NAP C algal particles
D. fine particles E. scattering F. b/a ratio
400 450 500 550 600 650 700 400 450 500 550 600 650 700 Wavelength (nm)
400 450 500 550 6(H) 650 700
Figure 3.4 Spectral profiles of inherent optical properties: the absorption coefficients of A. colored dissolved organic matter (CDOM), B. non-algal particulate matter (NAP), C. algal particles, and D. fine particles, E. the scattering coefficient of suspended particulate matter, and F. the ratio of total scattering to total absorption coefficient (b/a). All observations (n = 8) are shown.
93
0.8 1.0 1.2 1.4 1.6 1.8 2.0 K!(z;PAR) (m"1)
0
I
2
•S 3
Q 4
5
6 t — Measured — Estimated
0.8 1.0 1.2 1.4 1.6 1.8 2.0 - i , KJz;PAR) ( n i )
Figure 3.5 Vertical profiles of A. the estimated diffuse attenuation coefficient for PAR (Kd'(z;PAR)) for ail 8 sampling dates, and B. the estimated (solid line) and measured (broken line) diffuse attenuation coefficient for PAR (Kd'(z;PAR) and Kd(z;PAR), respectively) on 8 July 2008.
94
0.0 0.2 0.4 0.6 0.8 1.0 ad:i(z;PARJ (m1)
1.2
20 40 60 % contribution in KJ(z;PAR)
Figure 3.6 Vertical profiles of component-specific diffuse absorption/scattering coefficients for 8 July 2008: A. absolute values and B. proportional contribution to total diffuse attenuation coefficient for PAR in %.
95
2.0
H 1.5
aï _3 , 0
A. average over euphotic zone depth B average over euphotic zone depthin %
--. 0.5 -
00
pure water back scattering fine particles
algal particles NAP CDOM
llllllll C. at just below surface D. at just below surface in %
2.0 E. at euphotic zone depth F. at euphotic zone depth in %
S 1.5-
llllllllll *>. ^ _*> S > \°> *?> \°> --. y v̂ v̂ ^vv
\* _i% T> _**> \°» J> \°> "P
Date
Figure 3.7 Seasonal changes in the contribution of each absorption component to the diffuse attenuation coefficient for PAR (Kd'(z;PAR)): A. averaged over euphotic zone depth, C. just below the water surface, E. at the bottom of the euphotic zone. B, D, and F show the proportional contribution of each optical component to total Kd'(z;PAR) in %.
96
A. a/A) B. adJA)xE JA) at 0 m C.adJÀ)xEq:jA)at0.5m
D.aJAJxEJA) at 1.0 m E. adJA)xEqJA) at 2.0 m 0.4 /
° " _ r _ _* y 500 600 71
ilA___^___ls_^==--=^
2 --V-^A 2
M) 500 600 /
0.04
0.0.
0.02
0.01
F.aJA)xEq.JA)at4.0m
0.00 400 500 600 700
Wavelength (nm) 400
Figure 3.8 Spectra of the component-specific, per volume absorption/scattering of photon flux (ad;i(z; X)xEq;d(z;X)) at different depths down the water column: A. Measurements of the inherent optical properties of the different optical components. B-F. ad;i(z; X)xEq;d(z;X) at 0, 0.5, 1.0, 2.0, and 4.0 m, respectively.
97
Chapitre 4 Optical closure in a shallow urban lake: steps toward modeling of remote sensing reflectance in optically complex inland waters
4.1 Résumé La « fermeture optique » fait référence aux relations théoriques entre les propriétés
optiques apparentes et inhérentes (AOP et IOP). La modélisation directe (forward
modeling) des AOP à partir des IOP représente un défi majeur pour les masses d'eau
continentales qui sont optiquement complexes. Nous avons effectué une analyse de
fermeture du rayonnement photosynthétiquement actif (PAR) dans le lac Saint-Augustin,
Québec, Canada, où des fleurs d'eau d'algues toxiques sont fréquemment signalées durant
l'été. La mise en place d'un système de surveillance à la fine pointe de la technologie
permettrait de mieux gérer cette ressource en eau. Une bonne concordance a été observée
entre la réflectance spectrale, obtenue à partir de mesures in situ par un système de
profilage vertical, et celle estimée par un modèle de transfert radiatif (HydroLight), en
exploitant les données d'éclairement ambiant et les IOP mesurées par spectrophotométrie
AC-S. L'analyse par modélisation a montré qu'il y avait une forte sensibilité des résultats
aux mesures des IOP et au choix du protocole de calcul. Les résultats soutiennent la validité
et le potentiel de l'approche de modélisation HydroLight pour les eaux continentales, mais
soulignent aussi l'importance du choix de la correction pour les erreurs d'estimation de
l'absorption, et la nécessité de mesures précises de l'absorption dans la région des
longueurs d'onde proche de l'infrarouge.
98
4.2 Abstract Optical closure refers to the testing of theoretical relationships between apparent
and inherent optical properties (AOPs and IOPs). The modeling of AOPs from the IOPs is
referred to as forward modeling, and is an ongoing challenge for optically complex inland
waterbodies. We conducted a closure analysis for photosynthetically available radiation
(PAR) in Lake St. Augustin, Québec City, Canada, where frequent blooms of toxic algae
have been reported and where advanced monitoring systems are urgently needed for water
resource management. Good agreement was observed between the spectral remote sensing
reflectance obtained from in situ measurements by an in-water profiling system and that
estimated by a radiative transfer model (HydroLight), with input of ambient irradiance
conditions and measured IOPs made by an AC-S spectrophotometer. This forward
modeling analysis showed that there was a strong sensitivity of the HydroLight results to
the IOP measurement and data processing protocols. The results support the validity and
potential of the HydroLight modeling approach for inland waters, but also underscore the
importance of appropriate scattering error corrections for the absorption estimates, and the
need for accurate measurements of absorption in the near-infrared wavelength region.
99
4.3 Introduction The optical characteristics of natural waters can be defined by two classes of
variables: apparent optical properties (AOPs) that are dependent on the optical
characteristics of the medium and ambient light conditions, and inherent optical properties
(IOPs) that depend only on the optical characteristics of the medium and that are
independent of ambient light (Kirk 1994; Mobley 1994). Optical closure refers to the
testing of theoretical interrelationships between these two classes of optical properties
(Zaneveld 1994), while modeling the AOPs from the IOPs is referred to as forward
modeling (International Ocean Color Coordinate Group (IOCCG) 2000). Forward
modeling is required to understand radiative transfer processes in optically complex
waterbodies, where a variety of optically active components vary independently (IOCCG
2000). In particular, the modeling of remote sensing reflectance (Rrs(X)) from the input of
IOPs is required for the development of robust remote sensing algorithms for such waters
(IOCCG 2000). Additionally, uncertainties and sensitivities in the optical measurements
can be examined through closure analysis (Gallegos et al. 2008). Despite its central
importance in hydrologie optics, forward modeling has been largely restricted to the open
ocean, with only a few notable exceptions of optically complex systems (Belzile et al.
2004; Gallegos et al. 2008; O'donnell et al. 2010). Validation of such models is therefore
required to improve our understanding of the optical processes in lakes, reservoirs and
coastal waters.
An essential requirement for the forward modeling of optically complex waters is
the production of high quality IOP data (Gallegos et al. 2008). Such data are now
increasingly obtained in océanographie optics using AC meters (AC-9 or AC-S; WET Labs
Inc., USA), which are in situ spectrophotometers that measure both absorption and beam
attenuation coefficients (e.g., Babin et al. 2003a; Belzile et al. 2004; Gallegos et al. 2008).
These instruments are equipped with two sets of light sources and detectors, enclosed by
flow tubes that are filled with the sampled water. One flow tube has a non-reflecting
interior surface coating that absorbs the scattered light, and the measured value is therefore
assumed to be the beam attenuation coefficient minus that for pure-water (ct.w(X)). The
second flow tube is coated by quartz and is assumed to reflect all scattered photons back to
100
the detector (i.e., the scattering loss of photons is negligible); thus, the measured value is
considered to be the absorption coefficient minus that for pure-water (at.w(X)), even for
media with scattering components (Pegau et al. 2003). The advantage of the instrument is
that both at.w(X) and ct.w(X) are measured simultaneously, and the scattering coefficient of
suspended particulate matter (bp(X)) can be obtained by difference.
AC spectrophotometers have now been applied to obtain IOPs for forward modeling
in optically complex waters within a limited range (e.g., Gallegos et al. 2008; O'donnell et
al. 2010). However, there are uncertainties associated with the at.w(X) measurements made
by these instruments (Leymarie et al. 2010; Pegau et al. 2003). Although it is expected that
all scattered photons are conserved and detected at the detector in the reflective tube,
backscattered photons may not reach the detector in the actual measurements (Kirk 1992;
Leymarie et al. 2010; Pegau et al. 2003; Zaneveld et al. 1994). This would result in
overestimation of at.w(X), and the error could be large for highly scattering media.
To overcome the problem of scattering error, several correction methods have been
proposed. Among the first of these, the null point method, the following equation is
applied:
at_w(A) = am(A)-am(ANlR) eq. 4.1,
where am(X) is the measured absorption coefficient at X nm, and am(XNin) is that in the near-
infrared radiation (NIR) range (Pegau et al. 2003). This method assumes that the absorption
of optically active components in the water sample is negligible at NIR wavelengths. Thus,
amiXmiù is considered to be scattering and is subtracted from the rest of the spectrum. This
scattering correction has been frequently applied to other spectrophotométrie measurements
in hydrologie optics (Mitchell et al. 2002).
A second approach towards the scattering correction is the proportional method,
implemented as:
101
a,_w(A) = am(A)-a m(-^NIR)
c m ( ^ N m . ) ~ a m ( ^ a m ) "[_W-_W] eq. 4.2
where cm(X) is the measured beam attenuation coefficient and cm(XNnO is that at NIR (Pegau
et al. 2003; Zaneveld et al. 1994). In addition to the assumption of negligible absorption at
NIR wavelengths, a spectral dependency of scattering is assumed for subtraction of the
scattering error from the rest of the spectrum. This is the manufacturer's recommended
method and is the most commonly applied (Pegau et al. 2003).
Gallegos et al. (2008) improved the proportional correction method by assuming
significant absorption at NIR region. They applied the NIR absorption value measured by a
laboratory bench-top spectrophotometer in the proportional method as follows:
at_w(A) = am(A)-a m i^NTR ) ~ aCary (^NDt )
* k (*)-«„(*)] eq. 4.3
where acary(XNiiO w a s the absorption coefficient at NIR measured by the laboratory bench-
top spectrophotometer. This method significantly improved their results of forward
modeling in optically complex inland waters.
Finally, Leymarie et al. (2010) further improved the proportional method via an
extensive computer modeling analysis. They found that the at.w(X) corrected by the
proportional method with consideration of significant absorption at NIR range still includes
a residual scattering error that is proportional to the ratio of the scattering coefficient to the
cm(X). They estimated the proportional error on the corrected at.w(X) (%aerror(X)) as:
%aerror(A) = 1-
0.367 ■11.303 eq. 4.4
102
where bc(X) is the scattering coefficient corrected by the proportional method. This method
has not been applied to measurements in natural waters, to our knowledge.
The overall objective of the present study was to conduct the first assessment of
optical closure for an eutrophic lake. We selected an urban lake, Lake St. Augustin at the
edge of Québec City, Canada, which has experienced advanced eutrophication over the last
few decades, and where there is an urgent need for monitoring, including the application of
optical techniques (Pienitz and Vincent 2003). We compared field measurements of remote
sensing reflectance with modeled results based on measured IOPs to test the validity of the
model. We also evaluated the above four different scattering error correction methods on
IOP measurements made by an in situ spectrophotometer to test their validity and identify
the best approach for future applications.
4.4 Materials and Methods 4.4.1 Study site and field sampling
Lake St. Augustin (46°42TSI, 71°22'W) is located 20 km west of Québec City,
Québec, Canada. It has a surface area of 0.7 km2, an average depth of 3.5 m and a
maximum depth of 6.1 m (Pienitz et al. 2006; Pienitz and Vincent 2003). Sampling was
conducted on 5 October 2010 at the center of the lake within 1 h of local solar noon. At the
time of sampling, the sky was clear and the water surface was smooth because of low wind
speeds. Surface water (approximately 50 cm below the surface) was taken with Nalgene
amber polyethylene bottles (Thermo Fisher Scientific Inc., USA) and kept refrigerated in
the dark until processing in the laboratory within 5 h of collection. Temperature (°C),
dissolved oxygen (%), and Chlorophyll a fluorescence (f-Chl a, pg L"1) profiles were taken
by a YSI 6600 sonde (YSI Inc., USA).
4.4.2 Apparent optical properties (AOPs) Spectral downward irradiance (Ed(z;X), W m"2 nm"1) and upward radiance (Lu(z;X),
W m"2 nm"1 sr"1) were obtained with a hyperspectral radiometer system (Satlantic Inc.,
Canada). Above-water downward irradiance (Ed(0+;Xj) was measured at 50 cm above the
103
water surface, and the upward radiance just below the surface (Lu(0';X)) was measured by
stably holding the instrument with the upward radiance detector touching the water surface.
The instrument was placed at approximately 1 m away from the boat side to avoid boat and
observer shading. This instrument manipulation was aided by calm surface water
conditions (no wind). The continuous data readings were recorded for 20 s, and averaged
over the sampling period. The instrument was then manually lowered at approximately
5 cm s"1 to obtain the vertical profiles of Ed(z;X) and Lu(z;X). Raw radiometer data were
linearly interpolated to obtain the spectral profiles from 352 to 797 nm at 1 nm intervals,
and the depth profiles at 10 cm intervals. Correction of instrumental self-shading effects on
Lu(z;X) was not conducted on the present data set. The suggested method for the latter
(Mueller 2002) was originally developed for highly absorbing but low scattering media
(Gordon and Ding 1992), and may not be appropriate for the highly scattering media of
inland waters.
Remote sensing reflectance (Rrs(X), sr"1) was calculated as:
Rrs(X) = 0.543 Lu(a;X)/Ed(0+;X) eq. 4.5
where, 0.543 is the upward radiance transmittance of the water surface for normal
incidence from below (Mueller 2002). Downward irradiance for PAR (Ed(z;PAR)) was
obtained by integrating Ed(z;X) over the PAR range (400-700 nm), and the attenuation
coefficient of downward irradiance for PAR (Kd(PAR)) was obtained by regressing the
natural log of Ed(z;PAR) against depth.
4.4.3 Limnological variables Seston samples were filtered onto GF/F glass fiber filters (Whatman Inc., USA) and
stored frozen (-80°C) until laboratory analysis for chlorophyll a (Chi a, pg L"1). Pigments
were extracted in ethanol and Chi a concentration was determined by fluorometry (Cary
Eclipse spectrofluorometer, Varian Inc., Canada) before and after acidification (Nusch
1980). Water samples were also filtered through pre-combusted, pre-weighed GF/F filters
to quantify suspended particulate matter. Filters were weighed after drying for 2 h at 60°C
104
to determine total suspended particulate matter (SPMT, mg L"1). Subsequently, filters were
combusted at 500°C for 2 h, and then inorganic suspended particulate matter (SPMi,
mg L"1) was quantified. Organic suspended particulate matter (SPMo, mg L"1) estimates
were obtained as the difference between SPMT and SPMi. Dissolved organic carbon (DOC,
mg L"1) concentrations were measured with a TOC-5000A carbon analyzer (Shimadzu Co.,
Japan) on filtrates passed through pre-rinsed nitrocellulose membrane filters (25 mm
diameter, 0.22 pm pore size; Millipore Co., USA).
4.4.4 Inherent optical properties (IOPs) 4.4.4.1 Absorption coefficient of colored dissolved organic matter
Water samples were filtered through GF/F filters and subsequently through 0.22 pm
membrane filters, and the resultant filtrates were refrigerated in amber glass bottles until
analysis of colored dissolved organic matter (CDOM). The absorbance (A(X),
dimensionless) of the filtrate against pure-water prepared by a EASY pure II UV/UF
system (Barnstead International, USA) was measured from 200 to 850 nm at 1 nm intervals
(spectral slit width of 2 nm) in a 1-cm quartz cuvette using a Cary 100 dual beam
spectrophotometer (Varian Inc., Canada). The absorption coefficients of CDOM (acDOM -̂),
m"1) were calculated as:
. . . 2.303.4(A) a cDOMW= ; e q - 4 - 6
where L is the path length of the cuvette (Bricaud et al. 1981; Mitchell et al. 2002). Null
point correction was conducted by subtracting the average of OCDOM(X) from 801 to 850 nm
where the absorption curves became flat. The values at shorter wavelengths have
commonly been used for oceanic samples with much lower CDOM absorptions; however,
selection of the null point should be carefully determined by observing spectral curve for
highly colored inland waterbodies (Mitchell et al. 2000; 2002). The exponential slope
parameters of the spectra (SCDOM) were calculated by non-linearly fitting the following
equation to the measured values from 300 to 600 nm:
105
aCDouW=acDOM(K)-e~SaMU 'x~^ eq. 4.7
where Xn is the reference wavelength (Bricaud et al. 1981; Stedmon et al. 2000). The
DOC-specific absorption coefficient of CDOM (OCDOM*(X), m g" ) was calculated as
acDOMfX) divided by DOC concentration.
4.4.4.2 Absorption coefficients of suspended particulate matter
Water samples were filtered onto GF/F filters and stored at -80°C until optical
analysis. The absorbance of particulate matter was measured by the wet filter technique in
the same spectrophotometer as above fitted with an integrating sphere (Labsphere Inc.,
USA), and following the method of Mitchell et al. (2002). The absorbance of non-algal
particles was measured by extracting the phytoplankton pigments with methanol (Kishino
et al. 1985; Mitchell et al. 2002). The absorption coefficients of total and non-algal
particulate matter (ap(X) and ONAP(X) in m"1, respectively) were calculated as:
2.303A(A)T B V a p ( A ) o r a N A P ( A ) = „ ^ r i eq. 4.8
where T is the filtered area, V is filtered volume, and B is the path length amplification
factor (Mitchell et al. 2002). A value of /? = 2 was used after Roesler (1998) (See, Annexe 3
for further discussion regarding /J). Null point correction was applied in the same manner as
above (null point range: 801-810 nm). The absorption coefficient of algal particles (a0(X),
m"1) was obtained by subtracting ONAP(X) from ap(X). Similar to OCDOM(X), the exponential
slope parameter of the ONAP(X) spectra (SNAP) was calculated by fitting equation 4.7 to the
data. Non-linear regression was conducted over the spectral range of 380-730 nm excluding
the ranges 400-480 and 620-710 nm to eliminate the residual pigment absorption (Babin et
al. 2003a; Belzile et al. 2004). The SPMT-specific absorption coefficient of non-algal
particulate matter (ONAP*(X), m2 g"1) and Chi a-specific absorption coefficient of algal
particles (a0*(X), m2 mg"1) were also calculated. The GF/F filtrate was filtered through the
0.22 pm membrane filters and the absorption coefficient of materials collected on the filters
(fine particles, af,„e(X), m"1) was determined using the same method as for ap(X)
106
quantification described above (Ferrari and Tassan 1996; Gallegos 2005). The average of
afme(X) from 771 to 780 nm was subtracted as the null point correction.
4.4.4.3 AC-S measurements
The absorption and beam attenuation coefficients excluding those of pure-water for
unfiltered water samples (at.w(X) and ct.w(X) in m"1, respectively) were measured using a
AC-S in situ spectrophotometer (25 cm path length, WET Labs Inc., USA) on the
laboratory bench (Watanabe et al. 2011). In addition to measurements on the raw water
sample, the absorption coefficient of CDOM (acDOM-Ac(X)) was measured on water samples
filtered by GF/F and subsequently by the 0.22 pm membrane filter. Pure-water calibration
values obtained immediately before the sampling period were subtracted from the raw
readings (Pegau et al. 2003; Sullivan et al. 2006; Twardowski et al. 1999). A temperature
correction was applied by using the spectral correction coefficients obtained specifically for
our instrument (Pegau et al. 2003; Sullivan et al. 2006; Twardowski et al. 1999). The
measured spectral ranges of our instrument were 403.0-751.7 nm and 401.6-750.8 nm for
a,.w(X) and c,.w(X), respectively. These obtained spectra were linearly interpolated for the
range 403-750 nm at 1 nm intervals, and also extrapolated to 400 nm by fitting the
measured values from 403 to 410 nm to the exponential model as in the equation 4.7. The
four different scattering error correction methods discussed above (eq. 4.1 -4) were tested on
the at.w(X). The absorption and beam attenuation coefficients at 750 nm were applied, as for
NIR corrections. The acary(Xmiù of the equation 4.3 was estimated as 0.071 m"1 by summing
the absorption coefficient at 750 nm measured by the laboratory bench-top
spectrophotometer as:
aCary(XNiR) = CICDOM(750) + ap(750) + afme(750) eq. 4.9
The scattering coefficients of suspended particulate matter (bp(X), m"1) were then obtained
by subtracting at.w(X) from ct.w(X) for each scattering error correction method described
above. The SPMT-specific scattering coefficients for suspended particulate matter (bp*(X), 0 1
m g" ) were obtained by dividing bp(X) by SPMT. Finally, the total absorption coefficient
(at(X), m"1) was obtained by adding the absorption coefficient of pure-water (aw(X), m" ;
107
Pope and Fry 1997) to the a,.w(X), and the total scattering coefficient (bt(X), m"1) was
obtained by adding the scattering coefficient of pure-water (bw(X), m"1; Buiteveld et al.
1994) to bpfA).
4.4.5 Radiative transfer modeling (RTM) The radiative transfer model HydroLight 5.0 (Sequoia Scientific Inc., USA), was
applied for forward modeling. The AC-S data (at.w(X) and ct.w(X)) corrected with the above
four different scattering error correction methods were considered as the IOP input. Built-in
values were used for the pure-water absorption and scattering coefficients. We used the
built-in Fournier-Forand scattering phase function with a bb(X):bp(X) ratio of 0.018 (Petzold
1972) for back scattering estimation. The f-Chl a concentrations profile measured by the
YSI sonde, and a0*(X) measured on the discrete water sample were provided for the
chlorophyll fluorescence estimation. CDOM fluorescence estimation was also added to the
model by applying OCDOM-AC(X). In addition to these fluorescence estimations, inelastic
scattering (Raman scattering) was also included. For incoming radiation modeling, field
measurements ofEd(0+;X), sampling latitude and longitude, and sampling time were applied.
4.5 Results 4.5.1 Ambient conditions and IOPs
Limnological conditions measured on the discrete surface water sample were: DOC
6.0 mg L"1, SPMT 6.5 mg L"\ SPMi 3.7 mg L"1, and Chi a 16.0 pg L *. The vertical profiles
measured by the sonde showed that the lake was weakly stratified at 1 m and there was a
subsurface chlorophyll maximum at the time of sampling (Fig. 4.2). The IOP spectra of the
surface water sample obtained by the bench-top spectrophotometer are given Figure 4.3.
For CDOM absorption (Fig. 4.3a), aCDOM<440) was 1.979 m"1, aCDOM*(440) was 0.330 m2
g"1, and SCDOM was 0.0155 m". Non-algal particles (Fig. 4.3b) showed an ONAP(440) value
of 0.523 m"1, aNAp*(440) 0.080 m2 g"1, and SNAP 0.0094 m"1. For algal particles (Fig. 4.4c),
a0(44O) was 0.242 m"1 with a0*(44O) of 0.015 m2 mg"1. Fine particles showed a similar
exponential curve as for CDOM and NAP (Fig. 4.3d) with the afme(440) of 0.346 m"1 and
Sfine 0.0130 m". The IOP spectra measured by the AC-S using the proportional method (eq.
4.2) are given in the Figure 4.4. The at.w(X) spectra had exponential slopes with a
108
chlorophyll absorption peak at 680 nm (solid line, Fig. 4.4a); The at.w(440) value was 2.672
m"1. The bp(X) spectra showed spectral dependency with a depression around the
chlorophyll absorption peak (broken line, Fig. 4.4a); bp(440) was 4.079 m"1 and bp*(440)
was 0.628 m2g"'. The ratio of b,(X) to a,(X) peaked around 580 nm (Fig. 4.4b) with
b,/a,(580) of 6.519 and btlat(440) of 1.525 (dimensionless).
4.5.2 Optical closure The Rrs(X) spectrum obtained by the in-water radiometer system (Fig. 4.5) showed a
typical spectral shape for waterbodies containing a mixture of CDOM, non-algal particles,
and algal particles in moderate concentrations (IOCCG 2000). The peak was observed
around 580 nm (Rrs(580) = 0.00482 sr"1). The measured Rrs(X) spectrum is compared with
modeled Rrs(X) in the Figure 4.6. The AC-S data with the null point method underestimated
Rrs(X) across most of the spectral region (Fig. 4.6a). The root mean square error (RMSE)
was 5.66 x IO"4 sr"1, and relative root mean square error (RRMSE) was 19.9 %. When
modeled Rrs(X) was linearly regressed against measured values, the slope was 0.89 with an
r2 of 0.89. The proportional method considerably improved the result relative to the null
point method (RMSE = 3.87 x 10"4 sr"1, RRMSE = 9.6 %; Fig. 4.6b). However, it
overestimated Rrs(X), with a regression slope of 1.23 (r = 0.99). The Gallegos et al. (2008)
method further improved the result (Fig. 4.6c). The RMSE was 2.92 x 10"4 sr"1 and RRMSE
was 10.7 %. The regression slope was 0.96 with r2 of 0.99. Finally, the Leymarie et al.
(2010) method produced the best closure result, with an RMSE of 1.95 x 10"4 sr"', and
RRMSE of 8.2 % (Fig. 4.6d). The regression slope was 1.02 and r2 was 0.99. When the
Rrs(X) spectra were normalized to the maximum value in the range from 400 nm to 700 nm,
the spectral shapes were similar among measured and modeled values, with the exception
of the spectrum modeled with IOPs corrected by the null point method (Fig. 4.7).
The inclusion of fluorescence and inelastic scattering estimation to the modeling
was tested using the AC-S data corrected with Leymarie et al. (2010) method. The
chlorophyll and CDOM fluorescence estimates resulted in small changes in modeled Rrs(X)
spectra and improved the modeled result (Fig. 4.8). Inclusion of the Raman scattering
estimate did not result in a significant change in the modeled spectra.
109
4.6 Discussion 4.6.1 Limnological conditions
Lake St. Augustin has been identified as a hypereutrophic lake, with frequent
reports of nuisance algal blooms over the past decade (Bouchard-Valentine et al. 2004;
Pienitz et al. 2006). The Chi a concentration at the time of sampling was relatively low by
comparison with previously observed summer values (Bouchard-Valentine et al. 2004),
possibly due to the low water temperatures in late fall, but within the range for eutrophy.
The vertical profile showed that the f-Chl a concentrations had a subsurface peak at 2 m
depth. This higher concentration is unlikely to contribute significantly to the surface
reflectance, since the estimated depth range where 90 % of Rrs originates (ZOQ, Gordon and
McCluney 1975) was 0.59 m for PAR (Kd(PAR) = 1.691 m"1).
4.6.2 IOPs for optically active components The IOPs for the surface water sample provided a description of the optically active
components in the lake. The high OCDOM*(X) and low SCDOM values indicate that the
dissolved organic matter in this water is highly colored (Morris et al. 1995; Stedmon et al.
2000). Outside the bloom period of the lake, this highly colored dissolved organic matter
may play a dominant role in underwater absorption processes. Thus, it is important to
understand the detailed characteristics of the component such as its source and seasonal
dynamics for future optical monitoring and modeling of the lake ecosystem. The non-algal
particles in this lake are also highly absorbing relative to those previously reported (Babin
et al. 2003b; Belzile et al. 2004; Binding et al. 2008). This may be related to the size of the
particles or their composition (Babin et al. 2003b; Binding et al. 2008; Stramski et al. 2007),
and further analyses are required. Algal particles played a minor role in underwater light
attenuation at the time of sampling. The Chi a-specific absorption was comparable to that
of previously reported values for inland waters (Dekker et al. 2002; Zhang et al. 2010).
Fine particles were a significantly absorbing component in Lake St. Augustin, with
absorption coefficients comparable to those of non-algal particles. The strong exponential
slope with a small peak around 680 nm (Fig. 4.3d) implies that particles in this size fraction
are dominated by non-algal particles with only a minor contribution of small-sized algal
110
particles that passed through the glass fiber filters (GF/F). Due to their small size, these fine
particles may have much higher mass-specific absorption relative to larger particles (Babin
et al. 2003a); however, this could not be evaluated from the present data set. The particles
in this size range are particularly important to understand for the future modeling of
underwater light and reflectance in optically complex inland and coastal waterbodies
(Gallegos 2005; Watanabe et al. 2011), but have received little attention to date. Thus,
further evaluation of their characteristics is required in future research.
4.6.3 Optical closure In the forward modeling analysis, the four different scattering error correction
methods on IOP measurements yielded considerable variation in resulting Rrs(X). The best
result was obtained with the Leymarie et al. (2010) method (Fig. 4.6). These results clearly
demonstrated the importance of the scattering error correction for IOP measurements made
by AC meters, and its subsequent influence on optical modeling. The null point method
makes no adjustment based on NIR absorption nor does it take into account the wavelength
dependency of scattering (eq. 4.1), and this approach provided the least successful fit to the
observed data. It showed poor agreement with measured Rrs(X), not only in magnitude (Fig.
4.6a) but also in spectral shape (Fig. 4.7). The other correction methods, incorporating
wavelength dependency of scattering, produced a spectral shape that was consistent with
the measured Rrs(X) (Fig. 4.7), but that differed in absolute values (Figs. 4.6b-d). These
differences in modeled Rrs(X) among three different AC-S data inputs originated from the
small differences in a,.w(X) (Fig. 4.9a). The differences in at.w(X) between the proportional
and Gallegos et al. (2008) methods were 0.117 and 0.076 m"1 at 440 and 700 nm,
respectively. The difference between the proportional and Leymarie et al. (2010) methods
at the same wavelength were 0.085 and 0.053 m" , respectively. These small differences
subsequently resulted in change of the bp(X) that is obtained as the difference between
ct.w(X) and a,.w(X), and they ultimately resulted in a relatively large change in the b/at(X)
ratio (Fig. 4.9b), an important determinant of Rn(X) (Kirk 1994; Mobley 1994).
I l l
4.6.4 Uncertainties in IOP measurements 4.6.4.1 Near-infrared absorption
The differences in at.w(X) produced by the Gallegos et al. (2008) and Leymarie et al.
(2010) methods relative to the proportional method were the result of the application of the
significant absorption at NIR range as a key factor (acary(XNift) of the equation 4.3; Gallegos
et al. 2008). Determination of this value is, therefore, critical because even a small change
in acaryiXmp) translates into changes over the entire a,.w(A) spectra, with considerable
resultant change in modeled Rrs(X). In the present study, the NIR absorption coefficient was
estimated as the sum of three different absorption coefficients measured by the bench-top
spectrophotometer as shown in equation 4.9. The actual values were 0.048, 0.021, and
0.002 m"1 for acDOMf/50), ap(750), and afme(750), respectively. Among them, the
QCDOM(750) may not include the source of uncertainty that produces the large differences in
the results; however, two particulate absorption measurements (ap(750) and afine(750)) may
incorporate uncertainties as discussed below.
4.6.4.2 Null point correction of the quantitative filter technique (QFT)
The ap(X) was measured by the quantitative filter technique (QFT), also known as
the wet filter technique, and this approach has been widely adopted for a great variety of
natural waters (Kishino et al. 1985; Mitchell et al. 2002; Yentsch 1962). The uncertainties
of this absorption measurement are associated with two factors: near-infrared absorption
(ap(XNiiO) and the B factor. Similar to other spectrophotométrie measurements, null point
correction subtracting ap(XNiR) from the rest of the spectral range has been widely accepted
and applied for ap(X) analysis (Mitchell et al. 2002). However, selection of the wavelength
range for ap(XmR) must be done with caution depending on absorption characteristics of the
particles (Mitchell et al. 2000). In the present study, determination of the null point range
by observing the raw absorption spectra was difficult because the spectra showed weak
spectral dependency over the measured range (Fig. 4.10a). Therefore, the average of
aP(Xmp) between 800 and 810 nm where CDOM absorption spectra flattened (Fig. 4.10b)
was applied as the null point value, and this gave the ap(750) as 0.021 m"1 as applied to the
scattering error correction of AC-S values above. The ap(750), however, could vary
depending on the wavelength range chosen for the null point. If the shorter wavelength
112
range from 751 to 760 nm was chosen, the ap(750) would be 0.002 m"1. On the other hand,
if the longer wavelength range from 841 to 850 nm was applied, the value would be 0.032
m" . This potential variation in ap(750) could be a critical issue for scattering error
correction of AC-S measurements and subsequent forward modeling. Therefore, careful
analysis of particulate absorption in NIR range is required for the development of robust
methods.
4.6.4.3 The path length amplification factor
In addition to the null point correction, the path length amplification factor, B, can
be a significant source of uncertainty. This coefficient corrects the overestimation of the
absorption value caused by multiple scattering that occurs in the glass fiber filter (Mitchell
et al. 2002). Various methods to approximate B have been proposed based on theoretical
(Roesler 1998) and empirical analyses (See, the equations 4.6a and b, and table 4.1 in
Mitchell et al. 2002), and a /? value of 2 has been proposed as an acceptable approximation
(Bricaud and Stramski 1990; Roesler 1998). The value could, however, vary depending on
the measured absorbance value (Bricaud and Stramski 1990). Also, a number of factors are
likely to cause variation in /?, such as the configuration of the spectrophotometer and the
composition of the suspended particles. Studies on the B factor for the quantitative filter
technique (QFT) and its variability have mostly been conducted on either algal culture or
natural waters with the seston dominated by algal particles (open oceans). There are only a
few studies on B for inland and coastal waters where non-algal particles dominated
(Gallegos and Neale 2002; Tassan and Ferrari 2002). In the present data set, the usual B
value of 2 was adopted and this requires close evaluation in the future for optically complex
waterbodies.
4.6.4.4 The absorption coefficient of fine particles
The afine(X) measurements also incorporate uncertainties associated with NIR
absorption and the B factor as for those discussed for the QFT with glass fiber filters. The
characteristics of these errors would be significantly different, because optical
characteristics of the 0.22 pm membrane filters likely differ from those of GF/F filters.
113
4.6.4.5 Future improvement opportunities of particulate absorption measurements
Widely used absorption measurements incorporate uncertainties associated with
scattering, which would significantly affect subsequent modeling. A number of alternative
methods to minimize the error have been proposed: the transmittance-reflectance method of
QFT (Tassan and Ferrari 1995); the filter freeze and transfer technique (Tassan and Allali
2002; Tassan and Ferrari 1995); a sample filled cuvette placed at the entrance port of an
integrating sphere (e.g., Laurion et al. 2003); a sample filled cuvette placed at the center of
an integrating sphere (Babin and Stramski 2002; Nelson and Prezelin 1993; Tassan and
Ferrari 2003); and the point source integrating cavity absorption meter (PSICAM; Kirk
1997; Rottgers et al. 2007; Rottgers et al. 2005). Among this great variety of methods, a
sample filled cuvette placed at the center of an integrating sphere (Babin and Stramski
2002; Nelson and Prezelin 1993; Tassan and Ferrari 2003) appears to be the best method to
measure the absorption coefficient for samples with high concentrations of non-algal
particles. The configuration of the sample in the sphere allows all scattered photons to be
detected, including those that are back-scattered (Storm et al. 1998), and thus the scattering
error is minimized. Also, particles are all kept in suspension, which is a condition much
closer to natural water compared to the filter pad methods. Thus, the absorption coefficient
determined by this method may be considered as close as possible to the true absorption
coefficient of natural waters. This method has been tested for particle absorption
measurements of natural water samples (Babin and Stramski 2002; Nelson and Prezelin
1993; Tassan and Ferrari 2003). However, the published results are still limited and a full
comparison with conventional methods is not yet available.
4.6.5 Importance of fluorescence estimations on the RTM Correction for chlorophyll a fluorescence, estimated from the vertical profile of
f-Chl a concentration and the a0*(X), further improved the Rrs(X) estimation in the near-
infrared region. This is consistent with previously reported results for inland waterbodies
(Tzortziou et al. 2006), and implies the importance of fluorescence in estimation of Rrs(X)
of optically complex waters in the wavebands between 600 and 700 nm. Correction for
CDOM fluorescence also gave an improvement in Rrs(X) in the 450 to 650 nm range (Fig.
4.8). The adjustment was small in proportional terms, but the absolute change was
114
comparable to that for the chlorophyll fluorescence correction, and may be important in
highly colored inland waters. Raman scattering made little difference to the modeling
throughout the PAR wavelength range (Fig. 4.8), and may not have to be applied in this
type of water. Back scattering, which was estimated by a built-in model (Fournier-Forand
scattering phase function with the bb(X):bp(X) ratio of 0.018), may vary in optically complex
waters, and can be an important determinant of Rrs(X) (Gallegos et al. 2008). The variation
would be controlled by the composition (Boss et al. 2004; O'donnell et al. 2010) and size
distribution of particles (Snyder et al. 2008). Careful evaluation of this variable will be
required for future research.
4.7 Conclusions The results of this optical closure analysis showed that the radiative transfer model
HydroLight provides a useful tool to estimate the spectral remote sensing reflectance from
input of measured IOPs in optically complex inland waters. The comparison of modeled
results produced by different scattering error correction methods on the IOP measurements
underscored the sensitivity of the HydroLight modeling to small differences in IOP inputs.
This result indicates the importance of adopting the appropriate correction method to
perform valid forward modeling of highly scattering inland waters. Future modeling will
require an improved understanding of particulate absorption, especially in the near-infrared
region.
115
- - - - Extent of drainage basin (catchment)
Figure 4.1 Map showing the location and bathymetry of Lake St. Augustin. The figure is taken from Pienitz et al. (2006)
116
0 -
1 — • — - Chlo ^ \ _
1 - Temp / - Temp / £
j 2 ■ — A - - DO J /
a. __>
Q 3 -
4 A A I
— i i i
»
0 2 8 10 12 14 ChlaCugL")
50 1
60 i i
70 80 %DO
1
00
1
1
100
1
12.0 12.5 13.0 13.5 14.0 14.5 15.0 Temperature (°C)
Figure 4.2 Vertical profiles of physical and biological variables in Lake St. Augustin (5 October 2010) measured by the YSI 6600 sonde. Circles are chlorophyll a concentrations (f-Chl a, pg L"1), crosses are temperature (Temp, °C), and triangles are concentrations of dissolved oxygen expressed as percentage saturation (DO, %).
117
a CDOM b. NAP 4 i 1.0 | 4 i
aC£)Cft/(440)-= 1.979 m-1 1.0 |
aV4p(440) = 0.523 m'1
V \ acoo \ * ( 4 4 ° ) = °- 3 3 0 m~' ^ 0.8
^ 0 . 6
v flv^*(440) = 0.080 m'1
\ W = 00155 ^ 0.8
^ 0 . 6 \ SNAP = 0.0094 R 2 —. Ss Si 0.4 i ,
tf l 5 0.2
0 0.0 0 0.0
c. algal particles d. fine particles 0.30
0.25
0.8 0.30
0.25 a0(44O) = 0.242 m"1
0.8 0^(440)= 0.346 m"1
. — > / ^ Y V ( 4 4 0 ) = 0.015 m-1 -_--0.6 v Sr =0.0130 ■g 0.20 'g \ fine "•"»>"»
" 0.15 -v0.4 R ^ s . 0.10 *_» <3 5>T).2
0.05 n on n n
400 450 500 550 600 650 700 Wavelength (nm)
400 450 500 550 600 650 700 Wavelength (nm)
Figure 4.3 The inherent optical properties of the surface water of Lake St. Augustin measured by the bench-top spectrophotometer, a. the absorption coefficient of colored dissolved organic matter (OCDOM(X), m"1), b. non-algal particles (ONAP(X), m'1), c. algal particles (a0(X), m"1), and d. fine particles (afme(X), ra'1).
118
5 i a. absorption and scattering
5 i s a ( \)
k 4
— . o 2 ^ 1 o1
n b. Total scattering to absorption ratio
400 450 500 550 600 650 700 Wavelength (nm)
Figure 4.4 The inherent optical properties of Lake St. Augustin surface water measured by the AC-S. a. the solid line shows the absorption coefficient excluding those of pure-water (at.w(X)) corrected by the proportional correction, and the broken line shows the scattering coefficients of suspended particulate matter (bp(X)). b. The ratio of total scattering coefficient to total absorption coefficient (b/at(X), dimensionless).
119
0.005 -
0.004
S 0.003
-,0.002 ft!
0.001
0.000 400 450 500 550 600 650 700
Wavelength (nm)
Figure 4.5 Spectral remote sensing réflectance (Rrs(X), sr"1) of the Lake St. Augustin measured by the in-water radiometers.
120
0.006
0.005
_„_. 0.004
■2* 0.003
-£0.002
0.001
0.000
a. Null correction
RMSE = 5 6 6 x 10" RRMES = 19.9
b. Proportional correction
rf*b O 0
0
0/ ^ - ^ _ °
^ 5 ^ RMSE = 3.87x IO"4
« ^ RRMES = 9.6
c. Gallegos et al. (2008)
RMSE =2.92x 10" RRMES = 10.7
d. Leymarie et al. (2010)
RMSE = 1 95x IO"4
RRMES=8 2
400 450 500 550 600 650 700 400 450 500 550 600 650 700 Wavelength (nm) Wavelength (nm)
Figure 4.6 Results of the optical closure analysis. The solid lines show the spectral remote sensing reflectance (Rrs(X), sr"1) as measured by the in-water radiometers, and the circles represent Rrs(X) modeled with the AC-S measurements and corrected by a. the null correction method; b. the proportional correction method; c. the Gallegos et al. (2008) method; and d. the Leymarie et al. (2010) method. Root mean square error (RMSE) and relative root mean square error (RRMSE) are shown.
0.0
121
Proportional Gallegos et al. (2008) Leymarie etal. (2010)
, , , , , 450 500 550 600 650 700
Wavelength (nm)
Figure 4.7 Normalized remote sensing reflectance spectra (Rrs(X), sr" ) for Lake St. Augustin. The thick solid line shows the measured Rrs(X), the thin solid line is the Rrs(X) modeled with the AC-S measurements and null correction; the broken line is by the proportional correction; the dotted line is by the Gallegos et al. (2008) method, and the broken line with dots is after correction by the Leymarie et al. (2010) method.
122
<*< e
10
8
6
CDOM fluorescence /\ Chlorophyll fluorescence1
Raman Scattering
400 450 500 550 600 650 700 Wavelength (nm)
Figure 4.8 Results of the sensitivity test on forward modeling. The proportional changes in remote sensing reflectance (Rrs(X)) are shown. The solid line includes the addition of chlorophyll fluorescence to the model, the broken line includes CDOM fluorescence, and the broken line with dots includes Raman scattering. All modeling was conducted with AC-S data corrected by the Leymarie et al. (2010) method.
123
a. Error on a tJA)
-0.5
-0.6
2.0
Gallegos et al. Leymarie et al.
b. Error on br a/A)
400 450 500 550 600 650 700 Wavelength (nm)
Figure 4.9 Differences in inherent optical properties measured by the AC-S as a result of the different correction methods. Difference between the proportional method and other methods in a. the absorption coefficient excluding those of pure-water (at.w(X), m"1), and b. the ratio of total scattering to absorption (btlat(X), dimensionless). The solid line is after null correction, the broken line is after Gallegos et al. (2008) correction, and the broken line with dots is after Leymarie et al. (2010) correction.
124
0.0
0.5
0.0
Particulate matter
Null point value (0.135 m )
b. CDOM
a C D O . \ / ^ - I , Null point value (0.264 m" )
650 700 750 800 Wavelength (nm)
850
Figure 4.10 Raw absorption spectra in the near-infrared radiation (NIR) region. Solid lines are raw absorption coefficients of a. colored dissolved organic matter (CDOM; OCDOM(X), m"1) and b. total particulate matter (ap(X), m"1). Broken lines are the calculated null value to be subtracted in null point correction.
125
Chapitre 5
Conclusion générale Dans la présente étude, les processus optiques ont été caractérisés dans un ensemble
contrasté d'eaux continentales optiquement complexes en utilisant des instruments de
pointe et des outils de modélisation développés lors d'études océaniques. L'atténuation de
la lumière dans l'eau, le rayonnement résiduel sortant de l'eau et leurs facteurs de contrôle
ont été analysés quantitativement dans des écosystèmes lacustres diversifiés, incluant des
mares thermokarstiques subarctiques, un lac tempéré mésotrophe et un lac hypereutrophe
urbain. Ces analyses spectrales fournissent de nouvelles perspectives pour étudier le régime
de lumière des eaux de type 2 et élargissent ainsi les connaissances en optique
hydrologique, surtout développées pour les eaux océaniques de type 1.
Les mares thermokarstiques turbides, abondantes dans les zones de pergélisol
discontinu au Nunavik, ont montré une grande diversité de conditions optiques et des
différences marquées dans la couleur de l'eau. Cette variation de la couleur a été décrite
avec succès pour différentes formes des mesures de réflectance spectrales révélées grâce à
l'exploitation d'une sonde spectroradiométrique de terrain. Ces mesures spectrales ont été
converties en valeurs définies selon un système de coordonnées de couleur, fournissant une
description quantitative de la couleur de l'eau. La relation entre les caractéristiques optiques
et les conditions limnologiques a révélé que la gamme de couleurs était due aux variations
de la concentration de deux composantes de la colonne d'eau: le carbone organique dissous
et les particules non-algales en suspension. L'analyse des IOP et des spectres d'absorption
et de dispersion des composantes optiquement actives a confirmé que ces deux
composantes étaient les principaux acteurs expliquant les conditions optiques des mares
thermokarstiques. L'analyse des IOP a également montré que l'influence de ces
composantes optiques était similaire dans les mares, c'est-à-dire que la diversité optique
dépendait des différences dans leur concentration et non pas dans leur composition. Ce
dernier résultat implique que l'estimation quantitative de ces deux composantes majeures
serait réalisable par des observations optiques par télédétection des mares thermokarstiques.
126
L'analyse d'une image satellitaire multispectrale à haute résolution spatiale a montré
que ces images peuvent être utilisées pour l'observation des mares thermokarstiques en
région subarctique, en particulier pour estimer à distance le carbone organique dissous et
les particules en suspension en exploitant la modélisation multivariée. Compte tenu de
l'importance de ces deux constituants pour les cycles biogéochimiques en milieu aquatique
(notamment la structure thermique de la colonne d'eau, la concentration en oxygène et la
production microbienne), la surveillance par télédétection a le potentiel de fournir de
précieuses observations synoptiques des mares thermokarstiques dans les vastes régions
subarctiques, permettant ainsi l'extension des mesures locales des flux de gaz à effet de
serre aux échelles régionale et circumpolaire.
Les recherches futures sur les mares thermokarstiques subarctiques devraient
inclure une évaluation plus poussée des propriétés optiques de la matière organique
dissoute (y compris les caractéristiques de fluorescence) et des particules non-algales en
suspension. Il serait également important de valider le modèle de télédétection sur un
ensemble de données plus large, couvrant des zones géographiques diverses, où les mares
possèdent différentes caractéristiques limnologiques et optiques, notamment les mares
thermokarstiques associées à la dégradation des paises organiques. Enfin, des recherches
supplémentaires sont nécessaires pour évaluer l'absorption spectrale et les caractéristiques
de diffusion des particules fines, ainsi que leur quantification en masse sèche et leur
composition chimique. Ces particules jouent un rôle notoire dans les systèmes
thermokarstiques (et peut-être ailleurs), et sont pourtant encore mal comprises, en
particulier dû aux méthodes employées pour filtrer l'eau.
Dans le lac Saint-Charles, réservoir d'eau potable en région tempérée, les processus
optiques sous l'eau et leurs facteurs de contrôle ont été décrits en détail dans la présente
étude par une série d'analyses optiques. Selon les mesures spectrophotométriques et leurs
relations avec les variables limnologiques, les caractéristiques d'absorption des
composantes optiquement actives (le CDOM et différents types de particules en
suspension) se sont avérées stables pendant toute la période d'échantillonnage (de juin à
septembre). Cela indique que les changements dans les propriétés optiques du lac durant
127
l'été proviendraient de changements dans la concentration plutôt que dans la composition
de ces composantes. Cette caractéristique sera d'une grande aide pour la modélisation
optique et la surveillance de la qualité de l'eau de cet écosystème lacustre dans le futur. Les
analyses ont également montré une absorption significative par les particules fines,
phénomène ayant déjà été observé à ce jour dans des systèmes fortement turbides, mais
jamais pris en compte dans d'autres types d'eau naturelle.
Le calcul du budget de photons décrit clairement le gradient vertical du PAR et des
facteurs de contrôle sur l'atténuation spectrale au lac Saint-Charles, en combinant la
diminution de l'éclairement spectral et les IOP des composantes optiquement actives. Les
résultats ont montré que le CDOM domine l'atténuation du PAR dans le lac, alors que les
autres composantes n'ont joué qu'un rôle mineur. Cela suggère que la transparence de l'eau,
traditionnellement basée sur des estimations telles que la profondeur du disque de Secchi
ou le coefficient d'atténuation diffuse du PAR, serait un guide trompeur pour suivre les
changements de l'état trophique de ce lac et des écosystèmes aquatiques semblables des
régions tempérée ou boréale, où les apports élevés en CDOM terrigènes sont importants.
Les résultats de la présente étude ont également montré comment les coefficients
d'atténuation spécifiques des composantes optiquement actives et leur contribution relative
à l'atténuation changent dans le temps (saisons) et avec la profondeur dans la colonne d'eau.
Ces gradients affectent la moyenne des coefficients d'absorption in situ des composantes
optiques de la zone euphorique. Cette caractéristique est bien illustrée par la grande
variabilité des mesures du coefficient d'absorption in situ de l'eau pure. En effet, ce
coefficient a des valeurs moyennes dans la zone euphorique dix fois plus élevées que celles
qui ont été mesurées dans les systèmes océaniques au large des côtes. Ces résultats
soulignent l'importance de considérer les caractéristiques spectrales de la lumière ambiante
aussi bien que des composantes optiquement actives pour bien définir le processus
d'atténuation du PAR sous l'eau. L'utilisation de valeurs constantes pour les coefficients
d'atténuation du PAR par l'eau et les autres composantes doit être soigneusement évaluée
en examinant les caractéristiques spectrales de la masse d'eau ciblée. En particulier,
l'application des résultats obtenus dans l'océan à des zones côtières ou aux eaux
128
continentales devrait être évitée étant donné que le régime de lumière spectral de ces eaux
est bien différent de celui des systèmes océaniques.
La méthode du budget de photons a été introduite à la fin des années 80 dans l'océan
au large des zones côtières, mais peu appliquée par la suite. Tel que démontré dans cette
étude, cette méthode est un outil utile pour comprendre l'atténuation du PAR, applicable
aux eaux continentales étant donné la disponibilité accrue des spectroradiomètres. Ces
résultats fourniront des informations utiles pour la modélisation biologique et
biogéochimique, et ainsi contribueront à une meilleure gestion des ressources en eau.
La fermeture optique de la réflectance spectrale détectée à distance a été réalisée
pour le lac Saint-Augustin, un lac hypereutorophe urbain. Les résultats ont montré
notamment l'importance d'appliquer une correction appropriée des erreurs d'estimation des
IOP. La mesure de la réflectance spectrale détectée à distance (Rrs(X)) par un système de
profilage dans l'eau donnait un spectre avec un pic autour de 580 nm, ce qui est typique
pour les plans d'eau présentant un mélange de CDOM, de particules non-algales et de
particules d'algues à des densités modérées. Le transfert radiatif du modèle HydroLight a
permis d'estimer Rrs(X) avec succès à partir des IOP mesurées et des données sur les
conditions ambiantes. La comparaison des Rrs(X) modélisés en utilisant différentes entrées
d'IOP a montré la sensibilité de la modélisation pour la correction de diffusion sur les
mesures d'IOP obtenues par spectrophotométrie AC-S. Ces différences dans les Rrs(X)
modélisés provenaient de petites différences dans les coefficients d'absorption corrigés
obtenus par les différentes méthodes. Ces petites différences dans les coefficients
d'absorption ont par la suite produit des différences relativement importantes dans le
rapport entre la diffusion et l'absorption, un facteur important dans le calcul de Rrs(X). Le
choix de la meilleure correction d'absorption pour le lac Saint-Augustin, et pour des eaux
similaires fortement diffusives, s'appui sur l'hypothèse d'une absorption significative dans
la région proche-infrarouge, alors qu'elle est normalement considérée comme nulle dans les
eaux océaniques transparentes. Une évaluation plus poussée de l'absorption dans cette
gamme spectrale pour les plans d'eau optiquement complexes est requise pour améliorer le
protocole de mesure des IOP.
129
En conclusion, les eaux intérieures montrent des conditions optiques très
diversifiées, reflétant de grandes variations dans leurs caractéristiques biologiques,
géochimiques et physiques. Cette complexité dans les processus optiques offre une gamme
de conditions expérimentales fort intéressantes permettant d'étendre nos connaissances en
optique hydrologique. Des progrès importants dans ce domaine peuvent être réalisés en
appliquant aux eaux continentales les connaissances et les techniques développées pour les
eaux océaniques, où la caractérisation des procédés optiques est plus avancée. Toutefois, tel
que démontré dans la présente étude, les lacs et les mares sont des milieux hautement
diffusifs et présentant des propriétés spectrales très différentes de celles de l'océan. C'est
pourquoi les coefficients dérivés d'un milieu ne sont pas directement applicables à un autre
milieu. Ces progrès conduiront à une amélioration dans notre compréhension des
écosystèmes aquatiques, et fourniront des outils indispensables de modélisation et de
surveillance des précieuses ressources en eaux douces de la planète.
130
131
Références bibliographiques Arlen-Pouliot, Y., and N. Bhiry. 2005. Palaeoecology of a palsa and a filled thermokarst
pond in a permafrost peatland, subarctic Quebec, Canada. Holocene 15: 408-419.
Arrigo, K. R., G. L. Van Dijken, and S. Bushinsky. 2008. Primary production in the Southern Ocean, 1997-2006. J. Geophys. Res.-Oceans 113: C08004, doi: 08010.01029/02007 JC004551.
Arst, H. 2003. Optical Properties and Remote Sensing of Multicomponental Water Bodies. Praxis Publishing. Chichester, UK. 238pp.
Babin, M., A. Morel, V. Fournier-Sicre, F. Fell, and D. Stramski. 2003a. Light scattering properties of marine particles in coastal and open ocean waters as related to the particle mass concentration. Limnol. Oceanogr. 48: 843-859.
Babin, M., and D. Stramski. 2002. Light absorption by aquatic particles in the near-infrared spectral region. Limnol. Oceanogr. 47: 911-915.
Babin, M., D. Stramski, G. M. Ferrari, H. Claustre, A. Bricaud, G. Obolensky, and N. Hoepffher. 2003b. Variations in the light absorption coefficients of phytoplankton, nonalgal particles, and dissolved organic matter in coastal waters around Europe. J. Geophys. Res. 108: 3211, doi: 3210.1029/2001JC000882.
Belzile, C, W. F. Vincent, C. Howard-Williams, I. Hawes, M. James, M. Kumagai, and C. Roesler. 2004. Relationships between spectral optical properties and optically active substances in a clear oligotrophic lake. Water Resour. Res. 40: doi: 1029/2004WR003090.
Berwald, J., D. Stramski, C. D. Mobley, and D. A. Kiefer. 1995. Influences of absorption and scattering on vertical changes in the average cosine of the underwater light field. Limnol. Oceanogr. 40: 1347-1357.
Binding, C. E., J. H. Jerome, R. P. Bukata, and W. G. Booty. 2008. Spectral absorption properties of dissolved and particulate matter in Lake Erie. Remote Sens. Environ. 112: 1702-1711.
Boss, E., W. S. Pegau, M. Lee, M. Twardowski, E. Shybanov, G. Korotaev, and F. Baratange. 2004. Particulate backscattering ratio at LEO 15 and its use to study particle composition and distribution. J. Geophys. Res.-Oceans 109: C01014, doi: 01010.01029/02002 JC001514.
Boss, E., L. Taylor, S. Gilbert, K. Gundersen, N. Hawley, C. Janzen, T. Johengen, H. Purcell, C. Robertson, D. W. H. Schar, G. J. Smith, and M. N. Tamburri. 2009. Comparison of inherent optical properties as a surrogate for particulate matter concentration in coastal waters. Limnol. Oceanogr. Meth. 7: 803-810.
132
Boss, E. S., R. Collier, G. Larson, K. Fennel, and W. S. Pegau. 2007. Measurements of spectral optical properties and their relation to biogeochemical variables and processes in Crater Lake, Crater Lake National Park, OR. Hydrobiologia 574: 149-159.
Bouchard-Valentine, M., W. F. Vincent, and R. Pienitz. 2004. Stratification à court terme et dynamique de mélange dans un lac eutrophe: implications pour les études de suivi de la qualitié de l'eau. Eau et changements climatiques: comprendre pour mieux s'adapter. Proceedings of the annual conference of the Canadian Water Research Association, Montreal, Quebec. CWRA.
Bouchard, F., P. Francus, R. Pienitz, and I. Laurion. 2011. Sedimentology and geochemistry of thermokarst ponds in discontinuous permafrost, subarctic Quebec, Canada. J. Geophys. Res.-Biogeosci. 116: G00M04.
Bowers, D. G., G. E. L. Harker, and B. Stephan. 1996. Absorption spectra of inorganic particles in the Irish Sea and their relevance to remote sensing of chlorophyll. Int. J. Remote Sens. 17: 2449-2460.
Breton, J., C. Vallieres, and I. Laurion. 2009. Limnological properties of permafrost thaw ponds in northeastern Canada. Can. J. Fish. Aquat. Sci. 66: 1635-1648.
Bricaud, A., M. Babin, A. Morel, and H. Claustre. 1995. Variability in the chlorophyll-specific absorption-coefficients of natural phytoplankton - Analysis and parameterization. J. Geophys. Res.-Oceans 100: 13321-13332.
Bricaud, A., A. Morel, and L. Prieur. 1981. Absorption by dissolved organic-matter of the sea (yellow substance) in the UV and visible domains. Limnol. Oceanogr. 26: 43-53.
Bricaud, A., and D. Stramski. 1990. Spectral absorption-coefficients of living phytoplankton and nonalgal biogenous matter - a comparison between the Peru upwelling area and the Sargasso Sea. Limnol. Oceanogr. 35: 562-582.
Buiteveld, H., J. H. M. Hakvoort, and M. Donze. 1994. The optical properties of pure water. SPIE Proceedings on Ocean Optics XII2258: 174-183.
Bukata, R. P., J. H. Jerome, K. Y. Kondratyev, and D. V. Pozdnyakov. 1995. Optical Properties and Remote Sensing of Inland and Coastal Waters. CRC Press. Boca Raton, FL. 362pp.
Bukaveckas, P. A., and M. Robbins-Forbes. 2000. Role of dissolved organic carbon in the attenuation of photosynthetically active and ultraviolet radiation in Adirondack lakes. Freshw. Biol. 43: 339-354.
Calmels, F., M. Allard, and G. Delisle. 2008a. Development and decay of a lithalsa in Northern Quebec: A geomorphological history. Geomorphology 97: 287-299.
133
Calmels, F., G. Delisle, and M. Allard. 2008b. Internai structure and the thermal and hydrological regime of a typical lithalsa: significance for permafrost growth and decay. Can. J. Earth Sci. 45: 31-43.
Caplanne, S., and I. Laurion. 2008. Effect of chromophoric dissolved organic matter on epilimnetic stratification in lakes. Aquat. Sci. 70: 123-133.
CIE. 1931. CIE 1931 Standard Colorimetric Observer.
CIE. 1986. CIE Standard Illuminant D65 relative spectral power distribution (CIE 15.2-1986 Table 1.1, Part 2).
Cohen, W. B., T. K. Maiersperger, S. T. Gower, and D. P. Turner. 2003. An improved strategy for regression of biophysical variables and Landsat ETM+ data. Remote Sens. Environ. 84: 561-571.
Cole, J. J., S. R Carpenter, J. F. Kitchell, and M. L. Pace. 2002. Pathways of organic carbon utilization in small lakes: Results from a whole-lake C-l3 addition and coupled model. Limnol. Oceanogr. 47: 1664-1675.
Davies-Colley, R. J., W. N. Vant, and D. G. Smith. 2003. Colour and Clarity of Natural Waters : Science and Management of Optical Water Quality. Blackburn Press. Caldwell, NJ. 310pp.
Dekker, A. G., R. J. Vos, and S. W. M. Peters. 2002. Analytical algorithms for lake water TSM estimation for retrospective analyses of TM and SPOT sensor data. Int. J. Remote Sens. 23: 15-35.
Doxaran, D., K. Ruddick, D. Mckee, B. Gentili, D. Tailliez, M. Chami, and M. Babin. 2009. Spectral variations of light scattering by marine particles in coastal waters, from visible to near infrared. Limnol. Oceanogr. 54: 1257-1271.
Duarte, C. M., S. Agusti, M. P. Satta, and D. Vaque. 1998. Partitioning particulate light absorption: A budget for a Mediterranean bay. Limnol. Oceanogr. 43: 236-244.
Duntley, S. Q. 1963. Light in the Sea. J. Opt. Soc. Am. 53: 214-233.
Falkowski, P., and J. Raven. 2007. Aquatic Photosynthesis, 2nd ed. Princeton University Press. Princeton, NJ. 484pp.
Ferrari, G. M., and S. Tassan. 1996. Use of the 0.22 pm Millipore membrane for light-transmission measurements of aquatic particles. J. Plankton Res. 18: 1261-1267.
Ferrari, G. M., and S. Tassan. 1999. A method using chemical oxidation to remove light absorption by phytoplankton pigments. J. Phycol. 35: 1090-1098.
Flink, P., T. Lindell, and C. Ostlund. 2001. Statistical analysis of hyperspectral data from two Swedish lakes. Sci. Total Environ. 268: 155-169.
134
Frenette, J. J., and W. F. Vincent. 2003. Bio-optical variability in the littoral zone: local heterogeneity and implications for water quality monitoring, p. 41-59. In M. Kumagai and W. F. Vincent [eds.], Freshwater Management - Global versus Local Perspectives. Springer-Verlag. Tokyo, Japan. 233pp.
Fujii, M., E. Boss, and F. Chai. 2007. The value of adding optics to ecosystem models: a case study. Biogeosciences 4: 817-835.
Gallegos, C. L. 2005. Optical water quality of a blackwater river estuary: the Lower St. Johns River, Florida, USA. Estuar. Coast. Shelf Sci. 63: 57-72.
Gallegos, C. L., R. J. Davies-Colley, and M. Gall. 2008. Optical closure in lakes with contrasting extremes of reflectance. Limnol. Oceanogr. 53: 2021-2034.
Gallegos, C. L., and P. J. Neale. 2002. Partitioning spectral absorption in case 2 waters: discrimination of dissolved and particulate components. Appl. Optics 41: 4220-4233.
Giles-Guzman, A. D., and S. Alvarez-Borrego. 2000. Vertical attenuation coefficient of photosynthetically active radiation as a function of chlorophyll concentration and depth in case 1 waters. Appl. Optics 39: 1351-1358.
Gordon, H. R., and K. Y. Ding. 1992. Self-shading of in-water optical-instruments. Limnol. Oceanogr. 37: 491-500.
Gordon, H. R., and W. R Mccluney. 1975. Estimation of the depth of sunlight penetration in the sea for remote sensing. Appl. Optics 14: 413-416.
Heiskanen, J. 2006. Estimating aboveground tree biomass and leaf area index in a mountain birch forest using ASTER satellite data. Int. J. Remote Sens. 27: 1135-1158.
IOCCG. 1998. Minimum Requirements for an Operational, Ocean-colour Sensor for the Open Ocean. International Ocean-Colour Coordinating Group. Dartmouth, NS. 50pp.
IOCCG. 2000. Remote Sensing of Ocean Colour in Coastal, and Optically-complex Waters. International Ocean-Colour Coordinating Group. Dartmouth, NS. 140pp.
IOCCG. 2008. Why Ocean Colour? The Societal Benefits of Ocean-Colour Technology. International Ocean-Colour Coordinating Group. Dartmouth, NS. 141pp.
Johnson, R. A., and D. W. Wichern. 2007. Applied Multivariate Statistical Analysis, 6th ed. Prentice Hall. Upper Saddle River, NJ. 800pp.
Jones, J. R, D. V. Obrecht, B. D. Perkins, M. F. Knowlton, A. P. Thorpe, S. Watanabe, and R. R. Bacon. 2008. Nutrients, seston, and transparency of Missouri reservoirs and oxbow lakes: An analysis of regional limnology. Lake Reserv. Manage. 24: 155-180.
135
Kalff, J. 2002. Limnology: Inland Water Ecosystems. Prentice Hall. Upper Saddle River, NJ, USA. 592pp.
Karlsson, J., P. Bystrom, J. Ask, P. Ask, L. Persson, and M. Jansson. 2009. Light limitation of nutrient-poor lake ecosystems. Nature 460: 506-509.
Kirk, J. T. O. 1981. Monte Carlo study of the nature of the underwater light field in, and the relationships between optical properties of, turbid yellow waters. Aust. J. Mar. Fresh. Res. 32: 517-532.
Kirk, J. T. O. 1992. Monte-Carlo modeling of the performance of a reflective tube absorption meter. Appl. Optics 31: 6463-6468.
Kirk, J. T. O. 1994. Light and Photosynthesis in Aquatic Ecosystems, 2nd ed. Cambridge University Press. Cambridge, U.K. 509pp.
Kirk, J. T. O. 1997. Point-source integrating-cavity absorption meter: Theoretical principles and numerical modeling. Appl. Optics 36: 6123-6128.
Kishino, M., M. Takahashi, N. Okami, and S. Ichimura. 1985. Estimation of the spectral absorption-coefficients of phytoplankton in the sea. Bull. Mar. Sci. 37: 634-642.
Knowlton, M. F., and J. R. Jones. 2000. Non-algal seston, light, nutrients and chlorophyll in Missouri reservoirs. Lake Reserv. Manage. 16: 322-332.
Kutser, T., H. Arst, T. Miller, L. Kaarmann, and A. Milius. 1995. Telespectrometrical estimation of water transparency, chlorophyll-a and total phosphorus concentration of Lake Peipsi Int. J. Remote Sens. 16: 3069-3085.
Laurion, I., F. Blouin, and S. Roy. 2003. The quantitative filter technique for measuring phytoplankton absorption: Interference by MAAs in the UV waveband. Limnol. Oceanogr. Meth. 1: 1-9.
Laurion, I., W. F. Vincent, S. Macintyre, L. Rétamai, C. Dupont, P. Francus, and R. Pienitz. 2010. Variability in greenhouse gas emissions from permafrost thaw ponds. Limnol. Oceanogr. 55: 115-133.
Lewis, M. R. 2008. Measurement of apparent optical properties for diagnosis of harmful algal blooms, p. 207-236. In M. Babin, C. S. Roesler and J. J. Cullen [eds.], Realtime Coastal Observing Systems for Marine Ecosystem Dynamics and Harmful Algal Blooms: Theory, Instrumentation and Modelling. UNESCO Publishing. Venice, Italy. 830pp.
Lewis, W. M. 1983. Temperature, heat, and mixing in Lake Valencia, Venezuela. Limnol. Oceanogr. 28: 273-286.
Leymarie, E., D. Doxaran, and M. Babin. 2010. Uncertainties associated to measurements of inherent optical properties in natural waters. Appl. Optics 49: 5415-5436.
136
Li, R. R., Y. J. Kaufman, B. C. Gao, and C. O. Davis. 2003. Remote sensing of suspended sediments and shallow coastal waters. IEEE Trans. Geosci. Remote Sensing 41: 559-566.
Lorenzen, C. J. 1972. Extinction of light in the ocean by phytoplankton. J. Cons. Int. Explor. Mer 32: 262-267.
Markager, S., and W. F. Vincent. 2000. Spectral light attenuation and the absorption of UV and blue light in natural waters. Limnol. Oceanogr. 45: 642-650.
Markager, S., and W. F. Vincent. 2001. Light absorption by phytoplankton: development of a matching parameter for algal photosynthesis under different spectral regimes. J. Plankton Res. 23: 1373-1384.
Matsuoka, A., Y. Huot, K. Shimada, S. I. Saitoh, and M. Babin. 2007. Bio-optical characteristics of the western Arctic Ocean: implications for ocean color algorithms. Can. J. Remote Sens. 33: 503-518.
Miller, R. L., C. E. Del Castillo, and B. A. Mckee. 2005. Remote Sensing of Coastal Aquatic Environments: Technologies, Techniques and Applications. Springer. Norwell, MA. 345pp.
Mitchell, B. G., A. Bricaud, K. Carder, J. Cleveland, G. M. Ferrari, R. Gould, M. Kahru, M. Kishino, H. Maske, T. Moisan, L. Moore, N. Nelson, D. Phinney, R. Reynolds, H. Sosik, D. Stramski, S. Tassan, C. Trees, A. Weidemann, J. Wieland, and A. Vodacek. 2000. Determination of spectral absorption coefficients of particles, dissolved material and phytoplankton for discrete water samples, p. 125-153. In G. S. Fargion and J. L. Mueller [eds.], Ocean Optics Protocols for Satellite Ocean Color Sensor Validation, Revision 2. National Aeronautics and Space Administration.
Mitchell, B. G., M. Kahru, J. Wieland, and M. Stramska. 2002. Determination of spectral absorption coefficients of particles, dissolved material and phytoplankton for discrete water samples, p. 39-64. In J. L. Mueller and G. S. Fargion [eds.], Ocean Optics Protocols for Satellite Ocean Color Sensor Validation, Revision 4. National Aeronautics and Space Administration.
Mobley, C. D. 1994. Light and Water: Radiative Transfer in Natural Waters. Academic Press. San Diego, CA. 592pp.
Mobley, C. D. 1999. Estimation of the remote-sensing reflectance from above-surface measurements. Appl. Optics 38: 7442-7455.
Morel, A. 1988. Optical modeling of the upper ocean in relation to its biogenous matter content (Case-I waters). J. Geophys. Res. 93: 10749-10768.
Morel, A. 1991. Light and marine photosynthesis - a spectral model with geochemical and climatological implications. Prog. Oceanogr. 26: 263-306.
137
Morel, A., and D. Antoine. 1994. Heating rate within the upper ocean in relation to its biooptical state. J. Phys. Oceanogr. 24: 1652-1665.
Morris, D. P., H. Zagarese, C. E. Williamson, E. G. Balseiro, B. R Hargreaves, B. Modenutti, R. Moeller, and C. Queimalinos. 1995. The attenuation of solar UV radiation in lakes and the role of dissolved organic carbon. Limnol. Oceanogr. 40: 1381-1391.
Mueller, J. L. 2002. In-Water radiometric profile measurements and data analysis protocols, p. 7-20. In J. L. Mueller and G. S. Fargion [eds.], Ocean Optics Protocols for Satellite Ocean Color Sensor Validation, Revision 4. National Aeronautics and Space Administration.
Mueller, J. L., C. Davis, R. Arnone, R. Frouin, K. Carder, Z. P. Lee, R. G. Steward, S. Hooker, C. D. Mobley, and S. Mclean. 2002. Above-water radiance and remote sensing reflectance measurement and analysis protocols, p. 21-32. In J. L. Mueller and G. S. Fargion [eds.], Ocean Optics Protocols for Satellite Ocean Color Sensor Validation, Revision 4. National Aeronautics and Space Administration.
Nelson, N. B., and B. B. Prezelin. 1993. Calibration of an integrating sphere for determining the absorption coefficient of scattering suspensions. Appl. Optics 32: 6710-6717.
Nusch, E. A. 1980. Comparison of different methods for chlorophyll and phaeopigment determination. Arch. Hydrobiol. 14: 14-36.
O'donnell, D. M., S. W. Effler, C. M. Strait, and G. A. Leshkevich. 2010. Optical characterizations and pursuit of optical closure for the western basin of Lake Erie through in situ measurements. J. Gt. Lakes Res. 36: 736-746.
Obrador, B., and J. L. Pretus. 2008. Light regime and components of turbidity in a Mediterranean coastal lagoon. Estuar. Coast. Shelf Sci. 77: 123-133.
Oyama, Y., B. Matsushita, T. Fukushima, K. Matsushige, and A. Imai. 2009. Application of spectral decomposition algorithm for mapping water quality in a turbid lake (Lake Kasumigaura, Japan) from Landsat TM data. ISPRS-J. Photogramm. Remote Sens. 64: 73-85.
Oyama, Y., and A. Shibahara. 2009. Simulation of water colors in a shallow acidified lake, Lake Onneto, Japan, using colorimetric analysis and bio-optical modeling. Limnology 10: 47-56.
Patterson, J. C , and P. F. Hamblin. 1988. Thermal simulation of a lake with winter ice cover. Limnol. Oceanogr. 33: 323-338.
Payette, S., A. Delwaide, M. Caccianiga, and M. Beauchemin. 2004. Accelerated thawing of subarctic peatland permafrost over the last 50 years. Geophys. Res. Lett. 31: L18208, doi: 18210.11029/12004GL020358.
138
Pegau, S., J. R. V. Zaneveld, B. G. Mitchell, J. L. Mueller, M. Kahru, J. Wieland, and M. Stramska. 2003. Inherent optical properties: Instruments, characterizations, field measurements and data analysis protocols. In J. L. Mueller, G. S. Fargion and R. McClain [eds.], Ocean optics protocols for satellite ocean color sensor validation; Revision 4. National Aeronautics and Space Administration.
Perkins, M., S. W. Effler, C. Strait, and L. Zhang. 2009. Light absorbing components in the Finger Lakes of New York. Fundam. Appl. Limnol. 173: 305-320.
Petzold, T. J. 1972. Volume scattering functions for selected ocean waters. Scripps Inst. Oceanogr.: Report SIO 72-78.
Pienitz, R, P. T. Doran, and S. F. Lamoureux. 2008. Origin and geomorphology of lakes in the polar regions, p. 25-41. In W. F. Vincent and J. Laybourn-Parry [eds.], Polar Lakes and Rivers: Limnology of Arctic and Antarctic Aquatic Ecosystems. Oxford University Press. New York, NY. 327pp.
Pienitz, R, K. Roberge, and W. F. Vincent. 2006. Three hundred years of human-induced change in an urban lake: paleolimnological analysis of Lac Saint-Augustin, Quebec City, Canada. Can. J. Bot. 84: 303-320.
Pienitz, R, and W. F. Vincent. 2000. Effect of climate change relative to ozone depletion on UV exposure in subarctic lakes. Nature 404: 484-487.
Pienitz, R., and W. F. Vincent. 2003. Generic approaches towards water quality monitoring based on Paleolimnology, p. 61-82. In M. Kumagai and W. F. Vincent [eds.], Freshwater Management - Global versus Local Perspectives. Springer-Verlag. Tokyo, Japan. 233pp.
Ping, C. L., G. J. Michaelson, M. T. Jorgenson, J. M. Kimble, H. Epstein, V. E. Romanovsky, and D. A. Walker. 2008. High stocks of soil organic carbon in the North American Arctic region. Nature Geoscience 1: 615-619.
Pope, R. M., and E. S. Fry. 1997. Absorption spectrum (380-700 nm) of pure water .2. Integrating cavity measurements. Appl. Optics 36: 8710-8723.
Prairie, Y. T., D. F. Bird, and J. J. Cole. 2002. The summer metabolic balance in the epilimnion of southeastern Quebec lakes. Limnol. Oceanogr. 47: 316-321.
Preisendorfer, R. W. 1961. Application of radiative transfer theory to light measurements in the sea. 10: 11-30.
Preisendorfer, R. W. 1976. Hydrologie Optics. NOAA Pacific Mar. Environ. Lab. Seattle, WA, USA.
R Development Core Team. 2009. R: a language and environment for statistical computing. R foundation for statistical computing.
139
Roesler, C. S. 1998. Theoretical and experimental approaches to improve the accuracy of particulate absorption coefficients derived from the quantitative filter technique. Limnol. Oceanogr. 43: 1649-1660.
Roesler, C. S., and E. Boss. 2008. In situ measurement of the inherent optical properties and potential for harmful algal bloom detection and coastal ecosystem observations, p. 153-206. In M. Babin, C. S. Roesler and J. J. Cullen [eds.], Real-time Coastal Observing Systems for Marine Ecosystem Dynamics and Harmful Algal Blooms: Theory, Instrumentation and Modelling. UNESCO Publishing. Venice, Italy. 830pp.
Rottgers, R., C. Hase, and R. Doerffer. 2007. Determination of the particulate absorption of microalgae using a point-source integrating-cavity absorption meter: verification with a photometric technique, improvements for pigment bleaching, and correction for chlorophyll fluorescence. Limnol. Oceanogr. Meth. 5: 1-12.
Rottgers, R, W. Schofeld, P. R. Kipp, and R. Doerffer. 2005. Practical test of a point-source integrating cavity absorption meter: the performance of different collector assemblies. Appl. Optics 44: 5549-5560.
Schanda, J. 2007. CIE Colorimetry, p. 25-78. In J. Schanda [ed.], Colorimetry : Understanding the CIE System. Wiley-Interscience. Hoboken, NJ. 467pp.
Schiebener, P., J. Straub, J. Sengers, and J. S. Gallagher. 1990. Refractive-index of water and steam as function of wavelength, temperature and density. J. Phys. Chem. Ref. Data 19: 677-717.
Scully, N. M., and D. R. S. Lean. 1994. The attenuation of ultraviolet radiation in temperate lakes. Ergebn. Limnol. 43: 135-144.
Simis, S. G. H., S. W. M. Peters, and H. J. Gons. 2005. Remote sensing of the cyanobacterial pigment phycocyanin in turbid inland water. Limnol. Oceanogr. 50: 237-245.
Smith, R. C , and K. S. Baker. 1978. The bio-optical state of ocean waters and remote sensing. Limnol. Oceanogr. 23: 247-259.
Smith, R. C , J. Marra, M. J. Perry, K S. Baker, E. Swift, E. Buskey, and D. A. Kiefer. 1989. Estimation of a photon budget for the upper ocean in the Sargasso sea. Limnol. Oceanogr. 34: 1673-1693.
Snyder, W. A., R. A. Arnone, C. O. Davis, W. Goode, R. W. Gould, S. Ladner, G. Lamela, W. J. Rhea, R. Stavn, M. Sydor, and A. Weidemann. 2008. Optical scattering and backscattering by organic and inorganic particulates in US coastal waters. Appl. Optics 47: 666-677.
Sobek, S., L. J. Tranvik, and J. J. Cole. 2005. Temperature independence of carbon dioxide supersaturation in global lakes. Global Biogeochem. Cy. 19.
140
Stainton, M. P., M. J. Capel, and F. A. J. Armstrong. 1977. The Chemical Analysis of Freshwater, Canada Fisheries and Marine Service Miscellaneous Special Publications, No. 25, 2nd ed. Freshwater Institute. Winnipeg, MB. 180pp.
Stedmon, C. A., S. Markager, and H. Kaas. 2000. Optical properties and signatures of chromophoric dissolved organic matter (CDOM) in Danish coastal waters. Estuar. Coast. Shelf Sci. 51: 267-278.
Storm, S. L., A. Springsteen, and T. M. Ricker. 1998. The use of center mount sample holders in reflectance spectroscopy. In Labsphere Inc. [ed.], Application note No.2.
Stramski, D., M. Babin, and S. B. Wozniak. 2007. Variations in the optical properties of terrigenous mineral-rich particulate matter suspended in seawater. Limnol. Oceanogr. 52: 2418-2433.
Stramski, D., and S. B. Wozniak. 2005. On the role of colloidal particles in light scattering in the ocean. Limnol. Oceanogr. 50: 1581-1591.
Sullivan, J. M., M. S. Twardowski, J. R. V. Zaneveld, C. M. Moore, A. H. Barnard, P. L. Donaghay, and B. Rhoades. 2006. Hyperspectral temperature and salt dependencies of absorption by water and heavy water in the 400-750 nm spectral range. Appl. Optics 45: 5294-5309.
Sun, D. Y., Y. M. Li, Q. Wang, C. F. Le, C. C. Huang, and L. Z. Wang. 2009. Parameterization of water component absorption in an inland eutrophic lake and its seasonal variability: a case study in Lake Taihu. Int. J. Remote Sens. 30: 3549-3571.
Svab, E., A. N. Tyler, T. Preston, M. Presing, and K. V. Balogh. 2005. Characterizing the spectral reflectance of algae in lake waters with high suspended sediment concentrations. Int. J. Remote Sens. 26: 919-928.
Tarnocai, C. 2009. The Impact of Climate Change on Canadian Peatlands. Can. Water Resour. J. 34: 453-466.
Tassan, S., and K. Allali. 2002. Proposal for the simultaneous measurement of light absorption and backscattering by aquatic particulates. J. Plankton Res. 24: 471-479.
Tassan, S., and G. M. Ferrari. 1995. An alternative approach to absorption measurements of aquatic particles retained on filters. Limnol. Oceanogr. 40: 1358-1368.
Tassan, S., and G. M. Ferrari. 2002. A sensitivity analysis of the 'Transmittance-Reflectance' method for measuring light absorption by aquatic particles. J. Plankton Res. 24: 757-774.
Tassan, S., and G. M. Ferrari. 2003. Variability of light absorption by aquatic particles in the near-infrared spectral region. Appl. Optics 42: 4802-4810.
141
Tremblay, R., S. Légaré, R. Pienitz, W. F. Vincent, and R. Hall. 2001. Étude paléolimnologique de l'histoire trophique du lac Saint-Charles, réservoir d'eau potable de la communauté urbaine de Québec. Rev. Sci. Eau. 14: 489-510.
Twardowski, M. S., J. M. Sullivan, P. L. Donaghay, and J. R. V. Zaneveld. 1999. Microscale quantification of the absorption by dissolved and particulate material in coastal waters with an ac-9. J. Atmos. Ocean. Technol. 16: 691-707.
Tyler, A. N., E. Svab, T. Preston, M. Presing, and W. A. Kovacs. 2006. Remote sensing of the water quality of shallow lakes: A mixture modelling approach to quantifying phytoplankton in water characterized by high-suspended sediment. Int. J. Remote Sens. 27: 1521-1537.
Tyler, J. E. 1964. Colour of Ocean. Nature 202: 1262-1964.
Tzortziou, M., J. R. Herman, C. L. Gallegos, P. J. Neale, A. Subramaniam, L. W. Harding, and Z. Ahmad. 2006. Bio-optics of the Chesapeake Bay from measurements and radiative transfer closure. Estuar. Coast. Shelf Sci. 68: 348-362.
Vàhàtalo, A. V., and R. G Wetzel. 2004. Photochemical and microbial decomposition of chromophoric dissolved organic matter during long (months-years) exposures. Mar. Chem. 89: 313-326.
V.-Balogh, K., B. Nemeth, and L. Voros. 2009. Specific attenuation coefficients of optically active substances and their contribution to the underwater ultraviolet and visible light climate in shallow lakes and ponds. Hydrobiologia 632: 91-105.
Vincent, W. F., J. J. Dodson, N. Bertrand, and J. J. Frenette. 1996. Photosynthetic and bacterial production gradients in a larval fish nursery: The St Lawrence river transition zone. Mar. Ecol.-Prog. Ser. 139: 227-238.
Vincent, W. F., and J. Laybourn-Parry [eds.]. 2008. Polar Lakes and Rivers: Limnology of Arctic and Antarctic Aquatic Ecosystems. Oxford University Press. New York, NY. 327pp.
Walter, K. M., S. A. Zimov, J. P. Chanton, D. Verbyla, and F. S. Chapin. 2006. Methane bubbling from Siberian thaw lakes as a positive feedback to climate warming. Nature 443: 71-75.
Watanabe, S., M. F. Knowlton, W. F. Vincent, and J. R. Jones. 2009. Variability in the optical properties of colored dissolved organic matter in Missouri reservoirs. Verh. Internat. Verein. Limnol. 30: 1117-1120.
Watanabe, S., I. Laurion, K. Chokmani, R. Pienitz, and W. F. Vincent. 2011. Optical diversity of thaw ponds in discontinuous permafrost: a model system for water color analysis. J. Geophys. Res.-Biogeosci. 116: G02003, doi: 02010.01029/02010JG001380.
142
Wetzel, R. G. 2001. Limnology: Lake and River Ecosystems, 3rd ed. Academic Press. San Diego, CA. 1006pp.
Yentsch, C. S. 1962. Measurement of visible light absorption by particulate matter in the ocean. Limnol. Oceanogr. 7: 207-217.
Zaneveld, J. R. V. 1994. Optical closure: From theory to measurement. In R. W. Spinrad, K. L. Carder and M. J. Perry [eds.], Ocean Optics. Oxford Monographs on Geology and Geophysics. Oxoford University Press. New York. 304pp.
Zaneveld, J. R V., J. C. Kitchen, and C. M. Moore. 1994. The scattering error correction of reflecting-tube absorption meters, p. 44-55. In J. S. Jaffe [ed.], Ocean Optics XII. Proc. SPIE.
Zaneveld, J. R V., J. C. Kitchen, and H. Pak. 1981. The influence of optical water type on the heating rate of a constant depth mixed layer. Journal of Geophysical Research-Oceans and Atmospheres 86: 6426-6428.
Zepp, R. G., G. C. Shank, E. Stabenau, K. W. Patterson, M. Cyterski, W. Fisher, E. Battels, and S. L. Anderson. 2008. Spatial and temporal variability of solar ultraviolet exposure of coral assemblages in the Florida Keys: Importance of colored dissolved organic matter. Limnol. Oceanogr. 53: 1909-1922.
Zhang, Y. L., L. Q. Feng, J. S. Li, L. C. Luo, Y. Yin, M. L. Liu, and Y. L. Li. 2010. Seasonal-spatial variation and remote sensing of phytoplankton absorption in Lake Taihu, a large eutrophic and shallow lake in China. J. Plankton Res. 32: 1023-1037.
Zimov, S. A., E. A. G. Schuur, and F. S. Chapin. 2006. Permafrost and the global carbon budget. Science 312: 1612-1613.
143
Annexe 1
Error analysis for optical measurements
Al.l Introduction Each of the optical variables that was analyzed in this study is subject to
measurement error, which determines the precision of the estimates and the number of
significant figures to be reported. This Annexe presents approaches towards determining
the measurement errors for the optical instruments used in this study, specifically the bench
top spectrophotometer, the in situ spectrophotometer, and the field radiometer system.
A1.2 Methods Al.3.1 Varian Cary 100 bench-top spectrophotometer
The measurement error of the absorption coefficient of colored dissolved organic
matter (acDOMfX)) measured by a bench top spectrophotometer (Caryl00, Varian, Inc.,
USA) was evaluated in three ways. Firstly, the technical specifications for this instrument
were consulted for the manufacturer's statement of measurement accuracy. Secondly, the
absorbance values of a blank sample (pure water) taken prior to the actual sample analysis
conducted on 19 September 2007 were evaluated for the spectral range 400 to 700 nm.
Finally, differences between duplicate measurements of natural water samples from
thermokarst thaw ponds (see, Chaiptre 2) were evaluated over the same spectral range.
Al.2.2 WET Labs AC-S in situ spectrophotometer
The measurement error of an in situ spectrophotometer (AC-S, WET Labs, Inc.,
USA) for the beam attenuation (ct.w(X)) and absorption (at.w(X)) coefficient analyses were
evaluated in two ways. The technical specifications were consulted for the manufacturer's
estimate of accuracy. Differences among multiple readings on natural water samples were
also evaluated. For this evaluation, subsets of data from Chapitre 2 (thermokarst thaw
ponds: KWK2, KWK6, KWK11, KWK16) and Chapitre 3 (Lake St. Charles: 10 June, 22
July, and 19 September 2008) were chosen. These data sets represented the range of optical
conditions observed in the present study. Sample analysis was conducted five times for
144
thaw pond samples and three times for Lake St. Charles samples, and differences among
these readings were analyzed as the standard deviation for multiple readings.
Al.2.3 Satlantic radiometer system
The measurement errors of the Satlantic radiometer system for downward irradiance
(Ed(X)) and upward radiance (LU(X)) were analyzed in three ways. Firstly, the technical
specifications were consulted for the manufacturer's estimate of accuracy. Secondly, the
noise level of the dark counts during profiling was evaluated. Thirdly, the error during field
measurements was evaluated as the standard deviation of above-water measurements at
Lake St. Augustin, made on 5 October 2010 (see, Chapitre 4 for details).
A1.3 Results and discussion Al.3.1 Varian Cary 100 bench-top spectrophotometer
The manufacturer's technical specifications report that the accuracy of the
absorbance measurement is 0.00016 (dimensionless). This value is equivalent to the
QCDOM(X) of 0.037 m"1 when a cuvette with 1 cm path length is applied. The observed noise
of the blank reading was within ±0.0003 (±0.068 m"1 in acDOMfX), Fig. Al.l). The standard
deviation of this noise was 0.00001 (0.023 m"1 in acDOM(X)). The differences between
duplicate sample readings showed no evidence of spectral dependency (Fig. A1.2). The
difference fell within ± 0.0005 (±0.115 m"1 in OCDOM(X)), and mostly in the range ± 0.0002
(±0.046 m'1 in acDOMfX)). The standard deviation of all observed differences was 0.00013
(0.030 m"1 in OCDOM(X)). Therefore, it is appropriate to report values with two digits after
the decimal point (0.01 m"1 level) for the acDOM(X) measured by this instrument with a 1cm
cuvette. It was similar for the absorption coefficient of particulate matter analysis by the
same instrument equipped with an integrating sphere. All absorption measurements
reported in the present study should therefore be rounded to two decimal places.
Al.3.2 WET Labs AC-S in situ spectrophotometer
The manufacturer's technical specifications report that the accuracy of the
instrument is ±0.01 m . Measurement error analysis showed that the error is spectrally
dependent (Fig. A1.3). The error for both ct.w(X) and at.w(X) were mostly within ±0.4 m"1 for
145
thaw ponds (Fig. A 1.3a). It was higher for a sample from an extremely turbid pond
(KWK16), which gave a standard deviation of 1.5 m"1 at the blue end of the spectrum;
however, this was still less than 3 % of the measured value (>60 m"1). The error of ct.w(X)
fell in the range from 0.015 - 0.045 m"1 (solid lines, Fig. A1.3b), and that ofa,.w(X) was less
than 0.02 m"1 (dashed lines, Fig.A1.3b). The much higher error observed in the thaw pond
data relative to Lake St. Charles data would be due to the extremely high measured values
being close to instrumental limit, requiring special sample manipulation (dilution) for the
analysis. In conclusion, the measurements made by the AC-S for thermokarst thaw ponds
were reliable at 0.1 m"1 level, and that for the north temperate lakes (Chapitres 3 and 4)
were at 0.01 m"1 level. The AC-S data in the present study should therefore be rounded to
one (thaw ponds) or two (other lakes) decimal places.
A1.3.3 Satlantic radiometer system
The manufacturer's technical specifications report that the accuracies of the
instrument are 0.000015 W m"2 nm"1 for Ed(X) and 0.0000009 W m"2 nm"1 sr"1 forLu(X). The
observed noise of the dark count fell within ±0.004 W m"2 nm"1 for Ed(X) (Fig. A1.4a) and
±0.00002 W m"2 nm"1 sr"1 for LU(X) (Fig. A 1.4b). The measurement error ofEd(X) was within
±0.03 W m"2 nm"1 (Fig. A1.5a) and that of Lu(k) was ±0.00016 W m"2 nm"1 sr"1 (Fig. A1.5b)
for the above-water measurement under stable incident light conditions. These values were
less than 3 % of measured values. Therefore, our radiometric measurements should be
reported at 0.001 W m"2 nm"1 for Ed(X), and ±0.00001 W m"2 nm"1 sr"1 for LU(X), with
consideration of 2-3 % error. This relative error would be considerable under low
irradiances, for example at the limit of the euphotic zone.
146
c_ c/5 -__ C 0
' Î Â
a OJ
OJ o e -e o ai
-Q <
0.0004 0.0003 0.0002 -I 0.0001 0.0000 -I
-0.0001 -0.0002 -0.0003 -0.0004
400 450 500 550 600 650 700 Wavelength (nm)
Figure Al.l Observed noise on pure water (blank) readings for a Varian Cary 100 spectrophotometer. Absorbance (dimensionless) measured on pure water was normalized by the average of the value in the range from 400 to 700 nm.
147
-0.0006 400 450 500 550 600 650 700
Wavelength (nm)
Figure A1.2 Observed difference between duplicate readings for a Varian Cary 100 spectrophotometer. All 14 observations for the thermokarst thaw ponds (see, Chapitre 2) are shown.
148
a. Thaw ponds
b. Lake St. Charles 0.05
£ 0.04 g I 0.03 ' I -o 0.02 -a = 0.01 03
0.00 400 500 600 700
Wavelength (nm)
Figure A 1.3 Measurement error for the WET Labs AC-S spectrophotometer. Standard deviation of multiple measurements for a. four thermokarst thaw ponds (KWK2, KWK6, KWK11, and KWK 16), and b. three sampling date of Lake St. Charles (10 June, 22 July, and 19 September 2008) are shown. Solid lines represent the beam attenuation coefficients (ct.w(X)) and broken lines the absorption coefficients (at.w(X)).
149
0.004 0.003
'E G
S
0.002 0.001 0.000
'E G
S -0.001
5 -0.002 -0.003 -0.004
2e-5
le-5
E 0
-le-5
-2e-5
a. downward irradiance
b. upward radiance
400 450 500 550 600 650 700 Wavelength (nm)
Figure A 1.4 Observed noise on the dark count for a Satlantic spectroradiometer system for: a. the downward irradiance (Ed(X)) sensor; and b. the upward radiance (LU(X)) sensor.
150
a. downward irradiance "g 0.035 c
? 0.030 £ 0.025 -
c 0.020 \ o 1 0.015 ] >
•a 0.010 -I -o 03 -a = 0.005 ♦_f
in 0.000 ^ _ s
1.8e-4
s 1.6e-4
s 1.4e-4 -1 ■fl 1.2e-4 < 1.0e-4
8.0e-5 fi o 6.0e-5 03 4.0e-5 > OJ
-a ■H m -a
2.0e-5 0.0
-2.0e-5
b. upward radiance
■ j j 400 450 500 550 600 650
Wavelength (nm) 700
Figure A 1.5 Measurement error for the Satlantic radiometer positioned above the water. Standard deviation of a. downward irradiance (Ed(X))\ and b. upward radiance (LU(X)).
151
Annexe 2
Particle size distribution analysis of thaw ponds
A2.1 Introduction Non-algal suspended particles play a major role in optical processes in thermokarst
thaw ponds in the subarctic region (Watanabe et al. 2011), and thus an improved
knowledge about their optical characteristics is required to understand the limnological and
biogeochemical characteristics of these ecosystems. The grain size distribution of the
particles is an important determinant of their optical properties (Babin et al. 2003; Reynolds
et al. 2010) that strongly influence their absorption and scattering of light in the water
column. To improve the understanding of attenuation and reflectance processes in these
ponds, we examined the particle size distribution of suspended particulate matter in a
subset of sampled ponds.
A2.2 Methods Surface water of six ponds was sampled in the discontinuous permafrost region near
Whapmagoostui-Kuujjuarapik, Nunavik, Canada between 28 June and 7 July 2007 (see,
Chapitre 2). Sampled waters were preserved with glutaraldéhyde (0.5 %) and stored in the
cold and dark until analysis at the Laboratoire d'analyse des particules et des surfaces,
Institut des sciences de la mer de Rimouski, Université du Québec à Rimouski, Rimouski,
Québec. The size distribution of suspended particulate matter was determined using a
laboratory laser diffractometer (LSI3320 laser diffraction particle size analyzer, Beckman
Coulter, Inc., USA). The instrument was calibrated by standards prior and after analysis.
The diffraction and scattering patterns were measured on 250 ml of water samples in the
aqueous liquid module. The obtained patterns were then converted to particle size
distribution (as percent of total volume) using Mie theory, assuming a real index of
refraction of 1.8 and an imaginary index of refraction of 0.3.
152
A2.3 Results and discussion All six particle size distribution spectra were similar (Fig. A2.1). There were two
peaks: one fell in the range from 0.1 to 0.2 pm, and the other 1.5 to 2 pm, implying that the
particles consist of fine clays within two size ranges that differ by an order of magnitude.
These results, however, are highly dependent on the assumption of the refractive index. It
has been shown that the shape of the particle size spectrum of fine grained lake sediments
and soils obtained by laser diffractometory changes drastically depending on the refractive
index chosen (Sperazza et al. 2004). The results obtained here must therefore be considered
tentative until accurate estimates of refractive indexes are available. In the future, it also
would be useful to explore other methods for particle characterization in the thaw pond
systems, such as direct microscopic analysis (Andrews et al. 2010), electrical impedance
particle sizing (Coulter Counter; Reynolds et al. 2010), or Brownian motion analysis (e.g.,
NanoSight®, NanoSight Ltd., UK).
A2.4 References Andrews, S., D. Nover, and S. G. Schladow. 2010. Using laser diffraction data to obtain
accurate particle size distributions: the role of particle composition. Limnol. Oceanogr. Meth. 8: 507-526.
Babin, M., A. Morel, V. Fournier-Sicre, F. Fell, and D. Stramski. 2003. Light scattering properties of marine particles in coastal and open ocean waters as related to the particle mass concentration. Limnol. Oceanogr. 48: 843-859.
Reynolds, R. A., D. Stramski, V. M. Wright, and S. B. Wozniak. 2010. Measurements and characterization of particle size distributions in coastal waters. J. Geophys. Res.-Oceans 115.
Sperazza, M., J. N. Moore, and M. S. Hendrix. 2004. High-resolution particle size analysis of naturally occurring very fine-grained sediment through laser diffractometry. J. Sediment. Res. 74: 736-743.
Watanabe, S., I. Laurion, K. Chokmani, R. Pienitz, and W. F. Vincent. 2011. Optical diversity of thaw ponds in discontinuous permafrost: a model system for water color analysis. J. Geophys. Res.-Biogeosci. 116: G02003, doi: 02010.01029/02010JG001380.
153
e
i O
KWK3 KWK6 KWK 16 KWK21 KWK23 KWK33
0.01 0.1 1
Particle diameter (\xm) io
Figure A2.1 Particle size distribution of six thaw ponds (see, Chapitre 2 for pond descriptions). Note that the particle diameters are plotted on a log scale.
154
155
Annexe 3
The path length amplification factor (beta factor) for the
quantitative filter technique
A3.1 Introduction The quantitative filter technique (QFT, Yentschl962) has long been used to
measure the absorption coefficient of suspended particulate matter (ap(X)) in a variety of
natural waters, and was the approach adopted in the present study. This method, however,
incorporates uncertainty associated with multiple scattering of photons in the filter pad
structure, which causes an causing over-estimation that is referred to as the path length
amplification (Mitchell et al. 2002, and references therein). The correction of the
overestimation has been conducted by applying a factor in the conversion of the measured
absorbance to the absorption coefficient, and this factor is referred to as the path length
amplification factor or the beta factor (fi in equations 2.8, 3.4, and 4.8). In the present study,
this factor was assumed to be a constant (fi = 2), following Roesler (1998). However, it has
been suggested that the beta factor can vary depending on the absorbance value (i.e., the
factor can vary with particle load on the filter), and several different estimation methods
have been proposed (see, Mitchell et al. 2002). The objectives of the analysis presented in
this Annexe were to evaluate the QFT beta factor by comparing the resultant ap(X) with
different beta factors with ap(X) values obtained by spectrophotométrie analysis using the
WET Labs AC-S.
A3.2 Methods 3.2.1 The quantitative filter technique
The surface water of Lake St. Augustin was sampled at the center of the lake on 5
October 2010 (see, Chaptire 4), and the water was refrigerated in dark until laboratory
analyses. The water samples were filtered onto GF/F filters and stored at -80°C until optical
analysis. The absorbance of particulate matter was measured in a Cary 100 dual beam
spectrophotometer (Varian Inc., Canada) fitted with an integrating sphere (Labsphere Inc.,
156
USA), and following the method of Mitchell et al. (2002). The absorption coefficients of
suspended particulate matter (ap(Xj) were calculated as:
. , , 2.303__(;i)-r p B V
where T is the filtered area, V is filtered volume, and B is the path length amplification
factor (Mitchell et al. 2002). To estimate the R, three different methods were applied: 1) B
was assumed to be a constant value (fi = 2; Roesler 1998); 2) It was estimated by a
quadratic function (fi = 0.406 ± 0.519 A(X)~l) as suggested by Tassan and Ferrari (1995); 3)
it was estimated by a power function (fi = 1.630 A(X)~0220) as suggested by Bricaud and
Stramski (1990).
3.2.2 WET Labs AC-S analysis
The ap(X) was also measured using an AC-S in situ spectrophotometer (25 cm path
length, WET Labs Inc., USA) on the laboratory bench (see, Watanabe et al. 2011 for
details). The absorption coefficients were measured both on unfiltered water and on the
filtrate that passed through a glass fiber filter (GF/F; Whatman Inc., USA), and ap(X) was
obtained as their difference. To test the influence of scattering correction, three different
scattering correction methods were applied to obtain the absorption coefficients (see,
Methods of Chapitre 4): the null correction, the proportional correction (Pegau et al. 2003),
and the modified proportional correction (Gallegos et al. 2008),
A3.3 Results and discussion The three different B factor estimations provided considerably different ap(X) spectra
(Fig. A3.1, blue lines). The values were comparable at blue end of the spectrum but the
difference was pronounced at the red end of the spectrum. The constant beta (fi = 2),
applied in the present study, gave lowest values (blue solid line), the Tassan and Ferrari
(1995) method (blue dotted line) gave the highest values, and Bricaud and Stramski (1990)
produced intermediate values (blue dashed line). For example, the resultant absorption
157
coefficient at 440 nm was: 0.762 m"1 for the constant beta, 1.108 m"1 for the Tassan and
Ferrari (1995) approach, and 0.863 m"1 for the Bricaud and Stramski (1990) approach.
Although it was expected to validate these results by comparing them with the value
measured by method (AC-S), our ap(X) values from the latter instrument also incorporated
considerable uncertainties (Fig. A3.1, red lines). It was, therefore, not possible to identify
the best B estimation method from the present data set. This difficulty was also true for the
determination of the absorption coefficient of fine particle by QFT with nitrocellulose
membrane filters (pore size 0.22 pm). The constant beta factor (fi = 2) was also applied in
the present research; however, this could include uncertainties needed to be evaluated. Also,
the characteristics of errors would be significantly different, because the optical
characteristics of the 0.22 pm membrane filters likely differ from those of GF/F filters.
Thus, careful evaluation of the beta factor both for GF/F and membrane filters are required
in future research. The determination of particulate absorption values with minimum error
will ultimately be required to achieve this challenging goal (see, Discussion of Chapitre 4
for further details).
A3.4 References
Bricaud, A., and D. Stramski. 1990. Spectral absorption-coefficients of living phytoplankton and nonalgal biogenous matter - a comparison between the Peru upwelling area and the Sargasso Sea. Limnol. Oceanogr. 35: 562-582.
Gallegos, C. L., R. J. Davies-Colley, and M. Gall. 2008. Optical closure in lakes with contrasting extremes of reflectance. Limnol. Oceanogr. 53: 2021-2034.
Mitchell, B. G., M. Kahru, J. Wieland, and M. Stramska. 2002. Determination of spectral absorption coefficients of particles, dissolved material and phytoplankton for discrete water samples, p. 39-64. In J. L. Mueller and G. S. Fargion [eds.], Ocean Optics Protocols for Satellite Ocean Color Sensor Validation, Revision 4. National Aeronautics and Space Administration.
Pegau, S., J. R. V. Zaneveld, B. G. Mitchell, J. L. Mueller, M. Kahru, J. Wieland, and M. Stramska. 2003. Inherent optical properties: Instruments, characterizations, field measurements and data analysis protocols. In J. L. Mueller, G. S. Fargion and R. McClain [eds.], Ocean optics protocols for satellite ocean color sensor validation; Revision 4. National Aeronautics and Space Administration.
158
Roesler, C. S. 1998. Theoretical and experimental approaches to improve the accuracy of particulate absorption coefficients derived from the quantitative filter technique. Limnol. Oceanogr. 43: 1649-1660.
Tassan, S., and G. M. Ferrari. 1995. An alternative approach to absorption measurements of aquatic particles retained on filters. Limnol. Oceanogr. 40: 1358-1368.
Watanabe, S., I. Laurion, K. Chokmani, R. Pienitz, and W. F. Vincent. 2011. Optical diversity of thaw ponds in discontinuous permafrost: a model system for water color analysis. J. Geophys. Res.-Biogeosci. 116: G02003, doi: 02010.01029/02010JG001380.
Yentsch, C. S. 1962. Measurement of visible light absorption by particulate matter in the ocean. Limnol. Oceanogr. 7: 207-217.
159
! X
2.0
1.5
1.0 *
0.5
0.0 400 500 600 700
Wavelength (nm)
Figure A3.1 Comparison of the absorption coefficients of suspended particulate matter (ap(Xj) obtained by different methods. Blue lines show ap(X) measured by the quantitative filter technique: the solid line is that calculated with the constant beta factor (fi = 2), dotted line is that with the beta factor estimated by the Tassan and Ferrari (1995) method, and dashed line is by the Bricaud and Stramski (1990) method. Red lines show ap(X) measured by the WET Labs AC-S; solid line is the value estimated by the null correction, dotted line is the proportional correction, and the dashed line is the proportional correction modified by Gallegos et al. (2008). Details of the scattering correction methods are given given in Chapitre 4.
