Optical assessment of colored dissolved organic matter and its related parameters in dynamic coastal water systems

Accepted Manuscript Optical assessment of colored dissolved organic matter and its related parameters in dynamic coastal water systems Palanisamy Shanmugam, Theenathayalan Varunan, S.N. Nagendra Jaiganesh, Arvind Sahay, Prakash Chauhan PII:

S0272-7714(16)30087-7

DOI:

10.1016/j.ecss.2016.03.020

Reference:

YECSS 5080

To appear in:

Estuarine, Coastal and Shelf Science

Received Date: 31 October 2015 Revised Date:

1 February 2016

Accepted Date: 26 March 2016

Please cite this article as: Shanmugam, P., Varunan, T., Jaiganesh, S.N.N., Sahay, A., Chauhan, P., Optical assessment of colored dissolved organic matter and its related parameters in dynamic coastal water systems, Estuarine, Coastal and Shelf Science (2016), doi: 10.1016/j.ecss.2016.03.020. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT

Optical assessment of colored dissolved organic matter and its related parameters in dynamic coastal water systems Palanisamy Shanmugam1*, Theenathayalan Varunan1, S.N. Nagendra Jaiganesh1, Arvind Sahay2,

1

RI PT

Prakash Chauhan2 Ocean Optics and Imaging Laboratory, Department of Ocean Engineering, Indian Institute of

Technology Madras, Chennai 600036, India 2

Space Applications Centre, Ahmedabad 380015, India

SC

*Corresponding author: Phone: 91-44-22574818, Email: [email protected]

Abstract: Prediction of the curve of the absorption coefficient of colored dissolved organic matter

M AN U

(CDOM) and differentiation between marine and terrestrially derived CDOM pools in coastal environments are hampered by a high degree of variability in the composition and concentration of CDOM, uncertainties in retrieved remote sensing reflectance and the weak signal-to-noise ratio of spaceborne instruments. In the present study, a hybrid model is presented along with empirical methods to remotely determine the amount and type of CDOM in coastal and inland water environments. A large set of in-situ data collected on several oceanographic cruises and field campaigns from different regional waters was used to develop empirical methods for studying the distribution and dynamics of CDOM,

TE D

dissolved organic carbon (DOC) and salinity. Our validation analyses demonstrated that the hybrid model is a better descriptor of CDOM absorption spectra compared to the existing models. Additional spectral slope parameters included in the present model to differentiate between terrestrially derived and marine CDOM pools make a substantial improvement over those existing models. Empirical algorithms to derive

EP

CDOM, DOC and salinity from remote sensing reflectance data demonstrated success in retrieval of these products with significantly low mean relative percent differences from large in-situ measurements. The

AC C

performance of these algorithms was further assessed using three hyperspectral HICO images acquired simultaneously with our field measurements in productive coastal and lagoon waters on the southeast part of India. The validation match-ups of CDOM and salinity showed good agreement between HICO retrievals and field observations. Further analyses of these data showed significant temporal changes in CDOM and phytoplankton absorption coefficients with a distinct phase shift between these two products. Healthy phytoplankton cells and macrophytes were recognized to directly contribute to the autochthonous production of colored humic-like substances in variable amounts within the lagoon system, despite CDOM content being partly derived through river run-off and wetland discharges as well as from conservative mixing of different water masses. Spatial and temporal maps of CDOM, DOC and salinity products provided an interesting insight into these CDOM dynamics and conservative behavior within the 1

ACCEPTED MANUSCRIPT

lagoon and its extension in coastal and offshore waters of the Bay of Bengal. The hybrid model and empirical algorithms presented here can be useful to assess CDOM, DOC and salinity fields and their changes in response to increasing runoff of nutrient pollution, anthropogenic activities, hydrographic variations and climate oscillations.

RI PT

Key words: Optical assessment, CDOM, DOC and Salinity, Coastal system, Remote sensing, Land-ocean interaction 1.

Introduction

Dissolved organic matter (DOM) is a highly abundant form of organic matter resulting from naturally

SC

occurring humic and fulvic acids (mainly soluble fractions) and by-products from the decomposition of organisms of both terrestrial and aquatic origin, thus representing the largest pool of carbon in the coastal

M AN U

ocean and associated inland water systems (Vodacek et al., 1997; Nelson et al., 1998; Hansell and Carlson, 2002; Steinberg, 2003; Laanen, 2007; Loiselle et al., 2009; Zhang et al., 2010; Romera-Castillo et al., 2010). DOM introduced to these systems, which are subjected to nutrient enrichment caused by ever-increasing coastal industrialization, urbanization, and agricultural activities, is mainly through river run-off and wetland discharges. DOM plays a key role in a broad range of land-ocean interaction processes and climate-related biogeochemical cycles (e.g., carbon, oxygen, hydrogen, nitrogen, sulfur, and phosphorus) and has direct implications on coastal ecosystem health and productivity (Laanen, 2007;

TE D

Mannino et al., 2008; Shanmugam, 2011a; Brezonik et al., 2015). These intense physical and biological processes cause many coastal margin ecosystems to become hot spots of carbon cycling. DOM (that contains chromophores) can be optically measurable because of its strong absorption of UV and visible light, hence the term Chromophoric (or Colored) Dissolved Organic Matter (CDOM). This fraction of

EP

DOM has sources that emit a range of long wavelengths (fluorescence, hence the term FDOM) when exposed to short wavelength light. Thus, CDOM content is typically measured on the basis of absorbance or fluorescence and often used as a substitute value for DOM (Hansell and Carlson, 2002; Steinberg,

AC C

2003; Zhang et al., 2010; Romera-Castillo et al., 2010) and an alternative proxy for dissolved organic carbon (DOC) (Hansell and Carlson, 2002; Steinberg, 2003; Mannino et al., 2008; Del Castillo and Miller, 2008; Monteith et al., 2007; Roulet and Moore, 2006; Dupont and Aksnes, 2013). CDOM is used in many important applications such as quantification of carbon transport and continuous monitoring of wastewater discharge as the CDOM fluorescence which is related to total organic carbon TOC (Smart et al., 1976). CDOM can act as a natural tracer and indicate the dispersion, transport, and mixing of water mass (Wiley and Atkinson, 1982; Boss et al., 2001). Thus, it can be exploited in the tracing of special water bodies and assessment of physico-chemical water quality. CDOM is also an important quantity for other studies related to primary production, underwater imaging and communication, CDOM cycling and 2

ACCEPTED MANUSCRIPT

degradation processes and energy budgets (Vodacek et al., 1997; Nelson et al., 1998; Hansell and Carlson, 2002; Steinberg, 2003; Laanen, 2007). Consequently, the capability to routinely quantify CDOM levels in coastal and inland waters systems would greatly increase the effectiveness of monitoring efforts, improve satellite retrievals of phytoplankton pigment and other water constituents, and help explain

RI PT

events such as a sudden decrease in primary productivity, algal blooms, phytoplankton regime shifts, and related changes in an aquatic environment.

The spectral characteristics of CDOM – the knowledge required for bio-optical modeling studies to investigate the spatial and temporal dynamics of CDOM in aquatic ecosystems – are primarily determined

SC

by the molecular structures of humic and fulvic acids. The absorption of light by CDOM (aCDOM(λ)) is strongest in the UV and blue wavelengths and approaches to near zero in the red and near-infrared region. This spectral behavior is approximately described by the exponential equation (Jerlov, 1976; Bricaud et

M AN U

al., 1981). The slope S of the exponential function that describes the shape of the CDOM absorption curve is commonly used as an indicator for the molecular properties (composition) of the humic substances present in a water sample (Hansell and Carlson, 2002; Laanen, 2007; Helms et al., 2008; Song et al., 2013; Brezonik et al., 2015), and has also been previously used to trace changes in the CDOM pool from production and degradation mechanisms as well as from conservative mixing of different water masses (Hansell and Carlson, 2002; Stedmon and Markager, 2001; Bracchini at al., 2006). Thus, current CDOM

TE D

absorption models (e.g., single exponential model as described by Bricaud et al. (1981), Schwarz et al. (2002), and Loiselle et al. (2009); double exponential model by Stedmon and Markager (2001); a hybrid exponential model by (Shanmugam, 2011a)) make use of this slope for predicting the spectral absorption curves of CDOM in marine environments (with relatively low CDOM). CDOM absorption spectra of inland and turbid productive coastal and lagoon waters, typically with higher CDOM contents, can show a

EP

combination of a steep slope below 420 nm and a gentle slope beyond 420 nm due to a high degree of variability in the composition and concentration of CDOM influenced by terrestrial and local production

AC C

processes (Helms et al., 2008). Their measurement data from a variety of samples (marsh, riverine, estuarine, coastal, and open-ocean) exhibit a relatively more complex quasi-exponential decay of the absorption of CDOM (summarized with amplitude and shape) with increasing wavelength. This behavior of the CDOM absorption implies that a single spectral slope associated with those exponential models is not adequate for describing CDOM variability in coastal and inland waters, where its spectral slope can vary spatially and seasonally (Carder et al., 1989; Stedmon and Markager, 2001; Blough and Del Vecchio, 2002). Thus, there is a need to develop a generalized model that fits to the measured absorption spectrum over a broad range of wavelengths (i.e., 350-650 nm) and such spectra must be free from measurement noise in clear waters and scattering errors in turbid coastal waters (Stedmon and Markager, 2001; Laanen, 2007; Shanmugam, 2011a). A systematic validation of such a model with a sufficiently 3

ACCEPTED MANUSCRIPT

large measurement data is also necessary to evaluate its accuracy and stability over a wide range of inland and marine environments. In-situ CDOM measurements traditionally involve point sampling and laboratory analysis providing limited information in both space and time. Spectrophotometric techniques are often used to measure

RI PT

CDOM absorption coefficients but suffer from problems associated with sample handling (concentration and dilution) and storage and transport for subsequent laboratory analysis. Such measurement methods prevent the routine collection and analysis of CDOM samples from remote waters (Miller et al., 2002; Laanen, 2007). Alternatively, remote sensing methods are more cost-effective than traditional in-situ

SC

methods and can provide spatial and temporal information, which is a significant advantage over discrete in-situ sampling (Hansell and Carlson 2002; Laanen, 2007; Mannino et al., 2008). In the past decades, several empirical and semi-analytical algorithms have been developed and used with some success in

M AN U

coastal and oceanic waters. Empirical algorithms that employ band-ratios of reflectance at specific wavelengths in the visible domain require adequate in-situ data to parameterize the model for certain regional applications (Table 1). By contrast, semi-analytical/quasi-analytical models (Lee et al., 2002; Siegel et al., 2002; IOCCG, 2006; Zhu and Yu, 2013; Dong et al., 2013; Le and Hu, 2013) that incorporate both the empirical parameters and bio-optical models (based on radiative transfer calculations) require prior knowledge about specific inherent optical properties and the absorption slopes

TE D

of CDOM and non-algal particles (Brando and Dekker, 2003; Zhu et al., 2014). These semi-analytical models provide absorption coefficients of the CDOM and detritus together (collectively named as CDM, IOCCG (2006)) because of their similar shape and the difficulty in differentiating them in absence of a mechanical (namely a filtering) treatment (Morel and Gentili, 2009). Moreover, some semi-analytical models often use a constant slope for CDM, which cannot capture the CDOM spectral variability spatially

EP

(nearshore to offshore) and seasonally. This eventually limits our knowledge of CDOM dynamics and factors controlling its distributions and the capability to monitor DOC distribution spatially and

AC C

temporally in inland and coastal oceanic waters (Shanmugam, 2011a). There is also a lack of knowledge regarding the optical characteristics of different CDOM sources (e.g., autochthonous CDOM derived from algae and macrophytes and allochthonous CDOM derived from organic soil and humic substances of terrestrial sources) that present spatially and spectrally distinct CDOM absorption properties in coastal and inland systems. Moreover, the use of remote sensing in coastal and inland water systems, where CDOM absorptivity is often spatially and temporally very diverse (Helms et al., 2008; Song et al., 2013; Zhu et al., 2014; Brezonik et al., 2015), is limited because of the failure of atmospheric correction algorithms due to the interference of water constituents (Shanmugam, 2012; Singh and Shanmugam, 2014).

4

ACCEPTED MANUSCRIPT

Because DOC and CDOM can covary with salinity within coastal regions influenced by freshwater discharge and ecosystem processes (Mantoura and Woodward, 1983; Del Vecchio and Blough, 2004), past efforts took advantages of the relationships between these variables and demonstrated some success in satellite retrieval of DOC and salinity from CDOM-based algorithms derived using limited field

RI PT

measurements. For instance, Mannino et al. (2008) developed empirical algorithms to first derive the CDOM absorption coefficients (aCDOM(412) and aCDOM(355)) from Rrs ratios for MODIS-Aqua and SeaWiFS and then derive DOC through the relationship of CDOM-DOC in coastal waters within the US Middle Atlantic Bight. Del Castillo and Miller (2008) developed similar empirical relationships to derive DOC, CDOM, and salinity from monthly composites of SeaWiFS imagery in the Mississippi River

SC

plume. The existence of such relationships is indeed confirmed by other studies (Ferrari, 2000; Bowers et al., 2000, 2004; Binding and Bowers, 2003; Chaichitehrani, 2012; Ahn et al., 2008; Chaichitehrani et al.,

M AN U

2013). These approaches are found to be effective methods to derive DOC variations both spatially and temporally and enhance our understanding of the carbon cycling in the coastal oceanic regimes, although this relationship is found a poor indicator of DOC in manmade reservoirs due to exogenous DOC inputs from multiple sources (Hestir et al., 2015).

SeaWiFS, MODIS-Aqua and other similar ocean color sensors provide reliable ocean color information in oceanic waters, but these are broad band sensors with a limited dynamic range (leading to sensor

TE D

saturation over coastal and inland water bodies) and insufficient spatial resolution for measurements on coastal and inland water bodies. On the contrary, The Hyperspectral Imager for the Coastal Ocean (HICO) sensor is specifically designed to sample the coastal oceans (including estuaries, rivers, lagoons and other shallow-water areas). The HICO acquired hyperspectral images over a high dynamic range (14

EP

bits, translating into 16383 potential grey levels) with a high spatial resolution (90 m) that enables us to derive important coastal products including suspended sediments, CDOM, and algal pigments, water clarity, bottom types, bathymetry and on-shore vegetation maps.

AC C

In the present work, we use a large in-situ data set (consisting of physicochemical, optical, biological and biogeochemical measurements) in conjugation with remote-sensing observations and modeling approaches to develop a reliable and general-purpose CDOM model and algorithms for CDOM retrievals from satellite imagery in coastal and associated inland water systems. This CDOM model is an extended version of our previous model (Shanmugam, 2011b), and is evaluated using different in-situ data sets from both marine and inland waters (lakes, lagoons, river mouths and estuaries). These data sets are independent of those used for deriving the required model parameters (such as chlorophyll, exponential and hyperbolic slopes, CDOM values at reference wavelengths). Since CDOM behaves conservatively over the time scale of mixing processes in many coastal environments and thus exhibits a strong

5

ACCEPTED MANUSCRIPT

correlation with salinity and DOC, we develop empirical algorithms for retrieval of these products using a larger in-situ data set gathered from estuarine, riverine and coastal lagoonal regions. The practical utility of these algorithms is further investigated using HICO (Hyperspectral Imager for the Coastal Ocean) data from the coastal region of the Bay of Bengal. This region includes a lagoon-coastal system with different

RI PT

hydrologic and land-use characteristics and the potential CDOM sources (aquatic vegetation, sewage, marsh vegetation, agricultural drainage, swamps, and lagoons), wherein field observation data are used to corroborate a hypothesis that reveals an apparent phase shift between phytoplankton and CDOM. Data and methods

2.1.

Study site

SC

2.

Measurements were performed at various locations in Muttukadu lagoon (a coastal lagoonal system

M AN U

bounded by the latitude and longitude 12° 48' 2" N; 80° 14' 36.6" E) and coastal waters around Chennai (13° 7' 37" N; 80° 22' 9" E) on the southeast part of India (Fig. 1). The in-situ measurements in Muttukadu lagoon included seven stations from the landward side to the seaward side of the catchment, depending on the different hydrologic and land-use characteristics of the study area and the potential DOM sources (e.g., sewage, aquatic vegetation, agricultural drainage, marsh wetland, swamps). These insitu measurements were performed during 10 November 2013, 16 December 2013, 19 March 2014, 18 April 2014, and 27 July 2014 (the seasons characterized by largely different precipitation and river

TE D

discharges/run-off with contrasting seasonal changes in water flow and biogeochemical exchanges, Dev and Shanmugam (2014)). Field measurements made on 16 December 2013 and 18 April 2014 were coincident with HICO measurements over this study site (used for validation), whereas field measurements conducted two days and 9 days before the HICO measurements during November 2013

EP

and March 2014 were used for general comparison purpose. The temporal difference between the HICO measurements and the field measurements was 30 days for the July 2014 campaign and thus the matchups from these data were not used for validation. Our field measurements indicated that because the

AC C

Muttukadu lagoon is typically a highly turbid productive water system, CDOM, chlorophyll and turbidity always reached high values of the order of those typically measured in highly eutrophic coastal lake and estuarine systems. This suggests that this region is a good test site for investigating the spatial and temporal variability of CDOM, DOC and salinity and studying the transport of carbon flux from the inland areas to the offshore. Similar measurements were conducted in coastal waters of Chennai during these periods which are usually characterized by relatively low dissolved and particulate materials. To date, no studies have been conducted to describe observed variability in the bio-optical data sets from Muttukadu lagoon and coastal waters and this leaves the spatial and temporal distributions of climatically important biogeochemical and physical variables unknown. 6

ACCEPTED MANUSCRIPT

2.2.

In-situ data from coastal and inland waters around India

Remote sensing reflectance (Rrs) is an important optical property commonly used to develop remotesensing algorithms for deriving information regarding the water constituents and studying the upper ocean biogeochemical processes (Shanmugam, 2011a & b; Dev & Shanmugam, 2014). It is the ratio of the

RI PT

water-leaving radiance (Lw (λ)) to the downwelling irradiance (Ed (λ)), i.e., Rrs (λ) = Lw (λ)/ Ed (λ), which is dependent on the ratio of two important inherent optical properties of the water: i.e., the absorption coefficient a and the backscattering coefficient bb as well as on the light field in it measured (Morel and Prieur, 1977). However, the remote sensing reflectance is not a directly measurable quantity and its

SC

estimation is based on the determination of water-leaving radiance (e.g., Lw = Lt − ρLsky (Mobley, 1999), where Lt= total radiance, ρ= Fresnel reflectance, Lsky= sky radiance; ρLsky represents the surface reflected light caused by the sky). Here, above-surface spectra including the total water-leaving radiance,

M AN U

sky radiance and downwelling irradiance were measured using RAMSES-TRIOS hyperspectral radiometers (ARC and ACC). These data were used to determine the remote sensing reflectances as outlined in (Mobley, 1999). The Trios radiometers provide high quality spectral data for aquatic environments (Dev and Shanmugam, 2014). The reflectance measurements were made from a small boat in the middle of the Muttukadu lagoon, where water depth (up to 2.5 m in the channel) was sufficiently large and water clarity sufficiently small (Secchi depth <5 cm due to phytoplankton blooms or high

sediments (Tables 2-3).

TE D

detrital and CDOM contents) that these measurements were not biased by light reflectance from bottom

Water samples collected simultaneously during these measurements were filtered and particles collected by filtration on GF/F filters (0.7 µm pore size / 47 mm for chlorophyll measurements, 0.7 µm / 25 mm for

EP

particulate absorption measurements) were used for the determination of chlorophyll concentration and absorption coefficients of particulate matters. Another set of water samples was collected and filtered through 0.45 and 0.2 µm filters and the filtered water samples were stored in glass flasks in the dark at

AC C

4°C. Upon returning to the laboratory, these samples were analyzed for the measurements of absorption coefficients of CDOM (aCDOM(λ)) using the standard spectrophotometric method (further details on the sampling procedures and protocols for CDOM measurements can be found in Shanmugam, 2011a). For the determination of chlorophyll concentration, the chlorophyll pigments were extracted with 90% acetone and measured in a Shimadzu UV-Visible Spectrophotometer – UV-2600 according to Jeffrey and Humphrey (1975) (e.g., ((11.85 × OD664 ) − (1.54 × OD647 ) − (0.08 × OD630 )) × v V ) , where OD - optical density, v - volume of acetone, and V - volume of water samples filtered). The absorption coefficients of phytoplankton and detrital matter were measured with the above spectrophotometer following the methods presented in Ahn and Shanmugam (2007). 7

ACCEPTED MANUSCRIPT

2.3.

In-situ data from other regional and global waters

For assessing the performance of the hybrid exponential model of this study and developing empirical algorithms to derive the required model parameters for remote sensing applications, a global in-situ data set – the NASA Bio-optical Marine Algorithm Data Set (NOMAD; Werdell and Bailey, 2005) – was

RI PT

obtained from the website: http://seabass.gsfc.nasa.gov/. The NOMAD in-situ data set contains pigment and optical data, salinity and ancillary information collected simultaneously over a significant variety of oligotrophic, mesotrophic and eutrophic waters. In this data set, chlorophyll-a, aCDOM(λ), Rrs(λ) and salinity data were extracted and used in this study

.

SC

Since there is a strong relationship between CDOM, DOC and salinity as a result of the impact of freshwater distributions in coastal waters (Mantoura and Woodward, 1983; Del Vecchio and Blough, 2004; Mannino et al., 2008; Zhu et al., 2014), our approach was to obtain coincident field measurements

M AN U

of these parameters from different waters to acquire a sufficiently large in-situ data set for developing and validating satellite algorithms. Chaichitehrani (2012) and Chaichitehrani et al. (2013) reported field data of CDOM spectral absorption coefficients, DOC concentration, and salinity collected on several cruises (during 2005 and 2007–2009) around the birdfoot Delta, Barataria Bay, Terrebonne Bay, Atchafalaya Bay and Vermilion Bay in the northern Gulf of Mexico influenced by the Mississippi River and Atchafalaya River plumes. D’Sa (2008) made salinity and CDOM absorption measurements at various locations

TE D

between the 10 and 20 m depth during July 2005 in a region influenced by a circulation feature dominated by the westward coastal flow of the lower salinity waters of the Atchafalaya River. In-situ data collected from this region exhibits a large gradient in salinity and CDOM absorption coefficient. Mannino et al. (2008) obtained DOC and CDOM from multiple cruises (during 2005 and 2006) within the U.S. Middle

EP

Atlantic Bight and across the Chesapeake Bay and Delaware Bay mouths and plume regions to investigate the impact of freshwater discharge on coastal ocean distributions of these properties (DOC and CDOM). These data represent a wide range of waters profoundly influenced by highly variable bio-

AC C

optical properties, primary productivity, and carbon flux. Many of these in-situ data from regional and global waters (NOMAD) were extracted representing a wide range of optical conditions and a large variability in CDOM, DOC and salinity within continental margins. 2.4.

Satellite data

The retrieval algorithms developed here were applied to Hyperspectral Imager for the Coastal Ocean (HICO) measurements over the coastal environments of the Bay of Bengal (particularly coastal waters and turbid productive lagoon waters around Chennai on the southeast part of India) to illustrate their applicability and success. HICO is the first demonstration of environmental characterization of the coastal zone using a spaceborne maritime hyperspectral imager. It provides hyperspectral images at 90 m with 8

ACCEPTED MANUSCRIPT

full spectral coverage (most sensitive channels from 380 to 960 nm sampled at 5.7 nm interval) and a very high signal-to-noise ratio to enable retrieval of environmentally relevant quantities in the coastal ocean (http://hico.coas.oregonstate.edu/). Due to cloud cover and environmental conditions, we could acquire only a limited number of HICO scenes at the time of our field measurements in the study region during 22

RI PT

November 2013, 16 December 2013, 22 March 2014, 18 April 2014 and 28 July 2014. Two of the HICO scenes acquired over the coastal waters and turbid productive lagoon waters around Chennai on 16 December 2013 and 18 April 2014 were coincident with our in-situ measurements, and hence they were used for validating the derived products. Other images were processed and used for general comparison purpose. The fundamental quantity in aquatic remote sensing is the water-leaving radiance which is

SC

retrieved after removing the undesired signal (atmospheric signal and surface-reflected light) from the total signal recorded by the space-borne sensors (this procedure is known as “atmospheric correction”).

M AN U

Atmospheric correction, however, still remains a challenging issue for HICO sensor, particularly in optically complex coastal and inland waters. In this paper, we used the SSAS (Singh and Shanmugam Aerosol and Sunglint correction algorithm) algorithm with fine tuning required for both water and land applications. SSAS includes a new scheme for estimating aerosol radiance and extrapolating it across all the visible wavelengths (Singh and Shanmugam, 2014a) and a scheme for estimating sunglint radiances at all visible wavelengths (Singh and Shanmugam, 2014b). The SSAS algorithm builds upon the knowledge gained from the previous studies by Shanmugam and Ahn (2007), Shanmugam (2012), Shanmugam et al.

TE D

(2013) and Dev and Shanmugam (2014). SSAS is robust in terms of yielding physically realistic waterleaving (or surface-leaving) radiances for all visible bands. The standard NIR and SWIR schemes within SeaDAS software are generally applicable for open ocean waters, but produce negative water-leaving radiances across all the visible wavelengths in productive coastal and inland waters (Singh and

2.5.

EP

Shanmugam, 2014a).

Performance assessment

AC C

Assessment of the performance of models is based on six statistical parameters comparing the modelderived values of aCDOM(λ), DOC and salinity with the field measurements. These parameters include the Mean Relative Error (MRE), Root Mean Squared Error (RMSE), Bias, Intercept, Slope and R2 (regression) (definitions can be found in IOCCG, 2006; Varunan and Shanmugam, 2015). A good model should have errors close to zero and correlation and slope coefficients close to unity. 3.

Results and discussion

3.1.

A hybrid model

9

ACCEPTED MANUSCRIPT

The absorption of light by CDOM is characterized by smoothly varying spectral dependence with amplitude tending to decrease from high values in the UV-blue region to near-zero values in the red-near infrared region. This spectral behavior is approximately described by a negative exponential function

aCDOM (λ ) = aCDOM (λi ).e − S (λ −λi ) and also by a hyperbolic function (Twardowski et al., 2004),

(1)

−γ

(2)

SC

 λ   aCDOM (λ ) = a CDOM (λ 412 ).  λ 412 

RI PT

(Jerlov, 1976; Bricaud et al., 1981),

where aCDOM (λi) is the absorption coefficient at a reference wavelength λi (usually at short wavelengths,

M AN U

e.g., 412 nm, because of the higher magnitude of aCDOM), and S and γ are the slopes of the exponential and hyperbolic functions respectively. Both the amplitude and spectral slope of aCDOM depend on the composition of the CDOM pool and these in turn depend on a variety of source and sink processes (Babin et al., 2008; Nelson et al., 1998). The slope value (S) was reported to vary from 0.01 to 0.042 and γ was assumed as 6.92 (Twardowski et al., 2004). Note that these slopes may vary depending on the spectral range of regression and CDOM composition and variation. For semi-analytical algorithms that produce absorption coefficient adg (detrital and gelbstoff CDOM) rather than the aCDOM, S was set as 0.015 in order

TE D

to reduce bias (Blough and Del Vecchio, 2002; Shanmugam et al., 2010; Dong et al., 2013; Zhu et al., 2014). However, understanding the impact of these slope parameters on the determination of aCDOM(λ), Shanmugam (2011a) derived empirical methods based on the in-situ data to determine S and γ values for a diverse range of coastal and oceanic waters. His predictive equations are expressed as −0.9677

a (412)  γ = 2.9332 ×  CDOM   aCDOM (350) 

−0.7506

AC C

EP

a (412)  S = 0.0058 ×  CDOM   aCDOM (350) 

(3)

(4)

In Eqs. 3 and 4, the ratio of aCDOM at two wavelengths (caution exercised to limit the lower boundary at 350 nm because of the influence of absorption by other compounds as discussed in a later paragraph) can be related to CDOM molecular weight (MW) and to photochemically induced shifts in MW (Helms et al., 2008). These empirical equations produce a greater range of variability in S (0.006~0.042) and γ (3.6~14) falling within the range of spectral slopes (S) already reported in the literature (see Table 1 in Twardowski et al. 2004)). Of course, there are factors (heterogeneity between different studies in the aCDOM spectral 10

ACCEPTED MANUSCRIPT

band used to fit CDOM spectra as well as methodological differences) that affect the S, but the variability of this spectral slope is reasonably large although not sufficient to fully capture a high degree of in-situ variability in the composition and content of CDOM (i.e., origin and its sensitivity to environmental forcing such as photobleaching and bacterial degradation processes) (Nelson et al., 1998, Stedmon and

RI PT

Markager, (2001)). This implies that accurate prediction of the shape of the CDOM curves may require a rigorous understanding of the spectral character of absorption by CDOM and its influencing factors. To accurately detect significant changes in CDOM variability within coastal regions dominated by terrestrial inputs, a new exponential model was introduced with an additional slope parameter γ

SC

CDOM composition (Shanmugam, 2011a). This model is reproduced as

aCDOM (λ ) = aCDOM (λi ).e ( − S (λ −λi )−γ

γo = γ

o

o

a cdom (350) − (1 / γ ) a cdom (350) + (1 / γ )

)

related to the

(5)

is expressed as

M AN U

where aCDOM (λi ) is the absorption of CDOM at 350 nm and the γ

o

o

(6)

is a parameter rather similar to “S”, but more sensitive to the content and composition of CDOM

(Shanmugam, 2011a; Stedmon and Markager, 2001). While the new exponential equation was

TE D

successfully applied to typical oceanic and coastal waters with relatively low CDOM contents (Shanmugam, 2011a), its utility to make predictions of CDOM across a range of varying water quality conditions in complex inland (fresh and brackish water conditions) and coastal water (conservative mixing) environments with high CDOM contents with diverse spatial and temporal variability is still

EP

limited. This could be attributed to the fact that in these complex coastal environments, the autochthonous origin of CDOM (due to phytoplankton, aquatic macrophytes and periphyton) often dominates the

AC C

allochthonous origin of CDOM (largely of humic substances of terrestrial plant origin driven by surface runoff) with more optically active and higher humic acids. These components make productive coastal and inland water bodies more intensely colored (Thurman, 1985; Zhang et al., 2011) and result in CDOM absorption spectra with a combination of a steep slope below 420 nm and a gentle slope beyond 420 nm (Laanen, 2007; Helms et al., 2008). CDOM from such waters can vary significantly in quantity and o

quality and often represents up to 75% of DOC (Thurman, 1985). This implies that the slopes S and γ of the above exponential function that approximately describe the shape of the CDOM curve are not adequate (Fig. 2(a-c)) to predict its magnitude caused by the molecular properties of the humic substances present in these water bodies.

11

ACCEPTED MANUSCRIPT

CDOM has strong influence in the UV-blue region (overlaps with the absorption of chlorophyll and nonalgal particles) and lacks strong distinguishable spectral features like the absorption peaks and trough of chlorophyll. In this study, the modified (Shanmugam (2011a)) form of the single exponential (Jerlov,

RI PT

1976) model is further transformed into a hybrid model with an additional wavelength-chlorophylldependent term ε (λ ) to describe the general shape (quasi-exponential) and amplitude of CDOM spectra. The hybrid model takes the form,

[

(

)

]

(7)

SC

a CDOM (λ ) = a CDOM (λi ) × exp − S × (λ − λi ) − ε × γ o × 0.0001 × (λ − λi )

where aCDOM (λi ) is the CDOM absorption preferred at a reference wavelength of 412nm ( λi = 412 ).

M AN U

ε (λ ) is the parameter (similar to the exponential slope) accounting for the optically significant and higher humic acids plus the residual dissolved material (i.e., the autochthonous origin of CDOM) in inland and productive water bodies. It is defined as a function of the chlorophyll concentration (C) and wavelength (λ) as follows,

ε (λ ) = (−0.01× C ) + β (λ )

(8)

TE D

where β (λ ) = 86.46 × exp(− 0.0058 × λ )

ε (λ ) and β (λ ) give a unique solution for accurate prediction of CDOM variability in turbid productive inland and coastal waters. As shown in Fig. 2(d) and (e), β is the spectral function closely representing

EP

CDOM’s spectral properties, while ε (λ ) is the wavelength-dependent β and amplitude term determined by the chlorophyll concentration. In many productive coastal and lagoon waters, phytoplankton and

AC C

CDOM often exhibit spatial and seasonal trends (will be discussed in a later section), and decomposition and degradation of phytoplankton turn out to be an important source of CDOM (Zhang et al., 2009). Thus,

ε (λ ) is derived by taking into account the combined effect of chlorophyll and humic acid so that it is sensitive to changes in the composition of the CDOM pool originating from production and degradation of phytoplankton and macrophytes as well as from conservative mixing of different water masses (Fig. 2(e)). ε (λ ) is generally higher and nearly constant ( ε (350) = 11-11.4 for chlorophyll values 0.3 – 25 mg m-3) for coastal and open-ocean waters (low chlorophyll cases), however it is highly variable in productive coastal and lagoon waters ( ε (350) = 4.89-10.3 for chlorophyll values 109.37-647.56 mg m-3) where the autochthonous origin of CDOM with relatively higher molecular weight of humic acids (also 12

ACCEPTED MANUSCRIPT

residual fulvic acids from humic acids of lower molecular weight) is the main source resulting from the production and degradation of phytoplankton, aquatic macrophytes and periphyton (Laanen, 2007 and references therein; Menken et al., 2006; Helms et al., 2008). With these new terms, the hybrid exponential model can be effective in terms of adequately describing the UV-visible absorption spectrum of CDOM

RI PT

in both inland and marine waters. Such an exponential function can be used to fit to the measured CDOM absorption spectra in order to reduce measurement noise (in clear waters) and scattering errors due to small particles (still present in the sample after filtration of samples from sediment-dominated coastal waters) (Laanen, 2007; Shanmugam, 2011a). Reducing these measurement noises and scattering error can in turn improve the accuracy of CDOM absorption spectrum and thus enlarge the accuracy of bio-optical

SC

modeling studies.

It should be mentioned that since the above model is applicable to a wide range of inland and marine

M AN U

environments, it is preferred to denote the CDOM absorption as aCDOM instead of a , a cdm or a g throughout this work. Further, caution must be exercised when modeling the CDOM absorption at wavelengths below 350 nm where compounds other than CDOM in the dissolved fraction of seawater, such as oxygen, bromide, nitrate, and sea-salts begin to absorb intensely (Shifrin, 1988). 3.2.

Estimation of CDOM, DOC and salinity

Several researchers have demonstrated the possibility of estimating CDOM from remote sensing data

TE D

with various levels of success as was done globally for chlorophyll-a (Siegel et al., 2002; Bailey and Werdell, 2006; Mannino et al., 2008; Shanmugam, 2011a; also see Table 1). From these studies, it becomes evident that CDOM products can be derived from satellite data through empirical methods (relating CDOM absorption coefficients and reflectance band ratios) and such products can be further

EP

used to derive other important products such as DOC and salinity in coastal oceanic waters, where CDOM and DOC behave conservatively and have strong correlations with salinity (Mannino et al., 2008;

AC C

Del Castillo and Miller, 2008). Thus, CDOM-based algorithms that incorporate the relationships between CDOM, DOC, and salinity in such waters are highly desired. To predict the spectral behavior of the CDOM absorption using a hybrid exponential model as described in the previous section, CDOM absorption coefficients at two key wavelengths (i.e., 350 and 412) and chlorophyll-a concentration must be estimated using remote sensing data. Though many algorithms developed in the past (Zhu et al., 2014; Tiwari and Shanmugam, 2011, Shanmugam, 2011 and references therein) require as input Rrs at several wavelengths across the visible spectrum (e.g., 410, 440, 490, 555, and 667 nm), we choose satellitedetecting bands for CDOM absorption as 490, 620, 670, and 713 nm to better describe its variability in both inland and marine water environments. It should be noted that algorithms based on the ratio of reflectance in the blue-green spectral region (e.g., 410, 440, 490, 555 nm) may yield valid retrievals at 13

ACCEPTED MANUSCRIPT

low CDOM and pigment contents (e.g., aCDOM (412) = < 1 m-1; chlorophyll-a = < 5 mg m-3), especially in coastal and open-ocean waters where concentrations of CDOM, non-algal particles, and chlorophyll-a are often closely correlated and reflectance in the above spectral region is more sensitive to these constituents. In turbid productive coastal and inland waters, concentrations of CDOM, chlorophyll-a and non-algal

RI PT

particles are usually very high and often uncorrelated, and have strong overlapping absorption features in the blue region (Gitelson et al., 2000; Menken et al., 2006), which makes the blue reflectance very weak and less sensitive to variations in CDOM content. Furthermore, at increasingly high chlorophyll concentrations, the variation in reflectance is minimum in the visible region and maximum in the red-NIR region (Shanmugam et al., 2013). This suggests that accurate estimation of the CDOM absorption

SC

coefficient based on the conventional blue-green reflectance ratios is difficult at high chlorophyll-a levels (Menken et al., 2006). Nonetheless, Strömbeck and Pierson (2001) reported that absorbance of red light

M AN U

can be significant at high CDOM concentrations, which is well supported by this study and other studies (e.g., Helms et al., 2008). To avoid the influence of strong overlapping absorption features and confounding factors in the blue spectral region, algorithms presented here take advantage of absorption and reflectance features in the red and near infrared wavelengths. This spectral region is relatively free from the atmospheric correction issues particularly in optically complex coastal and inland waters (Singh and Shanmugam, 2014). Besides, the chosen wavelengths are within the band set of many current and future ocean color satellite sensors (e.g., OCM-3). Based on the review of the various correlations

TE D

between Rrs band ratios and aCDOM(λ) reported in the previous studies (e.g., Del Castillo et al., 1999; Mannino et al., 2008; Tiwari and Shanmugam, 2011; Griffin et al., 2011), relationships between the aCDOM(350) and aCDOM(412) and Rrs band ratios (Rrs(490)/Rrs(670), Rrs(620)/Rrs(713)) are established here for wider applicability. The nonlinear power-law curve-fitting techniques are applied to log-transformed

EP

data from the field measurements of Rrs (band ratios) and discrete determinations of aCDOM from nearsurface samples. Results show strong correlations for both inland and marine waters (Fig. 3a and b),

AC C

yielding the following equations,

if Rrs (678) > Rrs (748) (for coastal and oceanic waters)

 R (490)  aCDOM (350) = a × α × rs  Rrs (670)  

b

 R (490)  aCDOM (412) = c × α × rs  Rrs (670)   where α = 1 , a = 2.2 ; b = −1 ; c = 1 ; d = −0.95

14

(9)

d

ACCEPTED MANUSCRIPT

if Rrs (748) > Rrs (678) (for turbid productive inland waters)

 R (620)  aCDOM (412) = c × α × rs  Rrs (713)  

d

where α = 0.25 , a = 1.8756 ; b = −0.822 ; c = 0.7605 ; d = −0.82

b

(10)

RI PT

 R (620)  aCDOM (350) = a × α × rs  Rrs (713)  

SC

Since CDOM can be used as a rapid proxy for DOC in coastal-riverine systems (e.g., Griffen et al., 2011), CDOM and DOC measured from field samples of the river-influenced coastal systems were useful in deriving an empirical algorithm to map DOC. Fig. 3c shows the goodness-of-fit of the CDOM-DOC

M AN U

relationship with a high correlation (R2 = 0.81, RMSE = 0.09). This relationship indicates strong conservative behavior between these two properties for the summer and winter periods with variable mixing across the riverine-estuarine-coastal system. The empirical equation relating DOC and aCDOM(412) is statistically significant and given by,

DOC = 259 × [aCDOM (412)]

0 .55

(11)

TE D

The strong relationship (between aCDOM(350) or aCDOM(412) and DOC) is consistent with the results from the Chesapeake Bay mouth within the U.S. Middle Atlantic Bight (MAB) (Mannino et al., 2008), within the Orinoco River plume in the Caribbean Sea (Del Castillo et al., 1999), Delaware Bay plume (Del Vecchio and Blough, 2004), the Mississippi River Plume (Del Castillo and Miller, 2008; Chaichitehrani, 2012; Catalá et al., 2013) and other inland waters (Song et al., 2013; Brezonik et al., 2015). The strong

EP

correlation suggests that the algorithm presented here is valid for similar coastal riverine-estuarine systems and coastal inland waters where CDOM distributions are driven primarily by freshwater

AC C

discharge and in-situ biological processes representing up to 75% of DOC (Thurman, 1985). Currently salinity retrievals are enabled based on microwave sensing from the Aquarius/SAC-D and Soil Moisture and Ocean Salinity (SMOS) missions at spatial scales of 30-100 km much larger than required for resolving the rapidly changing dynamics of coastal waters and river plumes (Vandermeulen et al., 2014). In near-coastal regions, salinity retrievals are severely compromised because of the close proximity to the land. In contrast, ocean color sensors (e.g., MODIS-Aqua, MERIS, OCM2, GOCI and other new sensors) offer increased spatial and temporal coverage of coastal and estuarine regions compared to these microwave sensors. Salinity retrievals are typically achieved by empirical algorithms that make use of the strong inverse relationship of salinity and aCDOM in surface waters within continental

15

ACCEPTED MANUSCRIPT

margins (e.g., Ahn et al., 2008; Del Castillo and Miller, 2011). The major source for CDOM absorption is river flow and coastal runoff. For this study, field observation data collected from a variety of waters within coastal environments (including rivers, estuaries, lagoons, coastal and open ocean waters; i.e., Indian in-situ data; NOMAD in-situ data and other regional data sets including Del Castillo and Miller

RI PT

(2008)) were used to derive an empirical algorithm to map surface salinity. Fig. 3d demonstrates the strongest inverse relation for the entire range of salinity values (2-35 PSU) (R2 = 0.86; RMSE = 0.06),

Salinity = 38 × exp[α × aCDOM (412) × (−0.678)]

(12)

SC

where α = 1 , if Rrs (678) > Rrs (748)

α = 0.4 , if Rrs (748) > Rrs (678)

M AN U

In turbid productive coastal-lagoon waters, aCDOM(412) values are largely determined by the components associated with phytoplankton and aquatic macrophytes. This additional source for CDOM absorption introduces a seasonal variability in the observed relationship, although this variability in CDOM is found to be significantly smaller than that in phytoplankton (e.g., Boss et al., 2001). Thus, a constant factor α is introduced to account for this variability (mainly in highly turbid productive water bodies within coastal environments) and assure the stability of this relationship under these circumstances. Clearly, aCDOM(412)

TE D

decreases with increasing salinity, as oceanic waters dilute the riverine and terrestrial input. This relationship is consistent with the various previous studies in other regions (Monahan and Pybus, 1978; Boss et al., 2001; Bowers et al., (2000, 2004); Binding and Bowers, 2003; Moon et al., 2006; Ahn et al., 2008; Del Castillo and Miller, 2008; Chaichitehrani, 2012; Chaichitehrani et al., 2013; Catalá et al., 2013). Since surface salinity in estuarine, coastal and shelf waters is important for the parameterization of

EP

the dynamic physical processes that are responsible for the transport of terrestrial carbon into the ocean and thus dominating an important pathway for many biogeochemical processes (Vandermeulen et al.,

AC C

2014), CDOM-based measurements of salinity is appropriate for those coastal environments dominated by the terrestrial influences. However, caution must be exercised when using the above relationship which is empirical and cannot be generalized to a universal relationship because of certain physical and environmental reasons (see Boss et al., 2001). 3.3.

Evaluation of the hybrid model for CDOM

In an effort to evaluate the relative performance of our hybrid model (NM), three other models – commonly used in various bio-optical modeling studies (i.e., single exponential model with variable slopes (EVS) and with a constant slope (ECS) (Siegel et al., 2002; IOCCG, 2006; Dong et al., 2013; Zhu et al., 2014) and hyperbolic model with a constant slope (HCS) (Twardowski et al., 2004; Shanmugam, 16

ACCEPTED MANUSCRIPT

2011a) – were chosen for this comparison of aCDOM(λ) in the VU and visible wavelengths (350-650 nm). The ECS and HCS models were applied with typical slope values (S = 0.015; γ = 6.92) as recommended by Blough and Del Vecchio (2002) and Twardowski et al. (2004), while the EVS model was applied with flexible slopes of Shanmugam (2011a). In total, four different outputs from these models were evaluated

RI PT

based on a diverse set of UV-visible absorption spectra from different locations and trophic conditions (inland, estuarine, coastal, shelf and open ocean conditions), in order to determine the most suitable model which describes aCDOM(λ) and its natural variability and composition in those water bodies. Statistical measures and criteria were used to determine the potential of chosen models and this was done using two in-situ data sets (NOMAD in-situ data from coastal and open ocean waters, and Indian in-situ

SC

data from coastal and productive waters, N = 938). Since the NOMAD in-situ data set does not contain aCDOM(350), we converted aCDOM(412) to aCDOM(350) assuming its absorption characterized by smoothly

M AN U

varying spectral dependence with amplitude, using the samples collected from Indian waters, by plotting aCDOM (350) versus aCDOM (412). The conversion formula is derived through the following relationship,

aCDOM (350) = 2.4 × [a CDOM (412)]

0.99

(13)

This relationship yielded almost negligible errors with R2 value of 0.99, and hence it was used to derive aCDOM(350) values only for the NOMAD data. The new hybrid model that was applied to the in-situ data based on the inputs of the reference aCDOM (350) and aCDOM (412) values and chlorophyll concentration

TE D

from in-situ measurements provided remarkably good results, with the predictions closely matching with in-situ measurements across the entire visible wavelengths (Fig. 4). To avoid confusion, results of the other models are not shown on these scatter plots.

Fig. 5 shows the results of statistical significance (MRE, RMSE, Bias at UV and visible wavelengths) for

EP

the chosen models (slope and R2 values are close to unity with a little difference for different models and hence not shown here for brevity). All the models, excepting the hyperbolic model with a constant slope

AC C

(HCS), exhibit significant deviations at longer wavelengths although displaying nearly consistent trends on the aCDOM error plots. The HCS model yielded the highest errors especially at UV and red wavelengths (The UV bands are important here as they will help improve the separation of chlorophyll and non-algal absorption, Yoder et al. (2011)), thus it is ranked four in this analysis. The ECS model provided slightly improved results over the HCS model in fitting the CDOM spectra, ranking 3rd out of the four cases, whereas the EVS model is the second best performer as its errors are relatively small. It should be noted that the single or variable slopes associated with HCS, ECS and EVS cannot fully elucidate the CDOM variability in a wide variety of waters within coastal and inland water environments (also refer Fig. 2a). On the contrary, the hybrid model outperformed the exponential and hyperbolic models and yielded least

17

ACCEPTED MANUSCRIPT

errors across the spectral range and water types considered (MRE = 0.032 (3.2%); RMSE = 0.026 (2.6%); Bias = 0.0001; Intercept = -0.004; Slope = 0.99; R2 = 0.96 regardless of the wavelength). Though the EVS model appears as an effective method of describing the CDOM spectra, knowledge of its spectral slope alone is insufficient to quantify and characterize the CDOM pool (varying spatially and seasonally

RI PT

depending on the locations), and to trace changes resulting from the production and removal of CDOM as well as the mixing of different pools (e.g., CDOM of terrestrial and marine origin) (Stedmon and Markager, 2001; Shanmugam, 2011a). On the contrary, the hybrid model simultaneously derives multiple parameters including S, γ°, ε (λ ) and β (λ ) to characterize CDOM’s optical properties within the continental margin and open oceans. Since these parameters are derived based on the global in-situ data

SC

sets from clear, turbid, and turbid productive waters, they should be applicable to similar inland and marine environments. However, one should ensure the reliability of these products when derived from

M AN U

satellite imagery over the optically complex coastal environments, where the standard atmospheric correction algorithm mostly fails to retrieve probable water-leaving radiances (Singh and Shanmugam, 2014) and the standard bio-optical algorithms often deliver inaccurate chlorophyll-a products (Shanmugam, 2011b). 3.4.

Validation of CDOM, DOC and salinity retrievals

The performance of the algorithms of CDOM and salinity was also assessed based on the in-situ data

TE D

from different regional waters. Prior to this analysis, the CDOM absorption coefficients at 350 and 412 nm predicted by the empirical equations (Eqs. 9 and 10) were evaluated (Fig. 6). This validation analysis demonstrated a close correspondence between the estimated aCDOM (350) and aCDOM (412) and independent in-situ data (Table 4). When including all the available data (those already used in deriving

EP

the relationships of CDOM versus Rrs) the results were further improved. CDOM absorption coefficients at 350 and 412 nm (using Eqs. 9 and 10) and chlorophyll-a (using the ABI

AC C

(Algal Bloom Index) algorithm of Shanmugam (2011b) with modified parameters applicable for both marine and inland water applications) were used as input for the hybrid model to output aCDOM(λ) values. The model-derived aCDOM(λ) values were compared with in-situ observations, which were obtained from turbid productive lagoon waters and coastal and open ocean waters around Chennai and Point Calimere on the southeast part of India. Fig. 7 shows the scatter plots of the model-derived aCDOM(λ) and in-situ aCDOM(λ) at some key wavebands (412, 443, 489, 510, 555, and 670 nm) (independent data shown in color and the dependent NOMAD and Indian in-situ data in black). This validation demonstrated a close correspondence between the model-derived and in-situ aCDOM values regardless of the water types and regions. Statistical analyses performed on these data showed small errors and high slope and determination coefficients for the independent in-situ data (Table 4). The one-to-one correspondence with 18

ACCEPTED MANUSCRIPT

significantly small errors across the entire visible region suggests that the new model is capable of producing accurate aCDOM(λ) values for remote sensing and bio-optical modeling studies. Similarly, the performance of the salinity algorithm was assessed based on the independent in-situ data from riverine, estuarine, coastal and oceanic waters (Fig. 8). Interestingly, there was one-to-one

RI PT

correspondence between estimated and in-situ observations resulting in significantly low errors (MRE = 0.016; RMSE = 0.048; Bias = 0.024; Intercept = 0.12; Slope = 0.90; R2 = 0.89). These results indicate that the inverse relationship can be useful for estimating surface salinity in coastal and estuarine environments.

SC

Considering the wide range of waters and large sample size, the overall uncertainty associated with CDOM and salinity estimations (in terms of mean relative error) was found to be within ±5%, which is relatively low when compared with that recommended for the remote sensing retrieval of chlorophyll-a in

3.5.

Satellite application and validation

M AN U

open ocean waters (Bailey and Werdell, 2006; Hooker et al., 2007).

To further examine the potential of our CDOM model and empirical algorithms, they were tested on some HICO images over a range of varying water quality conditions and complex brackishwater and marine ecosystems around Chennai (on the southeast part of India) during 22 November 2013, 16 December 2013, 22 March 2014, 18 April 2014 and 28 July 2014. All HICO images were processed using SeaDAS

TE D

(version 7.1) with the new aerosol and sunglint schemes. The water-leaving radiance spectra retrieved from HICO data over Muttukadu lagoon waters using the new atmospheric correction algorithm (Singh and Shanmugam, 2014) were physically realistic and comparable with measured data (Singh and Shanmugam, 2016). For the typical marine waters off Chennai, we followed the satellite validation

EP

protocols described by Bailey and Werdell (2006) but analyzed the 3× 3 pixels centered on the field stations due to the heterogeneity in coastal waters. This procedure was however not possible to be

AC C

adopted for Muttukadu lagoon waters, where there was greater spatial heterogeneity in spatial features and some water pixels were affected by the adjacency effect due to the reflection from contiguous pixels (Sterckx et al., 2011).

The present algorithms were applied to the HICO images of coastal and lagoon waters of Chennai and Muttukadu in order to derive the products such as aCDOM(350), γ°, salinity, DOC and chlorophyll-a. To increase the confidence level, HICO-derived estimates of aCDOM(λ) and salinity were compared with field observations (Fig. 9). Note that there were only a few match-ups available for validating these products, because obtaining coincident data sets of HICO and field measurements of large sample size was a challenge due to limited HICO scenes for this region. Nevertheless, our validation match-ups covered a

19

ACCEPTED MANUSCRIPT

wide range of conditions within the inland and coastal regions of Chennai and Muttukadu representing the informative measure of algorithm performance. As expected, the validation (Fig. 9) match-ups of CDOM showed good agreement between HICO-derived and measured aCDOM(λ) values (e.g., error of less than 20%; bias = 0.07; intercept = 0.05; slope = 0.99; R2 = 0.89 – average statistics of six key wavelengths).

RI PT

The salinity validation match-ups also showed excellent agreement between HICO retrievals and field observations with significantly small errors (e.g., errors of less than 8%, bias = 0.03; intercept = 0.19; slope = 0.86; R2 = 0.94). The overall mean relative percentage errors for these products are 20% and 8% respectively which are well within the benchmark for a validated uncertainty of ±35% endorsed for the remote sensing retrieval of chlorophyll-a in oceanic waters (Bailey and Werdell, 2006; Hooker et al.,

SC

2007; Mannino et al., 2008). Slight discrepancies observed between the HICO retrievals and in-situ data could arise from the errors associated with the in-water algorithms, sub-pixel heterogeneity, band-to-

M AN U

band differences between in situ and HICO sensor (particularly in lagoon waters), spatial and temporal differences between in-situ and HICO observations (caused due to heterogeneity in regional features and water sampling methods) (Mannino et al., 2008). HICO-derived DOC concentration was not validated in the absence of in-situ data, but this product is considered reliable because of the observed strong correlations between DOC and CDOM, CDOM and salinity, and salinity and DOC (D’Sa et al., 2002; Del Vecchio and Blough, 2004; D’Sa, 2008; Chaichitehrani, 2012; Mannino et al., 2008; Matsuoka et al., 2012). Thus, good agreement between HICO-derived CDOM and salinity and field observations support

TE D

the validity of DOC products for this region.

Fig. 10 displays the spatial and temporal distributions of chlorophyll-a, aCDOM(350), γ°, salinity and DOC, wherein CDOM and DOC covary with salinity spatially and temporally due to diverse inputs and losses (including in-situ primary production, anthropogenic and terrigenous inputs, removal processes and

EP

exchange of fresh/brackishwater and seawater). These HICO products demonstrate much greater information on the spatial and temporal variabilities of CDOM, DOC, and salinity within the Muttukadu

AC C

lagoon and its extension in coastal waters of the Bay of Bengal; i.e., high CDOM and DOC levels associated with low salinity (5-17 PSU) within the Muttukadu lagoon, relatively low CDOM and DOC levels with high salinity (20-30 PSU, caused by mixing processes) in its coastal areas, and very low CDOM with typical seawater salinity (34-35PSU) in offshore waters of the Bay of Bengal.

The

composition-related parameter γ° (Fig. 2(b and c)) is close to minus one (-1) within the Muttukadu lagoon, which indicates that CDOM is mainly of the autochthonous origin resulting from phytoplankton and other aquatic macrophytes and partly of allochthonous origin. High correlations among CDOM, DOC and salinity and field measurements obtained at several locations indicated a conservative mixing or dilution behavior for the derived properties. Similar processes and variability were previously reported in

20

ACCEPTED MANUSCRIPT

the Chesapeake Bay mouth and nearby inner-shelf regions (Mannino et al., 2008), Orinoco River plume in the Caribbean Sea (Del Castillo et al., 1999), Delaware Bay mouth to the Sargasso Sea (Vodacek et al., 1997), and Mississippi River and Atchafalaya River plumes (D’Sa, 2008; Chaichitehrani, 2012). Our analyses of the HICO products display strong seasonal variability in chlorophyll-a, CDOM, DOC

RI PT

and salinity in the Muttukadu lagoon and its coastal region. In general, their gradients are well consistent with field observations during November 2013 to July 2014. Higher levels of CDOM and DOC along the coast coincided with periods of elevated freshwater outflow from the lagoon during the winter monsoon (September to November 2013). The alongshore and offshore transport of the CDOM and DOC is most

SC

pronounced during November and December 2013, which demonstrates that aCDOM is a useful tracer of terrigenous DOM in the continental margin. The gradient of low to high salinity from the Muttukadu lagoon and its coastal area to the open sea is also well consistent with CDOM and DOC distributions

M AN U

(high to low) as supported by field observations. On the other hand, aCDOM(350) substantially decreased within the Muttukadu lagoon and its coastal areas from November 2013 to March 2014 (except April 2014, the period influenced by intermittent precipitations; HICO image of July 2014 not shown here because of cloud cover), which could be due to the partial closure of the mouth of the Muttukadu lagoon and sunlight-induced photooxidation and degradation processes commonly reported in coastal and open ocean waters (Vodacek et al., 1997; Nelson et al., 1998). Substantial seasonal and spatial variability in

TE D

CDOM, DOC and salinity products from similar environments – with higher CDOM and DOC values and lower salinity – were previously reported in other studies (e.g., Boss et al., 2001; Del Vecchio and Blough, 2004; Mannino et al., 2008; Del Castillo and Miller, 2008). 3.6.

CDOM dynamics within the Muttukadu lagoon and its coastal system

EP

To further understand the local production and degradation processes within the Muttukadu lagoon system, remote sensing reflectance and absorption coefficients of CDOM (aCDOM), phytoplankton (aph) and detrital particles (ad) were measured at several stations within lagoon during 10 November 2013, 16

AC C

December 2013, 18 April 2014, and 27 July 2014 (Fig. 11). Our field observations showed considerable variations regarding the shape and magnitude of reflectance due to their optically related water quality characteristics between November 2013 and July 2014. In particular, the shape of the Rrs spectra remained consistent with a trough at 443 nm and a peak around 560 nm, but its magnitude varied greatly in the redNIR region (670-850 nm) between these two periods. Because all optically significant constituents (dissolved and suspended particulate matters) contribute to reflectance in the blue-green region, reflectance is less sensitive to phytoplankton density (Gitelson et al., 2000). In contrast, the NIR reflectance generally decreased with decreasing phytoplankton biomass and increasing CDOM content in November 2013 and reached the highest magnitude similar to land vegetation (with a red-NIR shift, 21

ACCEPTED MANUSCRIPT

Shanmugam et al., 2013) attributable to high algal biomass in July 2014. This trend is obvious during November 2013 to July 2014. Higher CDOM and lower particulate matter in November 2013 apparently decreased reflectance at NIR wavelengths. These in-situ data also exhibited some unique features (beyond the green spectral region) on the observed reflectance: a trough near 625 nm and 675 nm and a distinctive

RI PT

peak in the NIR (>700 nm). Notable depressions in these spectra (particularly the depth of troughs) are caused by the pigment absorption of phycocyanin around 620–630 nm and chlorophyll around 670-675 nm, which increase with higher pigment concentration (Gitelson et al., 2000; Shanmugam et al., 2013). The reflectance peak around 650 nm represents phycocyanin fluorescence and beyond 700 nm the combined effect of high backscattering and minimal absorption of phytoplankton (Mishra et al., 2009;

SC

Shanmugam et al., 2013). Similar features were evident on the HICO reflectance (not shown for brevity). Field observations of a significant seasonal cycle in the CDOM and phytoplankton distributions in

M AN U

Muttukadu lagoon waters further showed interesting patterns in aCDOM(443), aph(443) and ad(443), with a lag observed between the phytoplankton and CDOM at all seven sampling stations (Muttukadu lagoon waters) during November 2013 to July 2014 (Fig. 12). Note that aCDOM(443) was generally elevated during these periods, but its maxima followed the aph(443) between November 2013 and July 2014 with lag times varying from eight to twelve weeks for the winter maxima in aCDOM(443) and summer maxima in aph(443). As previously observed on HICO images, CDOM was noticeably enhanced within the deep

TE D

channels of Muttukadu lagoon in November 2013 because of the increased runoff input caused by the monsoonal precipitation. As summer progressed, the mouth of Muttukadu lagoon was partially closed resulting in lower salinity within the lagoon, which provided favorable conditions for the growth of phytoplankton and macrophytes. CDOM content considerably decreased during this period because of possible bleaching and photodegradation as UV light intensity increased in a seasonal intensification,

EP

although intermittent precipitations tended to influence the observed patterns (e.g., the image of 18 April 2014 showed slightly enhanced CDOM contents within the deep channels of Muttukadu lagoon).

AC C

However, CDOM content gradually increased (see in Fig. 11) as a result of cell degradation from summer to winter (our filed observations indicated that the water color changed from green to dark brown and chocolate color). Storm water runoff flows into this lagoon during the winter monsoon (SeptemberNovember) led to the opening of its mouth and subsequent exchange of freshwater and seawater, resulting in a new salinity regime within the Muttukadu lagoon. The increase in CDOM content is partially caused by these complex land-ocean-atmospheric interaction processes. Note that there is no time lag between aph(443) and ad(443) because during stationary and decline phases in summer the majority of larger particles was phytoplankton and the absolute amount of detrital particles increased. However, a time lag in aCDOM(443) and aph(443) provides indirect evidence that a significant portion of CDOM in Muttukadu lagoon waters is derived from the development and decline of phytoplankton blooms, aquatic 22

ACCEPTED MANUSCRIPT

macrophytes and periphyton (mainly autochthonous origin). This observation is strongly supported by the previous hypothesis based on ocean color observations that revealed an apparent phase shift between marine phytoplankton and CDOM (Hu et al., 2006). Summary and conclusion

RI PT

4.

A hybrid model was developed to predict the CDOM absorption curves in a wide variety of waters within inland and marine environments. The validity of this model was assessed based on the large in-situ data sets from regional (Indian in-situ data) and global waters (NOMAD) and its relative performance was examined relative to the existing models (EVS, ECS and HCS). In-situ validation results demonstrated

SC

considerably small errors for the hybrid model (within 3.2%) compared to those for the exponential and hyperbolic models. These results indicate that spectral slopes of these models are not adequate to characterize the CDOM variability in inland and productive coastal waters. On the contrary, the hybrid

M AN U

model simultaneously derives multiple parameters (including S, γ°, ε (λ ) and β (λ ) ) for better characterization of CDOM’s optical properties in these water bodies. From this study we found that the hybrid model is a better descriptor of CDOM absorption curves, which generally decrease in a nearexponential manner from the UV to the far visible wavelengths and eventually decline to near zero between 650 and 700 nm (Stedmon and Markager, 2001; Loiselle et al., 2009). For remote sensing applications, CDOM-based algorithms that incorporate the relationships between

TE D

CDOM, DOC, and salinity were developed and thoroughly validated using in-situ data from inland, coastal and oceanic waters. These algorithms were tested on four HICO images acquired over the Muttukadu lagoon and its coastal areas during 22 November 2013, 16 December 2013, and 22 April 2014 and 18 July 2014. Validation analyses based on the in-situ data and HICO-match-ups demonstrated

EP

successful retrieval of CDOM (at wavelengths from 350-700 nm), DOC and salinity from inland lagoon and coastal waters. Both the in-situ validation and validation match-ups (HICO) yielded low errors and

AC C

high slopes and determination coefficients, indicating good performance of these algorithms. A slight deviation of the HICO-derived estimates of CDOM and salinity with in-situ observations was observed likely due to the discrepancy arising from temporal differences between field observations and HICO observations.

Spatial maps of CDOM, DOC and salinity demonstrated substantial spatial and temporal aspects of CDOM dynamics within the Muttukadu lagoon and its extension in coastal waters of the Bay of Bengal. It was found that DOC-specific CDOM absorption was typically high in Muttukadu lagoon waters and its coastal environments compared to the typical marine DOC concentrations in the Bay of Bengal. Composition-related slope parameters (S, γ°, ε (λ ) ) were observed very low within the Muttukadu lagoon,

23

ACCEPTED MANUSCRIPT

suggesting that CDOM is of terrestrial origin or the combination of both terrestrial and aquatic origin. Concurrent field observations of reflectance, chlorophyll concentration, and absorption coefficients of CDOM and phytoplankton provided strong evidence of the phase shift in the CDOM and phytoplankton absorption coefficients (CDOM peaks and troughs typically followed phytoplankton peaks and troughs by

RI PT

approximately eight to sixteen weeks during November 2013 and July 2014; note that this was based on monthly observations). These results confirm that in addition to the influence of terrestrial processes, local production and degradation of phytoplankton and aquatic macrophytes contribute significantly to the CDOM and DOC variability in the lagoon and its coastal areas.

SC

Temporal patterns in HICO-derived products further demonstrated that during transport from the inland to the sea, CDOM optical properties showed strong gradients and conservative mixing between the freshwater/brackishwater and marine end-members. Estimates of surface salinity obtained from the

M AN U

CDOM absorption coefficient were consistent with this pattern of a conservative mixing behavior of these waters. However, in open ocean environments, caution must be exercised when using the salinity-CDOM relationship for the open-ocean waters where many different factors and mechanisms (e.g. photooxidation, flocculation and sorption) affect CDOM and DOC distributions (Carder et al., 1999; Siegel et al., 2002; Najjar, 2007; Shanmugam, 2011a). One recent approach that employs a temperaturebased algorithm to remotely estimate DOC distributions in major ocean basins was already discussed by

TE D

Siegel et al. (2002). This approach is not suitable for estimating DOC distributions in the continental margins that are largely influenced by complex terrestrial effects, freshwater discharges, ecosystem processes not essentially linked to sea surface temperature distributions (Mannino et al., 2008). In these environments, the hybrid CDOM model, CDOM-DOC, and CDOM-salinity relationships derived from this study (Caution needs to be exercised regarding the universality of CDOM-DOC and CDOM Salinity

EP

relationships) will have great potential for remote sensing to quantify carbon fluxes and their export from rivers or estuaries into the coastal ocean. These empirical algorithms can be applied to both hyperspectral

AC C

and multispectral satellite data to provide retrievals of surface CDOM, DOC and salinity within continental margins, to trace changes in the CDOM pool from production and degradation mechanisms as well as from conservative mixing of different water masses, and to investigate their spatial and temporal variability in response to increasing runoff of nutrient pollution, anthropogenic activities, hydrographic changes and climate oscillations. Acknowledgments This research was supported by the Space Application Centre (SAC, ISRO), Ahmedabad under the grant (OEC1314119SACXPSHA). We would like to thank D. Rajasekhar, The Head, Vessel Management Cell (VMC), and Director of National Institute of Ocean Technology (NIOT) for providing the Coastal 24

ACCEPTED MANUSCRIPT

Research Vessels Sagar Manjusha and Sagar Paschimi to Indian Institute of Technology (IIT) Madras, Chennai, India for making various bio-optical measurements in coastal waters around Chennai and Point Calimere. We gratefully acknowledge the contributors and scientists who contributed to NOMAD and Ocean Biology Processing Group of NASA for the distribution of the NOMAD in-situ data and the

RI PT

support of the SeaDAS Software. We would like to thank the HICO project team at both OSU (Prof. Curtiss O. Davis for his support and Dr. Jasmine Nahorniak for her kind help and assistance) and NRL for targeting, processing and distribution of HICO L1B data over the Muttukadu lagoon and Chennai coastal regions. All HICO data were provided by the Naval Research Laboratory through the Oregon State University,

College

of

Earth,

Ocean,

and

Atmospheric

Sciences

HICO

website

SC

(hico.coas.oregonstate.edu). We are thankful to Prof. M Elliott, Editor, for his guidelines and suggestions, and the two anonymous reviewers for their helpful comments to improve the manuscript.

M AN U

References

Ahn, Y.H., Shanmugam, P., 2007. Derivation and analysis of the fluorescence algorithms to estimate chlorophyll a concentrations in ocean waters. Journal of Optics A: Pure and Applied Optics, 9, 352–362. Ahn, Y. H., Shanmugam, P., Moon, J. E., Ryu J. H., 2008. Satellite remote sensing of a low-salinity water plume in the East China Sea. Annales. Geophysicae, 26, 2019−2035.

Babin, M., Roesler, C. S., Cullen, J. J., 2008. Real-time coastal observing systems for marine ecosystem

France.

TE D

dynamics and harmful algal blooms. Oceanographic Methodology Series, UNESCO Publishing, Paris,

Bailey, S. W., Werdell, P. J., 2006. A multi-sensor approach for the on orbit validation of ocean color

EP

satellite data products. Remote Sensing of Environment, 102, 12–23. Binding, C. E., Bowers, D. G., 2003. Measuring the salinity of the Clyde Sea from remotely sensed ocean

AC C

color. Estuarine, Coastal and Shelf Science, 57, 605−611. Blough, N.V., Vecchio, R. D., 2002a. Chromophoric DOM in the coastal environment. In Hansell, D., Carlson C.A., (Eds.). Biogeochemistry of Marine Dissolved Organic Matter, 509–546. Blough, N. V., Vecchio, D. R., 2002b. Distribution and dynamics of chromophoric dissolved organic matter (CDOM) in the coastal environment. In: Hansell, D.A., Carlson, C.A. (Eds.), Biogeochemistry of Marine Dissolved Organic Matter, Academic Press, New York, p. 500. Boss, E., Pegau, W. S., Zaneveld, R. V., Barnard, A. H., 2001. Spatial and temporal variability of absorption by dissolved material at a continental shelf. Journal of Geophysical Research, 106, 9499−9507.

25

ACCEPTED MANUSCRIPT

Bowers, D. G., Harker, G. E. L., Smith, P. S. D., Tett, P., 2000. Optical properties of a region of freshwater influence (the Clyde Sea). Estuarine, Coastal and Shelf Science, 50, 717e726. Bowers, D. G., Evans, D., Thomas, D. N., Ellis, K., Le, P. J., Williams, B., 2004. Interpreting the color of

RI PT

an estuary. Estuarine, Coastal and Shelf Science, 59, 13-20. Bracchini, L., Dattilo, A. M., Loiselle, S. A., Cozar, A., Tognazzi, A., Azza, N., Rossi, C., 2006. The role of wetlands in the chromophoric dissolved organic matter release and its relation to aquatic ecosystems optical properties. A case of study: Katonga and Bunjako Bays (Lake Victoria, Uganda). Chemosphere, 63, 1170–1178.

SC

Brando, V. E., Dekker, A. G., 2003. Satellite hyperspectral remote sensing for estimating estuarine and coastal water quality. IEEE Transactions on Geoscience and Remote Sensing, 41, 1378–1387.

M AN U

Brezonik, P. L., Olmanson, L. G., Finlay, J. C., Bauer, M. E., 2015. Factors affecting the measurement of CDOM by remote sensing of optically complex inland waters. Remote Sensing of Environment, 199–215. Bricaud, A., Morel, A., Prieur, L., 1981. Absorption by dissolved organic matter of the sea (yellow substance) in the UV and visible domains. Limnology and Oceanography, 26, 43−53. Carder, K. L., Steward, R. G., Harvey, G. R., Ortner, P. B., 1989. Marine humic and fulvic acids: Their effects on remote sensing of ocean chlorophyll. Limnology and Oceanography, 34, 68−81.

TE D

Carder, K. L., Chen, F. R., Lee, Z. P., Hawes, S. K., Kamykowski, D., 1999. Semi-analytic MODIS algorithms for chlorophyll a and absorption with bio-optical domains based on nitrate-depletion temperatures. Journal of Geophysical Research, 104, 5403−5421.

EP

Castillo, C. E. D., Coble, P. G., Morel, J. M., Lopez, J. M., Corredor, J. E., 1999. Analysis of the optical properties of the Orinoco River plume by absorption and fluorescence spectroscopy. Marine Chemistry, 66, 35–51.

AC C

Castillo, C. E. D., Miller, R. L., 2008. On the use of ocean color remote sensing to measure the transport of dissolved organic carbon by the Mississippi River Plume. Remote Sensing of Environment, 112, 836−844.

Castillo, C. E. D., Miller, R. L., 2011. Horizontal and vertical distributions of colored dissolved organic matter during the Southern Ocean Gas Exchange Experiment. Journal of Geophysical Research, 116, C00F07, doi: 10.1029/2010JC006781. Catalá, T. S., Mladenov, N., Echevarría, F., Reche, I., 2013. Positive trends between salinity and chromophoric and fluorescent dissolved organic matter in a seasonally inverse estuary. Estuarine, Coastal and Shelf Science, 133, 206-216. 26

ACCEPTED MANUSCRIPT

Chaichitehrani, N., 2012. Investigation of colored dissolved organic matter and dissolved organic carbon using combination of ocean color data and numerical model in the Northern Gulf of Mexico. Ph.D Thesis, Department of Oceanography and Coastal Sciences, Louisiana State University, USA. Chaichitehrani, N. C., D’Sa, E. J., Osburn, C. L., Bianchi, T. S., Schaeffer, B. A., 2013. Chromophoric

RI PT

Dissolved Organic Matter and Dissolved Organic Carbon from Sea-Viewing Wide Field-of-View Sensor (SeaWiFS), Moderate Resolution Imaging Spectroradiometer (MODIS) and MERIS Sensors: Case Study for the Northern Gulf of Mexico. Remote Sensing, 5, 1439-ar1464, doi:10.3390/rs5031439.

Dev, P. J., Shanmugam, P., 2014. A new theory and its application to remove the effect of surface-

SC

reflected light in above-surface radiance data from clear and turbid waters. Journal of Quantitative Spectroscopy & Radiative Transfer, 142, 75–92, doi:10.1016/j.jqsrt.2014.03.021.

M AN U

Dong, Q., Shang, S., Lee, Z., 2013. An algorithm to retrieve absorption coefficient of chromophoric dissolved organic matter from ocean color. Remote Sensing of Environment, 128, 259–267. D’Sa, E. J. D., Hu, C., Muller-Karger, F. E., Carder, K. L., 2002. Estimation of colored dissolved organic matter and salinity fields in case 2 waters using SeaWiFS: Examples from Florida Bay and Florida Shelf, Proceedings in Indian Academy of Science (Earth Planetary Science), 111, 197–207. D’Sa, E. J., 2008. Colored dissolved organic matter in coastal waters influenced by the Atchafalaya River,

TE D

USA: effects of an algal bloom. Journal of Applied Remote Sensing, 2, 023502. Dupont, N., Aksnes, D. L., 2013. Centennial changes in water clarity of the Baltic Sea and the North Sea. Estuarine, Coastal and Shelf Science, 131, 282-289.

Ferrari, G. M., 2000. The relationship between chromophoric dissolved organic matter and dissolved

EP

organic carbon in the European Atlantic coastal area and in the west Mediterranean Sea (Gulf of Lions). Marine Chemistry, 70, 339−357.

AC C

Ficek, D., Zapadka, T., Dera, J., 2011. Remote sensing reflectance of Pomeranian lakes and the Baltic. Oceanologia, 53, 959–970.

Gitelson, A. A., Yacobi, Y. Z., Schalles, J. F., Rundquist, D. C., Han, L., Stark, R., 2000. Remote estimation of phytoplankton density in productive waters. Arch. Hydrobiologia - Special Issues Advancements in Limnology, 55, 121–136. Griffin, C. G., Frey, K. E., Rogan, J., Holmes, R. M., 2011. Spatial and inter-annual variability of dissolved organic matter in the Kolyma River, East Siberia, observed using satellite imagery. Journal of Geophysical Research - Bioscience, 116, 12.

27

ACCEPTED MANUSCRIPT

Hansell, D. A., Carlson, C. A., 2002. Biochemistry of Marine Dissolved Organic Matter, Elsevier Science, Academic Press. Helms, J. R., Stubbins, A., Ritchie, J. D., Minor, E. C., 2008. Absorption spectral slopes and slope ratios

Limnology and Oceanography, 53(3), 955–969.

RI PT

as indicators of molecular weight, source and photobleaching of chromophoric dissolved organic matter.

Hestir, E. L., Brando, V., Campbell, C., Dekker, A., Malthus, T., 2015. The relationship between dissolved organic matter absorption and dissolved organic carbon in reservoirs along a temperate to tropical gradient. Remote Sensing of Environment, 156, 395–402.

SC

Hooker, S.B., McClain, C.R., Mannino, A., 2007. NASA Strategic Planning Document: A Comprehensive Plan for the Long-Term Calibration and Validation of Oceanic Biogeochemical Satellite

M AN U

Data. NASA Special Pub, 2007-214152, 31 pp., NASA Goddard Space Flight Cen., Greenbelt, Md. Hu, C., Lee, Z., Muller-Karger, F. E., Carder, K. L., Walsh, J. J., 2006. Ocean Color Reveals Phase Shift Between Marine Plants and Yellow Substance. IEEE Geoscience and Remote Sensing Letters, 3(2), 262266.

Jeffrey, S.W., Humphrey, G.F., 1975. New Spectrophotometric Equations for Determining Chlorophylls a, b, c1 and c2 in Higher Plants, Algae and Natural Phytoplankton. Biochemical Physiology, 167, 191-194.

TE D

Jerlov, N. G., 1976. Optical oceanography, New York: Elsevier. Kahru, M., Mitchell, B. G., 2001. Seasonal and non-seasonal variability of satellite derived chlorophyll and colored dissolved organic matter concentration in the California Current. Journal of Geophysical

EP

Research, 106, 2517−2529.

Kutser, T., Pierson, D. C., Kallio, K. Y., Reinart, A., Sobek, S., 2005. Mapping lake CDOM by satellite remote sensing. Remote Sensing of Environment, 94, 535–540.

AC C

Laanen, M., 2007. Improving the remote sensing of Colored Dissolved Organic Matter in inland freshwaters, Ph.D thesis, University of Amsterdam, The Netherlands. Le, C., Hu, C., 2013. A hybrid approach to estimate chromophoric dissolved organic matter in turbid estuaries from satellite measurements: A case study for Tampa Bay. Optic Express, 21, 18849-18871. Lee, Z. P., Carder, K. L., Arnone, R., 2002. Deriving inherent optical properties from water color: A multi-band quasi-analytical algorithm for optically deep waters. Applied Optics, 41, 5755–5772. Lee, Z. P., 2006. Remote sensing of inherent optical properties: Fundamentals, tests of algorithms, and applications, IOCCG report.

28

ACCEPTED MANUSCRIPT

Loiselle, S. A., Bracchini, L., Dattilo, A. M., Ricci, M., Tognazzi, A., 2009. Optical characterization of chromophoric dissolved organic matter using wavelength distribution of absorption spectral slopes. Limnology and Oceanography, 54, 590−597. Mannino, A., Russ, M. E., Hooker, S. B., 2008. Algorithm development and validation for satellite-

RI PT

derived distributions of DOC and CDOM in the U.S. Middle Atlantic Bight. Journal of Geophysical Research, 113, C07051, doi: 10.1029/2007JC004493.

Mantoura, R. F. C., Woodward, M. S., 1983. Conservative behavior of riverine dissolved organic carbon in the Severn Estuary: Chemical and geochemical implications. Geochim. Cosmochim. Acta., 47, 1293-

SC

1309.

Matsuoka, A., Bricaud, A., Benner, R., Para, J., Sempere, R., Prieur, L., Belanger, S., Babinm, M., 2012.

M AN U

Tracking the transport of colored dissolved organic matter in Southern Beaufort Sea waters, Canadian Arctic: relationship with hydrographical characteristics. Bioscience, 9, doi: 10.5194/bg-9-925-2012, 925940.

Menken, K. D., Brezonik, P. L., Bauer, M. E., 2006. Influence of Chlorophyll and Colored Dissolved Organic Matter (CDOM) on Lake Reflectance Spectra: Implications for Measuring Lake Properties by Remote Sensing, Lake and Reservoir Management, 22 (3), 179-190.

TE D

Miller, R. L., Belz, M., Castillo, C. D., Trzaska, R., 2002. Determining CDOM absorption spectra in diverse coastal environments using a multiple path length, liquid core waveguide system. Continental Shelf Research, 22, 1301–1310, doi: 10.1016/S0278-4343(02)00009-2. Mishra, S., Mishra, D.R., Schluchter, W.M., 2009. A novel algorithm for predicting phycocyanin

EP

concentrations in Cyanobackteria: A proximal hyperspectral remote sensing approach. Remote Sensing, 1, 758-777, doi: 10.3390/rs1040758.

AC C

Mobley, C. D., 1999. Estimation of the remote-sensing reflectance from above-surface measurements. Applied Optics, 38, 7442-7455. Monahan, E., Pybus, M., 1978. Color, ultraviolet absorbance and salinity of the surface waters off the west coast of Ireland. Nature, 274, 782–784. Monteith, D. T., Stoddard, J. L., Evans, C. D., Wit, H. A., Forsius, M., Hogasen, T., Wilander, A., Skjelkvale, B. L., Jeffries, D. S., Vuorenmaa, J., Keller, B., Kopacek, J., Vesely, J., 2007. Dissolved organic carbon trends resulting from changes in atmospheric deposition chemistry. Nature, 450, 537-540.

29

ACCEPTED MANUSCRIPT

Moon, J. E., Yang, C. S., Ahn, Y. H., 2006. Salinity estimation based on inherent optical property (IOP) in the southern Yellow Sea and northern East China Sea, Proceeding in 4th JKWOC meeting, 18– 21 December 2006, Jeju National University, Jeju Island, Korea. Morel, A., Prieur, L., 1977. Analysis of variation in ocean color. Limnology and Oceanography, 22,

RI PT

709−722.

Morel, A., Gentili, B., 2009. A simple band ratio technique to quantify the colored dissolved and detrital organic material from ocean color remotely sensed data. Remote Sensing of Environment, 113, 998−1011.

SC

Najjar, R. G., 2007. Impact of circulation on export production, dissolved organic matter, and dissolved oxygen in the ocean: Results from Phase II of the Ocean Carbon-cycle Model Inter-comparison Project

M AN U

(OCMIP-2). Global Biogeochemical Cycles, 21, GB3007, doi: 10.1029/2006GB002857. Nelson, N. B., Siegel, D. A., Michaels, A. F., 1998. Seasonal dynamics of colored dissolved material in the Sargasso Sea. Deep-Sea Research I, 45, 931−957.

Romera-Castillo, C., Sarmento, H., A´lvarez-Salgado, X. A., Gasol, J. M., Marrase, C., 2010. Production of chromophoric dissolved organic matter by marine phytoplankton. Limnology and Oceanography, 55(1), 446–454.

TE D

Roulet, N., Moore, T. R., 2006. Browning the waters. Environmental chemistry, 444, 283-284. Schwarz, J. N., Kowalczuk, P., Cota, G. F., Mitchell, B. G., Kahru, M., Chavez, F. P., Cunningham, A., McKee, D., Gege, P., Kishino, M., Phinney, D.A., Raine, R., 2002. Two models for absorption by colored

EP

dissolved organic matter (CDOM). Oceanologia, 44, 209−241. Shanmugam, P., Ahn, Y. H., 2007. New atmospheric correction technique to retrieve the ocean color from SeaWiFS imagery in complex coastal waters. Journal of Optics A: Pure and Applied Optics, 9(5),

AC C

511–530, doi:10.1088/1464-4258/9/5/016. Shanmugam, P., Ahn, Y. H., Ryu, J. H., Sundarabalan, V. B., 2010. An evaluation of inversion models for retrieval of inherent optical properties from ocean color in coastal and open seawaters around Korea. Journal of Oceanography, 66(6), 815–830. Shanmugam, P., 2011a. New models for retrieving and partitioning colored dissolved organic matter in the global oceans. Implications for remote sensing. Remote Sensing of Environment, 115, 1501–1521. Shanmugam, P., 2011b. A new bio-optical algorithm for the remote sensing of algal blooms in complex ocean waters. Journal of Geophysical Research, 116(C4), C04016, doi: 10.1029/2010JC006796.

30

ACCEPTED MANUSCRIPT

Shanmugam, P., 2012. CAAS: An atmospheric correction algorithm for the remote sensing of complex waters. Annales Geophysicae, 30(1), 203–220. Shanmugam, P., Suresh, M., Sundarabalan, B., 2013. OSABT: An Innovative Algorithm to Detect and Characterize Ocean Surface Algal Blooms, IEEE Journal of Selected Topics in Applied Earth

RI PT

Observations and Remote Sensing, 6(4), 1879–1892, doi: 10.1109/JSTARS.2012.2227993.

Shifrin, K. S., 1988. Physical Optics of Ocean Water American Institute of Physics, College Park, MD, 285.

Siegel, D. A., Maritorena, S., Nelson, N. B., 2002. Global distribution and dynamics of colored dissolved detrital

organic

materials.

Journal

of

Geophysical

Research,

107(C12),

SC

and

10.1029/2001JC000965.

r3228,

doi:

M AN U

Singh, R. K., Shanmugam, P., 2014a. A novel method for estimation of aerosol radiance and its extrapolation in the atmospheric correction of satellite data over optically complex oceanic waters. Remote Sensing of Environment, 142, 188–206, doi:10.1016/j.rse.2013.12.001. Singh, R. K., Shanmugam, P., 2014b. A robust algorithm for minimizing glint effects in satellite ocean color imagery. Ocean Science. Discussion, 10, 1-20, doi: 10.5194/os-10-1-2014. Singh, R.K., Shanmugam, P., 2016. A multidisciplinary remote sensing ocean colour sensor: Analysis of

TE D

user needs and recommendations for future developments. In Press: IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing.

Smart, P. L., Finalyson, B. L., Rylands, W. D., Ball, C. M., 1976. The relation of fluorescence to the

EP

dissolved organic carbon in surface waters. Water Research, 10, 805-811. Song, K. S., Zang, S. Y., Zhao, Y., Li, L., Du, J., Zhang, N. N., Wang, X. D., Shao, T. T., Guan, Y., Liu, L., 2013. Spatiotemporal characterization of dissolved carbon for inland waters in semi-humid/semi-arid

AC C

region, China. Hydrology and Earth System Sciences, 17, 4269–4281, 10.5194/hess-17-4269-2013. Stedmon, C. A., Markager, S., 2001. The optics of chromophoric dissolved organic matter (CDOM) in the Greenland Sea: an algorithm for differentiation between marine and terrestrially derived organic matter. Limnology and Oceanography, 46, 2087-2093. Steinberg, C. E. W., 2003. Ecology of Humic Substances in Freshwater, 1st Ed, Springer. Sterckx, S., Knaeps, E., Ruddick, K., 2011. Detection and correction of adjacency effects in hyperspectral airborne data of coastal and inland waters: the use of the near infrared similarity spectrum. International Journal of Remote Sensing, 32(21). 6479-6505, doi: 10.1080/01431161.2010.512930.

31

ACCEPTED MANUSCRIPT

Strömbeck, N., Pierson, D. C., 2001. The effects of variability in the inherent optical properties on estimations of chlorophyll a by remote sensing in Swedish freshwaters. Science of Total Environment, 268, 123-137.

516 p. http://dx.doi.org/10.1007/978-94-009-5095-5.

RI PT

Thurman, E. M., 1985. Organic geochemistry of natural waters. Netherlands: Martinus Nijhoff, Dr. Junk.

Tiwari, S. P., Shanmugam, P., 2011. An optical model for the remote sensing of colored dissolved organic matter in coastal/ocean waters. Estuarine, Coastal and Shelf Science, 93, 396-402.

Twardowski, M. S., Boss, E., Sullivan, J. M., Donaghay, P. L., 2004. Modeling spectral absorption by

SC

chromophoric dissolved organic matter (CDOM). Marine Chemistry, 89, 69−88.

Vandermeulen, R. A., Arnone, R., Ladner, S. Martinolich, P., 2014. Estimating sea surface salinity in

M AN U

coastal waters of the Gulf of Mexico using visible channels on SNPP-VIIRS. Proceedings of SPIE 0277786X, V. 9111.

Varunan, T., Shanmugam, P., 2015. A model for estimating size-fractioned phytoplankton absorptioncoefficients in coastal and oceanic waters from satellite data. Remote Sensing of Environment, 158, 235-254.

Vecchio, D. R., Blough, N. V., 2004. On the origin of the optical properties of humic substances.

TE D

Environmental Science and Technology, 38, 3885−3891. Vodacek, A., Blough, N. V., DeGrandpre, M. D., Peltzer, E. T., Nelson, R. K., 1997. Seasonal variation of CDOM and DOC in the Middle Atlantic Bight: Terrestrial inputs and photo-oxidation. Limnology and

EP

Oceanography, 42, 674−686.

Werdell, P. J., Bailey, S. W., 2005. An improved in-situ bio-optical dataset for ocean color algorithm development and satellite data product validation. Remote Sensing of Environment, 98(1), 122–140.

AC C

Wiley, J.D., Atkinson. L. P., 1982. Natural fluorescence as a tracer for distinguishing between Piedmont and Coastal Plain River water in the nearshore waters of Georgia and North Carolina. Estuarine, Coastal and Shelf Science, 14, 29-59.

Yoder, J. A., Antoine, D., Castillo, C.E.D., Evans, R.H., Mobley, C., Sarmiento, J.L., Sathendranath, S., Schueler, C.F., Siegel, D.A., Wilson, C., 2011. Assessing requirements for sustained ocean color research and operations, National Research Council of the National Academics, The National Academics Press, 500 Fifth Street. N.W. Washington, DC.

32

ACCEPTED MANUSCRIPT

Zhang, Y., van Dijk, M. A., Liu, M., Zhu, G., Qin, B., 2009. The contribution of phytoplankton degradation to chromophoric dissolved organic matter (CDOM) in eutrophic shallow lakes: field and experimental evidence. Water Research, 43(18), 4685-4697, doi: 10.1016/j.watres.2009.07.024. Zhang, Y., Zhang, E., Yin, Y., van Dijk, M. A., Feng, L., Shi, Z., Liu, M., Qin, B., 2010. Characteristics

RI PT

and sources of chromophoric dissolved organic matter in lakes of the Yungui Plateau, China, differing in trophic state and altitude. Limnology and Oceanography, 55(6), 2645–2659.

Zhang, Y., Yin, Y., Feng, L., Zhu, G., Shi, Z., Liu, X., Zhang, Y., 2011. Characterizing chromophoric dissolved organic matter in Lake Tianmuhu and its catchment basin using excitation-emission matrix

SC

fluorescence and parallel factor analysis. Water Research, 45, 5110-5122. PMid: 21840562.

Zhu, W. N., Yu. Q., 2013. Inversion of chromophoric dissolved organic matter (CDOM) from EO-1

M AN U

Hyperion imagery for turbid estuarine and coastal waters. IEEE Transactions on Geoscience and Remote Sensing, 51, 3286–3298.Zhu, W., Yu, Q., Tian, Y. Q., Becker, B. L., Zheng, T. Carrick, H. J., 2014. An assessment of remote sensing algorithms for colored dissolved organic matter in complex freshwater

AC C

EP

TE D

environments. Remote Sensing of Environment, 140, 766–778.

33

ACCEPTED MANUSCRIPT

Table. 1. Summary of the methods using the relationships of the CDOM or DM (combined detrital and dissolved) absorption coefficients and remote sensing reflectance ratios. S. No

CDOM or DG vs Rrs ratio

References

RI PT

aDG(443) vs Rrs(412)/ Rrs(551) and Rrs(443)/ Rrs(551);

Carder et al. (1999)

1. aDG(443) vs Rrs(443)/ Rrs(551),

SC

Rrs(488)/ Rrs(551) and Rrs(667)/ Rrs(551).

M AN U

aCDOM(412) vs Rrs(412)/ Rrs (510), 2.

D’Sa and Miller (2003)

Rrs (443)/ Rrs(510), and Rrs(510)/ Rrs(555)

aCDOM(300) vs Rrs(443)/ Rrs(510)

Kahru and Mitchell (2001)

TE D

3.

aCDOM(420) vs Band 2 (525–605 nm) and Band 3 (630–690 nm) ratio 4.

Kutser et al. (2005)

EP

from ALI (Advanced land

AC C

Imager)

5.

aCDOM(440) vs Rrs(490)/ Rrs(555)

Mannino et al. (2008)

aCDOM(400), vs Rrs(412)/ Rrs(555) 6.

Ahn et al. (2008) aCDOM(412)

34

ACCEPTED MANUSCRIPT

aCDOM at specified wavelength (400, 412, 440, 490, and 555 nm) 7.

Morel and Gentili (2009)

RI PT

vs Rrs(412)/ Rrs(443) and Rrs(490)/ Rrs(555)

SC

aCDOM(440) vs Rrs(570)/ Rrs(655) 8.

Ficek et al. (2011)

M AN U

(for inland waters)

aCDOM (350), aCDOM (412) vs Rrs(443)/ Rrs(555)

aCDOM (400) vs TM3(630–690nm)+

Shanmugam (2011a)

Griffin et al. (2011)

TM2(520–600 nm)/ TM1(450–520 nm)

AC C

EP

10.

TE D

9.

Table 2. In-situ data used for deriving and validating the CDOM model. For better clarity, sampled waters are categorized into four types: CW – clear waters (Chl-a: 0.15-0.4 mg m-3; SS: 0.45-0.85 g m-3), MTW – moderately turbid waters (Chl-a: 0.41-3.5 mg m-3; SS: 0.86-5.6 g m-3), TW – turbid waters (Chl-a: 0.86-2 mg m-3; SS: 13-35 g m-3), and TPW – turbid productive waters (lagoon, Chl-a: 109.36-1356.92 mg m-3;

Nature of

aCDOM

Min/

Rrs Chl-a

Max

350

412

443

489

510

555

35

670

489

670

620

715

# of

Model

Samples

SS: 4-30g m-3).

ACCEPTED MANUSCRIPT

waters

nm

nm

nm

nm

nm

nm

nm

nm

nm

-

-

(sr-1) -

-

-

0.0003

0.00005

(mg m-3) -

-

-

4.584

1.92

-

-

-

-

-

0.063

0.02

-

-

-

min

8.11

3.04

-

-

-

-

-

-

-

0.002

0.005

-

max

22.64

8.62

-

-

-

-

-

-

-

0.02

0.05

-

min

0.089

0.051

-

-

-

-

-

0.0025

0.0002

-

-

-

0.568

-

-

0.19

0.052

-

-

1.27

0.813

-

-

15.78

5.32

-

-

17.47

6.39

-

min

0.024

0.009

0.006

0.003

max

1.738

0.722

0.403

min

8.11

3.043

1.849

22.64

min

0.19

-

-

0.02

45

26

0.01

-

-

-

-

-

-

0.002

0.0002

-

-

-

-

-

-

0.05

0.046

-

-

-

-

-

-

-

-

0.004

0.041

-

-

35

5

-

-

-

-

0.011

0.032

-

0.002

0.001

0.00007

0.0004

0.00005

-

-

0.025

0.175

0.119

0.052

0.011

0.01

0.001

-

-

6.303

0.957

0.74

0.448

0.075

-

-

0.002

0.005

109.36

124

45

8.623

5.285

2.883

2.241

1.296

0.309

-

-

0.022

0.055

1281.53

0.05

0.01

0.001

0.034

0.006

0.003

0.002

0.0002

-

-

0.352

AC C

max

-

-

SC

1.172

M AN U

max

RI PT

max

TE D

CW/MTW

0.009

nm

565

TW NOMAD TPW dataset TPW TW CW/MTW TPW

0.024

EP

dataset

NOMAD

min

TPW

(Fig. 3a and b) (Fig. 7, black scatter points) scatter points)

Validation of aCDOM (λ) (Fig. 7, color

Validation of aCDOM (λ)

Validation of aCDOM (350, 412) (Fig. 6)

aCDOM (350, 412) vs Rrs

(m-1)

nm

35

max

1.27

0.813

0.663

0.49

0.426

0.297

0.066

0.05

0.04

-

-

3.36

min

0.089

0.05

0.03

0.01

0.013

0.006

0.0009

0.002

0.0003

-

-

0.015

max

1.172

0.568

0.476

0.354

0.309

0.227

0.057

0.02

0.009

-

-

47.35

min

15.78

5.327

3.149

1.496

1.091

0.568

0.08

-

-

0.004

0.032

1100.32

max

17.4

6.39

3.86

1.92

1.425

0.761

0.123

-

-

0.011

0.041

1356.92

26

5

36

ACCEPTED MANUSCRIPT

RI PT

Table 3. In-situ data used for deriving and validating DOC and salinity products. CW – clear waters, MTW – moderately turbid waters, TW – turbid waters, and TPW – turbid productive waters (lagoon).

Min -1

[µmolCL-1]

[PSU]

0.035

43.75

-

open sea waters from

(Fig. 3c)

different regions

Estuarine, coastal and

different regions Salinity model

4.94

735.7

-

min

0.027

-

2.6

max

3.72

-

36.8

min

0.009

-

25.49

max

0.938

-

36.83

Min

2.32

-

13.2

max

3.81

-

17.7

min

0.017

-

26.08

EP

155

AC C

TPW

15

CW/MTW

Salinity model validation

TPW

Samples

177

NOMAD dataset

(Fig. 3d)

# of

460

max

TE D

open sea waters from

M AN U

model

min

Regional waters

Estuarine, coastal and

Salinity

(412) [m ] Max

DOC

DOC

aCDOM

/

Nature of waters

SC

Model

40 max

0.435

-

34.59

min

2.17

-

12.9

max

4.05

-

19.9

(Fig. 8)

37

6

ACCEPTED MANUSCRIPT

Table 4. Statistical comparisons between the model results and in-situ data (retrievals of aCDOM(λ) from remote sensing reflectance data in marine and inland waters).

Wavelength

BIAS

RMSE

R2

SLOPE

INTERCEPT

350

0.249

0.063

0.237

0.898

0.919

0.037

412

-0.018

-0.011

0.219

0.896

0.968

-0.031

412

-0.064

-0.043

0.317

0.874

0.991

-0.006

443

-0.050

-0.043

0.312

0.874

0.991

-0.004

489

-0.028

-0.031

0.324

0.860

0.991

0.013

data

510

-0.017

-0.021

0.334

0.852

0.991

0.034

(Fig. 7)

555

0.031

0.042

0.331

0.822

0.991

0.087

0.354

0.744

0.991

0.176

CDOM

(nm)

retrievals

data (Fig.

retrievals from Rrs

0.009

0.020

AC C

EP

670

TE D

CDOM

M AN U

6)

SC

from Rrs

RI PT

MRE

Products

38

M AN U

SC

RI PT

ACCEPTED MANUSCRIPT

TE D

Fig. 1. Map showing the sampling sites on the east coast of southern India - Chennai, Point Calimere and

AC C

EP

Muttukadu coastal lake waters.

39

ACCEPTED MANUSCRIPT

0.016

1

(a)

-0.96

(b)

0.0155

(c) -0.965

0.5

0.015

-0.97 -γo

-0.5

0.014

-0.975

RI PT

0.0145

-γo

-0.98 -1

0.0135

-0.985

0.013 0.01

0.1

1

10

100

-1.5

-1

In-situ aCDOM (350) [m ]

0

5 In-situ a

10 CDOM

15

20

16

(d)

12

8

8

5

10

In-situ a

4

M AN U

ε

β

0

(350) [m -1]

12

4

-0.99

25

SC

S

0

CDOM

15

20

25

(350) [m -1]

(e)

0

-4

0 350 400 450 500 550 600 650 700

Wavelength (nm)

-8

0

100 200

300 400

500 600

700

-3

Chl-a [mgm ]

Fig. 2(a-c). Variation of the slope S values with aCDOM (350), and γo with aCDOM (350) (black unfilled

TE D

circle represents the clear and turbid waters and green filled circle represents the eutrophic waters; this variation of γo with aCDOM (350) is shown separately in Fig. 2c for better clarity). Note that S does not always vary inversely with aCDOM (350) in productive coastal and inland waters (green circles). (d) The variation of β as a function of wavelength for inland and marine waters, (e) The effect of chlorophyll on

EP

ε at different wavelengths for a wide variety of waters. Violet to red marker indicates the key

AC C

wavelengths 350, 412, 443, 489, 510, 555 and 670 nm.

40

10 10

-1 -2 -3

10

-2

10

3

-1

0

1

10 10 ratio * R α rs

10

2

10

3

10 10

-1 -2 -3

10

-2

10

-1

0

In-situ Salinity [PSU]

2

1

10 -2 10

10

-1

In-situ a

10

0

10

-1

CDOM

(412) [m ]

10

2

10

3

(d)

30 20 10 0

1

1

10 10 ratio * R α rs

40

(c)

10

10

0

M AN U

In-situ DOC [ µmolCL -1]

10

10

(b)

1

RI PT

10

0

10

2

SC

10

(a)

1

10

(412) [m -1]

10

2

CDOM

10

In-situ a

In-situ a

CDOM

(350) [m -1]

ACCEPTED MANUSCRIPT

0

0.5

1

1.5

x * In-situ a

2

CDOM

2.5

3

3.5

4

(412) [m -1]

TE D

Fig. 3a and b. Relationships between the aCDOM(350) and aCDOM(412) and remote sensing reflectance ratios (Rrs(490)/Rrs(670) and Rrs(620)/Rrs(713)) from the in-situ data (NOMAD and Indian in-situ data sets) (N= 610; R2 = 0.95 for aCDOM(412) and R2 = 0.94 for aCDOM(350)). (c) Relationship between the dissolved organic carbon (DOC) and aCDOM(412) measured from different riverine systems, estuaries and

EP

coastal regions (N = 460; R2 = 0.81). (d) Relationship between the salinity and aCDOM(412) measured from different riverine systems, estuaries and coastal regions (NOMAD, Indian and other regional in-situ data)

AC C

(N = 347; R2 = 0.86).

41

(443) [m -1] CDOM

-3

10

-3

10

-2

10

10 10 10 10

-4

10

10 10 10 10 10

-2

10

CDOM

-1

10

0

(489) [m -1]

0

-1

-2

-3

-4 -4

10

-3

10

10

CDOM

10

-1

10

10 10 10 10

1

0

-1

-2

10

-3

10

10

0

10

1

(555) [m -1]

10 10 10 10 10 10 10

-1

10

0

10

1

-1

CDOM

(443) [m ]

-4

10

EP

In-situ a

-2

-2

-4

10

-3

10

In-situ a

1

10

10

1

(670) [m -1]

(555) [m -1]

10

-3

In-situ a

-3

10

-3

10

-2

-3

10

-2

(412) [m ]

-1

In-situ a

CDOM

CDOM

0

-4

10

-1

-1

1

10

Model a

10

2

(510) [m -1]

10

10

1

CDOM

10

10

0

Model a

Model a

CDOM

(489) [m -1]

In-situ a

-1

10

0

RI PT

10

-2

10

1

SC

10

-1

Model a

10

0

10

M AN U

10

1

CDOM

10

2

Model a

10

TE D

Model a

CDOM

(412) [m -1]

ACCEPTED MANUSCRIPT

-2

10

CDOM

-1

10

0

10

1

(510) [m -1]

0

-1 -2 -3 -4 -5 -6

10

-6

10

-5

10

-4

In-situ a

10

-3

CDOM

10

-2

10

-1

10

0

(670) [m -1]

NOMAD data set

Clear/moderately turbid waters of Chennai

Turbid waters – off Point Calimere

AC C

Turbid productive waters

Fig. 4. Scatter plots showing the comparison between measured and estimated values of aCDOM(λ) from the hybrid model of this study (N = 938). Note that the modeled aCDOM(λ) values are based on the inputs of the reference aCDOM(350) and aCDOM(412) values from the in-situ measurements. Since aCDOM(350) values were not available in the NOMAD dataset, it was derived using the relationship of aCDOM(350) and aCDOM(412) determined on water samples collected from clear, turbid and eutrophic waters. The comparison results for other models are not shown here for clarity, but their statistical values are shown in Fig. 5 for comparison purposes.

42

EP

TE D

M AN U

SC

RI PT

ACCEPTED MANUSCRIPT

Fig. 5. Results of the statistical analyses performed on the in-situ aCDOM(λ) and model aCDOM(λ) data from

AC C

marine and inland waters. NM – new model (hybrid model of this study), EVS – exponential model with varying slopes, ECS – exponential model with a constant slope, HCS – hyperbolic model with a constant slope.

43

ACCEPTED MANUSCRIPT

10

(412) [m -1]

0

-1

-2

10

-2

10

-1

CDOM

10

1

10

2

(350) [m -1]

10 10 10 10

1

0

-1

-2

-3

10

-3

10

-2

10

-1

In-situ a

CDOM

10

0

10

1

10

2

-1

(412) [m ]

M AN U

In-situ a

10

0

10

RI PT

10

1

2

SC

10

10

CDOM

10

2

Model a

Model a

CDOM

(350) [m -1]

10

TE D

Fig. 6. Comparison between the estimated and measured values of aCDOM(350) and aCDOM(412) for the independent in-situ data obtained from Indian waters (total N = 610 considering all the in-situ data; color scatter points indicate independent data N=66; R2 = 0.92 for aCDOM(350), R2=0.89 for aCDOM(412)). Dark filled and open circles indicate the data used for deriving the model parameters. These data are

AC C

and aCDOM(412).

EP

superimposed on these plots to show the trend and consistency of the models in estimating aCDOM(350)

44

-3

10

-3

10

-2

10

10 10 10 10

-1

-2

-3 -3

10

-2

10

10 10

10

1

10

0

0

-1

-2

10

-4

10

10

-3

10

In-situ a

-2

CDOM

10 10

RI PT

-1

-2

-3 -3

10

10

-1

10

10

-2

10

10 10 10 10

0

10

1

10 10 10 10

1

10

2

(443) [m -1]

-3 -3

10

-2

10

-1

CDOM

10

0

10

1

(510) [m -1]

0

-1

-2

-3

-4

-5 -5

10

-4

10

In-situ a

45

10

-2

10

(555) [m -1]

CDOM

0

-1

In-situ a

10

10

0

10

10

-1

1

(489) [m -1]

-3

-4

10

0

10

1

EP

10

-1

CDOM

AC C

(555) [m -1] CDOM

Model a

10

2

10

1

In-situ a

0

In-situ a

10

(412) [m -1]

1

10

10

CDOM

10

1

(510) [m -1]

10

10

0

CDOM

Model a

CDOM

(489) [m -1]

In-situ a

-1

10

2

SC

(443) [m -1] CDOM

Model a

10

-2

10

M AN U

10

-1

Model a

10

0

(670) [m -1]

10

1

CDOM

10

2

Model a

10

TE D

Model a

CDOM

(412) [m -1]

ACCEPTED MANUSCRIPT

-3

10

-2

10 -1

CDOM

(670) [m ]

-1

10

0

ACCEPTED MANUSCRIPT

Fig. 7. Comparisons of the modeled aCDOM(λ) with measured aCDOM(λ) values from clear, turbid and turbid productive waters (color scatter points indicate independent data, N=66). Dark filled and open circles indicate the data used for deriving the model parameters (i.e., NOMAD and Indian in-situ data sets, N = 169). These data are superimposed on these plots to show the trend and consistency of the models in

RI PT

estimating aCDOM(λ).

Model Salinity [PSU]

40

SC

30 20

0

0

10

20 30 In-situ Salinity [PSU]

40

Clear/moderately turbid waters of Chennai Turbid productive waters

TE D

NOMAD data set Regional waters

M AN U

10

Fig. 8. Comparison of the estimated and measured salinity using independent in-situ data from Indian waters (color scatter points indicate independent data N = 46, R2 = 0.89). Dark filled squares and open circles indicate the data used for deriving the model parameters (i.e., NOMAD in-situ data and those from

EP

other regional waters, N = 169). These data are superimposed on these plots to show the trend and

AC C

consistency of the models in estimating surface salinity.

46

10 10 10

1

0

-1

-2

-3

10

30

20

10

0

-3

10

-2

10

-1

In-situ a

10

0

10 -1

CDOM

[m ]

1

SC

10

40 Model Salinity [PSU]

10

2

M AN U

Model a

CDOM

[m -1]

10

RI PT

ACCEPTED MANUSCRIPT

10

2

0

10 20 30 In-situ Salinity [PSU]

40

TE D

Clear/moderately turbid waters of Chennai Turbid productive waters

EP

Fig. 9. Validation match-ups comparing HICO-derived aCDOM(λ) and salinity products with the corresponding in-situ measurements from coastal waters of Chennai and Muttukadu lagoon waters (N =

AC C

30 for aCDOM(λ); N = 21 for salinity).

47

ACCEPTED MANUSCRIPT

22 Nov 2013Muttukadu

16 Dec 2013

22 Mar 2014

18 Apr 2014

Chl-a

Chl-a

Chl-a

aCDOM (350)

aCDOM (350)

aCDOM (350)

γo

AC C

EP

γo

TE D

M AN U

aCDOM (350)

SC

RI PT

Chl-a

lagoon

48

γo

γo

ACCEPTED MANUSCRIPT

16 Dec 2013

22 Mar 2014

Salinity

Salinity

Salinity

18 Apr 2014 Salinity

DOC

DOC

DOC

TE D

M AN U

DOC

SC

RI PT

22 Nov 2013

EP

Fig. 10. HICO-derived images of chlorophyll-a, aCDOM(412), γ°, salinity and dissolved organic carbon (DOC) for 22 November 2013, 16 December 2013, 22 March 2014, and 18 April 2014. Some spots in the

AC C

marine region are due to the presence of clouds and should not be considered for interpretation.

49

ACCEPTED MANUSCRIPT

0.025

18

10 Nov. 2013

25

aCDOM (412): 0.8 ~ 4.14 (m-1) 15

Chl-a: 60.18 ~ 423.34 (mg m-3)

aph [m-1]

Rrs 0.01

aCDOM [m-1 ]

12

0.015

9 6

0.005

0 400 450 500 550

Wavelength (nm)

750

Wavelength (nm)

18 Apr. 2014

10

0 350 400 450 500 550 600 650 700 750

Wavelength (nm)

20

aph (443): 16.67 ~ 23.78 (m-1)

25

15

15

0.01

10

0.005

5

0 400 450 500 550

0 400 450 500 550 600 650 700 750 800 850

750

27 Jul. 2014

30 25

Rrs

aph [m-1]

TE D

aCDOM [m-1 ]

35

0.03

600 650 700

0.02

20 15 10

0.01

Chl-a: 331.11 ~ 555.1 (mg m-3)

10

5

0 350 400 450 500 550 600 650 700 750

Wavelength (nm)

Wavelength (nm) 0.05

0.04

M AN U

0.015

aCDOM [m-1 ]

20

aph [m-1 ]

Rrs

0.02

aCDOM (412): 5.87 ~ 6.54 (m-1)

SC

0.025

600 650 700

30

0.03

15

5

3

0 400 450 500 550 600 650 700 750 800 850

aph (443): 2.25 ~ 15.49 (m-1)

20

RI PT

0.02

EP

Wavelength (nm)

0 400 450 500 550

15 12

aCDOM (412): 5.32 ~ 7.04 (m-1) aph (443): 16.59 ~ 27.1 (m-1) Chl-a: 843.1 ~ 1356.92 (mg m-3)

9 6 3

5

0 400 450 500 550 600 650 700 750 800 850

Wavelength (nm) 18

600 650 700

750

0 350 400 450 500 550 600 650 700 750

Wavelength (nm)

Wavelength (nm)

AC C

Fig. 11. In-situ data of the remote sensing reflectance (Rrs, first column) and absorption coefficients of phytoplankton (aph, second column) and CDOM (aCDOM, third column) measured in Muttukadu lagoon waters on 10 November 2013, 18 April 2014, and 27 July 2014. These data show the growth stages of phytoplankton bloom as the NIR reflectance tends to increase with the increasing chlorophyll-a concentration from November 2013 to July 2014. Elevated levels of CDOM in November 2013 resulted from the degradation of phytoplankton and runoff inputs (from monsoonal rain).

50

ACCEPTED MANUSCRIPT

1.2

Phytoplankton

Jul.2014

Apr.2014

Apr.2014

Dec.2013

Jul.2014

Nov.2013

Apr.2014

Jul.2014

Apr.2014

Dec.2013

M AN U

Jul.2014

Nov.2013

Apr.2014

Dec.2013

Jul.2014

Nov.2013

Apr.2014

Dec.2013

Jul.2014

Nov.2013

Apr.2014

Nov.2013

Dec.2013

0

Dec.2013

0.2

SC

0.4

Dec.2013

0.6

Nov.2013

RI PT

CDOM

Jul.2014

0.8

Nov.2013

Normalized absorption

Detritus 1

Time

St-3

St-2

St-4

TE D

St-1

St-5

St-6

St-7

Fig. 12. In-situ data of the absorption coefficients of colored dissolved organic matter (aCDOM(443)), phytoplankton pigment (aph(443)) and detrital particles (ad(443)) measured in Muttukadu lagoon waters during the degradation and growth stages of phytoplankton bloom (10 November 2013, 16 December

AC C

EP

2013, 18 April 2014, and 27 July 2014). Results are shown for seven sampling sites within the lagoon.

51

ACCEPTED MANUSCRIPT

Research highlights A hybrid model is developed for CDOM characterization Its relation with DOC and salinity is investigated

RI PT

CDOM model results are validated with in-situ data Spatiotemporal dynamics of CDOM and related parameters is investigated

AC C

EP

TE D

M AN U

SC

New tools are useful in assessing CDOM dynamics in coastal and estuarine waters