Through technological and research advances, numerous methods and protocols have emerged to estimate spectral absorption of light by particles, ap, in an aquatic medium. However, the level of agreement among measurements remains elusive. We employed a multi-method approach to estimate the measurement precision of measuring optical density of particles on a filter pad using two common spectrophotometric methods, and the determination precision, or uncertainty, of the computational techniques for estimating ap for six ocean color wavelengths (412, 443, 490, 510, 555, 670 nm). The optical densities measured with the two methods exhibited a significant, positive correlation. Optical density measurement precision ranged from 0.061%-63% and exhibited a significant, positive correlation. Multi-method uncertainty ranged from 7.48%-119%. Values of ap at 555 nm and 670 nm exhibited the highest values of uncertainty. Poor performance of modeled ap compared to determined ap suggest uncertainties are propagated into bio-optical algorithms.
© 2015 Optical Society of America
The variability in the global ocean has been monitored for > 20 years by satellite-borne ocean color sensors. A number of critical uses for ocean color are of particular importance in today’s society. Remote estimates of algal biomass are essential in computations of ocean productivity and ultimately impact our characterization of the global carbon cycle . Satellite imagery has been used to monitor inter-annual variation in the timing and extent of phytoplankton blooms, which are connected to the survival of larval fish . Moreover, satellite-derived sea surface temperature (SST) and wave height information can help aquaculture developers plan where to establish new fish farms. Additionally, satellite imagery can be used to detect and monitor blooms of harmful algae and can inform monitoring efforts for delicate ecosystems, particularly in global coastal environments.
The spectrally-dependent backscattering (bb) and total absorption (a) coefficients (in units of m−1) of light by the dissolved and particulate constituents in water (the inherent optical properties, IOPs) ultimately control the color of the ocean [3–5]. The properties of this light are used to derive information about the biogeochemical components of the water column, such as the amount of carbon or the abundance of phytoplankton. Remote sensing reflectances (Rrs), the light that emerges from the ocean, are used to derive a with bio-optical algorithms and models, such as the Quasi-Analytical Algorithm , Garver-Siegel-Maritorena Algorithm  and the Generalized IOP Model , to name a few. The general relationship between Rrs and IOPs includes the effects of backscattering and can be approximated by Eq. (1) [7,9].
The total absorption, a, can separated into a set of components using Eq. (2), where aw is spectral absorption by pure water, ap is spectrally-dependent absorption by particles and ag is spectrally-dependent absorption by dissolved constituents.Eq. (3). The parameters ap and aφ can be derived analytically from using additional bio-optical models [e.g., 10,11].
Collection of in situ ap data can be accomplished in the field and in the laboratory in multiple ways [12,13]. One common method, the quantitative filter technique (QFT), was first developed by Yentsch  and later modified by Mitchell  where particles are concentrated onto a filter pad. Absorbance, or optical density of particles on a filter (ODf; typically defined on the base of the decadic logarithm), is measured with a spectrophotometer and is converted to ap using Eq. (4), where 2.303 is the natural log of 10, Afp is the clearance area of the filter pad, Vf is the filtration volume, Vf/Afp is the theoretical geometric pathlength and β is the pathlength amplification correction. The scattering of light by the filter pad and particles amplifies the photon pathlength and falsely increases the value of absorption . Therefore, β is included in the computation of ap to scale the absorption measured on the filter pad to a predicted value of absorption in suspension. Additionally, the absorbance of algal and non-algal particles can be de-convolved by measuring the samples before and after extraction of algal pigments and their corresponding coefficients can be calculated using Eq. (4) [17,18]. Direct measurements of ap in suspension can be made with a spectral absorption and attenuation meter or AC-meter (e.g. AC-S; WetLABS Inc.), or a Point Source Integrating-Cavity Absorption Meter (PSICAM) [19,20].
Three methods commonly applied to determine ap using particles collected on a filter pad include: the transmittance method (T) [15,21,22], the transmittance-reflectance method (T-R) [23,24] and the inside sphere method (IS) [25–27]. Each method requires a different geometric configuration and is susceptible to different levels of scattering error (Table 1). Theoretically, the configurations of the T-R and IS methods reduce scattering error (loss of photons scattered from the filter pad and particles) producing a more precise measurement [17,23,26,27].
Many βs have been determined empirically [e.g., 15,21], and two βs have been derived analytically [22,29]. Historically, β is the ratio of ODf to the optical density of particles in a dilute suspension (ODs), a true measurement of absorption. The source of these suspensions and filter pad samples were commonly from cultures . A power law or quadratic function and associated coefficients were calculated from a least squares regression of ODs and ODf, which can then be applied to field samples to derive a relationship. These βs do not necessarily encompass all particle sizes and types encountered in the ocean and can be subject to bias. Consequently, the β is a major source of uncertainty for determining particle absorption coefficients by the filter pad methods. Although not definitively proven, there are some indications that suggest β is influenced by particle size and scattering properties [21,25–28,30]. Other sources of analytical uncertainty have also been discussed .
In contrast, AC-meter uncertainties are in absolute terms, not percentage. The field calibration term is ~0.01 m−1, and the instrument precision is about half of the calibration term. Temperature effects can create uncertainties near the peak of pure-water absorption of 0.02 m−1. Lee  claims a ~0.005 m−1 difference in the blue part of the spectrum (350-415 nm) in pure water. By far the largest source of uncertainty is that of the scattering correction [32–34].
The satellite ocean color community uses field and laboratory measurements of absorption coefficients to develop, evaluate, and validate their remote-sensing bio-optical algorithms. To maximize the spatial and temporal extent of data included in such analyses, agencies such as NASA typically invest in a broad range of investigations and require that the data collected subsequently be permanently archived in a central database. NASA has invested in the development of rigorous quality and assurance metrics and measurement protocols to enable high quality data collection and to minimize investigator-to-investigator and instrument-to-instrument differences in the archived data. The NASA Ocean Optics Protocols for Satellite Ocean Color Validation [e.g., 13] is a standard set of protocols for radiometric and biogeochemical measurement, the purpose of which is to provide measurement consistency across ocean color missions and research groups with tenable error assessments. Moreover, strict adherence to the protocols increases the probability that biogeochemical and radiometric measurements provide sufficiently accurate measurements for satellite derived product validation. In spite of the rigorous protocols developed for the measurement and computation of ap, relative measurement uncertainties are inherent in the choices of sample analysis and ap computation within each of these protocols.
Biogeochemical algorithms and models require high quality observational ground-truth data for both development and validation. The SeaWiFS Bio-optical Archive and Storage System (SeaBASS) is a publicly available, community-driven archive of in situ bio-optical data maintained by the NASA Ocean Biology Processing Group (OBPG) at NASA Goddard [35,36]. Approximately 20,000 a and ap products derived from various field measurements can be found in SeaBASS and have been acquired multiple ways by any number of researchers. Despite the existence of community-vetted protocols and quality assurance metrics, a number of questions remain: What is the uncertainty of many researchers using different analytical methods and computational techniques for determining ap? To what degree are they different? Our goal here was to quantify precision of these measurements so that these uncertainties may be accounted for during model development and validation.
In this paper we address measurement agreement, or precision, among multiple methods for the determination of the same parameter. Three different sample sets were used to characterize uncertainties in a dynamic range of absorption from a high particle concentration with higher Chlorophyll a (Ca) concentration to lower particle concentration and lower Ca concentration. The ultimate goal was to quantify the uncertainty associated with the diverse analytical methods and computational techniques to determine ap, and to show a test case of how these uncertainties could affect model-measurement closure. Here we are identifying the lack of precision of determined ap as uncertainty. We cannot address the accuracy of ap determinations here because there is no existing suspended particle absorption standard, nor did we measure suspended particle samples in parallel with the QFT samples. Instead, we offer alternative approaches with which we endeavored to address uncertainty associated with:
- 1) Two analytical methods for collecting absorbance data (measurement precision)
- 2) Six computational techniques for determining ap (determination precision, hereafter referred to as uncertainty)
- 3) Comparison of the QFT determined ap to ap measured with an AC-S
- 4) Comparison of determined against modeled ap
Although similar method evaluations have been approached in previous studies [e.g., 26,27] in this paper we conduct a valuable performance and statistical analysis that is independent from the original method papers using real world samples.
2.1 Sample collection
Three sources of filter pad samples and AC-S measurements of ap were used in the comparison: one set collected off the coast of Hawaii (“Blue Water”) shown in Fig. 1(a), another set collected off the coast of the eastern United States (“Coastal Water”) shown in Fig. 1(b) and the third set from laboratory cultures. Table 2 shows station numbers and depths at which AC-S and filter pad samples were collected concurrently in the field. Filter pad samples were collected in duplicate by vacuum filtration (~127 mmHg) onto pre-combusted 25 mm Whatman GF/F filters. Samples were placed in HistoPrep tissue capsules, flash frozen in liquid nitrogen and stored at −80° C until analysis. Samples for Ca determination were collected similarly and stored in foil. Phytoplankton pigments were determined using high performance liquid chromatography (HPLC) following the procedures of  and further described in . Filter pad and AC-S measurements were collected in the laboratory from dilutions of three phytoplankton species: Thalassiosira weissflogii (CCMP 1387), Emiliania huxleyi (CCMP 371), and Nannochloris sp. (15 duplicate filters). The cultures were diluted with filtered and sterilized seawater to attain the following dilution series 100%, 42%, 20%, 10% and 5%. Only the three lowest dilutions (20%, 10% and 5%) of each culture were measured with the AC-S (a total of nine measurements). For each sample source, one replicate filter was analyzed using the T method and the other with the IS method. For the culture experiment, fresh filter pad samples were used for the measurements.
2.2 Instrument configurations and analysis techniques
Measurements of ODf with the T method were conducted using a Perkin Elmer Lambda 35 UV-Visible, dual beam spectrophotometer following the protocol of Roesler . Briefly, the sample and reference beams were balanced by placing a neutral density filter (ESCO optics; Density = 1.0) at the reference port. The blank and sample filters were placed in the path of the sample beam. Measurements of ODf with the IS method were conducted using a Cary 100 UV-Visible dual beam scanning spectrophotometer equipped with a 15 cm integrating sphere (Labsphere DRA-CA-30) and using the protocol of Stramski et al. [27,39]. The filters were placed in the center of the integrating sphere using a Plexiglas slide. For both methods, scans were performed between 290 and 800 nm with a 2 nm Slit Band Width (SBW), and 240 nm per minute scan speed. For both filter pad analysis methods, blank filters were moistened with 0.2 μm-filtered artificial seawater (ASW). Blank filter and air scans (for the IS method) were measured throughout the day to monitor instrument drift. For both configurations, blank filter scans were subtracted from the raw ODf spectra prior to ap computation.
A fraction of scattered light is not detected using the T method due to its geometric configuration. To reduce this scattering error, a null point correction is applied that assumes there is no absorption by particles in the near-infrared part of the spectrum [13,22]. Data from the T measurements were null-point corrected by subtracting the average absorbance between 750 nm and 800 nm from the entire spectrum.
2.3 ap determination
Three different βs were chosen for the T method and two βs were chosen for the IS method to compute ap and are listed in Table 3, although others exist in the literature. The technique for which ap was computed from each β will hereafter be referred to as a ‘computational technique’. We are not endorsing one β over another, as it is currently not possible to prove the accuracy of a given approach. To reiterate, our goal is to make a general statement about uncertainties of existing data and the most commonly used analytical methods and computational techniques.
2.4 AC-S measurements
Spectral particle and dissolved absorption (apg) was measured between 400 and 750 nm using a WETLabs AC-S instrument. In the field, the AC-S was calibrated with ultrapure water and a mean of three days’ calibrations was used to subtract the pure water offset. AC-S profiles were made at each station. For the Coastal Water measurements, the AC-S was mounted horizontally to the CTD rosette on an auxiliary ring. The mean spectra during the period of bottle firing, approximately 90 seconds, were calculated. For the Blue Water samples, the mean spectra at each depth +/− 1 m were used.
Post processing involved merging the CTD with the AC-S by time stamp, adjusting for the time required for water to move from the intake to the center of the flow tube. For all AC-S measurements, corrections for salinity, temperature, and instrument drift were made, using the methods of . For the Blue Water and Coastal Water apg measurements the proportional scattering correction [Eq. (4) in 41] was applied using an empirical ratio of absorption at 715 nm and correcting for attenuation as well as absorption. For the culture apg measurements, three dilutions of each culture (20%, 10%, and 5%) were passed through the AC-S flow tubes, and the subsequent absorption values were corrected for scattering using proportional correction .
Conventionally, ap is calculated by subtracting the absorption of 0.2 μm filtered seawater (ag) from apg, as was done for the laboratory culture experiment and the Blue Water data in this study. For the Blue Water, two AC-S meters were used: one with a 0.2 μm pre-filter and one without. In the laboratory, the same meter was used for both measurements, allowing for calibration-independent measurement of ap . During the Coastal Water cruise, a filter was not used, thus concurrent ag measurements were not performed. Instead, discrete ag samples were later measured using a bench top spectrophotometer and subtracted from apg to retrieve ap. The value of ap at 555 nm for Station 8 was removed from the analysis because it was negative, possibly an artifact of subtracting the laboratory-measured ag.
2.5 Modeling ap
In ocean optics, identification of uncertainty is important to assess instrument performance and is essential for the appropriate scaling of validation efforts for satellite-derived products. Model output is dependent upon empirical data. Therefore, a lack of precision between measurements and models could mean either the instruments, models, or both are faulty.
Our aim was to investigate if differences in ap computational techniques influenced model development and measurement-model closure using a case study approach with one candidate model. We chose a power law, Eq. (5), that estimates ap from the following relationship with chlorophyll [Eq. (1) in 10].43]. Bricaud  recommended that this model only be applied to Case 1 waters where the optical properties are determined mainly by phytoplankton and dissolved organic matter and when Ca concentration is maximally 10 mg/m3. The Coastal Water samples are not considered Case 1; however, we still applied the model as our Ca concentrations fell well within this range, although the concentration of suspended sediments is usually greater in the coastal waters.44]. Additionally, Root Mean Square Error (RMSE) was calculated as a measure of the error between the measured data and the model predictions using Eq. (7) where yi is an observation and Yi is the predicted value.
3. Results and discussion
The measurement precision of ODf and uncertainty of ap were evaluated using three different particle concentrations and types: low particle concentration natural samples (Blue Water) shown in Fig. 2(a), high particle concentration natural samples (Coastal Water) shown in Fig. 2(b), and phytoplankton-derived samples (10% dilution Emiliania huxleyi) shown in Fig. 2(c). A smoothing technique was not applied to these data. Note the differences in absorption magnitude (y axis) between the samples. Spectra measured with the T method were noisier than the spectra measured with the IS method because the geometric configuration of the T method allows for a greater loss of photons scattered from the filter pad and particles, thereby reducing the signal-to-noise ratio. The geometric configuration of the IS method favors the collection of most of the scattered photons, increasing the signal-to-noise ratio . Therefore, ODf measured with the IS method is approximately 50% greater in magnitude and less noisy than ODf measured with the T method.
3.1 Analytical uncertainty of ODf
Measurement precision of ODf measured between the T method (ODf (T)) and IS method (ODf (IS)) was evaluated by computation of the minimum (min), maximum (max), median and mean coefficient of variation (hereafter CV; standard deviation/mean*100%) of ODf (T) and ODf (S) for each data point (Table 4). Additionally, the mean bias (mean difference of ODf (T) and ODf (IS)) and mean ratio of ODf (T)/ODf (S) were computed. A null correction was applied to values of ODf (T).
The Pearson’s Linear Correlation Coefficient (r) analysis was applied to assess the relationship of ODf measured by both methods at each ocean color (OC) wavelength depending on particle concentration and sample type. The corrcoeff function in MATLAB 2014a was used for this analysis . The ODf values from both methods showed a significant (p<0.05), positive correlation at 412 nm (r = 0.8673), 443 nm (r = 0.8510), 490 nm (r = 0.8491), 510 nm (r = 0.9062), 555 nm (r = 0.9433) and 670 nm (r = 0.9014).
3.2 Uncertainty of QFT determined ap
The min, max, mean and median CV (uncertainty) for ap were calculated from all computational techniques for the QFT methods only (Table 3) at six OC wavelengths. We acknowledge that calculated uncertainties in this analysis include sample collection error without measuring this error distinctly (difference between replicates). In previous analyses, our range of replicate uncertainty for ap determined from the T and IS computational techniques typically ranged from <0.001% to 16% and 0.001% to 21%, respectively, although most values fell well below 20%.
The range of uncertainty of the QFT values of ap determined from all computational techniques was 7.07%-62.55% for all six wavelengths (Table 5). The values of ap at 555 nm showed the largest uncertainty overall with a maximum value of ~63%.
Mean ap was also calculated and, as shown in Table 6, the CV depended on absorption range and wavelength. The largest spread of ap between all computational techniques occurred at 555 nm between 0.015 and 0.05 m−1.
3.2.1 Uncertainty of QFT determined and AC-S measured ap
Mean ap as well as the min, max, mean and median CV (uncertainty) were calculated for both QFT determined and AC-S measured ap (Table 7). Here, only coincident filter pad and AC-S field measurements of ap (Table 1) and dilutions 20%, 10% and 5% of the laboratory cultures were included in the analysis. The addition of AC-S measured ap increased the uncertainty to 7.48-119% for the six wavelengths (Table 7). Mean ap was computed and at 555 nm and 670 nm showed the highest uncertainty amongst the six wavelengths at 119% and 61.38%, respectively (Table 8).
The increase in uncertainty when the AC-S data are included in the analysis may be attributed to low signal, particularly for the Blue Water samples, and/or data computation technique. We used an average absorption for each sample depth +/− 1 m. Another analytical technique may be used to improve accuracy and decrease uncertainties. A source of bias exists when comparing particle absorption acquired from a filter pad and from using an AC-meter, depending on the water type. The QFT uses nominal 0.5-0.7 μm pore size fiberglass filters  while ap is separated from a by measuring the ag from 0.2 μm filtered seawater. In this way, the 0.2-0.7 μm fraction would be measured by the AC-S but not with the QFT methods .
3.2.2 QFT and AC-S ap trends
The individual ap values from each sample were compared with the sample median to examine deviations of the method and computational differences from the median as indicated by their distance from the 1:1 line, shown in Fig. 3. The ap values determined from the IS computational techniques exhibited a smaller deviation from the median than those values determined from the T computational techniques in most cases. Values of ap from Roesler  and Mitchell  computations were consistently greater than the median across the absorption dynamic range for all wavelengths. Values of ap from the BrSB  computation were consistently lower than the median except for 443 nm where they were greater than the median at ap values > 1.0 m−1. Values of ap from Röttgers  and Stramski  were consistently similar to the median based on their close proximity to the 1:1 line.
The values of AC-S measured ap exhibited low values most noticeably in the Blue Water samples at 443 nm, which are in the lowest region of the absorption range (<0.10 m−1). Generally, AC-S measured ap compared reasonably well to the median, falling below the 1:1 line at values < 0.2 m−1 and on or above the 1:1 line at values > 0.2 m−1.
3.3 ap model evaluation
Model Skill Score, RMSE and the CV of measured and modeled ap (Tables 9–11) were calculated to evaluate measurement-model agreement for the Blue Water and Coastal Water data. One Coastal Water station was not used in this analysis because the Ca concentration exceeded the limit of 10 mg m−3 for this ap model. At 443 nm, the model performed “very good” to “good” compared to all ap computations from the Blue Water samples except for the AC-S where the model performed poorly (Table 9; SS = −0.4145). For the Coastal Water samples at 443 nm, the model performed poorly compared to all computed ap but performed good compared to AC-S measured ap (SS = 0.4247). At 555 nm, the model compared poorly to most of the Blue Water ap values except for Mitchell  and Roesler  (SS = 0.2306 and 0.6145, respectively; Table 10). The ap values for the Coastal Water samples also compared poorly except for BrSB  and the AC-S measured ap (SS = 0.3629 and 0.5715, respectively). At 670 nm, the model performed good to very good for both the Blue Water and Coastal Water ap values, except for Roesler  (SS = 0.0869; Table 11). In both water types the model performed the best in most cases when compared to ap using the BrSB  method, particularly at 443 nm and 670 nm. Generally, the model performed best when compares to ap computed from the T method.
Three different sample sets were used to characterize uncertainties in the determination of ap for a dynamic range of absorption. The ultimate goal was to estimate the measurement precision of the methods for measuring ODf, the uncertainty of ap determined from six computational techniques, and to show how these differences could affect model-measurement closure. Therefore, an understanding of measurement uncertainties is crucial and should be accounted for during model and algorithm development. Ignoring these uncertainties can result in poor model and algorithm performance. Moreover, measurement uncertainties can decrease the signal-to-noise ratio and a fraction of detected environmental variability may be lost to uncertainty. The desired result is to distinguish real variability from measurement error.
Variability in the magnitude of ap in the field may be attributed to natural variability and the systematic uncertainties that accumulate with sample collection and analysis. Minimization of the noise is key to accurately identifying changes in the optical properties of the global ocean. McKee et al.  suggest that incorporating measurement uncertainties, both random and systematic into the computation of Ca-specific absorption of marine phytoplankton may help to discriminate natural variability from measurement error. In this paper, we focused on the inherent systematic uncertainties exist amongst the various methods for determining ap. These uncertainties are important to understand and quantify as they propagate to algorithm and model development and validation.
Absorption at wavelengths in the blue-to-green region of the visible spectrum (443, 490, 510 nm) is mainly influenced by phytoplankton pigments and, therefore, may be used to derive information about phytoplankton community structure. Light absorption in the red part of the visible spectrum (555 nm and 670 nm; Table 12) is dominated by minerals and sediments. The 670 nm band also allows for the evaluation of chlorophyll fluorescence, which can be used to characterize primary production and nutrient stress of phytoplankton .
In situ measurements of ap are important for validation and development of inverse models where Rrs is used to derive IOPs a and bb. For example, the relationship between phytoplankton size and pigment composition results in an empirically derived, non-linear function of Ca and ap for use in algorithms [10,49]. These derived parameters can then be used to develop empirical bio-optical and ecosystem models. Additionally, satellite-derived values of ap and aφ are important input vectors for modeling phytoplankton size classes [9,50,51]. Characterizing phytoplankton functional types in the global ocean and understanding their role in global carbon cycling is of great interest to many researchers.
From the QFT comparisons we saw that the uncertainty of ap determined from the IS computational techniques was small and consistent across all sample types (deviating least from the median) compared to the those determined from the T computation techniques. The values of ap computed from Mitchell  and BrSB  computational methods performed consistently higher and lower, respectively, than the median absorption values. Values of ap determined from the IS analysis method showed lower uncertainty compared to the median. The uncertainty in determined ap was similar across all wavelengths. The maximum uncertainty of 119% at 555 nm across all methods could be attributed to the minimum absorption of pigments at that waveband and to the scattering differences between the reference and sample filters, which are amplified on applying the β to the absorbance spectra of the samples. Minimal absorption by phytoplankton occurs in the green region of the spectrum (555 nm). The addition of AC-S measured ap increased the uncertainty between the measurements. The large uncertainty in the measured ap would be propagated through to model development and model-derived products would carry that uncertainty.
The ODf measured by the IS method was almost 50% greater than the ODf measured by the T method. ODf from both methods exhibited high correlation at each of the OC wavelengths. The measurement uncertainty for these methods ranged from 0.061% at 412 nm to 63.55% at 670 nm, comparable range of uncertainty of ap determined from these methods. The maximum uncertainty of ap was not impacted by the inclusion of β. This work points to the sensitivity of the parameterization of β, suggesting that more work needs to be done in this area.
Output from the ap model performed best when compared to ap computed from the T method, particularly BrSB , the analytical and computational technique from which it was developed. However, most of the methods compared well with the model in the Coastal Water samples at 670 nm even though the model was not optimized for the coastal regime. In general, ap determined from both QFT and the AC-S, compared good to poorly with the model output. Empirical models such as this are typically only comparable to the water types and methods used for model development. Development of a bio-optical model or method to model ap in Case 2 waters would be advantageous.
Data collection and analysis methods change and improve over time as technology advances. Inclusion of raw data in a database, such as SeaBASS , allows for computational analysis flexibility in the future with technological advances. We now have a general idea of the estimate of the uncertainty of existing ap data. Ultimately the use of traceable standards could be advantageous for addressing the accuracy of the measurements. Additionally, the assessment of analytical precision may benefit from a round robin-like activity in the future.
This work was supported by funding from the Ocean Biology and Biogeochemistry program of the National Aeronautics and Space Administration (NNG11HP11C). This is contribution No. 5098 of the University of Maryland Center for Environmental Science. Appreciation is extended to all of the participants in the 2014 Absorption Workshop for their expertise and advice, in particular Collin Roesler, Rick Reynolds and Eurico D’Sa for their feedback on this work. Assistance and advice provided by Mike Novak and Jeremy Werdell were greatly appreciated.
References and links
1. M. J. Beherenfeld, E. Boss, D. A. Siegel, and D. M. Shea, “Carbon-based ocean productivity and phytoplankton physiology from space,” Global Biogeochem. Cycles 1006, 2299 (2005).
2. C. Wilson, J. Morales, S. Nayak, I. Asanuma, and G. Feldman, “Ocean-colour radiometry and fisheries,” in Why Ocean Colour? The Societal Benefits of Ocean-Colour Technology, (IOCCG Rep. Number 7, 2008).
3. R. W. Preisendorf, “Introduction,” in Hydrologic optics, Vol. 1 (U.S. Dept. of Commerce, 1976).
4. A. Morel and L. Prieur, “Analysis of variations in ocean color,” Limnol. Oceanogr. 22(4), 709–722 (1977). [CrossRef]
5. R. Zaneveld, A. Barnard, and Z.P. Lee, “Why are Inherent Optical Properties needed in ocean-colour remote sensing?’ in Remote Sensing of Inherent Optical Properties: Fundamentals, Tests of algorithms and Applications, (IOCCG Rep. Number 5, 2006).
6. Z. Lee, K. L. Carder, and R. A. Arnone, “Deriving inherent optical properties from water color: a multiband quasi-analytical algorithm for optically deep waters,” Appl. Opt. 41(27), 5755–5772 (2002). [CrossRef] [PubMed]
8. P. J. Werdell, B. A. Franz, S. W. Bailey, G. C. Feldman, E. Boss, V. E. Brando, M. Dowell, T. Hirata, S. J. Lavender, Z. Lee, H. Loisel, S. Maritorena, F. Mélin, T. S. Moore, T. J. Smyth, D. Antoine, E. Devred, O. H. d’Andon, and A. Mangin, “Generalized ocean color inversion model for retrieving marine inherent optical properties,” Appl. Opt. 52(10), 2019–2037 (2013). [CrossRef] [PubMed]
10. A. Bricaud, A. Morel, M. Babin, K. Allali, and H. Claustre, “Variations of light absorption by suspended particles with chlorophyll a concentration in oceanic (case 1) waters: Analysis and implications for bio-optical models,” J. Geophys. Res., C, Oceans 103(C13), 31033 (1998). [CrossRef]
11. E. Devred, S. Sathyendranath, V. Stuart, and T. Platt, “A three component classification of phytoplankton absorption spectra: Application to ocean-color data,” Remote Sens. Environ. 115(9), 2255–2266 (2011). [CrossRef]
12. J. L. Mueller, G. S. Fargion, and C. McClain, eds., Ocean Optics Protocols For Satellite Ocean Color Sensor Validation, Revision 4, Volume 4, (NASA Tech. Rep., 2003), 1–12.
13. B. G. Mitchell, M. Kahru, J. Wieland, and M. Stramska, “Determination of spectral absorption coefficients of particles, dissolved materials and phytoplankton for discrete water samples,” in Ocean Optics Protocols For Satellite Ocean Color Sensor Validation, Revision 4, Volume 4: Inherent optical properties: instruments, characterization, field measurements and data analysis protocols, (NASA Tech. Rep., 2003), 39–64.
14. C. S. Yentsch, “Measurement of visible light absorption by particulate matter in the ocean,” Limnol. Oceanogr. 7(2), 207–217 (1962). [CrossRef]
15. B. G. Mitchell, “Algorithms for determining the absorption coefficient of aquatic particulates using the quantitative filter technique (QFT),” in Ocean Optics X (1990).
16. W. L. Butler, “Absorption of light by turbid materials,” J. Opt. Soc. Am. 52(3), 292–299 (1962). [CrossRef]
17. M. N. Kishino, N. Takahashi, N. Okami, and S. Ichimura, “Estimation of the spectral absorption coefficients of phytoplankton in the sea,” Bull. Mar. Sci. 37, 634–642 (1985).
18. N. Hoepffner and S. Sathyendranath, “Bio-optical characteristics of coastal waters: Absorption spectra of phytoplankton and pigment distribution in the western North Atlantic,” Limnol. Oceanogr. 37(8), 1660–1679 (1992). [CrossRef]
19. R. Röttgers, W. Schönfeld, P.-R. Kipp, and R. Doerffer, “Practical test of a point-source integrating cavity absorption meter: the performance of different collector assemblies,” Appl. Opt. 44(26), 5549–5560 (2005). [CrossRef] [PubMed]
20. R. Röttgers, C. Häse, and R. Doerffer, “Determination of the particulate absorption of microalgae using a point-source integrating-cavity absorption meter: verification with a photometric technique, improvements for pigment bleaching and correction for chlorophyll fluorescence,” Limnol. Oceanogr. Methods 5, 1–12 (2007). [CrossRef]
21. A. Bricaud and D. Stramski, “Spectral absorption coefficients of living phytoplankton and non-algal biogenous matter: A comparison between the Peru upwelling area and the Sargasso Sea,” Limnol. Oceanogr. 35(3), 562–582 (1990). [CrossRef]
22. C. S. Roesler, “Theoretical and experimental approaches to improve accuracy of particulate absorption coefficients derived from the quantitative filter technique,” Limnol. Oceanogr. 43(7), 1649–1660 (1998). [CrossRef]
23. S. Tassan and G. M. Ferrari, “An alternative approach to absorption measurements of aquatic particles retained on filters,” Limnol. Oceanogr. 40(8), 1358–1368 (1995). [CrossRef]
24. S. Tassan and G. M. Ferrari, “A sensitivity analysis of the ‘Transmittance-Reflectance’ method for measuring light absorption by aquatic particles,” J. Plankton Res. 24(8), 757–774 (2002). [CrossRef]
25. H. Maske and H. Haardt, “Quantitative in vivo absorption spectra of phytoplankton: Detrital absorption and comparison with fluorescence excitation spectra,” Limnol. Oceanogr. 32(3), 620–633 (1987). [CrossRef]
26. R. Röttgers and S. Gehnke, “Measurement of light absorption by aquatic particles: improvement of the quantitative filter technique by use of an integrating sphere approach,” Appl. Opt. 51(9), 1336–1351 (2012). [CrossRef] [PubMed]
27. D. Stramski, R. A. Reynolds, S. Kaczmarek, J. Uitz, and G. Zheng, “Correction of pathlength amplification in the filter-pad technique for measurements of particulate absorption coefficient in the visible spectral region,” Appl. Opt. 54(22), 6763–6782 (2015). [CrossRef] [PubMed]
28. M. Babin and D. Stramski, “Light absorption by aquatic particles in near-infrared spectral region,” Limnol. Oceanogr. 47(3), 911–915 (2002). [CrossRef]
29. S. Lohrenz, “A novel theoretical approach to correct for pathlength amplification and variable sampling loading in measurements of particulate spectral absorption by the quantitative filter technique,” J. Plankton Res. 22(4), 639–657 (2000). [CrossRef]
31. D. McKee, R. Röttgers, G. Neukermans, V. S. Calzado, C. Trees, M. Ampolo-Rella, C. Neil, and A. Cunningham, “Impact of measurement uncertainties on determination of chlorophyll-specific absorption coefficient for marine phytoplankton,” J. Geophys. Res. Oceans 119, 9013–9025 (2014).
32. Z. P. Lee, J. Wei, K. Voss, M. Lewis, A. Bricaud, and Y. Huot, “Hyperspectral absorption coefficient of “pure” seawater in the 350–550 nm range inverted from remote sensing reflectance,” Appl. Opt. 54(3), 546–558 (2015). [CrossRef]
33. W. H. Slade, E. Boss, G. Dall’Olmo, M. R. Langner, J. Loftin, M. J. Behrenfeld, C. Roesler, and T. K. Westberry, “Methods for improving accuracy in measurement of spectral particulate absorption and attenuation,” J. Atmos. Ocean. Technol. 27(10), 1733–1746 (2010). [CrossRef]
34. R. Röttgers, D. McKee, and S. B. Woźniak, “Evaluation of scatter corrections for ac-9 absorption measurements in coastal waters,” Methods Oceanogr. 7, 21–39 (2013). [CrossRef]
35. P. J. Werdell, S. Bailey, G. S. Fargion, C. Pietras, K. Knobelspiesse, G. Feldman, and C. McClain, “Unique data repository facilitates ocean color satellite validation,” Eos Trans. AGU 84(38), 377–392 (2003). [CrossRef]
36. P. J. Werdell and S. W. Bailey, “An improved in-situ bio-optical data set for ocean color algorithm development and satellite data product validation,” Remote Sens. Environ. 98(1), 122–140 (2005). [CrossRef]
37. L. Van Heukelem and C. S. Thomas, “Computer-assisted high-performance liquid chromatography method development with applications to the isolation and analysis of phytoplankton pigments,” J. Chromatogr. A 910(1), 31–49 (2001). [CrossRef] [PubMed]
38. S. B. Hooker, L. V. Heukelem, C. S. Thomas, H. Claustre, J. Ras, L. Schluter, J. Perl, C. Trees, V. Stuart, E. Head, R. Barlow, H. Sessions, L. Clementson, J. Fishwick, C. Llewellyn, and J. Aiken, The Second SeaWiFS HPLC Analysis Round-Robin Experiment (SeaHARRE-2), (NASA Tech. Rep., 2003), 1–75.
39. G. Neukermans, R. A. Reynolds, and D. Stramski, “Contrasting inherent optical properties and particle characteristics between an under-ice phytoplankton bloom and open water in the Chukchi Sea,” Deep Sea Res. Part II Top. Stud. Oceanogr. 105, 59–73 (2014). [CrossRef]
40. J. M. Sullivan, M. S. Twardowski, J. R. Zaneveld, C. Moore, A. Barnard, P. L. Donaghay, and B. Rhoades, “The hyper-spectral temperature and salinity dependent absorption of pure water, salt water and heavy salt water in the visible and near-IR wavelengths (400 - 750 nm),” Appl. Opt. 45, 5294–5309 (2006). [CrossRef] [PubMed]
41. R. Röttgers, C. Dupouy, B. B. Taylor, A. Bracher, and S. B. Wozniak, “Mass-specific light absorption coefficients of natural aquatic particles in the near-infrared spectral region,” Limnol. Oceanogr. 59(5), 1449–1460 (2014). [CrossRef]
42. J. R. V. Zaneveld, J. C. Kitchen, and C. M. Moore, “The scattering error correction of reflecting-tube absorption meters,” Proc. SPIE 2258, 44–55 (1994). [CrossRef]
43. A. Bricaud, Marine Optics and Remote Sensing Lab, Laboratoire d’Oceanographie de Villefranche, Villefranche-sur-Mer, CEDEX, 06238, France (personal communication, 2013).
44. D. K. Ralston, W. R. Geyer, and J. A. Lerczak, “Structure, variability, and salt flux in a strongly forced salt wedge estuary,” J. Geophys. Res. 115, C06006 (2010).
45. MATLAB and Statistics Toolbox Release, 2014a. The MathWorks, Inc. Natick, Massachusetts, United States.
46. F. P. Chavez, K. R. Buck, R. R. Bidigare, D. M. Karl, D. Hebel, M. Latasa, L. Campbell, and J. Newton, “On the chlorophyll a retention properties of glass-fiber GF/F filters,” Limnol. Oceanogr. 40(2), 428–433 (1995). [CrossRef]
47. C. S. Roesler, Earth and Oceanographic Science Department, Bowdoin College, 120B Druckenmiller Hall,Brunswick, M.E., 04011 (personal communication, 2014).
48. NASA, “Pre-Aerosol, Clouds, and Ocean Ecosystem (PACE) Mission,” Science Definition Team Report, 2012, http://decadal.gsfc.nasa.gov/pace.html.
49. A. Chase, E. Boss, R. Zaneveld, A. Bricaud, H. Claustre, J. Ras, G. Dall’Ollmo, and T. K. Westberry, “Decomposition of in situ particulate absorption spectra,” Methods Oceanogr. 7, 110–124 (2013). [CrossRef]
50. E. Devred, S. Sathyendranath, V. Stuart, and T. Platt, “A three-component model of phytoplankton absorption spectra: application to ocean-color data,” Remote Sens. Environ. 115(9), 2255–2266 (2011). [CrossRef]
51. A. M. Ciotti and A. Bricaud, “Retrievals of a size parameter for phytoplankton and spectral light absorption by colored detrital matter from water-leaving radiances at SeaWiFS channels in a continental shelf region off Brazil,” Limnol. Oceanogr. Methods 4, 237–253 (2006). [CrossRef]