In vivo chlorophyll fluorescence (ChlF) can serve as a reasonable estimator of in situ phytoplankton biomass with the benefits of efficiently and affordably extending the global chlorophyll (Chl) data set in time and space to remote oceanic regions where routine sampling by other vessels is uncommon. However, in vivo ChlF measurements require correction for known, spurious biases relative to other measures of Chl concentration, including satellite ocean color retrievals. Spurious biases affecting in vivo ChlF measurements include biofouling, colored dissolved organic matter (CDOM) fluorescence, calibration offsets, and non-photochemical quenching (NPQ). A more evenly distributed global sampling of in vivo ChlF would provide additional confidence in estimates of uncertainty for satellite ocean color retrievals. A Saildrone semi-autonomous, ocean-going, solar- and wind-powered surface drone recently measured a variety of ocean and atmospheric parameters, including ChlF, during a 60-day deployment in mid-2018 in the California Current region. Correcting the Saildrone ChlF data for known biases, including deriving an NPQ-correction, greatly improved the agreement between the drone measurements and satellite ocean color retrievals from MODIS-Aqua and VIIRS-SNPP, highlighting that once these considerations are made, Saildrone semi-autonomous surface vehicles are a valuable, emerging data source for ocean and ecosystem monitoring.
© 2020 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
Phytoplankton are photosynthetically active microorganism that exist in varying concentrations throughout the global oceans and play a crucial role in biogeochemical processes and the carbon cycle [1–3]. Phytoplankton are the basis of the oceanic food chain and monitoring the spatiotemporal variability of phytoplankton can improve our understanding of the ocean ecosystem. While much of the global ocean has limited phytoplankton activity due to light limitation at depth and nutrient limitation near the surface, upwelling along eastern boundary currents bring nutrients to the well-lit upper ocean often resulting in a high biomass of phytoplankton, which attract fish and support coastal fisheries [1,3].
All phytoplankton contain the photosynthetic pigment chlorophyll-a (Chl), which is often used as a proxy for estimating phytoplankton biomass. The concentration of Chl can be measured in a variety of ways, both in situ and remotely. Remote sensing techniques of estimating Chl concentration leverage satellite observations of top-of-atmosphere radiances at several discrete wavelengths in the visible and near-infrared spectrums. Atmospheric correction algorithms are then applied to the observed signal to remove atmospheric and sea surface components in order to retrieve spectral remote sensing reflectance (Rrs(λ); sr−1), a normalized ratio of the reflected light by the water column . Satellite Rrs(λ) are then related via a retrieval algorithm to estimate of Chl concentrations [5,6]. In situ techniques include measuring the pigment concentration in vitro (e.g. – fluorometric, HPLC, and absorption) , measuring the in vivo absorption signal from in situ absorption meters [8,9], and measuring in vivo chlorophyll-a fluorescence (ChlF) from in situ fluorometers [10,11].
In situ collection and filtering of water samples to measure Chl concentrations are often costly and limited geographically to regions frequented by research cruises or established monitoring programs. In contrast, Chl fluorometers are relatively easy to obtain and deploy on a variety of in situ platforms, such as gliders , profiling floats , and semi-autonomous surface vehicles . In this paper, we focus on autonomous platforms, which provide high spatial and temporal sampling of the marine environment and can be deployed for a span of days to multiple weeks, enabling these platforms to obtain in situ measurements across a variety of water types and in remote regions where sustained monitoring is challenging. While ChlF is operationally easy to obtain from such platforms, there is much uncertainty in the global oceans between measured ChlF and phytoplankton biomass, which varies regionally with phytoplankton species, light intensity and history, nutrient and trace metal concentration, and temperature [1,2,10,15]. Obtaining quality measurements of Chl from these autonomous platforms is valuable not only for understanding biogeochemical processes in remote regions, such as the Southern Ocean or Bering Sea , but has potential to serve as a prime source of validation data for global ocean color satellites sensors, which routinely make retrievals of Chl concentration in these remote regions of the global oceans [17,18].
Operationally, ChlF is only useful as a proxy for phytoplankton biomass if the measured signal can be reliably related to the true concentration of the Chl pigment. However, the extent of systematic errors that can be introduced into Chl:fluorescence relationship (cellular – mentioned above, or due to the fouling of the instruments) can sometime supersede the natural variability in situ. Bio-fouling masks the natural variability of the fluorescence signal and occurs when the performance of the instrument deteriorates due to the growth of biofilm on the optical components [12,19]. The most dominant of the physiological drivers is non-photochemical quenching (NPQ) of the fluorescence signal, which occurs during daytime periods when the quantum yield of ChlF is lower than nighttime values. This mechanism offers protection to the phytoplankton’s photosynthetic apparatus from damage caused by high solar insolation [11,20]. NPQ occurs fast (O(min)) and presents in ChlF data as a period of suppressed values temporally coincident with periods of high solar insolation . NPQ is distinctly a physiological response to a physical stressor, a process independent of the instrument calibration . An adjustment, or correction, can often be derived for NPQ in ChlF measurements . Deriving and applying such a correction is important when comparing in vivo ChlF data with other types of Chl measurements, both in situ and remotely sensed, since the effects of NPQ will negatively bias the fluorescence measurements relative to other methods. This is particularly true when comparing to satellite Chl retrievals, which are measured during the daytime when the effects of NPQ are most-pronounced in the fluorescence measurements.
Saildrone Inc. builds, operates, and deploys semi-autonomous surface vehicles outfitted with fluorometers to measure ChlF. For this study, Saildrone Inc. deployed a vehicle for 60-days in the California Current region (Fig. 1) from 11 April 2018 to 11 June 2018 . This deployment’s primary objective was to assess the vehicle’s scientific capabilities . The California Current region is a complex, but well-studied, region along the west coast of the U.S. and Mexico, known for a strong north-to-south current, associated with an upwelling of deep, dense, nutrient-laden water [24,25]. This region is dominated by strong gradients of temperature, nutrients, and Chl due to the mixing of surface and upwelled waters with coastal and open-ocean waters along a narrow continental shelf . This region has economic and ecological significance and has been the subject of much research and many monitoring programs, making it a prime area to assess the Chl-measuring capabilities of the Saildrone. During this deployment, the vehicle sampled upwelling and diurnal events, sailing near moored in situ buoys and recurrent monitoring stations frequently visited by California Cooperative Oceanic Fisheries Investigation (CalCOFI) cruises.
For the 2018 Baja deployment, the Saildrone was outfitted with a standard configuration of a variety of atmospheric and oceanic environmental sensors, including a WET Labs Eco Triplet-w mounted on the vehicle’s hull at 0.25 m depth and a Teledyne Citadel CTD-NH mounted inside the hull at 0.6 m depth . The Eco Triplet-w instrument measured stimulated ChlF, colored dissolved organic matter (CDOM) fluorescence, and Volume Scattering Function, β(124°, 650 nm). The CTD measured contemporaneous water temperature and salinity. ChlF was measured by the Eco Triplet via the emission of blue light at 470 nm at and detection at 685 nm. To relate the measured fluorescence signal to Chl concentration, a manufacturer-derived calibration was applied to the instrument, which, on average, is incorrect . Alternatively, the measured fluorescence signal may be constrained through a regression against chlorophyll-a pigment concentrations derived from a high-performance liquid chromatography (HPLC) pigment analysis of concurrently collected discreet water samples [10,13]. The CalCOFI monitoring cruises routinely collect such data and collocations between CalCOFI cruises and the Saildrone are explored later in this manuscript.
A preliminary satellite-based evaluation of ChlF data collected the 2018 Saildrone deployment in the California Current system was presented by Gentemann et al. . Here, we build upon that comparative analysis to identify and quantify the effects of NPQ on the stimulated ChlF signal, with the goal of removing known biases and improving these in situ measurements for a better comparison with satellite data. Furthermore, NPQ-corrected and uncorrected Saildrone ChlF measurements are compared against satellite estimates of Chl to evaluate the utility of applying an NPQ-correction to Saildrone in vivo ChlF measurements and to constrain the in situ fluorometer’s calibration via contemporaneously collocated HPLC pigment measurements.
2.1 Non-photochemical quenching correction
NPQ can be identified in near-surface (depth < 2 m) ChlF measurements by comparing daytime and nighttime measurements collected in temporally consistent waters (Fig. 2). The concentration of ChlF is expected to be consistent between daytime and nighttime measurements, provided the vehicle is sampling a body of water whose constituent components have not significantly changed during the period of observation [22,27].
Once identified in ChlF measurements, NPQ can be corrected by relying on the relationship between β and ChlF [21,28] (Fig. 3). The correction is derived by interpolating temporally across periods of NPQ. Unaffected nighttime measurements of ChlF are related to β from the night before the NPQ event, the night after, or both, depending on the temporal homogeneity of the water being sampled. The median ratio of ChlF:β for unaffected nighttime periods is then multiplied by the daytime β measurements to derive an NPQ-corrected daytime ChlF estimate. This approach operates on the premise that β measurements, unlike stimulated in vivo ChlF, are unaffected by NPQ since they are related to changes in concentration, indices of refraction and internal morphology of the light-scattering constituents of the water (phytoplankton, mineral particles, and detritus). Additionally, changes in particulate scattering of live organisms occur on larger time scales than NPQ and are not driven by the physiological responses of those constituents to solar stress. However, other factors, including a change in the marine environment (light availability, mixed layer depth) and consequently, a change in the concentration and type of light-scattering components of the water, can alter the ChlF:β relationship, rendering a temporal interpolation an invalid estimate of ChlF during an NPQ event.
Coincident Saildrone measurements of water temperature and salinity (Fig. 3) from the CTD instrument are analyzed alongside ChlF and β measurements to determine if there has been a fundamental change to the body of water that the vehicle was sampling (e.g. – crossing a frontal boundary, encountering a river plume intrusion, moving into a region that is dominated by recently upwelled water, etc). An NPQ correction is only appropriate to apply to an aquatic environment with a temporally and spatially consistent composition. Our approach to assess whether an NPQ correction is appropriate, relies on a holistic evaluation of contemporaneous measurements of the vehicle’s environment. Changes to bulk parameters, temperature and salinity, that describe a body of water’s composition, often accompany changes in the phytoplankton community structure and composition. Therefore, discontinuities and sudden changes in the variability of temperature, salinity, β and the ChlF:β ratio could be indicative of changes in the water composition. For such situations, we did not apply a NPQ correction. We acknowledge that there may be some subjectivity in this determination and that our view of the targeted water mass is two dimensional only since we have no knowledge about vertical distribution of phytoplankton community, such as stratification and mixing, (factors that can influence variability in ChlF:β ratio). However, in unclear or questionable circumstances, we erred on the side of caution and chose to not derive an NPQ correction for potentially temporally, non-homogeneous waters.
2.2 Saildrone and satellite matchups
Saildrone ChlF measurements were matched up to co-incident geophysical, swath-based (level-2) satellite estimates of Chl derived from retrievals of Rrs(λ) using the ocean chlorophyll index (OCI) approach  from the NASA Moderate Resolution Imaging Spectroradiometer onboard Aqua (MODIS-Aqua)  and the Visible Infrared Radiometer Suite onboard Suomi NPP (VIIRS-SNPP) , referenced hereafter as MODIS and VIIRS, respectively, via the methods and exclusion criteria of Bailey and Werdell . A matchup was found if: 1) if temporal coincidence of the satellite retrievals and in situ measurement was within ±3 hours; 2) if the filtered median (via the interquartile range) of all unmasked pixels of a 5 × 5 extract of satellite values centered on the in situ measurement could be computed; and 3) if the coefficient of variation within these extracted pixels was lower than 0.15 with more than 50% of the 25 extracted pixels remaining valid. Satellite pixels were considered invalid according the level-2 quality filtering approach of Scott and Werdell , namely: 1.) if certain quality flags were triggered, 2.) if the solar zenith angle exceeded 70°, or 3.) if the satellite zenith angle exceeded 60°. Saildrone measurements were also combined into a median value if multiple in situ measurement fell spatially within a given satellite pixel, since assessing intra-pixel variability is beyond the scope of this discussion, yet within the plausible capabilities of the Saildrone vehicle.
2.3 Comparison metrics
Mean bias, mean absolute error (MAE), and unbiased percent difference (UPD) were chosen as metrics to evaluate Saildrone ChlF measurements relative to satellite-derived Chl estimates, since Chl data have non-Gaussian distributions . The selected metrics are defined as:
The distribution of Chl data from this mid-2018 Saildrone deployment exists across several orders of magnitude , and the uncertainties associated with Chl proportionally vary with data magnitude . Therefore, Chl is handled in log-transformed space, and mean bias and MAE are presented in their multiplicative forms (see Eqs. (1) and (2)). The multiplicative forms of these metrics are dimensionless, and as such, may be expressed as a percentage of the measurement. When interpreting the systematic errors being quantified by multiplicative mean bias, values closer to unity indicate lower bias, with values less than unity corresponding to a negative bias. The random error expressed by multiplicative MAE will always be positive.
3. Results and discussion
The Saildrone measured ChlF in oligotrophic (Chl ≤ 0.1 µg L-1), mesotrophic (0.1 < Chl ≤ 1 µg L-1), and eutrophic (Chl > 1 µg L-1) regimes [33,35] (Fig. 2). These types of regimes are common to the diverse California Current system due to river plume intrusion and the upwelling of cold, deep, and nutrient-rich water. Despite being deployed for 60 days, in productive, eutrophic waters, the vehicle’s ECO Triplet-w instrument did not experience any detectable biofouling of the optical components. This was apparent from the variability of the ChlF signal (Fig. 2) showing natural oscillations absent of any large instrumental drift or data discontinuities, common to data affected by biofouling .
Non-photochemical quenching was detected during this Saildrone deployment. One example of NPQ is evident on 20 Apr 2018 from the dip in the ChlF data (Fig. 3, top panel, black line) between 09:00 to 15:00 local time. Uncorrected Saildrone Chl measurements appear biased low, relative to the stable nighttime values and relative to the temporally coincident VIIRS satellite retrievals of Chl. Stability is seen the night before and the night after in the temperature, salinity, Chl, and ChlF:β ratio. There is a warming signal (∼1°C) seen in the daytime temperature measurements followed by nighttime cooling, which may be in part due to a diurnal warming event . This variability in the water temperature is not indicative that the composition of the water had changed. The backscatter data from the night before does contain some noise, but it is over a relatively small dynamic range and variability remains low in the ChlF:β ratio. Combined with no major variations in salinity, these factors indicate that the vehicle was sampling a uniform body of water and applying a correction to remove the effects of NPQ on 20 Apr 2018 is appropriate. After the correction is derived (Fig. 3, magenta line), ChlF data, which had previously been biased low relative to the nighttime measurements and the daytime VIIRS satellite retrievals, shows improved agreement with both (see next subsection).
Of this 60-day Saildrone deployment, an NPQ correction is appropriate for 57% of the days (34 days). During these days, it was determined that the vehicle sampled relatively homogenous water between the NPQ-affected daytime and one or more of the adjacent nights. This determination was reached by an analysis of the stability of the contemporaneous measurements of temperature, salinity, backscatter, and the ChlF:β ratio. Superimposed on the diel changes in fluorescence are short-term variations, some of which may result from the passage of clouds and from photodamage to the reaction centers of the phytoplankton, resulting from overexposure to above-optimal irradiance. However, the passage of clouds across the sky is slow enough to permit the changes to the effective cross sections of the photosynthetic apparatus, which will accordingly result in a change to the fluorescence yields. This phenomenon is readily observed in stimulated fluorescence profiles and in time series observations made under partially cloudy conditions [36–38]. Differencing the NPQ-corrected data with the uncorrected data reveals that the median NPQ adjustment for the 34 corrected days is + 0.0441 µg L-1 with an interquartile range of 0.0798 µg L-1. The NPQ-correction corresponds to a median percent increase of 86.8% for ChlF across the oligotrophic, mesotrophic, and eutrophic regimes sampled by the Saildrone.
Two further corrections are commonly applied to bring the ChlF values into better agreement with in situ measurements of true chlorophyll concentration. First, a correction may be applied to remove the contribution of CDOM to ChlF . The presence of CDOM can lead to overestimation of Chl by the in situ fluorometer. This overestimation occurs under two sets of conditions: in waters with high CDOM concentration or in waters with very low chlorophyll concentrations . A CDOM correction was not applied to the Saildrone ChlF data since this deployment was conducted across oligotrophic and mesotrophic conditions, with low-to-negligible CDOM concentrations.
A second commonly applied ChlF correction is the factory calibration slope correction . Factory calibrations are designed to constrain the Chl:fluorescence ratio based on the Chl regime in which the instrument is to be deployed. WET Labs factory calibrations of ECO fluorometers tend to introduce a factor of two bias, and recommendations have been made to use concurrent in situ discreet measurements of Chl as a correction [10,13]. However, this Saildrone deployment’s Chl:fluorescence ratio could not be constrained beyond the pre-cruise factory calibration because no same-day discreet measurements of Chl were located within a 5-km radius of the Saildrone vehicle. Two CalCOFI cruises occurred during this Saildrone deployment: the 1804 spring cruise in April 2018 and the 1806 summer cruise in June 2018. Both of these cruises collected discreet Chl concentration measurements via HPLC pigment analysis of discreet bottle samples of seawater. However, the closest same-day CalCOFI measurement was at a distance of 68 km from the Saildrone vehicle. Additional, robust sources of in situ Chl measurements were sought from the NASA SeaWiFS Bio-optical Archive and Storage System (SeaBASS) , but none were found to be sufficiently coincident in space and time with the Saildrone. In the absence of collocated discreet measurements of Chl, the overestimation of ChlF by WET Labs ECO fluorometers may be remedied by applying a factor-of-two correction to the instrument’s calibration slope, as suggested by Roesler et al. . However, applying this correction to the NPQ-corrected ChlF measurements from this Saildrone deployment decreased the value of our in situ measurements and did not improve the agreement between the ChlF measurements and satellite retrievals of Chl. Therefore, applying the Roesler et al. factor-of-two factory calibration slope correction  was not deemed appropriate for this ChlF dataset.
Satellite-to-in situ matchups were found between the Saildrone and the MODIS and VIIRS satellite instruments. For consistency when comparing corrected and uncorrected Saildrone ChlF data, satellite-to-in situ matchups are only considered if an NPQ-correction was derived. This is to conserve the number of matchups between corrected and uncorrected ChlF data sets, in order to evaluate the benefit of the NPQ-correction relative to the satellite data. This is a departure from the preliminary satellite-relative evaluation of the Saildrone ChlF data from this deployment presented by the authors in Gentemann et al. . Only considering matchups for where an NPQ correction could be derived reduced the number of unique matchups by 63.3% for MODIS (from 221 to 81) and by 72.8% for VIIRS (from 346 to 94). VIIRS and MODIS matchups with NPQ-corrected Saildrone data occurred on 5 and 6 unique days, respectively, and only one day (25 Apr 2018) contained both MODIS and VIIRS matchups.
The effects of NPQ relative to MODIS and VIIRS can be seen across oligotrophic, mesotrophic, and eutrophic regimes for MODIS and across mesotrophic and eutrophic regimes for VIIRS (Fig. 4). However, for both MODIS and VIIRS, the matchups have a bimodal distribution between relatively high and low Chl. The uncorrected Saildrone data have a lower mean bias of 296% and 195% than MODIS and VIIRS, respectively. The corrected Saildrone data, while still biased low relative to the satellite data, have mean bias of 78.8% and 26.11% when compared to MODIS and VIIRS, respectively. The MAE associated with the uncorrected Saildrone data is 295% relative to MODIS and 197% relative to VIIRS. The NPQ-correction reduces these MAE values to 100% and 41.7% compared to MODIS and VIIRS, respectively.
Applying an NPQ-correction to this Saildrone deployment resulted in a reduction of both systematic and random errors. As a measure of systematic error, the mean biases between the in situ Chl measurements and satellite Chl retrievals were reduced 75.5% relative to MODIS and 80.3% relative to VIIRS as a result of applying an NPQ-correction to the Saildrone data (Table 1). The random error decreased by 65.7% and 70.8% for MODIS and VIIRS, respectively. Overall, better agreement between Saildrone and satellite data can be seen in Fig. 4 where the satellite-to-in situ matchups falls closer along the 1:1 line for the NPQ-corrected Saildrone data (upper panels) compared to the uncorrected Saildrone data (lower panels), both relative MODIS and VIIRS.
A Saildrone semi-autonomous surface vehicle was deployed in the California Current region for 60-days from April to June in 2018, measuring a variety of ocean and atmospheric parameters including ChlF. A variety of factors, adjustments, and corrections relevant to ChlF measurements were investigated. The vehicle’s ChlF had no detectable biofouling during this deployment. The Saildrone ChlF measurements were affected diurnally by NPQ, but these effects were correctable for more than half of the cruise days (34 out of 60). The NPQ correction derived and applied to this deployment reduces the bias of the Saildrone measurements relative to MODIS and VIIRS satellite retrievals of Chl, which underscores the importance of correcting NPQ in ChlF data. The evaluation of Saildrone ChlF for NPQ, biofouling, and calibration adjustments is essential before comparing Saildrone in vivo ChlF to satellite data retrievals of Chl to remove known, spurious biases between these two types of ocean observations. These considerations are important to make for a Saildrone semi-autonomous surface vehicle to have utility for feature analysis (e.g., frontal or turbulent eddy ChlF gradients, etc), but a tighter constraint on the Chl:fluorescence, such as can be derived from temporally and spatially coincident HPLC measurements, must be derived before the full utility of the Saildrone ChlF data for satellite ocean color validation can be fully realized.
When collocated discreet measurements of Chl are unavailable, the utility of the Saildrone measurement platform is reduced. The Saildrone semi-autonomous surface vehicle has the potential to extend the global set of Chl measurements both temporally and spatially in remote oceanic regions that are not routinely sampled by other monitoring vessels or observing platforms. Given the variety of physiological conditions and stressors that affect ChlF, the Saildrone platform can only accurately estimate Chl concentration via ChlF measurements when the Chl:fluorescence ratio is well-constrained by in situ pigment concentration data. Future deployment of Saildrone vehicles would be more valuable when accompanied by spatially and temporally coincident measurements of HPLC-derived pigment concentrations for this purpose, ideally with accompanying measurements of pigment absorption, which is the truly interesting parameter from a mechanistic perspective.
National Aeronautics and Space Administration (PACE); Earth Sciences Division (Ocean Biology and Biogeochemistry Program); Goddard Space Flight Center (Summer Intern Program); Schmidt Family Foundation; Saildrone Inc. (2018 Saildrone Award).
We thank Saildrone, Inc. for cruise planning, for data collection and for processing of the Saildrone data set. We thank the NASA GSFC Ocean Biology Processing Group (OBPG), SeaDAS and OCSSW developers for providing ocean color satellite data and software.
The authors declare no conflicts of interest.
1. R. J. Geider, “Light and Temperature Dependence of the Carbon to Chlorophyll a Ratio in Microalgae and Cyanobacteria: Implications for Physiology and Growth of Phytoplankton,” New Phytol. 106(1), 1–34 (1987). [CrossRef]
2. J. E. Cloern, C. Grenz, and L. Vidergar-Lucas, “An empirical model of the phytoplankton chlorophyll: carbon ratio-the conversion factor between productivity and growth rate,” Limnol. Oceanogr. 40(7), 1313–1321 (1995). [CrossRef]
3. B. Riemann, P. Simonsen, and L. Stensgaard, “The carbon and chlorophyll content of phytoplankton from various nutrient regimes,” J. Plankton Res. 11(5), 1037–1045 (1989). [CrossRef]
4. C. D. Mobley, J. Werdell, B. Franz, Z. Ahmad, and S. Bailey, “Atmospheric Correction for Satellite Ocean Color Radiometry,” NASA Tech. Memo. 2016–21755, 1–73 (2016).
5. H. M. Dierssen, “Perspectives on empirical approaches for ocean color remote sensing of chlorophyll in a changing climate,” Proc. Natl. Acad. Sci. U. S. A. 107(40), 17073–17078 (2010). [CrossRef]
6. C. Hu, Z. Lee, and B. Franz, “Chlorophyll a algorithms for oligotrophic oceans: A novel approach based on three-band reflectance difference,” J. Geophys. Res.: Oceans 117(C1), 1–25 (2012). [CrossRef]
7. O. Holm-Hansen, C. J. Lorenzen, R. W. Holmes, and J. D. H. Strickland, “Fluorometric Determination of Chlorophyll,” ICES J. Mar. Sci. 30(1), 3–15 (1965). [CrossRef]
8. C. S. Roesler and A. H. Barnard, “Optical proxy for phytoplankton biomass in the absence of photophysiology: Rethinking the absorption line height,” Methods Oceanogr. 7, 79–94 (2013). [CrossRef]
9. E. Boss, M. Picheral, T. Leeuw, A. Chase, E. Karsenti, G. Gorsky, L. Taylor, W. Slade, J. Ras, and H. Claustre, “The characteristics of particulate absorption, scattering and attenuation coefficients in the surface ocean; Contribution of the Tara Oceans expedition,” Methods Oceanogr. 7, 52–62 (2013). [CrossRef]
10. C. Roesler, J. Uitz, H. Claustre, E. Boss, X. Xing, E. Organelli, N. Briggs, A. Bricaud, C. Schmechtig, A. Poteau, F. D’Ortenzio, J. Ras, S. Drapeau, N. Haëntjens, and M. Barbieux, “Recommendations for obtaining unbiased chlorophyll estimates from in situ chlorophyll fluorometers: A global analysis of WET Labs ECO sensors,” Limnol. Oceanogr.: Methods 15(6), 572–585 (2017). [CrossRef]
11. P. Falkowski and D. A. Kiefer, “Chlorophyll a fluorescence in phytoplankton: relationship to photosynthesis and biomass*,” J. Plankton Res. 7(5), 715–731 (1985). [CrossRef]
12. I. Cetinić, G. Toro-Farmer, M. Ragan, C. Oberg, and B. H. Jones, “Calibration procedure for Slocum glider deployed optical instruments,” Opt. Express 17(18), 15420–15430 (2009). [CrossRef]
13. N. Haëntjens, E. Boss, and L. D. Talley, “Revisiting Ocean Color algorithms for chlorophyll a and particulate organic carbon in the Southern Ocean using biogeochemical floats,” J. Geophys. Res.: Oceans 122(8), 6583–6593 (2017). [CrossRef]
14. C. L. Gentemann, J. P. Scott, P. L. F. Mazzini, C. Pianca, S. Akella, P. J. Minnett, P. Cornillon, B. Fox-Kemper, I. Cetinić, T. M. Chin, J. Gomez-Valdes, J. Vazquez-Cuervo, V. Tsontos, L. Yu, R. Jenkins, S. De Halleux, D. Peacock, and N. Cohen, “Saildrone: adaptively sampling the marine enviroment,” Bull. Am. Meteorol. Soc. submitted (n.d.).
15. M. J. Behrenfeld, Y. Hu, R. T. O’Malley, E. S. Boss, C. A. Hostetler, D. A. Siegel, J. L. Sarmiento, J. Schulien, J. W. Hair, X. Lu, S. Rodier, A. J. Scarino, R. T. O’Malley, E. S. Boss, C. A. Hostetler, D. A. Siegel, J. L. Sarmiento, J. Schulien, J. W. Hair, X. Lu, S. Rodier, and A. J. Scarino, “Annual boom-bust cycles of polar phytoplankton biomass revealed by space-based lidar,” Nat. Geosci. 10(2), 118–122 (2017). [CrossRef]
16. C. Meinig, N. Lawrence-Slavas, R. Jenkins, and H. M. Tabisola, “The use of Saildrones to examine spring conditions in the Bering Sea: Vehicle specification and mission performance,” in OCEANS 2015 - MTS/IEEE Washington (2015), pp. 1–6.
17. G. P. Gerbi, E. Boss, P. Jeremy Werdell, C. W. Proctor, N. Haëntjens, M. R. Lewis, K. Brown, D. Sorrentino, J. R. V. Zaneveld, A. H. Barnard, J. Koegler, H. Fargher, M. DeDonato, and W. Wallace, “Validation of ocean color remote sensing reflectance using autonomous floats,” J. Atmos. Oceanic Technol. 33(11), 2331–2352 (2016). [CrossRef]
18. P. J. Werdell, C. W. Proctor, E. Boss, T. Leeuw, and M. Ouhssain, “Underway sampling of marine inherent optical properties on the Tara Oceans expedition as a novel resource for ocean color satellite data product validation,” Methods Oceanogr. 7, 40–51 (2013). [CrossRef]
19. M. Lehaitre and C. Compère, “Biofouling and Underwater Measurements,” Real-Time Coast. Obs. Syst. Mar. Ecosyst. Dyn. Harmful Algal Bloom. Theory, Instrum. Model.463–493 (2008).
20. A. Matsuoka, V. Hill, Y. Huot, M. Babin, and A. Bricaud, “Seasonal variability in the light absorption properties of western Arctic waters: Parameterization of the individual components of absorption for ocean color applications,” J. Geophys. Res.: Oceans 116(C2), C02007 (2011). [CrossRef]
21. B. S. Sackmann, M. J. Perry, and C. C. Eriksen, “Seaglider observations of variability in daytime fluorescence quenching of chlorophyll-a in Northeastern Pacific coastal waters,” Biogeosciences Discuss. 5(4), 2839–2865 (2008). [CrossRef]
22. S. J. Thomalla, W. Moutier, T. J. Ryan-Keogh, L. Gregor, and J. Schütt, “An optimized method for correcting fluorescence quenching using optical backscattering on autonomous platforms,” Limnol. Oceanogr.: Methods 16(2), 132–144 (2018). [CrossRef]
23. Saildrone, “Saildrone Baja field campaign surface and ADCP measurements,” Ver. 1.0. (Accessed: 2019-06-04) (2018).
24. D. M. Checkley and J. A. Barth, “Patterns and processes in the California Current System,” Prog. Oceanogr. 83(1-4), 49–64 (2009). [CrossRef]
25. K. H. Brink, R. C. Beardsley, J. Paduan, R. Limeburner, M. Caruso, and J. G. Sires, “A view of the 8575–8604 California Current based on surface drifters, floats, and remotely sensed data,” J. Geophys. Res.: Oceans 105(C4), 8575–8604 (2000). [CrossRef]
26. P. T. Strub and C. James, “The large-scale summer circulation of the California Current,” Geophys. Res. Lett. 22(3), 207–210 (1995). [CrossRef]
27. L. Carberry, C. Roesler, and S. Drapeau, “Correcting in situ chlorophyll fluorescence time-series observations for nonphotochemical quenching and tidal variability reveals nonconservative phytoplankton variability in coastal waters,” Limnol. Oceanogr.: Methods 17(8), 462–473 (2019). [CrossRef]
28. M. J. Behrenfeld and E. Boss, “Beam attenuation and chlorophyll concentration as alternative optical indices of phytoplankton biomass,” J. Mar. Res. 64(3), 431–451 (2006). [CrossRef]
29. NASA, “Moderate Resolution Imaging Spectroradiometer onboard the Aqua Earth-observing satellite (MODIS-Aqua) Ocean Color Data,” (2018).
30. NASA, “Visible Infrared Radiometer Suite onboard the Suomi National Polar-orbiting Partnership weather satellite (VIIRS-SNPP) Ocean Color Data,” (2018).
31. S. W. Bailey and P. J. Werdell, “A multi-sensor approach for the on-orbit validation of ocean color satellite data products,” Remote Sens. Environ. 102(1-2), 12–23 (2006). [CrossRef]
32. J. P. Scott and P. J. Werdell, “Comparing level-2 and level-3 satellite ocean color retrieval validation methodologies,” Opt. Express 27(21), 30140–30157 (2019). [CrossRef]
33. B. N. Seegers, R. P. Stumpf, B. A. Schaeffer, K. A. Loftin, and P. J. Werdell, “Performance metrics for the assessment of satellite data products: an ocean color case study,” Opt. Express 26(6), 7404 (2018). [CrossRef]
34. J. W. Campbell, “The lognormal distribution as a model for bio-optical variability in the sea,” J. Geophys. Res. 100(C7), 13237–13,254 (1995). [CrossRef]
35. B. A. Franz, P. J. Werdell, G. Meister, S. W. Bailey, and R. E. EpleeJr., G. C. Feldman, E. Kwiatkowskaa, C. R. McClain, F. S. Patt, and D. Thomas, “The continuity of ocean color measurements from SeaWiFS to MODIS,” Proc. SPIE 5882, 58820W (2005). [CrossRef]
36. M. R. Abbott, P. J. Richerson, and T. M. Powell, “In situ response of phytoplankton fluorescence to rapid variations in light,” Limnol. Oceanogr. 27(2), 218–225 (1982). [CrossRef]
37. M. Stramska and T. D. Dickey, “Short-term variations of the bio-optical properties of the ocean in response to cloud-induced irradiance fluctuations,” J. Geophys. Res. 97(C4), 5713 (1992). [CrossRef]
38. M. Stramska and T. D. Dickey, “Short-term variability of the underwater light field in the oligotrophic ocean in response to surface waves and clouds,” Deep Sea Res., Part I 45(9), 1393–1410 (1998). [CrossRef]
39. C. W. Proctor and C. S. Roesler, “New insights on obtaining phytoplankton concentration and composition from in situ multispectral Chlorophyll fluorescence,” Limnol. Oceanogr.: Methods 8(12), 695–708 (2010). [CrossRef]
40. P. J. Werdell and S. W. Bailey, “The SeaWiFS Bio-Optical Archive and Storage (SeaBASS): Current Architecture and Implementation,” NASA Tech. Memo. 2002–21161, 1–50 (2002).