1. Introduction

Many shipboard data sets are collected using underway seawater systems over the duration of research cruises. Such underway optical and biogeochemical measurements can provide high quality data for use in developing and refining algorithms, and satellite data product validation activities for current ocean color missions (e.g., TARA Oceans data set) [1]. Deployment of these systems by different scientists funded by the NASA Ocean Biology and Biogeochemistry Program on research vessels across all oceans offers increased diversity of data in the NASA SeaWiFS Bio-optical Archive and Storage System (SeaBASS), an in situ data archive curated by the NASA Ocean Ecology Lab [2]. Resulting data sets offer high spatial and temporal resolution of optical and biogeochemical data that is crucial for the validation of new or future ocean color hyperspectral (NASA’s PACE), high-spatial (ESA’s Sentinel-3) or high-temporal (KIOST’s GOCI and NASA’s GEO-CAPE) missions.

Particle size distribution (PSD) in the ocean is a key measurement, required in oceanographic research and ocean optics. In the open ocean, PSD is directly related to plankton size distribution and therefore its functional role [3, 4], and is highly important for our understanding of ecosystem functioning and ocean biogeochemistry, as well as any changes those mechanisms might experience over time. The recent increase of new remote sensing algorithms (both multispectral and hyperspectral) for detecting different phytoplankton types and sizes has made optical and biogeochemical measurements collected using flow-through systems even more necessary, as they offer high temporal/spatial resolution needed for the validation of such algorithms.

For most underway systems, the seawater supply onboard is provided by impeller pumps through a fixed “uncontaminated science seawater” plumbing system (e.g., typical of UNOLS vessels). However, researchers within the oceanographic community have ceased using seawater provided by the ship’s intake systems due to the impact they have on the quality of data collected. That is also the case for chemical oceanographers, whose samples can be easily contaminated by the pumps and tubing material [5], altered by the removal or addition of material or gases due to the presence of microbial biofilms and filter-feeding organisms on the components of the permanently mounted systems [6], or the release of the accumulated fouling on the intake filters (e.g [7].). Our experience has suggested that some pump systems installed on research vessels for uncontaminated seawater supplies (usually impeller pumps) can cause morphological (and potentially physiological) damage to the plankton cells, an important consideration regarding the quality of the associated optical and biogeochemical data.

Particulate beam attenuation, the most common optical measurement in oceanography, is predominantly driven by the particles in the 0.5–20 μm-size range. Furthermore, the volume scattering function of larger particles causes high dependence on the acceptance angle of the instrument, causing variability in attenuation measurements depending on the model of instrument used [8]. Hence, change in particle size distribution due to the breakage of cells (or aggregates) might have a significant effect on the particulate attenuation measurements in the ocean [9]. Breakage might impact particulate backscattering even more: this important oceanic measurement, directly related to remote sensing reflectance, is highly dependent on size distribution, morphology and the particle’s index of refraction, all of which will be modified if breakage of the cells occurs within intake systems.

Using a data set collected in the North Atlantic, spanning different oceanic water types, we outline the differences observed in concurrent samples from two different flow-through systems. The first pump system is a ship’s current uncontaminated science seawater supply plumbing and an impeller pump; the second system is independent of the ship’s system, installed by the science party, employing shorter, clean tubing runs and a diaphragm pump. We evaluate the differences observed across several measured parameters, including phytoplankton size distribution, community composition, and the carbon and chlorophyll content. Our results outlined here should be used as a guide towards building best case practices when it comes to the deployment of similar flow-through systems in the field of optics. Additionally, these results may serve as a warning to researchers using previously collected data sets, where the methods used to collect data, could introduce a significant and highly variable ecosystem-dependent bias.

2. Materials and methods

2.1 Underway sampling system setup

This study was conducted during the AE1319 cruise aboard R/V Atlantic Explorer, spanning a period of five days sailing along the coast of Nova Scotia, Newfoundland, Canada to its initial sampling station in the Labrador Sea (Fig. 1(a)).

 

Fig. 1 A) Station locations (marked with x) superimposed on a MODIS Chlorophyll 8-day composite, 20-28, during the flow-through experiment (color scale represents Chlorophyll a (μg L⁻¹)). Continuous temperature and salinity, depicted on panel B, with station time points marked with gray lines. Each station is located within a different water mass, correlating with the observed changes in the nano/micro phytoplankton community. C) Homogeneity in the mixed layer on ST4 is visible in particulate attenuation (purple) and density (gray). Mixed layer depth (MLD) is marked with blue dashed line.

Download Full Size | PDF

We compared two flow-through seawater systems: (1) the permanently-installed ship’s uncontaminated seawater system (impeller pump system, IPS); and (2) a temporary uncontaminated seawater system installed for the purposes of the principle investigators’ measurements (diaphragm pump system, DPS).

The ship’s permanently-installed system (IPS) is used for measurements with a standard instrument suite (SBE-21 thermosalinograph, Sea-Bird Scientific; and chlorophyll fluorometer, Turner Designs) and also is used to deliver seawater to forward and aft labs for science party use. The system (as of September 2013 [10],) utilizes a magnetically-coupled impeller pump (Marathon Electric 15518510) rated for 3.34 × 10⁻³ m³ s⁻¹ (200 L min⁻¹) with intake on the starboard bow, approximately two meters below the mean water line. Wetted parts in the pump are polypropylene, Viton, ceramic, and carbon, and plumbing throughout the ship is 1-inch schedule 80 PVC pipe. A flow sensor is also included in the system at the outflow of the SBE21. The estimated plumbing run length from the bow to aft laboratory for this system is ~36 m.

The second (temporary) seawater system (DPS) was installed by the authors, similar to a system previously installed aboard R/V Atlantic Explorer [11], to make underway inherent optical property and biogeochemical property measurements. This system employed an air-operated diaphragm pump (Ingersoll-Rand ARO, PD07A-AAS-PTT) installed above the waterline with a flexible suction hose connecting to a rigid section of PVC pipe mounted on a frame (Straza tower). The latter part of the intake system was lowered into the moon pool and positioned with the intake (~3.5 m below surface) in a forward-facing direction such that flow while underway would assist in priming the pump. Priming was accomplished by filling the suction hose using a valve-tee fitting and connecting it to the ship’s flowing seawater. A suction strainer (2.36-mm mesh size, stainless steel) was placed on the intake. Seawater pumped through the diaphragm pump was plumbed to the nearby aft lab using black opaque ¾” semi-rigid tubing (LLDPE, food-grade, approximately 7 m tubing run; total plumbing run length of ~13 m, including rigid suction tubing between intake and pump). Compressed air was sourced from the same lab, through a filter-dryer-regulator, with air pressure used to control the flow rate of the pump. The poly tubing was wrapped in foam pipe insulation to minimize temperature change in the seawater before introduction to instruments, as well as to dampen the pulsation and noise inherent in the diaphragm pump. The ARO pump was constructed of aluminum, with primary wetted parts (i.e., diaphragms, constructed of Teflon). Seawater was supplied to instruments after passing through a debubbler (MSRC VDB-1 2-inch, Instrument Laboratory, Stony Brook University). The flow rate into the debubbler was approximately 7-10 L min⁻¹, with output to instruments approximate-ly 5 L min⁻¹.

At ST1, for comparison purposes, we have collected a surface sample using a ~20 L bucket, deployed over the side of the ship, at the same time samples were collected from the IPS and DPS systems.

2.2 Sample and data analyses

Temperature and salinity (derived from conductivity) were measured by Sea-Bird’s SBE 21 SeaCAT thermosalinograph, deployed as part of the R/V Atlantic Explorer underway instrument suite. Eight-day, 9-km MODIS Aqua Chlorophyll product (OCI algorithm) for the period of Aug 20-28, 2013 was obtained from the NASA Ocean Color archive [12]. Vertical profile of the hydrographic and optical properties was collected by a ship-deployed CTD rosette, containing a Sea-Bird’s SBE 11 + CTD, and a WETLabs C-star attenuation meter. Data was processed using manufactures prescribed protocols. The mixed layer (ML) was defined in two different ways: (1) as a layer in which density differs from the average of the top 10 m density by <0.05 kg m⁻³; and (2) as described in Zubkov, et al. [13], as the depth of maximum change in optical profile, i.e., particulate attenuation, relative to the change in depth.

Microbial phytoplankton cell concentrations and particle size distributions were determined from fresh samples at sea during the cruise [14]. A single sample was taken from each source (IPS, DPS or surface) for each of the stations. Cells larger than 5 µm were imaged and analyzed using a FlowCam imaging cytometer (Fluid Imaging Technologies, Scarborough, ME), with image collection triggered by chlorophyll a fluorescence. The imaging cytometer was equipped with a 4x objective and a 300-μm flow cell, and was operated in fluorescence trigger mode at a flow rate of approximately 2.5 mL min⁻¹. For each sample, 100 mL of seawater was processed by the FlowCam, followed by a rinse of filtered seawater. Only complete (chlorophyll containing) cells were counted using laser excitation (532 nm). Cell concentrations were determined based on the volume of the sample imaged by the FlowCam. Sample data was analyzed using Fluid Imaging Technologies Inc. VisualSpreadSheet software, and five sub-groups of nano/micro-phytoplankton were identified: diatoms, dinoflagellates (autotrophic and mixotrophic), ciliates, ‘small cells’ and ‘other’. The ‘small cell’ phytoplankton group includes all cells less than 10 μm (range 5-10 μm). The ‘other’ group includes all cells, larger than 10 μm, that were not identified as one of the target groups. Using the binary image of each particle, the area attributed to each particle is a sum of pixels that satisfied the predefined threshold, multiplied by the size of each pixel (1 pixel = 1 μm). The circular area obtained from each image was then used to derive the Area Based Diameter (ABD). Both the ABD and cell biovolume were determined using the Visual Spreadsheet Software. Cell carbon (phytoplankton biomass) was calculated from the derived biovolumes using the algorithms of Menden-Deuer and Lessard [15].

For chlorophyll determination, water samples (in triplicates) were filtered through Whatman GF/F filters and stored at −80° C till analysis (within 5 days). Filtered samples were extracted in 5 mL of 90% acetone at −20° C for 24 hours and analyzed fluorometrically on a Turner Designs Model 10-AU digital fluorometer, which was calibrated before and after the field experiment with Turner Designs Chlorophyll (Chl) standards. Chl concentrations were calculated following Joint Global Ocean Flux Study protocol [16].

2.2 Uncertainty estimates

Due to the lack of replication in the cell count data we explored different routes in determining the uncertainties associated with our measurements. First, we estimated the methodology error associated with the number of cells counted in each of the samples, following the approach outlined by Guillard and Sieracki [17] (on 95% confidence limits). Second, we assumed that, for ST1, surface and DPS samples are coming from the same population, and resulting difference is due to the natural variability of the population. Third, we applied the sample-specific variability encountered in chlorophyll measurements to the phytoplankton count numbers (following the idea that both are an estimate of biomass therefore should vary in a similar manner). Fourth, for method #4 and 5, we used variability estimate from previously published study, where phytoplankton abundance [18] was measured repeatedly over the temporal/special scales applicable to this experiment. We applied these error estimates, expressed as coefficient of variation, on the total cell counts and the derived phytoplankton carbon, and run appropriate hypothesis tests.

2.3 Optical property models and calculations

To examine the effects of changing particle packaging and size distribution, we computed the phytoplankton beam attenuation and backscattering using size distribution measured with FlowCam. The aggregate model is described in detail in [19]. Briefly, the model assumes that a porous, fractal, aggregate can be represented as the average of (i) a shelled sphere with outer diameter that of the enclosing volume of the aggregate, composed of an inner water core and outer shell composed of the particulate material; and (ii) a spheroid with volume equivalent to aggregate and index of refraction diluted according to the aggregate porosity.

Phytoplankton beam attenuation ($c_{phyto}$) and backscattering ($b_{b,phyto}$) were calculated using Mie theory and FlowCam-measured cell size distributions as

For Mie calculations, the relative index of refraction was assumed to be $m=1.05$, typical of organic phytoplankton particles, and to assess the uncertainty in refractive index, plots of the cumulative beam attenuation and backscattering include error bars showing variation in the calculations resulting from $m=1.05\pm 0.01$.

Mie theory calculations were made using a vectorized Mie-theory code (fastmie.m) in MATLAB (available at http://misclab.umeoce.maine.edu/software.php).

3. Results

Our experiment took place over several different oceanic regimes (Fig. 1), starting in the coastal waters off Nova Scotia and ending in the open oceanic waters of Labrador Sea/North Atlantic region. Based on the temperature and salinity measured by the ship’s flow-through systems, and the satellite derived Chl, each station represented a different ecosystem, including a different microphytoplankton species composition (Fig. 2). A single CTD profile collected at ST4 (Fig. 1(c)) demonstrates homogeneity in the upper, mixed layer (ML) of the water column with coefficient of variation (c.v.) of density in ML being 0.04%, and c.v. of attenuation 1.5%. The ML depth at this station, using both physical and optical measurements, was found to be at 17.6 m. Small scale horizontal variability in surface physical parameters over ± 500 m surrounding the sampling site was small in comparison with overall range encountered (Fig. 1(b)), with the area surrounding ST1 having the highest variability in surface parameters (as measured by the IPS thermosalinograph): maximum standard deviation in salinity of 0.02 (c.v. of 0.07%), and 0.03 °C for temperature (c.v. of 0.32%).

 

Fig. 2 Comparison of phytoplankton cell concentrations for samples obtained from the two different flow-through systems (DPS and IPS) at all stations (ST1-ST4). At the first station, samples were collected directly from the ocean’s surface and used for comparison with other two systems. Note: not all five phytoplankton groups were present in all the samples, as demonstrated by different coloring of the bar graphs (blue – diatoms, red – dinoflagellates, yellow – SC - “small cells”, gray – cili = ‘ciliates’ and –green – OT – ‘other’). Error bars are depicting: black – natural variability of the population, purple – method error (see section 2.2 and Table 1).

Download Full Size | PDF

Microphytoplankton collected at ST1 was dominated by dinoflagellates, ST 2 and 4 by smaller phytoplankton cells (including a visible diatom population), while ST 3 was associated with the warmest and least saline waters (probably associated with waters exiting the Gulf of St. Lawrence) and had high numbers of ciliates present in the sample.

Cell concentrations of the different microplankton groups varied between stations. Only at ST1 were samples from both underway flow-through systems compared with a sample collected directly from the surface (via a bucket). In the ST1 samples, both species’ composition (proportion of total count) and total cell concentration in the DPS sample were more representative of the surface (13% higher cell counts), while the IPS sample had 38% less cells than the surface sample and no ciliates were present. For all stations, a similar trend was observed when comparing the underway systems. We found an overall decrease in cell concentration when samples collected from the IPS system were compared to samples collected from the DPS system: 37%, 55%, 34% and 34% for ST1, ST2, ST3 and ST4, respectively. Average decrease of 40% was statistically significant (two tailed t-test, p = 0.0001).

Following the assumption that the difference between the surface sample and DPS sample collected on ST1 (13%) is due to the natural variability, the same difference can be applied to all the samples in the experiment to test the impact of the natural variability to the observed trend. After taking into consideration natural variability, the observed differences between the DPS and IPS remain statistically significant (Table 1, Fig. 2). Natural variability was observed in chlorophyll replicates as well, with average c.v. of 6%. Application of the sample specific variability in chlorophyll samples to the cell count numbers, once again yielded significant differences between DPS and IPS samples. Methodology associated uncertainties (here defined through sample specific accuracy related with number of cells counted), once applied to cell counts show significance of the observed decrease in cell counts between DPS and IPS samples for all stations but ST3, where total cell counts were order of magnitude smaller than counts on other stations (Table 1, Fig. 2). Uncertainties associated with the previously reported variability in phytoplankton counts [18] decreases the significance of the observed trend (Table 1), increasingly so when a larger c.v. is applied.

Table 1. Statistical Significance of the Trends Depicted in Fig. 1, Evaluated for Different Sources of Uncertainties^a

View Table

Decrease in cell counts varied across the different phytoplankton types. Significant decreases (t-test, alpha = 0.01) in cell abundances when comparing IPS and DPS (reported here as average) were visible with ciliates (78%), other (60%), diatoms (34%) and other small cells (34%). Dinoflagellate concentrations demonstrated a variable pattern: for ST1 and ST3, dinoflagellate cell concentrations were reduced in samples collected using the IPS system; however, in ST2 and ST4 dinoflagellate concentrations were higher than in the samples collected from the DPS system.

If we evaluate the change in PSD between the different underway systems, similar patterns are observed (Fig. 3). Comparisons were made only for bins (significant bins, n = 9, out of total of 19 size bins) in which at least one of the samples had a count higher than 1% of the total particle count for that sample (i.e., bin counts in excess of the customary “rare species” threshold). In samples collected at ST1, for all significant bins, DPS samples had (on average) 13% higher frequencies per bin than samples collected in surface waters, and 83% higher than IPS ones. IPS bin counts were on average 37% lower than the ones in surface waters. For all stations (except for ST3 where rare species threshold was never reached, probably due to the low cell counts), we encountered higher frequency of particles in all the significant size bins (39.4%) for the samples collected from the DPS system.

 

Fig. 3 First 4 panels depict frequency of particle size, binned to 5 μm size bins (5-100 μm), for samples taken from two different flow-through systems (DPS and IPS) for all stations (ST1-ST4). As with previous figures, the surface sample was collected on ST1 for comparison purposes. Gray dashed line depicts “rare species threshold”.

Download Full Size | PDF

Similar trends were found in phytoplankton carbon biomass numbers (Fig. 4(a)). Phytoplankton carbon biomass collected from the DPS system at ST1 was lower than the sample collected directly from the surface waters (9%), however this difference was not statistically significant (two tailed t-test, method 1: p = 0.32, method 2: p = 0.41). The samples collected from the IPS system at ST1 contained significantly lower phytoplankton biomass (33%) than compared to the surface sample (two tailed t-test, both methods p<0.02) and 26% lower when compared to the DPS sample (two tailed t-test, both methods p<0.05). A significant decrease in phytoplankton biomass was also observed in the IPS sample at ST3 (52%, two tailed t-test, both methods p<0.01). Samples collected at ST2 and ST4 displayed slightly lower, but not statistically significant, drop in biomass for samples collected from DPS vs. IPS (2% and 13% respectively, two tailed t-test, all p>0.35). A comparison of Chl concentrations across these experiments yielded different results Fig. 4(b), with lower Chl measured in samples collected from the DPS system on the first two stations; with significant decrease on ST1 (20%; two tailed t-test, p<0.005) and not statistically significant decrease on ST2 (15% ST2, two tailed t-test, p>0.05). Slight difference in Chl between two systems on ST3 (4%) and ST4 (2%), was not statistically significant (two tailed t-test, alpha = 0.5).

 

Fig. 4 A) Biomass as determined for nano-/micro-phytoplankton from biovolume measurements using imaging flow cytometry; B) Corresponding bulk chlorophyll concentrations obtained at each station along the transect, error bars are standard deviation (when triplicates were collected). Error bars on panel A are depicting: black – natural variability of the population, purple – method error (see section 2.2 and Table 1, Fig. 2).

Download Full Size | PDF

4. Discussion

The impact of the underway system on the phytoplankton community structure and its size distribution, as well as associated optical properties, has been a point of discussion in the ocean optics community; however, no comprehensive studies have been conducted prior to now. In this experiment, we have compared two underway systems to assess the impact of the systems’ components on the particle size distribution (PSD). Across four different ecosystems, a trend was consistent: phytoplankton numbers were lower in the IPS system installed on the research vessel using an impeller pump, as compared to the custom system using a diaphragm pump (DPS) that we deployed during the cruise.

The physical structure of the upper ocean determines the distribution of the phytoplankton, therefore, the difference in the intake depths of the IPS and DPS pumps could drive the observed differences in phytoplankton. Very often, phytoplankton distribution in the upper water column (i.e., mixed layer) is homogenized by mixing processes. For ST4, the only station on which CTD cast was performed, physical and optical data indicated that the mixed layer depth was 17.6 m (Fig. 1(c), data available at http://www.rvdata.us/catalog/AE1319), and low within-ML variability in attenuation (and density) suggests active mixing process. Similar or deeper mixed layer depth (up to ~50 m) during the period of our experiment was observed by four ARGO floats operating in that area (data not shown, available at http://www.ifremer.fr/lpo/naarc/#). Along-track variability in temperature and salinity, measured over 1 km centered on the sampling points, was smaller than the minimal variability in physical parameters known to cause significant changes in phytoplankton community structure [20]. These observations suggest that it is unlikely that vertical or horizontal heterogeneity of the phytoplankton distribution is the driver of the observed differences in our data. However, observed differences between DPS and surface sample at ST1 (13%) suggest that even in a homogenous physical environment (Fig. 1(c)), some variability in phytoplankton community will be present. Prior studies have demonstrated that the magnitude of such variability differs across ecosystems, spanning from minimal and statistically insignificant [21], to as high as a c.v. of 66% [18]. This literature based, high-end estimate of variability would indicate that our observed trend is insignificant, though low observed variability in physical environment suggests homogeneous distribution of phytoplankton in the sampled area. However, it is important to remember that observed variability in phytoplankton community (structure and abundance) cannot be explained solely by changes in environmental parameters, but biologically mediated processes as well, such as taxa specific growth and mortality ([22] and references therein).

Heterogeneity in the phytoplankton composition on these scales can be artificially driven by the presence of “rare species” in collected samples. Low sampling volumes, such as the ones we used in this experiment, lead to underestimation of the true proportion of “rare species” in the natural phytoplankton population [23], and these results introduce bias if rare species are present in replicate samples. In order to avoid this potential bias, we excluded the “rare species” (<1% sensu Rodríguez-Ramos et al. [23]) from our PSD. Such low sampling volumes can cause additional biases, especially in low biomass environments where it is hard to satisfy the statistical constraints of the desired confidence level [17]. When accuracy of our cell abundance estimates is taken in consideration (Table 1, Fig. 2 purple error bars), the difference in cell counts between DPS and IPS observed at ST3 becomes statistically insignificant, due to the high c.v. of 26% and 35%, respectively, however the trends for the other stations remain sound.

Ultimately, after taking into consideration the in situ spatial heterogeneity and potential method-based errors, the overarching trend of decline in cell counts from DPS to IPS system remains. Hence, we believe it is the specifics of the sampling systems, including differences in length and complexity of tubing, potential presence of filter-feeding organisms on the components of the permanently mounted systems, and the type of the pump used, are likely to be the primary driver for the observed differences in our data.

Our data suggest that the observed difference in cell counts is not dependent on the size of the organisms (Fig. 3), but is highly dependent on the type of the phytoplankton (Fig. 2). Ciliates are known to be highly sensitive to mechanical disturbance (such as centrifuging [24]), causing them to be the cells most affected by the choice of the flow-through systems. Photosynthetic ciliates are not considered an important part of the global phytoplankton community, although in some slope and shelf systems they tend to contribute substantially to local primary production [25]. As expected, diatoms, with their silica cell walls, were highly susceptible to breakage. Such breakage of diatom cells and chains was previously observed in samples collected from a similar IPS system by authors during the NAB08 experiment in 2008, and Dolcevita experiment in 2003 (unpublished data). However, our data does not demonstrate a clear trend when it comes to dinoflagellates. These phytoplankton are known to be susceptible to shear, as seen from their low abundances in highly turbulent parts of the ocean, and from experiments in bioreactors where shear can lead to the collapse of algal cultures ([26] and references therein). However, dinoflagellates’ response to shear is variable across the dinoflagellate size spectra, which might be a potential driver of the variable trend observed in our results. It is possible that the diaphragm pump, although it produces less shear than the impeller pump, is destructive enough to cause breakage of dinoflagellate cells.

Since no optical measurements were collected as part of this experiment, we used optical models to illustrate the potential impact of observed change in the PSD on the associated inherent optical properties. We calculated the particulate attenuation and backscattering due to phytoplankton (5-60 µm range) using the size distributions measured by FlowCAM (Fig. 5). In a way, results for both simulations replicate the trends observed in particle counts. All samples, including the surface sample in ST1, have significantly different attenuation than the other samples taken at that station (two tailed t-test, p <0.001). For backscattering, higher sensitivity of backscattering simulations to the change in the index of refraction resulted with larger uncertainty, which was visible in higher p values for sample to sample comparisons. Still, similar to the behavior observed for attenuation, trends mimic the ones observed in cell counts. In spite of this, no significant difference was found in backscattering modeled on samples collected at ST2 (two tailed t-test, p = 1.297). The fact that 55% decrease in phytoplankton counts at ST2 (DPS to IPS system) translated into 27% change in attenuation and 15% change in backscattering, underlines the importance of understanding how the change in PSD impacts the inherent optical properties. However, a limitation of our results is that this simulation only includes the change in PSD measured by FlowCAM, confined to chlorophyll containing particles within the 5-60 µm particle size range. Comparison with an in situ profile collected at ST4, suggest that the fraction of the PSD that we analyzed here contributes ~5% to the total particulate attenuation. Fractionation experiments conducted previously [27] and during this cruise (data not shown) demonstrate high variability in the contribution of large particles (>5 µm) to the total attenuation and backscattering signal, ranging from between 30% to 90%. This discrepancy can be explained by the PSD and type of particles analyzed in this study – suggesting that most of the bulk attenuation signal in our study is probably associated with detrital particles, and particles smaller than 5 µm, which is expected for a late summer, recycling type ecosystem.

 

Fig. 5 Optical properties calculated from the PSD data (5-60 µm), as proposed by Mie theory. Panel A depicts phytoplankton attenuation, and panel B backscattering coefficients. Error bars represent standard deviation of the calculated optical properties, for different indices of refraction. Color coding of the bars is same as in Figs. 3 and 4.

Download Full Size | PDF

If the impact of IPS system was the removal of phytoplankton from the evaluated size range only (e.g., size-selective filter-feeding organisms in the plumbing), then the resulting change in bulk attenuation signal would be minimal, ~2.5% (not accounting for potential impact on the detrital matter). The second hypothetical case is that the observed removal of the cells was due to the breakage of particles (e.g. via impeller pump). In that case, most of that particulate mass would be shifted towards the smaller size range (probably outside of detectable range), and the impact on inherent optical properties would be larger, and highly unpredictable due to the change in PSD, particle morphology, and index of refraction. Measurements of associated chlorophyll biomass suggest lesser impact of the sampling system on these parameters, and might speak in support of the latter hypothesis. Chlorophyll collected on a filter prior to measurement using the extracted fluorometric method would be mostly conserved even if the cells break during passage through the system (due to their position within the chloroplast membranes). HPLC pigment analysis, which is capable of distinguishing chlorophyll a from chlorophyll metabolites of similar fluorometric signature (such as chlorophyllide), might be able to detect the effects of the cell breakage. Chlorophyllide, a chlorophyll metabolite that is associated with diatom senescence, is also noted as a potential extraction artifact (due to the breakage of the cells during filtration [28]). Trends in phytoplankton carbon are similar to the trends observed in phytoplankton counts; however, the carbon concentrations were derived from FlowCam biovolumes, and are not considered an independent indicator. It is important to note that cell breakage (and associated inter-cellular carbon loss) could have a significant impact on the chemical measurements of the total or phytoplankton specific carbon biomass [29].

Large sampling profilers, such as CTD rosette, tend to cause turbulence that might lead to the breakage of large cells and organic aggregates, which ultimately may lead to a discrepancy in optical characteristics of the downward and upward measured profile, in a similar manner as described in our modeling experiment [30]. Smaller pumps, used to push/pull the water through an optical and imaging instrument, have been shown to alter the PSD in the samples filled with inorganic aggregates [9], which was reflected in associated optical properties. Then again, when comparisons between optical measurements collected with impeller-type flow-through system and vertical profiler (CTD rosette) are done on larger scales, impacts of the cell breakage are less significant [31].

The growing need for optical and biogeochemical data of higher spatial and temporal resolution, either for remote sensing or model validation (e.g. remote sensing based estimates of carbon export [32]), or other applications where high frequency data are needed (e.g., gas exchange [33], fine-scale plankton community structure [13]) is driving the increase in utilization of flow-through systems aboard research vessels. This study demonstrates the importance of increased diligence when it comes to selecting the sampling system. Although custom systems using lower-shear pumps and direct runs of clean tubing similar to ours have been utilized successfully in validation exercises [1], our results suggest that further exploration of other less aggressive pumping systems is advisable in order to maximize the quality of highly valuable optical and biogeochemical data sets obtained from ship-board flow-through systems.