A prototype shipboard filter skid (p1SFS) was designed and built to facilitate shipboard collection and concentration of ≥50 µm planktonic organisms from large volumes of water. The p1SFS consisted of two stainless steel filter housings, each containing a filter bag, arranged in parallel. Validation of the device examined the particle retention efficiency of filter bags (using inert polystyrene microbeads), the potential toxicity of filter skid materials and the capture efficiency of plankton collected with a filter skid versus a plankton net at two sample volumes (5 and 10 m3). Microbead recovery in filter bags was >89 and 100% for microbeads of 50 and 150 µm diameters, respectively. Exposure to the sealant used to close the filter bags' seams or stainless steel did not lead to mortality of two model zooplankton species. Overall, the concentration of ≥50 µm plankton in the p1SFS relative to concentrations in the plankton net (i.e. the capture efficiency, CE) was 108 ± 66% (mean ± 1 SD, n = 6). The p1SFS CE was higher in experiments with 5 m3 sample volume (147 ± 74%; n = 3) relative to experiments with a 10 m3 sample volume (69 ± 28%, n = 3), although the difference in CE between the sample volumes was not significant. Consequently, these experiments suggest this or similarly validated filter skids are appropriate for in-line sampling of plankton from relatively large volumes of water.
Planktonic organisms can be rare and unevenly or patchily distributed in aquatic environments (Wiebe, 1970). Consequently, relatively large volumes of water must be sampled to detect and quantify aggregated individuals or sparsely distributed organisms. Plankton nets, long-established devices for collecting and concentrating planktonic organisms, are designed for in situ sampling within the water column and are typically towed through the water (Wiebe and Benfield, 2003). A plankton pump is an alternative approach that uses mechanical pumps to drive water though the filtration area (Coughlan and Fleming, 1978). An advantage of plankton pumps is that the filtration unit can be easily accessible (i.e. on the deck of ships or docks) and they do not need to be towed through the water aboard a moving platform. Pump-driven sampling devices are ideal for in-line sampling from closed fluidics systems, such as those supplying cooling water reservoirs (Yocum et al., 1978) or ballast water tanks aboard ships.
In the past several decades, the link between transport of ballast water and the introduction of non-indigenous organisms has become increasingly evident (Carlton, 1985). Consequently, national and international standards have been proposed to limit the density of living organisms in discharged ballast water. One size class considered in these standards are organisms ≥50 µm in minimum dimension of the minor body axis, nominally plankton, such as microinvertebrates, large phytoplankton, heterotrophic protists, crustacean nauplii and larval stages of other invertebrates and fish. Proposed US and international standards stipulate that discharged ballast water should contain less than 10 living organisms ≥50 µm in minimum dimension per m3 of discharge water (IMO, 2004; US Coast Guard, 2009). Given the need to detect and quantify sparsely concentrated organisms, large volumes of sample water (e.g. cubic meters) must be examined to determine with statistical confidence whether the discharged ballast water meets the proposed standards to minimize the introduction of potentially invasive species (e.g. US EPA, 2010; Miller et al., 2011).
In the context of ballast water sampling onboard ships, plankton nets are cumbersome and difficult to handle. Deploying plankton nets through manhole covers limits the size of the plankton net and the volume of water that can be assessed, and such an approach, importantly, would not collect a sample that is representative of the entire volume of interest. When water is pumped through stationary plankton nets, the filtered water must be washed overboard, added to the ship's bilges or pumped back into the discharge line downstream of sampling. This arrangement, especially when using large volumes of water, would be unwieldy and potentially unsafe when sampling inside a vessel. To easily and efficiently filter the necessary volumes of water, a novel approach was taken: an array of closed-housing filtration units (a shipboard filter skid) was designed and constructed for the purpose of collecting large volume samples from water flowing through pipes (e.g. as would be found when moving ballast water into or out of a ship). Filter skids must be able to sample plankton at elevated pressures and rapid flow rates and must return effluent water to the piping system.
A prototype shipboard filter skid (p1SFS) consisting of two filter housings arranged in parallel was built, and its performance was evaluated at a land-based facility by determining the particle retention efficiency of filter bags (using inert polystyrene microbeads), evaluating the toxicity of filter skid materials, and comparing the density of live and dead plankton collected in the p1SFS to a plankton net at two sample volumes (5 and 10 m3). The overall goal of this work was to determine whether the p1SFS can be used as a substitute for plankton nets and can capture plankton efficiently without introducing high rates of plankton loss or mortality. The design and evaluation of an earlier version of the filter skid, with more housings used to filter larger volumes of water (i.e. 30–60 m3), is reported elsewhere (Drake et al., 2011).
All work was conducted at the Naval Research Laboratory (NRL) in Key West, FL, which is located on Fleming Key. The NRL facility consists of two full-scale, above-ground tanks (“treatment” and “control” tanks, 382 and 151 m3 capacity, respectively), a discharge tank (394 m3 capacity) and a piping system for filling and draining the tanks. Pressure and flow rates are monitored throughout the system with in-line sensors (GF Signet, El Monte, CA, USA), and physical data (e.g. flow rate, pressure) were collected every minute during sampling with an automated process control system (Experion PKS, Honeywell, Morristown, NJ, USA). Ambient seawater used for all experiments was pumped into the NRL seawater handling system through submerged intake pipes drilled with holes (184 holes, each 1.6 cm diameter). Water was collected from approximately 0.5 to 2.5 m below the surface, thus providing a depth-integrated sample of the majority of the water column at the intake (∼3 m).
The p1SFS was intended to sample water discharged from ships' ballast tanks, and therefore, was designed to meet several criteria. First, filtration of the water was to occur in a closed system that could be operated in confined spaces (e.g. a ship's engine or pump room). The p1SFS needed to operate at high pressure as the water filling and draining ballast tanks is moved at elevated pressures (∼30 kPa). In this sense, the p1SFS differed from other plankton pump units that operate at atmospheric pressure and use gravity to move water through the filtration membrane (Powlik et al., 1991). Secondly, the filter skid was to be constructed from readily available and durable components.
The p1SFS was constructed with two Eaton Topline™ housings (27 L volume, Eaton Co., Cleveland, OH, USA) arranged in parallel. Each filter housing weights approximately 42 kg; therefore, the device requires machinery for lifting and installation aboard a ship. Surfaces that would be exposed to seawater during sampling were composed of corrosion resistant steel (316 L stainless steel). Valves were added to piping upstream of each housing so that housings could be manually isolated, permitting the exchange of filter bags during sampling if necessary (Fig. 1). Absolute pressure was measured by in-line sensors at both the p1SFS inlet and outlet pipes (GF Signet; El Monte, CA, USA; resolution = 0.07 kPa). Pressure differentials (δP) were calculated as the difference between inlet and outlet pressure readings and were used to indicate resistance (i.e. clogging) in the filter skid.
Fig. 1.
Open in new tabDownload slideModel of the p1SFS, which consisted of two filter housings. Arrows show the direction of water flow; influent water was split into two parallel pathways and traveled to the top of the filter housing where it was distributed downwards into the main housing that contained a filter bag. Water was discharged from the bottom of the filter housings and plumbed into a waste stream.
Filter bags made of nylon monofilament mesh (18 cm diameter; 81 cm in length, Filter Specialists, Inc., Michigan City, IN, USA) were used to capture and collect ≥50 µm plankton from the flowing water. The filter bags' mouths had a plastic snap-ring that created a seal around the bags when they were seated properly in the filter housings. The tight seal formed between the filter bag and the housing prevented water from leaking around the filter bag (and losing organisms). Filter bags had a nominal mesh size of 35 µm; therefore, in theory, the largest object that would pass through a filter bag was 49.5 µm (the hypotenuse of the square mesh opening). Initial experiments with polystyrene microbeads (Lemieux et al., 2010) suggested that objects that should be retained in the mesh netting escaped through the seams of the filter bags. To prevent this loss, filter bag seams were glued with white 3M™ Marine Fast Cure 5200 Adhesive Sealant (3M, St Paul, MN, USA). The entire seam along the filter bag was coated in the sealant two times and allowed to cure for at least 1 day prior to use.
Microbeads (Chromosphere; Fisher Scientific, Pittsburg, PA, USA) were used as plankton proxies to determine the retention efficiency of ≥50 µm plankton. Spherical microbeads were composed of a cross-linked polystyrene divinylbenzene copolymer. Microbeads of two diameters (49 ± 1.5 µm and 150 ± 3.6 µm, coefficient of variation = 8 and 6%, respectively) were suspended in water purified by deionization and reverse osmosis (type II water). Experiments were performed using microbeads of only one diameter per test (i.e. microbead suspensions were not mixed sizes). Droplets of microbead suspensions were placed on a glass slide, and the microbeads in suspension were counted using a dissecting microscope (15–20× total magnification) and then rinsed into glass beakers to yield stocks with known quantities of microbeads (typically 100–600 per experiment). These microbeads were used to calculate the retention efficiency of the filter bags in both laboratory and field experiments.
Experiments were conducted in the laboratory by manually rinsing a microbead suspension using a squirt bottle filled with type II water into filter bags with either 25 or 35 µm mesh size (the hypotenuse of each mesh size is calculated as 35 µm and 49 µm, respectively). Microbeads retained within the filter bag were manually rinsed from the outside of the filter bag with a rinse bottle filled with either type II water or 0.22 µm filtered seawater to accumulate microbeads in the bottom of the filter bag. The filter bag was inverted, and the contents were rinsed into a beaker, yielding approximately 0.5 L of rinse water with suspended microbeads. The microbeads were counted as described below.
Preliminary field experiments were conducted with freshwater and a filter skid constructed of two fiberglass housings in series. The purpose of these experiments was to determine the retention of small objects in the filter bags during the typical operating conditions (i.e. high pressures and water flows). Preliminary trials showed that debris and microalgae from the piping system and the freshwater reservoir tank accumulated in the filter bags and obscured the small microbeads. Therefore, a 25 µm filter bag was placed in the first housing in series to act as a pre-filter to collect debris. A 35 µm mesh filter bag (40 cm in length) was placed in the second of two fiberglass filter housings arranged in series, and microbeads were rinsed into the second filter bag in the series with a rinse bottle filled with either type II water or 0.22 µm filtered seawater. Filter housings were closed, and 3.8 m3 of municipal freshwater was filtered through the housings at 95 L min−1. The typical stainless steel housing was not used in these preliminary field experiments, since it would have required an extensive modification of the facility to flow clean, freshwater through the skid. After the filtration, the second filter bag in the series, which contained the microbeads, was removed from the housing and processed following the procedure for the laboratory experiments.
The entire sample was analyzed by dispensing subsamples into serpentine Bogorov counting chambers (15 mL capacity) and examining the chambers using a dissecting microscope (20–30) with darkfield illumination. All materials that contacted the microbead suspension (e.g. pipettes, beakers, filter bags) were examined using the microscope to verify that microbeads were not lost during fluid transfers. Microbeads remaining in the laboratory equipment, if found, were added to the total count of microbeads retained in the filter bag. The retention efficiency of the filter bags was calculated as the number of microbeads recovered compared with the number of microbeads added to the filter bag.
The p1SFS was designed to collect samples to provide estimates of living plankton in the sample water. Therefore, it is critical that the sampling environment must not cause plankton mortality. We investigated potential toxic effects of two components of the filter skid: the sealant used on the filter bags' seams and the stainless steel housings.
Both A. franciscana (Brine Shrimp Direct, Ogden, UT, USA) and copepods collected from ambient seawater were used to determine whether exposure to the filter bag sealant (3M™ 5200) resulted in plankton mortality. Artemia franciscana were hatched from cysts incubated in aerated, artificial seawater (ASW, salinity = 36; Instant Ocean®; Aquarium Systems, Inc., Mentor, OH, USA) for 24 h in 25°C with a 12:12 light:dark cycle under fluorescent bulbs (72 µM Einsteins m−2 s−1). The hatched nauplii (300–400 µm long) used in these experiments were approximately 12 h old. Nauplii were removed from the culture in 1 mL aliquots and dispensed into each of eight Petri dishes (47 mm diameter) with 10 mL of 0.22 µm filtered seawater (FSW). Next, the nauplii were examined using a dissecting microscope to verify that all organisms were living (determined by movement). A small strip of the 25 µm filter bag (approximately 4 × 1 cm) with sealant (a 3 × 0.5 cm bead) that had been cured for 24 h was placed in each of four Petri dishes. Four Petri dishes without sealant served as controls. Samples were incubated for 2 h, the approximate amount of time plankton would be sequestered in a filter bag during collection. After the incubation, samples were counted to measure concentrations of live and dead A. franciscana. Organisms were scored as dead if they were not moving and did not respond to gentle prodding with a metal probe within 10s. Following the tally, all samples were fixed with Lugol's iodine solution and the total number of A. franciscana was counted. The concentration of living organisms was the difference between total and dead organisms. The experiment was repeated with ambient copepods, which were concentrated from a seawater hose onto a 31 µm sieve. Copepods of all life history stages (≥ 50 µm in minimum dimension) were removed from the sample individually using a pipette. Copepods (20 per Petri dish) were subjected to the same treatment and analysis described above.
A similar bioassay was conducted to verify that exposure to the stainless steel filter housing does not induce plankton mortality during sampling periods. Stainless steel is composed of heavy metals (e.g. chromium), which can have acute toxic effects on plankton (Hollibaugh et al., 1980). In this experiment, A. franciscana and ambient zooplankton were used to measure mortality of zooplankton during the short-term exposure to stainless steel. Artemia franciscana were prepared from cysts as described above; all nauplii (70–77 per subsample) were moving (i.e. living). Nauplii were added to 800 mL of 0.45 µm filtered seawater. Ambient zooplankton (primarily copepods) were concentrated from a seawater hose onto a 37 µm sieve. The filtrand (material collected on the sieve) was suspended in 0.45 µm filtered seawater, and both live and dead ≥50 µm zooplankton were counted as described above. A mixture of A. franciscana (200–400 individuals) and ambient zooplankton (20–40 individuals) was added to five beakers containing 800 mL of 0.45 µm FSW. To each of three beakers, 10 washers (grade 316 stainless steel, 1.9 mm thick and 38.1 mm in diameter with a hole 15.8 mm in diameter in the center) were added. Ten washers per beaker were used because their cumulative surface area (SA) in the beaker was calculated to equal the SA:volume ratio of water in a filter skid housing composed of stainless steel. Two beakers with only zooplankton served as controls. Beakers were arranged haphazardly on a shelf and kept in the dark in the incubator (described above) for 2.5 h. This incubation time is 0.5 h longer than incubation times used for to test the sealant and is the maximum time zooplankton would be sequestered in a stainless steel filter housing during collection. After the incubation, the washers were removed from the treatment beakers using forceps, and the contents of each beaker were gently poured through a 37 µm sieve, the filtrand rinsed with filtered seawater into a beaker and the entire volume transferred to a Petri dish. Numbers of live and dead zooplankton were determined as described above.
The relative effectiveness of the p1SFS in capturing ≥50 µm plankton was determined by comparing plankton collected in the p1SFS to plankton collected using a plankton net (PN, Sea-Gear, Co. Melbourne, FL, USA). Sample volumes for these experiments were either 5 or 10 m3; three replicate trials were performed for each volume. Two centrifugal pumps (30 hp) filled the tank at a rate of 3780 L min−1 through polyvinyl chloride (PVC, both 15 and 20 cm diameter) piping. In this system, sample ports were located upstream and downstream of the treatment tank to sample water during fill and drain operations, respectively. The sample port used in validation experiments described here consisted of a PVC wand (4.4 cm) with a 90° bend installed with the opening in line with flow of water in the main line. This configuration was consistent with previously determined guidance to ensure representative sampling of the flow (Richard et al., 2008).
Total water flow rates were 95 and 190Lmin−1 for experiments sampling 5 and 10 m3 of seawater, respectively. The flow rate was controlled by a diaphragm valve placed downstream of the sample port (but upstream of the p1SFS or PN), and the flow rate was measured with a magnetic flow meter. The maximum operating flow rate of 95 L min−1 per filter housing (equivalent to 190 L min−1 for the two housings of the filter skid) was determined to be the maximum flow rate; higher flow rates were found to yield large pressure differentials across the filter housings (Lemieux et al., 2010). Therefore, all of these experiments used flow rates less than or equal to the maximum flow rate.
Either the p1SFS or the PN was used to sample ambient seawater as the tank was filled. Both sampling devices were supplied with seawater from the sample port via a flexible hose. The tank was drained immediately (no samples were collected during the tank drain), and the other sampling device would be used during the subsequent tank fill. Sampling occurred during the fill operation to minimize stress on plankton, as sampling during the tank drain operation would expose plankton to additional sources of mortality (e.g. discharge pump impellers, time in the tank, etc.). Within a given experiment, sampling events using the p1SFS or PN were started within 3 h of each other to minimize variations in the plankton community due to tidal flow or currents. To minimize any difference in plankton communities resulting from changes over tidal cycles that occurred during an experiment, the order in which the p1SFS and PN were used to sample seawater was varied. Because the intake pipes to the seawater system at NRL collected water throughout the majority of the water column, diel vertical migrations of plankton were considered unlikely to result in differences in plankton concentrations between sampling events.
Sample analysis consisted of counting and classifying (to the lowest possible taxon) the plankton captured in either the p1SFS or PN. After 5 or 10 m3 was filtered through the p1SFS, the contents of each filter bag were gently rinsed into the bottom of the filter bag using a spray nozzle attached to a hose supplying seawater. The filter bags were rinsed from the outside to prevent the introduction of plankton from the rinse water into the sample. Each filter bag remained in its filter housing, submerged in seawater, until it was rinsed; care was taken to rapidly process the filter bags and thus minimize temperature changes. The filter bag was inverted and the contents deposited into a dry, clean plastic bucket. Material remaining in the filter bag was rinsed from the mesh using filtered seawater (<1.5 µm, FSW) prepared the day prior to experiments. The contents of all filter bags were pooled into the plastic bucket and transferred into a graduated flask, yielding between 0.5 and 2 L of concentrated plankton in seawater. If the volume of the pooled sample was >2 L, the sample was further concentrated on a 35 µm mesh sieve. All surfaces in contact with the concentrated sample (e.g. filter bags, the bucket and the sieve) were rinsed three times with 1.5 µm FSW to ensure plankton were not lost during the sampling process. Plankton net samples were collected in a similar fashion. The entire net was rinsed and the collecting cup (i.e. the cod end) was removed so plankton accumulated in the folds connecting the cod end to the mesh netting could be rinsed into the bucket and concentrated as above.
The concentrated sample from the p1SFS or PN was subsampled by capping and inverting the sample flask five times prior to removing a 1 mL sample with a serological pipette. This aliquot was then deposited in a 15 mL plastic conical tube. The process of mixing and sampling was repeated until 5 mL of sample was deposited in the tube. This process was performed five times to yield five subsamples representative of the sample for analysis.
An initial microscope count of plankton was performed using a Sedgewick Rafter counting chamber to determine whether the subsample aliquots required dilution prior to processing. Briefly, ≥50 µm plankton were counted from a portion of the Sedgewick Rafter chamber (1 mL total volume). Microbeads (49 µm diameter; Chromosphere; Fisher Scientific) were added to the chamber as a size reference. If the concentration of plankton in the concentrated sample was too dense to count accurately (determined from previous work to be >30 individuals mL−1), subsamples were diluted with FSW. When needed, samples were diluted to either 10 or 50% of the initial subsample concentration.
Samples were counted in Bogorov chambers (4 mL capacity, which is a smaller volume than was used for counting microbeads for the filter bag retention experiments) using light microscopy (20–30× total magnification). Microbeads were added as size reference; typically, >10 microbeads were visible in every field of view throughout the chamber. The entire chamber was scanned and all organisms ≥50 µm in minimum dimension were counted. Organisms were classified into major taxonomic groups (e.g. copepods, nauplii, ciliates, diatoms, etc.), and both live and dead organisms were tallied. Live organisms were identified by motion, and if an individual was not moving, a small metal probe was used to gently touch the organism. An organism was classified as living if this stimulus elicited movement within 10s. Some organisms (e.g. diatoms) would not be expected to respond to prodding and, therefore, were classified as living if cellular structures such as frustules and chloroplasts appeared intact.
−1) in the sample was calculated by the equation:(1)
−1), D the sample dilution factor (e.g. 0.1 for 10% of the initial concentration), C the concentrated sample volume (mL, e.g. 700 mL) and S the total sample volume (L, e.g. 5000–10 000L).At least five subsamples were analyzed for each concentrated sample. The concentration of ≥50 µm plankton (P; individuals L) in the sample was calculated by the equation:where A is the aliquot concentration counted in the Bogorov chambers (ind. mL), D the sample dilution factor (e.g. 0.1 for 10% of the initial concentration), C the concentrated sample volume (mL, e.g. 700 mL) and S the total sample volume (L, e.g. 5000–10 000L).
To determine whether the p1SFS restructured the plankton community by selectively retaining certain taxa, the concentrations of groups of abundant organisms were recorded in ambient grab samples (described in the next paragraph), the p1SFS samples and PN samples. Four broad taxonomic groups (crustacean nauplii, adult copepods, ciliates and diatoms) typically represented >90% of the total concentration of plankton. Concentrations in each of the groups in samples collected using the p1SFS and PN were normalized to concentrations collected in ambient grab samples.
Ambient grab samples were collected from the seawater surface ∼1 m away from the seawater intake pipes. Four grab samples (5 L each) were filtered through a 35 µm sieve to concentrate plankton. These samples were processed and analyzed as described above. Only live plankton were considered in this analysis.
(2)
Concentrated samples from the p1SFS and PN were time-integrated (i.e. collected over the entire period of tank fill) and were representative of the entire volume used to fill the tank. The subsamples collected from concentrated samples (p1SFS or PN) represented analytical replicates and measured the precision of sample analysis. Mean values and error measurements (standard deviations) were calculated from replicate experiments. The performance of a filter skid was measured by capture efficiency (CE, %): the concentration of plankton determined with the filter skid (P) divided by the concentration determined with the plankton net (P).
(3)
Dead and PLive are concentrations of dead and live plankton, respectively. Mortality was also measured in grab samples collected from the seawater surface and concentrated on a sieve (as described in the previous section). Although this process of collecting and concentrating plankton arguably causes some level of mortality, it does not expose organisms to potentially detrimental conditions in the piping system, such as hitting pump impellers, experiencing sheer stress and turbulence and undergoing rapid changes in direction. Therefore, evaluating the number of dead plankton in the grab samples provided a baseline in the ambient seawater at the point and time of uptake. For calculations involving multiple variables with errors (such as CE or mortality), the error in variables was propagated using standard formulas. A t-test (two samples assuming unequal variance, α = 0.05) was used to determine whether differences between groups or experiments were significant. Percent mortality and CE data were arcsine transformed prior to performing statistical comparisons.Plankton mortality (M, %, i.e. the relative abundance of dead plankton) in the filter skid or plankton net was a measurement of the total number of dead plankton and was calculated by the following equation:where Pand Pare concentrations of dead and live plankton, respectively. Mortality was also measured in grab samples collected from the seawater surface and concentrated on a sieve (as described in the previous section). Although this process of collecting and concentrating plankton arguably causes some level of mortality, it does not expose organisms to potentially detrimental conditions in the piping system, such as hitting pump impellers, experiencing sheer stress and turbulence and undergoing rapid changes in direction. Therefore, evaluating the number of dead plankton in the grab samples provided a baseline in the ambient seawater at the point and time of uptake. For calculations involving multiple variables with errors (such as CE or mortality), the error in variables was propagated using standard formulas. A t-test (two samples assuming unequal variance, α = 0.05) was used to determine whether differences between groups or experiments were significant. Percent mortality and CE data were arcsine transformed prior to performing statistical comparisons.
Filter bags were modified by sealing the bag seams, which resulted in high microbead retention efficiencies in both laboratory and field experiments. Microbeads with 150 µm diameters were completely recovered in both laboratory and field experiments (Table I). While the recovery of smaller microbeads (50 µm in diameter) was lower, a mean recovery of 89 ± 3% of all microbeads were recovered in the field experiments with 35 µm mesh filter bags (n = 2). Recovery rates of 50 µm diameter microbeads were higher for 25 µm filter bags (95 ± 4%; n = 3) than 35 µm filter bags (89 ± 3%; n = 2). However, the difference in recovery between bags with different mesh sizes was not significant.
Table I:
Experimenttype.
Filter bag mesh(nominal size, µm).
Microbeaddiameter (µm).
Retentionefficiency (%).
n.
Laboratory 25 50 90 ± 7 15 Laboratory 35 150 100 1 Field 25 50 95 ± 4 3 Field 35 50 89 ± 3 2 Field 35 150 100 ± 0 2 Experimenttype.
Filter bag mesh(nominal size, µm).
Microbeaddiameter (µm).
Retentionefficiency (%).
n.
Laboratory 25 50 90 ± 7 15 Laboratory 35 150 100 1 Field 25 50 95 ± 4 3 Field 35 50 89 ± 3 2 Field 35 150 100 ± 0 2 Open in new tabTable I:
Experimenttype.
Filter bag mesh(nominal size, µm).
Microbeaddiameter (µm).
Retentionefficiency (%).
n.
Laboratory 25 50 90 ± 7 15 Laboratory 35 150 100 1 Field 25 50 95 ± 4 3 Field 35 50 89 ± 3 2 Field 35 150 100 ± 0 2 Experimenttype.
Filter bag mesh(nominal size, µm).
Microbeaddiameter (µm).
Retentionefficiency (%).
n.
Laboratory 25 50 90 ± 7 15 Laboratory 35 150 100 1 Field 25 50 95 ± 4 3 Field 35 50 89 ± 3 2 Field 35 150 100 ± 0 2 Open in new tabThere was no apparent, negative effect of the 3M™ 5200 sealant on A. franciscana or ambient copepods. Specifically, there were no significant differences in the percentage of living zooplankton between control and treatment groups exposed to the sealant in either A. franciscana or copepod experiments (Fig. 2A). Similarly, there was no significant difference in the percentage of living zooplankton in control or treatment groups exposed to stainless steel (Fig. 2B).
Fig. 2.
Open in new tabDownload slideZooplankton response to exposure to the sealant used for filter bag seams (A) and to stainless steel (B). Two groups of zooplankton were examined: brine shrimp, A. franciscana, and ambient zooplankton, represented mostly by copepods. The dotted line indicates all zooplankton were living. Bars represent the mean of three replicate experiments; error bars show one standard deviation.
The quantities of live ≥50 µm plankton collected in the p1SFS relative to plankton nets (i.e. CE) was the critical metric used to evaluate whether filter skids are suitable sampling devices. The p1SFS captured greater quantities of living plankton than the PN for 5 m3 sample volume experiments: the CE for 5 m3 comparisons was 147 ± 74% (Fig. 3). At the larger 10 m3 sample volumes, the p1SFS had a lower CE (69 ± 28%). Neither CE was significantly different from 100% (one sample t-test, P > 0.05). Mortality was not significantly different between the p1SFS and the PN at either sample volume (Fig. 4; t-test, P > 0.05). Furthermore, mortality observed in the samples collected in the p1SFS and the PN (2.8 ± 1.5 and 2.8 ± 3.5%, respectively, for both volumes, n = 3) was not significantly different from mortality in the ambient grab samples (2.6 ± 2.8%; n = 26).
Fig. 3.
Open in new tabDownload slideCapture efficiency (% CE) of the shipboard filter skid (p1SFS) relative to a plankton net (PN) for experiment with sample volumes of 5 and 10 m3. Values greater than 100% (indicated by the dotted line) indicate ≥50 µm plankton concentrations greater in the p1SFS than the PN. Bars represent the mean of three replicate experiments; error bars show one standard deviation.
Fig. 4.
Open in new tabDownload slidePlankton mortality measured in samples collected from the shipboard filter skid (p1SFS) and the plankton net (PN). The dotted line indicates the mean abundance of dead plankton collected from surface grab samples (n = 26). Bars represent the mean of three replicate experiments; error bars show one standard deviation. The “Combined data” represent the grand mean of all p1SFS and PN trials using both sample volumes (n = 6).
Mean pressure differentials (δP) across the p1SFS were calculated from pressure measurements throughout each experimental trial; measurements were collected every minute throughout the sampling event, yielding 48–52 data points for each trial. Mean δP were significantly lower (t-test, P< 0.05) in the 5 m3 replicate trials (mean δP = 3.1 ± 0.8 kPa; n = 3) compared to the 10 m3 replicate trials (mean δP = 10.7 ± 0.9 kPa, n = 3). Increases in δP were noted in several of the replicate trials, and in some instances, they were coincident with low CE (Fig. 5). For example, trial 3 of the 5 m3 experiments and trial 3 of 10 m3 experiments, both trials exhibited high δP towards the end of the sampling event (final δP = 17.1 and 22.3 kPa, respectively; Fig. 5), and both trials yielded CE values <65%.
Fig. 5.
Open in new tabDownload slidePressure differentials (δP) measured throughout the sampling event and CE (capture efficiency) of experiments with the shipboard filter skid (p1SFS) using 5 m3 (A) and 10 m3 (B) sample volumes. Sampling events were approximately 50 min (range 48–52 min); absolute pressure readings were recorded every minute by pressure sensors located at the p1SFS inlet and outlet pipes. Pressure differentials were calculated as the difference between the inlet and outlet sensors. Negative δP and δP less than the sensitivity of the sensors (0.07 kPa) were set to zero.
Selective retention of plankton groups was examined by comparing the concentrations of major groups of plankton in the p1SFS and PN to ambient grab samples collected on the same day of the experiment. Plankton groups commonly encountered in these samples were copepod nauplii, adult and juvenile copepods, loricated ciliates and diatoms of various morphotypes (including both centric and pennate forms). Concentrations in the p1SFS and PN were only significantly different from concentrations in ambient grab samples in one instance, for ciliates in the experiments using 10 m3 volumes (Fig. 6; t-test, P < 0.05). In this case, ciliates in the p1SFS were 12 ± 3% of the ambient grab samples, and ciliates in the PN were 55 ± 26% of the ambient grab samples.
Fig. 6.
Open in new tabDownload slidePercentage of living plankton in abundant groups in sample volumes of 5 m3 (A) and 10 m3 (B) relative to ambient grab samples. Concentrations found in the shipboard filter skid (p1SFS) and plankton net (PN) were normalized to concentrations in ambient seawater. The four groups represented, on average, 91% of total plankton in the p1SFS and PN samples. The dotted line indicates the same percentage of organisms as found in ambient grab samples. Symbols represent the mean of three replicate experiments; error bars show one standard deviation.
The design features of plankton nets (PN) render them well suited to capturing and concentrating plankton in situ. The disadvantage of using plankton nets on land or onboard ships is that, for proper operation and water flow, the filter area must be submerged to ensure organisms remain suspended in water throughout the sampling period. Thus, a container of water with a volume greater than the volume of the plankton net is needed; the container is open to the atmosphere and thus, onboard a vessel, would have a “free surface” of water, which could potentially overflow. For instance, turbid water can slow the flow through the plankton net and result in water rapidly overflowing the net. Likewise, if used on a pitching or rolling vessel, the container could potentially overflow. This event might present a safety or operational hazard on a ship, and such a loss of concentrated plankton will nullify the sampling event. The p1SFS, on the other hand, was designed specifically for ex situ sampling of water as it is transported in a closed piping system. In this manner, the requirement for an open vessel of water and the risk of losing part of the sample are ameliorated. The p1SFS, however, must concentrate plankton (without inducing autogenous mortality) from water at pressures and flow rates that are dependent upon the conditions in the piping system to which it is attached.
This study evaluated the suitability of filter skids for sampling living plankton from piping systems (including those used to sample ships' ballast water), and the materials unique to the p1SFS (i.e. sealant used in filter bags' seams and stainless steel housings) did not appear to affect the survivability of plankton. The p1SFS incorporated refinements from previous versions of a filter skid designed for use at land-based ballast water test facilities (Lemieux et al., 2010). Initial work with a similar filter skid suggested that objects were lost in the filter bag seams, and applying sealant to the filter bag seams greatly improved the retention of plankton proxies (microbeads). In the experimental trials reported in this study, >89% of the microbeads washed into the filter bag were recovered.
We note that microbeads are imperfect replicas of plankton because they do not have spines or setae, which are typical of many zooplankters, and these features likely decrease their retention within nets relative to live plankton. On the other hand, microbeads are likely less compressible than plankton, so they would be better retained in nets in this regard. Nonetheless, they do provide a first-order approximation of ≥50 µm plankton retention in nets.
To be suitable for use on ships, the p1SFS capture efficiency of living plankton must be at least as great as that of a PN, the standard sampling device with a long history of use, including efforts to test the efficacy of shipboard ballast water treatment systems worldwide (e.g. Gollasch et al., 2003). In these experiments, when the relative concentration of plankton in the p1SFS was compared to concentrations in the PN, the p1SFS outperformed the PN in experiments using small volumes of water (5 m3). That is, plankton concentrations in the p1SFS were higher than in the PN. In contrast, the p1SFS showed lower capture efficiency than the plankton net in experiments using large volumes of water (10 m3).
The lower capture efficiency of the p1SFS during high-volume experiments was primarily due to poor retention of ciliates. In these cases, ciliates were similar in size to the hypotenuse of the filter bags' mesh, and it is possible that the higher pressure differentials in the p1SFS during 10 m3 trials forced the relatively delicate ciliates through the filter bags' mesh. Evidence of poor retention of ciliates is known to occur in monofilament mesh netting (Carrias et al., 2001). The malleable cell membrane likely allows the cells to pass through constricted openings. In contrast, microinvertebrates (nauplii and adult copepods) were well retained in the p1SFS. The complex and semi-rigid body structure (e.g. setae and antennae) probably prevent easy passage through constricted mesh opening and contribute to high retention efficiencies. Furthermore, higher sample volumes introduce more particles and lead to net clogging that, in turn, leads to high pressure differentials and the potential loss of malleable organisms with sizes near the nominal mesh size. The coincidence of high pressure differentials and low CE in replicate trials suggests that the p1SFS is most efficient at low pressure differentials. If it is possible to reduce the sample volume per filter bag, the cascading effects resulting from high sample volumes (i.e. mesh clogging, high pressure differentials and plankton loss) can be prevented.
While the overall sampling efficiency of the p1SFS was comparable to the PN, modifications to the p1SFS may improve its retention of plankton. The p1SFS can be optimized to target certain planktonic groups by varying two parameters: the mesh size of the filter bags and the total sample volume. This flexibility permits focused sampling of certain groups or subpopulations based upon experimental or observational goals (Omori and Hamner, 1982). For example, filter bags with mesh sizes smaller than 35 µm may capture smaller organisms (e.g. microplankton) with higher efficiency. However, smaller mesh size filter bags will retain more particles and can lead to high pressure differentials. Therefore, their value in increasing the retention of smaller plankton will depend upon the amount of suspended materials in the sample water. For sampling targeting soft-bodied organisms, such as rotifers (e.g. Chick et al., 2010), smaller mesh sizes and sample volumes may be required. Monitoring pressure in the p1SFS influent and effluent lines, which can alert operators to high pressure differentials, will be especially important for operating the p1SFS and may provide useful cues for switching or replacing bags.
The data from the experiments presented here show that with the appropriate operation and oversight, the filter skid described can be a valid substitute for plankton nets for concentrating living plankton. In the current case, this involved sampling plankton from piping systems, such as those used to fill and drain ballast tanks. Additional work is underway to validate the p1SFS aboard a vessel, under operational shipboard piping system pressures and flow rates.
This work was supported by the United States Coast Guard (USCG) Environmental Standards Division (CG-5224) (HSCG23-09-X-MMS028, HSCG23-10-X-MMS192) and does not represent official USCG policy. The work conducted at the Naval Research Laboratory (NRL) facility was supported by Diane Lysogorski.
We are grateful to Dr Richard Everett for advice and guidance with this work. Elizabeth Schrack assisted with data collection. Barron Stringham and William “B.J.” Kinee assisted with the construction and deployment of the gantry supporting the plankton net. Luke Davis assisted with the set up of the data logging capabilities for the ballast water test facility's water control system. We appreciate all of these efforts.
Carlton
J. T.
Transoceanic and interoceanic dispersal of coastal marine organisms: the biology of ballast water
,Oceanogr. Mar. Biol. Ann. Rev.
,1985
, vol.23
(pg.313
-371
), vol.(pg.
Carrias
J. F.
Thouvenot
A.
Amblard
C.
Dynamics and growth estimates of planktonic protists during early spring in Lake Pavin, France
,Aquat. Microb. Ecol.
,2001
, vol.24
(pg.163
-174
), vol.(pg.
Chick
J. H.
Levchuk
A. P.
Medley
K. A.
Underestimation of rotifer abundance a much greater problem than previously appreciated
,Limnol. Oceanogr. Meth.
,2010
, vol.8
(pg.79
-87
), vol.(pg.
Coughlan
J.
Fleming
J. M.
A versatile pump-sampler for live zooplankton
,Estuaries
,1978
, vol.1
(pg.132
-135
), vol.(pg.
Drake
L. A.
Lemieux
E. J.
Robbins-Wamsley
S. H.
Validation of filter skids to capture viable organisms ≥ 50 µm in minimum dimension (nominally zooplankton) for evaluating the biological efficacy of ships’ ballast water management systems
,2011
Washington, DC
Gollasch
S.
Rosenthal
H.
Botnen
H.
Species richness and invasion vectors: sampling techniques and biases
,Biol. Inv.
,2003
, vol.5
(pg.365
-377
), vol.(pg.
Hollibaugh
J. T.
Seibert
D. L. R.
Thomas
W. H.
A comparison of the acute toxicities of ten heavy metals to phytoplankton from Saanich Inlet, B.C., Canada
,Estuarine Coastal Mar. Sci.
,1980
, vol.10
(pg.93
-105
), vol.(pg.
International Maritime Organization (IMO).
Convention BWM/CONF/36 ‘International Convention for the Control and Management of Ships’ Ballast Water and Sediments, 2004
,2004
Lemieux
E. J.
Wier
T. P.
Steinberg
M. K.
Design and preliminary use of a commercial filter skid to capture organisms ≥50 µm in minimum dimension (nominally zooplankton) for evaluating ships’ ballast water management systems at land-based test facilities
,2010
Washington, DC
Miller
A. W.
Frazier
M.
Smith
G. E.
Enumerating sparse organisms in ships’ ballast water: why counting to 10 is not so easy
,Environ. Sci. Technol.
,2011
, vol.45
(pg.3539
-3546
), vol.(pg.
Omori
M.
Hamner
W. M.
Patchy distribution of zooplankton: behavior, population assessment and sampling problems
,Mar. Biol.
,1982
, vol.72
(pg.193
-200
), vol.(pg.
Powlik
J. J.
St. John
M. A.
Blake
R. W.
A retrospective of plankton pumping systems, with notes on the comparative efficiency of towed nets
,J. Plankton Res.
,1991
, vol.13
(pg.901
-912
), vol.(pg.
Richard
R. V.
Grant
J. F.
Lemieux
E. J.
Analysis of Ballast Water Sampling Port Designs Using Computational Fluid Dynamics
,2008
Groton, CT
U.S. Coast Guard Research and Development Center
US Coast Guard.
Standards for Living Organisms in Ships’ Ballast Water Discharged in U.S. Waters (Notice of Proposed Rulemaking)
,2009
, vol.74
pg.44632
, vol.pg.
US EPA.
Generic Protocol for the Verification of Ballast Water Treatment Technology, Version 5.1
,2010
Washington, DC
United States Environmental Protection Agency Environmental Technology Verification Program
Wiebe
P. H.
Small-scale spatial distribution in oceanic zooplankton
,Limnol. Oceanogr.
,1970
, vol.15
(pg.205
-217
), vol.(pg.
Wiebe
P. H.
Benfield
M. C.
From the Hensen net toward four-dimensional biological oceanography
,Prog. Oceanogr.
,2003
, vol.56
(pg.7
-136
), vol.(pg.
Yocum
W. L.
Evans
M. S.
Hawkins
B. E.
A comparison of pump sampling systems for live zooplankton collection
,Hydrobiologia
,1978
, vol.60
(pg.199
-202
), vol.(pg.
© The Author 2012. Published by Oxford University Press. All rights reserved. For permissions, please email: journals.permissions@oup.com
© The Author 2012. Published by Oxford University Press. All rights reserved. For permissions, please email: journals.permissions@oup.com