Read the instrument as a counter with a classifier, not a microscope
Marine flow cytometry passes particles or cells one by one through an illuminated measurement region, records scattered light and fluorescence, and groups events using chosen electronic and analytical gates. A field table might report Synechococcus, Prochlorococcus, picoeukaryotes or total autofluorescent cells in cells per millilitre. These numbers are not automatically species counts. They are counts of events assigned to operational populations by an instrument, a protocol and a gating scheme. The distinction matters because fluorescence strength, cell size, preservation and detector thresholds can alter whether an event is counted.
In an ordinary explanation, the reader needs four facts before interpreting a change: which light channels were measured, how gates were drawn, what sampled volume was analysed and which quality checks were passed. A plot of red fluorescence against side scatter may show clusters, but a cluster is not a named biological species until suitable evidence supports that identification. Some small oceanic cyanobacteria have weak red fluorescence, making them hard to detect if the trigger threshold is high. Other cells may be distinguished partly by orange fluorescence associated with phycoerythrin. The exact assignment depends on instrument lasers, optical filters and sample handling.
For example, an instrument may record 8,000 candidate events during a measured 0.40 mL run. If all events belong to a validated target gate and the sample was undiluted, that corresponds to 20,000 cells mL⁻¹. But if only half of the events belong to that gate, the population estimate is 10,000 cells mL⁻¹. If the counted volume was not actually measured or calibrated, the same event total cannot support either concentration without a valid reference. A clear article therefore keeps event number, sampled volume, gate membership and reported concentration separate.
What scatter and fluorescence really tell us
Forward-scattered and side-scattered light often correlate with particle properties such as size, shape, refractive index and internal structure. They are not direct measurements of physical diameter without calibration and assumptions. Two cells of similar diameter can scatter differently because their internal composition or optical contrast differs. Fluorescence channels report light emitted within selected spectral bands under selected excitation, not the total pigment content of each cell. A red channel can help detect chlorophyll-bearing cells; an orange channel can help distinguish phycoerythrin-containing groups under an appropriate optical configuration.
A familiar oceanic example separates Synechococcus-like events from other small phototrophs using orange fluorescence, red fluorescence and scatter together. Prochlorococcus can be particularly dim in red fluorescence, and populations from different depths may behave differently. Picoeukaryotic phototrophs can produce brighter red signals and different scatter distributions. Such group labels should be written as operational classifications unless the method demonstrates taxonomic specificity. A reader should not turn a cloud of optical events into a complete plankton census or infer that all uncounted cells are absent from the water.
A flow cytometer’s event plot is also shaped by electronic scaling. Many fluorescence axes are logarithmic; a one-unit movement on that axis may represent a multiplicative change, not a one-unit difference in light. Gates built around a reference population must be checked after instrument gain changes and between field campaigns. Colour names such as “green” or “red” can refer to channels with different bandpasses on different devices. Store the laser wavelength, detector filter and compensation or spectral handling method. Without these details, comparing fluorescence intensities between laboratories may be misleading even when both tables say “red”.
How gates become numbers
A gate is a defined region in one or more measured dimensions. It might include cells with above-background red fluorescence and scatter within a chosen range while excluding non-fluorescent debris, doublets or beads. Analysts often use sequential gates: a broad event-quality gate, then a phototroph gate, then narrower biological groups. Each step changes the denominator. Saying that a population makes up 25 percent of all events is different from saying it makes up 25 percent of phototrophs after debris exclusion. Both may be mathematically correct, but their ecological meanings differ.
Suppose a sample contains 5,000 accepted phototroph events, with 2,000 assigned to a Synechococcus-like gate, 1,500 to a picoeukaryote-like gate and 1,500 to other accepted phototrophs. The Synechococcus-like fraction among phototroph events is 2,000/5,000 = 40 percent. It would be wrong to report 40 percent of every particle in the water if the original dataset also contains uncounted heterotrophic cells, detritus and below-threshold particles. The gate definition and denominator must travel with the percentage.
Manual gates can be reproducible if their boundaries and rules are preserved. Automated clusters also require validation; an attractive machine-generated separation does not guarantee biological identity. It is valuable to retain representative event plots, reference beads and a versioned gating template. If a new run needs a substantially shifted gate to recover a population, investigate gain, optics, preservation, background and real community differences before simply dragging the boundary to force agreement.
Counting beads, measured volume and dilution factors
Some flow cytometers measure sample volume directly or through a calibrated volumetric flow and run duration. Other methods infer analysed sample volume with added counting beads of known concentration. These are distinct routes to the same desired denominator. If a sample contains a known number of beads per millilitre and the instrument detects both beads and target cells under validated assumptions, a target concentration can be estimated from the cell-to-bead event ratio multiplied by bead concentration, with any sample dilution properly applied. Bead concentration itself must be verified because preparation and storage can affect standards.
As a teaching calculation, a sample prepared with a known final bead concentration of 12,000 beads mL⁻¹ produces 2,400 valid bead events and 6,000 target-cell events. The target concentration in the analysed mixture is (6,000/2,400) × 12,000 = 30,000 cells mL⁻¹. If one part original water had been mixed with one part diluent before beads were brought to the stated final concentration, and no other sample changes occurred, the original water concentration would be 60,000 cells mL⁻¹. The dilution factor belongs to the pre-analysis sample preparation, not to the optical intensity.
The calculation only works if bead and cell populations are reliably recognized, mixing is sufficient and event losses or coincidence do not distort their ratio. If beads sit in a different acquisition gate from faint cells, changing a trigger threshold may disproportionately affect counts. If the instrument samples a measured volume instead, it may be inappropriate to apply an additional bead-derived correction. Readers should reconstruct the chosen counting model and not mix two partially incompatible corrections simply because both appear in software.
The sample can change between the sea and the laser
Collection from a bottle, intake or incubation vessel introduces a sampling chain. The volume placed in a tube might represent a single depth, a mixed carboy or a composite of several casts. Those are not equivalent sampling units. Delays, temperature changes, exposure to light, fixation, freezing and thawing can alter cell fluorescence or integrity. Some protocols count fresh material and others preserve samples under specific conditions. The method must say which was used, because a frozen sample is not necessarily optically identical to living seawater even when its cell abundance estimate remains useful.
A screen or prefilter may exclude larger plankton so the small-cell cytometer does not clog or become overwhelmed. That step should be reported, not hidden, because it defines the counted size fraction. A sample passing through a 200 µm mesh remains rich in smaller phytoplankton, but it is no longer an unmodified whole-water sample. Similarly, a flow cytometer that resolves picophytoplankton well may miss fragile chains, large aggregates or cells that break apart during collection. The phrase “total phytoplankton abundance” should not be used if the assay only covers a documented optical or size-defined subset.
Field replicates help separate sample heterogeneity from instrumental repeatability. Re-running the same tube tests a different source of variation from collecting a second bottle. A strong dataset can include both: duplicate acquisition to test counting and gating stability, and independent field samples to test representativeness. When disagreement occurs, a reviewer needs to know which level failed. Calling them all “duplicates” makes the error budget impossible to interpret.
Why an event is not always one cell
Flow cytometers resolve particles as they cross the illuminated region. At high event rates, more than one cell can be present in a measurement window; this is coincidence. The resulting optical pulse may be distorted, merged or lost, depending on the instrument and trigger scheme. Dilution can reduce coincidence, but excessive dilution also leaves fewer events for rare populations. The best operating rate is therefore a method-specific balance. The research literature on high-flow oceanic cytometry documents undercounting as event rates become very high, which is a warning against treating events per second as unlimited counting capacity.
A sample containing aggregates also challenges the one-event-one-cell assumption. A cluster of two or more attached cells might traverse the laser as one event, while broken fragments can form extra signals. A gate excluding doublet-like pulses may remove aggregates intentionally but also alters the population that remains. The reader should ask whether the reported quantity counts individual cells, optical particles or events meeting an agreed definition. Good descriptions use “events” when that is the only directly evidenced unit and reserve “cells per millilitre” for validated enumeration procedures.
Instrument background matters as well. Electronics, sheath fluid particles and debris can create small signals. A blank acquisition with appropriate clean fluid tests the background and may reveal contaminated tubing or a dirty flow cell. It cannot prove all remaining events are biological; discrimination still depends on fluorescence and gate criteria. A blank performed after a highly concentrated sample can be especially informative for carryover.
Worked example: a rare population with counting uncertainty
Suppose 64 accepted target events are counted from a verified 0.80 mL sampled volume, with no dilution and negligible background. The point estimate is 64/0.80 = 80 cells mL⁻¹. For an ideal independent Poisson count, a quick standard uncertainty approximation for the event count is sqrt(64) = 8, or 12.5 percent relative. Dividing by volume gives about 10 cells mL⁻¹ from counting statistics alone. This is not a complete laboratory uncertainty: flow-volume calibration, gate boundaries, preservation and representative sampling may contribute more.
Compare a second tube with only four accepted target events in the same volume. The point estimate is 5 cells mL⁻¹, but sqrt(4)/4 gives about fifty percent relative counting uncertainty even before systematic errors. The result may be reported with an appropriate counting-based interval or detection qualifier instead of an exact-looking 5.00 cells mL⁻¹. At very small counts, symmetric normal approximations are poor, so a suitable Poisson confidence interval is preferable if an inferential statement matters. The arithmetic illustrates why rare populations need longer acquisition or larger analysed volumes, not merely brighter scatter plots.
If the original sample was diluted threefold before analysis, the corrected original concentration from the first example is 240 cells mL⁻¹, assuming the 0.80 mL measurement refers to diluted mixture. The counting-statistics relative uncertainty remains about 12.5 percent under the simple model; multiplying the estimate by three also multiplies its absolute uncertainty. Uncertainty in the dilution factor itself would need an additional contribution. A reader should not apply a factor of three twice, once in software and once in a spreadsheet.
Worked example: interpreting a changed gate
Imagine two analysts examine the same event file. Analyst A draws a red-fluorescence threshold that accepts 1,800 dim-cell events; analyst B places it slightly higher and accepts 1,400. The difference is 400 events, or roughly 22 percent of analyst A’s total. If the acquisition volume is 0.20 mL, their estimates are 9,000 and 7,000 events per millilitre respectively. This is an analytical-definition difference, not a difference in seawater. It should be resolved by reference controls, documented gate rules and a reproducibility exercise rather than choosing the count that aligns with an expected seasonal pattern.
Gate uncertainty is often largest near overlapping or dim clusters. Review the background distribution, bead positions, compensation between channels and representative event density. Re-gating a past dataset can improve comparability if applied consistently with justification, but it changes the historical method. Keep version numbers and, where practical, both raw event files and processed tables. A scientist writing about trends should say when optical groups have been redefined across campaigns.
Fluorescence intensity is not a fixed biomass conversion
A cell’s red fluorescence depends on pigments, fluorescence yield, excitation and detector response. Light history, photophysiology, growth state and preservation can alter the measured intensity. A brighter red signal does not necessarily mean a cell contains exactly twice as much chlorophyll, and higher group counts do not necessarily mean proportionally higher biomass if average cell size changes. A volumetric cell count is one type of information; carbon biomass, pigment mass and productivity are other quantities requiring independent models or measurements.
Suppose a profile shows 10,000 picoeukaryote-like cells mL⁻¹ at the surface and 8,000 at depth, while their median red fluorescence is higher at depth. It is correct to say that the counted operational group was less abundant at depth in those samples but exhibited a brighter median signal under the specified settings. It is not justified to multiply fluorescence by abundance and call the product primary production without a validated relationship. Nor should the brighter signal alone be described as a deeper bloom. Biological interpretation needs context such as light, nutrients, cell size, pigment physiology and independent biomass measurements.
The associated underwater light-attenuation guide explains how light availability changes with depth. That physical context can help explain changes in fluorescence patterns, but it does not replace calibrated cytometry or pigment methods. Keep causal links conditional unless they are tested.
Comparing concentrations between depths, cruises and instruments
A depth profile of cells mL⁻¹ is not automatically a depth-integrated inventory. To estimate cells per square metre over a water-column interval, integrate concentration over depth after converting millilitres to cubic metres. One cell per millilitre equals one million cells per cubic metre. A uniform abundance of 2,000 cells mL⁻¹ over a ten-metre layer corresponds to 2,000 × 10^6 × 10 = 2 × 10^10 cells m⁻² under the assumed uniform profile. This calculation illustrates dimensional conversion rather than claiming that real profiles are uniform.
For irregular depths, a trapezoidal or other justified numerical integration may be used; the bins, interpolation assumptions and sample coverage should be stated. A single shallow bottle cannot support a whole-water-column inventory just because the scientist knows the local depth. Likewise, comparing abundance per millilitre between cruises requires checking that dilution, volume, thresholds and gating protocols are compatible. Apparent changes may reflect instrument settings or sampling times rather than biology. Strong comparisons present biological interpretation only after the measurement conventions align.
Daily timing may also matter. A population can change size, scatter or fluorescence during a diel cycle. Sampling at dawn on one cruise and late afternoon on another may produce different optical signatures even if cell abundance is similar. That does not make the measurements useless, but it cautions against interpreting every shift in cluster location as an ecological replacement. Metadata for local sampling time and light conditions are therefore part of a reusable cytometry record.
Bead standards are references, not miniature cells
Fluorescent beads can test optical alignment, detector response and count concentration, depending on the protocol. A bead diameter alone does not determine the scatter of an unknown cell because refractive index, shape and internal complexity affect scattering. It can be helpful to compare an event cluster against a reference bead cloud within the same run, but reporting cells as though they have the physical diameter of the nearest bead is often too strong. Calibrated optical size inference requires a defined model and validation with appropriate particles or independent microscopy.
Bead stability must be assessed. Settling, clumping, dilution error and pipetting imprecision alter the actual bead population in a tube. A nominal concentration printed on a vial may not be the concentration after a multistep dilution. If two preparations disagree, the beads may reveal a preparation problem rather than an ecological anomaly. Store the bead lot, nominal and working concentrations, preparation dates and acquisition settings. Treat beads as a measurement standard with its own uncertainty, not as an unquestionable denominator.
How to turn a cluster plot into an honest paragraph
A careful result might say: “Flow cytometry identified distinct phototrophic event populations using scatter, chlorophyll-related red fluorescence and a phycoerythrin-sensitive orange channel. Abundances were calculated from validated sample volume after dilution adjustment. The Synechococcus-like gate contained more events in surface samples, but taxonomic identification and per-cell carbon conversion were not inferred from the optical gate alone.” The paragraph is readable and retains the conditions that make the count meaningful.
An overstatement would be: “The instrument discovered a new species and proved the surface water contained twice as much living biomass.” Flow cytometry can provide powerful clues about population structure, but species discovery demands independent identification, and biomass requires cell-property conversion with uncertainty. The responsibility to communicate faithfully is greater, not less, when a cluster plot looks visually persuasive.
A four-stage quality-control sequence
Stage one examines acquisition quality: stable flow, correct detector settings, sensible event rate, clean sheath fluid and a documented sample volume. Record any interruptions or clogs. Stage two examines optical controls: blank event patterns, bead clouds, detector drift and the separation of expected fluorescence populations. Stage three checks the gating version and treatment of doublets, noise and overlapping populations. Stage four checks the calculation: dilution, bead concentration or measured volume, cell units, event totals, uncertainty and the reported denominator. A sample can pass one stage and fail another, so a single vague “QC passed” label hides useful information.
The sequence also guides repair. If a bead cluster shifts in fluorescence after an electronic gain adjustment, compare settings and revalidate gates rather than calling it a biological response. If a volume calibration fails, recollect or correct with a documented independent method; do not simply use total run time as volume without calibrated flow. If a sample clogs during acquisition, the event count and actual processed volume may no longer align. Report the interval as unreliable if there is no defensible recovery procedure. Data quality is a property of the full measurement chain, not just the software plot.
A laboratory can build a simple audit sheet with fields for sample identity, collection depth and time, preservative and storage, dilution factor, bead lot, gate template version, event rate, number counted, volume or bead reference, flags, computed concentration and analyst. That modest structure supports reproducibility across busy voyages when dozens of tubes are processed. It also reveals whether two apparent ecological changes were measured with comparable procedures.
Fully checked calculation: bead reference with a background subtraction
A teaching mixture contains 18,000 standard beads mL⁻¹ and yields 3,600 accepted bead events. In the target gate, 9,200 events are recorded; an equal-volume blank indicates an expected 200 background events in that gate under the declared correction procedure. The blank-adjusted target is 9,000 events. Its estimated concentration in the analysed mixture is (9,000/3,600) × 18,000 = 45,000 target events mL⁻¹. If the original seawater was diluted by a factor of four before the final bead concentration was established, the corresponding original-water estimate is 180,000 cells mL⁻¹, conditional on a valid one-event-one-cell classification.
This example is intentionally constructed. Real blanks may not be subtracted as a single exact expected count, and their uncertainty contributes to the result. One should not subtract a blank from a cell gate if the background process or protocol does not support that operation. Likewise, if the standard beads were added before dilution rather than at the end, the reference concentration in the measured mixture changes and must be recalculated. The point is to follow the material and volume through each preparation stage. A defensible answer can explain why every multiplier appears.
For a rough ideal count-only standard uncertainty, the cell-to-bead ratio depends on both counts. If cell and bead event counts behave as independent Poisson variables and the background is ignored for this simplified propagation, the relative standard uncertainty of the ratio is approximately sqrt(1/N_cell + 1/N_bead). With N_cell = 9,000 and N_bead = 3,600, that is about sqrt(0.000111 + 0.000278) ≈ 0.0197, or two percent. Bead stock uncertainty, blank subtraction, gating and dilution may be much larger; the two-percent figure is not a total method uncertainty.
What the denominator means in ecological comparisons
Suppose the proportion of orange-fluorescent events rises from twenty to thirty percent between two cruises. The reader must ask: twenty percent of which events? If the first table uses all acquired pulses and the second uses only accepted chlorophyll-positive phototrophs, the percentages cannot be compared as ecological proportions. Even with consistent gating, the total community may have changed, so a higher fraction does not guarantee a higher absolute abundance. A group can gain share while losing cells if the rest of the community declines faster.
For example, cruise A has 10,000 accepted phototrophs mL⁻¹ and the target comprises 20 percent, meaning 2,000 target cells mL⁻¹. Cruise B has 5,000 accepted phototrophs mL⁻¹ and a target share of 30 percent, meaning 1,500 target cells mL⁻¹. The share increased, but the abundance declined by 500 cells mL⁻¹ or 25 percent relative to the earlier target count. This distinction between shares and counts is central to translating ecological graphs for the public. If a table omits its denominator, the scientifically responsible action is to recover metadata before writing a trend headline.
Exercises that test transfer, with answers
Exercise one: A validated volume measurement shows 0.50 mL acquired and the gate contains 2,750 accepted events. What is the estimated concentration before any corrections? Answer: 2,750/0.50 = 5,500 events mL⁻¹. If the sample was diluted twofold immediately before acquisition, the original concentration would be 11,000 events mL⁻¹, assuming the gate truly represents single target cells and the stated volume is the diluted mixture. A solver who divides by two when undoing dilution has reversed the direction of the correction.
Exercise two: A gate contains 225 events from one sample and 900 from another, both with equal analysed volume. Which counting estimate has the smaller Poisson relative uncertainty? Answer: the 900-event count; the rough relative standard uncertainty is 1/sqrt(900) ≈ 3.3 percent, compared with 1/sqrt(225) ≈ 6.7 percent. This comparison does not say the larger population is ecologically more reliable in every sense. If the 900-event gate is poorly defined or the flow is unstable, systematic error may dominate its seemingly better counting precision.
Exercise three: A graph shows two clouds separated by red fluorescence but their orange channels overlap. Can the analyst automatically label one group Prochlorococcus and the other Synechococcus? Answer: no. Identification must follow a validated configuration, including appropriate excitation, detection, gates and supporting controls. Dim Prochlorococcus-like populations and phycoerythrin-bearing Synechococcus-like populations can behave differently, but visual separation alone does not prove taxonomy. Write “operational fluorescence populations” if the evidence is limited to optical clustering.
Exercise four: A sample has 1,200 target-cell events and 800 bead events; working beads are 10,000 mL⁻¹. The mixture estimate is (1,200/800) × 10,000 = 15,000 cells mL⁻¹. A second tube gives 600 target events and 400 beads under the same working concentration. Both ratios equal 1.5, hence both point estimates are 15,000 cells mL⁻¹. The second tube has fewer counts and therefore larger ideal count-related uncertainty. Equal point estimates do not imply equal measurement quality.
Exercise five: A sample with high cellular abundance begins showing a declining cell-to-bead event ratio as the acquisition flow is increased. A possible cause is coincidence or unequal detection efficiency at high event rate, not necessarily a rapid ecological change within the tube. Repeat at an appropriately diluted concentration, inspect rate stability and follow the instrument’s approved validation protocol. If the corrected ratio becomes stable at lower rates, document the operating range used. The key lesson is that a fast count is not always an accurate count.
Why confidence depends on controls outside the plot
An event file can contain hundreds of thousands of points and still fail to measure cell concentration if sample volume is unknown. Conversely, a sparse file with a well-documented volume may support a qualified low-count estimate. More dots do not automatically mean greater validity. Bead alignment, optical stability, biological controls, independent microscopy, defined gates and correct sampling metadata all help establish what the dots represent. The best figure shows representative clusters and reports a table of operational counts rather than relying on colour-coded visual authority.
Environmental samples are not uniform calibration fluids. Organic debris, viruses, heterotrophic bacteria, mineral grains and detrital fragments may appear in the same sample, often at concentrations unrelated to the phototrophic cells of interest. A strong gating scheme identifies how background or overlapping objects are handled. Some tiny cells may sit below the trigger threshold, and loss is not necessarily uniform across depths. This is why a study can report a robust trend within its defined detection window without claiming to measure all marine life.
Frequently asked questions and decisions
Does a red-fluorescent event always equal one phytoplankton cell? Not necessarily; aggregates, fragments, coincidence and background can violate the assumption. Do beads directly measure cell size? Not without a validated optical model; beads are reference particles with their own scattering properties. Does a bigger fluorescence cluster prove more carbon biomass? No; fluorescence and abundance are separate from carbon per cell. Can fixed and unfixed samples be compared? Only with a method demonstrating that preservation and storage effects are controlled for the relevant populations and settings.
Is a species name justified by a fluorescence gate? A gate commonly identifies an operational group, and a species claim needs stronger identification. Why keep raw event files? They permit gate review and preserve information lost in final counts. Why keep the sampling time? Population physiology and fluorescence can vary with environmental and diel conditions. What should be done if bead counts drift? Examine bead preparation, optics, acquisition rate and calibration before applying an adjustment. What if two methods disagree? Explain their detection ranges and denominators rather than forcing identical answers.
A careful summary might say: “The method enumerated fluorescence-defined marine phototroph groups using documented scatter and pigment channels. Counts were normalized by validated sampled volume or bead references and corrected for declared dilution. The reported populations are operational classifications and should not be interpreted as a complete species inventory or a direct biomass measurement.” That sentence preserves the practical value of a sensitive tool while honestly describing its boundaries.
How to diagnose a mismatch between two laboratories
Imagine laboratory A reports 15,000 Synechococcus-like cells mL⁻¹ and laboratory B reports 9,000 from aliquots of the same seawater. The difference is large enough to investigate, but it does not identify which laboratory is correct. Start with sample handling: were the aliquots mixed before splitting, and did storage time and preservative match? Compare acquisition settings, laser and filter configuration, the counting volume or bead reference, trigger thresholds and gate boundaries. Then review how overlapping events and low-fluorescence populations were treated. If one laboratory used a bead-normalized calculation and the other a measured-volume calculation, inspect both reference uncertainties before forcing an average.
A controlled exchange can help. Each laboratory can analyse common reference material and, where feasible, a shared sample under pre-agreed settings. They can compare raw event distributions, not only their final concentrations. If both instruments count a stable reference consistently but disagree on dim oceanic cells, the problem may lie in biological gate definitions or sensitivity rather than volumetric calibration. If bead positions drift between runs, instrument optics or bead preparation warrant attention. A useful result is sometimes a documented method-specific difference rather than one universal number. That is more scientifically honest than selecting whichever estimate supports the preferred hypothesis.
A repeatable reporting template
For each reported population, include the optical definition, gates, reference volume, any dilution correction, count totals and quality flag. For example, state that a Synechococcus-like group was identified by the configured orange and red fluorescence and scatter characteristics, rather than saying a DNA-confirmed species was measured. Provide abundance in cells per millilitre only when event-to-cell assignment is supported. For rare populations, show the absolute number counted as well as the normalized estimate, because an apparently precise concentration may rest on very few events.
If a future analyst wants to convert the count into carbon stock, the document should identify that as a separate model requiring defensible per-cell carbon factors and uncertainties. Those factors may vary among organisms and environmental conditions. Keeping cell abundance, fluorescence and estimated biomass in separate columns prevents a reader from treating one number as three independent observations. Good scientific communication is not a contest to fit the most labels in a graph legend. It is a way of preserving the route from an observed optical pulse to a qualified ecological statement.
Original methods and further reading
Bigelow Laboratory phytoplankton enumeration and sorting; Zubkov and colleagues, syringe-pumped high-speed flow cytometry of oceanic phytoplankton; BCO-DMO ocean flow-cytometry dataset description; and peer-reviewed size-fractionated picophytoplankton analysis. All numeric exercises above are constructed examples, not observations from these sources.