Quantifying the accuracy of estimating harmful algal bloom cell concentrations using OLCI imagery in the tributaries of lower Chesapeake Bay

Harmful algal blooms of Alexandrium monilatum and Margalefidinium polykrikoides are a common occurrence in the tributaries of lower Chesapeake Bay, United States of America. This study evaluated whether daily processed imagery from the Ocean and Land Colour Instrument (OLCI) can accurately measure chlorophyll a (chl a) concentrations in these turbid, bloom-prone tributaries. Since HAB cell concentrations provide a more direct measure of risk than chl a, a second objective was to quantify the accuracy of bloom cell concentrations derived from satellite chl a. Field collected chl a concentrations and processed OLCI data were used to evaluate the accuracy of satellite-derived chl a concentrations in the York and James Rivers. Field, laboratory, and literature data were used to quantify the precision of satellite-based cell estimates. Satellite-derived chl a estimates correlated well with laboratory extracted chl a from the field and were stable over time for the entire OLCI satellite image dataset. The results of laboratory experiments and field studies showed that variations in the pg chl a cell -1 used to convert chl a to cell estimates, often varied five- to hundred-fold, depending on prevailing environmental growth conditions and the chl a contributed by co-occurring phytoplankton. The satellite-derived chl a concentrations proved accurate over a wide range of chl a concentrations, confirming their usefulness in tracking high biomass blooms and guiding sampling efforts required to evaluate potential adverse impacts, and as a basis for potentially estimating relative HAB cell concentrations. However, field, literature, and laboratory data showed that chl a cell -1 estimates at a given chl a concentration can vary by up to 100-fold. Consequently, satellite-based cell concentrations will be highly inaccurate. More accurate HAB cell estimates might be possible if chl a cell -1 is measured periodically during a bloom and used to estimate cell concentration from satellite chl a in real-time.

The York River and James River estuaries in Virginia, USA are representative of the highly productive tributaries of the lower Chesapeake Bay. These estuaries are subject to late summer to early fall blooms of the toxic dinoflagellates Margalefidinium (formerly Cochlodinium) polykrikoides (Margalef) F. Gómez, Richlen and D.M.Anderson and Alexandrium monilatum (J.F.Howell) (Mulholland et al., 2018). These blooms pose a direct threat to the aquaculture and fisheries industry due to their documented ability to cause high fish and shellfish mortalities when present at high concentrations (Connell and Cross, 1950; Howell, 1953; Gates and Wilson, 1960; Williams and Ingle, 1972; Ho and Zubkoff, 1979; Perry et al., 1979; Norris, 1983; Kim et al., 1999; Kim et al., 2000; Mulholland et al., 2009; Tang and Gobler, 2009; May et al., 2010; Harris et al., 2023; Gaillard et al., 2024b). Goniodomin A (GDA), an antifungal polyether macrolide, is the primary toxin produced by A. monilatum (Hsia et al., 2006; Gaillard et al., 2024a). GDA has been shown to interfere with actin function (Furukawa et al., 1993; Yasuda et al., 1998), limit angiogenesis (Abe et al., 2002) and induce a loss of viability in rat hepatocytes and human neuroblastoma cells (Espiña et al., 2016). This species also produces a range of unidentified, toxic, lytic compounds (Gates and Wilson, 1960; Ray and Aldrich, 1966; Sievers, 1969; Wardle et al., 1975; Clemons et al., 1980; Bass et al., 1983; May et al., 2010; Long et al., 2021; Gaillard et al., 2024b). These toxins are primarily released when cells are broken during grazing or at the end of blooms when cells senesce. Margalefidinum polykrikoides produces reactive oxygen species that destroy membranes, particularly those of gills (Kim et al., 1999; Kim et al., 2009; Shahraki et al., 2013; Shahraki et al., 2014; Pérez-Morales et al., 2017; López-Cortés et al., 2019). It can also secrete large amounts of extracellular polysaccharides that clog gill structures leading to asphyxiation (Kim et al., 2002). Alexandrium monilatum and M. polykrikoides cell densities reported during fish mortality events are most commonly in the 1,000,000 cells L -1 range or higher though concentrations as low as ∼500,000 cells L -1 sometimes caused mortalities depending on species and conditions. Lethal cell concentrations for shellfish and other invertebrates’ range between 1,000,000 and 5,000,000 cells L -1 (Tang and Gobler, 2009; May et al., 2010; Reece et al., 2020). These blooms also exhibit highly patchy distributions and rapidly change location in response to fluctuating wind, river flow, tidal, and current conditions. Despite extensive sampling efforts conducted monthly by the Chesapeake Bay Program (CBP), Maryland and Virginia state agency monitoring programs, and other organizations in response to Harmful Algal Bloom (HAB) events, accurately tracking blooms both spatially and temporally across large areas is a challenge.

Satellite surveillance provides a potential means to overcome this limitation. Synoptically sampling chlorophyll a (chl a) biomass at high spatial resolutions, over large regions can provide early warning of high biomass algal blooms and their potential risk (Wolny et al., 2020). This approach has proven effective for Chesapeake Bay where The National Oceanic and Atmospheric Administration (NOAA) has been providing chl a and fluorescence products operationally for monitoring blooms in U.S. coastal regions for over a decade. An operational chl a product derived from the Ocean and Land Colour Instrument (OLCI) has been used as an estimate of phytoplankton biomass. This product was validated against field chl a in the mainstem of Chesapeake Bay by Wynne et al. (2022), but did not include tributary samples where colored dissolved organic matter (CDOM) and turbidity are high. OLCI is currently the most suitable sensor for monitoring large scale blooms in estuaries due to its unique temporal (daily), spatial (300 m at nadir), and spectral resolution. OLCI contains additional bands in the red and near-infrared portion of the spectrum which are less sensitive than traditional blue-green ratio algorithms, to CDOM and scattering due to sediment and other particles (Gilerson et al., 2010).

The objectives of this manuscript were two-fold. The first was to validate the accuracy of the OLCI chl a estimates in the tributaries of lower Chesapeake Bay, which are characterized by higher turbidity and CDOM than in the mainstem of Chesapeake Bay. Validation of the chl a algorithm in these waters has not been previously undertaken. Given that HABs are a common occurrence in the lower Chesapeake Bay tributaries (Wolny et al., 2020), having accurate, synoptic, high-resolution chl a data is critical for allowing environmental managers to identify and track blooms of concern, and to prioritize monitoring efforts aimed at assessing the potential adverse bloom impacts.

The second broader objective was to quantify how well, or poorly, field cell concentrations of the dinoflagellate HAB species A. monilatum and M. polykrikoides can be estimated from OLCI-derived chl a measurements. Having accurate cell count estimates is important because they better predict risk compared to chl a measurements. Though the general consensus among satellite oceanographers is that any chl a to cell count conversion will be highly inaccurate, the extent of that inaccuracy has seldom been quantified. To address this issue, OLCI-derived chl a, field A. monilatum and M. polykrikoides cell count and extracted chl a were examined. Laboratory growth studies were conducted to investigate the variability in chla cell -1 within cultures. These combined data were then used to quantify the variation in chl a to cell count estimates and to illustrate the physiological and other mechanisms responsible for the observed variation in satellite chl a to cell count estimates.

Several algorithms related to chl a concentration and relative fluorescence have been derived from OLCI imagery and are used in a routine monitoring system for HABs in Chesapeake Bay (Wolny et al., 2020). A coastal chl a algorithm, which utilizes the red-edge portion of the electromagnetic spectrum was found to be reasonably accurate and stable for measuring chl a in the mainstem of Chesapeake Bay (Wynne et al., 2022). A refinement was subsequently added for chl a concentrations below 10 μg L -1 (Wynne et al., 2026). Briefly, values exceeding this threshold were calculated using the red-edge algorithm. For samples below 10 μg L -1, the algorithm employed to calculate chl a switched to the standard OC4 global algorithm developed by O’Reilly et al. (1998), O'Reilly and Werdell (2019). This algorithm has been extensively validated for open ocean, Case 1 waters where phytoplankton and their degradation products dominate the optics. However, it is important to note that the combined red-edge and OC4 approach for measuring chl a using satellite-based sensors has not been validated for the shallow, highly turbid tributaries of lower Chesapeake Bay.

The accuracy of the satellite-derived chl a values were validated by comparing chl a values from the daily 300 m resolution OLCI imagery with temporally and spatially matched field extracted chl a values collected from the York and James River estuaries. The field chl a values were collected over the past decade by the CBP and are available for download (DataHub, 2026). Given that HABs are a common occurrence in the lower Chesapeake Bay tributaries (Wolny et al., 2020), accurate, synoptic, high-resolution chl a are used by environmental managers to identify and track blooms of concern and prioritize monitoring efforts aimed at assessing the potential adverse bloom impacts. However, an assessment as to how this correlates with M. polykrikoides or A. monilatum cell counts has not been addressed.

The accuracy of HAB cell concentration from satellite chl a is dependent on the potential variations in the chl a cell -1 conversion factor used in the estimation. The greater the variation in chl a cell -1 in a species, the less accurate the cell estimate. Consequently, to address the second study objective, we quantified variation in chl a cell -1 in field samples, as well as in cultures where changes in chl a cell -1 under various growth conditions could be accurately measured. Specifically, the field chl a cell -1 estimates were calculated using matched cell count and extracted chl a data obtained from various York River monitoring program collected over the past decade. Equivalent data from published studies were also analyzed. The laboratory studies measured changes in chl a cell -1 under a range of nutrient, salinity, light and temperature conditions that prevail during blooms of A. monilatum and M. polykrikoides. Since A. monilatum and M. polykrikoide s blooms occur in numerous estuarine and nearshore systems worldwide, the findings in this study are likely broadly applicable to other systems.

Study site: lower Chesapeake Bay

The field samples used in this study to validate the OLCI satellite chl a data were collected from representative tributaries of lower Chesapeake Bay (the York and James River estuaries, Virginia, USA) from 2016 to 2025 (Figure 1). Blooms of one or both A. monilatum and M. polykrikoides have been reported to occur in these estuaries between July and September over the past 2 decades (Marshall and Egerton, 2009; Mulholland et al., 2009; Mulholland et al., 2018; Gaillard et al., 2024b).

Validation of satellite chl a estimates using field collected samples

The validation of satellite chl a estimates using field collected samples is complex due to the patchy nature of estuarine algal blooms and their rapid movement caused by wind and tides. To demonstrate, Figure 2 shows satellite-derived images of an M. polykrikoides bloom in the York River, Virginia, from 13 August 2022 to 28 August 2022. These data were collected from the MultiSpectral Instrument (MSI) onboard the Sentinel-2 satellite. This sensor has 10–20 m spatial resolution and temporal refresh every 5 days over this region, given the two satellites in orbit. The images have been processed on the Earth Observations Browser (EOB, 2025) as a custom false color image. It is apparent that these blooms are patchy and rapidly shift location within the river due to tidal and wind forcing. Unfortunately, these highly resolved images are collected at a low frequency. Therefore, daily satellite imagery for these analyses was obtained from the Ocean Land Colour Instrument (OLCI) carried on the Sentinel-3 satellite operated by the European Organisation for the Exploitation of Meteorological Satellites. Though this product is less resolved spatially (300 m vs. 10s of meters), it better captured the bloom movement and was used for further analysis. It is the primary sensor used for monitoring blooms in Chesapeake Bay and provided by NOAA operationally, due to its temporal and spectral resolution. A more accurate chl a for coastal systems was processed by NOAA’s National Centers for Coastal Ocean Science (NCCOS) using a switch from the Red-Edge 10 (RE10 (Gilerson et al., 2010)) to the global OC4 algorithm at chl a values below 10 μg L -1 (Gilerson et al., 2024) through the processing methods outlined in Wynne et al. (2026) and will be referred to as the chl_switch.

The patchy nature of dinoflagellate blooms, along with changes in their spatial distribution throughout the day due to winds, tides, vertical mixing and migration, make comparisons of field chl a to satellite pixel chl a challenging. There is also a temporal mismatch between the timing of sample collection versus the flyover by the satellite. Additionally, field samples represent several liters of water at one location, whereas a satellite pixel encompasses a larger footprint (in the case of OLCI, approximately 90,000 m 2). Considering these limitations, the accuracy of the satellite-derived chl a was evaluated in two ways.

The first method used was a “point-to-point” matchup between field and satellite derived chl a concentrations. The field chl a were obtained from five sampling stations in the York and James River estuaries which are routinely sampled monthly by the CBP and available online (Blue squares in Figure 1) (DataHub, 2026). These chl a field samples were collected between 2016 and 2025 and the extracted chl a concentrations were measured using the Welschmeyer (1994) method. Subsamples measuring 25 mL for chl a analysis were immediately filtered through a Whatman GF/F filter (25 mm diameter; Cytiva, Marlborough, Massachusetts, USA) at <5 cm Hg, blotted dry, wrapped in numbered aluminum foil packets, and kept on ice in a plastic bag for a short period of time (1–2 h) until the samples were returned to the laboratory and stored at −80 °C. Just prior to analysis, the GF/F filter was sonicated for 1–2 min in 7.5 mL of 90% acetone: 10% water solution. The homogenate was collected in a clean, brown 15 mL conical tube and placed in a −20 °C freezer for 24 h. Prior to analysis the samples were brought to room temperature but kept in the dark. They were then centrifuged for 10 min at 3,000 rpm and the resulting supernatant was transferred to a disposable culture tube (Fisherbrand, Rockville, MD, USA) and the chl a concentration measured on a Turner 10 AU fluorometer (Turner Designs, San Jose, CA, USA) using disposable culture tubes (Fisherbrand, Rockville, MD, USA). The same methodology was used for determining chl a concentration in the field and laboratory experiments described below.

Next, the extracted field chl a measurements were directly compared with the median OLCI satellite chl a obtained from a 3 × 3 pixel (900 m × 900 m each) grid surrounding the field sampling site. The OLCI imagery was processed by NOAA’s NCCOS SeaDAS (https://seadas.gsfc.nasa.gov/about/) using the Automated Processing System (SAPS) RE10 program as described by Wynne et al. (2026). The satellite chl a measurements were adjusted for offshore waters using a switch at 10 μg L -1 (Equations 1, 2) from the RE10 algorithm which was validated in Chesapeake Bay for eutrophic estuaries (Wynne et al., 2022) to the NASA standard global OC4 algorithm:

Field and satellite chl a estimates collected within 24 h of each other were included in the analysis. Specific sample comparisons were omitted if one or more of the pixels in a 5 × 5 pixel window were obscured by cloud cover or land. The field data used in the point-to-point comparisons were obtained from the CBP for the routinely monitored stations in the York River (LE4.2 and LE4.3) and James River (LE5.3, LE5.4, and LE5.5) (Figure 1). Chlorophyll a was determined using the Welschmeyer (1994) method as described above. Once valid field and satellite chl a matchups were identified, a regression analysis was performed and the bias and Median Absolute Deviation (MedAD) were calculated, as they show the typical error in the satellite chl a measurements (Seegers et al., 2018).

The second approach to determine the stability of the chl a algorithm in the York River involved creating a time-series plot of the median of the 3 × 3 pixels from the same CBP stations (2016–2025) against the monthly field chl a measurements from those stations.

Determining chl a cell -1 variations using corresponding field chl a and HAB cell concentrations

Variations in the chl a cell -1 of A. monilatum and M. polykrikoides were calculated using corresponding cell count and chl a concentrations from water samples collected in the York and James River estuaries by the CBP, Virginia Institute of Marine Sciences (VIMS), Virginia Department of Health (VDH) and NCCOS. These samples were collected between 2016 and 2022 using a large bucket or Niskin bottle (General Oceanics, Miami, Florida, USA). The specific methods used are described below. Mixing of the collected sample followed by rapid removal (<1 min) of aliquots was particularly important to prevent the aggregation of rapidly swimming cells into dense clumps. Although this behavior is frequently observed, especially during M. polykrikoides blooms, its causes remain unknown.

A. monilatum and M. polykrikoides field cell enumeration

To estimate cell concentrations, a 125 mL subsample was removed from the collection vessel and placed in light-tight, dark brown plastic bottle filled to the top. The sample was then fixed with 4–5 drops of neutral Lugol’s as a preservative, capped, gently mixed, and stored at 4 °C. Prior to counting, samples were gently inverted three to four times to resuspend the cells. Cell concentrations were then determined using either of the following two protocols.

In the first protocol, if samples contained <500 cells mL -1, a 1 mL sample was added to a Sedgewick Rafter chamber (Cole-Parmer, Vernon Hill, Illinois, USA) and grids were counted under ×10 magnification using a Nikon Eclipse TS100 (Nikon Instruments, Melville, NY USA) or a total of 10 columns were counted for a total of 200 grids. Cells mL -1 = (Average Cells Grid -1 * Total Grids in Chamber)/Chamber volume (mL). Cell densities exceeding 500 cell mL -1 could not be counted accurately, particularly when cells were forming chains. These samples were diluted 1:10 and recounted.

In the second protocol, samples were again initially counted using a 1 mL sample Sedgewick Rafter, ×10 magnification and an Olympus IX51 inverted microscope (Evident Scientific, Waltham, MA USA) as described above. For the higher density bloom samples, three 10 µL aliquots were placed on a microscope slide and all cells in each drop were counted. Cells mL -1 were then calculated by dividing the total number of cells counted by three, then multiplied by 100. If the cell density in a diluted sample exceeded 5,000 cells mL -1, an additional 1/10 dilution was made, and the cells were again counted.

Chl a concentrations from laboratory samples were determined as described previously. Some variation in chl a cell -1 is expected due to differences in prevailing growth conditions and chl a contributed by co-occurring phytoplankton. The chl a contributed by other phytoplankton causes an over estimation of the chl a cell -1 values obtained by dividing the bulk chl a concentration in a given sample by the corresponding cell count.

Criteria for bloom dominance by A. monilatum and M. polykrikoides

The typical bloom sequence in the lower Chesapeake Bay tributaries usually begins with an M. polykrikoides bloom, followed by a mixed bloom, that sometimes transitions to an A. monilatum -dominated bloom (Marshall and Egerton, 2009; Mulholland et al., 2018; Fortin et al., 2022). Given potentially mixed blooms, the data were first apportioned into samples where the A. monilatum concentrations exceeded those of M. polykrikoides and vice versa. The former data were used in calculating chl a cell -1 values for A. monilatum dominated blooms and the latter for M. polykrikoides blooms. Samples having cell counts less than 25,000 cells L -1 were excluded because this density limit represents prebloom conditions where chl a concentrations are too low for accurate quantitation using OLCI. Further, it is important to note that neither A. monilatum or M. polykrikoides blooms are mono-specific, particularly those of M. polykrikoides. Numerous other co-occurring dinoflagellate species are typically present, often in high abundance, along with a subdominant diatom assemblage.

Data analysis of field data

The chl a and cell count data were used to calculate pg chla cell -1 for all the A. monilatum- and M. polykrikoides -dominated samples. Power regressions between cell concentrations versus chl a were conducted for both species by year when samples were collected. The resulting correlation coefficient (R 2) values were recorded. This allowed determination of how well increases in chl a biomass corresponded with measured cell concentrations. Equivalent analysis of chl a versus chl a cell -1 for each species were also produced to illustrate the total variation in chl a cell -1 at a given chl a concentration. This variation represents the range of potential conversion factors that could be used to convert satellite chl a measurements to cell equivalents. Separate power regressions were again performed for the data from each bloom year to illustrate interannual variation in chl a cell -1.

Literature survey of M. polykrikoides chl a and cell count data

Published field studies on M. polykrikoides from other ecosystems were collated and used to calculate the variations in chl a cell -1 for comparison with the data obtained from the York River. The average, standard deviation as well as the minimum and maximum valued for chl a cell -1 were also calculated from the data obtained in this study and from the published studies. These data offer quantitative estimates of the variation chl a cell -1 during geographically diverse M. polykrikoides blooms (Gárate-Lizárraga et al., 2004; Gobler et al., 2008; Noh et al., 2018; López-Cortés et al., 2014; Al-Azri et al., 2014; Tomas and Smayda, 2008).

Laboratory studies to quantify how temperature, light, salinity, and nutrient limitation impact on chl a cell -1

As part of the second objective of this study, laboratory studies were undertaken to quantify how different growth conditions affect chl a content per cell of A. monilatum and M. polykrikoides. The results provide insights into how the prevailing growth conditions during blooms occurring in diverse geographic areas might impact the chl a cell -1. This information was obtained by growing both species under different salinity, temperature and light conditions in batch cultures and measuring cell and chl a concentrations over two-to-three-weeks. The resulting chl a and cell count data were used to calculate how chl a cell -1 varied among the different treatments.

Experimental design

The temperature, light, salinity and nutrients chosen for growing cells were based on literature data regarding the conditions found during blooms of both species. Blooms of A. monilatum typically occur between salinities of 20 and 35 (Juhl, 2005; Calvo, 2005; Perry et al., 1979), while bloom of M. polykrikoides commonly occur when salinities are between 25 and 34, with blooms reported between 4.4 and 40.1 globally (Kim et al., 2004; Marshall et al., 2006; Mulholland et al., 2018; Carney-Almeida et al., 2024; Pérez-Vega et al., 2025). Alexandrium monilatum commonly blooms between 25 °C and 32 °C with autumn blooms persisting until the temperature drops below 20 °C (Perry et al., 1979; Kim et al., 2004; Mulholland et al., 2018; Killberg-Thoreson et al., 2021). Margalefidinium polykrikoides blooms most often occur between 22 °C and 30 °C, with blooms reported at temperatures ranging from 16 °C to 36 °C (Kim et al., 2004; Azanza et al., 2008). Both species grow well as at irradiances between 30 and 400 µmol photons m -2 s -1 (Kim et al., 2004; Pérez-Vega et al., 2025). The growth conditions selected for these experiments include the upper and lower salinity ranges where blooms occur, the higher water temperatures observed during a majority of blooms and light levels optimal for growth and commonly observed in the field. Although these conditions do not encompass the full range of growth conditions, particularly with respect to temperatures that might be encountered in the field, they do provide a representative estimate of how much chl a cell -1 can vary under different growth conditions.

The A. monilatum and M. polykrikoides clones used in these experiments were isolated from the York River Estuary and provided by Dr. Carmelo Tomas and Robert York, University of North Carolina at Wilmington (ARC253, A. mon YR–Almo0909-1 clone) and Bill Jones, VIMS (Group III VA clone), respectively. The stock cultures of each isolate were maintained in L1 medium (minus silica, Guillard and Hargraves, 1993), which was made of 0.2 µm filtered Gulf Stream seawater. Cells were grown in 250 mL tissue culture flasks with vented caps (BD Biosciences, Bedford, MA, USA), kept in a Percival Scientific incubator (Boone, IA, USA), and maintained at constant temperature on a 12:12 h light:dark cycle. Photosynthetically active radiation (PAR) was maintained by horizontally mounted fluorescent lamps (Vita-lite, Duro-test Inc., Philadelphia, PA, USA). Light levels were adjusted using neutral density screens. The seawater was mixed with deionized water to achieve a salinity of 20 or 33 and then used to make L1 media.

The different salinity L1 media were sterilized using microwave treatment, a method developed by Keller et al. (1988). The culture pH was monitored using a Thermo Orion 3-Star pH meter and a Ross ultra-combination pH electrode (Thermo Fisher Scientific, Waltham, MA, USA) to ensure that pH levels stayed between 8.1 and 8.4 to prevent CO 2 limitation. Growth experiments were conducted at 20 °C and 27 °C with both salinities. Cells were acclimated at each salinity for two to 3 weeks prior to the start of each experiment. Light levels tested ranged from 46 to 365 µmol photons m -2 s -1 and were measured using a Biospherical Instruments Inc. QSL-100 4π wand meter. Cultures were acclimated to the specific light, temperature and salinity conditions several weeks before initiating an experiment.

At the start of each experiment, acclimated cells were transferred into 150 mL fresh media and allowed to grow for a period of between two and three weeks. This duration allowed cells to transition from a nutrient-replete to a nutrient-limited status providing a means to determine how nutrient status would impact chl a cell -1. Sampling began one day after the transfer into fresh media and continued at three-to-four-day intervals. During each sampling, the culture was gently mixed and triplicate 1 mL subsamples were removed and processed using a Beckman Coulter Multisizer™ three particle counter (Beckman Coulter Inc., Brea, CA) equipped with a 280 µm aperture to determine cell concentration and biovolume cell -1. Simultaneously, triplicate, 10–12 mL subsamples were also collected for chl a analysis. Chl a samples were extracted and analyzed using the Welschmeyer (1994) method as described above.

Validation of satellite chl a in the York and James Rivers

While examining the accuracy of the satellite chl_switch to pixels in more turbid, tributary waters of the York and James Rivers, a bias relative to field chl a of approximately 5 μg L -1 was observed. To remove this bias, the chl a was adjusted by −5.0 μg L -1 in subsequent analyses (Figure 4), resulting in a MedAD of 2.58 and bias of 1.065 μg L -1. The daily satellite chl a was consistent throughout the satellite time series (spring 2016–2025) at all stations (Figure 5). Peaks in the satellite occurred in several instances, indicating bloom conditions. The estimated values of chl a were similar to one another with the satellite time series showing reasonable seasonal variability.

Variation in A. monilatum and M. polykrikoides chl a cell -1 from field samples collected in this study and from the literature

Power regressions were carried out for the data from each of the years in which A. monilatum and M. polykrioides blooms were sampled. The R 2 correlation coefficient for the regression analysis of each HAB species cell concentration versus chl a is as an indicator of how much chl a is contributed by other species and to a lesser extent any variations in growth conditions (Figures 6A,C,E). The lower the correlation coefficient, the higher the contribution of chl a from other co-occurring species, and vice versa. The matching R 2 values for chl a versus chl a cell -1 indicates whether there is a substantial increase or decrease in chl a cell -1 with increasing chl a concentrations. Low correlation indicates no substantial trends were found. In that case, the range in chl a cell -1 over the range of chl a concentration sampled reflects the full range of possible chl a cell -1 conversion factors that could occur at any chl a concentrations found in that bloom.

The A. monilatum power regression R 2 values for the log cell concentration versus log extracted chl a equaled 0.83 in 2016 but only 0.02 for 2017 (Figure 6A). The corresponding R 2 values from the power regression of log chl a versus chl a cell -1 were 0.1 and 0.45 for 2016 and 2017, respectively (Figure 6B). In the case of 2016, the trend was for decreasing chl a cell -1 as chl a concentrations increased. The range between the lowest and highest chl a cell -1 values ranged from approximately 10-fold in 2016 to 100-fold in 2017 (Figure 6B). The M. polykrikoides the R 2 values for log extracted chl a versus chl a cell -1 were 0.06, 0.01, 0.61, and 0.12 for 2016, 2017, 2021, and 2022, respectively (Figure 6C). The samples from bloom years 2018 and 2019 contained too few samples for a regression analysis. The R 2 values for log chl a versus chl a cell -1 were 0.0003, 0.74, 0.001, and 0.01 for 2016, 2017, 2021 and 2022, respectively. Similar to A monilatum the variation between the lowest and highest chl a cell -1 observed at a given chl a concentration typically again ranged from 10- to 100-fold depending on the bloom. There was also no consistent change in chl a cell -1 over a wide range of chl a concentrations which could be applied to all blooms (Figure 6D).

The power regression R 2 values between cell concentration and chl a concentration from the different literature studies of M. polykrikoides were also variable: Gárate-Lizárraga et al. (2004), R 2 = 0.98; López-Cortés et al. (2014), R 2 = 0.96; Noh et al. (2018), R 2 = 0.75; Tomas and Smayda (2008), R 2 = 0.26; Al-Azri et al. (2014), R 2 = 0.06 (Figure 6E). The Gobler et al. (2008) study contained only one data point and was omitted. The R 2 values between chl a concentration and pg chl a cell -1 (conversion factor) from the different literature studies were as follows: Gárate-Lizárraga et al. (2004), R 2 = 0.28; López-Cortés et al. (2014), R 2 = 0.05; Noh et al. (2018), R 2 = 0.17; Tomas and Smayda (2008), R 2 = 0.23; Al-Azri et al. (2014) (Figure 6F). The variation in chl a cell -1 at a given chl a concentration was generally between 5- and 80-fold and are consistent with those obtained from the York River estuary in this study.

Laboratory studied quantifying how combined temperature, light, salinity, and nutrient limitation impact chl a cell -1

The laboratory studies showed that chl a concentrations in A. monilatum and M. polykrikoides cultures increased with increasing cell concentrations, but to different degrees depending on environmental growth conditions (Figures 7A,B). The increase in chl a with cell concentration was much lower at a salinity of 33 compared to 20. This difference in chl a was often 100-fold or more. For cultures grown at the same salinity, higher light generally resulted in lower chl a at a given cell concentration, while the variations in temperature tested had no discernible impact. For M. polykrikoides, where the data were more limited in scope, the observed difference in chl a at a given cell concentration was five-fold or less. The higher light conditions generally produced lower chl a concentration at a specific cell concentration (Figures 7A,B).

The relationship between cell concentration and chl a cell -1 demonstrates the impact of increasing nutrient limitation of cell grown in batch culture (Figures 7C,D). In every treatment chl a cell -1 declined as nutrients (primarily N) became more limiting, which occurred as the culture was allowed to grow without the addition of new nutrients. From the start to the end of an experiment this decline was on the order of 100-fold or more for A. monilatum and 5- to 10-fold for M. polykrikoides depending on the salinity, light and temperature conditions.

The plots in chl a concentrations versus chl a cell -1 illustrate the relationship between measured chl a and the potential values of chl a cell -1 that can result from different growth conditions (Figures 7E,F). Dashed black lines were included to emphasize the chl a cell -1 concentrations at the beginning and end of each experiment in response to increasing nutrient limitation. Different colored boxes were added to emphasize the impact that growth salinity alone had on chl a cell -1. The variation in chl a cell -1 at a given chl a concentration within each salinity box is due to differences in light or temperature growth conditions. Except at the highest light treatments, where chl a cell -1 varied from 2- to 10- fold lower at a given chl a concentration, these factors had relatively little impact on modulating chl a cell -1 compared to nutrient limitation and salinity. These factors caused a 100-fold or more variation in A. monilatum chl a cell -1 and a 5- to 10-fold variation in M. polykrikoides. Overall, the species-specific mean variations in chl a cell -1 observed in both the laboratory and field samples analyzed in this study were similar to those from other published work (Table 1).

The results of this study confirmed that though OLCI imagery can be used to accurately measure chl a concentrations in the highly productive and turbid tributaries of lower Chesapeake Bay, and is a valuable resource for determining the extent and location of high biomass blooms to aid monitoring efforts (Figures 1–5), it cannot be used to estimate HAB cell concentration with precision. This is due to variations in the measured chl a cell -1, which are used to convert the satellite-derived chl a measurements to HAB species cell concentration. The field data showed that for samples with equivalent chl a concentrations, the corresponding chl a cell -1 values could vary by 5- to over 100-fold (Figure 6). This degree of variation was typically consistent across a broad range of chl a concentrations observed in a bloom. Since the specific chl a cell -1 conversion factor applicable to estimate cell concentrations cannot be determined, the actual HAB concentration can only be estimated as falling somewhere between that derived using the lowest chl a cell -1 estimate and the highest estimate (5- to 100-fold range) (Figure 6).

This current chl a validation study confirms that accurate satellite-derived chla measurements could be obtained in the turbid, high CDOM tributaries of Chesapeake Bay where A. monilatum and M. polykrikoides are often most intense, extending the application of methods previously used by Wynne et al. (2022) in the Chesapeake Bay mainstem (Wynne et al., 2022) to the rivers. That study showed that chl a concentrations retrieved with the RE10 algorithm are consistent with field measurements. As a result, NCCOS have routinely provided satellite image products for the past decade showing high-resolution chl a distributions in the Bay to successfully guide sampling by state environmental agencies in states bordering Chesapeake Bay (Wolny et al., 2020) and use satellite chl a to initiate models designed to predict M. polykrikoides distribution (Xiong et al., 2023; Yu et al., 2025). We showed here that while there was a 5 μg L -1 bias in the satellite imagery for the tributaries of lower Chesapeake Bay, once adjustments for this bias were made, the MedAD dropped to 2.58 μg L -1 with no bias. This confirmed that satellite chl a estimates are consistent with the results of Wynne et al. (2022), despite difficulties in comparing field to satellite chl a, related to sub-pixel variability and bloom patchiness in the field (Kutser, 2009). The ability to track the distribution and movement of these blooms can be used in conjunction with habitat models to help identify the likely HAB species present and aid decisions on how to most effectively prioritize sampling efforts.

The variation in chl a cell -1 observed in blooms has two primary causes. The first is physiological. The laboratory experiments showed that prevailing growth conditions such as salinity, temperature, light and nutrient status greatly alter the amount of chl a per cell (Figure 7). In this study, salinity and nutrient availability were found to have the most profound impacts on these relationships (Figure 7; Table 1). Chl a content per cell was consistently higher at a growth salinity of 20 compared to 33. The cells in all growth conditions (equivalent salinity, light, and temperature) contained progressively less chl a cell -1 as cells in the cultures became increasingly more nutrient limited (Figures 7C,D). Cells grown at lower light conditions where other growth conditions were equivalent usually contained more chl a per the same number of cells. These results are consistent with other physiological studies of dinoflagellates that show cells up or downregulate chl a cell -1 in response to nutrient and light availability, or at differing growth salinities (McLachlan, 1961; Prézelin, 1976; Prézelin and Alberte, 1978; Prézelin and Matlick, 1983; Dagenais-Bellefeuille and Morse, 2013; Li et al., 2021). When considered together, the various growth conditions experiments in this study and others, show the full range in possible chl a cell -1 that might be observed among blooms occurring in estuarine and coastal regions worldwide. However, within any given bloom, the prevailing growth conditions, will foster only a subset of this full range of possible chl a cell -1 variations. For example, salinity may cause the chl a cell -1 at a given chl a concentration to be generally higher or lower within a bloom (Figure 7E). Increasing nutrient limitation will substantially downregulate chl a cell -1 at both salinities, but over very different, non-overlapping ranges. Interestingly, the field data for the 2016 A. monilatum bloom and the study of M. polykrikoides by Gárate-Lizárraga et al. (2004) are the only ones to show a general decrease in chl a cell -1 as chl a concentrations increased (Figures 6B,F). This observation suggests most blooms were not severely nutrient limited, even at high cell densities.

The second driver of chl a cell -1 variation is chl a contributed by co-occurring species. The tendency when working with high density HABs species is to assume the blooms are mono-specific, which is seldom the case. Given this reality, when chl a cell -1 conversion factors are estimated by dividing the bulk chl a in a sample by the HAB species cell concentrations, the chl a from co-occurring phytoplankton is erroneously ascribed to the HAB species. The more extraneous chl a being contributed by other species, the less accurate the chl a cell -1 conversion factor. Though the data in this study do not allow a precise determination of how much environmental growth conditions versus chl a contributed by other species drives the variation in chl a cell -1 at a given chl a concentration, there are indications that chl a contributed by other species is important. Specifically, the laboratory studies show that in each of the separate growth experiments there was a strong relationship between increase in cell concentration and chl a concentration (Figures 7C,D). If the actual blooms were monospecific, the cell concentrations versus chl a relationship would be expected to be similarly strong, which was true for some blooms but not the majority (Figures 6A,C). Blooms where HAB cell concentrations and chl a correlated more strongly usually exhibited lower variation in the chl a versus chl a cell -1 relationship (Figure 6). However, when high variability was observed in chl a cell -1, it can be assumed that co-occurring phytoplankton are contributing more chl a relative to the target HAB species. Additionally, the growth conditions observed in the field study were unlikely to generate the full range of growth conditions and chl a cell -1 values observed in the laboratory. Despite this fact, the mean chl a cell -1 values from the laboratory and field were comparable (Table 1). This observation is again consistent with chl a contributed by co-occurring phytoplankton, in conjunction with prevailing growth conditions, causing the observed variations in field chl a cell -1 measurements.

Cumulatively, the results indicate that attaining the most accurate cell estimates from satellite chl a would require collecting and processing field chl a and corresponding phytoplankton community biomass estimates in near real-time during the course of the bloom. The resultant dataset would allow calculation of accurate chl a cell -1 conversion factors that quantify the chl a contributions from co-occurring phytoplankton under prevailing environmental conditions. These conversion factors could then be applied to convert the next OLCI-derived chl a measurement to cell estimates over a wide region. Sampling may need to be repeated every few weeks, as growth conditions change. The direct cell counts would provide a way to check the accuracy of the chl a to cell concentration estimated. This three-step method (satellite to chl a, direct chl a cell -1 measurement, and chl a to cell concentration estimates) is similar to that described in Stumpf et al. (2016) for estimates of microcystin concentration in cyanobacterial blooms.

This study demonstrated that processed OLCI satellite imagery produces accurate estimates of chl a concentrations for phytoplankton blooms occurring in the tributaries of lower Chesapeake Bay. The data from A. monilatum and M. polykrikoides bloom samples and from laboratory studies showed the variation in the chl a per cell conversion factor needed to calculate HAB cell concentrations from satellite derived chl a measurements can vary by 5- to 100-fold (Figures 6, 7). This finding is likely true for other HAB species and is due to chl a contributed by co-occurring species and differences in prevailing growth conditions, which impact variations in the chl a cell -1 estimates used to convert satellite-derived chl a measurements to cell estimates. Consequently, determining more accurate HAB cell estimates can only be achieved by directly sampling chl a and cell concentrations in an ongoing bloom to generate more accurate chl a cell -1 conversion factors. Modeling HAB cell concentration based on chl a cell -1 from other studies of that species will likely produce inaccurate results. More accurate results may be obtained by periodically sampling blooms to obtain current, bloom-specific chl a cell -1 conversion factors, and using those factors to estimate the prevailing cell concentration. Despite the demonstrated limitations on estimating cell concentrations from satellite chl a, direct detection of high chl a features from satellite has proven to be a useful monitoring tool by providing guidance on bloom location, extent, and relative change in concentration, as has been demonstrated in Chesapeake Bay over the past decade (Wolny et al., 2020). This allows managers to more effectively utilize their sampling resources to collect cell concentration and toxicity data needed to better assess risks to public health and aquaculture operations.