The Lafayette River, a sub-tributary of the lower James River in the lower Chesapeake Bay, is known to be a hotspot for M. polykrikoides bloom initiation (Morse et al., 2013; Mulholland et al., 2018). A time series site in the Lafayette River at the Norfolk Yacht and Country Club (NYCC; Figure 1), has been used for weekly to sub-daily sampling each summer (June to September) since 2012, to target the onset, proliferation, and decline of summer HABs, particularly M. polykrikoides. The average water depth at NYCC is 6 m, and it has a 1–2 m tidal range. As part of the routine time series sampling, water column profile data are collected and water samples are collected for several analyses, including chlorophyll a (hereafter abbreviated to Chl-a) concentration and phytoplankton identification and cell enumeration. In addition to the routine sampling, four separate “Diel Studies,” designed to capture and quantify the vertical migration patterns and rates of M. polykrikoides were undertaken during the summers of 2016 and 2021 when M. polykrikoides was greater than 80% of the total phytoplankton biomass. The data from the routine sampling were used to track the temporal progression of M. polykrikoides populations and to target the timing of the Diel Studies. The methods for all sampling and analyses are described in detail in the following sections. All times given are local, UTC – 4 h, and dates are reported in DD/MM/YYYY format.

The Diel Studies were short time series experiments designed to observe and quantify the DVM rate of M. polykrikoides in the field. Diel Studies were only conducted when a bloom of M. polykrikoides was observed to be in progress at NYCC, when M. polykrikoides biomass accounted for over 80% of the total phytoplankton biomass. To characterize the phytoplankton community, whole water samples were collected around 12:00 local time from 0.25 m below the surface in 50 mL conical centrifuge tubes using a Masterflex peristaltic pump and polyethylene tubing. Samples were preserved using 2% acid Lugol’s solution for later identification and enumeration of phytoplankton cells by light microscopy on a Nikon TS100 inverted microscope under brightfield illumination in the Phytoplankton Analysis Laboratory at Old Dominion University, hereafter referred to as the ODU Phyto Lab. Sample bottles were gently inverted 10 times before a Palmer Maloney Counting Cell was loaded with 0.1 mL of Lugol’s preserved whole water sample and allowed to settle for about 3–5 min. An initial scan at low magnification was conducted to ensure no bubbles were within the chamber, and that phytoplankton cells were settled properly in the viewing plane. Full taxonomic enumerations of each chamber were performed in a transect fashion while avoiding the loading channels (LeGresley and McDermott, 2010) under 300× magnification by the ODU Phyto Lab. This method was specifically chosen in order to enumerate phytoplankton from a wide range of size classes. Although a smaller magnification could be used to enumerate M. polykrikoides, the higher magnification aids in the enumeration of smaller cells in the densely populated and diverse samples from the estuarine environment of the Chesapeake Bay (Karlson et al., 2010). The carbon biomass for each taxa observed was calculated using the accepted Chesapeake Bay Program biovolumes and carbon conversions estimated for each species enumerated (Interstate Commission on the Potomac River Basin, 2008). At the same time, additional whole water samples (25–50 mL) were filtered onto Whatmann GF/F filters for Chl-a analysis. Chl-a concentrations were measured fluorometrically using the non-acidification method on a Turner 10-AU fluorometer after extraction in acetone (Welschmeyer, 1994) and the Chl-a data were used to calibrate the YSI fluorescence profiles (see Supplementary Figure S1). All discrete Chl-a samples were analyzed within 2 weeks of their collection.

A total of four Diel Studies were completed, three in 2016 and one in 2021, over the following dates and start/end times: 12:00 16/08/2016 to 17:00 17/08/2016 (DS1, 29 h); 12:00 22/08/2016 to 16:00 23/08/2016 (DS2, 28 h), and 16:30 09/09/2016 to 16:00 10/09/2016 (DS3, 23.5 h); 12:00 06/08/2021 to 09:00 07/08/2021 (DS4, 21 h). Each Diel Study was conducted over a period of roughly 24 h but ranged in duration from 21 to 29 h. During each Diel Study vertical profiles of temperature, salinity and Chl-a fluorescence were collected at 1–2 hourly intervals using a YSI Model 6600 sonde. Additionally, during DS1, DS2, and DS3, a subset of water samples were collected from a range of depths to characterize the phytoplankton community, albeit at a lower vertical and temporal resolution than the YSI profiles. This sampling scheme allowed us to resolve both vertical and temporal variability in water column conditions over the daily cycle.

Previous studies have estimated DVM rates by tracking the relative abundance of M. polykrikoides with depth over the diel cycle (Park et al., 2001; Kim et al., 2010). Since the abundance of M. polykrikoides varies spatially as a result of biological and physical processes unrelated to DVM, by tracking the maximum relative abundance with respect to depth, we can track the DVM behavior of M. polykrikoides independently of variations due to spatial heterogeneity. However, unlike previous studies, here we use Chl-a as a proxy for M. polykrikoides biomass and track the relative concentration of Chl-a with depth, defined by Equation 1. Using Chl-a as a proxy allows for greater depth resolution and a larger number of replicate studies as it greatly reduces the labor required to analyse the data from each Diel Study. Although Chl-a is representative of the entire photoautotrophic community, Diel Studies were only undertaken during periods when M. polykrikoides accounted for the majority (>80%) of the total phytoplankton biomass. This allowed for the use of Chl-a as a good proxy for M. polykrikoides abundance. Further, we assume that the DVM behavior of M. polykrikoides is the same for any given water parcel in the vicinity of the study site, so that even if water parcels are moving through the fixed point in space where the sampling occurs, the timing and rate of M. polykrikoides DVM should be relatively invariant. Given these assumptions and constraints, we use the Chl-a concentration at any given depth relative to the total depth integrated Chl-a to track the vertical migration of M. polykrikoides over the course of the Diel Studies:

where the Rel_CHL(z) is the proportion of total depth-integrated Chl-a at depth z. DVM rates, w_DVM (m h⁻¹), were calculated from the change in z_MAX, the depth where the maximum value of Rel_CHL was identified between adjacent time points, ∆z_MAX (m), over the length of time between adjacent time points, ∆t (hr):

Chl-a data were averaged into 0.2 m bins and the frequency of sampling ranged from 1 to 2 h. The method presented here to estimate DVM rates relies on being able to identify a Chl-a maximum at some depth in the water column. When it is not possible to unequivocally identify a Chl-a maximum, such as when the water column is very well mixed, this method cannot be applied. To systematically identify periods when Chl-a was homogeneously distributed throughout the water column, we calculated the coefficient of variation of Chl-a for each of the Diel Study profiles. Here the coefficient of variation of Chl-a quantifies the relative degree of dispersion of the Chl-a values measured along each profile with respect to the mean, so when its value is low, the variability of Chl-a with depth is low. We identified profiles where the coefficient of variation of Chl-a was ≤5% and excluded those low variability profiles from our calculations of w_DVM in order to minimize errors or biases that might be introduced by using data where no clear Chl-a maximum could be identified.

During each of the studies, the water column temperatures were clearly influenced by the diel cycle of heating and cooling (Figure 2), with the highest surface temperatures observed after noon and with their deepest penetration into the water column occurring later in the day. Water column temperatures were coolest in the early morning, although some variability was observed, likely due to differences in the tidal cycle and wind conditions between each of the studies. Water temperatures were most elevated during DS1 (Figure 2A), with water temperatures in the top 1 m exceeding 32°C in the late afternoon (16:00–17:00) on the first and second days of the study and reaching a maximum observed temperature of 32.3°C at 16:00 on 17/08/2016. The lowest water temperatures during DS1 were around 28.9°C and only found below 4.4 m depth on the morning of 17/08/2016 (08:00–10:00). Water temperatures during DS2 (Figure 2B) followed a similar pattern to DS1, but were cooler, with a maximum temperature of 29.6°C, and warm waters (>29.5°C) extending as deep as 3.6 m depth on 22/08/2016 (18:00). The minimum temperature observed during DS2 was 28.0°C and only seen below 4 m depth on the morning of 23/08/2016 (08:00). DS3 (Figure 2C) and DS4 (Figure 2D) both had much cooler water temperatures overall, with their maximum observed temperatures of 27.7°C and 27.4°C, respectively, both cooler than the lowest temperatures observed during DS1 and DS2. The lowest temperatures observed during DS3 and DS4, again were similar, at 26.6°C and 25.7°C, respectively. During each of the studies, we observed a period of relative water column temperature homogeneity, generally during the morning, coinciding with the lowest temperatures. This period was considerably longer during DS2 (Figure 2B) compared to the other studies, with relatively lower and more consistent temperatures with depth from 00:00 to 12:00. The widest range of temperatures was observed during DS1 at 3.4°C, and the lowest range of temperatures was observed during DS3 at 1.1°C. DS2 and DS4, despite the difference in the overall temperatures during the studies, had a similar range at 1.6°C and 1.7°C, respectively.

The Lafayette River is tidal, resulting in regular sub-daily variability in salinity at NYCC during the Diel Studies (Figure 3). We observed alternating periods of high/low salinity associated with high/low tides. The relatively fresher waters were concentrated near the surface, with higher salinity waters extending through the full depth of the water column, and sometimes persisting in the deeper waters during periods of low tide, for example during the first half of both DS3 and DS4 (Figures 3C,D). We observed very similar patterns in salinity during DS1 and DS3 with salinity maxima of 23.9 PSU and 23.1 PSU, respectively, and salinity minima of 21.1 PSU and 21.5 PSU, respectively. DS2 had the highest salinities (Figure 3B), with a maximum of 25.2 PSU observed primarily lower in the water column, but extending to the surface at 12:00 on 23/08/2016, the second day of DS2. By contrast, we observed the lowest salinities during DS4 (Figure 3D), with a salinity maximum of 22.3 PSU observed below 4 m depth at 15:00 on 6/08/2021, the first day of the study. The salinity minimum during DS4 was 14.9 PSU at the surface at 08:00 on 7/08/2021, the second day of the study, this was likely a shallow lens caused by a surface input of fresh water. Apart from this extreme low salinity feature, low salinities <20 PSU were observed in the top 1 m of the water column below the fresh lens and during the subsequent profile at 09:00. The widest range of salinity was observed during DS4 at 8.7 PSU. DS1 had an intermediate range of salinity at 2.8 PSU. DS2 and DS3 had a relatively lower range of salinity at 1.8 PSU and 1.6 PSU, respectively.

The Diel Studies were timed to coincide with periods during which M. polykrikoides blooms were present in the water column at NYCC. Daily abundances of phytoplankton were enumerated as part of the routine sampling at NYCC, and from that data, the proportion of the total phytoplankton biomass, as a function of carbon, represented by M. polykrikoides was estimated (Table 1). In most cases the abundance of M. polykrikoides was near or above 1,000 cells mL⁻¹, and M. polykrikoides accounted for over 80% of the total phytoplankton biomass, except for the second day of DS2 (23/08/2016), when M. polykrikoides abundance was only 530 cells mL⁻¹, and it accounted for 73.2% of total phytoplankton biomass in the surface. However, on the first day of DS2 (22/08/2016) when the study was initiated, M. polykrikoides abundance was above 10,000 cells mL⁻¹ and it accounted for over 94% of the total phytoplankton biomass. Similarly, on the first day of DS3 (9/09/2016 at 16:00), M. polykrikoides abundance was only 890 cells mL⁻¹, but it accounted for 84.8% of the total phytoplankton biomass, well above the 80% cutoff. For DS4, we report M. polykrikoides abundance data collected 4 days prior to, and 2 days after, the Diel Study was completed, and in both cases M. polykrikoides accounted for >94% of the total phytoplankton biomass. Based on the M. polykrikoides abundances and percentage of total biomass prior to and after DS4 as well as spot checks of the community composition undertaken with a PlanktoScope imaging microscope in the field, we are confident that M. polykrikoides was dominant during DS4. Unfortunately, we did not fully analyze the PlanktoScope image data, and so cannot provide abundance data from the spot checks.

Temporal changes in the vertical distribution of both Chl-a (Figure 4) and Rel_CHL (Figure 5) were observed in each of the Diel Studies. Broadly, we observed the highest Chl-a concentrations in the surface waters during the daytime, with the pattern reversing overnight when the highest Chl-a concentrations tended to be found near the bottom (Figure 4). Across all the Diel Studies, the highest Chl-a concentrations were observed during DS1, when they reached 455.4 μg L⁻¹ in surface waters around 17:00 on the first day of the study. This was an extreme high concentration, but Chl-a concentrations >200 μg L⁻¹ were found to be associated with the near surface and near bottom Chl-a maxima throughout DS1. The minimum observed Chl-a concentration during DS1 was 12.0 μg L⁻¹ in near bottom waters around 10:00 on the second day of the study (Figure 4A). Similar Chl-a maxima and minima were observed during the other three Diel Studies, with minima of 9.0 μg L⁻¹, 9.7 μg L⁻¹ and 3.7 μg L⁻¹ and maxima of 108.9 μg L⁻¹, 92.2 μg L⁻¹ and 116.5 μg L⁻¹ for DS2, DS3, and DS4, respectively (Figures 4B–D). As with DS1, the maximum Chl-a values for each study were associated with daytime near surface or nighttime near bottom maxima.

Diel patterns in the vertical distribution of Chl-a were more easily observed in the temporal evolution of Rel_CHL, and the variation in z_MAX during the Diel Studies (Figure 5). In most cases, high Rel_CHL values and z_MAX were found near the bottom at night and in the early morning hours, in the surface and near surface waters between early morning and early afternoon, and in mid-depth waters and near the bottom in the late afternoon and evening. This variation in the depth distribution of Rel_CHL and z_MAX over the daily cycle reflected M. polykrikoides’ daily vertical migration since M. polykrikoides vastly dominated the phytoplankton community when these studies were undertaken (Table 1). In most cases, we observed the onset of high Rel_CHL values in the surface waters (<2 m) as early as 09:00, persisting in the top 1-2 m of the water column until at least 16:00. Between 16:00 and 20:00, the highest values of Rel_CHL were observed at intermediate depths between the surface and the bottom, with high Rel_CHL values found mostly at the bottom from around 20:00 to 04:00 the following day. Based on these consistent patterns in the depth distribution of Rel_CHL and z_MAX over the daily cycle, we broadly defined M. polykrikoides’ ascent window as 04:00 to 09:00, and their descent window as 16:00 to 20:00. Although this diel pattern was relatively consistent across all four Diel Studies, we particularly highlight the absence of a clear ascent pattern in Rel_CHL during DS3 (Figure 5C), where z_MAX remains confined to the top 1 m of the water column from approximately 00:00 to 16:00 on 10/09/2016 when the study ended.

We examined the samples taken during DS1, DS2 and DS3 to determine whether there was a relationship between Chl-a concentration and either the abundance of M. polykrikoides (Supplementary Figure S1A) or its percentage of the total biomass (Supplementary Figure S1B) and did not find any strong correlation between Chl-a concentration and either of these metrics of M. polykrikoides. However, when we examined the relationship between Rel_CHL and the relative depth abundance of M. polykrikoides (Figure 6) we found a strong positive relationship reflected by a statistically significant (p < 0.01) Spearman rank correlation of r = 0.74. This strongly supports the use of Rel_CHL as a proxy for the vertical position of M. polykrikoides during bloom conditions.

M. polykrikoides DVM rates, w_DVM, for each of the Diel Studies were estimated based on the temporal evolution of z_MAX, the depth of the highest Rel_CHL, over the duration of each of the studies. Using Equation 2 we calculated a mean and maximum w_DVM for each ascent/descent window in each Diel Study (Table 2). The mean w_DVM was determined by averaging all values of w_DVM between the time point where M. polykrikoides was last observed at the surface/bottom, and the next time point where M. polykrikoides was observed at the bottom/surface. The maximum w_DVM was simply the highest observed ascent/descent rate between adjacent profiles during any given Diel Study, although in cases like DS2 when the ascent was only resolved by two sequential profiles, we did not report a maximum w_DVM. The mean w_DVM for ascents ranged from 1.2 m h⁻¹ for DS1 to 1.3 m h⁻¹ for DS2 and DS4 (no ascent was observed in DS3), and for descents ranged from 1.2 m h⁻¹ for DS1 and DS3, to 2.2 m h⁻¹ for DS4. The maximum w_DVM for ascents ranged from 2.4 m h⁻¹ for DS1 to 2.5 m h⁻¹ for DS4, and for descents ranged from 1.4 m h⁻¹ for DS1 and DS2 to 3.2 m h⁻¹ for DS4. Given the relatively small number of replicates, 3 ascents (2 for maximum w_DVM) and 4 descents, we cannot unequivocally say whether the mean or maximum ascents or descents were consistently faster. We also note that the fastest mean and maximum w_DVM values were observed during DS4, which had relatively higher resolution temporal sampling at 1.0 ± 0.2 h between casts compared to 1.9 ± 0.5 h, 1.6 ± 0.5 h and 2.1 ± 0.8 h for DS1, DS2 and DS3, respectively.

The method presented here relies on M. polykrikoides being the dominant phytoplankton species present in the water column. Other vertically migrating dinoflagellate species such as Akashiwo sanguinea and Gymnodinium sp. also bloom in the Chesapeake Bay during the summer months, however they typically do not bloom at the same time as M. polykrikoides (Mulholland et al., 2018). An examination of the surface and water column phytoplankton community composition data collected before, after and during the Diel Studies (provided in Section S1 of Supplementary material) confirm that although other vertically migrating dinoflagellate species were present during the Diel Studies, they generally accounted for a very small proportion (<20%) of the total phytoplankton biomass, were at much lower abundances than M. polykrikoides and where depth-resolved data were available, were offset in time and depth from the bulk of the M. polykrikoides depth-resolved biomass.