Physical controls and ecological implications of the timing of the spring phytoplankton bloom on the Newfoundland and Labrador shelf

The timing of spring phytoplankton blooms is crucial to many species that have adapted their development to benefit from the enhanced feeding opportunity they offer. Any change to their timing may affect the productivity of an entire ecosystem. This study explores the relationship between the ocean climate, the timing of the spring bloom and the secondary production on the Newfoundland and Labrador shelf. It is found that over interannual cycles, the ocean climate is significantly correlated with the timing of the bloom and the abundance of Calanus finmarchicus, a key zooplankton species for the ecosystem. It also appears that the spring bloom is initiated by the onset of ocean re‐stratification following winter mixing. Understanding how annual variation in climate relates to the timing of the spring bloom and zooplankton abundance, that is, the base of the marine food web, can inform the development of ecosystem‐informed models for higher trophic levels.

subsequently transferred up the food chain, exported to the deepest layers of the ocean, or simply released near the surface and back to the atmosphere.The timing and evolution of spring blooms influence which higher trophic-level species will most benefit from the nutrient pool and eventually participate to the carbon export to the deep ocean layers (Allen et al. 2005), a key mechanism to mediate anthropogenic climate change (Riebesell et al. 2007).
The ocean's physical properties are driven by annual variation of its climate, that is, differences in the air and ocean (surface and bottom) temperature, ocean salinity, sea ice, and so on.For example, in seasonally ice-covered areas, the sea-ice retreat which is influenced by air and ocean temperature, plays a key role in initiating the spring bloom by increasing the incoming light in previously ice-covered waters (Wassmann 2011;Cheng et al. 2016).Sea-ice melt also promotes surface layer stratification and phytoplankton accumulation as a result of freshwater release (Wu et al. 2007).In the context of anthropogenic climate change, blooms may be occurring earlier as a consequence of warmer spring temperature or, following the above example, earlier sea-ice retreat (Leu et al. 2011).This can in turn affect energy transfers through the ecosystem by creating a mismatch between the timing of peak phytoplankton biomass and the demand from higher trophic levels, for example, zooplankton (Cushing 1990;Head et al. 2000).The impact of an ocean altered by a changing climate on the timing and magnitude of phytoplankton blooms remains however unclear.
The changes in the timing of the phytoplankton spring bloom is explored here for the Newfoundland and Labrador (NL) shelf (Fig. 1), a region that supported iconic fisheries for centuries (Schijns et al. 2021).Located under the direct influence of Arctic and sub-Arctic outflow (Florindo-L opez et al. 2020), the region is at the forefront of anthropogenic climate change (Greenan et al. 2019), while undergoing strong natural fluctuations of its climate.These fluctuations of the ocean climate have recently been characterized using a climate index incorporating several environmental variables (described in detail below; Cyr and Galbraith 2021).The effect of the climate and the timing of the spring bloom on the abundance of the dominant, energy-rich, zooplankton species, Calanus finmarchicus is also explored.Finally, while the initiation of the spring bloom has long been related to the retreat of sea ice (Wu et al. 2007), we propose a new mechanism based on the onset of stratification.In this region where sea ice has been gradually declining over the recent decades (Cyr and Galbraith 2021), stratification seems to better explain the initiation of the bloom.

Method
All data used in this study are archived in a public repository accessible at https://doi.org/10.5061/dryad.stqjq2c7k(Cyr et al. 2023).The detailed methodology is available in the Supporting Information and briefly summarized here.

Newfoundland and Labrador Climate Index
The ocean climate over the NL shelf can be described using the Newfoundland and Labrador Climate Index (NLCI; Cyr and Galbraith 2021; blue curves in Fig. 2).The NLCI is the annual average of 10 standardized time series (air temperature, sea ice, sea surface temperature, bottom temperature, etc.) representative of different aspects of the ocean climate in the region.It is a robust indicator of the state of the NL environment, and has been used in numerous stock assessments in relationship to the state of the ecosystem (see Cyr and Galbraith 2021 and references therein).The NLCI is openly available, covers the time period between 1951 and the present, and is updated annually (Cyr and Galbraith 2020).color map shows the bathymetric features, where some isobaths are identified, including the 1000-m isobath highlighted here with a thicker contour, used here to identified the limit of the shelf.A sketch of the main surface currents is shown in black.NAFO divisions 2J, 3K, and 3LNO are shown in slate gray.The red dots aligned in sections (also identified) represent the hydrographic stations where zooplankton samples used in this study were collected.Sta.27 (47 32:8 0 N,52 35:2 0 W) is identified with a red star.

Timing of the phytoplankton bloom
Satellite imagery (4-km resolution) of ocean color was used to characterize the phenology of the spring phytoplankton bloom for the shelf portion of Northwest Atlantic Fisheries Organization (NAFO) assessment divisions 2J, 3K, and 3LNO (Fig. 1).We used the R Shiny application PhytoFit (Clay et al. 2021) to calculate the timing of the initiation and maximum intensity (maximum chlorophyll a [Chl a] concentration) of the bloom using a method adapted from Zhai et al. (2011) and briefly summarized here.
For each NAFO division, the annual evolution of the daily spatially averaged Chl a concentrations between the pre-and post-spring bloom minima was approximated using a Gaussian fit.The timing of the spring bloom initiation (t bi ) was determined as the day-of-the-year (DOY) the fit reached 20% of the curve amplitude.The timing of maximum intensity of the spring bloom (i.e., bloom peak timing; t bp ) was calculated as the DOY for which daily mean surface Chl a concentration was maximum.Standardized anomalies (e.g., Eq.S1 in Supporting Information) of the timing of the bloom maximum intensity was further averaged over 2J, 3K, and 3LNO (green curves in Fig. 2).A brief description of the phytoplankton community and its ecological role is also provided in the Supporting Information.

Abundance of C. finmarchicus
Zooplankton samples were collected seasonally (spring, summer, and fall) at standard stations along oceanographic sections (Fig. 1) of the Atlantic Zone Monitoring Program (AZMP; e.g., Maillet et al. 2022).The annual mean abundances of C. finmarchicus by surface area (C; indm À2 ) were obtained by fitting a linear model (Supporting Information Eqs.S2, S3).To control for the order of the variables in the model, annual means were estimated using adjusted sums of squares.
Standardized anomalies of C. finmarchicus abundance for each section and for Sta.27 (see map in Fig. 1 for location) were calculated following Supporting Information Eq.S1 using the 1999-2020 reference period.The standardized anomalies of the abundance of C. finmarchicus were further averaged to obtain one time series representative of the NL shelf and NAFO divisions 2J, 3K, and 3LNO (orange curves in Fig. 2).

Stratification and MLD
Stratification and MLD data were obtained from monitoring Sta. 27, one of the longest hydrographic time series in Canada with regular (near-monthly or better) occupations since the late 1940s.Sta.27 is located in NAFO division 3LNO (47 32:8 0 N,52 35:2 0 W; Fig. 1).Its local ocean parameters are considered representative of the climate of the NL shelf as a whole (Petrie et al. 1991(Petrie et al. , 1992;;Colbourne et al. 1994;Han et al. 2015).A total of 1198 vertical temperature and salinity profiles acquired since 1998 were used to calculate the density anomaly referenced to the surface (σ 0 ) using the TEOS-10 Gibbs Seawater toolbox (McDougall and Barker 2011).
The stratification (approximated by the density difference between 5 and 150 m) and the MLD (defined as the depth of the maximum buoyancy frequency N 2 ) were further derived for each individual profile collected at this station since 1998.The climatological seasonal cycle of the stratification and the MLD is presented in Fig. 3a.The temperature (T) and salinity (S) contributions to the stratification (see Supporting Information) are presented in Fig. 3b.
Stratification at Sta. 27 generally decreases from its maximum in late summer to its minimum during the spring of the following year in a parabolic-like shape (orange curve in Fig. 3a).For each year, the decrease and further increase of stratification between the fall and the summer was approximated using a parabolic fit (red lines in Figs.S2-S4 in Supporting Information).The DOY when the minimum stratification is reached in the spring (t s ) was derived using the minimum of this fit.The standardized anomalies of this time series ( e t s ) are presented in Fig. 4 (blue curve).Unfortunately, the prediction for the timing of the bloom based on restratification was not possible for the period 2017-2021 due to insufficient data collected during winter.

Climate, bloom, and secondary production
On average, there is a south-to-north progression of the bloom on the NL shelf (Table 1).The timing of the bloom in these regions is however subject to large interannual modulations, highlighted here by averaging the standardized anomalies of the bloom peak timing ( f t bp ) over 2J, 3K, and 3LNO (Fig. 2a).
The timing of the spring bloom across the NL shelf is modulated by the ocean climate and adequately explained by the NLCI.The NLCI is negatively correlated with both the timing of the bloom initiation (r = À0.59,p = 0.007; Supporting  Information Fig. S7) and maximum intensity (r = À0.73,p = 0.001; 53% of variance explained) over the NL shelf.
Because of the better correlation between the NLCI and the maximum intensity, only the latter was reported here in Fig. 2a.This stronger correlation is probably due to the fact that the bloom peak signal is easier to capture from satellite over the broader area considered here, a region that exhibits significant cloud coverage in the spring (Isaac et al. 2020).The negative correlation also signifies that the bloom is delayed (positive anomaly in f t bp ) during the colder years (negative NLCI), and vice versa.
The NLCI and the timing of the phytoplankton bloom are further compared to the abundance of C. finmarchicus (Fig. 2b,c).The abundance of C. finmarchicus was weakly and not significantly correlated with the timing of the bloom (r = À0.42,p = 0.08; 18% of the variance explained) but positively and significantly correlated with the NLCI (r = 0.52, p = 0.02; 27% of the variance explained).This suggests that throughout interannual cycles, the ocean climate partially drives the timing of the phytoplankton bloom and the abundance of a key zooplankton species for the NL shelf ecosystem.

Stratification and the timing of the bloom
The onset of stratification following the winter mixing of the water column is further explored as a mechanism responsible for the initiation of the phytoplankton spring bloom.At Sta. 27, the seasonal evolution of the temperature (Supporting Information Fig. S1a) shows that the water column is generally homogeneous and below 0 C from January to March.As a result of mixing, the salinity progressively increases in the top $100 m of the water column from the early winter to the late spring (Supporting Information Fig. S1b).On average, the initiation of the spring bloom in 3LNO (where Sta.27 is located) corresponds to a decrease in salinity and an increase in temperature in the top 10 m of the water column (Supporting Information Fig. S1c).
More precisely, the initiation of the spring phytoplankton bloom in 3LNO (t bi ¼ 82 AE 12; see also Fig. 3a) occurs just after the minimum stratification at Sta. 27 is reached near DOY 77 (18 March on a non-leap year).The bloom's maximum intensity is reached on average 1 month later (t bp ¼ 112 AE 10 or 22 April) when the stratification has started to increase in a near quadratic fashion, but long before the maximum stratification is reached in early September (DOY 250).
The temperature contribution to stratification is negligible early in the year (red curve in Fig. 3b plunging toward negative values between DOY 40 and 80) due to nearly uniform vertical temperature (Supporting Information Fig. S1a).Except for the late summer when the temperature and salinity contributions to the stratification are about equal (near DOY 250), the stratification is largely driven by the salinity at Sta. 27.The restratification is thus driven by a freshening of the water column likely caused by river runoff and sea-ice melt upstream.The increase in near-surface temperature rapidly follows the onset of re-stratification of the water column (Fig. 3b).
For any given year, re-stratification index e t s is a good predictor of the timing of the spring bloom ( f t bi ) in 3LNO (Fig. 4).It explains 44% (r = 0.66, p = 0.002; Fig. 4) and 36% (r = 0.60, p = 0.002; not shown) of the variance of the initiation and maximum intensity timing of the spring bloom, respectively.On average over this period, the initiation of the spring bloom occurs 12 AE 10 days after the minimum stratification is reached.

The MLD and the timing of the bloom
The MLD exhibits a relatively linear decrease from the early winter ($ 75 m on average near DOY 0) to the middle of the summer when the minimum ($ 20 m on average near DOY 230) is reached (black curve in Fig. 3a).No clear relationship exists between the shallowing of the MLD and the timing of the bloom as indicated by the large scatter in the data.In addition, a closer look at the interannual time series of the MLD at the time of the bloom confirms the large spread in the MLD associated with the initiation of the bloom (anywhere between 10 and 100 m; Supporting Information Fig. S5).On average, the spring bloom occurs when the MLD is 59 AE 30 m deep.Altogether, this suggests that the MLD is not the main trigger of the spring bloom.

Mechanisms triggering the spring bloom on the NL shelf
The NL shelf is a coastal region dominated by intense seasonal cycles and under the influence of large freshwater fluxes from Arctic and sub-Arctic regions.This study suggests that the initiation of the phytoplankton spring bloom occurs once the external forcing responsible for the breaking down of the stratification (e.g., wind mixing, heat and salinity fluxes, etc.) reduces and the level of turbulence in the water column decreases.At this time, the water column rapidly re-stratifies and plankton cells begin to accumulate in the upper layers of the ocean.This description is analogous to the critical turbulence hypothesis (Huisman et al. 1999;Taylor and Ferrari 2011), which suggests that the initiation of the spring bloom should coincide with the end of the convective turbulent mixing season.In this framework, the bloom starts when the phytoplankton growth rate overcomes the mechanical removal of plankton cells from the photic layer by vertical overturning.The critical turbulence hypothesis is, in our view, identical to the onset of stratification hypothesis (Chiswell 2011;Chiswell et al. 2015), which highlights a difference between mixed and actively mixing layers, and posits that blooms can occur in the thin surface layer despite a deeper MLD below.Together, these two hypotheses adequately explain the initiation of the bloom on the NL shelf and the weak link between the MLD and the timing of the bloom (Fig. 3; see also Evans and Pepin 1989).
Although both temperature increase and freshwater release contribute to changes in water density, Wu et al. (2007) attributed the latter to be the principal cause of changes in stratification, which was further hypothesized as the mechanism linking the sea-ice retreat to the spring bloom initiation on the NL shelf.The results of this study support this framework ("Stratification and the timing of the bloom" section).However, we argue that linking re-stratification (and therefore the timing of the bloom) to the timing of the retreat of sea ice from its southernmost extent (Wu et al. 2007) may be problematic for two reasons.First, Cyr et al. (2022) show that sea ice on the Newfoundland shelf is mainly advected from the Labrador shelf.Therefore, ice may begin to melt before reaching its southernmost extent.Second, sea-ice extent has decreased dramatically in recent years (Cyr et al. 2022), potentially weakening its contribution to stratification.These factors, combined with the analysis of a longer time series than the one that was available to Wu et al. (2007), may explain the weakening of the relationship between sea-ice retreat and the timing of the spring bloom in recent years (see Section 3 in Supporting Information).The direct use of the stratification measured at Sta. 27 presented here appears to be a more useful proxy for the timing of the spring bloom.

Interannual variability of the timing of the bloom
On the NL shelf, cycles of milder and colder winters generally occur on decadal time scales.Cold winters are usually accompanied by delayed spring conditions, colder ocean temperatures, and the presence of heavy sea ice and numerous icebergs near the coast.In contrast, milder winters are generally synonymous with early sea-ice retreat, lower numbers of icebergs, and fresher ocean conditions.Because the winter and spring conditions also influence the ocean conditions for the rest of the year (e.g., through the generation of large volume of cold water), these cycles of the ocean climate are generally well captured by the NLCI (Cyr and Galbraith 2021).
It is thus not surprising that the NLCI, which includes parameters influencing the spring re-stratification of the water column (e.g., the harshness of the winter, air temperature, sea-ice melt), is a good predictor of the timing of the spring bloom.This study shows that a robust negative correlation is found between the NLCI and the timing of the spring bloom over the NL Shelf (53% of variance explained; Fig. 2).This means that positive phases of the NLCI (warmer ocean climate) are associated with early spring blooms and negative phases of the NLCI (colder ocean climate) to late blooms.

Match-mismatch and implications for higher trophic levels
It is generally understood that adult C. finmarchicus copepods emerge from overwintering at depths in the spring to feed on the phytoplankton bloom and reproduce (Hirche 1996).Match-mismatch between the emergence of copepods and the timing of the spring bloom can greatly influence the abundance of zooplankton that are, in the absence of an abundant food source, victims of starvation and/or cannibalism (Cushing 1990;Head et al. 2000Head et al. , 2015;;Dufour et al. 2016).
This study suggests that the abundance of C. finmarchicus fluctuates with the ocean climate and the timing of the phytoplankton blooms, with higher abundances found when the ocean climate is warm and the bloom earlier (27% and 18% of the variance explained; Fig. 2b,c).Noting that a warmer climate is also associated with earlier blooms (53% of the variance explained; Fig. 2a), a possible causal relationship is that years with warmer ocean conditions lead to earlier blooms, which in turn leads to a better match with the emergence of C. finmarchicus from diapause as demonstrated in the nearby Labrador Sea (Head et al. 2000(Head et al. , 2015)).Changes in water mass composition resulting from changes in the ocean climate (different proportion of subpolar vs. subtropical waters) are also expected to influence species composition on the NL shelf (Pepin et al. 2011) and may explain Fig. 2 observations.Another explanation is that colder phases of the climate are associated with an inflated subpolar gyre and a potential rerouting of the secondary production off the shelf and toward the Northeast Atlantic (H atún et al. 2016).
Shifts in zooplankton biomass and composition likely cascade up the food chain (Durant et al. 2005), influencing the abundance of forage fish such as capelin (Mallotus villosus), a keystone forage species for the NL ecosystem (Buren et al. 2014).Murphy et al. (2021) found a relationship between the timing of capelin spawning and the NLCI which supports this notion, although the cascading effect for the rest of the ecosystem has not been broadly examined.Buren et al. (2014) also showed a strong relationship between the timing of sea-ice retreat on the NL shelf and capelin biomass from the 1980s to 2010, although this relationship has weakened over the last decade (Lewis et al. 2019).

Climate change considerations and implications for the NL shelf
The world's oceans are becoming increasingly stratified as a result of anthropogenic climate change (Pörtner et al. 2019).This increase in stratification is projected to continue and may have many negative ramifications for the world's ecosystems.For example, it can negatively influence fisheries productivity by promoting mismatch between fish spawning timing and phytoplankton blooms (Asch et al. 2019).
Although stratification at Sta. 27 has shown a 3.9% increase between the 1950-1990 and the 1991-2020 periods-a number slightly above the 2:3% AE 0:1% increase of the world's ocean surface between 1971-1990and 1998-2017(Pörtner et al. 2019)-we did not find any significant trends in the timing of the re-stratification, and thus the prediction for the timing of the spring bloom (Supporting Information Fig. S6).Rather, the time series exhibits important decadal changes, with the timing of the re-stratification being later during a period corresponding roughly to 1970-2000, and earlier in the 1950s and between the late 1990s to the early 2010s.

Conclusion
Phytoplankton spring blooms annually trigger a complex chain of interactions throughout the marine ecosystem.The life cycles and migration patterns of many species are adapted to this process due to the improved feeding opportunities.Any changes to the timing of these blooms can have severe consequences for ecosystem productivity.This study shows that the NL climate explains relatively well the timing of the bloom and the abundance of a key zooplankton species over interannual cycles.The spring onset of stratification is also proposed as a mechanism triggering the bloom, but exhibited no decadal-scale trend resulting, for example, from anthropogenic climate change.The established relationship among a simple climate index, the timing of the bloom, and the secondary production can help predict changes in the base of the marine food web, aiding the development of ecosysteminformed models for higher trophic levels.

Fig. 1 .
Fig. 1.Map of the study region in the Northwest Atlantic Ocean.The

Fig. 2 .
Fig. 2. From the NL climate to the timing of the bloom and the abundance of zooplankton between 1998 and 2020.(a) Comparison between the NLCI (Cyr and Galbraith 2021) in blue and the standardized anomaly of the timing of the bloom peak intensity averaged over 2J, 3K, and 3LNO in green.(b) Comparison between the NLCI (blue) and the abundance of Calanus finmarchicus copepods averaged over the NL shelf (in orange; starts in 1999).(c) Comparison between the standardized anomaly of the timing of the bloom peak from (a) (green) and the abundance of C. finmarchicus copepods from (b) (orange).The Pearson correlation coefficients between the curves in each panel (r ) are shown.

Fig. 3 .
Fig. 3. Climatological (1998-2020) seasonal evolution of the stratification at Sta. 27 for each DOY.(a) The dark blue dots are the daily climatological values of the stratification (average of all individual observation made on a specific date during the period 1998-2020; N = 283 d represented), while the orange curve is a 30-d moving average through these points.The dark-gray dots are the daily climatological values of the MLD (N = 253), while the black curve is a 30-d moving average smoothing within these points.(b) The red and blue curves are, respectively, the temperature and salinity contributions to the stratification (orange curve; same as in a).The vertical green and red bands in both panels represent respectively the climatological (1998-2020) timing of the bloom initiation and maximum intensity (AE 1 SD) in 3LNO derived from Modis and SeaWiFS sensors.

Fig. 4 .
Fig. 4. Time series of the standardized anomalies of the initiation of the spring bloom in NAFO divisions 3LNO from Modis and SeaWiFS sensors ( f t bi , orange curve) and the timing of the minimum stratification at Sta. 27 ( e t s , blue curve) determined in "Timing of the phytoplankton bloom" section.The Pearson correlation coefficient (r) is indicated on the figure.

Table 1 .
Mean timing (AE 1 SD) of the spring bloom initiation and maximum Chl a intensity for differentNAFO divisions over  1998NAFO divisions over   -2004.   .