Journal of Geophysical Research: Oceans

Modeling of the mesoscale structure of coupled upwelling/downwelling events and the related input of nutrients to the upper mixed layer in the Gulf of Finland, Baltic Sea

Authors


Abstract

[1] A high resolution numerical study is undertaken to simulate an upwelling event along the northern coast of the Gulf of Finland, 21–29 July 1999, which was documented well by in situ and remote measurements. The simulated sequence of SST maps shows a reasonably good resemblance to that of satellite infrared imagery, including both mesoscale coherent structures (filaments or squirts) and the whole process of post-upwelling relaxation of the temperature field. Upwelling along the northern coast of the Gulf is accompanied by downwelling along the southern coast so that two longshore baroclinic jets and related fronts are developed simultaneously. When the strong westerly winds producing the upwelling/downwelling weaken, the longshore jets become unstable and produce transverse jets, cold/warm water squirts. Using pseudo-random simulated fields of temperature and velocity of currents, the apparent lateral diffusivity due to squirts is directly estimated at 500 m2 s−1. The model is also applied to estimate nutrient transport. Simulated phosphate concentration in the surface layer at the cold side of upwelling front is found to be about 0.3 mmol m−3 which is consistent with observations. The total content of phosphorus and nitrogen in the upper 10 m layer of the Gulf introduced by the upwelling event is estimated to be 387 and 36 tons, respectively. It follows, that the upwelling event transports nutrients into the upper layer with clear excess of phosphorus (N:P = 36:387 = 0.093) compared to the Redfield ratio of 7.2. Therefore phosphorus input caused by upwelling during summer most likely promotes nitrogen-fixing cyanobacteria blooms.

1. Introduction

[2] Owing to the prevailing south-westerly (along-gulf) winds [e.g., Mietus, 1998; Soomere and Keevallik, 2003] the northern Gulf of Finland is an active upwelling area in summer in the Baltic Sea. This has been shown by both satellite sea surface temperature (SST) images [e.g., Kahru et al., 1995] and model simulations [e.g., Myrberg and Andrejev, 2003]. Closely spaced CTD profiling and nutrient sampling, and SST images presented by Vahtera et al. [2005] displayed the existence of intensive squirts in the Gulf formed during and immediately after an upwelling event that was observed along the northern coast of the Gulf in the period of 21–29 July 1999. Several squirts, or cold-water jets crossing the upwelling front, reaching almost the entire width of the Gulf, were visible in SST images. Such squirts might have a large contribution to lateral mixing in the Gulf of Finland. Model simulations by Zhurbas et al. [2004] demonstrated that mesoscale disturbances (meanders, filaments, and eddies) of the longshore upwelling jet in the southeast Baltic Sea were attributed to the topographic irregularities.

[3] Coastal upwelling typically brings cold and nutrient-rich deep water to the surface layer. Late-summer blooms of filamentous cyanobacteria possessing the capability to fix dissolved atmospheric nitrogen (diazotrophy) are common in the eutrophic and brackish Baltic Sea [e.g., Kahru et al., 1994; Finni et al., 2001]. The cyanobacteria blooms occur at the most nutrient-deplete time of the year, from the end of June to the end of August and the bloom biomass typically accumulates in the surface layer during calm periods. Primary productivity in the Baltic Sea is generally considered nitrogen limited during summer [e.g., Granéli et al., 1990; Kivi et al., 1993; Tamminen and Anderssen, 2007], whereas the filamentous diazotrophic cyanobacteria are either phosphorus or trace element limited during the blooms [e.g., Stal et al., 1999; Moisander et al., 2003; Kangro et al., 2007]. Phosphate input through upwelling events has been suggested as a phosphorus source for bloom formation and sustenance in the Baltic Sea [e.g., Larsson et al., 2001; Lignell et al., 2003].

[4] According to field measurements seasonal nutriclines in the western Gulf of Finland lie in the thermocline, the phosphacline being shallower than the nitracline [Laanemets et al., 2004]. An important consequence from the vertical separation of the phosphacline and the nitracline in the Gulf of Finland is that upwellings during summer bring mainly phosphate into the surface layer. Thus upwelling events during nutrient deplete and thermally stratified summer conditions may promote blooms of nitrogen-fixing cyanobacteria [e.g., Kononen et al., 2003; Vahtera et al., 2005].

[5] The objectives of this study are

[6] (1) to simulate mesoscale variability of temperature, velocity, and nutrient fields in the course of and following the upwelling event of 21–29 July 1999 in the Gulf of Finland and compare it with available remote and in situ observations;

[7] (2) using simulation results, to study the role of squirts (i.e., transverse jets) in the relaxation of coupled upwelling/downwelling processes in the Gulf and estimate corresponding lateral eddy diffusivity;

[8] (3) using simulation results, to estimate total input of nutrients (phosphorus and nitrogen) into the upper mixed layer within the Gulf caused by the upwelling event.

2. Model Setup

[9] We apply the Princeton Ocean Model (POM) [Blumberg and Mellor, 1983]. The POM is a primitive equation, sigma co-ordinate, free surface, hydrostatic model with a 2.5 moment turbulent closure sub-model embedded. The model domain includes the whole Baltic Sea closed at the Danish Straits; digital topography is taken from [Seifert and Kayser, 1995]. A model grid with variable grid spacing is applied and it is refined to 0.5 nautical miles in the Gulf (see Figure 1); there are 20 sigma layers in the vertical direction. The model resolves the mesoscale variability because the grid size falls below the first baroclinic deformation radius (about 2–5 km in the region under investigation [e.g., Fennel et al., 1991; Alenius et al., 2003]).

Figure 1.

Map of the Baltic Sea (right) and a close-up of the Gulf of Finland, where the model grid is refined to 0.5 nautical miles (left). Shown are the locations of Kalbådagrund (▴) and Utö (▾) weather stations and R/V ARANDA study area (grey rectangle).

[10] Atmospheric forcing (wind stress and heat flux components) for a 20–day period starting on 20 July 1999 was taken from a gridded meteorological data set established and maintained at SMHI (the space and time resolution is 1° and 3 h, respectively). To calculate winds at 10 m level from the SMHI geostrophic wind vectors, the latter were turned counter clockwise at 15° and multiplied by a factor 0.6. Then components of the10 m level wind along with other meteorological parameters from the SMHI data set were interpolated to the model grid. Before introduction to the model, wind data from the SMHI gridded meteorological data set were compared with wind measurements at the Kalbådagrund (25°36′E, 59°59′N), R/V ARANDA (study area of 22°50′-24°00′E, 59°30′-59°48′N), and Utö (21°22′E, 59°47′N) weather stations (see Figures 1 and 2) . It is clearly seen from Figure 2 that the wind stress calculated from measurements at R/V ARANDA and Utö weather stations fit well with that of the Kalbådagrund weather station. On the other hand, the x-component of wind stress calculated from the SMHI gridded meteorological data set is below that of the wind measurements at the weather stations by approximately a factor of 2. In such circumstances we rely on the direct wind measurements and multiply wind stress values calculated from the SMHI data set by a factor of 2.04 to fit cumulative wind stress for the upwelling period of 21–29 July 1999 estimated directly from the weather station data.

Figure 2.

Time series of x (solid) and y (dotted) components of wind stress for the period 20 July–9 August 1999 calculated from measurements at R/V ARANDA (thin lines on top panel) and the Kalbådagrund (thin lines on bottom panel) weather stations, and from interpolations of the SMHI gridded meteorological data set (bold lines). Gaps in the R/V ARANDA time series were filled with wind measurements at the Utö weather station.

[11] Initial thermohaline fields were constructed with the help of the Data Assimilation System (DAS) coupled with the Baltic Environmental Database established and maintained by Alexander Sokolov and Fredrik Wulff at the Stockholm University (see http://nest.su.se/das). Since interpolation of DAS data on 20 July yields approximately an upper mixed layer temperature of 16°C in the Gulf of Finland versus roughly 19°C measured aboard R/V ARANDA on 20–21 July 1999 [Vahtera et al., 2005], the initial temperature field constructed using DAS was increased by 3°C in the upper 10 m layer of the whole Baltic Sea.

[12] Two equations describing passive tracer balance were added to the POM. Those were used to simulate nutrient transport. Nutrients can be considered as a conservative passive tracer if one is interested to estimate the transport of nutrients from deep layers to the surface layer; posterior behavior of nutrients is however not conservative (e.g., due to rapid phytoplankton uptake). The equations describing nutrient balance were solved numerically within the POM code in the same way as those of temperature and salinity.

[13] To construct initial fields of nutrients we use data of phosphate and nitrate sampling performed aboard R/V ARANDA on 20 July 1999 (Figure 3). The maximum sampling depth was 60 m, and we extended the profiles to greater depths using data from previous cruises (actually behavior of nutrients below 60 m is of minor importance for the present study because so deep waters hardly reach the surface layer in the course of upwelling events in the Gulf; see results of simulation in the next section). Vertical profiles of nutrients presented in Figure 3 were uniformly extended to the whole Baltic Sea.

Figure 3.

Vertical profiles of (right) phosphate and nitrate concentration measured on 20 July 1999 at (23°40′E, 59°32′N) aboard R/V ARANDA and (left) temperature and salinity obtained by interpolation of DAS data to 20 July and (23°20′E, 59°25′N).

[14] It is known that circulation in the Baltic Sea is subject to drastic changes depending on wind conditions, and the adjustment of circulation pattern to the wind-forcing change occurs within a day or so [e.g., Krauss and Brügge, 1991]. For this reason, the Baltic Sea, in contrast to the ocean, is practically free of quasi-permanent mean baroclinic currents which might be reproduced by diagnostic calculations using seasonally averaged initial thermohaline fields. That is why we, following Krauss and Brügge [1991], took the liberty to skip spin-up of the model and start prognostic runs from the motionless state and zero surface elevation at 00:00 a.m. of 20 July 1999. Here we are going to study mesoscale variability which, being mostly associated with baroclinic instability, is unlikely sensible to the inaccuracy in initial conditions. The last statement can be checked a posteriori by the comparison of simulation results with observations.

3. Results of Simulation and Discussion

[15] Figure 4 presents the comparison of SST maps of the Gulf obtained from satellite imagery and simulated by the model. On 26 July within a day after culmination of the westerly winds, the area covered with cold upwelled water has about identical values in the satellite image and simulation plot. (Note that along the shallow northern coast of the Gulf the satellite-born SST is distorted by bottom closeness, which produces a masking effect underestimating the cold-water area.) Moreover, a good resemblance between observed and simulated coherent mesoscale structures including several cold water filaments/squirts whose tips display a tendency to cyclonic rotation can be observed. This corresponds to results of Zhurbas et al. [2006] who showed that instability of a longshore baroclinic jet associated with upwelling (downwelling) results in selective formation of mostly cyclonic (anticyclonic) eddies. By 3 August both satellite-born and simulated SST maps display equal relaxation of the upwelling process, resulting in substantial decrease of the cold water area and an increase in temperature of the upwelled water.

Figure 4.

Comparison of SST maps of the Gulf of Finland obtained from satellite imagery (Remote Sensing Laboratory, Stockholm University, the top panels) and simulated by the model at “true” wind stress field (mid panels) and that of scaled by a factor 0.5 (bottom panels).

[16] To illustrate the sensitivity of simulated SST maps to wind-forcing change, Figure 4, bottom panels, presents results of modeling when the “true” wind stress field was scaled by a factor 0.5. Such reduction of the wind-forcing results in a considerable shrinkage of the cold water area and loss of resemblance to the observed SST maps.

[17] The development of transverse jets/squirts due to instability of longshore baroclinic upwelling/downwelling jets is illustrated in Figure 5 where simulated maps of SST and current velocity at 1 m below the sea surface are shown in the central Gulf of Finland. Since the velocity field was strongly influenced by inertial oscillations, which masked “pure” squirt motions, the former were filtered out using a simple procedure:

equation image

where v(x, t) and equation image(x, t) are the velocity vectors before and after filtration, respectively, Ti = 0.58 day is the inertia oscillation period.

Figure 5.

Simulated sea surface currents and SST maps in the central Gulf of Finland displaying the development of transverse jets/squirts due to instability of longshore baroclinic upwelling/downwelling jets. The three white segments depict the position of transects used to estimate lateral eddy diffusivity.

[18] At t = 8.97 days (i.e., 3 days after the culmination of the westerly winds, but still within the upwelling wind period) one can observe several cold water squirts extending throughout entire Gulf's width with southward velocities up to 0.5 m s−1; a warm water squirt with northward velocity can be identified as well. The dominance of southward transport at t = 8.97 days is explained by the persistence of westerly winds resulting in southward Ekman transport. At t = 14.47 days the westerly winds have ceased, and the southward momentum transport produced by cold squirts is more or less balanced by the northward momentum transport by warm squirts. It is also seen clearly in the bottom panel of Figure 5 that instability of the downwelling baroclinic jet results in the formation of mostly anticyclonic eddies along the south shore of the Gulf, in accordance with Zhurbas et al. [2006].

[19] There is no doubt that cold and warm water squirts running back and forth across the Gulf contribute to lateral mixing. To perform direct estimates of lateral eddy diffusivity caused by squirts using simulated pseudo-random fields of temperature and velocity, three along-gulf transects were chosen (see Figure 5, the bottom panel). The along-gulf distance series of temperature T(xi) and the transverse component of current velocity at level z = 1 m, v(xi) were extracted from the simulation data. Note that now the x and y axes are directed along and across the Gulf, respectively (see Figure 5, the top panel). Then, by means of high-pass filtering series of temperature and transverse velocity fluctuations T′(xi) and v′(xi) were calculated; the value of filter half-width was varied within the range of 40–90 km. Finally, the lateral eddy heat diffusivity was estimated from the “direct” formula

equation image

where 〈…〉 means spatial averaging over the three transects.

[20] Time series of the lateral eddy diffusivity calculated at different values of the filter half-width are given in Figure 6, the top panel. Relatively high levels of fluctuations of KH in the plot are explained by poor statistics of mesoscale disturbances available: in accordance with Figures 45, one can observe no more than 4–5 squirts at once in the Gulf; this can be cured to some extent by time averaging (smoothing) of the KH(t) curve. During the first five days of the upwelling period (21–25 July) estimates of lateral eddy diffusivity are negligible varying around zero within a range of 30 m2 s−1, which corresponds to a relatively stable development of longshore baroclinic upwelling/downwelling jets. Then, KH displays a sudden increase to approximately 500 m2 s−1 within a couple of days. This corresponds to swelling of instability of the longshore baroclinic jets in the form of cold and warm squirts/filaments which eventually can be transformed to mushroom structures and eddies. The value of KH remains high even after the westerly winds cease, i.e., during the relaxation period of coupled upwelling/downwelling. Therefore instability of longshore baroclinic jets may be considered as the major process contributing to relaxation of upwelling/downwelling.

Figure 6.

Time series of the lateral eddy diffusivity in the surface layer of the central Gulf of Finland calculated from simulated pseudo-random fields of velocity of currents and temperature at different values of (top) filter half-width and (bottom) wind stress scaling factor.

[21] It seems worthwhile discussing how sensitive our estimates of KH are to the wind-forcing change. One may expect that the impact is minor since much of mesoscale variability is associated with the baroclinic instability of the jets. The point is that the main characteristics of the longshore baroclinic jets associated with fully developed upwelling/downwelling no longer depend on the wind-forcing: the jet's velocity and width are controlled by the densimetric velocity c = RIf and the internal (baroclinic) Rossby radius RI, respectively; f is the Coriolis parameter [Csanady, 1977; Zhurbas et al., 2006]. Time series of the lateral eddy diffusivity calculated at different values of wind stress scaling factor, s, and a constant value of the filter half-width (60 km) are given in Figure 6, the bottom panel. At weak winds (s = 0.25) KH remains below 100 m2 s−1 during whole 20-day period. At stronger winds (s ≥ 0.5) in the post-upweling period (t > 10 days) the estimates of KH are fluctuating somewhere around 500 m2 s−1 displaying very weak tendency to grow with the wind stress scaling factor. At t > 16 days and s > 1 the KH series display sudden growth of fluctuation amplitude likely due to further pauperization of the squirts statistics. One may suggest that the transition to fully developed upwelling regime takes place at a value of the scaling factor within the range of 0.25 < s < 0.5. Also it is worth noting that the rate of KH growth before achieving the “saturated” value of about 500 m2 s−1 increases with s. This corresponds to the finding by Zhurbas et al. [2006] who showed that the instability growth rate of a longshore baroclinic jet related to fully developed upwelling/downwelling increases with wind-forcing. The last effect was physically explained by the increase of jet detachment from the shoreline with wind-forcing, so that the stabilizing effect of the lateral wall is weakened.

[22] Maps of phosphate and nitrate concentration in the sea surface layer are given in Figure 7. Cold squirts identified in velocity and temperature fields in Figure 5 can be easily seen in the phosphate maps as well (Figure 7, left panels). The maximum of simulated phosphate concentration in the surface layer at the cold side of the upwelling front is about 0.3 mmol m−3 which is consistent with observations made by Vahtera et al. [2005].

Figure 7.

Simulated maps of (left) phosphate and (right) nitrate concentration in the sea surface layer.

[23] Since the nitracline is situated deeper than the phosphacline (see Figure 3), nitrate is introduced to the surface later than phosphate in the course of an upwelling event and, therefore, the sea surface area occupied by nitrate is smaller than that of phosphate (cf., the right and left panels in Figure 7). However, the simulated nitrate concentration maximum in the surface layer is about 1 mmol m−3, which is larger by a factor of 3 compared to that of phosphate. Before the upwelling event, nitrate concentrations above 1 mmol m−3 were found only at depths greater than 40 m (Figure 3), this would suggest that upwelled water originated from depths below 40 m.

[24] Distribution of nitrate concentration along the northern coast of the Gulf formed due to upwelling is found to display well-pronounced patchiness with concentration maxima related to some features in shoreline/bottom topography where upwelling is most intensive. One of such upwelling centers is situated at the tip of the Hanko peninsula. Note that the excess of nitrate concentration at the Hanko upwelling center has disappeared within several no-wind days following the upwelling event due to sinking of dense, nitrate-rich water (cf., Figure 7, right panels). In order to elucidate potential phosphorus sources of late summer blooms of filamentous cyanobacteria in the Gulf of Finland it is important to estimate the total input of phosphate and nitrate to the sea surface layer due to upwelling events (see Introduction). To arrive at such estimates, we have used results of our simulation to calculate total content of phosphate and nitrate in the upper 10 m layer in the area restricted in latitude and longitude by inequalities ϕ ≥ 23° E, 59° ≤ λ ≤ 60.7°N. Time series of total content of PO4-P and NO3-N in the upper 10 layer of the Gulf are shown in Figure 8. Being vanishingly small at the beginning, the surface layer content of PO4-P and NO3-N achieve the maximum values of 536 and 135 tons, respectively, by the time upwelling promoting winds die down at t = 10 days. The nitrate to phosphate ratio reaches a very low value of 135:536 = 0.25. However, during the relaxation period the surface layer content of nutrients decrease due to sinking of the upwelled cold and dense water, and the relative decrease of nitrate-nitrogen (from 135 to 36 tons by the end of the simulation period, i.e., a factor of 3.75) is found to be much larger than that of phosphorus (from 536 to 387 tons, i.e., a factor of 1.42). The preferable out-flow of nitrogen from the surface layer is explained by the fact that it came to the surface later than phosphorus, and, therefore, sank earlier because the undisturbed nitracline is situated at deeper depths than the phosphacline (see Figure 3). Again, the sea surface area occupied by upwelled nitrate was found to be smaller than that of phosphate (see Figure 7), which facilitates sinking of the former during relaxation of upwelling. As a result, one can conclude that the upwelling event transported nutrients into the upper layer with clear excess of phosphorus (N:P = 36:387 = 0.093) compared to the Redfield ratio of 7.2, displayed by rapidly growing algae (review by Sterner and Elser [2004]). The clear excess of phosphorus in upwelling water suggests that a large part of the upwelled phosphate might be utilized by the diazotrophic cyanobacteria due to nitrogen limitation of other phytoplankton groups. Annual total waterborne loads of phosphorus to the Gulf of Finland have been estimated to amount up to 6030 tons [HELCOM, 2004]. A monthly mean external input would thus be approximately 500 tons. The estimate of waterborne phosphorus by HELCOM [2004] includes both phosphate phosphorus and organic phosphorus compounds, both in approximately equal shares. Nausch and Nausch [2006] estimated 8–65 % of organic phosphorus compounds to be bioavailable in the Baltic Proper pelagic areas; if we apply an approximation that roughly half of the organic phosphorus compounds would be bioavailable the monthly external phosphorus load to the Gulf of Finland would be approximately 380 tons, roughly equaling the input by one upwelling event. Since external loads are to a large extent correlated with river runoff [e.g., Grimvall and Stålnacke, 2001] it can be assumed that this number is an overestimation since river runoff is small during late summer [Richter and Ebel, 2006]. The role of upwelling as a nutrient source to the nutrient depleted surface layer during summer is substantial and promotes occurrence of diazotrophic cyanobacteria through clear phosphorus excess.

Figure 8.

Time series of simulated total content of PO4-P (solid) and NO3-N (dotted) introduced to the upper 10 m layer of the Gulf of Finland due to upwelling event of 21–29 July 1999.

4. Conclusions

[25] (1) Mesoscale coherent structures (filaments or squirts) and the whole process of relaxation of temperature field observed in the Gulf after upwelling event were reasonably well reproduced by the model;

[26] (2) Numerical simulations showed that the relaxation of longshore baroclinic jets and related thermohaline fronts caused by coupled upwelling and downwelling events in the Gulf occurs in the form of cold and warm water squirts running back and forth across the Gulf and thereby contributing to lateral mixing. Using simulated pseudo-random fields of temperature and current velocity, the lateral eddy diffusivity in the surface layer due to squirts was directly estimated at 5 · 102 m2 s−1.

[27] (3) Estimated phosphate concentrations in the surface layer at the cold side of the upwelling front were about 0.3 mmol m−3, which are consistent with observations;

[28] (4) The total content of phosphorus and nitrogen in the upper 10 m layer of the Gulf of Finland introduced by the upwelling event was estimated to be 387 and 36 tons, respectively. It follows, that the upwelling event transported nutrients into the upper layer with clear excess of phosphorus (N:P = 38:387 = 0.093) compared to the Redfield ratio of 7.2. Thus phosphorus input caused by upwelling during summer most likely promotes diazotrophic cyanobacteria blooms.

Acknowledgments

[29] We are thankful to James Richman and anonymous reviewers for constructive recommendations. The work was sponsored by EU Structural Funds (ESF measure 1.1) and the Russian Foundation for Basic Research (grant 06-05-64412). Finnish Meteorological Institute has kindly provided weather station's wind data. Special thanks to Oleg Andrejev for provision of meteorological data.

Ancillary