The influence of increasing water turbidity on the sea surface temperature in the Baltic Sea: A model sensitivity study

Journal of Marine Systems 88 (2011) 323–331

Contents lists available at ScienceDirect

Journal of Marine Systems j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / j m a r s y s

The influence of increasing water turbidity on the sea surface temperature in the Baltic Sea: A model sensitivity study Ulrike Löptien ⁎, H.E. Markus Meier Swedish Meteorological and Hydrological Institute (SMHI), Folkborgsvägen 1, S-601 76 Norrköping, Sweden

a r t i c l e

i n f o

Article history: Received 18 January 2011 Received in revised form 1 June 2011 Accepted 7 June 2011 Available online 17 June 2011 Keywords: Baltic Sea Ocean modeling Water turbidity SST trends Secchi depth

a b s t r a c t The aim of the present study is to investigate the influence of enhanced absorption of sunlight at the sea surface due to increasing water turbidity and its effect on the sea surface temperatures (SST) in the Baltic Sea. The major question behind our investigations is, whether this effect needs to be included in Baltic Sea circulation models or can be neglected. Our investigations cover both, mean state and SST trends during the recent decades. To quantify the impact of water turbidity on the mean state different sensitivity ocean hindcast experiments are performed. The state-of-the art ocean model RCO (Rossby Centre Ocean model) is used to simulate the period from 1962 to 2007. In the first simulation, a spatially and temporally constant value for the attenuation depth is used, while in the second experiment a climatological monthly mean, spatially varying attenuation coefficient is derived from satellite observations of the diffuse attenuation coefficient at 490 nm. The inclusion of a spatially varying light attenuation leads to significant SST changes during summer. Maximum values of + 0.5 K are reached in the Gulf of Finland and close to the eastern coasts, when compared to a fixed attenuation of visible light of 0.2 m − 1. The temperature anomalies basically match the pattern of increased light attenuation with strongest effects in shallow waters. Secondary effects due to changes in the current system are of minor importance. Similar results are obtained when considering trends. In the absence of long-term basin wide observations of attenuation coefficients, some idealizations have to be applied when investigating the possible influence of long-term changes in water turbidity on the SST. Two additional sensitivity experiments are based on a combination of long-term Secchi depth station observations and the present day pattern of water turbidity, as observed by satellite. We show the potential of increased water turbidity to affect the summer SST trends in the Baltic Sea significantly, while the estimated effect is apparently too small to explain the overall extreme summer trends observed in the Baltic Sea. © 2011 Elsevier B.V. All rights reserved.

1. Introduction The Baltic Sea is located in central Europe and about 85 million people live in its drainage basin (e.g. Leppäranta and Myrberg, 2009). The shallow, brackish sea is of crucial importance for the bordering countries, e.g. concerning fishery and tourism. The Baltic Sea changed considerably during the recent decades, which is partly due to human influence. Several studies have indicated that the temperature of the Baltic Sea has risen substantially during the last 100 years while the reasons for this warming are still unclear (e.g. Belkin, 2009; Bradtk et al., 2010; Leppäranta and Myrberg, 2009; MacKenzie and Schiedeck, 2007; Siegel et al., 2006). Belkin (2009) found warming trends during 1982–2006 in the Baltic and North Sea that exceeded the surrounding seas by 0.5–0.6 K. Furthermore, the analysis of daily station data from long-term monitoring programs since the 1860s ⁎ Corresponding author now at: Leibniz Institute of Marine Sciences (IFM-GEOMAR), Düsternbrooker Weg 20, D-24105 Kiel, Germany. Tel.: + 49 431 600 4283; fax: + 49 431 600 4469. E-mail address: [email protected] (U. Löptien). 0924-7963/$ – see front matter © 2011 Elsevier B.V. All rights reserved. doi:10.1016/j.jmarsys.2011.06.001

reported an unprecedented warm period during the mid 1990s (MacKenzie and Schiedeck, 2007). From 1985 to the early 2000s the temperature rose on average 0.6 K with a significantly higher increase in summer (N1.4 K). Also, Siegel et al. (2006) report strong positive trends during the summer month for the period 1990–2004 with a maximum slope in the Bothnian Sea. According to their findings, the Gotland Sea warmed by 0.15–0.18 K per year (July–September) which is close to the values reported by MacKenzie and Schiedeck (2007) for the period 1985 to the early 2000s. This warming can partly be explained by the atmospheric conditions. However, in particular, the summer warming is higher than what could be expected on the basis of the consensus view of global increase of air temperatures. Due to the strong warming large ecological consequences might occur (MacKenzie and Schiedeck, 2007). Besides, one of the most serious environmental problems in the Baltic Sea is eutrophication (e.g. Boesch et al., 2006; Boesch et al., 2008; Jansson and Dahlberg, 1999; Nehring, 1992; Pawlak et al., 2009; Wasmund and Uhlig, 2003; Wulff et al., 1999). This is caused by increased nutrient input from the intensively cultivated catchment area, resulting in an increase in biomass and a reduction in water

324

U. Löptien, H.E.M. Meier / Journal of Marine Systems 88 (2011) 323–331

transparency (Sanden and Hå̊kansson, 1996). The decrease in water transparency is mainly illustrated by long-term Secchi depth measurements, which is the depth at which a white disk disappears from the observers' view as it is lowered into the water. Secchi depth measurements exist since the beginning of the last century at single locations (Aarup, 2002; Sanden and Hå̊kansson, 1996) and apparently Secchi depth experienced a remarkable decrease during the last 100 years (Leppäranta and Myrberg, 2009; Sanden and Hå̊kansson, 1996). While the present level is 5–10 m, some measurements indicate that it has almost halved at numerous locations since the beginning of the last century (Laamanen et al., 2004). Empirical studies found a good correlation between water transparency and the state of eutrophication (Elmgren and Larson, 2001; Nielsen et al., 2002; Sand-Jensen and Borum, 1991). Note, that the Baltic Sea is optically a multi-componental water (coastal water, Type 1–5) and the decrease in Secchi depth might possibly be influenced by changes in yellow substances or suspended organic matter, besides chlorophyll (Kratzer and Tett, 2009; Siegel et al., 2005). However, regardless of the reasons for the observed increase in water turbidity, it can be assumed to feed back on the sea surface temperature (SST). Kahru et al. (1993) report a substantial feedback of cyanobacterial blooms and their accumulation on SST. Satellite data indicate a local SST increase by up to 1.5 K due to cyanobacterial blooms. Also, it is well known that optically active substances, such as chlorophyll and yellow substances, strongly absorb light in the blue and red parts of the spectrum. Through this, the vertical distribution of radiative heating can be modified which results in enhanced surface warming and sub-surface cooling (Chang and Dickey, 2004; Oschlies, 2004). State-of-the-art Baltic Sea models neglect the temporal and spatial variability in the attenuation of light and assume for simplicity a fixed, mean attenuation (e.g. Meier and Faxen, 2002; Rudolph and Lehmann, 2006; Schrum et al., 2003; Wilhelmsson, 2002). Thus, the effect of changing water transparency is not included. The present study is a first attempt to introduce a more complex attenuation of light into a Baltic Sea climate model, investigate the effects and quantify the importance. To test the impact of spatial variability of light attenuation, we focus on the visible part, since water itself is highly absorbing in the near infrared spectrum. Our new parametrization of visible light (KDv) is based on the mean seasonal cycle of measurements of the diffuse attenuation coefficient at 490 nm (KD(490)), which is measured directly by satellites and closely related to Secchi depth. The conversion to KDv follows the results of Pierson et al. (2008) and is especially designed for the Baltic Sea. To estimate the potential of decreased Secchi depth to affect trends in SST, we perform two idealized experiments which are based on station data of Secchi depth. In the absence of basin wide measurements, the station data are combined with a spatial pattern of water turbidity which matches the present day satellite observations. Additionally, we calculate anomalous net heating rates in dependence of KDv theoretically and compare the results with our model. The paper is organized as follows. The calculation of net heating rates, a description of the three dimensional model, the experimental design and the source of observational data for model validation are described in Section 2. Section 3 addresses the results and is subdivided into three major subsections, comprising of the net heating rates, the mean seasonal cycle and trends. A comprehensive discussion follows in Section 4. 2. Methods 2.1. Net heating rates To quantify the uncertainties concerning the underlying parameters and the possible effects of varying attenuation coefficients, we consider as a first step the dependency of the anomalous net heating rates on the attenuation of visible light. Net heating rates estimate temperature changes of a water parcel due to radiation neglecting all

other physical processes (e.g. diffusion and advection). The calculation follows the formulation given by Sweeney et al. (2005). We consider a profile of downward radiative heat flux, which is represented as Iðx; y; zÞ = I0 ðx; yÞJ ðzÞ;

ð1Þ

where I0 is the downwelling shortwave radiative flux just below the sea surface (W/m 2) and J(z) is a dimensionless attenuation function exponentially decaying with depth, depending on an attenuation coefficient specifying the profile. J(z) is generally wavelength dependent. While sea water itself is highly absorbing in the near infrared spectrum, visible light penetrates much deeper and can be largely influenced by optically active substances. Thus the focus is, as in the three dimensional experiments, on the visible part of sunlight and an attenuation coefficient of 0.2 m − 1 is used for KDv as reference. This value is compared to different parameter settings for KDv, ranging from 0.15 to 0.36 m − 1. The corresponding shortwave heating affects the local temperature T in a Boussinesq fluid according to     ∂T z = − 1 = ρCp ∂ = ∂z ρCp F −I : ∂t

ð2Þ

Here, F z accounts for vertical processes such as advection and diffusion, ρ denotes the density and is set to a mean value. Cp denotes the heat capacity of sea water. From this equation it is possible to separate the net heating rate due to shortwave radiation of a water column between the depth levels Z1 and Z2 by,     Z2 ∂I 1 = ρCp ∫Z1 dz = 1 = ρCp ðIð Z2Þ−IðZ1ÞÞ: ∂z

ð3Þ

2.2. The ocean model The three-dimensional (3D) ice–ocean model used for all three dimensional sensitivity experiments is the Rossby Centre Regional Ocean model (RCO). RCO was used previously for various ocean and climate studies (e.g. Meier, 2005; Meier, 2006; Meier and Kauker, 2003b) and is described in more detail in Meier et al. (2003) and Meier and Kauker, (2003a). The ocean model is a regionalized version of the Ocean Circulation Climate Advanced Model (OCCAM) (Webb et al., 1997) implemented for the Baltic Sea. It is a Bryan–Cox–Semtner primitive equation model with a free surface coupled to a Hibler-type sea ice model. A prognostic two-equation turbulence closure scheme is embedded. The downward radiative flux in RCO is parametrized by two exponential functions describing the near infrared and visible part of the spectrum. Since the parametrizations vary between the sensitivity experiments, they are explained in more detail in Section 2.3. The model domain covers the Baltic Sea including Kattegat. RCO in this setup has a tendency to produce structured noise and thus different advection schemes were tested. Following Meier et al. (2003) in our study the third-order advection scheme (splitquick) by Webb et al. (1998) is used. This scheme results in comparatively smooth horizontal tracer distributions allowing the calculation of SST trends at a relatively low noise level. The horizontal resolution used here is 2 nautical miles and 41 levels in the vertical. Layer thicknesses ranging from 3 m close to the surface to 12 m near the bottom are used. The model depths are based on a bottom topography taken from Seifert and Kayser (1995). We use open boundaries in the northern Kattegat. In case of inflow, temperature and salinity values at the boundaries were nudged in all simulations towards observed climatological profiles in the southern Skagerrak. In case of outflow, a modified Orlanski radiation condition is utilized (Stevens, 1990). Such a simplified treatment of the open boundary in Kattegat is sufficient since the Kattegat deep water salinity of about 33.2 psu does not change significantly (Meier and Kauker, 2003b). Sea

U. Löptien, H.E.M. Meier / Journal of Marine Systems 88 (2011) 323–331

325

level elevation at the boundaries is prescribed from daily tide gage data. River runoff data are taken from Bergström and Carlsson (1994) and updated with results from a large-scale hydrological model (Graham, 2004). The crossover from observations to the hydrological model causes some change in variability for some rivers while the mean values remain basically unchanged. Air temperature, wind, cloud cover, sea level pressure, humidity and precipitation are required as surface boundary conditions. Surface restoring or data assimilation is not applied. Since both, common reanalysis data sets as well as global climate models have their limitations in regional detail, we use regional climate models with a limited model domain and higher resolution to downscale coarsely resolved data sets. Here, we use the Rossby Centre Regional Atmosphere model (RCA) at 25 km horizontal resolution (Jones et al., 2004; Samuelsson et al., 2011). The boundary conditions for the atmospheric model are taken from ERA40 reanalysis data (Uppala et al., 2005) updated with operational ECMWF data. This data set allows for hindcast simulations covering the period 1962–2007. The atmospheric model RCA has been used successfully as boundary condition for the ocean model RCO previously (Döscher et al., 2002). The ability of RCO to perform realistically for present day climate in the Baltic Sea has been demonstrated e.g. in Meier and Kauker (2003a). 2.3. Experiments The above configuration of RCO is used to perform a suite of sensitivity experiments that differ by the formulation of light attenuation only. The downward radiative flux of RCO is parametrized by two exponential functions describing the near infrared and visible parts according to Theakaekara (1974). Since water itself is highly absorbing in the near infrared spectrum, the absorption in the near infrared part will be fixed in the following experiments (attenuation coefficient = 0.95) while the attenuation of visible light (KDv) will differ. Based on different parametrizations, we performed four sensitivity ocean hind-cast experiments (1962–2007). All model experiments which are discussed below are listed in Table 1. In the first simulation (the reference run), we use a spatially and temporally constant value of 0.2 m − 1 for KDv, which relates to an observed mean Secchi depth of 8.5 m (Sanden and Hå̊kansson, 1996). The conversion between attenuation and Secchi depth follows the suggestions by Kratzer et al. (2003), who relate KDv to Secchi depth by multiplying the reciprocal attenuation coefficient by a factor of 1.7. This estimate is particularly designed for the Baltic Sea. In the second experiment, a climatological, monthly varying value for KDv is used (experiment CLIM). The measurements are based on satellite observations of the diffuse attenuation coefficients at 490 nm (KD(490)) measured by MODIS (Moderate-resolution Imaging Spectroradiometer) (2002–2007) (Werdell and Bailey, 2005). These values are transformed into KDv following the results of Pierson et al. (2008). Winter values form November to February were too sparse and a constant value of 0.17 was assumed, which relates to a Secchi depth of 10 m. This refers to typical winter values, when observed. Note, that the winter values for KDv can be assumed to have

Fig. 1. (a) Climatological mean value of KDv in summer (JJA) in m− 1. (b) Assumed trends in Secchi depth for the experiments TREND1 (blue line) and TREND2 (red line) in the Baltic proper in July.

a much weaker influence than the summer values, due to the comparable weak incoming solar radiation. During the remaining months missing values were substituted by linear interpolation and the data were smoothed afterwards using a boxcar smoother of 2 in xand 4 in y-direction. During the model simulation, the month to month transitions were also smoothed. The climatology of KD(490) shows maximum values in summer. The seasonal mean is depicted in

Table 1 Acronyms for model experiments discussed in the text (first column), a short description and the attenuation coefficient used for visible light. INCR1 and INCR2 denote some factors, increasing with time as described in the section ‘Experiments’. Exp.

Description

Reference run Constant attenuation coefficient CLIM Attenuation depends on clim. KD(490) measurements TREND1 Slow increase in Secchi depth TREND2 Fast increase in Secchi depth

KDv 0.2 m− 1 0.6677*KD(490)0.6763 m− 1 0.6677*KD(490)0.6763*INCR1 m− 1 0.6677*KD(490)0.6763*INCR2 m− 1

Fig. 2. Locations of the observational data stations considered in the following.

326

U. Löptien, H.E.M. Meier / Journal of Marine Systems 88 (2011) 323–331

The same climatology is used for the third and fourth experiments (experiment TREND1 resp. TREND2) which are designed to test a possible impact on SST trends. In contrast to experiment CLIM which uses a fixed climatology in time, a gradual increase in KDv is now assumed. Since Secchi depth measurements are not comprehensive in the Baltic Sea, we perform two idealized experiments which are based on station data of Secchi depth and a spatial pattern of water turbidity which matches the present day satellite observations of KD(490). The experiment TREND1 is based on an increase of the attenuation coefficient KDv by 0.4% per year until 1985, 1.3% from 1985 to 2000 and 0.4% after 2000. This increase is adapted to Secchi depth observations in the Bornholm Basin and the Bothnian Sea (Fig. 1b). Experiment TREND2 assumes a stronger increase of 3% from 1985 to 2002 and 0.4% per year otherwise matching observations of Secchi depth in the Northern Gotland Basin until 2002 (Leppäranta and Myrberg, 2009, page 300). This drastic increase in attenuation during the 90s can be regarded as an upper bound.

2.4. Data sources

Fig. 3. (a) Anomalous net heating rates in the upper 18 m [K/day] for different attenuation coefficient compared to an attenuation of 0.2 m− 1 based on a solar radiation (visible part) of 120 W/m2. (b) Anomalous net surface heating rate [K/day] integrated over the upper 10 m for different specific attenuation coefficients for visible light ranging from 0.15 to 0.36 m− 1.

Fig. 1a. Particularly large values occur close to the eastern coasts and in the Gulf of Finland (0.34 m − 1) while the summer mean values in the central Baltic proper are around 0.26–0.28 m − 1. Most extreme climatological summer values are reached in July where large areas of the Gulf of Finland show values up to 0.45 m − 1.

For a comparison of our model results with in situ data we use quality checked station data from the Swedish Oceanographical Data Center (SHARK) at the Swedish Meteorological and Hydrological Institute (SMHI). Measurements are provided daily (if available) and we considered all data up to 2 m depth. We chose six stations with high data coverage for the locations shown in Fig. 2. As a second data source we use satellite SST provided by the BSH (Bundesamt für Seeschifffahrt und Hydrographie, Germany) since 1990. This weekly database contains the closest available measurements from the 7 previous days of the week since Wednesday. It is thus unknown at which date and time during the day the observations were taken while the advantage is, that missing values due to cloud cover can be reduced to small areas.

Fig. 4. Mean seasonal cycle of SST (in °C) at different stations (1962–2007). The red line depicts modeled SSTs of the reference run and the black line observations from the SHARK database. The black bars depict the range of ± one standard deviation in the observations.

U. Löptien, H.E.M. Meier / Journal of Marine Systems 88 (2011) 323–331

3. Results 3.1. Net heating rate Fig. 3a is a conceptual figure about anomalous net heating rates in the upper 18 m in dependence of different attenuation coefficients compared to a reference value of 0.2 m − 1 for visible light. The visible part of the incoming solar radiation is assumed to be 120 W/m 2 which relates to typical monthly mean values during summer in the southern Baltic Sea (Dera and Wozniak, 2010). A conversion to different amounts of shortwave radiation is simply linear. The vertical profile in Fig. 3a shows a clear positive anomaly of the net heating rates close to the surface when KDv increased. The reversed effect occurs below the surface. In the surface layer, the relationship between KDv and the net heating rate is almost linear. An increase of 0.01 m − 1 in the attenuation coefficient leads on average to an increase of the net heating rate of 0.002 K/day in the considered parameter range (not shown). However, surface waters in the oceans are usually well mixed and it is the integrated effect over the mixed layer that is observed. In the Baltic Sea, the summer mixed layer depth can be assumed to be between 10 and 20 m (Leppäranta and Myrberg, 2009, page 71). The anomalous net heating rates integrated over the upper 10 m are much weaker than directly at the surface and in general non-linear (Fig. 3b). The largest changes occur at small values

327

of KDv. For instance, an increase from 0.15 to 0.16 m − 1 and from 0.3 to 0.31 m − 1 leads to increases in net heating rates of 0.16 K/month and 0.04 K/month, respectively. Thus, a decrease in Secchi depth from 10 to 5 m (resp. increase in the attenuation coefficient from 0.17 to 0.34 m − 1), as observed in the Baltic proper during the recent 60 years, leads to a substantial increase in the net heating rates in summer. Assuming three months of strong solar radiation in summer, this would result in a trend of approx. 0.5 K per decade. Assuming a deeper mixed layer of 20 m, this value is considerably smaller and of the order of 0.1 K per decade. 3.2. The seasonal mean sea surface temperatures 3.2.1. Model–data comparison (mean state) First, we compare the modeled SST of the reference run with station data. Here, monthly and seasonal means were calculated from all dates when observations were available. The mean seasonal cycles at different stations are depicted in Fig. 4. The largest biases occur at the northernmost station and are very likely due to a delay in icemelting (not shown). Otherwise, largest differences occur in autumn and spring and the model tends to simulate higher temperatures than observed. In summer, the difference between observations and modeled SSTs is rather small, apart from the northern most station. All deviations between modeled and observations are well within the

Fig. 5. Difference between satellite and modeled SST of the reference run (in K) in (a) spring (MAM), (b) summer (JJA), (c) autumn (SON) and (d) winter (DJF) during the period 1990–2007. White areas are due to missing data.

328

U. Löptien, H.E.M. Meier / Journal of Marine Systems 88 (2011) 323–331

range of one standard deviation of the observed values. Also, the seasonal mean modeled vertical temperature profiles show a satisfactory agreement with observations (not shown). A comparison between satellite data and modeled SST is shown in Fig. 5. Note, that the satellite measures skin temperature instead of SST and the data also contain some ‘cloud bias’ (only cloud free conditions are considered). Thus, some differences between in situ and satellite observations have to be expected. Also, the precise date and time of the observation is unknown. We thus consider all months where data were available for the computation of the climatological means. In contrast to the station data, the differences between model and satellite observations are smallest in spring (approx. 0.2–0.3 K) and relatively homogeneous over the whole Baltic Sea. In summer, the difference pattern is more ‘patchy’, showing slightly colder temperatures than the satellite data in the central Gotland Basin and Bothnian Sea, while the simulated temperatures are warmer in some coastal areas. As in the station data, the temperatures in the Bothnian Bay are too cold in summer and spring. In autumn the biases are largest, ranging from +0.2 K in the Baltic proper to −0.4 to −0.5 K in coastal regions and the Gulf of Finland. These coastal differences are very likely related to wind driven upwelling. In winter the satellite shows basin wide somewhat warmer values, reaching maximum values in the Gulf of Riga and Finland. Otherwise, the differences are in the range ±0.4 K almost everywhere during all seasons. 3.2.2. Modeled influence of varying water turbidity on seasonal mean SSTs Comparing the sensitivity experiment CLIM to the reference run reveals that the mean long-term response to spatial variations in KDv is straightforward in the Baltic Sea. Largest SST changes occur in summer when the solar radiation is strongest, the mixed layer is shallow and Secchi depth is minimal. Fig. 6a shows the mean SST differences between experiments CLIM and the reference run during summer (JJA). Anomalous warm SSTs match usually turbid regions and strongest differences occur in areas with high values of KDv. Maximum SST differences compared to the reference run occur in the Gulf of Finland, reaching values up to 0.5 K close to the coast and around 0.2 K elsewhere. According to a t-test the mean summer differences are significant almost everywhere except large parts of the Arkona Basin and the Bothnian Sea and smaller parts of the Bothnian Bay (not shown). Fig. 6b depicts a temperature section at 59.7°N. A strong subsurface cooling below the thermocline coincides with the surface warming. The sub-surface temperature anomaly matches the spatial pattern of KDv well and thus, secondary effects due to changes in the circulation, e.g. as observed in the tropical Pacific (Anderson et al., 2009; Löptien et al., 2009; Sweeney et al., 2005; Timmermann and Jin, 2002), seem to be of minor importance. In winter and spring, the effects of the modified light attenuation are very small. In spring, small SST anomalies occur in coastal regions. Elsewhere, the incoming radiation is apparently too weak and the mixed layer too deep to lead to significant changes in the open sea (not shown). In late autumn and beginning of winter, the sub-surface temperature anomalies are mixed into the mixed layer leading to slightly colder temperatures in the open sea than in the reference run (not shown). However, this effect is rather small and of the order of less than − 0.1 K, even though in autumn significant differences occur in the central Baltic Sea, the Gulf of Riga and the Gulf of Finland. Thus, in summary the inclusion of a spatially varying light attenuation leads to minor model improvement in the Baltic proper but cannot overcome the problems related to ice melting and coastal upwelling. 3.3. Sea surface temperature trends 3.3.1. Model–data comparison (trends) When considering linear trends of satellite observations (1990– 2007), both model and observations show strongest basin wide trends

Fig. 6. (a) Mean SST-differences (1962–2007) between the experiments CLIM and the reference run in summer (JJA). The data were smoothed with a moving average of two grid cells. The red line depicts the position of the section considered in (b). (b) Mean temperature difference between the same experiments at 59.7°N in summer.

during the summer season (not shown). This is in agreement with the findings of MacKenzie and Schiedeck (2007) and Siegel et al. (2006). Since, additionally changes in the attenuation of visible light in an observed range have been shown in Section 3.2.2. to have a considerable effect in summer only, we focus on the summer season in the following. Fig. 7 shows the basin mean (15° E–28° E, 56° N–66° N) summer temperatures for the period 1990–2007 and the corresponding linear trends. Despite a satisfactory simulation of the mean SSTs, model and observations show rather large discrepancies when considering the summer trends. The modeled trend is considerably weaker than the observed trend. This becomes especially apparent when comparing the spatial pattern. Fig. 8 shows the spatial distribution of linear trends in satellite and model data during the period 1990–2007 in summer. The order of magnitude of the satellite trends is in close agreement with the values reported by MacKenzie and Schiedeck (2007), while Siegel et al. (2006) found somewhat higher summer values at the stations ‘Arkona Sea’, ‘Gotland Sea’ and ‘Bothnian Sea’ for the shorter period 1990–2004. Nevertheless, also here the magnitude of the reported trends is similar. According to

U. Löptien, H.E.M. Meier / Journal of Marine Systems 88 (2011) 323–331

Fig. 7. Basin mean summer temperatures as observed by satellite (red line) and modeled by the reference run (green line) together with the corresponding linear trends [°C]. The modeled temperatures are depicted with an offset of 4 K.

Fig. 8 the modeled summer trends are in the order of 0.4–0.7 K too weak in large parts of the basin, especially in the Bothnian Bay, Bothnian Sea and in the Gotland Basin. This lack of the strong summer trends in the central and northern Baltic Sea in the model does not have an obvious explanation. We might be either missing some physical processes (e.g. the feedback of water turbidity on the SST) or model or forcing deficiencies might cause this difference.

3.3.2. Modeled influence of varying water turbidity on SST-trends Fig. 9 shows the differences in the modeled trends (1990–2007) between experiment TREND1 and the reference run as well as TREND2

329

and the reference run. The largest deviations from the trends in the reference run occur in experiment TREND2 in the Gulf of Finland, in the western part of the Gotland Basin and in the near coastal areas while the lowest values are seen in the central Bothnian Bay. Maximum differences are of the order of 0.06–0.25 K/decade. Considerably weaker values are obtained when considering experiment TREND1 instead of TREND2. For both experiments, largest values occur close to the coast and in shallow regions. Otherwise the values are considerably smaller and the pattern is relatively uniform despite the spatial variability in the pattern of KDv. Here, the differences of KDv are likely to compensate. On one hand largest trends in KDv appear where KDv is large but on the other hand changes of small values have a larger effect (see Fig. 9b). Due to the short time period the pattern is rather patchy and differences are more robust, when considering a longer period. The modeled long-term trends since 1962 are in agreement with MacKenzie and Schiedeck (2007), considerably weaker than the trends during the period 1990–2007 (Fig. 10a in comparison to Fig. 8b). The trends simulated by the experiments TREND2 compared to the reference run are shown in Fig. 10b. The spatial differences between both are similar to those in Fig. 9. These differences are significant in shallow regions close to the coast, the eastern part of the Gulf of Finland, the Bothnian Bay and large parts of the Baltic proper on the 5%-level (Mann–Kendall, not shown). The order of the differences is around 0.25 K/decade close to the coast and around 0.08–0.1 K/decade in the open Baltic. As expected, the trend difference between the experiments TREND1 and the reference run are even smaller. Thus, even though water turbidity has a considerable influence this influence is neither strong enough to explain the discrepancies between modeled and observed SST-trends nor the large observed contrast to the surrounding seas. Note, however, that the influence in shallow water close to the coast and in shallow regions can be substantial. 4. Conclusions SST changes in the Baltic Sea are an important issue and might have a large impact on the marine ecosystem (MacKenzie and Schiedeck, 2007). This holds true especially for harmful algae blooms of cyanobacteria that are known to depend strongly on the summer SST (Kanoshina et al., 2003; Lehtimaki et al., 1997). It was shown statistically by e.g. Suikkanen et al. (2006) that apart from nutrients the increasing summer temperature has an effect on phytoplankton communities, especially on cyanobacteria and even small changes are assumed to make a big difference.

Fig. 8. (a) Linear trends in satellite SST in summer (JJA, 1990–2007) and (b) modeled trends of the reference run for the same period [K/decade].

330

U. Löptien, H.E.M. Meier / Journal of Marine Systems 88 (2011) 323–331

Fig. 9. The differences between the simulated summer trends (1990–2007, JJA) between (a) experiment TREND2 and the reference run, (b) experiment TREND1 and the reference run in K/decade. The data shown here, were smoothed with a moving average of two grid cells.

As an attempt to improve the modeling of SSTs in the Baltic Sea, a spatially and climatological varying attenuation of light was included into a state-of-the-art ocean model. The effect is locally considerable (up to 0.5 K) while it is not strong enough to lead to a substantial model improvement in our setup. Furthermore, it was shown by the analysis of satellite data that strong summer SST trends exist in large parts of the basin during the period 1990–2007. Similar trends were discussed by other authors also (MacKenzie and Schiedeck, 2007). These trends are only partly reproduced by 3D-ocean model simulations with fixed attenuation of light and the modeled trend is about 0.4–0.7 K/decade too weak in large parts of the Gotland Basin and the Bothnian Sea. The reasons for the discrepancies are unclear and the question arises on whether the missing part might be due to changes in solar absorption, caused by increasing water turbidity. We show the potential of increasing water turbidity to influence summer SST in the Baltic Sea and the corresponding trends significantly. Idealized experiments, including the effect of increasing water turbidity during the last decades, show stronger summer SST trends than the reference run. A sensitivity experiment based on the strongest observed increase in water turbidity in the past (experiment TREND2) leads to the best agreements with observed trends in large parts of the

basin, while it is unlikely that the bulk of the missing signal is due to this mechanism. A local increase in summer SST of 0.04–0.1 K per decade due to increasing water turbidity seems likely while close to the coast this effect might be more than doubled. The order of magnitude of this increase in summer SST trends agrees well with theoretical consideration of net heating rates. Note, however, that these experiments are idealized due to limited data availability and also a two way coupling with the atmosphere is not included. Thus, some large uncertainties exist for these first estimates, which might be reduced in the future by an extension of databases including the optical properties of the oceans. Also, an extension of basin-wide SST measurements will reveal whether the recent trends turn out to be robust. Acknowledgments This study was performed within the projects ‘Sensor Networks to Monitor Marine Environment with Particular Focus on Climate Changes’ funded by the Swedish Governmental Agency for Innovation Systems (Vinnova, project-no. P29461-1) and ‘Advanced modeling tool for scenarios of the Baltic Sea ECOsystem to SUPPORT decision making

Fig. 10. Simulated linear trends in summer (JJA) for the period 1962–2007 in (a) the reference run and (b) in experiment TREND2 [K/decade].

U. Löptien, H.E.M. Meier / Journal of Marine Systems 88 (2011) 323–331

ECOSUPPORT’ jointly funded by the Swedish Environmental Protection Agency (Naturvå̊rdsverket, grant no. 08/381) and the European Commission within the BONUS+ program (www.bonusportal.org). The RCO model simulations were partly performed on the climate computing resources ‘Vagn’ and ‘Ekman’ that are operated by the National Supercomputer Centre (NSC) at Linköping University and the Centre for High Performance Computing (PDC) at the Royal Institute of Technology in Stockholm, respectively. These computing resources are funded by a grant from the Knut and Alice Wallenberg foundation. We are thankful for these supports. Satellite data were measured by MODIS (Moderate-resolution Imaging Spectroradiometer) as a part of the NASA-centered international Earth Observing System (EOS). Furthermore, we would like to thank Elin Almroth, Susanne Kratzer, Manu Thomas, Christian Dieterich and the anonymous reviewers for helpful and encouraging comments. References Aarup, T., 2002. Transparency of the North Sea and the Baltic Sea — a Secchi depth data mining study. Oceanologica 44 (3). Anderson, W., Gnanadesikan, A., Wittenberg, A., 2009. Regional impacts of ocean color on tropical Pacific variability. Ocean Sci. Discuss. 6, 243–275. Belkin, I., 2009. Rapid warming of large marine ecosystems. Prog. Oceanogr. 81, 207–213. Bergström, S., Carlsson, B., 1994. River runoff to the Baltic Sea: 1950–1990. Ambio 23, 280–287. Boesch, D., Hecky, R., O'Melia, C., Schindler, D., Seitzinger, S., 2006. Eutrophication of Swedish Seas. Swedish Environmental Protection Agency, Report No.5509, pp. 1–72. Boesch, D., Hecky, R., O'Melia, C., Schindler, D., Seitzinger, S., 2008. Eutrophication of seas along Sweden's West Coast. Swedish Environmental Protection Agency, Report No.5898, pp. 1–78. Bradtk, K., Herman, A., Urbanski, J.A., 2010. Spatial and interannual variations of seasonal sea surface temperature patterns in the Baltic Sea. Oceanologia 52, 345–362. Chang, G.C., Dickey, T.D., 2004. Coastal ocean optical influences on solar transmission and radiant heating rate. J. Geophys. Res. 109. doi:10.1029/2003JC001821. Dera, J., Wozniak, B., 2010. Solar radiation in the Baltic Sea. Oceanologia 52, 533–582. Döscher, R., Willén, U., Jones, C., Rutgersson, A., Meier, H.E.M., Hansson, U., 2002. The development of the coupled ocean–atmosphere model RCAO. Boreal Environ. Res. 7, 327–334. Elmgren, R., Larson, U., 2001. Eutrophication in the Baltic Sea area: integrated coastal management issues. Science and Integrated Coastal Management. Dahlem University Press, Berlin, pp. 15–35. Graham, L.P., 2004. Climate change effects on river flow to the Baltic Sea. Ambio 33, 235–241. Jansson, B.O., Dahlberg, K., 1999. The environmental status of the Baltic Sea in the 1940s, today, and in the future. Ambio 28, 312–319. Jones, C., Willén, U., Ullerstig, A., Hansson, U., 2004. The Rossby Centre Regional Atmospheric Climate Model part I: model climatology and performance for the present climate over Europe. Ambio 33, 199–210. Kahru, M., Leppänen, J.M., Rud, O., 1993. Cyanobacterial blooms cause heating of the sea surface. Mar. Ecol. Prog. Ser. 101, 1–7. Kanoshina, I., Lips, U., Leppänen, J.M., 2003. The influence of weather conditions (temperature and wind) on cyanobacterial bloom development in the Gulf of Finland (Baltic Sea). Harmful Algae 2, 29–41. Kratzer, S., Tett, P., 2009. Using bio-optics to investigate the extent of coastal waters: a Swedish case study. Hydrobiologica 629, 169–186. Kratzer, S., Hå̊kansson, B., Sahlin, C., 2003. Assessing Secchi and photic zone depth in the Baltic Sea from satellite data. Ambio 32 (8), 577–585. Laamanen, M., Fleming, V., Olsonen, R., 2004. Water transparency in the Baltic Sea between 1903 and 2004. HELCOM indicator Report 2004. Lehtimaki, J., Moisander, P., Sivonen, K., Kononen, K., 1997. Growth, nitrogen fixation, and Nodularin production by two Baltic Sea cyanobacteria. Appl. Environ. Microbiol. 63, 1647–1656. Leppäranta, M., Myrberg, K., 2009. Physical Oceanography of the Baltic Sea. Springer. ISBN: 978-3-540-79702-9, pp. 85,296–85,303. Löptien, U., Eden, C., Timmermann, A., Dietze, H., 2009. Effects of biologically induced differential heating in an eddy-permitting coupled ocean–ecosystem model. J. Geophys. Res. 114. doi:10.1029/2008JC004936. MacKenzie, B.R., Schiedeck, D., 2007. Daily ocean monitoring since 1860s shows record warming on northern European seas. Glob. Chang. Biol. 13 (7), 1335–1347. Meier, H.E.M., 2005. Modeling the age of Baltic Sea water masses: quantification and steady state sensitivity experiments. J. Geophys. Res. 110. doi:10.1029/2004JC002 607.

331

Meier, H.E.M., 2006. Baltic Sea climate in the late twenty-first century: a dynamical downscaling approach using two global models and two emission scenarios. Clim. Dyn. 27, 39–68. Meier, H.E.M., Faxen, T., 2002. Performance analysis of a multiprocessor coupled ice– ocean model for the Balit Sea. J. Atmos. Oceanic Technol. 19, 114–124. Meier, H.E.M., Kauker, F., 2003a. Modeling decadal variability of the Baltic Sea: 2. Role of freshwater inflow and large-scale atmospheric circulation for salinity. J. Geophys. Res. 108. doi:10.1029/2003JC001.799. Meier, H.E.M., Kauker, F., 2003b. Sensitivity of the Baltic Sea salinity to the freshwater supply. Clim. Res. 24, 231–242. Meier, H.E.M., Döscher, R., Faxén, T., 2003. A multiprocessor coupled ice–ocean model for the Baltic Sea: application to salt inflow. J. Geophys. Res. 108. doi:10.1029/ 2000JC000.521. Nehring, D., 1992. Eutrophication in the Baltic Sea. Sci. Total Environ. Suppl. 673–682. Nielsen, S.L., Sand-Jensen, K., Borum, J., Geertz-Hansen, O., 2002. Phytoplankton, nutrients, and transparency in Danish coastal waters. Estuaries Coasts 25 (5), 930–937. Oschlies, A., 2004. Feedbacks of biotically induced radiative heating on upper-ocean heat budget, circulation, and biological production in a coupled ecosystemcirculation model. J. Geophys. Res. 109. doi:10.1029/2004JC002430, 2004. Pawlak, J.F., Laamanen, M., Andersen, J.H., 2009. Eutrophication in the Baltic Sea — an integrated thematic assessment of the effects of nutrient enrichment in the Baltic Sea region. An executive summary. Baltic Sea Environment Proceedings No. 115A. Helsinki Commission (Baltic Marine Environment Protection Commission), pp. 1–18. Pierson, D.C., Kratzer, S., Strömbeck, N., Hå̊kanson, B., 2008. Relationship between the attenuation of downwelling irradiance at 490 nm with the attenuation of PAR (400 nm–700 nm) in the Baltic Sea. Remote Sens. Environ. 112, 668–680. Rudolph, C., Lehmann, A., 2006. A model-measurements comparison of atmospheric forcing and surface fluxes of the Baltic Sea. Oceanologia 48, 333–360. Samuelsson, P., Jones, C.G., Willen, U., Ullerstig, A., Gollvik, S., Hansson, U., Jansson, C., Kjellström, E., Nikulin, G., Wyser, K., 2011. The Rossby Centre Regional Climate model RCA3: model description and performance. Tellus A 63, 4–23. Sanden, P., Hå̊kansson, S., 1996. Long-term trends in Secchi depth in the Baltic Sea. Limnol. Oceanogr. 41, 346–351. Sand-Jensen, K., Borum, J., 1991. Interactions among phytoplankton, periphyton, and macrophytes in temperate, freshwaters and estuaries. Aquat. Bot. 41, 137–175. Schrum, C., Hünbar, U., Jacob, D., Podzun, R., 2003. A coupled atmosphere/ice/ocean model for the North Sea and the Baltic Sea. Clim. Dyn. 21, 131–151. Seifert, T., Kayser, B., 1995. A high resolution spherical grid topography of the Baltic Sea. Meereswiss. Ber., 9. Warnemünde, Germany. Siegel, H., Gerth, M., Ohde, T., Heene, T., 2005. Ocean colour remote sensing relevant water constituts and optical properties of the Baltic Sea. Int. J. Remote. Sens. 26 (2), 315–330. Siegel, H., Gerth, M., Tschersich, G., 2006. Sea surface temperature development of the Baltic Sea in the period 1990–2004. Oceanologia 48, 119–131. Stevens, D.P., 1990. On open boundary conditions for three dimensional primitive equation ocean circulation models. Geophys. Astrophys. Fluid Dyn. 51, 103–133. Suikkanen, S., Laamanen, M., Huttunen, M., 2006. Long-term changes in summer phytoplankton communities of the open northern Baltic Sea. Ambio 37 (2), 86–92. Sweeney, C., Gnanadesikan, A., Griffies, S.M., Harrison, M.J., Rosati, A.J., Samuels, B.L., 2005. Impacts of shortwave penetration depth on large-scale ocean circulation and heat transport. J. Phys. Oceanogr. 35, 1103–1119. Theakaekara, M., 1974. Extraterrestrial solar spectrum, 3000–6100 Å at 1-Å intervals. Appl. Opt. 13, 518–522. Timmermann, A., Jin, F.F., 2002. Phytoplankton influences on tropical climate. Geophys. Res. Lett. 39 10.10129/2002GL15434. Uppala, S., Kallberg, P.W., Simmons, A.J., Andrae, U., Da Costa Bechtold, V., Fiorino, M., Gibson, J.K., Haseler, J., Hernandez, A., Kelly, G.A., Li, X., Onogi, K., Saarinen, S., Sokka, N., Allan, R.P., Andersson, E., Arpe, K., Balmaseda, M.A., Beljaars, A.C.M., van de Berg, L., Bidlot, J., Bormann, N., Caires, S., Chevallier, F., Dethof, A., Dragosavac, M., Fisher, M., Fuentes, M., Hagemann, S., Holm, E., Hoskins, B.J., Isaksen, L., Janssen, P.A.E.M., Jenne, R., McNally, A.P., Mahfouf, J.-F., Morcrette, J.-J., Rayner, N.A., Saunders, R.W., Simon, P., Sterl, A., Trenberth, K.E., Untch, A., Vasiljevic, D., Viterbo, P., Woollen, J., 2005. The ERA-40 reanalysis. Q. J. R. Met. Soc. 131, 2961–3012. Wasmund, N., Uhlig, S., 2003. Phytoplankton trends in the Baltic Sea. ICES J. Mar. Sci. 60, 177–186. Webb, D.J., Coward, A.C., de Cuevas, B.A., Gwilliam, C.S., 1997. A multiprocessor ocean circulation model using message passing. J. Atmos. Oceanic Technol. 14, 175–183. Webb, D.J., de Cuevas, B.A., Richmond, C.S., 1998. Improved advection schemes for ocean models. J. Atmos. Oceanic Technol. 15 (5), 1171–1187. Werdell, P.J., Bailey, S.W., 2005. An improved in-situ bio-optical data set for ocean color algorithm development and satellite data product validation. Remote Sens. Environ. 98, 122–140. Wilhelmsson, T., 2002. Parallelization of the HIROMB Ocean Model. Royal Institute of Technology, Stockholm. ISBN 91-7283-296. Wulff, F., Stigebrandt, A., Rahm, L., 1999. Nutrient dynamics of the Baltic Sea. Ambio 19, 126–133.