7 July 2021
The Baltic Sea, located in northern Europe, is a semi-enclosed, shallow and tideless sea with seasonal sea-ice cover in its northern sub-basins. Its long water residence time contributes to oxygen depletion in the bottom water of its southern sub-basins. In this study, recently performed scenario simulations for the Baltic Sea including marine biogeochemistry were analysed and compared with earlier published projections. Specifically, dynamical downscaling using a regionally coupled atmosphere–ocean climate model was used to regionalise four global Earth system models. However, as the regional climate model does not include components representing terrestrial and marine biogeochemistry, an additional catchment and a coupled physical–biogeochemical model for the Baltic Sea were included. The scenario simulations take the impact of various global sea level rise scenarios into account. According to the projections, compared to the present climate, higher water temperatures, a shallower mixed layer with a sharper thermocline during summer, less sea-ice cover and greater mixing in the northern Baltic Sea during winter can be expected. Both the frequency and the duration of marine heat waves will increase significantly, in particular in the coastal zone of the southern Baltic Sea (except in regions with frequent upwellings). Nonetheless, due to the uncertainties in the projections regarding regional winds, the water cycle and the global sea level rise, robust and statistically significant salinity changes could not be identified. The impact of a changing climate on biogeochemical cycling is predicted to be considerable but still smaller than that of plausible nutrient input changes. Implementing the proposed Baltic Sea Action Plan, a nutrient input abatement plan for the entire catchment area, would result in a significantly improved ecological status of the Baltic Sea, including reductions in the size of the hypoxic area also in a future climate, which in turn would increase the resilience of the Baltic Sea against anticipated climate change. While our findings regarding changes in heat-cycle variables mainly confirm earlier scenario simulations, they differ substantially from earlier projections of salinity and biogeochemical cycles, due to differences in experimental setups and in input scenarios for bioavailable nutrients.
The Baltic Sea is a shallow, semi-enclosed sea located in northern Europe
(Fig. 1). It has a mean depth of 54 m, but due to its strongly varying bottom
topography it can be divided into several sub-basins, with limited transport
between them (Sjöberg, 1992). In particular, water exchange between the
Baltic Sea and the North Sea is hampered because of two shallow sills
located in narrow channels connecting these two water bodies. Thus, large
saltwater inflows occur only sporadically, on average once per year, mainly
during the winter season but never during summer (Mohrholz, 2018). Furthermore, because the Baltic Sea is embedded within a catchment area that
is about 4 times larger than the Baltic Sea surface, annual freshwater
inputs are large relative to the volume of the Baltic Sea (Bergström and
Carlsson, 1994). The volume of the Baltic Sea is
Bottom topography of the Baltic Sea (depth in metres). The Baltic Proper comprises the Arkona Basin, Bornholm Basin and Gotland Basin. The border of the analysed domain of the Baltic Sea models is shown as a black line in the northern Kattegat. The tide gauges in Klagshamn (55.522
Some 85 million people, in 14 countries, currently live in the catchment area of the Baltic Sea, and anthropogenic pressure on the marine ecosystem is accordingly high (HELCOM, 2018). Insufficiently treated wastewater, pollutant emissions, overfishing, habitat degradation and intensive marine traffic, including oil transport, place a heavy burden on the Baltic Sea ecosystem (Reckermann et al., 2022). One consequence is oxygen depletion of the Baltic Sea's deep waters, such that bottom areas lack higher life forms (e.g. Carstensen et al., 2014; Meier et al., 2018b). In 2018, the area of dead bottom was equal to that of the Republic of Ireland,
Selected ensembles of the scenario simulations for the Baltic Sea carried out in international projects (AR is the IPCC Assessment Report, GCM is a general circulation model, RCSM is a regional climate system model, RCAO is the Rossby Centre Atmosphere Ocean model, RCA4 is the Rossby Centre Atmosphere model version 4, NEMO is the Nucleus for European Modelling of the Ocean, REMO is a regional model, MPIOM is the Max Planck Institute Ocean Model, and HAMSOM is the Hamburg Shelf Ocean Model).
Projections of the Baltic Sea's climate at the end of the 21st century were among the first to be made for coastal seas worldwide (Meier and Saraiva, 2020). Already at the beginning of the 2000s, the first scenario simulations were carried out for selected time slices in present and future climates (e.g. Haapala et al., 2001; Meier, 2002a, b; Omstedt et al., 2000). In the dynamical downscaling approach used for those simulations, regional climate models (RCMs) were employed to refine predictions of global climate change to regional and local scales, in this case for the Baltic Sea (e.g. Rummukainen et al., 2004; Döscher et al., 2002). However, those first projections were based on scenarios consisting of a single global climate model (GCM) and a single greenhouse gas (GHG) concentration (150 % increase in equivalent
For the second assessment of climate change in the Baltic Sea region (BACC II Author Team, 2015), continuously integrated transient simulations from present to future climates became available and even included marine biogeochemical modules (e.g. Eilola et al., 2013; Friedland et al., 2012; Gräwe and Burchard, 2012; Gräwe et al., 2013; Gröger et al., 2019, 2021b; Holt et al., 2016; Kuznetsov and Neumann, 2013; Meier et al., 2011b, c, 2012a, c, d; Neumann, 2010; Neumann et al., 2012; Omstedt et al., 2012; Pushpadas et al., 2015; Ryabchenko et al., 2016; Skogen et al., 2014) and higher trophic levels (e.g. Bauer et al., 2019; Ehrnsten et al., 2020; Gogina et al., 2020; Holopainen et al., 2016; MacKenzie et al., 2012; Niiranen et al., 2013; Vuorinen et al., 2015; Weigel et al., 2015). The BACC II Author Team (2015) concluded that “recent studies confirm the findings of the first assessment of climate change in the Baltic Sea basin”. A key finding of their report was that “No clear tendencies in saltwater transport were found. However, the uncertainty in salinity projections is likely to be large due to biases in atmospheric and hydrological models. Although wind speed is projected to increase over sea, especially over areas with diminishing ice cover, no significant trend was found in potential energy …” (a measure of energy to homogenise the water column). “In accordance with earlier results, it was found that sea level rise has greater potential to increase surge levels in the Baltic Sea than does increased wind speed. In contrast to the first BACC assessment (BACC Author Team, 2008), the findings reported in this chapter are based on multi-model ensemble scenario simulations using several GHG emissions scenarios and Baltic Sea models. However, it is very likely that estimates of uncertainty caused by biases in GCMs are still underestimated in most studies” (BACC II Author Team, 2015).
Since the early 21st century, transient simulations for the period 1960–2100 using regional ocean (Holt et al., 2016; Pushpadas et al., 2015) and regionally coupled atmosphere–ocean models, so-called regional climate system models (RCSMs; Bülow et al., 2014; Dieterich et al., 2019; Gröger et al., 2019, 2021b), have been available for the entire combined Baltic Sea and North Sea system. An overview was given by Schrum et al. (2016) as part of the North Sea Region Climate Change Assessment Report (NOSCCA, Quante and Colijn, 2016) and by Gröger et al. (2021a) within the Baltic Earth Assessment Reports (BEAR) project (this thematic issue).
There is a notable difference in the salinity projections between the first two assessments (BACC Author Team, 2008; BACC II Author Team, 2015) and recent scenario simulations (Meier et al., 2021). The first Baltic Sea scenario simulations, driven by nine RCMs and five GCMs, showed a pronounced negative ensemble mean change in salinity because two of the GCMs included a significant increase in the mean west wind component (Meier et al., 2006). These pronounced changes in the large-scale atmospheric circulation were not a feature of later studies (Saraiva et al., 2019a). However, as the natural variability was poorly sampled, this finding may be coincidental.
The large spread in river discharge did not decrease between the studies,
ranging from
Salinity projections assessed by the BACC Author Team (2008), BACC II Author Team (2015) and BEAR (this study). Salinity changes depend on the changes in the wind field (in particular, the west wind component), river discharge and sea level rise (SLR). The changes refer to the mean differences between historical and future periods. (Data sources: Meier et al., 2006, 2011b, 2021).
In the following, we provide an overview of the projections performed since 2013, i.e. after the last assessment of climate change for the Baltic Sea basin, and compare recent results with previous findings by the BACC II Author Team (2015). We focus on projections for the marine environment from both physical and biogeochemical perspectives. Among the analysed variables are temperature, salinity, oxygen, phosphate, nitrate, phytoplankton, primary production, nitrogen fixation, hypoxic area and Secchi depth (measuring water transparency). An accompanying study by Christensen et al. (2021) investigated atmospheric projections in the Baltic Sea region. For an overview of the development of RCSMs and their applications, the reader is referred to Gröger et al. (2021a). In our comparisons of the various scenario simulations, we analyse only published data (Table 1), with a focus on two recently generated sets of scenario simulations: BalticAPP and CLIMSEA (Table 1; see Saraiva et al., 2019a, b; Meier et al., 2019a, 2021). These are compared with the previous ECOSUPPORT scenario simulations (Meier et al., 2014) assessed by the BACC II Author Team (2015). Investigations of the impact of climate change on primary production in the Baltic Sea that did not utilise a RCM (Holt et al., 2016; Pushpadas et al., 2015) are not addressed herein, nor are nutrient input reduction scenarios under present climate, e.g. as described by Friedland et al. (2021). To our knowledge, further coordinated experiments aimed at projections for the coupled physical–biogeochemical system of the Baltic Sea after 2013 have not been published. Uncoordinated scenario simulations performed prior to 2013 (including Ryabchenko et al., 2016) and their uncertainties were previously discussed by Meier et al. (2018a, 2019b).
The paper is organised as follows. In Sect. 2, the dynamical downscaling method, the catchment and Baltic Sea models, the experimental setup and the analytical strategy are introduced. In Sect. 3, the historical and future climates results of the three scenario simulations, ECOSUPPORT, BalticAPP and CLIMSEA, are compared. Tables 1, 3 and 4 provide an overview of these (Tables 3 and 4) and other (Table 1) scenario simulations from the literature. A consideration of knowledge gaps and a summary of our findings conclude the study. Abbreviations used in this study are defined in Table 5.
List of scenario simulations of three ensembles. From left to right, the columns show the Earth system model (ESM), the regional climate system model (RCSM), the Baltic Sea ecosystem model, the greenhouse gas (GHG) emission or concentration scenario, the nutrient input scenario, the sea level rise (SLR) scenario and the simulation period, including historical and scenario periods. The four nutrient input scenarios were the Baltic Sea Action Plan (BSAP), reference (REF), business-as-usual (BAU) and worst case (WORST) scenarios. For the three SLR scenarios in the CLIMSEA ensemble, the mean sea level changes at the end of the century are given in metres.
Summary of the characteristics of the ECOSUPPORT, BalticAPP and CLIMSEA scenario simulations discussed in this study. For further details, the reader is referred to Tables 1 and 3. Abbreviations are defined in Table 5.
List of abbreviations (in alphabetical order), their definitions and references.
Continued.
Continued.
Dieterich et al. (2019) produced an ensemble of scenario simulations with a
coupled RCSM, called RCA4-NEMO, which was introduced by Wang et al. (2015).
Gröger et al. (2019, 2021b) and Dieterich et al. (2019) validated and
analysed the different aspects of the RCA4-NEMO ensemble discussed herein.
The atmospheric component, RCA4 (Rossby Centre Atmosphere model version 4),
was run at a resolution of 0.22
This RCSM was applied to downscale eight different Earth system models (ESMs), each one driven by three Representative Concentration Pathways (RCPs). For the Baltic Sea projections, four ESMs (MPI-ESM-LR, EC-Earth, IPSL-CM5A-MR, HadGEM2-ES; see Gröger et al., 2019, and references for the ESMs therein) and the GHG concentration scenarios RCP4.5 and RCP8.5 were selected (Table 3). The four ESMs were part of the fifth Coupled Model Intercomparison Project (CMIP5; Taylor et al., 2012) and their results were assessed in the fifth IPCC Assessment Report (AR5; IPCC, 2013).
Surface variables of the atmospheric component were saved at hourly to 6-hourly frequencies to allow for an analysis of means and extremes in present and future climates. As RCA4-NEMO does not contain model components for terrestrial and marine biogeochemistry, two additional models forced with the atmospheric surface fields of RCA4-NEMO, i.e. a catchment and a marine ecosystem model, were employed (Fig. 2).
Dynamical downscaling approach for the Baltic Sea region. The models for the various components of the Earth system are explained in Sect. 2. (Source: Meier et al., 2021.)
For the ECOSUPPORT scenario simulations, dynamical downscaling was performed with the regional Rossby Centre Atmosphere Ocean (RCAO) model (Döscher et al., 2002). RCAO consists of the atmospheric component RCA3 (Samuelsson et al., 2011) and the oceanic component RCO (Meier et al., 2003; Meier, 2007), with horizontal grid resolutions of 25 km and 6 nautical miles (11.1 km), respectively. In the vertical, the ocean model has 41 levels with layer thicknesses ranging between 3 m close to the surface and 12 m at 250 m depth. The latter was the maximum depth in the model.
In BalticAPP/CLIMSEA and ECOSUPPORT, the catchment model E-HYPE
(Hydrological Predictions for the Environment,
In CLIMSEA, two nutrient input scenarios, defining plausible future pathways of nutrient inputs from rivers, point sources and atmospheric deposition, i.e. the BSAP and reference (REF) scenarios (Saraiva et al., 2019a, b), are used (Fig. 3). In BalticAPP, nutrient input scenarios follow BSAP, REF and worst case (WORST) scenarios (Saraiva et al., 2019a, b; Pihlainen et al., 2020). Finally, in ECOSUPPORT, instead of WORST, a business-as-usual (BAU) scenario is applied (Gustafsson et al., 2011; Meier et al., 2011b).
Projections of river discharge and nutrient inputs from land and
atmosphere into the entire Baltic Sea according to the BalticAPP and CLIMSEA
scenario simulations. Upper panel: low-pass-filtered runoff data (in m
In the BSAP scenario in CLIMSEA and BalticAPP, nutrient inputs linearly decrease from the actual values in 2012 (i.e. the average for 2010–2012) to the maximum allowable input in 2020 defined by the mitigation plan. Thereafter, nutrient inputs remain constant until the end of the century. A similar temporal evolution is defined in ECOSUPPORT but with a reference period of 1997–2003 (Gustafsson et al., 2011; their Fig. 3.1).
In the REF scenario, in CLIMSEA and BalticAPP, nutrient inputs are calculated using E-HYPE, which considers the impact of changing river flow on nutrient inputs but neglects any changes in land use or socioeconomic development. These inputs correspond on average to the observed mean inputs during the period 2010–2012.
Projected ensemble mean changes in total (land and atmosphere)
bioavailable annual phosphorus (
The two additional, above-mentioned scenarios on future projections, BAU and WORST, are not compared because the corresponding input assumptions differ (see Meier et al., 2018a). However, both are characterised by population growth and intensified agricultural practices such as land cover changes and fertiliser use (HELCOM, 2007; Zandersen et al., 2019; Pihlainen et al., 2020). In this study, they are discussed only for the sake of completeness.
A comparison of the historical (1980–2005) and future (2072–2097) periods
reveals that the reductions in nutrient inputs under the BSAP scenario are
smaller in ECOSUPPORT than in BalticAPP and CLIMSEA (Meier et al., 2018a;
their Fig. 3). In ECOSUPPORT and BalticAPP/CLIMSEA, using the same physical–biogeochemical model (RCO-SCOBI), input changes of bioavailable
phosphorus amount to
This study used data from three different Baltic Sea models. The Swedish Coastal and Ocean Biogeochemical model coupled to the Rossby Centre Ocean model (RCO-SCOBI) is driven by the atmospheric surface field data calculated by either RCAO or RCA4-NEMO and by the river runoff and nutrient input scenarios derived from either STAT or E-HYPE projections and atmospheric deposition (Fig. 2). Atmospheric deposition is assumed to be constant or reduced as in the BSAP (Fig. 3). RCO is a Bryan–Cox–Semtner-type ocean circulation model with horizontal and vertical grid resolutions of 3.7 km and 3 m, respectively (Meier et al., 1999, 2003; Meier, 2001, 2007). SCOBI is a biogeochemical module of the nutrient–phytoplankton–zooplankton–detritus (NPZD) type; it considers state variables such as phosphate, nitrate, ammonium, oxygen concentration, the phytoplankton concentrations of three algal types (diatoms, flagellates and others, cyanobacteria) and detritus (Eilola et al., 2009; Almroth-Rosell et al., 2011, 2015). RCO-SCOBI has been used in many Baltic Sea climate applications (for an overview, see Meier and Saraiva, 2020), evaluated with respect to measurements and compared with other Baltic Sea models (Eilola et al., 2011; Placke et al., 2018; Meier et al., 2018a).
The Ecological ReGional Ocean Model (ERGOM; see
The BAltic sea Long-Term large-Scale Eutrophication Model (BALTSEM) spatially resolves the Baltic Sea into 13 dynamically interconnected and horizontally averaged sub-basins with high vertical resolution (Gustafsson et al., 2012). For further details of these and other available Baltic Sea ecosystem models, the reader is referred to Meier et al. (2018a).
In CLIMSEA, we analysed the ensemble of 48 RCO-SCOBI scenario simulations for the period 1976–2098 (Table 3) that was produced following the dynamical downscaling approach described in Sect. 2.1–2.3 (Fig. 2) and presented in Meier et al. (2021). Unlike in previous studies (Meier et al., 2011b; Saraiva et al., 2019a), the CLIMSEA scenario simulations also consider various scenarios of global SLR. In the three SLR scenarios starting from the year 2005 that were applied by Meier et al. (2021), the mean sea level changes relative to the seabed projected by the year 2100 are (scenario 1) 0 m, (scenario 2) the ensemble mean of RCP4.5 (0.54 m) and RCP8.5 (0.90 m) IPCC projections (IPCC, 2019b; Hieronymus and Kalén, 2020) and (scenario 3) the 95th percentiles of the lowest case (1.26 m, here combined with RCP4.5) and highest case (2.34 m, here combined with RCP8.5) scenarios following Bamber et al. (2019; Table 3). A deepening of the water depth at all grid points every 10 years increases the relative sea level linearly. The spatially varying land uplift was not considered. For details, the reader is referred to Meier et al. (2021).
The CLIMSEA ensemble simulations are compared with earlier ensemble scenario
simulations by Meier et al. (2011b, 2012c) and Neumann et al. (2012) (ECOSUPPORT), and by Saraiva et al. (2019a, b) and Meier et al. (2019a) (BalticAPP). Both ECOSUPPORT and BalticAPP rely on a downscaling approach similar to that used in the CLIMSEA projections (Fig. 2). However, the scenario simulations of ECOSUPPORT are based upon different global and
regional climate models, three coupled physical–biogeochemical models for
the Baltic Sea and previous GHG emission scenarios as detailed by the fourth
IPCC Assessment Report (AR4; Table 1). Compared to BalticAPP, the CLIMSEA
ensemble is enlarged by three SLR scenarios (Table 3), whereas previous
projections assumed no change in the mean sea level relative to the seabed.
The inclusion of SLR scenarios followed the finding that the relative sea
level above the sills in the entrance area limits transport and controls
salinity in the entire Baltic Sea (Meier et al., 2017). As the relative SLR
during the period 1915–2014 was estimated to be 0–1 mm yr
In this study, the model results of the BalticAPP and CLIMSEA scenario simulations during the historical period were evaluated by calculating the annual and seasonal mean biases during the historical period obtained with RCO-SCOBI simulations and reanalysis data (Liu et al., 2017). Liu et al. (2017) utilised the ensemble optimal interpolation (EnOI) method to integrate profiles of temperature, salinity and the concentrations of oxygen, ammonium, nitrate and phosphate determined by the Swedish environmental monitoring programme into the RCO-SCOBI model. As reanalysis data are available for the period 1971–1999, we limited our bias calculations to 1976–1999, the overlap period between the historical period of the scenario simulations and the reanalysis data. Model data of historical periods of BalticAPP and ECOSUPPORT scenario simulations were evaluated by Saraiva et al. (2019a, b) and Meier et al. (2011b, 2012c, d), respectively.
The mixed-layer depth (MLD) was calculated following de Boyer Montégut
et al. (2004), using a threshold value for the difference between the
near-surface water temperature at 10 m depth and the temperature at the MLD
of
Secchi depth (SD) is a measure of water transparency and is calculated from
SD
First, the monthly average of SST was computed from the model output every 48 h. The linear trend was then calculated using the Theil–Sen estimator (Theil, 1950; Sen, 1968). The trend computed with this method was the median of the slopes determined by all pairs of sample points. The advantage of this computationally expensive method is that it is much less sensitive to outliers. The significance of the SST trends was evaluated from a Mann–Kendall non-parametric test with a threshold of 95 %. The SST trends were computed by season and annually. In the latter case, the annual cycle was removed before the linear trend was computed.
Following Kniebusch et al. (2019), we performed a ranking analysis to identify the atmospheric drivers others than air temperature that are most
important for the monthly variability of SST in each ESM forcing of the CLIMSEA data set and in the RCP scenarios RCP4.5 and RCP8.5. The SST trend
is dominated by the trend in air temperature. Thus, to eliminate the air
temperature effect on SST, the difference between the SSTs and a linear regression between the SSTs and surface air temperatures (SATs) was calculated. This was followed by applying a cross-correlation analysis of
the residual SSTs to determine the main factor driving the SST trend. For
each grid point and variable (i.e. cloudiness, latent heat flux and
During recent decades, the Baltic Sea region has warmed faster than either
the global mean warming (Rutgersson et al., 2015; Kniebusch et al., 2019) or
any other coastal sea (Belkin, 2009), making it prone to marine heat waves (MHWs). Indeed, short periods of abnormally high water temperatures have been documented for the Baltic Sea (Suursaar, 2020). MHWs can be defined
with reference to the mean climatology (e.g. the 90th, 95th, 98th percentile
temperatures) or by temperatures exceeding absolute temperature thresholds,
defined with respect to the end-user application (Hobday et al., 2018). In
most cases, MHWs are defined by the number of periods, their intensity,
their duration and the specific purpose (Hobday et al., 2018). In this study, the focus was on the general impact of climate change and the sensitivity of ecosystem dynamics. Hence, MHWs are defined herein as periods of SST
Upper panels: annual and seasonal mean sea surface temperature (SST, in
The climate of the Baltic Sea region varies considerably, due to maritime and continental weather regimes. For the period 1970–1999, the annual mean SST was
On average, during the period 1976–2005, the climate in the CLIMSEA
simulations is warmer than the climate according to the reanalysis data
(Fig. 4). During spring and summer, the shallow coastal zone of the northern
and eastern Baltic Sea is too warm. The spatially averaged biases during winter, spring, summer and autumn and in the annual mean are 0.8, 0.9, 0.8,
1.0 and 0.9
In the ECOSUPPORT scenario simulations, there is also a systematic warm bias of the RCAO driven by GCMs at the lateral boundaries, such that winter water temperatures are too warm and sea-ice cover is too low (Meier et al., 2011d, 2012c, d). While these biases occur in all three applied Baltic Sea models (Table 3) forced with RCSM atmospheric surface fields, in the simulations driven by regionalised reanalysis data (ERA-40), the mean biases are smaller (Eilola et al., 2011).
Mixed-layer thickness calculated according to the criterion following de Boyer Montégut et al. (2004).
Figure 5 shows the seasonal MLD cycle calculated after de Boyer Montégut
et al. (2004). A deeper MLD with pronounced west–east gradients characterises the open ocean. This is related to the predominant southwesterly wind regime, with the larger wind fetches and higher significant wave heights in the eastern Gotland Basin causing wave-induced vertical mixing. Furthermore, a positive sea–atmosphere temperature contrast favours higher wind speeds (“positive winter thermal feedback loop”; Gröger et al., 2015, 2021b). In spring, a weakening wind regime, which reduces heat exchange (with a shift from heat loss to heat gain), together with the increased solar irradiance leads to a thinner MLD in the southern Baltic Sea, while melting sea ice and subsequent thermal convection and wind-induced mixing maintain a MLD
The ensemble model mean in CLIMSEA reproduces these dynamics and the spatial pattern relatively well. During the cold season, however, the MLD is somewhat shallower in the simulation than in the reanalysis data of Liu et al. (2017). This may be the result of air–sea coupling. Gröger et al. (2015, 2021b) demonstrated that the complex thermal air–sea feedbacks in winter are less well represented by stand-alone ocean models than by fully coupled ocean–atmosphere GCMs. This can result in SST biases and too-shallow MLDs (Gröger et al., 2015; Fig. 7a therein; Gröger et al., 2021b). However, the real reasons for the underestimated winter MLD are unknown.
In the literature, MLDs in the ECOSUPPORT scenario simulations have not been analysed.
Baltic Sea MHWs are defined herein as periods of
The first MHW index uses a fixed threshold that emphasises the environmental impact of the heat waves. In particular, diazotrophic nitrogen fixation becomes effective at higher temperatures. The spatial pattern of MHWs is strongly related to the simulated SST. Figure 6a shows that MHWs are mostly absent in the open sea of the Baltic Proper and further north in the Gulf of Bothnia, but they are highly abundant in shallow marginal bays such as the Gulf of Finland and Gulf of Riga as well as along the coasts. The MHWs produced by the RCO ensemble mean are generally more frequent and of longer duration than those of the reanalysis data set. Furthermore, the coastal signature of high abundance extends further offshore (Fig. 6a). For the Belt Sea and Bay of Lübeck, this leads to considerable deviations from the reanalysis data set.
The second index is based on a reference climatology, here defined as that of 1976–1999. The number of MHWs (Fig. 6c) correlates negatively with their average duration (Fig. 6d). This is somewhat more pronounced in the reanalysis data set. In general, the patterns obtained with the reanalysis data and the RCO are similar but the amplitude of spatial variance is higher in the former (Fig. 6c), perhaps as it includes small-scale regional observations. In the RCO (Fig. 6d), MHWs in the open sea are of the longest duration, with their interruption likely due to the vertical mixing induced by wind events.
Since MHWs in the Baltic Sea are predominantly a summer phenomenon, the stability of the seasonal thermocline is likely a key element in their dynamics such that processes related to vertical mixing can be considered a benchmark in their simulation by the models. Given that mixing is highly parameterised in current ocean models, the RCO reproduces the spatial patterns of the number and average duration of MHW reasonably well.
In the literature, MHWs in the ECOSUPPORT scenario simulations have not been analysed.
The annual mean sea surface salinity (SSS) distribution shows a large
north–south gradient mirroring both the input of freshwater from rivers, mostly located in the northern catchment area, and saltwater inflows from the North Sea (Fig. 7). The SSS drops from about 20 g kg
Upper panels: annual mean sea surface salinity (SSS) and bottom
salinity (BS) (in g kg
Probably due to differences in the data of the hydrological model (E-HYPE)
compared to observations, SSS in the coastal zone and the Kattegat is on
average lower in the CLIMSEA climate models than in the reanalysis data of
Liu et al. (2017) (Fig. 7). The spatially averaged, annual mean bias is
In the ECOSUPPORT scenario simulations, SSS is overestimated in the entire Baltic Sea, in particular in its northern and eastern regions (Meier et al., 2011c, 2012c). In both, the ensemble mean bottom salinity and vertical stratification are also overestimated, while the bottom salinity in the eastern Gotland Basin is well reproduced (Meier et al., 2012c).
Due to the seasonal cycle in wind speed, with wind directions predominantly
from the southwest, the sea level in the Baltic Sea varies considerably
throughout the year, with the highest levels (
The differences in the mean sea level between the CLIMSEA climate models and
the reanalysis data are small (Fig. 7), and the spatially averaged winter mean bias is only
Monthly mean sea level according to a hindcast (driven by a regionalised reanalysis of atmospheric surface fields, i.e. RCA4 driven by ERA-40; hindcast 388), reanalysis with the data assimilation of Liu et al. (2017) (hindcast 888) and four climate simulations following Saraiva et al. (2019a) (Run_A001, …, Run_D001), the ensemble mean and observations for the historical period 1976–2005 at the sea level stations Klagshamn, Landsort, Hamina and Kalix (for the locations, see Fig. 1). The 95 % confidence interval of the observations is shown as a grey-shaded area.
In the ECOSUPPORT scenario simulations, sea levels were not systematically analysed. In one of the three models (RCO-SCOBI), seasonal mean biases were comparable to the biases in the CLIMSEA scenario simulations (Meier et al., 2011a).
Since the 1950s, nutrient inputs into the Baltic Sea have increased due to population growth and intensified fertiliser use in agriculture (Gustafsson et al., 2012; Fig. 3). Nutrient inputs reached their peak in the 1980s but have steadily declined following the implementation of nutrient input abatement strategies. Nonetheless, since the 1960s, the bottom water of the Baltic Sea below the permanent halocline has been characterised by oxygen depletion and large-scale hypoxia (Fig. 9).
Upper panels: summer (June–August) mean bottom dissolved oxygen (DO) concentrations (in mL L
Consistent with the stratification biases in the deeper sub-basins of the
Baltic Sea, summer bottom oxygen concentrations in the Bornholm Basin are
higher and those in the Gotland Basin lower in the CLIMSEA/BalticAPP climate
simulations than in the reanalysis data of Liu et al. (2017) (Fig. 9). The
stronger vertical stratification, especially at the halocline depth, hampers
vertical fluxes of oxygen, causing prolonged residence times and lower bottom oxygen concentrations. Spatially averaged biases during winter, spring, summer and autumn and in the annual mean are small but systematic:
In the ECOSUPPORT scenarios, the ensemble mean deep-water oxygen concentration in the eastern Gotland Basin is slightly higher (but within the range of natural variability) and that in the Gulf of Finland significantly lower than determined from observations (Meier et al., 2011b, 2012d).
Nutrient (i.e. phosphorus and nitrogen) content in the surface layer during winter is a good indicator of the intensity of the following spring bloom. Sea surface mean winter concentrations of phosphate and nitrate are highest in the coastal zone, in particular close to the mouths of the large rivers in the southern Baltic Sea that transport elevated inputs of nutrients into the sea (Fig. 9).
For the historical period of 1976–1999, winter surface phosphate concentrations according to the climate simulations are relatively close to
those of the reanalysis data (Fig. 9). The concentrations differ substantially only in those coastal regions influenced by large rivers, such
as those affected by discharges of the Oder, Vistula and Pärnu rivers. Spatially averaged biases are largest during summer and autumn, with an average bias in summer of
Likewise, winter surface nitrate concentrations in the simulations are close
to those in the reanalysis data but in coastal regions they differ due to
differences in the inputs from large rivers (Fig. 9). This is exemplified by
the Gulf of Riga and the eastern Gulf of Finland, where the large differences between them are due to inputs from the Neva River. Spatially averaged biases during winter, spring, summer, autumn and in the annual mean are rather small but systematic:
In the ECOSUPPORT scenario simulations, the simulated profiles of phosphate, nitrate and ammonium are within the range of observations for 1978–2007, except in the case of phosphate in the Gulf of Finland (Meier et al., 2012d). According to hindcast simulations, the biases in the coupled physical–biogeochemical models of the Baltic Sea relative to the standard deviations of observations are larger for the northern Baltic Sea than for the Baltic Proper (Eilola et al., 2011).
During the period 1976–1999, dense phytoplankton blooms were confined to the coastal zone, i.e. the area with the highest nutrient concentrations (Fig. 10). Water transparency, measured by Secchi depth, is lower in the
Baltic Sea than in the open ocean (Fleming-Lehtinen and Laamanen, 2012), and
for the period 1970–1999 the annual mean Secchi depth averaged for the
entire Baltic Sea, including the Kattegat, was only
Upper panels: annual mean phytoplankton concentrations (CHL; in mg Chl m
Due to nutrient concentration biases, the annual mean surface phytoplankton
concentrations of the simulations are close to those of the reanalysis data
of Liu et al. (2017) but they deviate in coastal regions (Fig. 10). Spatially averaged biases during winter, spring, summer and autumn and in the annual mean are relatively small:
Similar results are found for the mean biases in the simulated Secchi depths
(Fig. 10). In climate simulations, Secchi depths are systematically deeper
in the regions south of Gotland and at the entrance to the Gulf of Finland
(northeastern Gotland Basin) than elsewhere in the Baltic Sea. Spatially
averaged biases during winter, spring, summer and autumn and in the annual
mean are
Compared to the Secchi depth data from HELCOM (HELCOM, 2013a; their Table 4.3) and Savchuk et al. (2006; their Table 3), the CLIMSEA climate simulations under- and overestimate the Secchi depth in the southwestern and northern Baltic Sea, respectively, while in the Gotland Basin the model results well fit the observations (Meier et al., 2019a).
In the ECOSUPPORT scenario simulations, Secchi depth was not compared with observations.
An evaluation of biogeochemical fluxes, such as primary production and nitrogen fixation, is difficult because observations are lacking. An exception is the study by Hieronymus et al. (2021), in which historical simulations with RCO-SCOBI were compared with in situ observations of nitrogen fixation. The RCO-SCOBI model includes a cyanobacteria life cycle (CLC) model (Hense and Beckmann, 2006, 2010) driven by reconstructed atmospheric and hydrological data. The authors found a satisfactory agreement, with the results mainly within the uncertainty range of the observations. However, simulated monthly mean nitrogen fixation during 1999–2008 showed a prolonged peak period in July and August, whereas according to observations the peak was mostly confined to July. It should be noted that the RCO-SCOBI version used in the scenario simulations discussed here (e.g. Saraiva et al., 2019a) does not contain a CLC model.
Ensemble mean changes in sea surface temperature (SST; in
In Figs. 11–12 and Table 7, annual and seasonal mean SST changes
between 1976–2005 and 2069–2098 in RCO-SCOBI are depicted and quantified
respectively. The maximum seasonal warming signal propagates between winter
and summer from the Gulf of Finland via the Bothnian Sea into the Bothnian
Bay (Fig. 11). Maximum warming occurs during summer in the Bothnian Sea and
Bothnian Bay. The seasonal patterns of RCP4.5 and RCP8.5 are similar although warming is greater in the latter. As SLR has almost no impact on SST changes, BalticAPP and CLIMSEA scenario simulations yield similar results (not shown). The warming level according to ECOSUPPORT is between that predicted by CLIMSEA/BalticAPP RCP4.5 and RCP8.5 because the GHG emissions of the A1B scenario, which forces the ECOSUPPORT ensemble One of the scenario simulations of ECOSUPPORT is driven by the A2 scenario, which due to higher GHG emissions is generally warmer than the A1B scenario. However, this particular simulation of the ECHAM5/MPIOM GCM is exceptional and at the end of the 21st century the temperature is not much warmer than that obtained with the corresponding run based on the same model under the A1B scenario.
Changes in seasonal mean SST as simulated by the CLIMSEA ensemble. From left to right, mean SST changes (in
From left to right, changes in the mean SST (
In the CLIMSEA/BalticAPP RCSM projections, the annual mean SST changes in the Baltic Sea driven by four ESMs, i.e. MPI-ESM-LR, EC-EARTH, IPSL-CM5A-MR, HadGEM2-ES, under the RCP8.5 scenario are
While the spatial patterns of the SST changes in the scenario simulations of ECOSUPPORT (e.g. Meier et al., 2012c) and CLIMSEA (e.g. Saraiva et al., 2019b) are similar, the uncertainties due to the applied global (Meier et al., 2011a) or regional (Meier et al., 2012b) model are in some cases considerable. Of note is the summer ensemble range of the various GCMs (Meier et al., 2011a). The strong dependence on forcing is seen by comparing the different warming levels in the RCP4.5 and RCP8.5 scenarios shown in Fig. 12.
Since SLR and nutrient input scenarios have a negligible impact on SST changes, only the RCP4.5 and RCP8.5 scenarios in CLIMSEA/BalticAPP are compared. The multi-model mean of the annual mean SST trends averaged over
the Baltic Sea is
Multi-model mean (MMM) of annual
As seen in Fig. 14, the relative SST trends indicate faster warming of the
northern than the southern Baltic Sea (0.02 and 0.04
MMM of the annual SST trends relative to the
spatial average (in
At an annual timescale, the variability in the air temperature, through the sensible heat fluxes, is the main driver of the Baltic Sea's SST (Kniebusch et al., 2019), illustrated here by the high variance of SST explained by air temperature (between 0.85 and 0.95, Fig. 15). The minimum of variance explained is located in the Bothnian Bay, where the sea-ice cover isolates seawater from the air in winter.
MMM explained variance (in percent) between the monthly mean SST and the forcing air temperature over the period 2006–2099 in the
The processes responsible for the SST trends were analysed using a rank
analysis of atmospheric variables (i.e. latent heat fluxes, cloud cover and
Results of the cross-correlation analysis of the detrended SST
(monthly mean) with the wind components, latent heat flux and cloudiness.
Maps of the atmospheric drivers with the highest cross-correlations in the
RCP4.5
In the vertical, temperature trends are larger in the surface layer than in the winter water of the Baltic Sea above the halocline, thus causing a more intense seasonal thermocline (see Sect. 3.2.2). Surface layer trends are largest in spring and summer (not shown). Elevated trends also characterise deep water, due to the influence of saltwater inflows that will be warmer in a future climate because they originate from the shallow entrance area and occur mainly in winter. Hence, in sub-basins that are sporadically ventilated by lateral saltwater inflows, such as the Bornholm Basin and the Gotland Basin, the deep water below the halocline will warm more than the overlaying intermediate layer water.
In the literature, trends in ECOSUPPORT scenario simulations have not been analysed.
Figure 17 shows the changes in the MLD. During winter, reduced sea-ice cover
in the Bothnian Sea and Bothnian Bay favours a widespread deepening of their
MLDs, likely caused by wind-induced mixing. In spring, the most pronounced
feature is a strong shallowing of the MLD in the Bothnian Sea, probably
attributable to the radiative fluxes that warm the surface layer and to less
thermal convection (Hordoir and Meier, 2012). During the historical period,
water temperatures in this area were between 2.0 and 3.0
Mixed-layer depth calculated according to the criterion of de Boyer Montégut et al. (2004). Shown are the ensemble average changes of four different ESMs between 1976–2005 and 2069–2098 with the mean sea
level rises
The changes during summer are less pronounced. In contrast to winter, there is an overall shallowing in the entire Baltic Sea. This is in agreement with a shallower, more intense thermocline in warming scenarios, as suggested by Gröger et al. (2019), and it is a common feature among the projections, because the changes in wind speed are small (Christensen et al., 2021). Autumn is primarily characterised by a prolongation of the thermal stratification, leading to an overall shallower MLD than during the historical period.
While Hordoir et al. (2018, 2019) speculated that these changes in thermocline depth during summer will impact the vertical overturning circulation, the meridional overturning circulation in the Baltic Proper does not show a clear signal but rather a northward expansion of the main overturning cell (Gröger et al., 2019). Indeed, the effect is expected to be small (Placke et al., 2021).
The number of MHWs within climatological 30-year time slices is shown in
Fig. 18. Under historical climate conditions, MHWs are virtually absent in
open ocean areas. They are most frequent in shallow regions and more abundant along the eastern (Baltic states) than the western (Swedish) coasts, which may reflect the greater frequency of coastal upwelling events along the western than the eastern coasts of the Baltic Sea. Even under the RCP4.5 scenario, wide areas of the Baltic Proper are affected by MHWs roughly once a year. The strongest response is projected for the high-emission RCP8.5 scenario and specifically in marginal basins such as the Gulf of Riga and the Gulf of Finland, where in the future MHWs will occur two or three times per year. Not only the frequency but also the average duration of the MHWs will increase with climate warming. Under RCP8.5, MHWs of
MHWs can also be analysed by calculating them with respect to the 95th percentile temperature of the historical reference climate (Fig. 19). For the historical climate, the average duration of MHWs in most regions is
Ensemble mean changes in annual mean sea surface salinity (SSS; in g kg
In the CLIMSEA ensemble, salinity changes are not robust; i.e. the ensemble
spread is larger than the signal (Meier et al., 2021). Under both RCP4.5 and
RCP8.5, the ensemble mean salinity change is small because the impact on
salinity of the projected increase in total river runoff from the entire
catchment (Fig. 3) is approximately compensated by the impact of larger
saltwater inflows due to the projected SLR (Table 8). Hence, compared to
previous studies such as those by Meier et al. (2011b; ECOSUPPORT) and
Saraiva et al. (2019a; BalticAPP; Fig. 12), the ensemble mean salinity
changes in CLIMSEA are much smaller (Table 8). In idealised sensitivity
experiments performed with the RCO-SCOBI model for the period 1850–2008
(Meier et al., 2017, 2019d), the change in the average Baltic Sea salinity (1988–2007) increased linearly with SLR and at a rate of
Salinity changes averaged for the Baltic Sea in 1988–2007 relative
to 1850 as a function of sea level rise (SLR). In the reference simulation,
the mean salinity is 7.42 g kg
Following global sea level changes, SLR in the Baltic Sea will accelerate
(Hünicke et al., 2015; Church et al., 2013; Bamber et al., 2019; Oppenheimer et al., 2019; Weisse and Hünicke, 2019), albeit at a somewhat slower rate than the global mean because of the remote impact of the melting Antarctic ice sheet (Grinsted, 2015). Changes in SLR in the North Atlantic (and the Baltic Sea) will be larger in response to the melting of the Antarctic ice sheet than to the melting of Greenland, due to gravitational effects. For a mid-range scenario, SLR in the Baltic Sea is projected to be
In BalticAPP and CLIMSEA scenario simulations, sea level changes are small (Fig. 12, Table 8), whereas in ECOSUPPORT scenario simulations they are larger, particularly in spring, because one member of the multi-model ensemble considers Archimedes' principle (not shown). Note that the sea level changes shown in Fig. 12 consider only changing river runoff, changing wind and melting sea ice as affecting the sea level according to Archimedes' principle (only in the ECOSUPPORT ensemble); as neither the global mean SLR nor land uplift is included, they have to be added (e.g. Meier, 2006; Meier et al., 2004a).
In CLIMSEA, there are no statistically significant seasonal changes in the
SLR (Fig. 20). In both GHG concentration scenarios, the largest changes are
only about
Same as in Fig. 18 but for heat waves defined as periods of
Monthly mean sea level changes between 1976–2005 and 2069–2098 at Klagshamn, Landsort, Hamina and Kalix (for the locations, see Fig. 1) for RCP4.5
In response to the global mean SLR, the sea level extremes in the Baltic Sea that are rare today will become more common in the future (e.g. Hieronymus and Kalén, 2020). However, changes in sea level extremes relative to the mean sea level will not be statistically significant because wind velocities are projected to remain unchanged (Christensen et al., 2021). The exceptions are areas with sea-ice decline since they are linked to a decrease in atmospheric stability accompanied by increased wind velocities, the result of increases in temperature and turbulent fluxes (Meier et al., 2011c). These increases will mostly translate as changes from calm to light wind conditions as the stable atmospheric boundary layer becomes less stable. For stronger wind conditions related to high sea level extremes, the impact of stratification effects on mixing is small. In addition, open water areas after sea-ice loss have a smaller surface roughness than ice-covered areas, with the reduced surface friction leading to an increase in wind velocities.
As in Table 8 but showing the ensemble mean changes in the summer
mean bottom oxygen concentration (in mL L
As sea level extremes also depend on the path of low-pressure systems over the Baltic Sea area (Lehmann et al., 2011; Suursaar and Sooäär, 2007), which in a future climate do not show systematic changes (Christensen et al., 2021), changes in sea level extremes relative to the mean sea level are not expected. In addition, a large internal variability at low frequencies prevents the detection of climate-warming-related changes in sea level extremes (Lang and Mikolajewicz, 2019).
Projected changes in bottom oxygen concentrations differ considerably between ECOSUPPORT and BalticAPP/CLIMSEA scenario simulations, as illustrated for summer (Figs. 21 and 22, Table 10), whereas the differences between BalticAPP (SLR
Ensemble mean changes in the bottom dissolved oxygen
concentration (mL L
As in Fig. 21 but for CLIMSEA RCP4.5 (upper panels) and CLIMSEA
RCP8.5 (lower panels) under a high sea level rise scenario, i.e. 1.26 m
(RCP4.5) and 2.34 m (RCP8.5). Left and right columns show the BSAP and REF
scenarios, respectively. (Source: Meier et al., 2021; their Fig. 5 distributed under the terms of the Creative Commons CC-BY 4.0 License,
Most of the differences in the oxygen concentration changes between the ECOSUPPORT and BalticAPP/CLIMSEA ensembles can be explained as follows. In ECOSUPPORT, changes in nutrient input relative to the historical period 1961–2006, including the observed nutrient inputs averaged from the period 1995–2002, were applied (Gustafsson et al., 2011; Meier et al., 2011b). For the historical period 1980–2002, these inputs were lower than in BalticAPP/CLIMSEA scenario simulations because in the latter the observed monthly nutrient inputs, including the pronounced decline from the peak in the 1980s until the much lower recent values, were used as the forcing (Meier et al., 2018a). Furthermore, in ECOSUPPORT, future nutrient inputs under the BSAP scenario were calculated as relative changes, resulting in higher future inputs than in BalticAPP/CLIMSEA, in which absolute values of the BSAP were applied.
Hence, the reductions between future and historical nutrient inputs are smaller in ECOSUPPORT under the BSAP than in BalticAPP/CLIMSEA (Table 6) and result in a smaller response of biogeochemical cycling. We argue that the more realistic historical simulation, including a spin-up since 1850, based on observed or reconstructed nutrient inputs as used in the BalticAPP and CLIMSEA ensembles result in a model response that is more realistic than that of the ECOSUPPORT scenario simulations.
In ECOSUPPORT, the hypoxic area is projected to increase under REF and BAU nutrient input scenarios (Meier et al., 2011b). Only under BSAP is there a slight decrease compared to the early 2000s.
In CLIMSEA under REF, the hypoxic area is projected to decrease slightly until about 2050, as a delayed response to nutrient input reductions, and then increase again towards the end of the century, presumably in response to increased nutrient inputs and warming (Fig. 23). Larger hypoxic areas are calculated under RCP8.5 than under RCP4.5. Under BSAP, the hypoxic area is projected to considerably decrease. At the end of the century, the size of the hypoxic area is expected to be 22 %–78 % smaller than the average size during the period 1976–2005. This range represents the results of the various scenario simulations.
From top to bottom: hypoxic area (in km
In accordance with previous studies, such as Saraiva et al. (2019b) and Meier et al. (2021), the impact of warming (reduced oxygen solubility, increased internal nutrient cycling, increased riverine inputs) and of increasing stratification (decreased ventilation) will be an amplified depletion of oxygen that enlarges the hypoxia area in the Baltic Sea and partially counteracts nutrient input abatement strategies such as the BSAP. However, in all available scenarios, the impact of climate change is smaller than the impact of nutrient input changes.
As in Fig. 21 but for annual mean surface phytoplankton concentration changes (mg Chl m
While in ECOSUPPORT scenario simulations of future climate the projected surface phosphate concentrations in winter increase under all three nutrient input scenarios (except in the Gulf of Finland in BSAP), in BalticAPP projections the surface phosphate concentrations in winter decrease almost everywhere (except in the Oder Bight and adjacent areas in REF and WORST) (not shown). In contrast to the nearly ubiquitous changes in the surface phosphate concentration, larger nitrate concentration changes are usually confined to the coastal zone and differ in their signs. In ECOSUPPORT projections, the increases in winter surface nitrate concentrations in REF and BAU are largest in the Gulf of Riga, the eastern Gulf of Finland and along the eastern coasts of the Baltic Proper (not shown). In BalticAPP projections, the increases in winter surface nitrate concentrations in REF and WORST are largest in the Bothnian Bay and the Odra Bight, while in the Gulf of Riga and the Vistula lagoon nitrate concentrations decrease. Overall, the differences in surface nutrient concentrations between the two sets of scenario simulations are considerable (not shown) and can be explained by the large differences in nutrient inputs from land. Thus, while the projected changes in inputs in ECOSUPPORT refer to the average inputs during 1995–2002, in BalticAPP scenario simulations the observed historical changes include a decline in nutrient inputs since the 1980s (Meier et al., 2018a).
As in Fig. 21 but for changes in the annual mean Secchi depth (m). (Source: Meier et al., 2011b; Saraiva et al., 2019a.)
Annual mean changes in surface phytoplankton concentration (expressed as chlorophyll concentration) follow the changes in nutrient concentrations and are confined to the productive zone along the coasts (Fig. 24). In ECOSUPPORT projections, annual mean Secchi depths decrease in all scenario simulations (see Fig. 25 and Table 11). In the BalticAPP projections, the area-averaged Secchi depths generally increase, except in the combined RCP8.5 and BAU scenario (Table 11), indicating a general improvement of the water quality in future compared to the present climate. The most striking changes occur in the BSAP scenario, in which the Secchi depth increases by up to 2 m in the coastal zone of the eastern Baltic Proper. Changes in stratification (illustrated by the differences between BalticAPP and CLIMSEA ensembles and between the CLIMSEA ensemble mean and high SLR scenarios) have only a minor impact on water transparency (Table 11). The overwhelming driver of the changes in the Secchi depth is the nutrient input scenario (illustrated by the differences between ECOSUPPORT and BalticAPP/CLIMSEA ensembles and highlighted by, in some cases, contradictory signs in the changes).
As in Table 8 but showing the ensemble mean changes in the annual Secchi depth (in m) in the ECOSUPPORT, BalticAPP RCP4.5, BalticAPP RCP8.5, CLIMSEA RCP4.5 and CLIMSEA RCP8.5 scenario simulations averaged for the Baltic Sea including the Kattegat. The projected changes depend on the nutrient input scenario: BSAP, REF, BAU or WORST. (Data sources: Meier et al., 2011b, 2021; Saraiva et al., 2019a.)
Ensemble 10-year running mean North Atlantic Oscillation (NAO)
index
In CLIMSEA under the BSAP, primary production and nitrogen fixation are
projected to considerably decrease in a future climate (Fig. 23). According
to this scenario, the interannual variability declines. Under REF, nitrogen
fixation is projected to slightly decrease until
The dominant large-scale atmospheric pattern controlling the climate in the Baltic Sea region during winter is the North Atlantic Oscillation (NAO; Hurrell, 1995). However, its influence is not stationary but depends on other modes of variability, such as the Atlantic Multidecadal Oscillation (AMO; Börgel et al., 2020). During the past climate, the relationship between the NAO index and regional climate variables in the Baltic Sea region, such as SST, changed over time (Vihma and Haapala, 2009; Omstedt and Chen, 2001; Hünicke and Zorita, 2006; Chen and Hellström, 1999; Meier and Kauker, 2002; Beranová and Huth, 2008).
Figure 26 shows the calculated ensemble mean winter (December–February) NAO index for the period 2006–2100. For the RCP4.5 emission scenario, the NAO shows high interannual variability. Following a wavelet analysis, the calculated NAO index exhibits decadal variability, which differs for every model (not shown). A comparison of RCP4.5 with the high-emission scenario RCP8.5 shows that the spread of the ensemble increases with increasing GHG concentrations. Figure 26 also depicts the running correlation between the NAO index and the area-averaged SST. The correlation remains positive but it is not constant in time. Also evident from a comparison of RCP4.5 and RCP8.5 is that there are no systematic changes in the two emission scenarios, although for RCP8.5 the ensemble spread is slightly larger.
In the largest set of scenario simulations of this study, the CLIMSEA ensemble, only four ESMs were regionalised using only one RCSM; consequently, this ensemble is still too small to estimate the uncertainties caused by ESM and RCSM differences. While nine ESMs with the same RCSM were recently regionalised, they did not include running modules for terrestrial and marine biogeochemistry (Gröger et al., 2021b), such that these simulations were not considered in our assessment. The uncertainties related to unresolved physical and biogeochemical processes in the Baltic Sea and on land were also not considered, because only one Baltic Sea and one catchment model were used. Although the CLIMSEA ensemble is larger than the ensembles in previous studies, it is still too small to estimate all sources of uncertainty.
In addition to the uncertainties related to global and regional climate and impact models, pathways of GHG and nutrient emissions are thus far unknown and the role of natural variability versus anthropogenic forcing is not well understood (Meier et al., 2018a, 2019b, 2021). Recent studies suggest that the impact of natural variability, such as the low-frequency AMO, is larger than hitherto estimated. For instance, in palaeoclimate simulations the AMO affected Baltic Sea salinity at timescales of 60–180 years (Börgel et al., 2018), which is longer than the simulation periods of available scenario simulations. Furthermore, the AMO may also influence the centres of action of the NAO (Börgel et al., 2020). Lateral tilting of the positions of the Icelandic Low and Azores High explains the changes in the correlations between the NAO and regional variables such as water temperature, sea-ice cover and river runoff in the Baltic Sea region (Börgel et al., 2020). Despite indications that the AMO is affected by climate states such as the Medieval Climate Anomaly and Little Ice Age (Wang et al., 2017; Börgel et al., 2018), how future warming would affect these modes of climate variability is unclear.
Changes in sea-ice cover were not analysed in this study because in the recent scenario simulations of the CLIMSEA ensemble sea-ice cover is systematically underestimated. However, we found that future sea-ice cover is projected to be considerably reduced, with an on-average ice-free Bothnian Sea and western Gulf of Finland. Recent results by Höglund et al. (2017) confirmed earlier results by Meier (2002b) and Meier et al. (2019d, 2014); see BACC Author Team (2008).
The various scenario simulation sets have in common that plausible nutrient input changes have a bigger impact on changes in biogeochemical variables, such as nutrient, phytoplankton and oxygen concentrations, than of either the projected changes in climate, such as warming, or changes in vertical stratification. The latter would be caused by increased freshwater inputs, SLR or changes in regional wind fields, assuming RCP4.5 or RCP8.5 scenarios. Long-term simulations of past climate support these results. Although historical warming had an impact on the size of the present-day hypoxic area, model results suggest that hypoxia in the Baltic Sea is best explained by the increases in nutrient inputs due to population growth and intensified agriculture since 1950 (Gustafsson et al., 2012; Carstensen et al., 2014; Meier et al., 2012a, 2019c, d). Hypoxia is also a feature of the Medieval Climate Anomaly (Zillén and Conley, 2010). However, a preliminary attempt to simulate the past 1000 years could not explain the low-oxygen conditions without substantial increases in nutrient inputs (Schimanke et al., 2012). Thus, the sensitivity of state-of-the-art physical–biogeochemical models to various drivers can be questioned and it is clear that the models do not reproduce all important processes.
As outlined in previous assessments, current and future bioavailable nutrient inputs from land and atmosphere are unknown and were consequently classified as one of the largest uncertainties (Meier et al., 2019b). For a more detailed discussion of uncertainties in Baltic Sea projections, the reader is referred to Meier et al. (2018a, 2019b, 2021).
As shown in Sect. 3, the latest published scenario simulations confirm the
findings of the first and second assessments of climate change in the Baltic
Sea region (BACC Author Team, 2008; BACC II Author Team, 2015), namely that,
in all projections driven by RCP4.5 and RCP8.5 and by four selected ESMs of
CMIP5, water temperature is projected to increase and sea-ice cover to decrease significantly. In the two RCP scenarios, the ensemble mean annual
changes in SST between 1978–2007 and 2069–2098 are 2 and 3
The spatial patterns of seasonal SST trends projected for 2006–2099 are similar to those of historical reconstructions of the period 1850–2008, although in most regions the magnitude of the simulated trends is larger. The largest trends are those in summer in the northern Baltic Sea (Bothnian Sea and Bothnian Bay) and thus in regions where under a warmer climate sea ice would melt earlier or disappear completely due to the ice–albedo feedback. This implies that, with increasing warming, SST trends in the northern Baltic Sea will become larger than those in the southern Baltic Sea. Accordingly, in contrast to the present climate, in which mean SSTs considerably decline from south to north, in a future climate the north–south temperature gradient will weaken.
In contrast to previous scenario simulations, recent scenario simulations considered the impact of the global mean SLR on Baltic Sea salinity, which for the ensemble mean salinity would more or less completely compensate for the effects of the projected increasing river runoff. However, as future changes in all three drivers of salinity (wind, runoff and SLR) are highly uncertain, the spread in the salinity projections of the various ESMs is larger than any signal.
In agreement with earlier assessments, we conclude that SLR has a greater potential to increase surge levels in the Baltic Sea than does changing wind speed or direction. For the latter, there have been no statistically significant changes during the 21st century thus far.
In agreement with earlier studies, changes in nutrient input according to the BSAP or REF scenarios will have a larger impact on biogeochemical cycling in the Baltic Sea than will a changing climate driven by RCP4.5 or RCP8.5 scenarios. Furthermore, the impact of climate change will be more pronounced under higher than under lower nutrient conditions. Hence, without further nutrient input reductions, as suggested by the BSAP, eutrophication and oxygen depletion will worsen. However, the response determined in recent studies differs considerably from the responses reported in previous studies, because of more plausible assumptions regarding historical and future nutrient inputs. In some cases, this has led to opposite signs in the response of bottom oxygen concentrations. The new scenarios suggest that implementation of the BSAP would lead to a significant improvement in the ecological status of the Baltic Sea regardless of the applied RCP scenario.
However, recent studies identified SLR as a new global driver. Depending on the combination of SLR and RCP scenarios, the impact on the bottom oxygen concentration may be significant. A higher mean sea level relative to the seabed at the sills would cause increased saltwater inflows, a stronger vertical stratification in the Baltic Sea and a larger hypoxic area. The relationship between vertical stratification and the size of the hypoxic area was confirmed in historical measurements. Nevertheless, recent studies suggest that the difference in future nutrient emissions between the BSAP and REF scenarios is a more important driver than the projected changes in climate with respect to changes in hypoxic area, phytoplankton concentration, water transparency (Secchi depth), primary production and nitrogen fixation.
The currently available ensembles of scenario simulations are larger than those in previous studies. Consequently, the uncertainty range covered by the assessed ESMs and, in turn, the spread of the results are also larger. However, the ensemble size might still be too small and model uncertainty is very likely underestimated. Moreover, natural variability might be a more important source of uncertainty than previously considered for applications in the Baltic Sea.
In the present climate, the climate variability of the Baltic Sea region during winter is dominated by the impact of the NAO. However, in the past, the correlation between the NAO and regional variables such as water temperature or sea ice varied in time. The low-frequency changes in this correlation are projected to continue. Furthermore, systematic changes in the influence of the large-scale atmospheric circulation on regional climate and on the NAO itself could not be detected. While a northward shift in the mean summer position of the westerlies at the end of the 21st century compared to the 20th century was reported (Gröger et al., 2019), it was based upon a limited set of simulations with a few ESMs.
This is a review article and no new data were generated. The sources of data displayed in the figures and tables have been cited. All data are publicly available.
HEMM designed the review article, coordinated the writing and produced the tables. Analyses of publicly available data resulting in additional figures were performed by MG, CD, KS, FB and HEMM. All authors contributed to the writing of the article with text and comments.
The contact author has declared that neither they nor their co-authors have any competing interests.
Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
This article is part of the special issue “The Baltic Earth Assessment Reports (BEAR)”. It is not associated with a conference.
During the preparation of this paper, shortly before its submission, our co-author, Christian Dieterich, passed away (1964–2021). This sad event marked the end of the life of a distinguished oceanographer and climate scientist who made important contributions to climate modelling for the Baltic Sea, North Sea and North Atlantic regions.
This study belongs to the series of Baltic Earth Assessment Reports (BEAR)
of the Baltic Earth Program (Earth System Science for the Baltic Sea Region; Baltic Earth, 2022) and is dedicated to Christian Dieterich. The work was financed by the Copernicus Marine Environment Monitoring Service through the CLIMSEA project (Regionally downscaled climate projections for the Baltic and North seas, CMEMS 66-SE-CALL2: LOT4) and by the Swedish Research Council for Environment, Agricultural Sciences and Spatial Planning (Formas) through the ClimeMarine project within the framework of the National Research Programme for Climate (grant no. 2017-01949). Regional climate scenario simulations were conducted on the Linux clusters Krypton, Bi, Triolith and Tetralith, all operated by the National Supercomputer Centre in Sweden (NSC,
The work was financed by the Copernicus Marine Environment Monitoring Service through the CLIMSEA project (Regionally downscaled climate projections for the Baltic and North seas, CMEMS 66-SE-CALL2: LOT4) and by the Swedish Research Council for Environment, Agricultural Sciences and Spatial Planning (Formas) through the ClimeMarine project within the framework of the National Research Programme for Climate (grant no. 2017-01949). Regional climate scenario simulations were conducted on the Linux clusters Krypton, Bi, Triolith and Tetralith, all operated by the National Supercomputer Centre in Sweden (NSC,
This paper was edited by Marcus Reckermann and reviewed by Vladimir Ryabchenko and Boris Chubarenko.