Comparison of two light attenuation parameterization focusing on timing of spring bloom and primary production in the Baltic Sea

Ecological Modelling 259 (2013) 40–49

Contents lists available at SciVerse ScienceDirect

Ecological Modelling journal homepage: www.elsevier.com/locate/ecolmodel

Comparison of two light attenuation parameterization focusing on timing of spring bloom and primary production in the Baltic Sea Zhenwen Wan a,∗ , Hongsheng Bi b , Jun She a a b

Centre for Ocean and Ice, Danish Meteorological Institute, Lyngbyvej 100, DK-2100 Copenhagen, Denmark Chesapeake Biological Laboratory, University of Maryland Center for Environmental Science, P.O. Box 38, Solomons, MD 20688, USA

a r t i c l e

i n f o

Article history: Received 5 October 2012 Received in revised form 11 March 2013 Accepted 14 March 2013 Available online 23 April 2013 Keywords: Ecosystem model Parameter optimization Light attenuation Baltic Sea Timing of spring bloom Primary production

a b s t r a c t The physical–biogeochemical coupled model HMB–ERGOM is used to investigate the effects of light attenuation on the timing of spring bloom (TSB) in the Baltic Sea. When light attenuation was not included, the predicted TSB was earlier than observed values in shallow areas (<50 m) and the predicted primary production tended to be lower, especially in the open-sea areas. Tuning the value of related parameters could not resolve these two discrepancies simultaneously. In the present study, a new light attenuation parameter was introduced to incorporate the effects of inorganic suspended particulate matter (SPM) using bathymetry depth and vertical turbulent diffusivity. A variable optimal photosynthesis irradiance in ERGOM was replaced with a constant value. The new parameterization led to improvement in three aspects of modeled results: nutrients and chlorophyll concentrations, TSB, and primary production. However, insufficient light utilization and under-estimation of primary production in some coastal regions remain problematic. The present study demonstrates the possibility of examining the potential impacts of inorganic SPM without explicitly coupling a complicated SPM model and highlights the importance of inorganic SPM modulating TSB in shallow areas. The new parameterization could be used to examine spatial variation of TSB in the Baltic Sea. © 2013 Elsevier B.V. All rights reserved.

1. Introduction The 3D physical–biogeochemical coupled model HBM–ERGOM is currently providing operational service for the Baltic Sea (Wan et al., 2011, 2012b). A previous assessment showed that the model could effectively capture the observed general seasonal patterns and vertical distribution of the targeted variables: dissolved inorganic nitrogen (DIN), dissolved inorganic phosphorus (DIP) and chlorophyll a (Chl), but the modeled TSB was earlier than observations in shallow areas (depth <50 m) and the modeled primary production was generally lower than the observed values, especially in open-sea regions (Wan et al., 2012b). Maar et al. (2011) also found that the ERGOM underestimated the primary production in the Baltic Sea. To resolve the discrepancy in TSB, we increased the parameter value of optimal photosynthesis irradiance in ERGOM (unpublished), which led to better estimates for the TSB, but worse performance in the modeled primary production, i.e., much lower than the observed values. Conversely, decreasing the value of the irradiance parameter led to better estimates of primary production, but worse performance in the predicated TSB and nutrient

∗ Corresponding author. Tel.: +45 3915 7284; fax: +45 3915 7300. E-mail address: [email protected] (Z. Wan). 0304-3800/$ – see front matter © 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.ecolmodel.2013.03.010

concentrations, i.e., the predicted TSB was much earlier than the observed values and the predicted nutrient concentrations in winter in coastal areas were also much lower. Therefore, the discrepancies cannot be resolved by parameter optimization, which implied some mechanistic problems with the light attenuation parameterization in ERGOM (Neumann, 2000). Previous model validation conducted by Neumann et al. (2002) did not find the problems with the predicted TSB (too early) and nutrient concentrations in winter (too less) in shallow areas, because their validation mostly focused on three offshore stations in the relatively deep Baltic proper and the observed data for validation were too sparse to resolve the TSB. One of the major differences between shallow coastal waters and deep offshore waters is inorganic suspended particulate matter (SPM), which could be important for photosynthesis process. However, the inorganic SPM was not included in the ERGOM. In fact, it is well noted that the SPM affects the underwater light conditions which in turn play a crucial role in predicting the TSB (Xu et al., 2005; Allen et al., 2007; Arndt et al., 2007; Tian et al., 2009). The addition of a sub-model to simulate the dynamics of SPM explicitly will increase the complexity of the model system, as the SPM model usually include several state variables (Soulsby, 1997; Puls et al., 1997; Pleskachevsky et al., 2005; Gayer et al., 2006). In the present study, we propose a procedure to feature the impact of SPM on light attenuation. In this new procedure, the SPM is a function

Z. Wan et al. / Ecological Modelling 259 (2013) 40–49

41

Fig. 1. Topography of the Baltic Sea (unit: m) and location of time-series observational stations A–R (*).

of bathymetry depth and vertical turbulent diffusivity. Meanwhile, we replaced the less common variable optimal photosynthesis irradiance in the ERGOM (Garrada et al., 1983; Stigebrandt and Wulff, 1987) with a constant value. The objective of this manuscript is to compare the modeled results with and without the new procedure on light attenuation. If the updated model significantly improves the modeled TSB, it will be a useful tool to investigate the spatial and temporal dynamics of spring bloom and nutrients in the Baltic Sea. 2. Models, data and methods 2.1. Physical model The physical model is the HIROMB-BOOS ocean circulation model (HBM) (Berg and Poulsen, 2012). The source code used in this study is tagged as MyOV2. HBM is based on the primitive geophysical fluid dynamics equations for the conservations of volume, momentum, salt and heat. The wind, air pressure, air temperature, humidity, evaporation/precipitation and cloud cover are taken into account in the parameterizations of surface boundary conditions. Water levels of tides and surges and monthly climatology of temperature and salinity are imposed as outer lateral boundary conditions. River runoff is included as an inner lateral boundary condition. The model is set up to cover both the Baltic Sea and the North Sea though. our targeted area is only the Baltic Sea (Fig. 1). The model setup and configuration are the same as Wan et al. (2011). 2.2. Ecosystem model The applied version of ERGOM is similar to the original version by Neumann (2000), Neumann et al. (2002), and Conkright et al. (2002). ERGOM originally adopted Redfield ratio for the phytoplankton stoichiometry. Wan et al. (2011) documented that a non-Redfield ratio is more suitable in the Baltic Sea than the Redfield ratio. Moreover, Wan et al. (2012a) demonstrated that a spatially variable N/P ratio is closer to the real phytoplankton stoichiometry in the Baltic Sea than a fixed non-Redfield ratio. The rest values of model parameters are based on Neumann (2000) with minor changes (Wan et al., 2011).

Fig. 2. Vertical profiles of light limitation efficient from two parameterization of light attenuation defined in Section 2.3. Horizontal axis stands for a typical summer day (July 1), unit: hour; vertical axis stands for depth, unit: m.

2.3. Parameterization of light attenuation ERGOM adopts the relation of light limitation to photosynthesis according to Steel (1962): rL =

I Iopt



exp

1−

I



Iopt

,

(1)

where rL , I, Iopt stands for light limitation, irradiance intensity and optimal photosynthesis irradiance, respectively. Irradiance intensity I depends on surface irradiance Is and light attenuation coefficient k: I = Is exp(−k · z),

(2)

where z is the depth. ERGOM assumes Iopt is adjustable and dependent of Is (Garrada et al., 1983; Stigebrandt and Wulff, 1987): Iopt = max(0.25Is , Imin ),

(3)

where Imin is a constant of minimum optimal photosynthesis irradiance. Comparing the variable Iopt in Eq. (3) with the constant Iopt of Steel (1962), the variable Iopt can turn an overly irradiance depressing photosynthesis to an optimal irradiance. Even though Eq. (3) moves the light condition at the sea surface to favor photosynthesis, it reduces the light utilization in the whole water column as shown in Fig. 2 which compares the profiles of light limitation between two parameterizations on a typical summer day. ERGOM assumes: k = kw + k c · C

(4)

where k is the total light attenuation coefficient, kw is the light attenuation coefficient of pure water, and kc is the percentage of light attenuation attributed to organic SPM, and C is the concentration of organic SPM. Now we try to include the inorganic SPM in Eq. (4). Without introducing too much complexity of SPM modeling, we

42

Z. Wan et al. / Ecological Modelling 259 (2013) 40–49

Fig. 3. Comparison of seasonal evolutions of DIN (panels a–d, mmol m−3 ), DIP (panels e–h, mmol m−3 ) and Chl (panels i–l, mg m−3 ) in surface layer between model Case N (red curves) and Case O (green curves) against observations (black dashed cycles) at four representative Stations D (panels a, e, and i), I (panels b, f, and j), M (panels c, g, and k), P (panels d, h, and l). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of the article.)

try to reflect the main effects of inorganic SPM. The inorganic SPM chiefly comes from the re-suspension of sediments due to strong mixing, as the inorganic SPM from rivers mostly settle down in limited estuarine areas. We assume the re-suspension of inorganic SPM is correlated positively with the turbulence diffusivity but negatively with the bathymetry depth. In doing so, we avoid adding a sub-model to simulate explicitly the dynamics of total SPM, but capture the main effect of inorganic SPM on light attenuation. After trial simulations, we select this relation: D/D0 0 k = kw = k c · ks · (log/ ) 10 − log2

(5)

ks

where is the light attenuation ratio of inorganic SPM,  and  0 are the vertical turbulence diffusivity and its reference value, and D and D0 are the bathymetry depth and its reference value. Based on preliminary runs, the new parameters ks ,  o and D0 are set 0.12 m−1 , 0.005 m2 s−1 and 64 m, respectively. 2.4. Simulation The inorganic SPM can have large effect on TSB but little impact on photosynthesis in summer (Tian et al., 2009), because the concentration of inorganic SPM in summer is negligible except a few

estuarine areas. Thus, the goal to include inorganic SPM is to improve model performance for the predicted TSB in shallow areas (<50 m). Another change is expected for improving the estimated primary production by replacing a variable optimal irradiance to deepen light penetration. The problem in the predicted TSB in shallow areas (<50 m) is less likely related to the variable Iopt , because Is is rarely larger than four times of during spring blooms. Based on Eq. (3), when the surface irradiance Is is smaller than four times of Imin , Iopt equals the constant Imin , otherwise equals which mostly occurs in summer. Because the impacts of two changes are not interacted with each other, we examine the potential impacts of two changes in a single simulation. Two scenarios are simulated: Case N, new parameterization, including the impact of inorganic SPM and using a constant of optimal photosynthesis irradiance (Eqs. (1), (2) and (5)); and Case O, old parameterization, without including the impact of inorganic SPM and using a variable of optimal photosynthesis irradiance (Eqs. (1)–(4)). The simulation period ranges from January 1, 2007 to June 30, 2009. The simulation covers years of 2007 and 2008 so that results can be compared with our previous studies (Wan et al., 2012b). The simulation is limited to 2009 because observations are not available afterwards.

Z. Wan et al. / Ecological Modelling 259 (2013) 40–49

2.5. Data sources The in situ data for Chl, DIN and DIP are downloaded from the database of International Council for the Exploration of the Sea. The data of satellite derived Chl are downloaded from the website (http://marcoast.dmi.dk/chlorophyll.php). The satellite derived data of primary production are downloaded from the website (http://www.science.oregonstate.edu/ocean.productivity/ standard.product.php) which were processed with the method (Behrenfeld and Falkowski, 1997). 2.6. Algorithm to determine TSB and calculate primary production We define the TSB as the peak timing when Chl concentration in the surface layer reaches maxima from February 1 to June 1. The minimum data required for determining TSB are at least 4 observational records during this period. If the data do not meet the requirement, we mark no data for that year. If the data meet the requirement, we use a cubic-spline function to regress the observations. The TSB is determined as the date when Chl reaches the maxima of the regressed curve. The ERGOM (Neumann, 2000) is a nitrogen-based model. We calculate primary production as the amount of DIN up-taken by phytoplankton multiplying a ratio of C:N, which depends on nutrient limitation (Geider et al., 1998; Henriksen et al., 2002). 3. Results 3.1. Seasonal variations of targeted state variables in the surface layer The modeled results from Case N are closer to the observed values in shallow areas (<50 m) than Case O. Specifically, Case N performed better in the predicted TSB (later), nutrient concentrations in winter (higher), and Chl concentration during spring blooms (higher). Four stations (D, I, M, P, see Fig. 1 for the locations) are selected to represent the shallow area in the southern Baltic Sea, the Baltic Proper, the coastal zone in Gulf of Finland, and the Bothnia coast, respectively. In Case N, the spring decline of DIN curve from high values in winter, reflected in Chl, which indicates the TSB, occurs later than in Case O at Stations D, M and P, but no much difference at Station I. Correspondingly, DIN concentration in winter in Case N is higher than in Case O (Fig. 3a–d). For DIP, the change from Case O to Case N is similar to DIN (Fig. 3e–h). In Case N, the TSBs at Station D, M, P are later than in Case O, and the peak values of Chl during spring blooms are also higher and closer the observed values except Station I (Fig. 3i–l). The peak values of Chl in late summer and fall in Case N are lower than in Case O. 3.2. General patterns of targeted state variables The comprehensive model validation scheme (Wan et al., 2011) is used to recover the overall pattern of model performance (Fig. 4). Regarding the seasonal evolution of Chl (Fig. 4 a), the winter concentrations in Case N are lower, the peak values during spring blooms are higher, and the timing of the recovery from winter trough to spring peak in Case N is later than Case O. Regarding DIN, the winter concentrations in Case N are higher than in Case O (Fig. 4b). Regarding DIP, the winter concentrations in Case N are higher than in Case O (Fig. 4c). These changes show that the modeled results in Case N are more consistent with observations. However, the DIN concentrations in other seasons in Case N, especially from March to December in 2008, are worse than in Case O. Regarding the vertical profile of Chl (Fig. 4d), the modeled concentrations in Case N are lower than those of Case O and closer to

43

Table 1 Statistical measures of model-observation comparison using all available observations. Case

DIN DIP Chl

NS

R2

RMSE

15,657 16,419 7916

N

O

N

O

4.09 0.50 2.66

4.00 0.52 2.72

0.33 0.84 0.17

0.27 0.83 0.14

Abbreviations: NS, number of samplers; RMSE, root mean square error (units: DIN – mmol m−3 , DIP – mmol m−3 , Chl – mg m−3 ); R2 , the square of correlation coefficient, i.e., coefficient of determination.

observations in the upper 20 m layer, however, the modeled concentrations in Case N are slightly higher than those of Case O and deviate away observations around the depth 35 m. Regarding the vertical profile of DIN (Fig. 4e), the modeled concentrations in Case N are higher than Case O and closer to observations in the upper 20 m layer, however, the modeled concentrations in Case N are higher than Case O and deviate away observations below 60 m. The vertical profile of modeled DIP in Case N does not differ much from Case O (Fig. 4f). In statistics (Table 1), the root mean square error (RMSE) and the coefficient of correlation (R2 ) show the model results of DIN, DIP and Chl in Case N are closer to observations than that in Case O, except for the RMSE of DIN. This documents that the improvement of Case N is comprehensive, not only at the selected four stations. 3.3. TSB The TSBs at 18 stations (see Fig. 1 for locations) determined with in situ observations and model results are plotted for years 2007–2009 (Fig. 5). The observed TSBs varied among stations and interannually. The maximum delay in TSB occurs from Kattegat (Station B) to the Bothnian Bay (Station R), up to 60 days. The observed TSB occurs mostly around April 7, the standard deviation of TSB among 18 stations is about 17 days in years 2007–2009 (Table 2). The modeled TSB in Case N is consistent with the observed TSB. The modeled TSB in Case N occurs mostly around April 7, the standard deviation of TSB among 18 stations is about 15 days in years 2007–2009. The coefficient of determination (R2 ) of modeled TSB in Case N is 0.33. In Case O, the modeled TSB is earlier than the observed TSB at most of stations. The statistics show that the modeled TSB of Case O occurs mostly around March 28, the standard deviation of TSB among 18 stations is about 12 days in years 2007–2009. R2 of modeled TSB in Case O is just 0.21. All statistical measures in Case N are better than in Case O. Regarding the horizontal distribution, the satellite derived TSB is ∼April 1 from Skagerrak to east of Arkona (Fig. 6a). In the southern Baltic Sea, the TSB starts from Arkona in early March and develops eastwards. The TSB in the southeastern Baltic Sea occurs in late April. From the north of Baltic Proper (59.5◦ N) to the Gotland Deep, the TSB shows a southward development trend. The TSB in Gulf of Finland tends to be the latest. The TSB in the Bothnia Sea/Bay is generally late around the end of April. Table 2 Statistical measures for comparing TSB between two model cases against in situ observations.

Observations Model case N Model case O

NS

Mean (days)

SD (days)

RMSE (days)

R2

47 47 47

6.0 5.9 −4.3

16.9 15.5 12.5

– 14.8 18.7

– 0.33 0.21

Abbreviations: NS, number of samplers; SD, standard deviation; RMSE, root mean square error; R2 , the square of correlation coefficient, i.e. coefficient of determination.

44

Z. Wan et al. / Ecological Modelling 259 (2013) 40–49

Fig. 4. Comprehensive comparison between model Case N (red curves) and Case O (green curves) against observations (black dashed cycles). Panels A–C depict the seasonal pattern of model biases for Chl, DIN and DIP, respectively, and Panels D–F show their vertical profiles (vertical axes for depth, unit: m). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of the article.)

In Case N, the modeled TSB starts from Skagerrak in early March, and successively develops in Kattegat, Arkona, Bornholm and the southeastern Baltic Sea (Fig. 6 b). In the Baltic proper from 56◦ N to 60◦ N, the TSB shows a southeastward developing trend, starting from late March and ending in late April. In the Gulf of Finland, the TSB starts from the central area (around 60◦ N, 26◦ E) in late March and then develops eastwards till early May. In the Bothnia Sea, the TSB starts from the northern area (around 63◦ N, 19◦ E) in mid-March and develops southwards in late April. In the Bothnia Bay, the TSB starts from the central area in early April and develops toward coasts until late April. In Case O, the horizontal distribution of the TSB south of 60◦ N shows a general pattern related to bathymetry depth (Fig. 6 c). TSB

in coastal regions can start in late February. The TSB in the deep area of the Baltic Proper occurs around mid-April. The TSB in Gulf of Finland starts along southern coasts in mid-March and then develops northeastwards around mid-April. The TSB in the Bothnia Sea is generally later than in the Baltic Proper, but early than in the Bothnia Bay. The TSB south of 60◦ N in Case O is generally one or two weeks earlier than in Case N. 3.4. Light utilization We use the vertical profile of modeled DIN to check the change of light utilization. The vertical profiles of DIN from two cases are compared to in situ observations at four representative stations

Z. Wan et al. / Ecological Modelling 259 (2013) 40–49

45

has its annual peak in June–July. The model simulation in Case N is more consistent with the satellite derived data in terms of the timing of the start and the end of the growth season than Case O. As to the horizontal distribution of annual mean primary production, the satellite derived results show that the primary production in coastal regions is generally higher than in offshore regions, and higher in the Baltic Proper than in the Bothenian Sea (Fig. 8b). The highest primary production occurs in the Gulf of Finland and the Gulf of Riga. The modeled result in Case N in the Baltic Proper is also higher than the Bothenian Sea, but does not show a clear correlation with the offshore distance (Fig. 8c). The modeled results in Case N are generally higher in offshore regions (>50 m) than Case O, but lower in coastal regions (<20 m) (Fig. 8d). Compared to the satellite derived primary production, the modeled primary production in Case N is consistent with satellite derived estimates with improvement in offshore regions, but deviation in coastal regions.

4. Discussion 4.1. Light attenuation parameterization

Fig. 5. Comparison of modeled TSB between Case N (red squares) and Case O (green squares) against observations (black cycles). Letters (A–R) at horizontal axis correspond to observation stations (Fig. 1). Panels A–C are for years 2007–2009. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of the article.)

(Fig. 7). For Case N, light utilization is enhanced in deep offshore regions (>50 m), but not much impacted or even slightly deteriorated in coastal regions (<20 m). In the offshore region of average depth (50 m), the depth of modeled light penetration in summer in Case N is 5–10 m deeper than in Case O using the contour of 0.5 mmol m−3 as a reference (Fig. 7e and i). The observed DIN is generally depleted 35–45 m (Fig. 7a), but the modeled DIN depletion depths in both cases are shallower than observations at Station D with a deeper depth for Case N. In deep offshore regions, the depth of light penetration is improved in Case N, close to the observed light penetration (Fig. 7b and f). In the coastal region of the Gulf of Finland, the light penetration in Case N is slightly different the observed values (Fig. 7g and k). In the coastal region of Bothnia Sea, the light penetration is similar between Cases N and O (Fig. 7h and l). 3.5. Primary production The modeled primary productions are comparable to the satellite derived data (Fig. 8). The modeled result in Case N is 50–100 mg C m−2 d−1 higher than in Case O in April–August, but 20–50 mg C m−2 d−1 lower than in Case O in September–March (Fig. 8a). The modeled primary production reaches its annual peak during spring blooms, but the satellite derived primary production

The modeled results from two parameterization procedures are compared by examining the TSB, light utilization, primary production, seasonal dynamics of DIN, DIP and Chl in surface layer and overall performance. The comparison shows that the new parameterization (Case N, Eqs. (1), (2), (5)) is more suitable than the original parameterization of ERGOM (Case O, Eqs. (1)–(4)). If we examine two statistical measures (R2 , RMSE) for three targeted state variables (DIN, DIP, Chl) (Table 1), results in Case N outperform Case O in five out of six statistical measures. Meanwhile, all the statistical measures for the TSB (Table 2) show that the modeled results in Case N are better than Case O. The new parameterization improves the light utilization in offshore regions, e.g. at Station D and I, but with slight deviation in some coastal regions, e.g. at Station M (Fig. 7). Finally, the mean primary production averaged over the entire Baltic Sea in Case N is more consistent with the satellite derived data than Case O (Fig. 8a). Therefore, we believe that the new parameterization is closer to the reality of light attenuation than the original ERGOM parameterization (Neumann, 2000). Since most ecological models use a constant optimal irradiance with respect to phytoplankton functional group (Steel, 1962), why did ERGOM use a variable optimal irradiance dependent of the irradiance below the surface water? ERGOM adopted this parameterization from Stigebrandt and Wulff (1987), who adapted an earlier practice (Garrada et al., 1983) which used a minimum optimal irradiance 25 W m−2 . The variable optimal irradiance (Eq. (3)) has the advantage of keeping optimal light condition in the top 10 m layer only if the light intensity is higher than the minimum optimal irradiance (Fig. 2b). It means no light depression would occur for photosynthesis no matter how strong the irradiance is. However, the variable optimal irradiance could limit light penetration and light utilization, which thus decreases primary production. For example, the light penetration of a constant optimal irradiance is deeper than that of the variable one, even if a relatively high value of constant optimal irradiance 50 W m−2 is used (Fig. 2a). Obviously, if the model validation is focused on the surface layer, it will be difficult to identify the disadvantages for the variable optimal irradiance. The onset of spring bloom had been believed subject to a balance between the light penetration depth and the depth of the mixed layer, i.e. the Critical Depth hypothesis (Sverdrup, 1953). However, the recent progress documented the Critical Depth hypothesis could not be true and that zooplankton grazing could predominantly modulate the timing of phytoplankton blooms (Behrenfeld,

46

Z. Wan et al. / Ecological Modelling 259 (2013) 40–49

Fig. 6. Horizontal distribution of TSB averaged over years 2007–2009, unit: days after April 1. Panels: (A) satellite derived; (B) model case N; (C) model case O.

2010). If the Critical Depth hypothesis is true, the TSB should have a certain relation with the bathymetry depth and the offshore distance, as these two factors generally relate to the mixing depth. However, the observed TSB does not show any clear correlation with the bathymetric depth and the offshore distance (Fig. 6). The complexity of the observed TSB implies the potential influences of other factors. At the end of winter before the onset of spring bloom, Chl and organic detritus are low in water column. Inorganic SPM can play a major role modulating TSB (Xu et al., 2005; Allen et al., 2007; Arndt et al., 2007; Tian et al., 2009). Compared to Case O, the new parameterization improves TSB (Table 2). The significant model improvement of TSB indicates that the new parameterization captures the effect of inorganic SPM on light attenuation. The factors affecting the light attenuation include not only Chl, organic and inorganic SPM, but also chromophoric dissolved organic matter (CDOM) (Kelble et al., 2005; Kowalczuk et al., 2006). In the Baltic Sea, CDOM is related with salinity (Kowalczuk et al., 2006; Stedmon et al., 2010) Kelble et al. (2005) suggests the impact of CDOM to light attenuation is much smaller than the impact from inorganic SPM. However, the parameterization (Eqs. (1), (2), and (5)) ignoring the effect of CDOM likely cause some uncertainty.

The new parameterization (Eqs. (1), (2), and (5)) leads to a general improvement in model performance, but it causes some problems in coastal regions, as indicated by the deviations in the light penetration at Station M (Fig. 7g and k) and the primary production in coastal regions (Fig. 8b and c). The parameterization of inorganic SPM might be effective for estimating TSB, however, it might be unrealistic in coastal regions in other seasons. In principle, a mechanistic model should be able to represent more realistic dynamics of SPM, but a mechanistic model for SPM is likely computationally expensive and requires a detailed forcing condition to achieve more accurate results (Soulsby, 1997; Puls et al., 1997; Pleskachevsky et al., 2005; Gayer et al., 2006). 4.2. TSB in the Baltic Sea Although spring blooms have long been investigated in the Baltic Sea, the distribution of TSB in basin scale has not been fully investigated yet. The feature of TSB development from west to east in the southern Baltic Sea was commonly recognized (Banse, 1956; Kaiser and Schulz, 1978; Kahru and Mömmann, 1990; Wasmund et al., 1998; Fennel, 1999; Neumann, 2000). Kahru and Mömmann (1990) suggested that the spring bloom starts from the more

Z. Wan et al. / Ecological Modelling 259 (2013) 40–49

47

Fig. 7. Comparison of vertical profiles of DIN (mmol m−3 ) between Case N (panels e–h) and Case O (panels i–l) against observations (panels a–d) at stations D (panels a, e, and i), I (panels b, f, and j), M (panels c, g, and k), P (panels d, h, and l).

stratified areas and evolves toward the less stratified central areas. Wasmund et al. (1998) found that the earliest blooms occurred in shallow bights. In general, the spring blooms in the Baltic Sea are delayed from south to north (Kahru and Mömmann, 1990). A precondition triggering the spring bloom is that the gross growth exceeds the gross lose within the mixing layer. Sverdrup (1953) described that the spring bloom initiated when the thickness of the mixing layer was reduced to a critical level. Riley (1967) suggested the spring bloom commenced when the depth-averaged irradiance was increased up to a critical intensity 20.9 W m−2 . Although these thresholds provide ways to distinguish the onset of spring bloom, they cannot explain the complicated pattern of TSB in the Baltic Sea because the mixing/stratification depth in spring is complicated and other factors, e.g. zooplankton grazing (Behrenfeld, 2010), can also play important roles. When the water temperature is lower than the temperature of maximum density, the heating in surface layer leads to vertical convection; only when the water temperature is higher than the temperature of maximum density, the seasonally heating helps to reduce the mixing layer. In fact, the temperature of maximum density varies with salinity. Therefore, temperature, salinity, mixing and heating, zooplankton grazing and inorganic SPM jointly modulate the TSB. Both the model (Case N) and the satellite results capture the commonly recognized feature: the TSB evolves from west to east in the southern Baltic Sea (Fig. 6a and b). It is also generally consistent

in the Baltic proper south of 59◦ N with the finding that the spring bloom progresses from the more stratified areas toward the less stratified central areas (Kahru and Mömmann, 1990). The model also captures the area of the minimum value of the satellite derived TSB between the northern Gotland Sea and south of 60◦ N. In general, the model and the satellite results have good consistency south of 59◦ N and in the Bothnia Bay, but show inconsistency in the Bothnia Sea. 4.3. Primary production in the Baltic Sea Primary production in the Baltic Sea has long been investigated using both measurements (Elmgren, 1984; Osterroht and Thomas, 2000; Wasmund et al., 2001; Struck et al., 2004; Wasmund et al., 2005; Savchuk and Wulff, 2007) and model results (Maar et al., 2011; Wan et al., 2011). Data on primary production are still insufficient to investigate seasonal and spatial features (Wasmund et al., 2005). A few investigations have estimated the annual mean primary production in the Baltic proper. Wasmund et al. (2005) estimated the net primary production in the eastern Gotland Sea was 6454 mmol C m−2 , which equals 211 mg C m−2 d−1 , during the most productive period (March 28, 2001–July 13, 2001) which they believed were close to the annual net primary production. Struck et al. (2004) calculated the annual new production in the

48

Z. Wan et al. / Ecological Modelling 259 (2013) 40–49

Fig. 8. Primary production, unit: mg C m−2 d−1 . (A) Seasonal evolution of global means of satellite derived (blue), model Case N (red) and Case O (green); (B–D) annual means of satellite derived (B), model Case N (C) and Case O (D). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of the article.)

eastern Gotland Sea ranging from 247 to 588 mmol N m−2 yr−1 , i.e., 112 to 266 mg C m−2 d−1 , if the Redfield ratio and the f-ratio 0.48 (Wasmund et al., 2005) were adopted. Two model estimates from Maar et al. (2011) and Wan et al. (2011) range from 110 to 220 mg C m−2 d−1 in the Baltic proper. In the present study, the model estimates range from 120 to 300 mg C m−2 d−1 in the Baltic proper (Fig. 8c) which are consistent with other studies, but lower than the satellite estimates (240–360 mg C m−2 d−1 ; Fig. 8b). The modeled primary production (Case N) is consistent with the satellite derived results on the timing of the starting and ending of growth season and the magnitude of peak value (Fig. 8a). However, the satellite results show the timing of peak value is in midsummer, rather than during spring blooms like the modeled results. Primary production during spring blooms was believed dominant in the annual primary production (Smetacek, 1980; Wasmund et al., 2005). However, there are also observations showing that the peak of annual cycle of primary production appears in summer rather than during spring blooms in the Danish Straits (Lyngsgaard et al., 2012). The annual mean of primary production of model Case N does not show a horizontal distribution pattern decreasing away from coastlines, which is inconsistent with the satellite derived results and the model Case O results. This might be one limitation of new parameterization. 5. Summary This study was motivated to improve the model system which is providing operational biogeochemical service for the

Baltic Sea based on a physical–biogeochemical coupled model HMB–ERGOM. The targeted issues are to resolve the discrepancies for the modeled TSB (too early) in shallow waters and the primary production (generally low). The former is attributed to the lack of the effect of inorganic SPM, while the later was attributed to insufficient light utilization by Wan et al. (2012b). We introduce a new parameterization procedure of light attenuation for the former and correct the latter by using commonly accepted constant optimal photosynthesis irradiance. The new light attenuation parameterization leads to a general improvement of model results in three aspects: nutrients and chlorophyll concentrations, the TSB, and primary production. However, insufficient light utilization and under-predicted primary production in some coastal regions remain problematic. This study highlights the importance of inorganic SPM modulating the TSB in coastal zones and the possibility to reflect the main impact of inorganic SPM without explicitly coupling a complicated SPM model. Acknowledgements We would like to thank Per Berg for technical assistance with the HBM model code and setups, and to thank the Swedish Meteorological and Hydrological Institute and the Bundesamt für Seeschifffahrt und Hydrographie in Hamburg, Germany for providing the river data. This work was supported by European Commission FP7 projects MYCOEAN (Contract No. 218812), OPEC (Contract No. 283291) and MEECE (Contract No. DK18159104).

Z. Wan et al. / Ecological Modelling 259 (2013) 40–49

References Allen, J.I., Holt, T.J., Blackford, J., Proctor, R., 2007. Error quantification of a highresolution coupled hydrodynamic–ecosystem coastal-ocean model: part 2, chlorophyll-a, nutrients and SPM. Journal of Marine Systems 68, 381–404. Arndt, S., Vanderborght, J., Regnier, P., 2007. Diatom growth response to physical forcing in amacrotidal estuary: coupling hydrodynamics, sediment transport, and biogeochemistry. Journal of Geophysical Research 112 (C5), c05045. Banse, K., 1956. Ergebnisse eines hydrographisch-produktionsbiologischen Längsschnitts durch die Ostsee im Sommer II. Verteilung von Sauerstoff, Phosphat und. Suspendierter Substanz. Kiel Meeresforsch 13, 186–201. Behrenfeld, M.J., 2010. Abandoning Sverdrup’s Critical Depth Hypothesis on phytoplankton blooms. Ecology 91, 977–989. Behrenfeld, M.J., Falkowski, P.G., 1997. Photosynthetic rates derived from satellitebased chlorophyll concentration. Limnology and Oceanography 42, 1–20. Berg, P., Poulsen, J.W., 2012. Implementation details for HBM. DMI Technical Report No. 12-11, ISSN: 1399-1388, Copenhagen. Conkright, M.E., Locarnini, R., Garcia, H., O’Brien, T., Boyer, T.P., Stephens, C., Antonov, J., 2002. World ocean atlas 2001: Objective Analyses, Data Statistics and Figures, CDROM Documentation. National Oceanographic Data Center, Silver Spring, MD, p. 17. Elmgren, R., 1984. Trophic dynamics in the enclosed, brackish Baltic Sea. Rapports et Proces-verbaux des Réunions. Conseil International pour l’Éxploration de la Mer 183, 152–169. Fennel, K., 1999. Convection and the timing of phytoplankton spring blooms in the Western Baltic Sea. Estuarine Coastal and Shelf Science 49, 113–128. Garrada, G., Hopkins, T., Jeftie, L., Morcos, S., 1983. Quantitative analysis and simulation of Mediterranean coastal ecosystems: the Gulf of Naples, a case study. UNESCO Reports in Marine Sciences 20, 158. Gayer, G., Dick, S., Pleskachevsky, A., Rosenthal, W., 2006. Numerical modeling of suspended matter transport in the North Sea. Ocean Dynamics 56, 62–77. Geider, R.J., MacIntyre, H.L., Kana, T.M., 1998. A dynamic regulatory model of phytoplanktonic acclimation to light, nutrients, and temperature. Limnology and Oceanography 43 (4), 679–694. Henriksen, P., Riemann, B., Kaas, H., Sorensen, H.M., Sorensen, H.L., 2002. Effects of nutrient-limitation and irradiance on marine phytoplankton pigments. Journal of Plankton Research 24 (9), 835–858. Kaiser, W., Schulz, S., 1978. On the causes for the differences in space and time of the commencement of the phytoplankton bloom in the Baltic. Kiel Meeresforsch Sonderh 4, 161–170. Kelble, R.C., Ortner, B.P., Hitchcock, L.G., Boyer, N.J., 2005. Attenuation of photosynthetically available radiation (PAR) in Florida Bay: potential for light limitation of primary producers. Estuaries 28 (4), 560–571. Kowalczuk, P., Stedmon, A.C., Markager, S., 2006. Modeling absorption by CDOM in the Baltic Sea from season, salinity and chlorophyll. Marine Chemistry 101 (1–2), 1–11. Kahru, M., Mömmann, S., 1990. The phytoplankton spring bloom in the Baltic Sea in 1985, 1986: multitude of spatio-temporal scales. Continental Shelf Research 10 (4), 329–354. Lyngsgaard, M.M., Markager, S., Richardson, K., 2012. Primary production in relation to water column characteristics in the Baltic Sea transition zone: vertical, horizontal and temporal distribution patherns. Unpublished results. Maar, M., Møller, E.F., Larsen, J., Kristine, S.M., Wan, Z., She, J., Jonasson, L., Neumann, T., 2011. Ecosystem modeling across a salinity gradient from the North Sea to the Baltic Sea. Ecological Modelling 222, 1696–1711. Neumann, T., 2000. Towards a 3D-ecosystem model of the Baltic Sea. Journal of Marine Systems 25, 405–419.

49

Neumann, T., Fennel, W., Kremp, C., 2002. Experimental simulations with an ecosystem model of the Baltic Sea: a nutrient load reduction experiment. Global Biogeochemical Cycles 16 (3), 1033–1051. Osterroht, C., Thomas, H., 2000. New production enhanced by nutrient supply from non-Redfield mineralisation of freshly produced organic material. Journal of Marine Systems 25, 33–46. Pleskachevsky, A., Gayer, G., Horstmann, J., Rosenthal, W., 2005. Synergy of satellite remote sensing and numerical modeling for monitoring of suspended particulate matter. Ocean Dynamics 55 (1), 2–9. Puls, W., Pohlmann, T., Suendermann, J., 1997. Suspended particulate matter in the southern North Sea: application of a numerical model to extend NERC North Sea project data interpretation. Deutsche Hydrographische Zeitschrift 49 (2/3), 307–327. Riley, G.A., 1967. The plankton in estuaries. In: Lauff, G.H. (Ed.), Estuaries. American Association for the Advancement of Science Publication 83, Washington, DC, pp. 316–326. Savchuk, O.P., Wulff, F., 2007. Modeling the Baltic Sea eutrophication in a decision support system. AMBIO 36, 141–148. Smetacek, V., 1980. Annual cycle of sedimentation in relation to plankton ecology in western Kiel bight. Ophelia Supplement 1, 65–76. Soulsby, R., 1997. Dynamics of Marine Sands: A Manual for Practical Application. Thomas Telford Services, London, p. 249. Stedmon, A.C., Osburn, L.C., Kragh, T., 2010. Tracing water mass mixing in the Baltic-North Sea transition zone using the optical properties of colored dissolved organic matter. Estuarine Coastal and Shelf Science 87 (1), 156–162. Steel, J.H., 1962. Environmental control of photosynthesis in the sea. Limnology and Oceanography 7, 137–150. Stigebrandt, A., Wulff, F., 1987. A model for the dynamics of nutrients and oxygen in the Baltic proper. Journal of Marine Research 45, 729–759. Struck, U., Pollehne, F., Bauerfeind, E., Von Bodungen, B., 2004. Sources of nitrogen for the vertical particle flux in the Gotland Sea (Baltic Proper)–results from sediment trap studies. Journal of Marine Systems 45, 91–101. Sverdrup, H.U., 1953. On conditions for the vernal blooming of phytoplankton. Journal du Conseil International pour I’Exploration de la Mer 18, 287–295. Tian, T., Merico, A., Su, J., Staneva, J., Wiltshire, K., Wirtz, K., 2009. Importance of resuspended sediment dynamics for the phytoplankton spring bloom in a coastal marine ecosystem. Journal of Sea Research 62, 214–228. Wan, Z., Jonasson, L., Bi, H., 2011. N/P ratio of nutrient uptake in the Baltic Sea. Ocean Science 7, 693–704. Wan, Z., Bi, H., She, J., Maar, M., Jonasson, L., 2012a. Model study on horizontal variability of nutrient N/P ratio in the Baltic Sea and its impacts on primary production. nitrogen fixation and nutrient limitation. Ocean Science Discussion 9, 385–419. Wan, Z., She, J., Maar, M., Jonasson, L., Baasch-Larsen, J., 2012b. Assessment of a physical-biogeochemical coupled model system for operational service in the Baltic Sea. Ocean Science 8, 683–701. Wasmund, N., Nausch, G., Matthäus, W., 1998. Phytoplankton spring blooms in the southern Baltic Sea – spatio-temporal development and long term trends. Journal of Plankton Research 20, 1099–1117. Wasmund, N., Nausch, G., Schneider, B., 2005. Primary production rates calculated by different concepts: an opportunity to study the complex production system in the Baltic Proper. Journal of Sea Research 54, 244–255. Wasmund, N., Voss, M., Lochte, K., 2001. Evidence of nitrogen fixation by nonheterocystous yanobacteria in the Baltic Sea and re-calculation of a budget of nitrogen fixation. Marine Ecology Progress Series 214, 1–14. Xu, J., Hood, R., Chao, S., 2005. A simple empirical optical model for simulating light attenuation variability in a partially mixed estuary. Estuaries and Coasts 28 (4), 572–580.