evaluating the effects of nutrient load reductions on the...

BOREAL ENVIRONMENT RESEARCH 6: ¶–¶ ISSN 1239-6095Helsinki ¶¶ © 2001

Evaluating the effects of nutrient loadreductions on the biomass of toxic nitrogen-fixing cyanobacteria in the Gulf of Finland,Baltic Sea

Mikko Kiirikki1), Arto Inkala2), Harri Kuosa3), Heikki Pitkänen1),Minna Kuusisto1) and Juha Sarkkula1)

1) Finnish Environment Institute, P.O. Box 140, FIN-00251 Helsinki, Finland2) Environmental Impact Assessment Centre of Finland Ltd. (EIA Ltd.),

Tekniikantie 21 B, FIN-02150 Espoo, Finland3) Finnish Institute of Marine Research, P.O. Box 33, FIN-00931 Helsinki,

Finland

Kiirikki, M., Inkala, A., Kuosa, H., Pitkänen, H., Kuusisto, M. & Sarkkula, J.2001. Evaluating the effects of nutrient load reductions on the biomass oftoxic nitrogen-fixing cyanobacteria in the Gulf of Finland, Baltic Sea. BorealEnv. Res. 6: ¶–¶. ISSN 1239-6095

The effects of nutrient load reductions on the biomass of N-fixing cyanobacteria wereevaluated in the scale of the Gulf of Finland. The two analysed reduction scenarioswere Finnish national agenda and the improvement of phosphorus removal in thepresent purification plants of St. Petersburg. The effects of load reduction scenarioswere tested by using a 3D-ecosystem model with a horizontal resolution of 5 km.According to the results the Finnish national agenda cannot decrease the biomass ofN-fixing cyanobacteria, but it seems to be able to reduce the total phytoplanktonbiomass in the coastal waters. The phosphorus reduction in St. Petersburg decreasesthe N-fixing cyanobacteria significantly in the central parts of the Gulf of Finland.Improvement of the phosphorus purification efficiency in the present sewage treat-ment plants offers us an opportunity to try to control the intensity of the toxic bloomsof nitrogen-fixing cyanobacteria in the scale of the Gulf of Finland.

Introduction

During the first week of July 1997, the Gulf ofFinland (GOF) became covered by floating ac-

cumulations of cyanobacteria. Two main generaresponsible for the bloom were Nodularia andAphanizomenon, which both are capable offixing atmospheric nitrogen. Nodularia-species

2 Kiirikki et al. • BOREAL ENV. RES. Vol. 6

are found to be commonly toxic in the GOF(Sivonen et al. 1989). The bloom lasted almost amonth, over the most famous summer holidayseason in July. The whole open sea became sodensely covered by the potentially toxic accu-mulations that swimming as well as many otherrecreational activities were largely hampered.All this happened simultaneously with one ofthe warmest and sunniest summers of the centu-ry. The bloom resulted in a public debate aboutthe ecological state of the GOF and possibleactions to prevent similar blooms in the future.

The blooms of N-fixing cyanobacteria haveprobably been a natural part of the Baltic Seaecosystem for thousands of years (Bianchi et al.2000). One of the first bloom reports originatesfrom the year 1884, when the prince of Monaco,Albert I, was exploring the Baltic Sea. Hedescribed the scenery in the following words:“Algal vegetation covered the whole Baltic Seaspanning from Gotland to Preuss and to theentrance of the Gulf of Finland colouring waterolive green.” (Pouchet and de Guerne 1885).

The growth of these blooms is controlled bythe phosphorus availability (Kononen et al. 1994).The main natural phosphorus sources in theBaltic Sea are in the deep anoxic sediment areas(e.g. Fonselius 1969). The last big bloom in theGOF was preceded by an oxygen deficiency inthe bottom water in late summer 1996 (Pitkänenand Välipakka 1997). Under anoxic conditionsin the sediment–water interface phosphate isreleased from the sediments mostly due to re-duction of oxidised iron compounds. This phe-nomenon is called internal loading and in 1996 itwas estimated to correspond to one-year’s an-thropogenic load of phosphorus from the drain-age area to the GOF.

The main reason for the sudden oxygendeficit was the lack of vertical mixing in winter1995–1996 (Pitkänen and Välipakka 1997). Inthe following throughout mixing in winter 1996–1997, the released phosphorus load entered thesurface layer increasing the average DIP con-centrations by 10%–30% in the whole GOF.When the spring bloom 1997 was over, therewas still plenty of excess DIP available in thesurface water for the growth of N-fixing species.

Even though the hydrographical and meteor-ological factors largely control the outcome of

the blooms of cyanobacteria, anthropogenic eu-trophication of the GOF is supposed to have aclear effect on the intensity of the blooms.Despite some decrease in the nutrient load dur-ing the 1990s, GOF is still receiving a 3 to 5times higher nutrient load than the Baltic Sea onaverage (Pitkänen 1994) and there is no doubtthat the origin of the phosphorus buried in thesediment is mainly anthropogenic. Furthermore,the nutrient load keeps the new biomass produc-tion and sedimentation high (Heiskanen andKononen 1994). Decomposition of this materialconsumes oxygen both in the deeper water lay-ers and on the sediment surface and acceleratesthe formation of anoxia during the periods ofpronounced stratification.

In this paper we aim to evaluate the effectsof nutrient load reduction scenarios on the bio-mass of the potentially toxic nitrogen-fixingcyanobacteria. The task requires developmentand testing of an ecosystem model. A realisticecosystem model for the GOF includes at leastdescriptions of the dynamic water exchangewith the Baltic Proper, oxygen concentrationdependent sediment processes, a detailed pelag-ic food web as well as a high spatial resolutionalong the coastal zone and capability for longmodel simulations. Due to limiting computercapacity and comprehensive knowledge of thekey processes all of these requirements are diffi-cult to reach. Our choice has been to focus onone algal group, the nitrogen-fixing cyanobacte-ria with high spatial resolution and reasonablylong model simulations. Our approach do nottake into account the phenomena caused bywater exchange with the Baltic Proper and chang-es in the oxygen conditions, but in this paper weprove that a realistic cyanobacterial model forthe Gulf of Finland is possible to construct witha less complicated approach.

Material and methods

Load reduction scenarios

The only bloom promoting factor which can becontrolled is the anthropogenic nutrient load.The extensive bloom of summer 1997 speededup the preparation of a Finnish national agenda

3BOREAL ENV. RES. Vol. 6 • Effects of nutrient load reductions

“Water Protection Targets to 2005” (Anony-mous 1998), which aims at cutting 50% of theFinnish anthropogenic load of both N and Pbefore the year 2005. The reference years for theload reduction are set to the beginning of 1990sand therefore the targets have already partlybeen reached. However, the agenda is still esti-mated to cut the GOF entering bioavailable N-discharges by 2100 tonnes y–1 and P-dischargesby 62 tonnes y–1.

The main anthropogenic nutrient source forthe GOF is the Russian city of St. Petersburg withits 4.5 million inhabitants. According to the infor-mation provided by the Russian authorities, onethird of the sewage is lead directly to the GOFwithout any treatment. Two thirds are treatedbiologically. The reduction of biological oxygendemand is good in the sewage treatment process,but only ca 50% of total phosphorus is removed.According to studies on nutrient limitation in theGOF (Kivi et al. 1993, Pitkänen and Tamminen1995) phosphorus seems to be the main limitingnutrient in the easternmost estuarine waters,while the role of nitrogen becomes more impor-tant towards west. However, the potentially toxicblooms of N-fixing cyanobacteria in the middleand western parts of GOF are also controlled byphosphorus (Kononen et al. 1994).

Treatment of the last third of the sewagerequires building of a new large sewage treat-ment plant. The new plant would cut both thenitrogen and phosphorus load to the GOF sig-nificantly. Preliminary plans already exist butthe financing is still open. Meanwhile, morephosphorus could be removed at the presentsewage treatment plants by optimising the bio-logical process (Rantanen 1994) or by usingferrosulphate precipitation. These should be rel-atively cheap procedures which need only minortechnical changes to the present treatment plants.These processes could result in a reduction of570 tonnes y–1 of bioavailable phosphorus. TheFinnish national agenda can only reach roughly10% of this.

Model development

The working principles of our ecosystem modelfollow the oceanic phytoplankton model de-

scribed by Tyrrell (1999). Phytoplankton is de-fined as two competing groups of species: nitro-gen-fixing cyanobacteria and the other phyto-plankton. The other phytoplankton grows fasterand out-competes the N-fixers in all tempera-tures if both DIN and DIP are available. WhenDIN is consumed close to zero but DIP is stillabundant, the N-fixers gain a competitive advan-tage. N-fixation is an energy consuming process,thus the growth of N-fixers is not effective inlow temperature. Their optimum temperature forgrowth varies between 16–28 °C in the GOF(Kononen and Leppänen 1997). The availabilityof trace elements needed for the N-fixing en-zyme complex is not limiting their growth in thebrackish waters of GOF unlike in the openoceans (Falkowski 1997). Both algal groups usenutrients in the model according to the Redfieldratio (Redfield 1958).

All losses of biomass are described by oneloss rate term which is temperature dependent.The maximum loss rate is lower for the nitrogenfixers because they are avoided by grazers dueto their toxins (Sellner 1996). The nutrient andlight limitation of growth rate is calculated ac-cording to the Michaelis–Menten kinetics andtemperature limitations according to Frisk (1982).Light availability is limited by the presence ofice cover. The growth is further limited by shelf-shading when the actual biomass is reaching themaximum biomass describing the carrying ca-pacity of the area. When the actual biomassreaches the minimum biomass, which mimicsthe over-wintering stages of the phytoplanktonspecies, the loss rate approaches to zero and thealgae cannot get extinct during the unfavourableseason.

When the algal cells die, they are convertedinto detritus N and P, which starts to settle at aconstant speed and release DIN and DIP backinto the water column at temperature depend-ent rates. The rate of phosphorus regenerationis roughly twice as fast as nitrogen regenera-tion (Garber 1984). When detritus N and Preach the lowest grid cell of the 3D model, theystart to be converted into sediment at constantrates, permanently out the nutrient cycle. Deni-trification is not described as a separate processbut included to the sedimentation term of detri-tus nitrogen. Ecosystem model equations, vari-


ables and parameters are presented in Tables1–3. The previous version of the ecosystemmodel with only one algal group not capablefor N-fixing has been presented for the GOF byKuusisto et al. (1998).

The ecosystem model has been built on thetop of a 3D-water quality model (Virtanen et al.1986, Koponen et al. 1992). The model is ap-plied to the GOF with a horizontal grid resolu-tion of 5 km and thirteen vertical layers. TheBaltic Proper facing border of the GOF is closedand no inflow is possible. There is only outflowfrom the GOF equal to the flow of the mainrivers Neva, Narva and Kymi.

The model is run by using synoptic windobservation from the coastal weather station ofthe Finnish Meteorological Institute (FMI). Wa-ter temperature data for the ecosystem model arecollected from the intensive monitoring sites ofthe Finnish Environment Institute (FEI) andlight intensity as total radiation data in Helsinkiweather station by FMI. Data on ice cover isprovided by the Finnish Institute of MarineResearch (FIMR). The transport of all variablesis calculated 3-dimensionally, except for thebiotic interactions between dissolved nutrientsand algal biomass in the mixed surface layer.The algal biomass is presented as biomass perarea [g (ww) m–2]. There is no algal growth andno interactions between the dissolved nutrientsand algal biomass below the mixed layer. The

reason for this partly 2D-approach is the lack ofinformation on the depth distribution of algae inthe data used for the model calibration.

Calibration

The data for model calibration were collected atfive intensive monitoring sites by FEI and FIMR(Fig. 1). Biomass [g (ww) m–3] was calculatedfrom chlorophyll a measurements (mg m–3) witha statistical formula (Kuusisto et al. 1998) basedon parallel measurements in the GOF. The origi-nal measurements were combined samples cov-ering the whole euphotic zone defined as twicethe Secchi depth. Thus the biomass concentra-tions (g m–3) were further converted to biomassper area (g m–2) by multiplying them by thedepth of the euphotic zone (m).

The calibration was started with a 0-dimen-sional version of the growth model using thetwo algal groups. All data available from thefive calibration sites (Fig. 1) for 1990s wereused. Biomass data of N-fixing cyanobacteriawas not available and therefore the model wascalibrated by using total phytoplankton biomass.The model was developed with the AQUASIMprogramme version 2.0 (Reichert 1995, Reichertet al. 1995). AQUASIM is a computer pro-gramme for dynamic model development inaquatic systems. It can be used for estimation ofmissing or uncertain parameters in cases wheresufficient material is available. The programmeis able to utilise several independent data sets inthe estimation process.

The estimated parameters were transferred tothe 3D-model and the calibration process wascontinued by using total phytoplankton biomass,surface DIN and DIP data for three years: 1995,1997 and 1998. The years were selected torepresent completely different conditions for thegrowth of cyanobacteria. Year 1995 was the lastyear when practically all phosphorus was usedby the spring bloom of phytoplankton and therewas no marked excess DIP left for the N-fixingcyanobacteria. Year 1997 was the extensivebloom summer when the high DIP-concentra-tions coincided with sunny weather and excep-tionally high surface water temperatures. Year1998 started like the previous year with equally

Table 1. Model variables.————————————————————————Symbol Definition Unit————————————————————————cC Biomass of cyanobacteria

(wet weight) g m–2

cA Biomass of the other algae(wet weight) g m–2

cDIN DIN concentration mg m–3

cDIP DIP concentration mg m–3

cNdet Detritus nitrogen mg m–3

cPdet Detritus phosphorus mg m–3

LDIN DIN load mg m–3 d–1

LDIP DIP load mg m–3 d–1

I Total radiation MJ m–2 d–1

T Temperature °CIce Ice-cover (0.1) –t Time dh Depth of grid cell m————————————————————————


high DIP concentrations. The summer weatherwas completely opposite with occasional heavywinds and rain combined with low surface tem-perature. Due to the weather conditions in 1998the blooms remained moderate compared to the

previous summer. Calibration results for thewesternmost (Längden) and easternmost (Huo-vari) sites are presented in Fig. 2. The calibratedparameters together with references are present-ed in Table 2.

Table 2. Model parameters—————————————————————————————————————————————————Symbol Definition Reference Value Unit—————————————————————————————————————————————————NinC Nitrogen in cyanobacteria Redfield 1958 0.0193 –PinC Phosphorus in cyanobacteria Redfield 1958 0.00268 –NinA Nitrogen in the other algae Redfield 1958 0.0193 –PinC Phosphorus in the other algae Redfield 1958 0.00268 –µCmax Maximal growth rate of cyanobacteria calibration 0.5 d–1

µAmax Maximal growth rate of the other algae Olli et al. 1996 0.7 d–1

RCmax Maximum loss rate of cyanobacteria calibration 0.1 d–1

RAmax Maximum loss rate of the other algae calibration 0.15 d–1

KNC Half-saturation coefficient of DIN for Tyrrell 1999 0 mg m–3

cyanobacteriaKPC Half-saturation coefficient of DIP for Kononen & Leppänen

cyanobacteria 1997 2 mg m–3

KNA Half-saturation coefficient of DIN for the calibration 7 mg m–3

other algaeKPA Half-saturation coefficient of DIP for the calibration 1 mg m–3

other algaeKIC Half saturation coefficient of radiation for calibration 20 MJ m–2 d–1

cyanobacteriaKIA Half saturation coefficient of radiation for the calibration 15 MJ m–2 d–1

other algaeCmin Minimum biomass of cyanobacteria calibration 0.5 g m–2

Amin Minimum biomass of the other algae calibration 0.01 g m–2

Amax Maximum total biomass of algae calibration 300 g m–2

β0 Maximal detritus nitrogen mineralisation rate Garber 1984 0.018 d–1

γ0 Maximal detritus phosphorus Garber 1984 0.043 d–1

mineralisation rateνNdet Settling rate of detritus nitrogen Heiskanen & Tallberg 1 m d–1

1999νPdet Settling rate of detritus phosphorus Heiskanen & Tallberg 1 m d–1

1999SNdet Sedimentation rate of detritus nitrogen calibration 0.16 m d–1

SPdet Sedimentation rate of detritus phosphorus Lehtoranta 1998 0.00 m d–1

Topt Optimal temperature Kononen & Leppänen 25 °Cfor the growth of cyanobacteria 1997for the growth of the other algae calibration 15 °Cfor losses calibration 25 °Cfor detritus nitrogen mineralisation Garber 1984 18 °Cfor detritus phosphorus mineralisation Garber 1984 18 °C

aT Coefficient for temperature limiting factorfor the growth of cyanobacteria calibration 1.14 –for the growth of the other algae calibration 1.001 –for losses calibration 1.05 –for detritus nitrogen mineralisation Garber 1984 1.31 –for detritus phosphorus mineralisation Garber 1984 1.60 –

Ired Radiation attenuation by ice calibration 0.5 –hmix Depth of mixing layer calibration 20 m—————————————————————————————————————————————————


Table 3. Model equations, rates and limiting factors.—————————————————————————————————————————————————Equations

∂∂

µc

tR cC

C C C= −( ) (1)

∂∂

µc

tR cA

A A= −( )Α(2)

∂∂

β µ µc

tc N c h N c h LDIN

Ndet A inA A mix C inC C mix DIN= − − +− −1 1 (3)

∂∂

γ µ µc

tc P c h P c h LDIP

Pdet A inA A mix C inC C mix DIP= − − +− −1 1 (4)

∂∂

β νc

tN R c h N R c h c c h S c htNde

inA A A mix inC C C mix Ndet Ndet Ndet Ndet Ndet= + − − −− − − −1 1 1 1(5)

∂∂

γ νc

tP R c h P R c h c c h S c hPdet

inA A A mix inC C C mix Pdet Pdet Pdet Pdet Pdet= + − − −− − − −1 1 1 1(6)

Rates

µ µC Cmax CN DIN DIP CI AC A C= f c c f I f T f c c( , ) ( ) ( ) ( , ) (7)

µ µA Amax AN DIN DIP AI AC A C= f c c f I f T f c c( , ) ( ) ( ) ( , ) (8)

R R f T c C cC Cmax C C= −( )( )min(9)

R R f T c A cA Amax A A= −( )( )min(10)

β β= 0 f T( ) (11)

γ γ= 0 f T( ) (12)

Limiting factors

f c cc

c K

c

c KCN DIN DIPDIN

DIN NC

DIP

DIP PC

( , ) =+

×+

(13)

f c cc

c K

c

c KAN DIN DIPDIN

DIN NA

DIP

DIP PA

( , ) =+

×+

(14)

f T dTT

T

( ) exp ln=

∫ θ

opt

, where θ = + −a a T TT T opt( )1 (15)

f II I

I I KCIred

red IC

IceIce

( )( )

( )= −

− +1

1(16)

f II I

I I KAIred

red IA

IceIce

( )( )

( )= −

− +1

1(17)

f c cc c

AAC A CA C( , )

max

= − +1 (18)

—————————————————————————————————————————————————


L�ngden

HuovariA

B

C

D

E

F

G

H I

J

K

1

0 50 100 150 200 km

2 3 45678

9 1011 12

13

14

1516

17

18192021

Sweden

BalticSea

Estonia

Russia

Finland

Drainage

area

Fig. 1. The calculation area for the 3D model of the Gulf of Finland, the Baltic Sea. Grid size of the map is 5 km.The area is divided into 11 parts (A–K) for staring values of the nutrient concentrations, which are presented inTable 5. The most important point sources are presented as numbers (1–21) and their nutrient load is given inTable 4. The five sampling stations used for model calibration are presented as black circles. Calibration resultsof the westernmost (Längden) and the easternmost station (Huovari) are given in Fig. 2.

Validation

The N-fixing cyanobacteria are the only speciesforming floating accumulations in the open GOF.Most of these accumulations are formed byNodularia. Aphanizomenon is more commonclose to the coastline, but it often grows in thedeeper water layers contributing only a minorpart to the floating accumulations (Kononen etal. 1998). The Nodularia accumulations can bedetected in satellite images during cloudlessconditions (Kahru et al. 1994). The model wasvalidated by using NOAA satellite image inter-pretations from the on-line data bank of theStockholm University (Naturgeografiska institutio-nen) (http://wwwmarin.natgeo.su.se/~ab/). Rep-resentative sets of images were available for twosunny summers 1997 and 1999. The number ofimages with detected accumulations in the GOFwere 16 and 13, respectively. The combinationof these image interpretations was comparedwith the modelled mean biomass of N-fixingcyanobacteria for the corresponding periods.

Simulating the effects of nutrient load re-ductions

The calibration runs lasted only one year andeach of them was started with a separate set ofstarting values. There is no reason to run cali-

bration of several years in a sequence, becausethe model at its present state cannot simulate theoxygen dependent sediment processes like themassive phosphorus release in 1996–1997. Forsimulating the effects of nutrient load reduc-tions, one year is not a long enough period todemonstrate the effects in the scale of GOF,where the theoretical residence time of water isclose to 3 years. (e.g. Alenius et al. 1998). Thus,we have to extend the simulations to last clearlylonger than the calibration periods — even thoughthis increases the uncertainty of the results con-siderably. The uninterrupted six-year simula-tions were started with initial nutrient concentra-tions measured in January–March 1999. Thesimulations ended to the year 2005, which is thetarget year of the Finnish national agenda. Thesimulations were run by using real weather datafrom 1993–1998.

The water quality model uses precalculatedflow fields and wind data for the computation ofhydrodynamics. The precalculated fields are flowfields for certain steady wind conditions. Theyare used in combination with real wind data tocalculate an approximation of the real flow field.In addition to the wind induced flows, also theriverine fresh water imposes a flow componentto the GOF. Only the largest rivers are includedand their river flow dependent flow fields areadded to the calculation. The procedure is ex-plained in detail by Virtanen and Koponen (1985).


This approximation requires much less comput-ing time compared to online flow calculationmaking long simulation periods possible.

Both the wind induced flow fields and theriver flow fields are calculated separately forthermally stratified and non-stratified conditionsrepresenting summer and winter periods. In sum-mer (March–September) the thermocline is setpermanently to the depth of 12 meters. The endconcentration fields for algae, soluble nutrientsand detritus nutrients were saved when the set offlow fields was changed and these concentrationfields were used for the starting values forcalculating the following season.

To be able to run 6-year simulations we hadto make a presumption that with the presentnutrient load and without any significant chang-es in the internal loading, GOF is in a steadystate, in which the nutrient input and output areequal. When the load is cut, output increasesinput and the nutrient concentrations start todecrease. In the model, the steady state was setup by calibrating the sedimentation rate of detri-tus nitrogen so that the nitrogen levels remainedstable throughout the simulation period with thepresent nutrient load. The steady state assump-tion is close to the real situation in the 1990swhen the rise in the DIN concentrations have

250

200

150

100

50

0

50

40

30

20

10

0

100

80

60

40

20

0

Western

GulfofFinland(L�ngden)

Eastern

GulfofFinland(Huovari)

Biomass g mÐ2 Surface water

DIN (µg dmÐ3)

Total biomass Nitrogen-fixing species

1998

1997

1995

1998

1997

1995

Surface water

DIP (µg dmÐ3)

J F M A M J J A S J F M A M J J A S J F M A M J J A S

250

200

150

100

50

0

50

40

30

20

10

0

100

80

60

40

20

0

250

200

150

100

50

0

50

40

30

20

10

0

100

80

60

40

20

0

250

200

150

100

50

0

50

40

30

20

10

0

100

80

60

40

20

0

250

200

150

100

50

0

50

40

30

20

10

0

100

80

60

40

20

0

250

200

150

100

50

0

50

40

30

20

10

0

100

80

60

40

20

0J F M A M J J A S J F M A M J J A S J F M A M J J A S

Months Months Months

Months Months Months

Fig. 2. Example of the eco-system model calibrationat two intensive monitoringsites on the western andthe eastern Gulf of Finland(see Fig. 1). The calibratedparameters were total phyto-plankton biomass and sur-face water DIN and DIP.Observations are describedas black circles and modelresults as lines.


been stabilising or even a slight decrease havebeen observed (Perttilä et al. 1995, Anonymous1999, Kauppila and Bäck 2000). The main rea-son for this change is supposed to be a decreasedN-load due to the collapse of the Soviet Union.The net sedimentation rate of detritus phospho-rus was assumed to be close to zero (Lehtoranta1998) due to the poor oxygen conditions, so thatthe only output of phosphorus was the outflowfrom the GOF. This assumption made it possibleto keep the phosphorus concentrations in themodel simulation at the relatively high level ofthe late 1990s.

The simulation was first run with the presentload and then with reduced loads according tothe chosen scenarios. The average biomasses ofboth algal groups were recorded for the lastgrowing season (March–September) of the 6thcalculation year. The biomass of each scenariowas then compared with the biomass calculatedby using the present load. The results are pre-sented as relative change in the average biomassof the whole growing season. Due to the sim-plicity and possible semiquantitative nature ofthe model, we decided to present the results,instead of actual percentages, to slight (< 15%)or significant (> 15%) changes. Changes lowerthan 2% were neglected.

Nutrient load and starting values

Information on the loads of total N and P frompoint sources and via rivers from Finland to theGOF was obtained from the databases main-tained by FEI. The total nutrients were convert-ed into bioavailable nutrients by using results ofalgal tests (Ekholm 1994, Ekholm and Krogerus1998). The main nutrient sources to the GOF arethe from the city of St. Petersburg and the riverNeva discharges to the eastern part of the GOF.Information on their bioavailable nutrient dis-charges are based on the river flow measure-ments, nutrient analysis and availability testscarried out in co-operation between Russianresearch institutes, environmental authorities andFEI. Other nutrient load information from Rus-sian and Estonian sources are obtained directlyfrom the local environmental authorities underthe Estonian–Finnish–Russian co-operation

(Pitkänen et al. 1997). Atmospheric supply ofnitrogen is based on model calculations (Bar-tnicki et al. 1998). The annual nutrient discharg-es for 1998 are presented in Table 4 togetherwith the nutrient load reduction scenarios. Formodel input the discharges have been convertedto monthly values which take into account sea-sonal variation in river flow.

The starting values of DIN and DIP concen-trations were given for 11 sub-areas of GOF(Fig 1). Different sets of starting values wereused for years 1995, 1997/1998 and 1999 (Table5) because of the benthic DIP release in latesummer 1996. Starting values are based onwinter measurements by FEI and FIMR duringJanuary–March when the water column has beenvertically well mixed down to the permanenthalocline. Sampling stations do not cover thewhole GOF evenly. Best data sets were availa-ble from the Finnish coastal waters and from themiddle and western GOF. The available infor-mation from the Estonian and Russian coastalwaters was scarce.

Results

Validation

The nutrient starting values used for validation aswell as calibration were most comprehensivealong the Finnish coast. In 1997 some measure-ments were available from the Estonian andRussian territorial waters, but we lacked all near-coast measurements. In 1999 the lack of currentinformation was even worse and we had to usethe 1997/1998 starting values for the southernand eastern sides of the GOF. These values weremost likely too high for DIP, because in otherparts of the GOF a clear decreasing trend wasobserved. Therefore the quantitative validationwas carried out only for the northern side ofGOF were the best sets of starting values wereavailable.

The satellite detected frequency of the N-fixing cyanobacteria accumulations (mainly Nod-ularia) was generally twice as high in 1997 as in1999 (Fig. 3). The clearest change in the distri-bution of the accumulations between the valida-tion years can be seen in the coastal areas of the


Table 4. Bioavailable nutrient load to the Gulf of Finland (GOF). The present load is based on data and esti-mates for the year 1998. The load reductions “Finnish national agenda” and “P-removal in St. Petersburg” arecalculated on the base of this data. The load reduction data is presented only for those sources where reduc-tions take place.—————————————————————————————————————————————————Number in Source Type of discharge Present (1998) Finnish national P-removal inFig. 1 agenda St. Petersburg

———————— ———————— —————————DIN t y–1 DIP t y–1 DIN t y–1 DIP t y–1 DIN t y–1 DIP t y–1

—————————————————————————————————————————————————1 Karjaanjoki river 422 7 304 5 – –2 Kirkkonummi municipal 73 0 41 0 – –3 Espoo municipal 447 5 443 5 – –4 Helsinki municipal 1 311 13 1 197 7 – –5 Vantaanjoki river 934 24 708 17 – –6 Porvoo municipal 92 0 18 0 – –7 Mustijoki river 256 7 186 6 – –8 Porvoonjoki river 900 16 578 11 – –9 Koskenkylänjoki river 146 4 116 3 – –10 Kymijoki river 3 130 64 2 555 44 – –11 Kotka municipal/industrial 158 4 27 4 – –12 Virojoki river 102 3 84 2 – –13 Viipuri municipal/river 380 40 – – – –14 Neva river 22 000 500 – – – –15 St. Petersburg municipal 16 500 1 980 – – 16 500 1 41016 Koporskaja river 370 55 – – – –17 Luga river 1 580 45 – – – –18 Narva municipal/industrial 520 45 – – – –19 Narvajoki river 2 780 390 – – – –20 Kohtla-Järve municipal/industrial 2 090 30 – – – –21 Tallin municipal 1 100 35 – – – –

Finland non-point 2 103 83 1 712 64 – –Estonia non-point 5 300 245 – – – –Atmospheric 17 250 0 – – – –Total discharge to GOF 79 944 3 595 77 839 3 533 79 944 3 025Change –2 105 –62 0 –570

—————————————————————————————————————————————————

Table 5. Starting values for the bioavailable nutrients in the Gulf of Finland. Values are based on measurementcarried out between 1 January and 31 March and represent well mixed winter conditions. Missing values re-placed with earlier measurements are written in italics.—————————————————————————————————————————————————Area code No. of 1995 1997 and 1998 1999in Fig 1. sampling ————————————— ————————————— —————————————

stations DIN (mg m–3) DIP (mg m–3) DIN (mg m–3) DIP (mg m–3) DIN (mg m–3) DIP (mg m–3)—————————————————————————————————————————————————A 4 120 22 105 25 148 31B 4 89 18 97 23 108 30C 2 100 19 87 21 81 20D 40 194 30 188 34 167 37E 3 118 26 118 32 104 30F 1 141 25 109 26 109 26G 8 181 30 168 38 202 35H 3 142 29 156 41 160 33I 1 150 28 122 29 122 29J 3 220 31 200 33 200 33K 0 250 30 250 30 250 30—————————————————————————————————————————————————


eastern GOF. In 1997, a high frequency ofaccumulations followed the coastline to theFinnish–Russian border. In 1999, this area re-mained practically free of cyanobacteria accu-mulations. The model was generally able toreproduce the observed spatial distribution andannual variation in the accumulations. In theopen GOF, the model was able to explain 50%of the observed pattern (Fig. 4). Along the coastthe predicted biomass was generally too high.

3.2. Scenarios

According to the model results, the Finnishnational agenda doesn’t seem to be an effectiveway to fight against accumulations of cyanobac-teria (Fig. 5). The results indicate that there maybe a risk for slight biomass increase in the N-fixing cyanobacteria along the densely populat-ed part of the Finnish coastline. However, thenational agenda will decrease the total phyto-plankton biomass in the vicinity of the Finnishcoastline, but its effect to the whole GOF ispractically negligible. The coastal biomass de-crease takes place both during the spring and thesummer period (Fig. 6A)

The model results indicate that phosphorusremoval in St. Petersburg would affect the bio-mass of N-fixing cyanobacteria practically in the

whole GOF (Fig 5). Their biomass will decreasemainly in the central parts of GOF but thepositive effects might be detectable even at thewestern border of the calculation area. Theeffect will be negligible outside of St. Peters-burg, because this area is mainly phosphoruslimited and thus outside the distribution of N-fixers. The total phytoplankton biomass willdecrease in the whole open GOF except for anarrow zone of slight increase across the easternGOF. The spring bloom biomass is decreasingslightly even in this zone. The main increase

Fig. 3. Validation of N-fixing cyanobacteria dur-ing the bloom periods ofsummers 1997 and 1999.Modelled biomass and fre-quency of satellite detectedaccumulation in the Gulf ofFinland. The satellite im-ages are processed andinterpreted by the Natur-geografiska institutionenof the Stockholm Univer-sity. The area used forquantitative validation (Fig.4) is indicated by a rectan-gle.

0

2

4

6

8

10

12

14

16

0 5 10 15 20 25 30 35

Number of overlappingdata points1 2 3 4

y = 0.4x Ð 0.5

R2= 50%

n = 50

Modelled biomass of N-fixing cyanobacteria (g mÐ2)

Frequencyofsatellite

detected

accumulations

Fig. 4. Quantitative validation of N-fixing cyanobacteriaduring the bloom periods of summers 1997 and 1999.Modelled biomass versus frequency of satellite de-tected accumulation in the Gulf of Finland. The areaused for the quantitative validation is indicated by arectangle in Fig. 3.


takes place in the summer (Fig. 6B).If both nutrient reduction scenarios are com-

bined, their adverse effects seem to diminish.The possible increase in the biomass of N-fixingcyanobacteria due to the Finnish national agen-da can no longer be detected in the modelsimulation if the scenarios are combined. Fur-thermore, the increase in the total phytoplanktonbiomass in the zone across the GOF will de-crease.

Discussion

According to the validation made with satelliteimages, the outcome of our model appeared tobe reasonably good in simulating the biomass ofthe toxic genus Nodularia. Finding good valida-tion data proved to be difficult due to the ex-tremely patchy character of the blooms. Thusroutine water sampling rarely is suitable forestimating the magnitude of cyanobacterialblooms. The semiquantitative validation data

gathered by using satellite images proved to beuseful until other means of automated data col-lection are available.

The validation runs lasted only one growingseason and they were started with separate year-specific sets of nutrient starting values. Thegood results make us believe that the outcome ofthe blooms is highly dependent on the startingvalues, especially on the amount of excess DIPleft over when the spring bloom has consumednutrients according to the Redfield ratio. How-ever, if temperature is not high enough for thegrowth of cyanobacteria, all excess DIP cannotbe utilised and the bloom do not reach itsmaximum potential biomass. The starting valuedependence offers us a possibility to use themodel for short-term forecast purposes. Thewinter nutrient concentrations can be used toforecast the magnitude of the blooms for thecoming summer.

The most serious problem in the bloommodelling was obvious along the coastline. Inour validation data the blooms seemed to be

N-fixing cyanobacteria Total phytoplankton

Finnishnationalagenda

P-removalinSt.Petersburg

Both

scenariostogether

Relative biomass change

Helsinki

Finland

Tallin

Estonia

St. Petersburg

Russia

no change slight increaseslight decreasesignificant decrease significant increase

Time-series in Fig. 6B

Time-series in Fig. 6A

Fig. 5. The relative effectof the nutrient load reduc-tions on the biomass of N-fixing cyanobacteria andthe total phytoplanktonbiomass (N-fixing cyano-bacteria + the other phyto-plankton) in the Gulf of Fin-land. Finnish nationalagenda aims at 50% re-duction of both N and Pdischarges in Finland com-pared to the level of early1990s. P-removal in St.Petersburg reduces onlythe P-load but it does nothave any effect on the N-load.


absent along the coast where, however, highbiomasses were shown by the model. One ex-planation for the mismatch might be difficultiesto detect the biomass close to the coastline andin the archipelago from the satellite images. Themain genus close to the coastline is commonlyAphanizomenon, which do not form floatingaccumulations like Nodularia (Kononen et al.1998). If Aphanizomenon is not visible in thesatellite images, it is possible that the modelresults are not as far from the reality as appearsin our validation. An other possible explanationis the low sedimentation rate of detritus phos-phorus used in the model. This may cause toohigh recycling and availability of phosphorus inthe shallow areas, where detritus cannot settlebelow the productive zone, promoting the growthof N-fixing cyanobacteria.

The model results indicate that N-dominatednutrient load reductions, like the Finnish nation-al agenda, cannot decrease the biomass of N-fixing toxic cyanobacteria, but can even have aslight increasing effect. Some experiments sup-port this results (Elmgren and Larsson 1997).According to the model, the possible increasewill occur mainly in the coastal area where thebiomass of N-fixing cyanobacteria is generallylower than on the open sea. The reason for thelower biomass can be too high N:P ratio or alsothe influence of the bacterial production. Asbacteria are superior competitors with regard toDIP uptake and they are often P-limited in thecoastal waters, they may alter the P dynamicsconsiderably (Heinänen and Kuparinen 1992).Thus the interactions of bacteria and DIP aredifficult to predict in a simple model simulation.Near the coastline the combination of additionalDIP and a high availability of dissolved organicsubstances may at least partly compensate thesimulated increase by the increased phosphorusuptake and production of bacteria. An otherprocess able to compensate the increase in theN-fixing cyanobacteria is the influence of thedecreasing total phytoplankton biomass to theoxygen consumption and P-release from thedeep bottom areas. The present model do notdescribe this connection.

In winter, when the algal growth is close tozero, the nutrient retention capacity of the Nevaestuary is negligible in the model simulations.

The same phenomenon has already been demon-strated by Savchuk (2000). Especially duringthe ice covered period, when the wind inducedvertical mixing is excluded, the fresh water flowfrom the river Neva can often be detected alongthe Finnish coastline (e.g. Alenius et al. 1998).In spring when the ice-cover breaks-up and thephytoplankton production starts, the nutrientscarried by this flow are rapidly consumed by thespring bloom. The P-load reduction seems toalter the N:P ratio of this westward flowingwater mass so that less DIP is left for thecyanobacteria after the spring bloom of phyto-plankton. We believe that this is the main proc-ess explaining the model indicated significantdecrease in the cyanobacteria biomass in the

A

100

100

0

0

Finnish national agenda

Present

P-removal in St. Petersburg

PresentB

gm

Ð2

gm

Ð2

Mar Apr May Jun Jul Aug

Fig. 6. Time-series of the influence of the nutrient loadreductions to the seasonality of the other phytoplankton(total phytoplankton without N-fixing cyanobacteria).Both time series represent the last 6th calculation year.— A: the area where the influence of the Finnish na-tional agenda is clearly detectable. — B: the area withthe highest increase in the total phytoplankton biomasscaused by P-load reduction in St. Petersburg. Theexact locations of the time-series points are presentedin Fig 5.


central parts of GOF.According to the model results, both scenari-

os seem to decrease the spring bloom of phyto-plankton in their impact area. The biomass pro-duced during the spring bloom is not grazedeffectively and it is responsible for the mainoxygen consuming load to the sediment bottoms(Heiskanen and Kononen 1994). Thus the bothscenarios should have a positive effect also ondeep sediment bottoms. However, the springbloom dynamics is difficult to mimic in a simplephytoplankton model. In the Baltic Sea several ofthe most important cold-water diatoms and di-noflagellates produce cysts, which settle rapidlyat the end of the bloom (Heiskanen and Kononen1994). These cysts contain a large quantity ofhigh-energy compounds for the resting period.They are a mode of long-term burial of bothorganic carbon and nutrients, which complicatethe modelling of nutrient circulation. It is verydifficult to estimate what happens to the cystproduction when there are less nutrients availa-ble. It is even possible that the cyst production isstimulated leading to a higher loss of nutrientsand carbon out of the system (Kuosa et al. 1997).

The model results are sensitive to the N:P ratioof the algae. In our model this ratio is fixed to thewidely accepted and used Redfield ratio, 7.2 byweight. If the algal community is able to accom-modate its N:P ratio or to perform luxury uptake ofthe less limiting nutrient, the above describednegative effects of the nutrient load reductionsmay not be any more valid. This kind of alterna-tions from the Redfield ratio have been reported indiatoms, which are an important part of springbloom in the Baltic (Istvanovics et al. 1994, Glib-ert et al. 1995). Whether these processes areimportant in the case of the GOF is not known.Therefore the model results should be consideredrather as caricatures of the changes caused by theload reductions. At least the negative effects aremore likely to be over than underestimates.

Phenomena occurring in the spatial scaleclose to the 5 km horizontal model resolutioncannot be reliably simulated due to so callednumerical diffusion. This means that runningthe model in a coarse resolution induces excessdiffusion to naturally sharp boundaries found

e.g. in the coastal zone and archipelago. In thecase of the Finnish national agenda, the modelprobably spreads its effects too far from thecoast and therefore underestimates the biomasschanges along the coastal zone. The modelindicated slight reduction in the total phyto-plankton biomass is likely to be a clear underes-timate for areas close to the most importantFinnish point sources and river mouths.

In the model results, reduction in the phos-phorus load from St. Petersburg decreases nitro-gen retention in the phosphorus limited Nevaestuary during the growing season and thuspromote export of nitrogen westward in sum-mer. This process have been documented in asmaller scale in the Stockholm archipelago(Brattberg 1986). We believe that the zone ofincreased total phytoplankton biomass shown inFig. 5 is formed when this nitrogen load fromthe Neva estuary meets the nitrogen limitedparts of the GOF. The location of this zone isnot necessarily fixed to a certain part of theGOF. In our test runs with altered weather dataor starting values, it moves between the easternand the central parts of the GOF.

In recent model simulations dealing with theeffects of the nutrient load reductions in theGOF (Kuusisto et al. 1998, Inkala and Pitkänen1999, Savchuk and Wulff 1999, Savchuk 2000),the authors have concluded that the most effec-tive way to reduce eutrophication in the Gulf ofFinland is to cut nitrogen load. Only in the Nevaestuary the cutting of phosphorus load is foundto have positive effects. Our results are not inconflict with these previous results. In the main-ly nitrogen limited GOF, nitrogen removal iswithout doubt the most effective means to cutthe primary production and to slow down theprogress of eutrophication. However, the poten-tially toxic blooms of cyanobacteria in the openGOF are mainly phosphorus limited and theiroccurrence is dependent on the summertime DIPavailability in the surface water. Even thoughour model approach do not take into account thewater exchange with the Baltic Proper, it seemsto be clear that phosphorus removal will affectthe N:P ratio to a less favourable direction forthe nitrogen fixing cyanobacteria.


Management guidance

The effective nutrient load reduction in Finlandhas started by the chemical phosphorus precipi-tation. By now, phosphorus load from most ofthe point sources is minimal and only diffuseload mainly from agriculture can be furtherdiminished. The logical next step is reduction ofthe nitrogen load. Nitrogen dominated load re-ductions in Finland do not necessarily affect thebiomass of N-fixing toxic cyanobacteria, butthey decrease the total phytoplankton biomassclose to the coastline. The phosphorus load canstill be cut significantly in the city of St. Peters-burg. According to our results, this cuttingmight be a realistic option to decrease the bio-mass cyanobacteria in the GOF. Combination ofimproved nitrogen removal in Finland and phos-phorus removal in St. Petersburg would de-crease both the toxic blooms of cyanobacteriaand the total phytoplankton biomass along theFinnish coastline.

We have to remember that blooms of cyano-bacteria are a natural part of the Baltic Seaecosystem. We may be able to manipulate theirintensity but we cannot stop them. Their occur-rence is not only dependent on the absolutenutrient load or sea water nutrient concentra-tions but also the N:P ratio. Therefore the de-crease in the total nutrient load do not necessari-ly guarantee a decrease in their outcome.

References

Alenius P., Myrberg K. & Nekrasov A. 1998. The physi-cal oceanography of the Gulf of Finland: a review.Boreal Env. Res. 3: 97–125.

Anonymous 1998. Water Protection Targets to 2005.Suomen ympäristö 226: 1–82. [In Finnish with Eng-lish summary].

Anonymous 1999. Nutrients in European ecosystems.Environmental assessment report 4: 1–155. EuropeanEnvironment Agency.

Bartnicki J., Barrett K., Tryso S., Erdman L., Gusev A.,Dutchak S., Pekar M., Lükewille A. & Krognes T.1998. Atmospheric supply of nitrogen, lead, cadmi-um, mercury and lindane to the Baltic Sea. EMEP/MSC-W Note 3/98.

Bianchi T.S., Westman P. & Rolff C. 2000. Cyanobacteri-

al blooms in the Baltic Sea: Natural or human in-duced? Limnology and Oceanography 45: 716–726.

Brattberg G. 1986. Decreased phosphorus loading chang-es phytoplankton composition and biomass in theStockholm archipelago. Vatten 42: 141–153.

Ekholm P. 1994. Bioavailability of phosphorus in agricul-turally loaded rivers in southern Finland. Hydrobio-logia 287: 179–194.

Ekholm P. & Krogerus K. 1998. Bioavailability of phos-phorus in purified municipal wastewaters. Water Re-search 32: 343–351.

Elmgren R. & Larsson U. 1997. Himmerfjärden. Changesin a nutrient enriched coastal ecosystem. Naturvårds-verket Rapport 4565: 1–197. [In Swedish with Eng-lish summary].

Falkowski P.G. 1997. Evolution of the nitrogen cycle andits influence on the biological sequestration of CO2 inthe ocean. Nature 387: 272–275.

Fonselius S.T. 1969. Hydrography of the Baltic deepbasins III. Fishery Board of Sweden, Series Hydrog-raphy, Report 23.

Frisk T. 1982. An oxygen model for Lake Haukivesi.Hydrobiologia 86: 133–139.

Garber J.H. 1984. Laboratory study of nitrogen and phos-phorus remineralisation during the decomposition ofcoastal plankton and seston. Estuarine, Coastal andShelf Science 18: 685–702.

Glibert P.M. Conley D.J., Fischer T.R., Harding L.W.Jr.& Malone T.C. 1995. Dynamics of the 1990 winter/spring bloom in Chesapeake Bay. Marine EcologyProgress Series 122: 27–43.

Heinänen A. & Kuparinen J. 1992. Response of bacterialthymidine and leucine incorporation to nutrient (NH4,PO4) and carbon (sucrose) enrichment. Archiv fürHydrobiologie 37: 241–251.

Heiskanen A.-S. & Kononen K. 1994. Sedimentation ofvernal and late summer phytoplankton communitiesin the coastal Baltic Sea. Archiv für Hydrobiologie131: 175–198.

Heiskanen A.-S. & Tallberg P. 1999. Sedimentation andparticulate nutrient dynamics along a coastal gradientfrom a fjord-like bay to the open sea. Hydrobiologia393: 127–140.

Inkala A. & Pitkänen H. 1999. The effect of load reduc-tions on algal biomass in the eastern Gulf of Finlandestimated by the FINNALGA model. Boreal Environ-ment Research 4: 357–366.

Istvanovics V., Padisak J., Petterson K. & Pierson D.C.1994. Growth and phosphorus uptake of summerphytoplankton in Lake Erken (Sweden). Journal ofPlankton Research 16: 1167–1196.

Kahru M., Horstmann U. & Rud O. 1994. Satellitedetection of increased cyanobacteria blooms in theBaltic Sea: Natural fluctuation or ecosystem change?Ambio 23: 469–472.

Kauppila P. & Bäck S. 2001. State of the Finnish coastal


waters in the 1990’s. Finnish Environment VOL??[In press].

Kivi K., Kaitala S., Kuosa H., Kuparinen J., Leskinen E.,Lignell R., Marcussen B. & Tamminen, T. 1993.Nutrient limitation and grazing control of the Balticplankton community during annual succession. Lim-nology and Oceanography 38: 893–905.

Kononen K., Lahdes E.O. & Grönlund L. 1994. Physio-logical and community responses of summer planktonto nutrient manipulation in the Gulf of Finland (BalticSea) with special reference to phosphorus. Sarsia 78:243–253.

Kononen K. & Leppänen J.-M. 1997. Patchiness, scalesand controlling mechanisms of cyanobacterial bloomsin the Baltic Sea: Application of a multiscale researchstrategy. In: Kahru, M. & Brown, C.W. (eds.), Moni-toring algal blooms: New techniques for detectinglarge-scale environmental change. Landes Bioscience.Austin, pp. 63–84.

Kononen K., Hällfors S., Kokkonen M., Kuosa H.,Laanemets J., Pavelson J. & Autio R. 1998. Develop-ment of a subsurface chlorophyll maximum at theentrance to the Gulf of Finland, Baltic Sea. Limnologyand Oceanography 43: 1089–1106.

Koponen J., Alasaarela E., Lehtinen K., Sarkkula J., Simbi-erowicz P., Vepsä H. & Virtanen M. 1992. Modellingdynamics of large sea area. Publications of the Waterand Environment Research Institute 7: 1–91.

Kuosa H., Autio R., Kuuppo P., Setälä O. & Tanskanen S.1997. Nitrogen, silicon and zooplankton controllingthe Baltic spring bloom: an experimental study. Estu-arine, Coastal and Shelf Science 45: 813–821.

Kuusisto M., Koponen J. & Sarkkula J. 1998. Modelledphytoplankton dynamics in the Gulf of Finland. Envi-ronmental Modelling & Software 13: 461–470.

Lehtoranta J. 1998. Net sedimentation and sediment-waternutrient fluxes in the eastern Gulf of Finland (BalticSea). Vie Milieu 48: 341–352.

Olli K., Heiskanen A.-S. & Seppälä J. 1996. Developmentand fate of Eutreptiella gymnastica bloom in nutrientenriched enclosures in the coastal Baltic Sea. Journalof Plankton Research 18: 1587–1604.

Perttilä M., Niemistö L. & Mäkelä K. 1995. Distribution,development and total amounts of nutrients in theGulf of Finland. Estuarine, Coastal and Shelf Science41: 345–360.

Pitkänen H. 1994. Eutrophication of the Finnish coastalwaters: Origin, fate and effects of riverine nutrientfluxes. Publications of the Water and EnvironmentResearch Institute 18: 1–45.

Pitkänen H., Kondratyev S., Lääne A., Gran V., KauppilaP., Loigu E., Marcovets I., Pachel K. & RumyantsevV. 1997. Pollution load on the Gulf of Finland fromEstonia, Finland and Russia in 1985–1995. Suomen

ympäristökeskuksen monistesarja 105: 9–18. FinnishEnvironment Institute, Finland.

Pitkänen H. & Tamminen T. 1995. Nitrogen and phospho-rus as production limiting factors in the estuarinewaters of the eastern Gulf of Finland. Marine EcologyProgress Series 129: 283–294.

Pitkänen H. & Välipakka P. 1997. Extensive deep wateroxygen deficit and benthic phosphorus release in theeastern Gulf of Finland in late summer 1996. Suomenympäristökeskuksen monistesarja 105: 51–59. Finn-ish Environment Institute, Finland.

Pouchet G. & de Guerne J. 1885. Sur le faune pelagiquede la mer Baltique et du Golfe de Finlande. ComptesRendus des Seances de l’Academie des Sciences 100:919–921.

Rantanen P. 1994. Biological phosphorus removal studyat the Suomenoja research station. Vatten 50: 321–328.

Redfield A.C. 1958. The biological control of chemicalfactors in the environment. American Scientist 46:205–221.

Reichert P. 1995. Design techniques of a computer pro-gram for the identification of processes and the simu-lation of water quality in aquatic systems. Environ-mental Software 10: 199–210.

Reichert P., von Schulthess R. & Wild D. 1995. The useof AQUASIM for estimating parameters of activatedsludge models. Water Science and Technology 31:135–147.

Savchuk O. 2000. Studies of the assimilation capacity andeffects of nutrient load reductions in the eastern Gulfof Finland with a biogeochemical model. BorealEnvironment Research 5: 147–163.

Savchuk O. & Wulff F. 1999. Modelling regional andlarge-scale response of Baltic Sea ecosystem to nutri-ent load reductions. Hydrobiologia 393: 35–43.

Sellner K.G., Olson M.M. & Olli K. 1996. Copepodinteractions with toxic and non-toxic cyanobacteriafrom the Gulf of Finland. Phycologia 35 (suppl.):177–182.

Sivonen K., Kononen K., Esala A.-L. & Niemelä S.I.1989. Toxicity and isolation of the cyanobacteriumNodularia spumigena from the southern Baltic Sea.Hydrobiology 185: 3–8.

Tyrrell T. 1999. The relative influence of nitrogen andphosphorus on the oceanic primary production. Na-ture 400: 525–531.

Virtanen M. & Koponen J. 1985. Simulations of transportunder irregular flow conditions. Aqua Fennica 15:65–75.

Virtanen M., Koponen J., Dahlbro K. & Sarkkula J. 1986.Three-dimensional water-quality-transport model com-pared with field observations. Ecological Modelling31: 185–199.

Received 1 September 2000, accepted 6 November 2000

evaluating the effects of nutrient load reductions on the...

Documents