Non-parametric trend test of baltic sea data

Enuimnmetrics; 1991; 2(3):

263-278

Non-Parametric Trend Test of Baltic Sea Data P. Sanddd, L. Rahm2 and F. Wulff3

ABSTRACT The recently reported tendencies toward decreasing total amounts of silicate in the Baltic Sea are investigated by use of non-parametric trend analysis. The period 1968-1986 showed significant falling trends in surface waters from the whole system. The deepest parts of the Baltic proper did in contrast reveal strong increasing trends. These trends are more pronounced during the latter part of the analysis period which is characterized by stagnant conditions in the Baltic proper. These conditions have been prevailing since the major inflow in 1976. The causes of the observed changes are unclear. The increased load of nutrients and accompanying increase in primary production is, however, one factor. Another is coupled to the stagnation conditions of the Baltic Proper.

KEY WORDS: Non-parametric trend test; Baltic Sea; Silica. 1. INTRODUCTION Nature’s response to mans activities is often gradual, and can be hard to detect before serious and perhaps irreversible damage is imposed on the system. Establishment of trends at an early stage is therefore an important part of environmental research and policy making. Trends in marine chemistry data are, traditionally, analyzed with linear regression methods (e.g. Nehring et al. 1987 and 1990). Usually the data show several properties Swediih Meteorological and Hydrological Institute, $60176 Norrk6ping, Sweden. 33Department of Systems Ecology, University of Stockholm, S10691 Stockholm, Sweden. 1180-4009/91/030263-16$08.00

@John Wiley & Sons, Ltd.

Recerved August 1, 1992 Revised September 5, 1991

264

P. SANDEN,L. RAHMAND F. WULFF

that make them unsuitable for this method. These include inhomogeneous variance, dependence between observations, seasonality, extreme values, censored data and non-Gaussian distributions. Parametric methods for time series analysis have the prerequisite that the observations must be equidistant. This is unfortunately far from the case for oceanographic data. Large numbers of missing data exist in the data series and the sampling expeditions do not follow any strict schedule. The reasons for this are manifold, but it is partly due to the use of "ships of opportunity". There is also a clear bias toward sampling in the summer months with fair weather. Hirsch and co-workers proposed a non-parametric method for trend analysis of water quality records (Hirsch et al. 1982;Hirsch and Slack 1984), with some applications to river phosphorous data. As this method is robust against non-Gaussian distributions, seasonality, non-equidistant observations, extreme values and censored data it can be used for trend analysis in oceanographic records. This paper presents results from the application of the above mentioned non-parametric method t o silicate data from the Baltic Sea. This semienclosed sea (Figure 1) is one of the largest brackish water areas in the world. It has been subject to increased nutrient load over the last decades (Larsson et al. 1985) with higher concentrations of both nitrogen and phosphorous in virtually all parts of the Baltic Sea (Nehring 1987 and 1990). The effect is an increased concentration of nutrients in virtually all parts of the Baltic resulting in an increased primary production (Elmgren 1989; Stigebrandt 1991). Nitrogen is assumed to be the limiting production factor in these brackish waters (Rosenberg et al. 1990); only the oligotrophic and iron-rich Bothnian Bay is phosphorous limited. Silicate is here an interesting parameter in the context of eutrophication. The hypothesis that the decrease in silicate concentration is dependent on increasing primary production is supported by long-term declines in silicate levels and inverse relationships between silicate and nutrient load in both Lake Michigan (Schelske and Stroemer 1971) and the North Sea (van Bennekom et al. 1975; F'ransz and Verhagen 1985). Further, silicate may even become a limiting factor during the spring bloom with its intense diatom deposition on the relatively shallow bottoms. A consequence of an increase in net production would be that the total amount of silicate in the water mass should decrease over time as long as the load of silicate is unchanged. The expected decrease has been found in the calculation of total amounts of nutrients by Wulff and Rahm (1988). These estimates were, however, hampered by the few observations available, and the method used. To corroborate these results, this paper presents a trend analysis of time series from several hydrographic stations. It is of major importance that the validity of the estimates be determined, as these trend estimates may be the foundation for both determination of their causes and future policy making.

-

TREND TESTOF

\ \

THE

265

BALTIC

Significant increase

Significant decrease

r

r'

u

U --

/

I

I

I

U

Figure 1. Map showing both the location of the hydrographic stations used in the study and an overview of results from the trend test. The two regions in the Gulf of Bothnia are indicated by shadowed bands. Only those trends that were significant (p < 0.05) without the assumption of independent observations are shown.

266

P. SANDEN, L. RAHM A N D F.

WULFF

This is so since the long term effect of decreasing silicate concentrations in the surface layers may lead t o drastic changes in the population dynamics of the trophogenic ecosystem. A change in the diatom population as silicate becomes limited could lead to changes in the species composition of the primary producers. This could also have a large effect on other trophic levels of the system.

2. DATA The time series are based on observations of silicate over the period 1968-1986. The hydrographic data comes from the HELCOM and ICES databases at Helsinki and Copenhagen. Standard sampling and analytical methods have been used (HELCOM 1984) and the analysis technique has not been changed over the period. The results from different national laboratories yield a surprisingly high degree of conformity in the analysis, only during strongly anoxic conditions do the analyses differ. These conditions coincide with the formation of high concentrations of hydrogensulphide. Significant differences have only been found after 1986 and do not affect our analysis. For the Baltic proper and the southern Bothnian Sea, stations are located as shown in Figure 1. The lack of observations in the southern Bothnian Bay and the northern Bothnian Sea forced us to introduce two latitudinal bands of one-degree width covering these areas. They include all stations found in the respective bands (Figure 1). They should give a fair description of the spatio-temporal response of the system. Four different seasons were used (January-March, April-June, JulySeptember, and October-December). This was the highest number that could be justified by the sampling frequency. Fkom the surface to 110 meters depth intervals of 10 meters were used. Between 110 and 190 meters depth the interval was 20 meters. Three intervals were defined for the deepest part (192-241, 242-341, and below 342 meters). If more than one sample exists for a given depth interval and season, the average value was used. The one degree latitudinal areas in the Gulf of Bothnia was divided into 6 half-degree regions and each of these was treated as a sampling location.

3. STATISTICAL METHODS As a preliminary step in the data analysis, inspection of time series plots was carried out to check all data. There were no signs of systematic analytical errors or other inconsistencies in these data during the period considered. Examples from surface and deep layers at BY15 and SR5 (for

TRENDTESTOF

THE

BALTIC

267

location see Figure 1) are presented in Figures 2 and 3. The plots reveal the same composition; the surface layers are low in Si while the highest levels are found in the deeper parts. The increase with depth is, however, modest in the Gulf of Bothnia. The same pattern can be found in salinity. The presence of seasonal variation, extreme observations, non-equidistant sampling, missing observations and serial dependence, were factors that had to be taken into account in the data analysis. The scheme of data analysis adopted, was based on the ideas and methods proposed by Hirsch and co-workers (Hirsch et al. 1982; Hirsch and Slack 1984). A non-parametric trend test was used and point estimates of the slopes of the trends were computed. All tests were calculated as twesided tests since both upward and downward trends were of interest. The trend test primarily used has been described by Hirsch et al. (1982). In this test data from different seasons are analyzed separately. Hirsch and Slack (1984) have also modified the test to account for serial dependence. This modification does of course reduce the power of the test. Both tests were performed in the present work. Estimates of the slopes have been calculated using the seasonal Kendall slope estimator (Hirsch et al. 1982). This can be described as the median annual change adjusted for seasonal variation. The six half-degree regions of each band in the Gulf of Bothnia were pooled together according to the intrablock method proposed by van Belle and Hughes (1984). An individual trend analysis for each season was also performed at all stations. In this analysis the surface layer (0-30 m) was used. The depth intervals (0-4, 5-9, 10-14, 15-19, 20-29, 30-34) were treated as seasons according to the intrablock method. Samples taken from different depths at the same location and time are strongly correlated. Therefore the test assuming independent observations was abandoned in this case. Finally time series for the Baltic proper was divided into two parts, before and after 1977. This was done to investigate changes in the Pelagic system due to the large inflow of saline water that occurred during 1976. The Baltic deep water is renewed rather infrequently by these major inflows of water with high salinity and oxygen concentrations while a more continuous renewal occurs with lesser salinities for the intermediate layers. The major inflow of 1976 resulted in stagnant conditions in the largest subbasin of the Baltic Proper that has last well over the end of the analyzed period. This stagnation have resulted in well developed anoxic conditions. Lack of data prevented this split analysis for the Gulf of Bothnia.

4. RESULTS The results of the trend analysis are shown for the period 1968-1986 in Table 1. Here the significance of the trends is calculated assuming independent observations. Since this assumption is questionable, corresponding

268

P.

i

S A N D ~ N L. ,

RAHMAND F. WULFF

1

Figure 2. Time plots of silicate (thin curve with bars) and salinity (thick curve) for BY15. The time period is 1968-1986. The bars denote the 95% confidence interval for the mean values.

TREND TESTOF

THE

BALTIC

SRS O - 2 5 m

269

1

"I

I

s

4

SR5 > l o o m t

Figure 3. Time plots of silicate (thin curve with bars) and salinity (thick curve) for

SR5. The time period is 1968-1986. The bars denote the 95% confidence interval for the mean values.

ns/m

pE ”’

Average

20

21

Average

61

ns/M

”p

100

51.1 -1 2 3 nS/ns

nslns

~~

60.9 -0.60 0.005/0.012

57.8 -1.21 0.013/0.024

47.8 -1.18 ns/m

nslm

35.3

w/ns

57.9 -0.83 0.018/0.052

nslns

56.7

52.5 -0.80 0.00010.004

nslns

40.5

27.7 -0.53 0.021/0.052

o.o94/ns

m/m

20.9

18.0 -0.21

14.1 -0.27 0.009/0.041

13.3 -0.30 0.002/0.016

~

0.000/0.00s

12.5 -0.39

12.6 -0.40 0.000/0.002

BY31

14.6

13.1 -0.28 0.035/m

0.009/0.067

-0.34

12.6

12.3 -0.39 0.001/0.014

12.3 -0.42 0.001/0.016

BY27

Table 1. (Continued on following page)

P

59.4

Average

91

P

nsfns

pP ”‘

90

50.2

0.o94Ins

nslm

59.5

42.0 -0.58

ns/ns

29.5

ns/ns

ns/ns 17.0

13.2

11.4 -0.28 ns/ns

10.7 -0.26 ns/ns

9.4 -0.16 O.O58/ns

54.0

Average

P’lP2

Slope

Average

ns/m

41.8

nslm

ns/m 26.6

17.3

ns/m

81

80

-

71

,S’E

,S’E

SO

70

Average

p”’~

Average

:&

50 51

41

40

14.3

31

Average

ns/m

30 p”@

13.4

12.6

Average

O.ooS/O.O63

ns/m

11

9.4 -0.26

BY15

12.8

pP ”’

Average

BY5

10

0

hted(m)

~~

.

59.7 -0.82 0.007/0.013

nslns

57.9

0.041~

48.6 -0.76

nsh

36.0

ns/ns

ns/ns 22.4

16.7

0.022/ns

14.1

0.002/0.031

13.4

12.8 -0.47 0.000/0.013

12.5 -0.43 0.000/0.004

BY32

0.001/0.029

61.2

59.0 -0.53 O.O55/ns

52.1 -0.79 O.O47/ns

42.4 -0.81 0.029/0.058

ns/m

28.4

4118

20.0

16.6 -0.50 nslm

15.7 -0.50 ns/m

~

ns/m

14.4

14.0 -0.19 O.O45/ns

BY38 ~~~~~

nS/M

25.1

Nlns

25.7

0.000/0.002

24.5

ns/ns

22.8

ILII/N

19.4

o.on/ns

16.5 -0.47

15.2 -0.53 0.017/0.058

14.7 -0.76 O.O33/ns

~

~~

17.3 -0.74

12.7

W/M

12.4

nslm

11.3

ndns

14.8

nS/M

~~

19.6 -0.75 0.001/0.029 21.6 -0.58 0.003/ns

~~~~

18.1 -1.01 0.001/0.035

O.OO0/0.029

~

14.2 -0.54 0.002/0.014 ~~

16.9 -0.73 O.OO0/0.016

O.OOO/O.ooS

14.3 -0.59

16.7 -0.79 O.OO0/0.001

~

~

19.4

~~

nd

nd

ns/ns

nd

nd

nd

m/m

28.1

O.WOfO.ooS

19.4

0.000/0.007

~~~

O.OOO/O.ooS

19.2 -0.78

0.000/0.009

20.0 -0.85

4 r -l

C

r

t3 4 0

-

Averwe

171

v

.

..

Averwe

ipE

A

~~

~~

~~~~~

nd

nd

nd

78.4 1.71 o.o0o/o.o08

o.o0o/o.oos

74.4 1.63

69.6

66.2 1.07 0.001/0.016

nd

’ : independence ~ ~ ~ u x n e d

p”PF

Aver-

nd

nd

nd

nd

nd

nd

56.0

ns/ns

nd

: independaxe not suumed M : not significant nd : no data

-

341

-~

241

240

191

;pF

q

m-

170

Is0

Avercwe

151

-

Epfi

150

I

Average

Average

131

111

Average

BY15

BY5

~

~~~~

nd

nd

nd

nd

66.4

O.oSs/lU

63.4 -0.42

M/M

63.2

62.5 -0.44 0.05/~

N/N

63.5

rn

nd

nd

nd

o.oss/o.o95

63.8 -0.72

65.6

0.009/0.0%

O.osS/na

~

62.4 -0.72

61.8 -0.40

66.6

61.7 -0.74 0.009/0.017

nd

BY32

-1.17 0.014/0.040

60.6 -0.36 0.086/ns

nd

nd

65.4 -1.22 0.005/0.012

BY31

BY27

nd

nd

nd

nd

nd

nd

nd

63.5 -0.48 0.020/0.031

BY38

nd

nd

nd

nd

nd

nd

31.2

nd

SR5

nd

nd

nd

nd

nd

nd

nd

nd

B Northern ~

nd

nd

nd

nd

nd

nd

nd

BothnianBay ~

~

Table 1. Trend analysis of silicate data at different stations and depth intervals. Average concentration (mmol m-3), slope (mmol ma3 yr-’) and p-values (two-sided test) are presented.

I

I

1

I

I

101

Inkrval(m)

z!

h)

2m

c)

E

el

U

Ez

el

~

s

272

P. S A N D ~ L. N , RAHMAND F. WULFF

calculations were also performed without this assumption. In both cases only those estimates with pvalues less than 0.10 are presented. A falling trend is detectable in the surface waters throughout Baltic Sea, with the exception of the southernmost station, the Bornholm Basin (BY5). The falling trend is most pronounced in the Bothnian Bay, while it decreases southward down to the central part of the Baltic Proper. In the deeper layers the picture becomes quite complex. There are no significant trends in the Gulf of Bothnia while falling trends are found in the northwestern part of the Baltic Proper. Only the Eastern Gotland Basin reveals positive trends. Examination of the trend test for individual seasons after pooling the different depth intervals indicates that the observations are strongly correlated, hence the assumption of independent observations must be omitted. The Bothnian Bay shows significant falling silicate concentrations in the surface layers during all seasons (Table 2), with the exception of the period January-March. In the northern and southern Bothnian Bay the lack of observations coupled to the strong dependence of data yields simultaneous significant decreasing concentrations during only one season. All trends found are negative. The major inflow of deep salt water (Figure 2) in 1976 represents the start of a more stagnant period than the previous one. The silicate concentration in the water from the Kattegat is lower than the concentrations in the deep water of the Baltic Proper. This may have a major effect on the concentration evolution as can also be deduced from Figure 2. The time period was therefore split into two, 1968-1976 and 1978-1986. Independent analyses were carried out for each period (Table 3) though lack of data from the Gulf of Bothnia hampered any investigation in that region. In the Baltic Proper only BY5 showed significant trends in the surface layers (without the assumption of independent observations) for the latter period. BY15 has positive trends in the deeper parts in both periods but they increased during the latter. BY31 on the other hand has falling trends during the first period (though they are based on the assumption of independent observations). The latter period shows no significant trends.

5 . DISCUSSIONS The silicate distribution in the Baltic Sea has already been described by Voipio (1961) from extensive summer cruises in 1954-1959. A clear gradient in the surface layers from 30-60 mmol m-3 in the Bothnian Bay to 20-30 mmol m-3 in the Bothnian Sea to 10-20 mmol m-3 in the Baltic Proper was found. Small vertical variations are found in the Gulf of Bothnia. The sharp halocline in the Baltic Proper, on the other hand, delimits the surface layers from the deep water with concentrations in the range 40-70 mmol

Average Slope Pl

Average Slope

pl

Jul

-

Oct -

Dec

ns

12.0

9.8 -0.28 0.032

ns

11.5

ns

ns

10.8

0.000

11.2 -0.49

11.5 -0.92 0.002

ns

17.6

BY32

15.6 -0.47 0.052

13.1 ns

12.2 -0.70 0.002

ns

12.8

ns

17.4

SR5

11.7 -0.21 0.027

11.5 -0.52 0.007

ns

20.7

BY38

15.0 -1.61 0.073

ns

17.6

16.0 -1.04 0.040

ns

nd

Norihern Bothnlan sea

15.1 -1.39 0.035

-0.50 0.044

20.6

24.1 -4.01 0.013

ns

3.9

Bothnian Bay

Table 2. n e n d analysis of silicate data at different stations and season. Depth interval is 0-30 rn. Average concentration (mrnol m-3), slope (rnrnol rn-' yr-l) and p-values (two-sided test) are presented.

ns

12.5

11.5 -0.34 0.039

9.5

ns

11.0 -0.34 0.041

10.7 -0.68 0.000

11.7 -0.53 0.009

9.8 -0.91 0.005

8.0 -0.32 0.025

ns

17.8

BY31

7.6 -0.27 0.021

ns

15.8

BY27

ns

13.9

BY15

: independence not assumed ns : not significant nd : no data

S P

p1

Jun

Slope

Average

Apr

-

p1

17.5

Average Slope

Jan

Mar

BY5

Season

151

-

131

131

-

111

31

-

21

P2

Slope p'

Average

nd

nd

63.9 1.36 0.065 ns

M

55.6

ns

nd

ns

M

11.1

P2

nd

M M

14.3

~~

ns

M

10.5

Slope p'

Average

P2

Slope p'

Average

M

M

P2

12.7 0.641 0.051 ns

11.7 0.62 0.001 0.035

13.6

Slope p'

ns ns

10.8

Table 3 . (Continued on following page)

I

21

-

Average

ns

P2

11

M

p'

11

11.8 0.57 0.001 0.042

13.9

Slope

Average

69.2 3.24 0.003 0.033

ns

na

55.6

ns

IIS

10.2

ns ns

8.2

ns

ns

7.9

BY15 1978-86 1968-76 1978-86

BY5

1968-76

-

0

Interval (m)

~~

0.083

68.9 -1.77 0.030 ns ns

61.8

us ns

M

ns

9.7

59.8

~

ns

ns

9.5

ns

0.067

9.3 -1.32

68.0

ns

M

13.9

ns

M

13.7

ns na

13.8

BY27 1978-86

-1.49 0.024 0.045

1-76

ns

0.049

ns

M

59.6

M

64.4 -1.13

M

0.053

58.7

M

nr

11.2

-~

ns

M

10.2

M

M

10.2

0.009

62.9 -1 .50

ns ns

14.9

ns ns

14.6

ns

ns

14.8

BY31 1968-76 1978-86

ns ns

65.2

M

64.3 -1.28 0.099

M N

14.4

ns

ns

14.1

ns

M

13.9

~~~

M

M

58.6

M M

58.4

m

M

11.9

M M

10.6

M

ns

10.4

BY32 1968-76 1978-86

M

ns

14.5

nd

nd

M M

16.0

M

nd

nd

N

0.026

15.5 0.849

~~~

14.1 0.58 0.014 0.078

13.7 0.36 0.014 ns

BY38 1978-86

15.0 -0.22 0.047

1-76

_

_

~

~

~

nd

~

nd

nd

~ _ _ _ ~~

nd

nd

nd

nd

nd

nd

nd

nd

-0.96

66.3

65.3 -1.47 0.057 0.081

0.033 0.056

64.9 -1.09

N

m

61.0

N N

61.7

N

nr

60.6

N N

M

61.6

ns

64.7

0.088

nd

_

nd

0.031

0.006

84.0 3.49

80.5 4.75 0.001 0.034

~~

nd

nd

nd

M

65.7 -1.62 0.079

~~~~~~~

nd

nd

nd

M M

61.5

BY31 BY32 1968-76 1 9 7 8 4 6 1968-76 1978-86

P2

~~

_

nd

75.6 2.40 0.007 0.041

ns

71.1 1.27 0.065

BY27 1968-76 1978-86

0.043

_

nd

nd

nd

BY15 1968-76 1978-86

Slope P '

Average

_

P2

p'

nd

nd

nd

BY5 1968-76 1 9 7 8 4 6

nd

nd

nd

nd

nd

nd

BY38 1968-76 1 9 7 8 4 6

Table 3. Tkend analysis of silicate data at different stations and depth intervals. Analysis performed separately for two time periods. Average concentration (mmol m-3), slope (mmol m-3 yr-') and p-values (twesided test) are presented.

-

341

~

341

Slope

Averxe

241

-

p' P2

241

Slope

Average

191

-

p' P2

Slope

Average

191

-

171

Interval (m)

m

X

c3

-¶

0

U

zz

c3

2 76

P.

SANDEN,

L. RAHMAND F. WULFF

m-3. The same picture is found today but the surface values have decreased as is evident in Table 3. The trend test gave more significant results than could be expected to occur by chance. For this study, even without the assumption of independence, the test gave more than 25% significant results and for the trophogenic layer (0-30 m) this figure becomes 60%. From this we can conclude that the silicate concentrations in the surface waters has been decreasing over the last two decades in the whole system. The strongest trends are found in the Gulf of Bothnia. The cause of the observed decrease in surface layer concentration could be attributed to several different factors. As there is no evidence of increasing concentrations in the intermediate layers and if we assume there is no marked decrease in allocthonous input (Ah1 1989, pers. comm.), the cause has to be sought in silica turnover either in the sediments, in the deep water or in the trophogenic layer. The Eastern Gotland Basin, the largest but not the deepest of the Baltic Proper subbasins, may act as a silicate trap due t o its strong stratification. This stratification hampers the diapycnal fluxes of dissolved substances. The rapid development of anoxia after the last major inflow of saline water in 1976 is just one indication on this situation. This may be compared with the non-stagnant conditions in the deepest of the Baltic proper subbasins, the Landsort Basin. Because its deep water supply comes, due to its sill configuration, from the intermediary layers of the Eastern Gotland Basin, it is not so strongly stratified. In fact, the deep water is rather well-mixed and homogeneous. No positive trends in silicate are found either. The pool of silicate in the deep layers of the Eastern Gotland Basin (below 130 m) is of the same order of magnitude as that of the overlaying surface layers (0-30 m) (x10 mmol m-3 in the surface layer of volume zz 1.2 x 10l2 m3 yields 12 Gmol; ~ 7 mmol 0 m3 in the deep layers of volume -2.5 x 10" m3 gives 18 Gmol, see also Table 1). Further, the mean yearly increase in total amounts in the deep layers :.( 1.6 mmol m-3 yr-' yields 0.4 Gmol yr-') gives about the same ratio t o the observed decrease in the corresponding surface layers (Z0.2 mmol m-3 yr-' yields 0.24 Gmol yr-'). The annual decrease in the surface layers calculated for the whole northern and central Baltic proper, x 0.3 mmol m-3 yr-' in 3 x 10l2 m3, yields 0.9 Gmol yr-'. These subbasins do not show any accumulation in their deep waters. To conclude, in the Eastern Gotland Basin the loss in the surface pool does not balance the total amount accumulation rate in its deep water. On the other hand the latter will only be half the decrease estimated in the surface pool of the central and northern Baltic Proper. No increase in concentration has been observed in the deeper parts of the Gulf of Bothnia. Hence this silicate trap cannot account for the whole decrease. The sediments

TRENDTESTOF THE BALTIC

277

of this basin have not been studied and could possibly accumulate silica in larger amounts during the stagnation period. The second hypothesis is that primary production of silica forming algae is increasing. This can either be caused by increasing nitrogen loads (Larsson et al. 1985) or by a change in population dynamics of this type of algae. The former hypothesis is supported by the tendency towards increasing Si concentrations during the winter (Table 2) when the biological activity is low. This may be caused by the increasing turnover due to increased sedimentation and thereby increasing sediment efflux (Yamada and D'Elia 1984). In fact the net annual primary production estimated by Stigebrandt (1991) shows an increase of about 2% yr-' for the Bothnian Sea during this period. This should be compared with an observed decrease in Si concentration of about 4% yr-'. The corresponding increase in new production for the Baltic Proper is roughly 3% yr-' while the Bothnian Bay only comes up to 1%yr-'. This should be compared with roughly 3% and 4% falling trends for the Baltic Proper and Bothnian Bay respectively. One should be aware that these extrapolations of single stations t o whole subbasins, though supported by independent estimates (Wulff and Rahm 1988) and sampling studies (Wulff and Rahm 1989), have some degree of uncertainty. This hypothesis could, however, be corroborated by the analysis of specific seasons (Table 2); unfortunately the cold seasons have too few nonsignificant trends. The increased primary production will also lead t o an increased mineralization and consequently an increased oxygen consumption. The low oxygen levels result in growing bottoms without macrobenthic fauna. This has, in turn, lead to the formation of laminated surface sediments (Jonsson et al. 1990) in large parts of the Baltic Proper and the Bothnian Sea. Consequently the silicate efflux from these sediments should decrease due t o lack of bioturbation while the sediment storage should increase. Unfortunately, this has not yet been verified by field observations. The decrease in the different basins is however about the expected order except for the Bothnian Bay. The reason for the latter is still unknown. There are two main conclusions t o be drawn from this study. First that the non-parametric test for trend used here is useful for evaluating water quality records in the marine environment. The other is that falling trends in Si exist throughout the Baltic Sea, except for some deep layers. Neither of the suggested processes can by themselves explain all observations but they can be mutually supportive. However, only a careful hypothesis test including budget estimates of the whole Baltic Sea can solve these questions.

2 78

P. S A N D ~ L. N , RAHMAND F. WVLFF

6. ACKNOWLEDGEMENTS The authors thanks Annika Anderson and Miguel Rodriguez-Medina (Dept. of Systems Ecology, University of Stockholm) for their help with retrieving the data. This work has been financially supported by the Swedish Environmental Protection Agency.

REFERENCES Elmgren, R. (1989), “Man’s impact on the ecosystem of the Baltic Sea: Energy flows today and at the turn of the century”. Ambw 18, 326-332. Fransz, H.G.and J.H.G. Verhagen (1985), “Modelling research on the production cycle of phytoplankton in the Southern Bight of the North Sea in relation to riverborne nutrient loads”. Netherlands Journal of Sea Research 19, 241-250. HELCOM (1984), “Guidelines for the Baltic Monitoring Programme for the Second Stage”. Baltrc Sea Envrronmentol Proceedings 12. Hirsch, R.M. and J.R. Slack (1984), “A nonparametric trend test for seasonal data with serial dependence”. Water Resouroes Research 20, 727-732. Hirsch, R.M.,J.R.Slack and R.A. Smith (1982),“Techniques of trend analysis for monthly water quality data”. Water Resources Research 18, 107-121. Jonsson, P.,R. Carman and F. Wulff (1990), “Laminated sediments in the Baltic - a tool for evaluating nutrient mass balances”. Ambio 19, 152-158. Larsson U.,R. Elmgren and F. Wulff (1985), “Eutrophication and the Baltic Sea: causes and consequences”. Ambio 14, 9-14. Nehring, D., G. Aertebjerg, P. Alenius, V. Astok, S. Fonselius, M. Hannu, V. Tkevo, A. Troszinska, P. Tulkki and A.K. Yurkovskis (1987). “Nutrients”. In First Periodic Assessment of the State of the Manne Environment of the Baltrc Sea Area, 1980-1985. Background document, J. Lassig (ed). Baltic Sea Environment Proceedings 17B, 3581. Nehring, D., H.P. Hansen, L.A. Jhrgensen, D. Khmer, M. Mazamachs, A. Pertill, A. Trzosinska, F. Wulff and A. Yurkovskis (1990), “Nutrients”. In Second Periodic Assessment of the State of the Marine Environment of the Baltrc Sea Arm, 1980-1985. Background document, J. Lassig (ed). Baltic Sea Environment Proceedings 35B, 109152. Rosenberg, R.,R. Elmgren, S. Fleischer, P. Jonsson, G. Persson H. Dahlin (1990), “Marine eutrophication case studies in Sweden”. Ambio 19, 102-108. Schelske C.L.and E.F. Stroemer (1971), “Eutrophication, silica depletion and predicted changes in algal quantity in Lake Michigan”. Science 173, 423-424. Stigebrandt, A. (1991), “Computations of oxygen fluxes through the sea surface and the net production of organic matter with application to the Baltic adjacent areas”. Submitted. van Bell, G. and J.P. Hughes (1984), “Nonparametric tests for trend in water quality”. Water Re~ourcesResenrch 20, 127-136. van Bennekom, A.J., W.W.C. Gieskes and S.B. Tijssen (1975), “Eutrophication of Dutch coastal waters”. Proceedings of the Royal Society London B189, 359-374. Yamada, S.S. and C.F. D’Elia (1984), “Silicic acid regeneration from estuarine sediment cores”. Marine Ecology Progress Series 18, 113-118. Voipio, A. (1961), “The silicatein the Baltic Sea”. Ann. Acad. Scient. Fennican. A11 106, 1-15. Wulff, F. and L. Rahm (1988), “Long-term, seasonal and spatial variations of nitrogen, phosphorous and silicate in the Baltic: An overview”. Morine Environmental Research 26, 19-37. Wulff, F. and L. Rahm (1989), “Optimizing the Baltic sampling programme: The effects of using different stations in calculations of total amount of nutrients”. Beitrige zur Meereskunde 60, 61-66.