Modelling temporal trends of 137Cs and 99Tc concentrations in Fucus vesiculosus from the eastern Irish coastline

Marine Pollution Bulletin 62 (2011) 2337–2344

Contents lists available at SciVerse ScienceDirect

Marine Pollution Bulletin journal homepage: www.elsevier.com/locate/marpolbul

Modelling temporal trends of 137Cs and from the eastern Irish coastline

99

Tc concentrations in Fucus vesiculosus

S. Cournane a,⇑, L. León Vintró a, P.I. Mitchell a, C.A. McMahon b, K. Smith b, S. Long b a b

School of Physics, University College Dublin, Belfield, Dublin 4, Ireland Radiological Protection Institute of Ireland, 3 Clonskeagh Square, Clonskeagh Road, Dublin 14, Ireland

a r t i c l e

i n f o

Keywords: Fucus vesiculosus 137 Cs 99 Tc Seasonal variation Sellafield Modelling

a b s t r a c t Time series of 137Cs and 99Tc activity concentrations in the brown seaweed Fucus vesiculosus and seawater, gathered at three locations on the eastern Irish coastline during the period 1988–2008, have been modelled using a novel approach incorporating a variable uptake rate in the seaweed. Seasonal variations in the time series, identified using spectral analysis, were incorporated into the model which was used to determine transfer kinetic parameters and to predict 137Cs and 99Tc concentrations in seaweed, as influenced by levels in ambient seawater. An optimisation method combining evolutionary and grid search minimisation techniques was adopted to determine the best values for the model parameters, from which concentration factors (CF) and biological half-lives (tb1/2) for 137Cs and 99Tc in F. vesiculosus were calculated. CF values of 170–179 and 1.1  105 l kg1 (dry weight) were obtained for 137Cs and 99Tc, respectively, while the corresponding tb1/2 values were 39–47 and 32 days, respectively. Ó 2011 Elsevier Ltd. All rights reserved.

1. Introduction 1.1. Seaweed as a bio-indicator The brown algae Fucus vesiculosus is known to effectively concentrate radionuclides and heavy metals from seawater and, as a result, has been widely used as a bio-indicator in marine and estuarine pollution studies (Fuge and James, 1974; Germain et al., 1995; Keogh, 2006; Kershaw et al., 2005). These studies have shown that analysis of radionuclide and heavy metal concentrations in seaweed can provide an accurate picture of concentrations in the surrounding waters (Black and Mitchell, 1952; Bryan and Hummerstone, 1973; Burger et al., 2006; Carlson and Erlandsson, 1991; Morita et al., 2010; Nawakowski et al., 2004). Furthermore, concentrations in archived seaweeds have been used to hindcast trends in the discharge rate of radionuclides where there is incomplete discharge information from their principal sources (Aarkrog et al., 1995; Fievet and Plet, 2003; Raisbeck et al., 1995). The increased sensitivity that can be achieved using Fucus species, owes to its high concentration factor for many different pollutants. Indeed, data relating to the uptake by some common Fucus species of certain trace contaminants have shown F. vesiculosus to have the greatest degree of accumulation when compared to other members of the fucoid family (McCartney and Rajendran, 1997). A concentration factor (CF, l kg1), defined as the ratio of the con⇑ Corresponding author. Tel.: +353 1 416 2716. E-mail address: [email protected] (S. Cournane). 0025-326X/$ - see front matter Ó 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.marpolbul.2011.08.039

centration in fresh seaweed (Bq kg1) to that in filtered seawater (Bq l1), is employed to describe the degree of accumulation in seaweed relative to seawater, with CF values varying with different species and contaminants. For F. vesiculosus, CF values for 137Cs and 99Tc have been reported to be approximately 32 and 3  104 l kg-1 (wet weight), respectively (Kershaw et al., 2005). A potential drawback of the use of bioindicators for monitoring purposes is the reported seasonality in the uptake, irrespective of water concentrations (Fuge and James, 1974; Kershaw et al., 1999; Nawakowski et al., 2004; Riget et al., 1995; Villares et al., 2002). In some cases, higher concentrations have been observed to occur during the winter months, when seaweed undergoes a period of slow growth. In F. vesiculosus, this effect has been reported for 99Tc concentrations (Kershaw et al., 1999) and 54Mn (Carlson and Erlandsson, 1991). In contrast, 137Cs in F. vesiculosus appear to follow an opposite trend, with the radionuclide accumulating to a higher extent in the seaweed during the early summer months (Carlson and Erlandsson, 1991). Increased concentrations during the summer period have been attributed to increased contaminant levels in receptacles and new growth in the algae. It is reported that at this time, the receptacles are mature and accumulate Cl-linked monovalent cations, such as Na+ and K+ (Munda and Hudnik, 1986), which could also explain the observed increase in the uptake of the Cs+. While divalent heavy metal ions such as Cd2+, Cu2+, Zn2+, Pb2+, Cr3+ and Hg2+ are taken up by passive transport (Davis et al., 2003), there is some uncertainty as to whether K+ and Cs+ ions are taken into algal cells by active transport, by an electropotential or a Donnan system (Lewin, 1962).

2338

S. Cournane et al. / Marine Pollution Bulletin 62 (2011) 2337–2344

In this study, time series of 137Cs and 99Tc activity concentrations in F. vesiculosus and seawater (affected by discharges from Sellafield), gathered at different sampling locations on the eastern Irish coastline in the period 1988–2009 were measured and their seasonality analysed. A novel modelling approach, taking into account a variable uptake rate by the seaweed, was then adopted with a view to (i) identifying any possible seasonality in the uptake of these radionuclides; and (ii) calculating appropriate values for the respective concentration factors (CF) and biological half-lives (tb1/2).

137

Cs levels in each sample were determined by high-resolution gamma spectrometry using an n-type germanium detector with relative efficiency of 30% and resolution of 1.70 keV (FWHM) at 1.33 MeV. The counting times for the seawater and seaweed samples were typically 1 and 3.5 days, respectively. 99Tc concentrations in seaweed and ambient (filtered) seawater were analysed using a radiochemical separation technique in accordance with the method described by Harvey et al. (1991), followed by beta spectrometry using a gas-flow proportional counter. 2.2. Analytical and optimisation techniques

2. Materials and methods 2.1. Sampling and analysis Sampling of seawater and seaweed was undertaken by staff of the Radiological Protection Institute Ireland (RPII) as part of their programme of radioactivity monitoring of the Irish marine environment. Adult individuals of F. vesiculosus, as well as seawater samples, were collected from three separate locations (Greenore, Balbriggan and Bull Island, see map in Fig. 1) over a period of 20 years from 1988 to 2008 inclusive. Seaweed samples were collected on a regular basis, typically once a month per location, while seawater samples were collected less frequently, typically every 2– 3 months per sampling site. Fucus individuals, having reached the Gametophyte stage, were collected at low tide and washed in nearby seawater to remove any extraneous material. Samples (250 g) were then oven dried at 80 °C for 24 h, ground, and thoroughly homogenised. For seawater, 137Cs was separated from the seawater (50 l) using ASG, a caesium exchanger consisting of silica gel impregnated with ammonium molybdophosphate. The water was initially filtered through a (0.45 lm) filter and adjusted to pH 2 with nitric acid prior to passing it through ASG exchange resin. The ASG resin was then placed in a well-defined and calibrated geometry prior to radiometric measurement (Baker, 1975).

2.2.1. The Lomb periodogram Spectral analysis of each time series was achieved through use of the Lomb–Scargle periodogram (Lomb, 1976; Scargle, 1982), a spectral technique for detecting periodicities in unevenly sampled time series. The periodogram performs a general linear least squares regression of the data to a sine/cosine series of different frequencies. The Lomb normalised periodogram, with spectral power dependent on angular frequency with x  2pf > 0 and frequency f, is defined by: 8h i2 hP i2 9 > P ðh  hÞ >  cosðxðt  sÞÞ  = j j j j ðhj  hÞ sinðxðt j  sÞÞ 1 < P P N ð xÞ  þ P 2 2 ðxðt  sÞÞ > cos 2 r2 > j : ; j j sin ðxðt j  sÞÞ ð1Þ

where s is defined by the equation:

P tanð2xsÞ  P

j

j

sinð2xt j Þ ; cosð2xt j Þ

ð2Þ

where r2 is the variance of (hj  h), and h and hj are defined to be the mean concentration value and the individual concentration values measured at time tj, respectively. The periodogram is a normalised function due to the inclusion of r2, with the significance of the frequencies exhibiting an exponential probability distribution. The probability that P(x) will be

Fig. 1. Sampling locations on the east coast of Ireland (Balbriggan, Bull Island and Greenore,) of Fucus vesiculosus and ambient seawater over a period of 20 years from 1988 to 2008, inclusive.

S. Cournane et al. / Marine Pollution Bulletin 62 (2011) 2337–2344

between some positive values z and z + dz is given by exp(z)dz. Where M independent frequencies are scanned, the probability that none return values larger than z is given by (1  ez)M, and thus, the false-alarm probability of the null hypothesis, related to the significance level of any peak in PN(x), is given by Eq. (3). A small value for the false-alarm probability is indicative of a highly significant period.

Pð> zÞ  1ð1  ez ÞM :

ð3Þ

In such cases where the data is equally spaced and the sampled frequencies fill the frequency range from 0 to the Nyquist frequency, fc, given as:

fc 

1 ; 2D

ð4Þ

where D is the time sampling interval, M is reported to be approximately equal to N, the number of data points (Horne and Baliunas, 1986). 2.2.2. Genetic algorithms Genetic algorithms (GA) are optimisation techniques based on the concepts of natural selection, genetics and evolution, and have been applied successfully in a wide range of disciplines (Haldenbilen and Ceylan, 2005; Ng and Perera, 2003; Nyarko and Scitovski, 2004; Wang and Wu, 2004). GA can be classified as a particular type of evolutionary algorithms that make use of concepts inspired by evolutionary biology such as fitness, inheritance, mutation, selection and crossover (also called recombination). The genetic algorithm process begins with an initial population where each individual of the population is represented by a randomly generated set of parameters. These sets of parameters provide solutions, in the given search space, which are characterised by an objective ‘fitness’ function that evaluates the goodness of fit, termed the ‘fitness’, for each individual. In keeping with the concept of natural selection, the fittest individuals, providing the most optimal solutions, are selected as parent individuals for the next generation and, subsequently, are randomly paired or ‘mated’ with each other. The next generation individuals are formed through a crossover operation, whereby the model parameters of the paired parent individuals are combined and exchanged to produce two new parameter sets, and thus, the next generation retains some of the fitter characteristics of the parents. In addition, a mutation operator is employed to change or ‘mutate’ a randomly selected small percentage of the new generation model parameters, in order that the complete search space be explored, thereby preventing the convergence to local non-optimum solutions. Once a new generation has been formed, the fitness of each individual is evaluated and the cycle continues, through several generations, until the overall fitness of the population converges to a level whereby no further improvement in the optimisation ‘fitness’ function occurs, thus providing the optimum solution parameters. 2.2.3. Method of least squares and the grid search method The method of least squares determines the values of the parameters aj that minimise the function,

6qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi7 27 N 6 6 ðeðxi ÞÞ  f ðxi ÞÞ 7 1X 4 5; N i¼0 eðxi Þ

ð5Þ

where e(xi) and f(xi) are the experimental and fitted values, respectively, and N is the number of data points in the time series. In this study, Eq. (5) was used as the fitness function of the GA to compare the difference between the radionuclide concentration measurements and the modelled data.

2339

Once the GA technique had been used to explore the whole search space and find non-local optimum parameters, a grid search method was used to further refine the search for optimum parameters. If the parameters are independent of each other, the optimum values for the least squares fit can be determined by minimising the fitness function with respect to each parameter separately. With successive iterations, locating the local minimum for each parameter in turn, the absolute minimum is found. A disadvantage of this method arises when parameters are not independent of each other leading to slow convergence towards the minimum. The uncertainty (±r) in each parameter (aj) is the change (Daj) which gives a value of the minimum cost function + 1/N (Bevington and Robinson, 1992). 3. Time series analysis and modelling 3.1. Detrending data for spectral analysis 137 Cs time series were prepared for spectral analysis by the application of a regression method to detrend the data. The detrending of data is a statistical manipulation that attempts to remove long-term biases and other distortions that can occur in analysis, allowing for more focus on the short-term fluctuations in the data. A least squares regression of the form,

CðtÞ ¼ a  ebt

ð6Þ

was applied to the data, where C(t) is the time dependent trend-line and a and b are constants. The equation was taken to be of this form given the almost exponential average decrease in 137Cs concentrations in seawater observed over time in the eastern Irish Sea (Mitchell et al., 1999). The time-series was then divided by this trend-line in preparation for the implementation of the Lomb periodogram, which calculates the dominant spectral components. A similar exponential regression and detrending technique has previously been successfully applied in the case of 137Cs activity concentrations following the Chernobyl disaster (Viswanathan et al., 2000). 3.2. Incorporation of a seasonal factor in a continuous compartmental model A novel approach was used to incorporate a seasonal factor into the simple classical compartmental model described by Fievet and Plet (2003). In this work, two similar models are proposed to describe seasonal variation, one employing an oscillatory input and the other simulating an oscillatory uptake rate. The classical single compartmental model, describing uptake by a biological compartment, can be represented as a first order differential equation (Fievet and Plet, 2003) in the form:

dCðtÞ ¼ ki  C s ðtÞ  ko  CðtÞ dt

ð7Þ

where C(t) and Cs(t) denote the radionuclide concentration inside and outside the biological compartment as a function of time (t) and ko and ki denote the rate of radionuclide output and input, respectively. In the case where we consider the seawater, Cs(t), we use:

C s ðtÞ ¼ C 0 þ C 1 sinðxtÞ

ð8Þ

with C0 and C1 considered to be the average seawater concentration and constant amplitude, respectively, and x the period of oscillation. For constant background concentration, the solution is proportional to exp(kot), and thus, in the case of a time-dependant Cs(t), a solution of the form C(t) = A(t) exp(kot) is sought. It is therefore necessary to determine A(t), and substituting this into Eq. (7) returns

2340

S. Cournane et al. / Marine Pollution Bulletin 62 (2011) 2337–2344

dAðtÞ ¼ ki C 0 eko t  ki C 1 sinðxtÞeko t ; dt

ð9Þ

which can be integrated over the interval t’  [0,t],

AðtÞ ¼ ki C 0

Z

t

0

0

eko t dt  ki C 1

0

Z

0

t sinðxtÞeko t dt

0

ð10Þ

0

to return the solution:

þ ko eko t sinðxtÞÞ:

ð11Þ

By multiplying by exp(kot) and assuming that kot  1, so that the transient terms can be neglected, C(t) becomes:

ki C 0 ðtÞ ki C 1 þ 2 ðko sinðxtÞ  x cosðxtÞÞ ko ko þ x2

ð12Þ

ki C 0 ðtÞ ko ki c 1 þ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ðsinðxt  /ÞÞ ko 2 ko þ x2

ð13Þ

or

CðtÞ ¼

where tan(/) = x/ko. This equation is used for the case of the oscillatory seawater input where ki/ko can be considered the time-averaged concentration factor (CF) over the period x, while C1 is related to the amplitude of the oscillation of the ambient seawater concentration. For the case of a seasonally varying uptake rate, Eq. (7) becomes:

dC i ðtÞ ¼ C 0 ðtÞðki þ Dki sinðxtÞÞ  ko  C i ðtÞ dt

ð14Þ

where Dki = ki C1/C0(t), yielding the solution:

0

4. Results and discussion 4.1.

ki C 0 ko t ki C 1 AðtÞ ¼ ðe  1Þ þ 2 ðx  xeko t cosðxtÞ ko ko þ x2

CðtÞ ¼

ing the cost function with respect to the parameters independently and alternately until convergence of all the parameters to their optimum value occurred.

1

k o Dk i B ki C CðtÞ ¼ C 0 ðtÞ@ þ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ðsinðxt  /ÞÞA ko 2 ko þ x

ð15Þ

Finally, the biologically half-life can be estimated using the parameter ko, similar to the simple compartmental case (Fievet and Plet, 2003). 3.3. Implementing the GA optimisation technique Genetic algorithms were employed to find the best-fit parameters of a function modelling the variations in time series 137Cs and 99 Tc activities in F. vesiculosus due to transfers from a corresponding time series in the ambient seawater using the proposed model. Each individual solution of the model was represented by a binary string of 45 bits with each of the 3 parameters encoded as a 15 bit binary number, respectively. The initial randomly generated population was set at 100 while the simple biased roulette wheel method was employed as a selection process with the fitness function defined as 10,000  v2 (Goldberg, 1989). Increased values of the fitness function require low values of the chi-squared cost function and, thus, imply a better fitted model. The implemented mutation rate was 2% with a two-point crossover operator used in the recombination process. This mutation rate was established as it proved adequate to both explore the complete search space and allow for convergence of the parameters. Termination of the algorithm was determined when the mean and standard deviation of the fitness of the whole population reached appropriate values. This ensured the increase of the average population fitness and the convergence of the parameters of the whole population to an optimal solution. In addition, to ensure the optimum parameters were achieved and to calculate their uncertainties, a grid search was used on the fittest individuals from the final generation, minimis-

137

Cs time series

The measured data consisted of concurrent time series of 137Cs concentrations in F. vesiculosus and ambient seawater sampled at three locations along the eastern Irish coastline (Fig. 1), respectively. The 137Cs measured concentrations in the seaweed and seawater, as well as the fitted trendlines used for detrending purposes at each location, are shown in Fig. 2 and Fig. 3, respectively. Increased concentrations of 137Cs were observed in F. vesiculosus during the summer months, in agreement with Carlson and Erlandsson (1991), showing the highest annual levels to occur in July and August and the minimum concentrations during winter. As expected, spectral analysis of the detrended time-series for the seaweed revealed a clear seasonal variation in the data, with a prominent period (significance >0.9) of one year for each sampling site (Fig. 4). For the seawater, depicted in Fig. 5, there was also evidence of a 1-year periodicity in the concentrations for each of the time-series. The probability, although still significant, was lower than that of the seaweed, ranging from 0.5 to 0.99, possibly due to the smaller number of measured data points. Nevertheless, the seasonally higher CF observed for samples taken in the summer and seasonally lower taken during the winter, calculated directly from concurrent 137Cs measurements in Fucus and seawater, indicate that the variation observed in the 137Cs concentration cannot be wholly attributed to the seawater variation. This CF variation for Balbriggan, Bull Island and Greenore ranged between 59 and 595, 68 and 385 and 493 and 53 l kg1 dry wt., respectively, with the higher values typically during the summer months and the lower CF values during winter. A possible explanation for the seasonal changes in 137Cs seawater concentrations could be the annual formation of a summer gyre in the western Irish Sea. The gyre is responsible for causing stratification within the water column which leads to the isolation of a dome of cold dense bottom water centred on the western Irish Sea mud patch during the summer months. As a result, less dense, low salinity, radionuclide rich waters, originating from the eastern Irish Sea, tend to remain close to the surface, leading to an inhomogeneous distribution of radionuclides within the water column (Leonard et al., 2004). Studies of 137Cs seawater distributions during the 1970s showed the surface concentrations to be between 1.3 and 1.5 times those of the bottom water (Jefferies et al., 1982). The predominant flow of water from the eastern Irish Sea exits via the North Channel; however, with the formation of a summer gyre some of the north-bound currents can be redirected to the western Irish Sea causing a southerly flow along the eastern Irish coastline (Aldridge et al., 2003; Hill et al., 1997; Horsburgh et al., 2000; Leonard et al., 1997). The combination of the stratification and redirection of radionuclide-rich currents previously observed may explain the annual periodicity detected from analysis of the seawater time series. Concentrations in Fucus were modelled using the variable uptake rate model, obtaining a best-fit solution for each time series through the use of a genetic algorithm and grid search methods. Fig. 2 presents the measured F. vesiculosus concentrations, model predictions and trendline for each of the sampling sites. The optimum parameters calculated for each time series, in addition to the regression coefficient and the annual CF and tb1/2 values are listed in Table 1. There is excellent agreement amongst the annual 137 Cs CF values calculated for the different sampling sites, with val-

2341

S. Cournane et al. / Marine Pollution Bulletin 62 (2011) 2337–2344

30

30

20

Trend-line

15

137Cs

(Bq kg-1 dry wt.) 137Cs

Variable Uptake Rate model

(Bq kg-1, dry wt.)

Measured data (Balbriggan)

25

10

25

Measured (Balbriggan)

20

Trendline

15 10 5

5 0 1988

0 1988 1992

1996

2000

2004

1992

1996

2000

2004

2008

Year

2008

35

Year

30

137Cs

(Bq kg-1 dry wt.)

30

Variable Uptake Rate model

25

Trend-line

20

(Bq kg-1, dry wt.)

Measured data (Bull Island)

25

137Cs

Measured (Bull Island)

35

10

Trendline

20 15

15

5

10

0 1988

1992

1996 Year

5

2000

2004

30 0 1988

1992

1996

2000

2004

25

35

Measured data (Greenore)

137Cs

Variable Uptake Rate model

25

137Cs

(Bq kg-1 dry wt.)

30

(Bq kg-1, dry wt.)

Year

Trend-line 20

Measured (Greenore) Trendline

20 15 10 5

15

0 1988

10 5 0 1988

1992

1996

2000 Year

2004

2008

Fig. 3. Measured 137Cs concentrations and trendline of seawater at three different sampling locations: Balbriggan, Bull Island and Greenore.

1992

1996

2000

2004

2008

Year 137

Fig. 2. Measured and modelled Cs concentrations, and trendline in Fucus vesiculosus from the three different sampling locations: Balbriggan, Bull Island and Greenore, with the error bars representing the 95% confidence interval.

ues ranging from 171 to 179 l kg1 dry wt. In order to compare this range with reported values, it is necessary to express the CF in the same units by multiplying by a factor of 0.18, given as the mean dry:wet weight ratio for the samples (Radiological Protection Institute of Ireland, 1988–2008). This returns a CF range of approximately 30.7–32.2 l kg1 fresh weight, comparable to 32 l kg1 found in other studies (Kershaw et al., 2005). In the case of the biological half-life, there is also good agreement amongst the different sampling sites, with the values ranging from 39 to 47 days. Boisson et al. (1997), in controlled F. vesiculosus depuration experiments at 2 and 12 °C, determined 134Cs biological half-life values of 96 and 54 days respectively, of the same order as those computed in this study (temperature in Irish coastal waters typically range from 6 °C in winter to 14 °C in later summer). The phase shift parameter (/) gives an indication as to when the uptake rate of F. vesicu-

losus approaches a yearly maximum and also when the rate is at its lowest. In this study, the phase shift of the sine function (/), ranging from 0.77 to 0.86, suggests that the Fucus experiences an annual uptake rate maximum during the late spring/early summer months and minima during the late autumn/early winter months, in agreement with the findings of Carlson and Erlandsson (1991). Interestingly, both the seaweed and seawater data sets display maximum yearly activity levels in the later summer months, in July and August,when the seasonal gyre is prominent, while the rate of uptake of 137Cs by the Fucus is at a maximum in the earlier summer months. A Digital Signal Processing (DSP) inspired model, as detailed by Fievet and Plet (2003), was also employed for each of the time series, with the optimum solution obtained using the Solver™ tool of MicroscoftÒ Excel™. The CF and biological half-life values returned by the DSP method ranged between 137 and 168 l kg1 fresh weight and 78–280 days, respectively, having a much larger range than those calculated using the variable uptake rate model. It is worth noting that the DSP model demands evenly spaced data as an input, and thus, an interpolation of the measured seawater and seaweed time series concentrations is necessary to provide a

2342

S. Cournane et al. / Marine Pollution Bulletin 62 (2011) 2337–2344

7 6

20 Balbriggan 99.99% 99.5% 95% 50%

15 10 5

Power Spectrum (PN(T))

Power Spectrum (PN(T))

25

0 0

1 Period (T)

Balbriggan 90% 50% 10% 1%

5 4 3 2 1 0

2

0

1

2 Period (T)

10 9 Bull Island 99.99% 99% 75% 25%

15 10 5

Power Spectrum (PN(T))

Power Spectrum (PN(T))

20

8 Bull Island 99% 95% 50% 5%

7 6 5 4 3 2 1

0 0

1 Period (T)

0

2

1

2 Period (T)

10

9

9

8 Greenore 99% 90% 25% 1%

7 6 5 4 3 2

Power Spectrum (PN(T))

10

Power Spectrum (PN(T))

0

8 Greenore 97.50% 90% 50% 5%

7 6 5 4 3 2 1

1

0

0 0

1

2 Period (T)

0

1

2 Period (T)

Fig. 4. Power spectrum of the 137Cs Fucus vesiculosus concentration time series with significance levels from the three different sampling locations: Balbriggan, Bull Island and Greenore.

Fig. 5. Power spectrum of the 137Cs seawater concentration time series with significance levels from the three different sampling locations: Balbriggan, Bull Island and Greenore.

constant sampling interval T of 1 month. This is the most likely cause of the discrepancies between the calculated CF and tb1/2 values returned by the two models given that the number of points per year in each time series is low. The half-life values calculated using the variable uptake rate model proposed in this study should not be influenced by the sampling period of data and hence, the model presents as a very useful tool where limited time series data are available.

surface and bottom waters in the western Irish Sea have reported higher 99Tc concentrations in surface water samples, being on average 1.3-fold greater than of the bottom waters during summer months. The inhomogeneous distribution of radionuclides is said to be due to the stratification of water caused by the presence of the gyre, an effect which may explain the 1-year periodicity of the seawater (Leonard et al., 1997). It should, however, be noted that following the commissioning of the Enhanced Actinide Removal Plant (EARP) in 1994, there were significant increases in 99 Tc discharges from Sellafield, with great fluctuations witnessed in monthly discharges during the period 1994–1998 (Leonard et al., 2004). These increased discharges offer an explanation for the other significant periods witnessed in both the Fucus and seawater time series. Further spectral analysis of the monthly discharges during the period 1994–2004 returns no significant annual spectral component, suggesting the 1-year periodicity evident in the Balbriggan seawater time series to be as a result of the seasonal pool gyre rather than fluctuations in monthly discharges. Concentrations of 99Tc in F. vesiculosus were modelled using the variable uptake rate model with the best-fit solution parameters

4.2.

99

Tc time series

In the case of 99Tc, the measured experimental data consisted of concurrent time series of 99Tc concentrations in F. vesiculosus and ambient seawater sampled from Balbriggan during the period 1995–2006, inclusive. The measured 99Tc concentrations in the seaweed and seawater at each location are shown in Figs. 6 and 7, respectively. The Lomb normalised periodogram was used to perform spectral analysis on both 99Tc time series. A 1-year periodicity with a probability of 0.75 was observed in the case of the seawater; however, the spectrum also contained a number of other significant peaks. Studies on the distribution of 99Tc between

2343

S. Cournane et al. / Marine Pollution Bulletin 62 (2011) 2337–2344 Table 1 The optimum parameters calculated for each

137

Cs time series, in addition to the regression coefficient and the annual CF and tb1/2 values.

Sampling site

ki

ko

Dki

/

CFs ± r, l kg1 (dry weight)

tb1/2 ± r, day

Regression coefficient (r)

Balbriggan Bull Island Greenore

1130 ± 69 1151 ± 110 927 ± 94

6.3 ± 0.5 6.5 ± 0.7 5.4 ± 0.5

97 ± 67 91 ± 63 209 ± 151

0.78 0.77 0.86

179 ± 15 177 ± 20 171 ± 18

40 ± 4 39 ± 4 47 ± 5

0.90 0.91 0.92

DSP model may be due to the relatively low number of concurrent data points, averaging approximately 5.5 per year.

80

Measured data (Bull Island)

70

99Tc

(mBq l-1)

60

5. Conclusions

50 40 30 20 10 0 1995

1997

1999

2001

2003

2005

2007

Year Fig. 6. Measured sampling site.

99

Tc (±r) seawater concentrations collected from the Balbriggan

10000

Measured data (Balbriggan)

99 Tc

(Bq kg-1 dry wt.)

8000

Variable Uptake Rate model 6000

4000

2000

0 1995

1997

1999

2001

2003

2005

2007

Year Fig. 7. Measured and modelled 99Tc (±r) concentrations of Fucus vesiculosus collected from the Balbriggan Island sampling site.

calculated by employing a genetic algorithm and grid search method. The measured and modelled F. vesiculosus concentrations are presented in Fig. 7, where the optimum parameters calculated were: ki = 8.8  105 ± 1.1  105, ko = 7.9 ± 0.9 and Dki = –3.1  105 ± 1.9  105. The CF values for F. vesiculosus using the variable uptake rate model were calculated to be 1.1  105 ± 0.15  105 l kg1 dry wt. or 2  104 ± 0.23  104 l kg-1 wet weight, which is in good agreement with a previously reported value of 3  104 l kg-1 wet weight (Kershaw et al., 2005). Similarly, the biological half-life for 99Tc, calculated to be 32 ± 5 days, is also in good agreement with that (39 days) reported by Fievet and Plet (2003) for 99Tc, albeit for Fucus serratus. The calculated phase shift value suggests the maximum uptake rate to occur in the winter months in good accord with previous studies (Kershaw et al., 1999). The DSP model was also employed for the Balbriggan 99Tc time series, returning CF and biological half-life values of 1.21  105 l kg1 dry weight or 2.2  104 l kg-1 wet weight and 144 days, respectively. While the CF values calculated by each model in this study are in good agreement, the biological half-life values differ substantially. Once again, the larger tb1/2 value returned by the

Seasonal variations of 137Cs concentrations in F. vesiculosus have been identified and incorporated into a variable uptake rate model using genetic algorithms as an optimisation method. Spectral analysis of the time series provides evidence of significant 1-year seasonal variations in the concentrations, most likely as a result of the formation of the seasonal western Irish Sea gyre, which transports radionuclide-rich seawater to the sampling sites during the later summer months. Implementation of the variable uptake rate model also suggests there to be an inherent variation in the uptake of 137 Cs by F. vesiculosus, peaking in the late spring/early summer months. This observed seasonality can be attributed to the development of new vegetation in the algae, resulting in higher levels in the receptacles at this time. Modelling of the data yielded 137 Cs CF values of 30.7–32.2 l kg1 fresh weight and a biological half-life of 39–47 days for the seaweed, showing good agreement with other studies (Kershaw et al., 1999). Similar analysis was conducted on the 99Tc time-series for seaweed and seawater, with no apparent seasonal spectral signal observed for F. vesiculosus, while a significant 1-year variation was evidenced in the seawater. Once again, the gyre is charged at causing the flow of radionuclide-rich currents southwards along the eastern Irish coastline, an effect which explains the observed periodicity. The variable uptake rate model, using genetic algorithms as an optimisation technique, was employed to calculate a 99Tc CF value of 2  104 l kg1 wet weight, and a biological half-life of 32 days for F. vesiculosus. Accordingly, a novel model incorporating a variable uptake rate has been successfully used to simulate the seasonal variations of 137 Cs and 99Tc levels in F. vesiculosus. It is important to account for such seasonality in Fucus as these levels can fluctuate by up to a factor of 1.8 in a given year, a change which may not necessarily be followed to the same extent in the ambient seawater. Furthermore, this model may prove useful for modelling concentrations levels in other bioindicators where seasonal variations in uptake are exhibited. Acknowledgements We gratefully acknowledge Dr. Zoltán Neufeld of the School of Mathematical Sciences, UCD for his assistance with the mathematical modelling. References Aarkrog, A., Dahlgaard, H., Hansen, H., Holm, E., Hallstadius, L., Rioseco, J., Christensen, G., 1995. Radioactive tracer studies in the surface waters of the northern North Atlantic including the Greenland, Norwegian and Barents Seas. Rit Fiskideildar 9, 37–42. Aldridge, J.N., Kershaw, P., Brown, J., Young, E.F., McCubbin, D., Leonard, K.S., 2003. Transport of Plutonium (239,240Pu) and Caesium (137Cs) in the Irish Sea: comparison between observations and results from sediment and contaminant transport modelling. Continental Shelf Research 23 (9), 869–899. Baker, C.W., 1975. The determination of radiocaesium in sea and fresh waters. Ministry of Agriculture, Fisheries and Food, Directorate of Fisheries Research, Lowestoft, Aquatic Environment Protection: Technical Report Number 16.

2344

S. Cournane et al. / Marine Pollution Bulletin 62 (2011) 2337–2344

Bevington, P.R., Robinson, D.K., 1992. Data Reduction and Error Analysis for the Physical Sciences, 2nd ed. McGraw-Hill, New York. Black, W.A.P., Mitchell, R.L., 1952. Trace elements in the common brown algae and seawater. Journal of Marine Biological Association of the UK 30, 575–584. Boisson, F., Hutchins, D.A., Fowler, S.W., Fisher, N.S., Teyssie, J.-L., 1997. Influence of temperature on the accumulation and retention of 11 radionuclides by the marine alga Fucus vesiculosus. Marine Pollution Bulletin 35 (7–12), 313–321. Bryan, G.W., Hummerstone, L.G., 1973. Brown seaweed as an indicator of heavy metals in estuaries in south-west England. Journal of Marine Biological Association of the UK 53, 705–720. Burger, J., Gochfeld, M., Kosson, D.S., Powers, C.W., Jewett, S., Friedlander, B., Chenelot, H., Volz, C.D., Jeitner, C., 2006. Radionuclides in marine macroalgae from Amchitka and Kiska Islands in the Aleutians: establishing a baseline for future biomonitoring. Journal of Environmental Radioactivity 91 (1–2), 27–40. Carlson, L., Erlandsson, B., 1991. Seasonal variation of Radionuclides in Fucus vesiculosus L. from the Öresund, Southern Sweden. Environmental Pollution 73, 53–70. Davis, T.A., Volesky, B., Mucci, A., 2003. A review of the biochemistry of heavy metal biosorption by brown algae. Water Research 37, 4311–4330. Fievet, B., Plet, D., 2003. Estimating biological half-lifes of radionuclides in marine compartments from environmental time-series measurements. Journal of Environmental Radioactivity 65, 91–107. Fuge, R., James, K.H., 1974. Trace metal concentrations in Fucus from the Bristol Channel. Marine Pollution Bulletin 5, 9–12. Germain, P., Leclerc, G., Simon, S., 1995. Transfer of polonium-210 into Mytilus edulis (L) and Fucus vesiculosus (L.) from the baie de Seine (Channel coast of France). The science of the total environment 164, 109–123. Goldberg, D.E., 1989. Genetic Algorithms in Search, Optimization, and Machine Learning. Addison–Wesley, University of Alabama. Haldenbilen, S., Ceylan, H., 2005. Genetic algorithm approach to estimate transport energy demand in turkey. Energy Policy 33, 89–98. Harvey, B.R., Ibbett, K.D., Williams, J.K., Lovett, M.B., 1991. The determination of technetium-99 in environmental materials. Aquatic Environment Protection: Analytical Methods Number 8, Directorate of Fisheries Research, Lowestoft. Hill, A.E., Brown, J., Fernand, L., 1997. The summer gyre in the western Irish Sea: shelf sea paradigms and management implications. Estuarine, Coastal and Shelf Science 44 (A), 83–95. Horne, J.H., Baliunas, S.L., 1986. A prescription for period analysis of unevenly sampled time series. Astrophysical Journal 302, 757–763. Horsburgh, K.J., Hill, A.E., Brown, J., Fernand, L., Garvine, R.W., Angelico, M.M.P., 2000. Seasonal evolution of the cold pool gyre in the western Irish Sea. Progress in Oceanography 46, 1–58. Jefferies, D.F., Steele, A.K., Preston, A., 1982. Further studies on the distribution of 137 Cs in British coastal waters – I. Irish Sea. Deep Sea Research 29, 713–738. Keogh, S., 2006. Carbon-14 and iodine-129 in the irish marine and terrestrial environment. PhD Thesis. National University of Ireland, Dublin, 101p. Kershaw, P.J., McCubbin, D., Leonard, K.S., 1999. Continuing contamination of north Atlantic and Arctic waters by Sellafield radionuclides. Science of the Total Environment 237, 119–132. Kershaw, P.J., McMahon, C.A., Rudjord, A.L., Smedley, C., Leonard, K.S., 2005. Spatial and temporal variations in concentration factors in NW European Seas – secondary use of monitoring data. Radioprotection 40 (1), 93–99.

Leonard, K.S., McCubbin, D., Brown, J., Bonfield, R., Brooks, T., 1997. Distribution of technetium-99 in UK coastal waters. Marine Pollution Bulletin 34 (8), 628– 636. Leonard, K.S., McCubbin, D., McDonald, P., Service, M., Bonfield, R., Conney, S., 2004. Accumulation of technetium-99 in the Irish Sea? Science of the Total Environment 322 (1–3), 255–270. Lewin, R.A., 1962. Physiology and Biochemistry of Algae. Academic Press, New York, London, 929. Lomb, N.R., 1976. Least-squares frequency analysis of unequally spaced data. Astrophysics and Space Science 39, 447–462. McCartney, M., Rajendran, K., 1997. 99 Tc in the Irish Sea: recent trends. Radioprotection – Colloques 32 (c2), 359–364. Mitchell, P.I., Condren, O.M., León-Vintró, L., McMahon, C.A., 1999. Trends in plutonium, americium and radiocaesium accumulation and long-term bioavailability in the western Irish Sea mud basin. Journal of Environmental Radioactivity 44 (2–3), 223–251. Morita, T., Fujimoto, K., Kasai, H., Yamada, H., Nishiuchi, K., 2010. Temporal variations of 90Sr and 137Cs concentratinos and 137Cs/90Sr activity ratio in marinebrown algae, Undaria pinnnatifida and Laminaria longissima, collected in coastal areas of Japan. Journal of Environmental Monitoring 12, 1179–1186. Munda, I.M., Hudnik, V., 1986. Growth responses of Fucus vesiculosus to heavy metals, singly and in dual combinations, as related to accumulation. Botanica Marina 29, 401–412. Nawakowski, C., Nicholson, M.D., Kershaw, P.J., Leonard, K.S., 2004. Modelling99 Tc concentrations in Fucus vesiculosus from the north-east Irish Sea. Journal of Environmental Radioactivity 77 (2), 159–173. Ng, A.W.M., Perera, B.J.C., 2003. Selection of genetic algorithm operators for river water quality model calibration. Artificial Intelligence 16, 529–541. Nyarko, E.K., Scitovski, R., 2004. Solving the parameter identification problem of mathematical models using genetic algorithms. Applied Mathematics and Computation 153, 651–658. Radiological Protection Institute of Ireland, 1988–2008. Radioactivity Monitoring of the Irish Marine Environment 1988–2008. Raisbeck, G.M., Yiou, F., Zhou, Z.Q., Kilius, L.R., 1995. 129 I from nuclear fuel reprocessing facilities at Sellafield UK and La Hague (France); potential as an oceanographic tracer. Journal of Marine Systems 6 (5–6), 561–570. Riget, F., Johansen, P., Asmund, G., 1995. Natural seasonal variation of cadmium, copper, lead and zinc in brown seaweed (Fucus vesiculosus). Marine Pollution Bulletin 30 (6), 409–413. Scargle, J.D., 1982. Studies in astronomical time series II: statistical aspects of spectral analysis of unevenly spaced data. The Astrophysical Journal 263, 835– 853. Villares, R., Puente, X., Carballeira, A., 2002. Seasonal variation and background levels of heavy metals in two green seaweeds. Environmental Pollution 119, 79–90. Viswanathan, G.M., Buldyrev, S.V., Garger, E.K., Kashpur, V.A., Lucena, L.S., Shlyakhter, A., Stanley, H.E., Tschiersch, J., 2000. Quantifying nonstationary radioactivity concentration fluctuations near Chernobyl: a complete statistical description. Physical Review E 62, 4389–4392. Wang, H.F., Wu, K.Y., 2004. Hybrid genetic algorithm for optimization problems with permutation property. Computers and Operations Research 31, 2453– 2471.