The disappearance of the European eel from the western Wadden Sea

Journal of Sea Research 66 (2011) 434–439

Contents lists available at SciVerse ScienceDirect

Journal of Sea Research journal homepage: www.elsevier.com/locate/seares

The disappearance of the European eel from the western Wadden Sea Jaap van der Meer ⁎, Henk W. van der Veer, Johannes IJ. Witte NIOZ, Royal Netherlands Institute for Sea Research, P.O. BOX 59, 1790 AB Den Burg, The Nertherlands

a r t i c l e

i n f o

Article history: Received 28 February 2011 Received in revised form 30 August 2011 Accepted 30 August 2011 Available online 22 September 2011 Keywords: Wadden Sea Extinction Fisheries Population Dynamics Dynamic Energy Budget Anguilla anguilla

a b s t r a c t A cohort model for the European eel is presented, which enables the interpretation of observed catches of yellow eel and silver eel in the western Wadden Sea in terms of recruitment data of glasseel. The model builds on various assumptions on length-dependent mortality and silvering rates and on the standard Dynamic Energy Budget (DEB) model which predicts length growth. DEB parameter values are estimated on the basis of literature data. The model predictions are generally in good agreement with the data, though the final decline in numbers in the 1980s occurs earlier than predicted. This suggests that the decrease in eel stock is not just a consequence of lower glasseel immigration but that local conditions must have impoverished, a phenomenon earlier observed in fresh water. © 2011 Elsevier B.V. All rights reserved.

1. Introduction Reconstructed long-term series on landings of the European eel Anguilla anguilla (L.) in Northern European countries show a steady decline since the mid 1960s (Dekker, 2004a). The immigration of glasseel, as indicated by long-term monitoring programmes in Sweden, Germany, the Netherlands and France, only started to decline in the early 1980s (Dekker, 2003). This difference in the onset of the decline, points to a decreasing yield per recruit, i.e. yield per immigrating glasseel, starting in the 1960s. In theory, a declining yield per recruit may be due to a severely increasing fisheries mortality resulting in growth overfishing. Long-term records on fishery effort and mortality are lacking for most of Europe, but the few series that are available, such as those for the IJsselmeer (sometimes translated as Lake Yssel), indeed point to a strong increase in the number of fykes used (Dekker, 2004b). A fyke is a fish trap consisting of a net suspended over a series of hoops, laid horizontally in the water, and is routinely used for catching eel. However, research surveys from the IJsselmeer, one of the largest fresh-water lakes of western Europe, show that undersized eel in the range between 10 and 25 cm, also started to decline in the 1960s, whereas the smallest size class, that is smaller than 10 cm, more or less followed the recruitment signal and only started to decline in the early 1980s. Recall that the arriving glasseel are around 7 cm long and that the minimum landing size is 28 cm. Hence, overexploitation can be ruled out as the main cause of the decline. In an extensive analysis, Dekker (2004a) also rejected other explanations for the observed declining trend, such as habitat loss, barriers to ⁎ Corresponding author. E-mail address: [email protected] (J. van der Meer). 1385-1101/$ – see front matter © 2011 Elsevier B.V. All rights reserved. doi:10.1016/j.seares.2011.08.007

migration, introduced parasites and changes in ocean climate. He concludes that the decline of the IJsselmeer stock is still a mystery. Not all glasseel arriving along the European shores migrate into fresh water, and a minor part remains in the coastal waters and estuaries during the entire yellow eel stage (Daverat et al., 2006; Cairns et al., 2009). At the end of this period, which may last up to 20 years, the animals mature, metamorphose and return as silver eels to their breeding grounds in the Sargasso Sea, without ever having entered fresh water. Here we present so far unpublised data on eel catches in a research fyke for the period 1960–2009 from such a coastal water, namely the shallow areas around the island of Texel in the western Wadden Sea. Apart from describing the temporal trend in abundance of this local population, we examine by using a quantitative model approach whether the fyke catches can be linked to the recruitment data of the neighboring monitoring site at the Den Oever sluices by simply assuming constant mortality and growth conditions during the observation period. If a similar decline in yield per recruit has occurred in the coastal waters as in the fresh water, such simple linkage would not be possible. This way the analysis aims to shed more light on the mystery of the declining yield per recruit in the fresh water habitat since the mid 1960s. Recently, Henderson et al. (in press) reported a population collapse over the last 30 years in Bristol Channel, another estuarine stock. Linking recruitment to catch is merely a matter of good bookkeeping of growth, size-related natural and fisheries mortality, and size-related escapement from the local population as silver eel. Eels catched in the research fyke were not aged, so we had to estimate growth in an indirect way. Basis of the growth model applied here is the standard Dynamic Energy Budget (DEB) model as developed by Kooijman (2010), and see also Van der Meer (2006). Under constant food conditions the DEB model simplifies to the well known

J. van der Meer et al. / Journal of Sea Research 66 (2011) 434–439

Bertalanffy growth model, but where the ultimate size depends upon the food conditions. All basic DEB model parameters are estimated using independent datasets from the literature, generalizing our results. This only leaves the food condition and the mortality to be estimated from the catch data. Strong evidence exists that the European eel is a panmictic species (Als et al., 2010), and for this reason we have refrained from building a closed population model. Such model would be of little use to describe the local Wadden Sea situation. Hence, recruitment data are model input and no stock-recruitment relationship was involved. 2. Material and methods 2.1. DEB model parameter estimation Using a variety of datasets on various life-history parameters, on growth in the field (Tesch, 2003), growth in the laboratory (Angelidis et al., 2005), and on length–mass relationship (Tesch, 2003), a total of 10 DEB model parameters are estimated according to a procedure described in this volume by Lika et al. (2011). Parameters on natural mortality, net selection, and escapement are taken from the literature, but two parameters, that is the scaled functional response f, describing the food conditions, and the instantaneous fishery mortality rate F are estimated by maximizing the likelihood of these parameters given the observed catch data. The DEB parameter estimation procedure as advocated by Lika et al. (2011) starts with 10 so-called zerovariate data points and 7 pseudo-data points (Table 1). These pseudo-data are in fact standard values for a selection of DEB parameters and these values are taken from Kooijman (2010), Table 8.1. Zero-variate data on age, length and mass at birth and at puberty, on maximum reproductive rate and on maximum length, mass and age are all taken from Tesch (2003), and references therein. Growth and development are probably rather similar in A. anguilla and the related North-American species A. rostrata and Asian species A. japonica. Hence, if no or only few data are available for A. anguilla, data from one of the related species are also presented. Tesch (2003) refers to work of Bezdzenyezhnykh et al. (1983) on the European eel, who obtained 3 mm long larvae 50–60 h after fertilization. Unfortunately no temperature at which the larvae were reared is Table 1 Data used to fit the standard Dynamic Energy Budget model. Data in the upper part of the table are true zero-dimensional data, the middle part gives pseudo-data. These pseudo-data are a-priori estimates of the parameters and are needed to avoid overfitting. The lower part gives true univariate data. The last column gives the model equations that link the data to the model parameters; Liref is a reference length and equals 1 cm, z is the zoom factor, δM the shape coefficient. Equation and section numbers refer to Kooijman (2010). Variable

Explanation

Model

ab ap am Lb Lp Li Wb Wp Wi R˙ i

Age at birth Age at puberty Maximum life span Physical length at birth Physical length at puberty Ultimate physical length Dry mass at birth Dry mass at puberty Ultimate dry mass Maximum reproductive rate Energy conductance Allocation fraction to soma Reproduction efficiency Volume-specific somatic maintenance rate Area-specific somatic maintenance rate Maturity maintenance rate coefficient Growth efficiency Physical length versus age Wet mass versus age Wet mass versus physical length

Eq. 2.39 Eq. 2.53 Sn 6.1.1 Eq. 2.46 Eq. 2.54 Lref i z=δM Sn 1.2.3; Eq. 3.3 Sn 1.2.3; Eq. 3.3 Sn 1.2.3; Eq. 3.3 Eq. 2.58

v˙ κ κR ½ p˙ M  f p˙ T g k˙ J κG L, a Ww, a Ww, L

See text See text Sn 1.2.3; Eq. 3.2

435

given. Pedersen (2004) observed hatching of the European eel 48–52 h after fertilization at 20–21 °C. Larvae were 2.35 mm long. Yamamoto et al. (1974), who worked with A. japonica observed hatching 38–45 h after fertilization at 23 °C at a length of 4.8 mm. American eel larvae hatched 32–43 h after fertilization at 20 °C, and larval size at hatching was 2.7 mm (Oliveira and Hable, 2010). We have chosen an age at birth of 50 h and a length at birth of 2.7 mm. In order to arrive at the dry mass of the newly hatched larva, we use a wet mass–cubic length conversion factor of 0.0014 g/cm 3, which has been reported for glasseel (Tesch, 2003). We have not been able to find a specific conversion factor for the youngest larval stage. We use a single conversion factor dry mass–wet mass of 0.4 (Tesch, 2003), though some evidence exists that this factor changes during lifetime (Boëtius and Boëtius, 1985). Puberty is defined as the stage when fully differentiated immature testes and ovaries appear, which happens at a length of 30 cm (Colombo and Grandi, 1996). In northern European waters, this length is reached around 5 years after immigration. The animals weigh then around 44 g (Tesch, 2003). Maximum length and age are taken from Tesch (2003) and Dekker (2004b). Tesch (2003) reports that eels as long as 130 cm were caught in a Swiss lake and Dekker (2004b) mentions a maximum length of 133 cm for the IJsselmeer. Tesch (2003) further reports an average wet mass of 1622 g (range from 1360 to 1884 g) for eel within the size class 98–103 cm. This points to a wet mass-cubic length conversion factor of 0.0016 g/cm3. Using the same factor implies a maximum wet mass of around 3500 g for the Swiss eels. Tesch (2003) refers to the work of Edel (1975) who counted 1.3–1.5 million eggs in mature A. rostrata weighing 560 g. However, Barbin and McCleave (1997) report a much higher fecundity for American eel of that mass, namely between 4 and 5 million eggs. Boëtius and Boëtius (1980) report that European eels of 560 g produce slightly less than one million eggs. We have chosen a maximum of 2 million eggs for an eel of 560 g, resulting in a maximum reproductive rate of slightly more than 900 eggs/day. This mass corresponds to a length of 66 cm, which in northern European waters is reached about 11 years after immigration, and thus 6 years after reaching puberty. Eel can become very old, and in Denmark a specimen was held for 55 years (Walter, 1910), which we took as the maximum lifespan. The standard DEB model provides predictions for all zero-variate data. We do not repeat the relevant equations, but refer the reader to Kooijman (2010). Detailed references to the equations are given in Table 1. For completeness, several conversion coefficients should be mentioned here: volume-wet mass 1 g/cm 3, volume-dry mass 0.4 g/cm 3 (Tesch, 2003), energy per C-mol in the reserves 550 kJ/C-mol (Kooijman, 2010), and dry mass 23.9 g/C-mol (Kooijman, 2010). Combined these coefficients reveal an energy content within the reserves of 9.2 kJ/g wet mass. The procedure of Lika et al. (2011) further allows for the use of any kind of so-called univariate data, such as length data obtained at a various ages. Age-length data from the field are taken from Tesch (2003), who reports in his Table 3.10 last row, average Northern European data. We set the average temperature for these waters at 10 °C. Data of glasseel growth in the laboratory in the form of agewet mass data, which were obtained at ad-libitum food conditions and at a much higher temperature of 21.5 °C, were taken from Angelidis et al. (2005). These growth data are only concerned with animals that have already entered the glasseel stage. Food conditions and temperature may be very different for the earlier oceanic life stage, but because duration of the oceanic stage is still unknown (Tesch, 1998) and relevant information on the environmental conditions that the leptocephali larvae experience is lacking, growth during this stage was not explicitly modeled. Instead an additional time-lag parameter τ was used, resulting in the growth equation L ¼ fLi −ð fLi −Lb Þe

r˙ B ðtþτÞ

ð1Þ

J. van der Meer et al. / Journal of Sea Research 66 (2011) 434–439

t is time since reaching the glasseel stage, and T is ambient temperature in K. All other parameters are explained in Tables 1 and 3. Finally, length-mass data from the field were taken from Tesch (2003). The goodness-of-fit function that is minimized in the procedure by Lika et al. (2011) is a weighted sum of squares of the form rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi   ˆ 2 ∑wi Yi −Yi Y i , where wi is the weight coefficient for observation i, and Yi and Yˆ i are the observed and predicted value, respectively. The choice of the weights wi is partly arbitrary, and we have chosen to put more emphasis on true data than on pseudo-data, and more on mass data than on length data. Weights for the true zero-variate data are wi = 100 or 1000, for the pseudo-data they are 1 or 100, and for the univariate data they are 100 or 1000. As the error in the univariate data is probably not entirely proportional to the observed value, we used higher weights for the larger animals, i.e. wi = 10, 000 for animals longer than 65 cm or heavier than 30 g. The DEB model routines and the parameter estimation routine are part of the software packages DEB tool and Add-my-pet and can be downloaded from the Theoretical Biology website of the VU University Amsterdam http://www.bio.vu.nl/thb/deb/deblab/debtool. 2.2. Population model The population model describes the development of each annual cohort of eel, in terms of their age, size and total number. All animals follow the same growth trajectory, and all cohorts follow the same survival curve. Average temperature is assumed constant and is set at 10 °C, the same value as we used for the Northern European inland dataset provided by Tesch (2003). Cohorts only differ in the time of arrival and the initial number arriving. The instantaneous natural mortality rate is set to 0.138/year, according to Dekker (2004a). Instantaneous fisheries mortality rate and silvering or escapementrate are related to length by logistic equations of the form

μ ¼ μm

ex 1 þ ex

ð3Þ

μF ð1−Si Þ: μM þ μF þ μE

ð5Þ

Ci . It is assumed that half of The catch rate per period is defined as Δt i the arriving glasseels will become female and the other half male. Since survival fractions within a period may differ between males and females as a result of different escapement, the decrease in numbers should be calculated separately for the two sexes. For each sex the total number of animals in each cohort after n periods, that is at time ∑ Δti, is given by half the initial abundance times the cumulative survival product

Nn;sex ¼

N0 ∏Si;sex : 2

ð6Þ

The initial abundance of each cohort is only known in relative terms, that is in a glasseel immigration index (Fig. 1). We further do not know total catches, but only catches of our constant effort research program. Observed catches will only reflect total catches under the assumption of a constant relative effort. Hence, both predictions and observations are only relative figures, with unknown conversion coefficients. In order to link the predicted to the observed catches, predictions are therefore constrained such that the sum of all predicted catches equals the sum of all observed catches. As said earlier, maximum instantaneous fisheries mortality rate and the scaled functional response f were estimated from the data. Parameter values were chosen that maximized the Poisson likelihood Σyi log yˆ i − yˆ i , where yi and yˆ i are the observed and the predicted counts, respectively. The Poisson likelihood can be approximated by means of a cubic root transformation of the counts (McCullagh and Nelder, 1983), and for illustrative purposes the correlation between the cubic root of the observed and predicted counts was used as a simple and reliable goodness-of-fit measure. Observed counts are daily catches from a fyke that has been used for scientific research since 1960 (Van der Meer et al., 1995). Sampling is mainly carried out during spring (March–June) and autumn (August–October). The fyke is located at the southern tip of the Frisian island Texel, close to the Royal Netherlands Institute for Sea Research. Since 1972 length measurements of each fish have been taken, though in the period 1972–1979 fish were only classified in three length classes (smaller than 20 cm, between 20 and 40 cm, and larger than 40 cm). After 1979 fish were recorded in cm classes. As catches are rather low in the last period, this detailed information was not used, and fishes were only categorized in the three crude

0

where the linear predictor x is given by (L−L50)/b, L is the length, L50 the length at which the rate μ is 50% of the maximum rate μm, and b indicates how quickly the rate increases with length. Based upon sizefrequency data of the catches, it appeared that L50 = 31 cm and b = 50 cm are reliable estimates for the fisheries mortality. L50 is thus slightly higher than the minimum allowed landing size of 28 cm. For escapement the same value for b is used, and values for L50 are set to 45 cm for males and 75 cm for females. Maximum instantaneous escapement rate is set equal to 0.6/year. The lifespan of each cohort is split up in periods in which an individual passes through a one centimeter size class. If such period should fall in two separate calendar years, then it is split up in two periods. This approach simplifies a comparison of model predictions and available data in the form of annual catches per cm class. For each period i the surviving fraction, or actually the fraction remaining in the area, is given by

Ci ¼

120

ð2Þ

100

T

− TA

80

TA

eTref

60

!−1

40

ref 1 fzLi ½E  þ G r˙B ¼ 3 v˙ ½ p˙ M 

Δti stands for the duration of the period. The fraction caught in each period equals

20

where L is physical length, r˙ B is the Bertalanffy growth coefficient, given by

Index

436

−ðμM þμF þμE ÞΔti

Si ¼ e

ð4Þ

where μM, μ F, and μE stand for the instantaneous natural mortality rate, fisheries mortality rate, and escapement rate, respectively. These rates are, as explained above, functions of the size of the eel.

1940

1960

1980

2000

Year Fig. 1. Long-term trend in the glasseel index of the Den Oever sluices. The index is based on nightly catches during spring with a shore-operated dipnet (Dekker, 1998).

0

1000

Mass g

40

Length cm

0 0

2000

4000

20

Age days

40

60

80

Length cm

The goodness of fit at the estimated parameter values was remarkably good for all zero- and univariate data (Table 2, Fig. 2). The DEB model is able to describe the very different growth rates in Northern European freshwaters and in the laboratory, and points to temperature as the main reason for this difference. The estimated time lag parameter τ equaled 22 days, which is, however, much shorter than the oceanic period. Various other parameter estimates showed rather large standard errors (Table 3), which points to overfitting and suggests that more and different types of data are required to obtain reliable estimates. In the present context, however, the growth-related parameters are most relevant. In fact, only the compound parameters ultimate length and the Bertalanffy growth coefficient were used in the population model. These compound parameters were much more reliably estimated than the underlying basic parameters, which appeared to be strongly correlated. The ultimate physical length estimate Li equaled 130.5 cm and the Bertalanffy coefficient r˙ B equaled 0.0558/year at 10 °C. These values are within the range reported earlier by Tesch (2003), where the ultimate length for normal sized eel varied between 45 and 143 cm and the Bertalanffy coefficient between 0.013 and 0.29/year. Initially, the population model contained two estimable parameters, namely the scaled functional response parameter f and the maximum instantaneous fisheries mortality μFm. The first parameter was estimated at f = 1.12, which falls outside the feasible range 0 b f b 1. The estimation procedure was therefore repeated with f set equal to one, resulting in an estimate of the only remaining estimable parameter of μFm = 0.225/year. The survival rate, or better the ‘remaining’ rate, is initially high, but starts to decrease when the fish approach the minimum allowable size of 28 cm. It further drops down around Table 2 Observed versus predicted zero-dimensional data. Data in the upper part of the table are true data, the lower part gives pseudo-data. Weight refers to the weighing of the squared differences in the optimization procedure. Variable Observed

Predicted Unit

ab ap

2.083 1825

1.399 1679

d d

100 100

am Lb Lp

20,080 0.27 30

21,340 0.2361 29.65

d cm cm

100 100 100

Li

133

130.5

cm

100

Wb

1.102e−5 8.917e−6 g

Wp Wi R˙ i

17.6 1400 913.2 0.02 0.8 0.95 18 0 0.002 0.8

v˙ κ κR ½ p˙ M  fp˙ T g k˙ J κG

17.66 1506 862.2 0.04947 0.9882 0.95 45.95 0 0.002 0.9271

Weight Source

1000

g 1000 g 1000 d− 1 100 cm d− 1 1 – 100 – 1 J d− 1 cm− 3 1 −1 −2 Jd cm 1 d− 1 1 – 100

Yamamoto et al. (1974) Colombo and Grandi (1996), Tesch (2003) Walter (1910) Yamamoto et al. (1974) Colombo and Grandi (1996), Tesch (2003) Dekker (2004b), Tesch (2003) Tesch (2003), Yamamoto et al. (1974) Tesch (2003) Tesch (2003) Edel, (1975), Tesch (2003)

40 0

3. Results

Mass g

80

length classes. Hence, goodness-of-fit was based on a comparison of observed and predicted total catches in the period before 1972, and on comparing observed and predicted catches per crude length class in the period after 1972. More details on the fyke can be found in Van der Meer et al. (1995). Glasseel data come from a scientific sampling program at the Den Oever sluices, which started in 1938. Sampling is carried out at night at two-hourly intervals during spring using a 1 m 2 dipnet with a mesh size of 1 mm 2. For details we refer to Dekker (1998, 2002, 2010).

437

80

J. van der Meer et al. / Journal of Sea Research 66 (2011) 434–439

0

100

200

300

Age days Fig. 2. Observations and predictions of the Dynamic Energy Budget (DEB) model for length versus age, field data from Northern European inland waters (Tesch, 2003), wet mass versus length, field data (Tesch, 2003), and wet mass versus age, laboratory data on glasseel growth (Angelidis et al., 2005). See text for further details.

40 cm, when the males start to escape from the area, and rises again when all males have disappeared. At a length larger than about 70 cm, that is when the females start to silver, survival rate decreases again (Fig. 3). The correlation between the cubic root of the observed and predicted counts equaled r = 0.93, pointing to a very reasonable fit (Fig. 4). However, a comparison of the time course of the predictions and observations shows that the decline in the observations precedes the decrease in the predictions (Fig. 5). The lag is about two to five years. The predicted small recovery in the catches around 2002, following the minor recruitment peak in 1997, did not occur. Note that the two most striking maxima in the predicted catch, which occurred in 1967 and 1983, also followed five years after the peaks in the recruitment of 1962 and 1978.

4. Discussion Estimates for several primary DEB parameters, such as the Arrhenius temperature, the zoom factor, the shape parameter, and the allocation

Table 3 Parameter estimates of the standard Dynamic Energy Budget model. A standard error SE equal to zero means that the parameter is not estimated, but fixed a-priori. All rates are given for a reference temperature Tref of 293 K. Parameter

Estimate

TA f z δM v˙ κ κR ½ p˙ M 

13,310 1 11.93 0.0914 0.0495 0.988 0.95 45.95

fp˙ T g

0

0

J d− 1 cm− 2

k˙ J

0.002

0

d− 1

[EG] EHb EHp h˙ a

9026 0.00108 2669 1.306e−8 0 22.25

sG τ

SE 3744 0 7.121 0.0584 0.103 0.157 0 84.15

10,390 0.0121 27,590 0.009179 0 12.36

Unit

Explanation

K – – – cm d− 1 – – J d− 1 cm− 3

Arrhenius temperature Scaled functional response Zoom factor Shape parameter Energy conductance Allocation fraction to soma Reproduction efficiency Volume-specific somatic maintenance rate Area-specific somatic maintenance rate Maturity maintenance rate coefficient Specific costs for structure Maturity at birth Maturity at puberty Weibull aging acceleration Gompertz stress coefficient Time lag of the growth curve

J cm− 3 J J d− 1 – d

20

−0.2 20

60

100

100

5 0

0.8 0.4

1960 0.0

Relative abundance

0.6 0.3

Escapement rate

0.0 20

60

Length cm

60

100

Length cm

Fig. 3. Model predictions of the fate of a single cohort of eel in terms of catch rate, silvering or escapement rate, log survival rate and relative abundance versus length. All rates are expressed per year.

15 10 5 0

Cubic root of observed annual catch

fraction to soma, were reasonably accurate. Other parameter estimates showed, however, large standard errors. The population model required only the compound parameters maximum physical length and Bertalanffy growth rate, which could reasonably well be estimated. Hence, the present application did not suffer from the fact that such did not hold for all primary DEB parameters. Yet, if reliable estimates are required for all primary DEB parameters more information on, for example, growth rate and reproduction rate in response to various food levels should become available. Perhaps, unpublished data from eel hatcheries can be used. The oceanic part of the eel life cycle, i.e. the first and last stages, has so far remained mainly outside the reach of eel biologists. They predominantly work in freshwater habitats, or when working in the marine environment study the upstream movement of eels migrating through estuaries into freshwaters (Creutzberg, 1958; Creutzberg, 1961). For example, though growth rates of leptocephali larvae have been estimated on the basis of otolith microstructure (Castonguay, 1987), there is still much debate about the underlying assumption that the observed growth increments are daily rings (Tesch, 1998). The reported growth rate of 0.38 mm/day by Castonguay (1987) implies that the metamorphosis length of 60 mm is reached in about 5 months. Indirect estimates of the oceanic stage

5

1970

1980

1990

2000

2010

Year 20

Length cm

0

15

−0.6

100

10

60

Length cm

Catch per day

20

−1.0

0.15

log(Survival rate)

J. van der Meer et al. / Journal of Sea Research 66 (2011) 434–439

0.00

Catch rate

438

10

15

Cubic root of expected annual catch Fig. 4. Predicted versus observed catch of eel in the Wadden Sea research fyke, expressed as the cubic root of the average number of eels caught per day. Each datapoint refers either to the total annual catch (before 1972) or to the annual catch of large or intermediate-sized eels (after 1972). Large eels are larger than 40 cm, intermediatesized eels have a length between 20 and 40 cm.

Fig. 5. Model predictions and observed catches of two length classes of eel in the period 1960–2009, Wadden Sea research fyke. The solid line and the circles refer to the overall catch and the dotted line and triangles to fish smaller than 40 cm. Population model is explained in detail in the text.

duration by size-frequency analysis of leptocephali larvae point to an oceanic period between 3/4 to more than 3 years (Boëtius and Harding, 1985; Tesch, 1998). As the logistic problems of oceanic eel research are huge, it seems that only if breeding eel in captivity becomes successful, and some promising results have been obtained recently, further information on these elusive stages will become available. Our estimated time lag parameter τ equaled 20 days, which is much shorter than the actual length of the oceanic period, even if that is only as short as 5 months. Apparently, growth rate during this period is much lower than would be expected on the basis of later growth. The population modeling exercise points to a discrepancy between the predicted and observed timing of the decline. A first reason for such discrepancy might be that the underlying assumption of constant catchability is invalid. Relative catchability may be lower when stocks are low, for example because the area around the single fyke may be earlier deserted than other parts of the local area. Although we cannot rule out such explanation, the fact that other fykes in the area in the period 1966–1973 showed more or less parallel trends for the eel (Van der Meer et al., 1995), suggests that such local variability is not very likely. One might further argue that the earlier observed decline could be due to a shorter time lag between recruitment and catch as a result of an increase in the growth rate, which by itself may be due to better food conditions. A model that assumes constant growth conditions is by definition not able to capture such changing conditions. However, the estimated scaled functional response parameter f was estimated at a value even slightly higher than one, i.e. pointing to excellent food conditions for the entire period. Such good food conditions in marine environments are in accordance with recent observations that eel growth is higher in saline than in fresh water environments (Cairns et al., 2009). An improvement in food conditions is thus rather unlikely. Furthermore, improved food conditions cannot explain the not coming off of a small peak in abundance in the early 2000s. A good fit between model predictions and catches would not have been expected when migrating silver eels leaving the large inland waters would have contributed much to the catches. Size-frequency distributions separated for spring and autumn, however, suggest that the fyke only catches local animals and migrating silver eels are believed to pass much further off-shore through the deeper parts of the Marsdiep tidal inlet. Summing up, rather similar to the situation for the IJsselmeer, we are confronted with a so far unexplained impoverishment of the Wadden Sea as a habitat for the eel since the 1980s. The disappearance of the eel from the Wadden Sea has mainly, but not only occurred through decreasing numbers of glasseel arriving.

J. van der Meer et al. / Journal of Sea Research 66 (2011) 434–439

Acknowledgments We would like to thank two anonymous reviewers and Leon Heinsbroek for their helpful comments that greatly improved the manuscript. Thanks are also due to Siemen Gieles for his strenuous fishing efforts. References Als, T.D., Hansen, M.M., Maes, G.E., Castonguay, M., Riemann, L., Aarestrup, K., Munk, P., Sparholt, H., Hanel, R., Bernatchez, L., 2010. All roads lead to home: panmixia of European eel in the Sargasso Sea. Molecular Ecology 20, 1333–1346. Angelidis, P., Pournara, I., Photis, G., 2005. Glass eels (Anguilla anguilla) growth in a recirculating system. Mediterranean Marine Science 6, 99–106. Barbin, G.P., McCleave, J.D., 1997. Fecundity of the American eel Anguilla rostrata at 45° N in Maine, U.S.A. Journal of Fish Biology 51, 840–847. Bezdzenyezhnykh, V.A., Prochorchik, A.M., Petrikov, A.M., Petukhov, V.B., Plyuta, M.V., 1983. Obtaining larvae of Anguilla anguilla L. (Pisces, Anguillidae) under experimental conditions. Dokl. Akad. Nauk 268, 1264–1266. Boëtius, I., Boëtius, J., 1980. Experimental maturation of female silver eels, Anguilla anguilla. Estimates of fecundity and energy reserves for migration and spawning. Dana 1, 1–28. Boëtius, I., Boëtius, J., 1985. Lipid and protein content in Anguilla anguilla during growth and starvation. Dana 4, 1–17. Boëtius, J., Harding, E.F., 1985. A reexamination of Johannes Schmidt's eel examinations. Dana 4, 129–162. Cairns, D.K., Secor, D.A., Morrison, W.E., Hallett, J.A., 2009. Salinity-linked growth in anguillid eels and the paradox of temperate-zone catadromy. Journal of Fish Biology 74, 2094–2114. Castonguay, M., 1987. Growth of American and European eel leptocephali as revealed by otolith microstructures. Canadian Journal of Zoology 65, 875–878. Colombo, G., Grandi, G., 1996. Histological study of the development and sex differentiation of the gonad in the European eel. Journal of Fish Biology 48, 493–512. Creutzberg, F., 1958. Use of tidal streams by migrating elvers (Anguilla vulgaris Turt.). Nature 181, 857–858. Creutzberg, F., 1961. On the orientation of migrating elvers (Anguilla vulgaris Turt.) in a tidal area. Netherlands Journal of Sea Research 1, 257–338. Daverat, F., Limburg, K.E., Thibault, I., Shiao, J.C., Dodson, J.J., Caron, F.O., Tzeng, W.N., Iizuka, Y., Wickström, H., 2006. Phenotypic plasticity of habitat use by three temperate eel species, Anguilla anguilla, A. japonica and A. rostrata. Marine Ecology Progress Series 308, 231–241. Dekker W., 2002. Monitoring of glass eel recruitment. Report C007/02-WD, Netherlands Institute of Fisheries Research, IJmuiden.

439

Dekker, W., 2004a. What caused the decline of the Lake IJsselmeer eel stock after 1960? ICES Journal Marine Science 61, 394–404. Dekker, W., 2004b. Slipping through our hands. Population dynamics of the European eel. Thesis University of Amsterdam, Amsterdam, ISBN 90-74549-10-1. Dekker, W., 2010. Report on the eel stock and fisheries in the Netherlands 2008. Report of the 2009 Session of the Joint EIFAC/ICES Working Group on Eels. ICES CM 2009/ACOM:15. International Council for the Exploration of the Sea, Copenhagen, pp. 140–181. Dekker, W., 1998. Long-term trends in the glasseels immigrating at Den Oever, the Netherlands. Bulletin Francais de la Peche et de la Pisciculture 349, 199–214. Dekker, W., 2003. Did lack of spawners cause the collapse of the European eel Anguilla anguilla? Fisheries Management & Ecology 10, 365–376. Edel, R.K., 1975. The induction of maturation of female American eels through hormone injections. Helgoländer Wissenschaftliche Meeresuntersuchungen 27, 131–138. Henderson, P.A., Plenty, S.J. Newton, L.C., Bird, D.J., 2011. Evidence for a population collapse of European eel (Anguilla anguilla) in the Bristol Channel. Journal of the Marine Biological Association (United Kingdom). Published online doi:10.1017/ S002531541100124X. Kooijman, S.A.L.M., 2010. Dynamic Energy Budget Theory for Metabolic Organisation, Third ed. Cambridge University Press, Cambridge. Lika, K., Kearney, M.R., Freitas, V., van der Veer, H.W., van der Meer, J., Wijsman, J.W.M., Pecquerie, L., Kooijman, S.A.L.M., 2011. The “covariation method” for estimating the parameters of the standard Dynamic Energy Budget model I: Philosophy and approach. Journal of Sea Research 66, 270–277 (this volume). McCullagh, P., Nelder, J.A., 1983. Generalized Linear Models. Chapman and Hall, London. Van der Meer, J., Witte, J.IJ., van der Veer, H.W., 1995. The suitability of a single intertidal fish trap for the assessment of long-term trends in fish and epibenthic invertebrate populations. Environmental Monitoring and Assessment 36, 139–148. Van der Meer, J., 2006. An introduction to Dynamic Energy Budget (DEB) models with special emphasis on parameter estimation. Journal of Sea Research 56, 85–102. Oliveira, K., Hable, W.E., 2010. Artificial maturation, fertilization, and early development of the American eel (Anguilla rostrata). Canadian Journal of Zoology 88, 1121–1128. Pedersen, B.H., 2004. Fertilisation of eggs, rate of embryonic development and hatching following induced maturation of the European eel Anguilla anguilla. Aquaculture 237, 461–473. Tesch, F.W., 1998. Age and growth rates of North Atlantic eel larvae (Anguilla spp.), based on published length data. Helgoländer Wissenschaftliche Meeresuntersuchungen 52, 75–83. Tesch, F.W., 2003. The Eel, Fifth ed. Blackwell, Oxford. Walter, E., 1910. Der Flussaal, eine biologische und fishwirtschaftliche monographie. Verlag J. Neumann, Neudamm. Yamamoto, K., Morioka, T., Hiroi, O., Omori, M., 1974. Artificial maturation of Japanese eels by the injection of salmonid pituitary. Bulletin of the Japanese Society of Scientific Fisheries 40, 1–7.