Relationship between phytoplankton production and cod production on the Faroe Shelf

Abstract

Phytoplankton production on the Faroe Shelf varied dramatically from 1990 to 2003 and a close empirical relationship between phytoplankton production and cod production is demonstrated in this study. The large fluctuations of the biomass of the Faroe Plateau cod stock in the 1990s seem to be explained by variations in primary production, i.e. by a fluctuating carrying capacity of the Faroe Shelf ecosystem. The causal mechanism seems to be a positive relationship between phytoplankton production, zooplankton production and production of food organisms for cod (e.g. benthic crustaceans and especially sandeels). The year-class strength of Faroe Plateau cod seems to be determined during their second winter, corresponding to the time when they leave the nearshore nursery areas and enter the feeding areas of adult cod. The regulating factor seems to be abundance of suitable food organisms for small cod, which seems to be a result of total food availability (determined by the phytoplankton production) and competition from older cod, i.e. density-dependent recruitment. The individual growth of cod older than 2 years seems to be a result of the total food availability and the stock size of cod, i.e. density-dependent growth. Implications for stock assessment and management are discussed.

Introduction

The Faroe Plateau is located approximately midway between Iceland and Scotland ( Figure 1 ). The size of the shelf (<120-m bottom depth) is about 60 × 80 nautical miles, covering an area of about 10 000 km ² . It is separated from other shallow areas by depths exceeding 400 m in all directions. Below about 500 m, cold deep Norwegian Sea water dominates. The upper layers (above ∼500-m depth) are dominated by Atlantic water from the North Atlantic Current, which flows in a northeasterly direction ( Hansen, 1985 ; Hansen and Østerhus, 2000 ).

Figure 1

The location of the Faroe Islands, bottom topography, and typical flow fields. The letters refer to stations mentioned in the text.

On the Faroe Shelf, extremely strong tidal currents lead to intense mixing of the water column, resulting in homogeneous water masses in the shallow shelf areas ( Hansen, 1992 ; Gaard et al. , 1998 ; Hansen et al. , 1998 ). The well-mixed shelf water is separated relatively well from the offshore water by a persistent tidal front, which surrounds the shelf at about the 100–130-m bottom depth contour ( Gaard et al. , 1998 ; Larsen et al. , 2002 ). In addition, residual currents have a persistent clockwise circulation around the islands ( Hansen, 1992 ; Simonsen, 1999 ; Gaard and Hansen, 2000 ). These circulation patterns lead to a relatively uniform shelf ecosystem on the Faroe Shelf that is different from the offshore environment. The ecosystem has its own planktonic communities which, in terms of species composition and production, are distinct from those of the surrounding oceanic environment ( Gaard, 1996 , 1999 , 2000 ). The shelf is also a habitat for a variety of benthic fauna and several fish stocks (e.g. cod, Gadus morhua ;e.g. Joensen and Tåning, 1970 ; Jákupsstovu and Reinert, 1994 ; Gaard and Steingrund, 2001 ; Gaard and Reinert, 2002 ). Furthermore, a large number of seabirds breed on the Faroe Islands ( Bloch and Sørensen, 1984 ) and take most of their food from the shelf waters.

The catch of cod has fluctuated between 20 000 and 40 000 t per year, except during the Second World War and a period from 1990 to 1994 ( Figure 2 ). During this time, the catch of other commercial fish species (haddock, Melanogrammus aeglifinus ; plaice, Pleuronectes platessa ; lemon sole, Microstomus kitt ) also declined to unusually low levels, causing severe economic problems for the fishing industry and national economy on the Faroe Islands. Analytical assessments of cod and haddock have shown that the stock sizes were the lowest on record for the period covered by the assessment (1961–2003; ICES, 2004 ) and most likely the whole period covered by catch statistics (1903–2003). Prior to the collapse of the cod fishery, growth and recruitment had been poor for several years and fishing mortality was high ( ICES, 2004 ).

Figure 2

Faroe Plateau cod annual landings, biomass of ages 2+, and spawning-stock biomass. Prior to 1960 the landings from Faroe Bank (Mean: ∼8% of total) are included in the figures.

The cod stock rapidly recovered, however, during the period 1993–1995 and catches increased from 6000 t in 1993 to 40 000 t in 1996 ( Figure 2 ). The main reasons for the increased catches were strong recruitment, increased individual growth, and increased fishing effort ( Steingrund et al. , 2003 ). This “comeback” of the cod came as a big surprise to both fisheries scientists and local people. Local people explained the variations in the catches by large scale migrations, but tagging experiments do not support this hypothesis ( Strubberg, 1916 , 1933 ; Tåning, 1940 ; unpublished data). Fisheries scientists, on the other hand, had to revise the spawning stock-recruitment relationship. Prior to 1992 a small spawning-stock biomass was never observed to give strong recruitment, but this changed when the strong 1992 and 1993 year classes were observed to come from extremely small spawning-stock biomasses.

The revision of the stock-recruitment relationship suggested that the role of environmental factors needed to be examined. Water temperature is known to affect the metabolic rates of poikilotherms, and Brander (1995) has found quite a good correlation between long-term variability in Faroe Plateau cod growth rates and water temperature on the Faroe Shelf and the weather station M in the Norwegian Sea. However, during the 1990s, the short-term variability in growth rates did not correlate with temperature variability ( Gaard et al. , 2002 ). Nor was the cod recruitment on the Faroe Shelf influenced by variability in temperature. The temperature, which normally ranges between 6 and 10°C ( Smed, 1952 ; Hansen, 2000 ; Heath et al. , 2000 ) is close neither to the upper nor to the lower tolerance limits for cod, and the relatively small, long-term interannual temperature variability observed on the shelf does not correlate with the variability in cod recruitment ( Planque and Frédau, 1999 ; Gaard et al. , 2002 ). Temperature fluctuations are also not supposed to have influenced the distribution of cod on the Faroe Shelf, as is the case in the Barents Sea ( Nakken and Raknes, 1987 ) and northern Gulf of St. Lawrence ( Dutil et al. , 1999 ). In those regions, sea temperatures are much closer to the lower tolerance limit of cod.

Recent studies have shown that there is a good interannual correlation between plankton production, fish reproduction and growth rates, and seabird reproduction on the Faroe Shelf. Variability in weight-at-age and total production of cod and haddock and seabird recruitment co-fluctuated, and correlate very well with interannual variability in primary production ( Gaard et al. , 2002 ; Steingrund et al. , 2003 ). In this paper we explore in detail the annual variability of cod production, its relationship with primary production in the ecosystem, and possible reasons for the relationship. Factors causing the annual variations in primary production are only briefly dealt with in this study.

Material and methods

Hydrography

Salinity was measured between 22 and 30 June 2000 on 48 stations, which were evenly distributed on the shelf and slope, using a Seabird Electronics SBE 911 plus CTD. The salinity was calibrated against water bottles analysed on a salinometer.

Nitrate

Nitrate was measured on 22–48 stations, which were evenly distributed on the shelf and slope in late June 1990–2003.

Frequent measurements of nitrate are also carried out on water from a coastal station (Station C on Figure 1 ), where water is pumped from 18-m depth in a well-mixed water column. The samples were collected approximately twice a week after 1994.

The nitrate samples that were taken in 1990 were stored in a refrigerator and analysed a few days after sampling. In 1991–1994, they were frozen immediately after sampling and analysed ashore. Since 1995 the samples that were collected using a research vessel were analysed onboard, and the samples that were taken on the coastal station (S) were preserved with 12 drops of chloroform per 100 ml of sample. Nitrate and nitrite was measured with an autoanalyser according to the method of Grasshoff et al. (1983) .

An approximation of relative nitrate assimilation (potential new primary production) in the shelf water ecosystem during the high productive period was calculated based on reduction of nitrate concentrations from winter levels to a fixed date (26 June each year) on stations S1 and S2 plus net influx of nutrients from the surrounding offshore water during the investigated time:  

The net inflow of nitrate was calculated as  

where [NO ₃⁻ ] _offshore and [NO ₃⁻ ] _shelf are the nitrate concentrations in the offshore and the shelf water, respectively. The average renewal rate is estimated to about 1/75 dg ⁻¹ ( Gaard, 2003 ).

Cod production

Cod production is defined as weight increase of the cod population ( Pitcher and Hart, 1982 ). To estimate the production it was necessary to know the abundance and individual growth rate of the cod stock. Both of these were available in the stock assessment of Faroe Plateau cod presented by ICES (2004) . The assessment (Extended Survivor Analysis, XSA) covered the period 1961–2003 and the input was based on catch-at-age data (commercial catch split into age groups) and weight-at-age data (corresponding weights of the age groups). The XSA was tuned by two groundfish surveys: a spring survey, where 100 fixed stations of 1-h duration were trawled in March 1994–2004, and a summer survey, where 200 fixed stations of 1-h duration were trawled in August–September 1996–2003. The assessment gave an estimate of cod abundance (number of fish of ages 2–9 years) at the beginning of each calendar year.

The primary production on the shelf is mainly restricted to the period May–July ( Gaard et al. , 2002 ). The resulting production of food organisms for cod occurs some time after the spring bloom. Therefore, cod production was calculated from midyear to midyear. For example, the phytoplankton production in 1996 was related to the production of cod in the period from midyear 1996 to mid year 1997.

According to Pitcher and Hart (1982) , the production of a population during a period may be calculated as the average number of the population multiplied by the individual weight increase, leading to:  

where W _t+1 − W _t denotes the individual weight increase from summer in year t to summer the next year. N _t+1 (stock population numbers on 1 January) was used as a proxy for the average number of cod in the time period. For example, the production of the 1993 year class resulting from the 1996 plankton production was calculated as the number of cod at the beginning of 1997 (4 year old) multiplied by the weight increase from summer 1996 to summer 1997 (weight of 4-year-old cod in 1997 minus 3-year-old cod in 1996). The oldest age (age 9) was not included in the analysis because sample sizes were small and therefore inaccurate. Furthermore, they represented a very small part of the total cod production.

Since small fish are numerous in a population compared with large fish, there was a need to include as many young ages as possible in the calculations. The youngest age included in the stock assessment was age 2, i.e. the number of 2 year olds in the beginning of the year was known as well as the individual weight of 2.5-year-old cod. In order to calculate production of cod of ages 1.5–2.5 years, it was necessary to estimate the weight of 1.5-year-old cod. No adequate data were available from the catch records, so they were obtained from the summer groundfish survey (c. 10 August to c. 5 September, i.e. 1.7 years old). Length frequencies were analysed in order to estimate average length, which was then converted to weight using year specific condition factors [Fulton condition factor = 100 × (weight in g) × (length in cm) ⁻³ ]. Prior to 1996, the weight was set at a constant value, using an average length and an average condition factor for the period 1996–2003.

An attempt was also made to calculate the production of cod of ages 0.5–1.5 years. Data from the summer groundfish surveys suggested a mean length of 0.7-year-old cod of 10 cm, corresponding to 10.2 g. The numbers of 1-year-old cod at the beginning of the year were calculated from the number of 2-year-old cod the year after, having a natural mortality of 1: N _{1 year} = N _{2 years} × e ¹ . The estimate of natural mortality is rather imprecise and is obtained from stomach data in the period 2000–2003 (unpublished results).

Thus, the production of cod, haddock, and saithe was obtained for following periods: 0.5–1.5, 1.5–2.5, 2.5–3.5, …, 7.5–8.5 years. These contributions were summed in different ways. When all periods were included, it was denoted “ages 0.5+” or “0.5 years and older”. The youngest ages were sometimes excluded, and were denoted “ages 1.5+” and “ages 2.5+”, respectively. On one occasion, the production of cod was split into (i) recruiting cod (1.5–2.5 years), (ii) medium-sized cod (2.5–3.5, 3.5–4.5, 4.5–5.5 years), and (iii) large cod (5.5–6.5, 6.5–7.5, 7.5–8.5 years).

The individual growth was based on the weights in the commercial catches. Since these may be biased as a result of factors such as changing fleet composition, individual weights were also calculated based on the summer groundfish surveys alone. The age–length keys were calculated using Gaussian smoothing. The overall length distributions of cod (scaled to total cod catch at each station) were pooled by stratum ( Kristiansen, 1988 ), and divided by the number of stations in each stratum. Having the mean length distribution by stratum, an average was taken of all 15 strata weighted by stratum area. Using this length distribution and the age–length key, the age–length distribution was calculated (for more details see ICES, 2002a ). From the relationship: individual weight = K(length) ³ , the total weight of cod by length and age was calculated using specific values for K for each year and each 10-cm-length group of cod. By summing the weights for each age and dividing by the corresponding number of fish, the mean weight-at-age was obtained.

Cod recruitment

When relating the primary production to recruitment of cod, estimates from the XSA were used for all relevant year classes, i.e. the 1989–2001 year classes. The estimates of the 2000 and 2001 year classes may be unprecise, because the XSA is not converged for the last 2 years.

Relationship between the Faroe Shelf area and deeper areas

As data on phytoplankton production are available for the Faroe Shelf area only (within the 130-m isobath), it was necessary to calculate the abundance of cod in the Faroe Shelf area compared with the rest of the Faroe Plateau. This analysis was based on the summer groundfish survey. An unsmoothed age–length key was calculated by pooling all stations and having length groups of 5 cm. The length distributions (5-cm-length groups) per station (scaled to total catch of cod), were pooled for Faroe Shelf stations (<130 m) and the deeper stations separately, and corresponding age–length distributions were calculated. Because of small sample sizes, the age–length key was usually inadequate for the youngest ages (ages 0 and 1), and assumptions about the age were made from the length distribution alone. The results were presented as the proportion of cod caught in the Faroe Shelf area to the total catch in each age group.

In order to compare variability in the productivity of the Faroe Shelf area and the deeper areas, the Fulton condition factor was calculated for fish species that were abundant in both areas. These included cod, haddock, whiting ( Merlangus merlangus ), plaice, lemon sole, and Norway haddock ( Sebastes viviparus ). The Fulton condition factor was obtained for all individuals and the mean value calculated, thus giving one data point a year for each species. Since individual weights were scarce for some species in 1996, only seven data points could be calculated. For plaice the data were scarce (between 7 and 191 fish in each area by year), but for the other species the number of fish was usually more than 100 and frequently over 1000.

Statistics

When analysing the relationship between primary production and fish production, we first calculated a critical correlation coefficient based on the assumption that the observations were independent. This corresponded to a simple linear regression ( Wonnacott and Wonnacott, 1985 ). The critical correlation coefficient (r _crit = 0.55 having α = 0.05 and 13 − 2 = 11 degrees of freedom) was obtained from statistical table R in Rohlf and Sokal (1995) . Simple linear regression was also performed on cod production that was segregated into age groups.

There were, however, signs of autocorrelation in the data ( Figure 14 ). We calculated the effective number of degrees of freedom by the “modified Chelton method” proposed by Pyper and Peterman (1998) , which is outlined the Equations (1)–(3) , and the results are shown in Table 1 .  

(1)

where r _XX (j) is the autocorrelation coefficient of time-series X at lag j, N = 13 is the sample size, and X _arit is the arithmetic mean of time-series X. The same formula was used to calculate r _YY (j).  

(2)

where N* is the “effective” number of degrees of freedom.  

(3)

where r _crit is the critical correlation coefficient, α = 0.05 (two tails), and N* − 2 denotes the number of degrees of freedom.

Table 1

Results from the statistics with autocorrelated data. Index of primary production was the independent variable (x) and the dependent variables (y) were: (i) cod (ages 2.5 and older), (ii) cod (ages 1.5 and older), (iii) cod (ages 0.5 and older). N* denotes the corrected number of degrees of freedom. Rxy denotes the correlation coefficient between y and x.

	Primary production vs.
Parameter	Cod ages 2.5+	Cod ages 1.5+	Cod ages 0.5+
1/N*	0.101	0.110	0.111
N*	9.95	9.05	8.98
t α,N*−2	2.36	2.36	2.45
Critical Rxy	0.64	0.66	0.68
Observed Rxy	0.76	0.72	0.61

Another way of dealing with the autocorrelation (the long-term low frequency signal) was to relate short-term changes (from one year to the next) in primary production to short-term changes in cod production. In all, 12 such comparisons or “trials” were possible for the 13-year-long time-series. We counted the number of “successes” (decrease or increase in both variables) and “failures” (decrease in one variable and increase in the other). The probability of a failure is 0.5 if the null-hypothesis (no correlation between primary production and cod production) is correct. We observed only two failures out of 12, which gives a p-value of 0.02 according to the binomial distribution.

Results

Hydrography

Owing to precipitation, retention of the water mass, and shallow bottom depths the salinity is always somewhat lower on the Faroe Shelf than in the surrounding ocean, as shown in the example in Figure 3a . Based on the isohalines, an approximate position of the front between the oceanic and the shelf water can be identified. It usually follows the bottom contour and is generally situated between the 100- and 150-m isobaths.

Figure 3

(a) Salinity at 50-m depth and (b) mean nitrate concentrations (μM) at 5–60-m depth, 23 June–1 July 2000.

Plankton production

The retention of the Faroe Shelf water masses results in the amount of nutrients being limited to the amounts that are pooled in the shelf water at spring plus the net influx of nutrients during the productive season. During summer, the nutrient concentrations may decrease down to very low levels in the shelf water ( Figure 3b ).

Frequent nitrate measurements in the shelf water since 1995 have shown that the concentrations in most years start to decrease in May. It decreases rapidly during the spring bloom, and in most years it decreases much more than offshore. It usually reaches a minimum in July and then slowly increases again. By November, it again has reached winter levels, which are around 12–12.5 μM ( Figure 4 ). There is large interannual variability in the timing of the decrease in nitrate concentrations (timing of the spring bloom) as well as the minimum level in nitrate concentrations in summer. Generally, the years with earliest spring blooms coincide with those that have the lowest nitrate concentrations in summer and vice versa.

Figure 4

Nitrate concentrations (μM) at station C between May 1995 and December 2003.

There was very high interannual variability in nitrate concentrations in summer during the 1990–2003 period ( Figure 5 ). In 1990–1992 and 2002–2003, the nitrate concentrations remained at a high level during summer. In other years (mainly 1994, 1995, 2000, and 2001), the nitrate concentrations decreased much in summer.

Figure 5

Mean nitrate concentrations (μM) at stations S1 and S2 on 26 June, 1990–2003.

Based on the reduction in nitrate concentrations in the central shelf water and calculated net influx rates of nitrate, an approximate index of a potential new primary production from spring to midsummer each year is calculated ( Figure 6 ). This index shows that the potential new primary production has fluctuated by a factor of about 5 during the period between 1990 and 2003.

Figure 6

Index of the potential new primary production on the Faroe Shelf from winter to 26 June 1990–2003.

The y-axis in Figure 6 expresses an index only representing calculated approximate losses of nitrate (μmol l ⁻¹ ) from spring to midsummer. Using a Redfield ratio of C:N = 106:16 and assuming a mean bottom depth of 80 m, the index corresponds to a potential new primary production ranging from about 18 gC m ⁻² in 1990 to about 103 gC m ⁻² in 2000 and 2001, or a mean value of ∼60 gC m ² between spring and midsummer. Since the shelf water covers an area of about 10 000 km ² , the mean potential new primary production in this water mass from spring to midsummer may have been about 6 × 10 ⁵ t C between 1990 and 2003. Estimation of the total annual primary production is difficult and can only be approximate. However, assuming that the spring–midsummer period stands for about 60–70% of the annual production and the total primary production is about twice the new production, the mean annual primary production on the shelf can be estimated to about 2 × 10 ⁶ t C year ⁻¹ .

Cod production

Cod production in the period 1961–2002 is shown in Figure 7 and varied by a factor of 9 (between ca. 8000 and ca. 70 000 t), the most productive period being 1972–1976. Prior to and during the collapse in the catches during the early 1990s, there were 8 successive years (1985–1992) of average or low production.

Figure 7

Biomass (ages 2+), individual growth rates (arithmetic mean for ages 2.5–3.5, 3.5–4.5, 4.5–5.5, and 5.5–6.5), and production of Faroe Plateau cod (ages 1.5+) 1961–2002. Note the logarithmic scale, which implies that a fixed distance equals a fixed proportion.

Splitting cod production into age groups showed that the youngest ages contributed most to the total cod production ( Figure 8 ). The production of the youngest age group (recruiting fish: period 1.5–2.5 years old) was large when total cod production was large, and vice versa. The production from recruiting fish made up a larger proportion of the total cod production when the total cod production was large. In other words, when total cod production was low, most of it came from old cod (2.5 years or older) – i.e. a low recruitment of 2-year-old cod those years. For most of the time-series, the production of recruiting cod (1.5–2.5 years old), medium-sized cod (2.5–5.5 years), and large cod (5.5–8.5 years) varied in the same way (simple linear regression, p < 0.05) ( Figure 8 ).

Figure 8

Production of Faroe Plateau cod split into age groups. Note the logarithmic scale.

There was a positive relationship between primary production and cod production ( Table 1 , Figures 9 and 10 , p < 0.05 in binomial test). Using individual growth rates obtained from the summer survey gave the same pattern in production as that obtained when growth rates from commercial catch were used.

Figure 9

Relationship between primary production, cod recruitment, and cod production. Production of cod based on survey weights is also shown. Each data series is standardized to its own mean by dividing by the average value for the whole period. The data series are placed one unit apart from each other.

Figure 10

Cod production vs. index of primary production.

Individual growth rate

No relationship was found between individual growth rate and stock abundance (linear regression, r ² = 0.10, p > 0.05) ( Figure 7 ). A significant (linear regression, r ² = 0.36, p < 0.05) positive correlation was, however, found between individual growth rate (year t to t + 1) and stock abundance the following year (beginning of year t + 2). As expected (since production is calculated as number of fish multiplied by individual growth), a significant positive relationship (linear regression, r ² = 0.70, p < 0.05) was found between total biomass and production of cod ( Figure 7 ), and between individual growth and production of cod (linear regression, r ² = 0.46, p < 0.05).

Primary production, 0-group index and recruitment

No correlation was found between a 0-group index and year-class strength ( Figure 11 ). Relating primary production to year-class strength of cod (recruiting 2 years later to the fishery) gave no correlation (not shown), but there was a positive correlation [r ² = 0.49 fitted with an exponential function, p (corrected for autocorrelation) < 0.05] between primary production and recruitment, as measured by the number of 2-year-old cod on 1 January the following year ( Figure 12 ). The abundance of 1- and early 2-group cod was not always indicative of year-class strength ( Figure 13 ), as observed for the 1997 and 2001 year classes.

Figure 11

0-group index and year-class strength of cod on Faroe Plateau.

Figure 12

Relationship between the index of primary production and recruitment of 2-year-old cod at the beginning of the following year. Closed circles indicate year classes 1989–1999. Open circles indicate the 2000 and 2001 year classes.

Figure 13

Relationship between primary production (12 months before) and catch per unit effort (numbers per trawl hour) of 1-group cod in August and 2-group cod of the same year class in March and August the following year.

Faroe Shelf area and deeper areas

The index of primary production mainly applies to the Faroe Shelf area (within the 130-m isobath), where the majority of young cod (ages 1–3), which are contributing most to the total cod production, were caught, although there was a considerable variation between years ( Figure 14 ). There was a positive relationship between the proportion of age 3 and most of the older ages (simple linear regression, p < 0.05).

Figure 14

Proportion of cod caught in the Faroe Shelf area (bottom depth < 130 m) in the summer groundfish survey.

Although cod 4 years and older, haddock 1 year old, and saithe 3 years and older were most abundant outside the Faroe Shelf area ( Table 2 ) they did not ruin the relationship between primary production and fish production. It was therefore investigated whether food availability inside the Faroe Shelf area and the deeper areas could have co-varied. Analysis of interannual variation in Fulton condition factors for species abundant in both areas indicates that this may be the case ( Figure 15 ).

Figure 15

Fulton condition factors of fish caught inside the Faroe Shelf area (depth < 130 m) vs. fish caught outside (depth > 130 m) during the summer groundfish survey. The values are standardized to own mean and each point represents 1 year.

Table 2

Proportion of fish (percentages) caught in the Faroe Shelf area (inside the 130-m isobath) in the Faroese summer groundfish survey 1996–2003. Blank cells indicate no data.

Species	Year	Age 1	Age 2	Age 3	Age 4	Age 5	Age 6	Age 7
Cod	1996	98	83	51	33	31	32	35
	1997	63	88	80	69	45	22	25
	1998	86	63	49	38	28	11	10
	1999	62	75	67	61	53	46	17
	2000	93	69	43	28	20	20	18
	2001	74	75	66	60	54	51	49
	2002	94	89	82	70	61	51	48
	2003	92	69	58	47	32	26	18
	Average	82.8	76.4	62	50.8	40.5	32.4	27.5
Haddock	1996	38	61	78	75	72	74	73
	1997	22	51	64	72	69	68	71
	1998	60	51	55	67	70	66	66
	1999	46	59	71	63	63	62	60
	2000	68	74	80	82	78	76	76
	2001	50	73	68	65	62	60	60
	2002	56	77	81	73	70	73	74
	2003	53	61	83	82	74	67	70
	Average	49.1	63.4	72.5	72.4	69.8	68.3	68.8
Saithe	1996		19	9	6	1	0	0
	1997	100	29	22	16	10	3	1
	1998		61	8	3	1	1	0
	1999	100	92	49	10	2	1	0
	2000	0	88	5	4	1	0	0
	2001			2	2	2	1	0
	2002			40	13	7	2	0
	2003			2	2	1	0	0
	Average	66.7	57.8	17.1	7.0	3.1	1.0	0.1

Discussion

Overview

The collapse of the cod stock in the beginning of the 1990s and the rapid recovery in the mid-1990s was closely linked to the productivity of the Faroe Shelf ecosystem as measured by phytoplankton productivity. Phytoplankton production strongly affected recruitment (as measured by number of 2-year-old cod on 1 January the following year) and production of cod during a period of 1 year after the spring bloom. This was supported by two different statistical approaches.

Relationship between shallow (<50 m), intermediate – Faroe Shelf – (50–130 m), and deep (>130 m) areas

No data on cod productivity in the shallow areas are available. Thus, it is not known whether it fluctuates in the same way as for the Faroe Shelf area. On the other hand there are indications that the productivity of the deeper areas fluctuates in the same way as for the Faroe Shelf area, as shown in Figure 15 . The production of cod in the deeper area seems to be about 50% of the production in the Faroe Shelf area (about one-third of the total production), based on the distribution of cod in Table 2 . This may, however, be an overestimate since the proportion of mature cod in the Faroe Shelf area is higher in the spawning season. Whether the productivity of the deeper area is driven by the primary production in the Faroe Shelf area or other factors are affecting both systems in the same way is not known.

Primary production

Due to low irradiance, primary production levels are very low during winter but increase during spring and summer ( Gaard, 1996 , 2003 ). Since the water mass on the shelf and hence the total amounts of nutrients in this water mass are limited, the primary production affects the nutrient concentrations in the shelf water ( Figure 4 ). New primary production is primarily based on nitrate as nitrogen source ( Dugdale and Goering, 1967 ), and potential new primary production is therefore limited to the nutrient pool in this water plus the advection of nutrients into the shelf during the summer.

During the 14-year period for which nitrate data from the Faroe Shelf are available, the decrease in nitrate concentrations during summer has varied dramatically between years ( Figure 5 ). As well as levels of nitrate assimilation, variable exchange rates may also affect nitrate concentrations. However, the variability in nitrate influx (due to variable renewal rates of the shelf water) is exhibited markedly lower than the nitrate loss during spring and summer ( Gaard and Hansen, 2000 ; Gaard, 2003 ). Therefore, this suggests that the nutrient changes are mainly due to variable assimilation. The new primary production from spring to midsummer seems to have fluctuated by a factor of 5 during the 1990s. In the summers of 1995, 2000, and 2001 it seems to have been close to the upper limit for what is possible, based on the nutrient pool and supply in the ecosystem.

The mechanisms controlling the primary production on the shelf are not well understood. Earlier studies have shown no correlation between primary production (neither timing of spring bloom nor production level) and light irradiance ( Gaard et al. , 1998 ; Gaard, 2003 ), and modelling indicates that if light was the only controller the spring bloom would have occurred earlier than it does in nature. Hence, other controlling mechanisms apparently are present. Gaard et al. (1998) and Gaard (2003) have shown a clear inverse relationship between potential new primary production and zooplankton biomass (variable influx of the copepod Calanus finmarchicus from offshore areas) and it has been hypothesized that the observed plankton variability could be due to variable grazing from C. finmarchicus . Recent modelling studies do, however, indicate that a main controlling factor on timing of the primary production may be variable exchange rates of the shelf water ( Hansen et al. , submitted ).

Energy transfer through foodwebs

Fish (and seabird) production in the Faroe Shelf ecosystem (inside the ∼130-m depth contour) is around 100 000 t wet weight or about 5000–10 000 t C year ⁻¹ . This is the same magnitude as would be expected in the third trophic level, assuming a mean ecological efficiency of 0.15. Thus, primary production seems largely to be generated within the shelf ecosystem and it seems to be efficiently transferred up to higher trophic levels in the ecosystem. Furthermore, the relationship between primary production and fish production suggests that food is rapidly used in the ecosystem and that fish production is probably food limited.

Results from studies on the lowest trophic levels in the Faroe Shelf ecosystem since 1990 give an indication on how the variable phytoplankton production may be trophically connected to the co-occurring variability in cod production. Phytoplankton production (timing and magnitude) seems to be very variable between years ( Gaard et al. , 1998 ; Gaard 2003 ) and affects zooplankton reproduction and composition on the Faroe Shelf ( Gaard, 1999 , ( 2000 ). During years with low and late primary production in the early 1990s copepod reproduction occurred late in spring. There were high concentrations of large overwintered Calanus finmarchicus (advected from offshore) and lower concentrations of small-sized copepods (nauplii and young copepod stages of C. finmarchicus and of small-sized neritic species, mainly Acartia longiremis and Temora longicornis ). This changed markedly in years with high and early primary production ( Gaard, 1999 ), when the concentrations of small locally produced nauplii and copepodites during spring were substantially higher. During early spring, the cod larvae depend on small prey such as phytoplankton and copepod eggs and nauplii ( Gaard and Steingrund, 2001 ), and therefore their feeding conditions depend highly on plankton production in early spring. Although no feeding studies are made on sandeel larvae on the Faroe Shelf, their feeding conditions can be assumed to vary similarly to those for cod larvae, as they feed in the same areas and at the same time of the year, and recruitment of sandeel juveniles seems to a large extent to co-fluctuate with plankton production on the shelf (unpublished 0-group results).

When sandeels are abundant they are a preferred food item for cod on the shelf ( Rae, 1967 ; Du Buit, 1982 ) and hence affect the feeding conditions for demersal cod on the shelf already during the first year after recruitment of the sandeel. Stomach analysis show that in bottom depths less than 200 m on Faroe Plateau sandeels and benthic crustaceans ( Galathea sp., Paguridae, Hyas sp., and Portunidae) dominate the diet of cod ( Rae, 1967 ; Du Buit, 1982 ). When sandeels are abundant they form the principal food item for cod. In deeper areas where Norway pout is more abundant, it is an important prey ( Du Buit, 1982 ; Nicolajsen, 1993 ; unpublished data). Years with high cod production seem to be associated with a high abundance of sandeels ( Gaard et al. , 2002 ; unpublished data).

Variability in primary production (timing and magnitude) on the Faroe Shelf therefore seems to affect feeding conditions for fish larvae (e.g. cod and sandeel larvae) through production of small-sized zooplankton in spring, and also feeding conditions for post-settlement cod through variable sandeel production. Thus, two main energy pathways from phytoplankton to cod can be set up:

• Phytoplankton → zooplankton → sandeels or other zooplankton predators → cod.

• Phytoplankton → benthic organisms (mainly crustaceans) → cod.

If the relationship between phytoplankton production and cod production (during the first year after the phytoplankton bloom) is linked by these two energy pathways, it should be expected that most cod prey are less than 1 year old. This seems be the case for energy pathway 1, as the length of sandeels in August is 60–80 mm (unpublished data). Benthic crustaceans in energy pathway 2 are, however, most likely older than 1 year (unpublished data) and should thus be a disturbing factor in the relationship between phytoplankton production and cod production. It is possible that energy pathway 1 dominates over energy pathway 2 and ensures the relationship between phytoplankton production and cod production, but no analysis is yet available to confirm this. The rapid response of the cod production to the primary production (within less than a year) should be shorter than in areas such as the Barents Sea, where the most important prey for cod is 1–2-year-old capelin ( Jørgensen, 1992 ; ICES, 2002b ).

Several other factors may contribute to ensure a close relationship between phytoplankton production and cod production. The age groups of cod, which contribute most to the total cod production (ages 1–4), are largely inside the Faroe Shelf ecosystem ( Table 2 , Figure 14 ). Second, the opportunistic and flexible feeding behaviour ( Dill, 1983 ) of cod ( Brawn, 1969 ) may ensure that cod are able to catch the energy flow through the foodweb by preying on abundant prey species. Third, the fluctuations in primary production are very large (by a factor of 5).

Cod production vs. individual growth

Production of fish, as defined in this paper, is the product of two factors: abundance of fish and individual growth rate. The abundance of cod is, in turn, dominated by young age groups, especially the recruiting age group (age 2). Two extreme cases could be set up: (i) cod recruitment is the dominating factor, whereas the individual growth rate of older fish is constant. (ii) The individual growth rate of cod varies in the same way as the primary production, whereas the recruitment is constant.

The reality seems to lie between these extremes. Figure 8 shows that the production of small, medium-sized, and large cod covaries for most of the study period, most likely because the primary production affects all ages of cod. The small cod, newly recruited to the fishery, dominate the production, probably because the survival of recruiting cod is linked to food availability (see recruitment of cod later in the discussion). The year-class strength of cod older than 2 years is considered to be established well enough that variations in production are caused by variations in the individual growth rate. Thus, the individual growth rate of cod may be regarded to be a result of food availability and competition from other cod. Put otherwise, the individual growth is food and density-dependent. A positive relationship between individual growth rate and food availability is also observed in the Barents Sea ( Jørgensen, 1992 ) and Icelandic waters ( Astthorsson and Vilhjálmsson, 2002 ). In Canadian waters, size-at-age and condition factors of cod in the northern Gulf of St. Lawrence declined as the production of cod declined ( Dutil et al. , 1999 ).

It is often expected that there should be a negative relationship between individual growth rate and stock biomass of cod, i.e. a density-dependent effect, but this was not observed ( Figure 7 ). The reason seems to be the varying carrying capacity of the ecosystem (indicated by cod production) which overrides density-dependent effects on individual growth rates. In areas where the annual productivity is more stable, a negative relationship between individual growth rate and abundance of cod is sometimes observed, e.g. in the North Sea ( Houghton and Flatman, 1981 ) and in the Barents Sea ( Jakobsen, 1992 ).

Recruitment of cod

The peak spawning of Faroe Plateau cod is in the second half of March ( Jákupsstovu and Reinert, 1994 ), and eggs mainly hatch in April. The larvae drift with a residual current ( Hansen, 1992 ) that flows anticyclonically on the Faroe Shelf ( Gaard and Steingrund, 2001 ). In spring, the larvae prey mainly on small zooplankton (copepod eggs, nauplii, and small copepodites). As they grow during spring and summer, they gradually progress to larger prey ( Gaard and Steingrund, 2001 ). In late June, pelagic cod juveniles are found all over the Faroe Shelf area as well as outside, where they prey on copepods, Malacostraca, and fish larvae ( Gaard and Reinert, 2002 ). In July–August, the fry have settled to the bottom, and are found in dense vegetation of macroalgae in shallow waters (0–20 m) ( Tåning, 1943 ). The biology of juvenile cod is poorly known, but they feed on a variety of prey items, including benthic and pelagic crustaceans (Dánjal Petur Højgaard, Faroese Fisheries Laboratory, pers. comm.).

Juvenile cod recruit to the fishery at an age of 1–2 years (30–45 cm). They enter the fishery mainly during autumn (from October and onwards, unpublished results) and are mainly found within the Faroe Shelf area ( Table 2 ). Because no stations shallower than appoximately 60 m are sampled during the groundfish surveys, where the relative abundance of juvenile cod is expected to be highest, the percentages in Table 2 most likely are underestimated.

Food availability is crucial for survival of newly hatched yolk-sac larvae ( Cushing, 1995 ), and for some stocks, e.g. in the Barents Sea, year-class strength seems to be determined at an age of approximately 3 months ( Helle et al. , 2000 ). Density-dependent processes may, however, regulate survival later in life ( Fromentin et al. , 2000 ). Reinert (1988) found a weak positive relationship between a 0-group index 1974–1984 and the year-class strength of Faroe Plateau cod estimated from a traditional VPA as number of 1-year-old cod. However, expanding the time-series to include the most recent years, the overall relationship between the 0-group indices and the VPA estimates ( Figure 11 ) is not statistically significant ( ICES, 2004 ; Jákup Reinert, Faroese Fisheries Laboratory, pers. comm.). Results from the summer groundfish survey indicate that there are annual variations in the survival of juvenile Faroe Plateau cod as late as their second winter. There is normally a high correlation between the abundance of 1.7-, 2.3-, and 2.7-year-old cod (in August, and March and August the following year). However, when the abundance of 1.7-year-old cod is high and primary production is low (i.e. little food for 12 months), as observed for the 1997 and 2001 year classes, this relationship does not hold ( Figure 13 ). Preliminary investigations of stomach data during the period 2000–2003 indicate that natural mortality (cannibalism) of 1-group cod is high and may range between 0.24 and 2.4 (21–92%).

Thus, it seems likely that the year-class strength of Faroe Plateau cod could be determined at a relatively old age (late 1-group or early 2-group) compared with other areas ( Helle et al. , 2000 ). This may reflect the habitat shift from the shallow (<c. 20 m) nearshore areas to the deeper (>60 m) feeding areas of older cod. The habitat shift could by itself represent a bottleneck of juvenile cod survival, since the fish experience novel predators and prey. Alternatively, the food availability in the feeding areas of older cod could be a bottleneck of survival of recruits, which is supported by Figures 9 , 10 , 13 . The mechanism could be food-mediated mortality, i.e. slow growth–high mortality ( Nordeide et al. , 1994 ), and older cod could probably be an important predator (i.e. cannibalism), since both 0-group and 1-group cod are occasionally found in stomachs of cod on the Faroe Plateau (unpublished data).

Even though there is some doubt when the year-class strength of Faroe Plateau cod is determined, it seems clear that the primary production (and hence food availability some months later) is a major factor controlling recruitment ( Figure 12 ). The recruiting cod (age 2) usually dominate the cod production (age 1.5–2.5, Figure 8 ), thus ensuring the relationship between primary production and cod production. The primary production has, as stated earlier, a similar effect on the production of older age groups of cod ( Figure 8 ). The actual allocation of the total available food production to the age groups of cod (recruiting or older cod) depends on the outcome of the competition between them, but older cod (3 years or more) seem to be the stronger competitors since their part of the total cod production is larger when the total cod production (and primary production) is low. It should be noted that the diet of 2-year-old cod and 3–4-year-old cod usually is very similar (unpublished data). Thus, it can be concluded that the primary production affects all ages of cod ( Figures 8–10 , 13 ) and other fish species as well.

Implications for stock assessment and management considerations

Worldwide, there is a strong desire to include holistic, ecosystem-based information in fisheries management ( FAO, 2001 ; ICES, 2002c ). However, because of insufficient knowledge about most ecosystems, this holistic approach has not been generally adopted. The relationship between primary production and fish production on the Faroe Shelf makes this particular ecosystem a distinct candidate for such inclusive, ecosystem-based management practices.

Results presented here demonstrate links between primary production, recruitment and growth of Faroe cod that may have the potential to be included in more holistic stock assessments. This will require the development of new stock assessment procedures and should be preceded by simulations of likely improvements in management [see references in Basson (1999) ].

There is no basis to conclude from this study that the size of the spawning stock is of no importance to recruitment. The fact that no correlation can be identified between cod spawning stock size and recruitment does not by itself prove that no relationship exists. It might as well indicate that this relationship is highly interannually variable, due to variable food production, which obviously affects the recruitment variability more than stock size variability alone. If the spawning stock is reduced to a very low level, there might be a problem in attaining good recruitment, as observed for cod in Canadian waters (although some doubt still exists) ( Drinkwater, 2002 ).

In most years, the nearshore nursery areas seem to provide enough young cod to secure recruitment of 2-year-old cod. This is probably not the case for the deep areas (>130 m). A high fishing mortality may prevent the deeper areas to be fully utilized by cod, thus reducing total cod production. It may be speculated whether the production of cod during the very high phytoplankton production in 1999–2000 has been constrained by the high fishing mortality (F = 0.54 and 0.37, respectively; ICES, 2004 ) compared with the former productive period in 1994–1995 (F = 0.19 and 0.32, respectively) when cod production was higher ( Figure 9 ).

Almost every person working now or previously at Faroese Fisheries Laboratory and the research vessel “Magnus Heinason” has to some extent been involved in the process of generating the data used in this study. Thanks to Jákup Reinert, Faroese Fisheries Laboratory, for providing us with the 0-group indices. Thanks also to Einar Hjörleifsson at Marine Research Institute, Iceland, and two anonymous referees for valuable comments on earlier versions of the manuscript.