Estimating spawning stock biomass using egg surveys

Why do we carry out egg surveys?

Many fish species spawn huge numbers of eggs into the plankton. The eggs disperse in the plankton and pass through a series of well-defined stages of embryonic development before hatching into larvae. The egg stage for species such as cod and plaice can last up to two or three weeks in UK waters. Planktonic eggs are easy to sample using a variety of fine-mesh nets, and the catch rates are not affected by factors such as bottom type and fish behaviour that can affect catches of fish during trawl surveys. The total numbers of eggs spawned into the plankton by the mature stock is proportional to the spawning stock biomass, in accordance with the average fecundity of the fish (numbers of eggs produced by each individual female).

Egg surveys are therefore useful for:

• Mapping spawning grounds and seasonal spawning cycles;
• Identifying discrete spawning grounds that could represent sub-populations of fish;
• Estimating spawning stock biomass or trends over time

How do we sample planktonic eggs?

Some egg surveys, for example for anchovies and sardines, are carried out by vertically hauling a fine-meshed net from near the sea bed to the sea surface on a grid of stations covering the spawning grounds. This provides a large number of discrete estimates of egg abundance. Egg surveys for species such as mackerel, cod and plaice in European waters use a high-speed plankton sampler deployed at 3 - 5 knots, towed in a double-oblique haul through the water column from the surface to near the seabed and back again. The sampler filters a larger volume of water than is possible with a vertical haul, and the ship does not need to stop at each station, saving some time.

Fig. 1.Gulf-7 plankton sampler
used by Cefas for egg surveys

Figure 1 shows a typical sampler of the "Gulf-7" variety used by Cefas (Nash et al., 1998). The front of the sampler is fitted with a conical fibreglass cone tapering down to a 40cm opening. Behind this is a fine-mesh (270 micron mesh) net terminating in a removable pot that collects the filtered eggs and other plankton. The sampler is encased in a rigid frame that also carries electronic equipment to monitor depth and sea temperature during the haul. The front aperture of the net is made small enough to ensure that the volume of water entering the net per unit time is no more than the amount that can be filtered through the much larger area of fine-mesh netting in the same time, thus avoiding back-wash problems. Flow meters are mounted inside and outside the net to measure the actual volume of water flowing through the net. One can be seen in the front opening of the net pictured.

During a double-oblique tow, the net will pass through patches of eggs at various depths, as shown in Fig. 2:

Fig. 2.  Schematic representation of a double-oblique tow sampling several patches of eggs in the water column.

The numbers of eggs caught per cubic metre of water filtered is usually converted to mean numbers per m² of sea surface area using data on total volume filtered and the distance and depth towed. This is done to allow the mean numbers per square metre to be raised up to the total area of the spawning grounds covered by the survey.

Mapping egg distributions

Eggs occur in patches of varying sizes depending on the way that spawning fish aggregate and also the rate at which eggs disperse due to tides and wind. If the patches are small, many of the plankton samples will contain zero or very few eggs. If the patches are large, a larger fraction of the stations will have eggs and the distribution will appear more continuous. This is illustrated in Fig. 3. An actual survey distribution of eggs, from the triennial ICES mackerel egg survey, is shown in Fig. 4 (ICES, 2005).

Fig. 3. Illustration of how the sizes of egg patches (green circles) determine the appearance of egg distribution in a survey (crosses indicate plankton stations). Top: small patches, many zero and small egg catches. Bottom: larger patches, fewer zeros, distribution appears more continuous.

The drift of eggs with the tide can affect the apparent distribution, but is only a problem if a patch of eggs moved continuously in the same direction as the ship, and at the same rate, so that it was over-sampled. In practice, the eggs slop back and forth with the daily tidal cycle and we assume that over the period of a survey the drift of eggs has a random rather than systematic effect on catch rates. This contributes to the overall tow-to-tow variability in catch rates, which dictates the survey design and the amount of sampling needed in different areas.

Fig. 4: Daily egg production of western and southern mackerel in March-April 2004. Size of symbols is proportional to egg production; plus signs = zero (reproduced with permission from ICES).

Estimating the average density of eggs in a survey, and the total abundance

Egg surveys can sample only a small fraction of the spawning habitat in the vicinity of each station, and hence the total number of eggs in the area covered by the survey has to be inferred from the average catch rate in the plankton tows.

One approach is to take samples at completely random locations. If this procedure was repeated many times, the average numbers of eggs per m² from the repeat surveys would follow a bell-shaped curve. The extremes would represent surveys in which many of the random samples were located on high density or low density patches purely by chance, whereas the peak of the curve would lie very close to the true average density. The spread of the curve reflects the sampling error – the broader it is, the less precise a single survey is likely to be (i.e. a larger probability of obtaining a result quite different from the true value). Because the spread of the curve is related to the patchiness of the individuals, it is also reflected in the variability between stations in a single survey, and this variability is used to infer the precision of the survey.

Simple random sampling has the disadvantage that there may be large gaps between some stations, and other stations may be very close together. This is undesirable for mapping spawning distribution patterns, and a lot of time may also be spent surveying areas that have very low egg densities. The survey design can be improved by dividing the area into several areas or “strata” defining, for example, areas of low, medium or high egg abundance, based on prior knowledge of spawning patterns. Placing more stations per unit area in the high-density strata will improve precision because such strata will contribute more to the overall egg abundance and variability in abundance than the strata with few eggs.

A third approach is to place stations more regularly across the survey area rather than at random. This can also be done using separate strata with different station spacing according to expected egg abundance. The theoretical example in Fig. 3 follows this systematic/stratified approach, and is in fact the method currently used for egg surveys of Irish Sea cod, haddock and plaice and for NE Atlantic mackerel (Fig. 4). The method provides the best approach for mapping the distribution of eggs and also provides statistically valid estimates of abundance since the distribution of eggs itself has a random element due to fish distribution and egg drift.

The total abundance of eggs in the area covered by a survey is estimated by converting the catch rate of eggs at each station into numbers per square metre of sea surface, averaging over all stations in each stratum, then raising the averages to the surface area of each stratum. Summing across strata then gives an estimate of total abundance for the survey area. More elaborate modelling approaches can also be used to better represent the smooth “topography” of egg densities that is often apparent (Fox et al., 2000).

Converting to daily egg production using egg development rates

Typically, the fish eggs caught at a station represent survivors from several days or even weeks of spawning. The eggs will be a mixture of different development stages, as illustrated in Fig. 5. As we are interested in calculating the average numbers of eggs produced daily, in order that the seasonal pattern of egg production can be estimated from a series of surveys, it is necessary to know how long it takes an egg to pass through each recognisable stage of embryonic development. The development rate is closely related to ambient temperature, and detailed experiments have been conducted at Cefas and elsewhere to determine the relationship between temperature and stage-duration for species such as cod, plaice and mackerel.

Fig. 5. Cod eggs at the first development stage (1A) and a later stage (stage 2) when the embryo is more developed.

For example, if the earliest development stage (1A) lasted two days at the temperature recorded at a plankton station, then the average daily production of stage 1A eggs at that station would be the total number of stage 1A eggs divided by 2. This procedure is normally carried out on a station-by-station basis to allow for the spatial patterns in temperature:

Daily production at stage s = (nos of stage s eggs m^-2 / expected stage s duration)

In practice, egg production surveys of species such as mackerel and cod in European waters tend to use only the data from the earliest stage eggs, to avoid problems caused by high and variable mortality rates during subsequent development. For some other species, such as anchovy, where the egg stage is relatively brief, daily production is estimated for all the egg stages and a mortality curve is fitted to the production-at-stage data to improve the estimate of initial production at age zero.

The problem of egg identification

Some species have eggs that are uniquely identifiable from their size, shape and morphology of the embryo. However, species such as cod, haddock, whiting and witch have identical-looking eggs during early embryonic development. To apply egg surveys to these species, it is necessary to apply biochemical methods. Cefas, in conjunction with the University of East Anglia, has developed gene-probes to identify eggs of cod, whiting and haddock, and these have been used very successfully in large-scale egg surveys of the North Sea in 2004 and the Irish Sea in 2006.

The technical details of the genetic method for distinguishing eggs of cod, haddock and whiting are described in Taylor et al. (2002) and a previous application in the Irish Sea in Fox et al. (2005).

Calculating seasonal egg production

As spawning activity builds up during the spawning season, the abundance and often the spatial coverage of the eggs builds up to a peak then declines again. This is observed in the daily egg production estimates from a series of egg surveys conducted throughout the spawning season. An example is given in Fig. 6 for plaice in the Irish Sea in 1995 and 2000

Fig 6. Estimates of total daily egg production of plaice (stage IA) in the Irish Sea in 1995 and 2000.  Error bars denote 1 standard error. 

By integrating under this seasonal egg production curve, the total annual egg production can be estimated. It is this figure that can be related to the average fecundity of individual spawners to give an estimate of spawning stock biomass, as described in the next section. 

Calculating spawning stock biomass from annual egg production

Fish species can be classified as “determinate” or “indeterminate” spawners. Indeterminate spawners such as anchovy tend to have long spawning seasons, and the fish produce new batches of eggs at intervals during the season. Hence the total annual number of eggs to be spawned cannot be determined in advance. For determinate spawners, which includes cod, plaice and mackerel, each mature female produces all the eggs it will spawn during the season before spawning commences. By sampling the females at random immediately prior to the spawning season, an estimate can be obtained of the average potential fecundity (total number of eggs) of mature females in the population, i.e. the potential number that could be spawned during the year. This is done by taking small samples of ovary tissue (Fig. 7) and counting the number of yolked eggs (oocytes). Small eggs without yolk are not counted as these will not be spawned in the coming season. The number of eggs in the small tissue samples is raised up to the total number of eggs in the ovary, by multiplying the ratio of ovary weight to tissue sample weight.

This estimate of fecundity is called the “potential fecundity” because not all of the eggs will in fact be spawned. Some of the eggs are re-absorbed by the fish, a process called “follicular atresia”. Rates of atresia can be elevated when the eggs are developing, and it is best to sample for potential fecundity after this process has more or less stopped but before spawning has commenced. Atresia can also occur during the spawning season, and regular sampling is required to estimate the numbers lost this way. In the Irish Sea, atretic losses during spawning have been low in cod and plaice, but high in sole.

By sampling a range of fish sizes as randomly as possible immediately prior to the spawning season, the relationship between potential fecundity and fish size can be determined. Fig. 7 shows an example for Irish Sea cod in 2006. Total fecundity increases more or less linearly with fish weight. If expressed as numbers of oocytes per gram of fish weight, there is no clear relationship with size (length is shown in Fig. 7).

To estimate the spawning stock biomass of female fish from egg surveys, it is necessary to divide the estimate of total annual egg production from integrating under the seasonal egg production curve (e.g. from data as shown in Fig. 6) by the average fecundity per unit weight (e.g. as shown in Fig. 7), after correcting for losses due to atresia. This gives the average biomass of spawning females during the spawning season. To extend this process to estimate total spawning stock biomass of males and females it is necessary to have an estimate of the sex ratio in the population.

Fig. 7. Photo: removing a tissue sample from a cod ovary to estimate fecundity. Plots: potential fecundity (nos. of eggs) of Irish Sea cod in 2006. Top plot is total number of eggs vs fish weight. Bottom plot is the number of eggs per gram of fish weight, plotted against total length.

A worked example: Irish Sea plaice

During 2000, a series of eight ichthyoplankton surveys was carried out in the Irish Sea, together with sampling of adult fish, to estimate the spawning stock biomass of cod and plaice using the annual egg production method described above (Armstrong et al., 2001). Early-stage (1A and 1B) plaice eggs were found mainly over shallow sandy seabed types in the coastal region of the western Irish Sea and in the eastern Irish Sea (Fig. 8).

Fig. 8.  Abundance of stage 1A and 1B plaice eggs (nos./m²) during the first five surveys in 2000 in the Irish Sea (plaice spawning was negligible in the subsequent three surveys).

After converting egg abundance to daily egg production at each station, using the temperature-development relationship for plaice, the mean daily egg production per survey was calculated (nos/m²/day) and raised to the total for the surveyed area, and plotted against mean survey date to show the seasonal egg production cycle (Fig. 9). Egg production by the population in both areas of the Irish Sea peaked in early March. By integrating the estimates over the spawning season, an estimate of the annual egg production is obtained.

Fig. 9.  Mean daily production of stage 1A plaice eggs during the first five surveys in 2000, shown separately for the western and eastern Irish Sea and for the area as a whole. (Date is given as days from 1 January). Error bars are + 1 standard error.

Sampling of the adult population for fecundity showed a clear linear relationship between fish weight and total number of yolked eggs in the ovary immediately prior to the spawning season (Fig. 10). The relationships for the western and eastern Irish Sea were not significantly different. 

 
Fig. 10.  Relationship between potential fecundity of plaice and fish weight, for the western and eastern Irish Sea in 2000.

Based on these surveys and sampling results, the spawning stock biomass of Irish Sea plaice in 2000 was calculated as follows, using western Irish Sea data as an example:

The data and calculations for the western and eastern Irish Sea are summarised in Table 1.

Table 1. Estimation of spawning stock biomass of Irish Sea plaice using the annual egg production method in 2000. Terms are explained below the table. 

AEP = total annual egg production
F(eggs) potential = mean number of eggs per female before adjusting for atresia;
Fg(eggs g^-1) = potential fecundity expressed per gram female weight
AT (eggs g^-1) = average losses due to atresia during the spawning season, expressed per gram female weight
Fr(eggs g^-1) = realised fecundity (potential fecundity minus losses due to atresia)
SSB(t) females =  spawning stock biomass estimate for females
P = average proportion of spawning stock comprising female fish;
SSB(t) both sexes = total spawning stock biomass
CV(%) = coefficient of variation of mean (ratio of standard error to the mean, as percentage).

As also found during similar surveys in the Irish Sea in 1995, the SSB estimates for plaice from the egg surveys are well in excess of similar estimates calculated by ICES from fishery catch-at-age data. It can be difficult to reconcile such differences due to the very different way in which spawning stock biomass is estimated from egg surveys and catch-based assessments, and both approaches have strengths and weaknesses that need further evaluation (c.f. Hunter et al., 1998). There may be many factors related to the specific characteristics of the plaice stock in the Irish Sea that could be affecting the absolute biomass estimate from either the ICES assessment or the annual egg production estimates, and further work is continuing to resolve this.  For example, the catch-based analysis for plaice currently excludes discarded fish for which there are insufficient data, and may exclude significant quantities of mature fish. Also, any changes in maturity patterns over time are not accommodated. The egg surveys provide a more direct estimate of spawning stock biomass although there are still questions regarding the impact of early-stage egg mortality on the production estimates. By contrast, an estimate of spawning stock biomass of plaice in the North Sea, obtained using the annual egg production method in 2004, was much closer to the results of the ICES assessment which does include estimates of discards (ICES, 2006).

Nonetheless, the broad age profile of plaice in the commercial fishery catch in the Irish Sea indicates a declining and low rate of mortality, and the ICES assessment indicates the stock is lightly exploited with an increasing spawning stock biomass. Continued monitoring of the stock using egg surveys would provide a benchmark against which absolute biomass estimates from the ICES assessments can be evaluated once discards estimates are included. In the meantime, further insights into factors that could affect the accuracy of annual egg production estimates of SSB should be obtained as the method continues to be applied.

Some further reading

Armstrong M. J., Connolly P., Nash R. D. M., Pawson M. G., Alesworth E., Coulahan P. J., Dickey-Collas M., Milligan S. P., O'Neill M., Witthames P. R. and Woolner L. 2001. An application of the annual egg production method to estimate spawning biomass of cod (Gadus morhua L.), plaice (Pleuronectes platessa L.) and sole (Solea solea L.) in the Irish Sea. ICES J. Mar. Sci., 58, 183-203.

Fox, C.J., O’Brien, C.M., Dickey-Collas, M. and Nash, R.D.M. (2000).  Patterns in the spawning of cod (Gadus morhua L.), sole (Solea solea L.) and plaice (Pleuronectes platessa L.) in the Irish Sea as determined by generalized additive modeling.  Fisheries Oceanography, 9, 33-49.

Fox, C.J., Taylor, M.I., Pereyra, R., Villasana-Ortiz, M.I. and Rico, C. (2005).  TaqMan DNA technology confirms likely over-estimation of cod (Gadus morhua L.) egg abundance in the Irish Sea: implications for the assessment of the cod stock and mapping of spawning areas using egg based methods.  Molecular Ecology, 14, 879-884.

Hunter, J., Horwood, J., O’Brien, C., Pepin, P. and Powers, J. (1998).  Report of the meeting to examine the methodology of the Annual Egg Production Method as applied to Irish Sea and Bristol Channel stocks.  Contract MF0144 report, April 1998, Cefas: Lowestoft.

ICES. 2005.  Report of the Working Group on mackerel and horse mackerel egg surveys (WGMEGS). ICES CM 2005/G:09  www.ices.dk/products/CMdocs/2005/G/WGMEGS05.pdf

ICES. 2006. Report of the Planning Group on North Sea Cod and Plaice Egg Surveys in the North Sea (PGEGGS). ICES CM 2006/LRC:02. www.ices.dk/products/CMdocs/2006/LRC/pgeggs06.pdf

Nash, R.D.M., Dickey-Collas, M. and Milligan, S.P. 1998. Descriptions of Gulf-VII/PRO-NET and MAFF/Guildline unencased high-speed plankton samplers. Journal of Plankton Research, 20(10):1915-1926.

Taylor, M.I., Fox, C.J., Rico, I. and Rico, C. (2002).  Species-species TaqMan probes for simultaneous identification of cod (Gadus morhua L.), haddock (Melanogrammus aeglefinus L.) and whiting (Merlangius merlangus L.).  Molecular Ecology Notes, 2, 599-601.