CM 2005/K:05
Application and validation of otolith microstructure as stock identifier in mixed Atlantic herring (Clupea harengus) stocks in the North Sea and Western Baltic. Clausen, L.W1,4., Bekkevold, D2., Hatfield, E.M.C3. and Mosegaard, H1.
Abstract Herring (Clupea harengus) is a highly migratory species that perform seasonal migrations typically between spawning-, feeding-, and wintering areas. In the feeding areas of Kattegat, Skagerrak, and the Eastern North Sea different autumn-, winter- and spring spawning populations mix during parts of their life cycle. For stock assessments a split-factor based on population based mean vertebral count and an individual based, visual inspection, otolith microstructure analysis is applied to the catches in the area. The visual inspection method identifies hatch type by otolith microstructure pattern formed during the larval period assuming spawning time fidelity. To test the accuracy of the visual inspection method, otolith microstructure analyses by visual inspection on material from spawning populations were compared to the sampling season. Assuming spawning time fidelity a high accuracy would imply a high correspondence between estimated hatch type and sampling season for the spawning populations. Variation in hatch type determination may be due to reader related variability and besides this natural variability in the larval microstructure formed after hatch, overlapping of spawning seasons or mismatch of hatchand spawning time in individual herring (‘strayers’). To validate the visual inspection beyond the matter of reader variability and to resolve the influence of natural variability in both microstructure pattern and spawning season a validation method independent of the assumption of spawning type fidelity was performed analysing variability in otolith microstructure pattern in post-larval 0-ringer herring, hatched during different seasons. Keywords: Stock identification, otolith microstructure, herring 1
Danish Institute for Fisheries Research, dept. of Marine Fisheries, Charlottenlund Castle, DK-2920 Charlottenlund, Denmark 2 Danish Institute for Fisheries Research, dept. of Department for Inland Fisheries, 8600 Silkeborg, Denmark 3 FRS Marine Laboratory Aberdeen, PO Box 101, Victoria Road, Aberdeen AB11 9DB, Scotland, UK 4 Contact author: Lotte A. Worsøe Clausen, Danish Institute for Fisheries Research, dept. of Marine Fisheries, Charlottenlund Castle, DK-2920 Charlottenlund, Denmark. Phone: +45 33 96 33 64 E-mail:
[email protected]
CM 2005/K:05
Introduction Atlantic herring (Clupea harengus L) population dynamics are complex and the various stocks display high variation in terms of life history events (Jennings and Beverton 1991, McQuinn 1997a). Herring may perform extensive seasonal migrations typically between spawning-, feeding-, and wintering areas (see e.g. Slotte 1998) and different stock components mix with each other at e.g. feeding grounds (Rosenberg and Palmén 1981, Wheeler and Winters 1984, Husebø et al. 2005). There are various examples of mixed stocks of Atlantic herring that display variation in life history traits. The large Norwegian spring spawning herring stock (NSS) is distributed along the entire coast of Norway and displays extensive migrations (Slotte 1998). The spawning migration may be influenced by individual state; cost of migration and prospects of larval survival, but the season is the same (Slotte 2001). However, the small population of autumn spawning herring mix with the NSS in feeding and wintering areas and is not produces as a result of year class twinning but constitute a separate stock with different population dynamics (Husebø et al. 2005). In the North Sea there are a series of spawning sites along the UK coast. Previously the herring in the North Sea were divided into several minor stocks (Cushing 1967, Hulme 1995) but more recently the herring is divided into two main stocks; the North Sea Autumn Spawners (NSAS) and the English Channel winter spawning Down’s herring (ICES 2004). These populations mix in nursery and feeding grounds in the North Sea as well as subdivision IIIa (Cushing 1967, Rosenberg and Palmén 1981, Hulme 1995). Although meristic characters like vertebral counts and otolith microstructure does differ between these herring to some extend (Hulme 1995, Mosegaard and Madsen 1996) no significant genetic difference has been demonstrated between the stock components (Bekkevold et al. 2005). The preservation of such a complex stock structure necessitates knowledge of how these migratory components of various stocks overlap both spatially and seasonally. The ICES working group on Herring Assessment for the Area South of 62ºN (HAWG) has for individual stock assessments applied a split-factor to the catches in the area to separate North Sea autumn-winter spawners from western Baltic spring spawners. In the past herring stock components have been defined based on phenotypic differences in various meristic characteristics (Cushing 1955). Prior to 1996 the splitfactor used in the HAWG was calculated from a population based mean vertebral count. In the period from 1996 to 2001 splitting keys were calculated by using a combination of the vertebral count and otolith microstructure methods (ICES 2001). From 2001 and onwards, the splitting keys have been calculated based on a stock separation based solely on an otolith microstructure method (ICES 2004). The larval otolith microstructure is found in the central part of the adult herring otolith and may be analysed after either etching (Zhang and Moksness 1993) or grinding and polishing the otolith (Mosegaard and Madsen 1996). Analysis of otolith microstructure is a powerful tool for determination of life history trajectories, where reflection of hatch season and larval ambient environment is the key identification of individual population affiliation. Differences in otolith growth trajectories between herring larvae experiencing different temperature and feeding regimes have been identified in both field and laboratory studies (Moksness 1992, Fossum and Moksness 1993, Stenevik et al. 1996, Folkvord et al. 1997). Thus herring larvae hatched at different times of the year, experiencing different temperature and feeding regimes, will display different patterns of primary increments in their larval otolith. Herring larvae otolith microstructure was demonstrated to be a stock separator (Moksness and Fossum 1991) and differences in the larval otolith
CM 2005/K:05
microstructure have been identified in adult herring and successfully used to separate adult herring from different spawning stocks at the individual level (Mosegaard and Madsen 1996, ICES 2004). Assuming spawning time fidelity spawning individuals has historically been the basis for reference collections and validation of the applied splitting keys following the concept of pure stocks (Mosegaard and Madsen 1996). However, variation in microstructure among stocks and possibility of non-adherence of spawning season to hatch season make a revisiting of these assumptions appropriate. The present study was initiated to analyse the variability in otolith microstructure pattern in postlarval 0-ringer herring, hatched during different seasons, to achieve a validation method independent of the assumptions behind the pure stock concept. Herring 0-ringers were chosen according to the assumption that they would exhibit an unbroken series of daily increments from some period after hatch until capture. We develop an independent objective validation method combining back-tracking of date of birth with measurements of microstructure increment patterns and visual inspection of the larval otolith. The primary increments formed during the larval stage in herring have been shown to be daily in Norwegian spring spawners (Moksness 1992) consequently enabling a back-tracking of date of birth by counting the daily increments in an individual from the edge to the centre. However each population may require its own validation (Geffen 1982, Folkvord et al. 2000, Fox et al. 2004). The present study compares two methods for validation of the use of visual inspection of the larval otolith microstructure to separate herring from different spawning stocks at the individual level: 1) Accuracy of visual inspection as defined by hatch type assignment by visual inspection of otoliths from spawning individuals collected in the North Sea, English Channel and the Baltic area. 2) An objective method separating 0-group herring from the mixed stock in subdivision IIIa into hatch types based on measurements of the larval otolith microstructure and compares this with back-tracked hatch dates. For studies of population structuring the identification of individuals straying among populations with different spawning seasons is of focal interest but also presents a problem concerning validation of the otolith microstructure method. The otolith microstructure of straying individuals may not be detected as the apparent variability may be too high to detect the phenomenon. If on the other hand a specific hatch period gives rise to highly variable otolith microstructure some individuals may be falsely identified as coming from a different spawning season. We discuss the reasonable application of the visual inspection of otolith microstructure as stock separator in the light of the objective validation method beyond precision and natural variability of the otolith microstructure.
Materials and Methods Validation by visual inspection on spawning individuals assuming spawning time fidelity. To test accuracy of the visual inspection of hatch type, otolith microstructure analyses by two experienced readers on material from spawning populations sampled under the EU project HERGEN (QLRT – 2000 – 01370) were compared with the actual sampling season. A subset of all individuals in the ripe and running maturity stage (maturity stage 6) were selected from the sampled
CM 2005/K:05
material which gave a total of 697 individuals from spawning sites in the North Sea, English Channel and Western Baltic (figure 1). The otoliths were retrieved from archives and mounted with the sulcus side up in thermoplastic resin (Buehler Thermoplastic Cement no. 40-8100) at 150oC allowing for repeated relocation of the otolith for grinding and polishing on both sides. The ID of the individuals were coded in such a way that the readers were not able to detect which spawning population, the individuals originated from. The order of the individuals was random so the 3 possible spawning types were not appearing in groups. The otoliths were polished using a series of grinding and polishing films with decreasing grain sizes from 30 µm to 0.3 µm to optimise the visual resolution to a focal plane through the otolith nucleus and a transect from this to the edge. Hatching type was estimated on all individuals. Visual inspection was performed using a Leica™ DMLB compound light microscope with objective lenses of 20- and 40 times magnification and long distance between focus and lens to allow view of structures trough a microscope slide of 1.5 mm thickness. All selected otoliths were assigned to a hatch type of either spring, autumn or winter following the guidelines presented in table 1. The zone where incremental widths increased from less than 2 to more than 2.5 µm was used as a marker for the onset of spring increased growth conditions. Otoliths considered as unreadable by one or both readers were disregarded for that comparison. The accuracy of the visual inspection of hatch type was calculated as the correspondence between the assigned hatch type and the season within the spawning individual was sampled.
Validation by means of the objective separation of otolith microstructure pattern by image analysis. Since 1999 sagitta taken from herring caught in subdivision 20 and 21 and the transfer area in the North Sea (figure 1) has been polished and spawning type determined on a routine basis at the Danish Institute for Fisheries Research (DIFRES). A search in the DIFRES database on 0-group herring (fish caught before the onset of the first otolith annulus) from 1999 to 2003 was made and 300 individuals were selected from the years 2001 to 2003 and from different locations within the area (figure 1). The otoliths were retrieved from archives and remounted following the same procedure as described above for the otoliths from the spawning populations. The otoliths were subjected to visual inspection following the preparation process and were all classified as autumn, winter or spring hatch types as performed for the spawning individuals. Following the visual inspection images of a subset of 0-group herring sagittae (n=108) were digitized; each otolith was analysed taking several pictures following the longest axis along the postrostrum. Measurements of otolith microstructure were made with a Leica™ 350 F digital camera and ImagePro™ 5.0 image-analysis package for Windows™. The x-y-coordinates, distance from samplers origin and increment width were measured along a profile of grey values increasing from 0 (black) to 255 (white) with the Caliper tool in ImagePro using a profile width of 10 µm. The Caliper tool was set to identify the onset of an increment as the point at which the grey values changed at the fastest rate towards lower values. If the program failed to identify the start of and increment or produced increments where other structures gave the same increase in grey values (e.g. cracks in the otolith) these were manually altered to fit the real increments. In cases where the increments were too unclear to identify both by the Caliper and by eye, the distance from the last visible increment to the next visible increment was measured. A minimum acceptable increment
CM 2005/K:05
width was set at 0.5 µm to filter out the segments where false or no daily rings were visible along the measurement axis. All measurements were transferred to an MS Excel™ spreadsheet. Areas with no detectable ring structures were occasionally found in the trajectory from the otolith centre to the edge. Since these areas would appear as abnormally broad increments but only counting one day, a running median value mi= MED(wi-2,.., wi+2) was applied as a smother to yield a robust estimate of increment width at distance from centre. This median was then used to estimate duration in days di between observed increments wi, as di= wi/mi, independent of whether these were true daily increments or just zones with several non readable daily increments. A median over five successive increments was enough to screen out all non readable areas. This was indicated by the fact that no median increment exceeded 7 microns in the first 200 microns from the centre and no median at all was more than 14 microns wide. Further, only 5 successive pairs of medians out of 28300 had more than a 50% change in width between them. Measurements closer than 10 microns from the centre were disregarded (Folkvord et al. 2004) and the distribution of distances from the centre to the first increment was investigated. Since 94% of all otoliths had their first measured increment less than 25 microns from the centre, this distance was used as an initial reference to group otoliths according to date of formation. To estimate the initial increment widths from 10 microns to the first identified increment, the first seven increment widths were plotted against estimated Julian day, Ji, at the formation of the increment 25 microns from the centre. On a scale starting with day i = 200-365 = -165 and ending with day i = 200 a quite good quadratic relationship was found: mi= 1.67 + 0.0081* Ji + 0.000038* Ji2 (R2=0.73, p 300 microns from the centre are visible. Results Validation by visual inspection on spawning individuals assuming spawning time fidelity. The accuracy of the visual inspection of hatch type is presented in table 2. Assuming spawning time fidelity, the accuracy of the visual inspection was high with an overall correct classification of 91%. The individuals collected as ripe and running fish in winter (from November to December) were classified with the highest difficulty with a misclassification of 32% whereas the individuals collected during spring (from March to June) were hatch type determined with the lowest misclassification rate of 3%. The most apparent pattern in the misclassification was that spawning individuals from autumn and winter were the most frequently confused with each other whereas the spawning individuals from spring when misclassified were assigned to equally often to either of the two remaining hatch types (table 2).
Validation by means of the objective separation of otolith microstructure pattern by image analysis. The majority of the measured otoliths showed to have a whole unbroken transect of daily increments from the start of the measurements at 20 µm from the centre and to the edge of the otolith. The distribution of back-tracked hatch dates are shown in figure 3 where the smooth curves are the normal distribution of hatch dates calculated from the distributions found by the backtracked dates. The back-tracked hatch dates falls within 3 clearly separated groups, winter, spring and fall, however some overlap between the groups is evident especially between the autumn and winter hatch date groups. The periods for the spawning seasons were defined by the normal distributions to spring being from February 18th to July 9th, autumn being from July 9th to November 5th and winter being from November 5th to February 18th. The visual inspection of the mixed stock otoliths gave an overall correct classification of 89% when the classification by visual inspection was compared to the back-tracked hatch season of the individuals (table 3). The misclassification pattern repeated the pattern seen in the pure stock samples as the hatch types most frequently confused were the autumn and winter types. Autumn spawners were classified as winter spawners in 10% of the analysed individuals and winter spawners were classified as autumn spawners by visual inspection in 17% of the analysed individuals. The overlap between autumn and winter spawning seasons and the pattern of misclassification by visual inspection led to further examination of the division between these two spawning seasons. The seasons were forced to fixed periods so that winter was categorized to be from December 1st to
CM 2005/K:05
February 18th, spring was between February 18th and July 9th, autumn was set to be between July 9th and November 5th and late autumn to be within November 5th and December 1st. Using these categories the visual inspection results were re-analysed and it showed that 63% of the misclassified winter hatch types fell within the period of late autumn (table 4). The classification of spring hatch types did not as an effect of the new categorization of spawning seasons, whereas the correct classification of winter hatch types increased. Comparing the reading results in table 3 and 4 it is apparent that the misclassifications of autumn hatch individuals in the non-restricted division of spawning seasons were almost exclusively done of individuals hatched late in the autumn. Thus the visual inspection of hatch types may fail to classify individuals hatched in the periods of overlapping spawning seasons. Although three well separated hatch date groups were found, there was a significant within group difference in mean hatch date for both autumn (p=0.04) and winter (p=0.0018) groups among years. The development of increment width with distance from the otolith centre is shown for 4 hatch types (as determined by increment counts to winter, spring, autumn and late autumn) in figure 4. The spring hatched individuals clearly separates from the remaining hatch types by exhibiting increments wider than 2 µm from the beginning of measurements and the increments continue to increase in width over the whole measurement transect levelling out at approximately 6 µm at a distance of 400 µm from the samplers origin. The increment width development in autumn, late autumn and winter hatched individuals did overlap especially in the first part of the measurement transect. At a distance of 150 µm from the samplers origin the three types can be separated, the late autumn hatch type being the gradual change from autumn to winter type. To aid in visual inspection of these more difficult separated hatch types a series of sections along a transect from the otolith centre towards the edge were selected and the separation ability of the increment widths in these sections were tested in a multiple regression analysis. When mz from segments 1-7 of the otolith i.e. the area from 15-225 microns from the centre were analysed, the linear combination: Sk = 7.8 + 1.3×ln(m1) + 1.6×ln(m5) + 4.8×ln(m6) - 0.74 ×m6 (R-square=0.88) exhibited the best fit (figure 5). However, a large number of other section combinations also gave high prediction of hatch season, with section 5 (135-165microns from the centre) often showing up as the major influence. When segments >10 were analysed the best combination was: Sk = 1.6 + 2.6×ln(m11) + 3.0×ln(m14) + 0.45 ×m20 (R-square=0.76) with section 11 (315-345 microns from the centre) generally having the most influence in different combinations. This result shows that also increments formed after the larval period may be used in discriminating between hatch types, and aid as an additional tool to visual inspection when over grinding of an otolith has made the routine method problematic.
Discussion An ecosystem approach to fisheries management should consider intra-specific bio-complexity expressed as variability in population structure and correspondingly evolved life history variation. Knowledge of the stock integrity is of unequivocal importance for a sustainable fisheries management since variable compositions in mixed areas together with asynchronous population dynamics may lead to over fishing of individual stocks if not all components are considered. Atlantic herring displays a wide range of migration behaviour, homing and possible deviations from this (Husebø 2005, McQuinn 1997a,b, Slotte 1998). To ensure conservation of herring population diversity in the North Atlantic and their natural migration patterns all stock components must be
CM 2005/K:05
considered in the advice on the fishery. Thus the highest level of precision in the input data to the assessment of mixed stocks is needed and the methods of stock discrimination are of great importance. For routine separation of mixed herring catches based on individual hatch season a visual inspection method of otolith microstructure pattern has been developed and tested on so called pure stocks where the pure stock concept assumes that all herring spawn in the same season when they were hatched themselves (Mosegaard and Madsen 1996, ICES 2004). The high agreement between the assigned hatch type by visual inspection and the sampling season of spawning individuals observed in the present study confirms visual inspection of the larval otolith microstructure in spawning individuals as a valid discrimination method between hatch types. However, despite the high degree of correspondence between assigned hatch type and spawning season some reader variation is found. This variation may be categorised into i) within reader variation, ii) among reader variation and iii) consistent disagreement between reader assigned hatch type and spawning season. The reasons for misclassifications under iii) may be found in the natural variability of the larval otolith microstructure formed after hatch, potentially overlapping spawning seasons and possible straying of individuals not exhibiting spawning time fidelity (McQuinn 1997b, Slotte 1998, 2001). The objective classification of hatch types in 0 group herring developed by the present study provides a possibility to calibrate the visual inspection of hatch types in all herring allowing the method to include possible variation in environmental influence on the otolith microstructure allowing for variability of the pattern within each hatch type. The underlying assumption for this approach is that primary increments are sufficiently close to daily so that back-calculation of hatch season in 0-group herring is possible. A further development of the method assumes that increment measurements along radii at specific distance from the otolith core reflect seasonal and area specific environmental conditions during herring larval growth, thus enabling the imprinted otolith microstructure patterns to identify offspring from different spawning populations. Though it was not intended to calculate the absolute hatch date but to back-track the hatch season with a reasonable accuracy the presented validation technique has two important prerequisites; knowledge of the timing of the formation of the first daily increment and the successive daily deposition of micro increments in the larval otolith. The formation of the first discernable daily increment in herring larvae coincides with the onset of first feeding in the beginning of the post yolk-sac growth (Høie et al. 1999, Moksness 1992) which occurs in herring around 10 to 19 days from hatch depending on the herring population in question (Fox et al. 2004). However growth rate and temperature has been shown to be of high influence on the formation of the first discernable increment (Høie et al. 1997, Folkvord et al. 2000, Fox et al. 2003). Folkvord et al. 2004 showed no increase in size of sagitta from herring larvae reared at 4ºC up to 30 days whereas herring reared at 12ºC showed sagitta growth after 9 days. We calculated initial undetectable increment widths by a general curve-linear relationship from the best observed individuals, however there is no way of detecting the occurrence of no otolith growth under unfortunate environmental conditions; therefore we felt that adding a variable amount of days to the counts of daily increments, depending on some uncertain environmental forcing, would make the calculation of the absolute age more uncertain then necessary for the present purpose. In the present study it is likely that the ages of the winter hatch type individuals are under-estimated as these individuals will have experienced the lowest post-hatch temperatures of the three hatch types, But also individuals hatched during autumn could suffer from some underestimation depending on specific hatch time and annual variation in temperature.
CM 2005/K:05
The formation of daily increments in embryonic stages of herring has not been confirmed (McGurk 1984, Moksness et al. 1987), however for stages after the absorption of yolk sac otolith microincrements have been observed to be formed on a daily basis (McGurk 1987, Moksness and Wepestad 1989, Moksness and Fossum 1991, Moksness 1992). Growth rate, however, seem to influence the deposition of daily increments. Several studies demonstrate non-daily increment deposition in herring larvae with a growth rate
