Joint Norwegian-Russian monitoring of 0-group fish on autumn surveys in the Barents Sea, 1965-2023

A joint Norwegian-Russian survey of 0-group fish (here defined as fish hatched earlier in the same calendar year) in the Barents Sea was started in September 1965 with the motivation to provide initial information on year class strength of commercially important fish stocks (ICES 1965, Eriksen and Prozorkevich 2011). The survey initially used echosounders to record 0-group fish combined with trawl sampling to identify the composition of the acoustic backscatter (Dragesund and Olsen 1965). The joint 0-group survey was continued the following years with participation also by the United Kingdom from 1966 to 1976. Intercalibration of the echosounders was done before the start of the survey to improve comparability of results obtained by different research vessels (Dragesund 1970, Dragesund et al. 2008).

The acoustic information was used in a semiquantitative manner by classifying the echo-sounder paper recordings into 5 categories from no (0) to very dense (4) recordings (Dragesund et al. 2008). The number of fish caught in supporting trawl catches was additionally used to distinguish between scattered and dense concentrations on distribution maps (Haug and Nakken 1977). While trawling in the first period was guided by the echo-sounder results, ICES advised in 1980 on a standardized trawling procedure (stepwise in the upper 60 m; see later section) which has been followed from 1981 onwards. At the same time, the 0-group survey shifted from a combined acoustic-trawl survey to a standardized trawl survey (Dragesund et al. 2008, Eriksen and Prozorkevich 2011). From 1981 onwards, all vessles have used the same type of trawl, a fine-meshed commercial trawl (‘Harstad’) designed to catch capelin (Nakken and Raknes 1996, Dragesund et al. 2008). This trawl has a rectangular opening of about 20 by 20 m.

The results from the survey have been calculated and expressed as a set of 0-group fish abundance indices of the main commercial species of fish found in the Barents Sea (Dingsør 2005, Eriksen et al. 2009, Eriksen and Prozorkevich 2011). The abundance values have also been converted to 0-group fish biomass by multiplying numbers with mean weight of the 0-group fish that are recorded routinely during the surveys (Eriksen et al. 2011, 2017b).

0-group fish play dual roles in the ecosystem. They are the recruiting life stages of fish stocks that are of great ecological and economic importance, and variation in recruitment, as reflected at the 0-group stage, plays large roles for the dynamics of the fish stocks as well as the wider ecosystem through trophic interactions. In addition, the 0-group fish are planktonivorous and constitute a substantial component among the pelagic fish in the ecosystem. This is the case not only for true pelagic species, such as capelin and herring, but also for demersal species, such as cod and haddock, before they settle to near the seafloor later in autumn.

Time series of 0-group abundance and biomass have been used in descriptions and analyses of the Barents Sea ecosystem (e.g. Eriksen et al. 2017, ICES WGIBAR 2018). We are currently expanding these analyses to address in more detail the roles of 0-group fish in relation to recruitment variability and stock dynamics of major fish species, and for the structure and energy flow in food webs of the Barents Sea ecosystem. In the project ‘Trophic Interactions in the Barents Sea: steps towards Integrated Ecosystem Assessment’ (TIBIA) and ICES working group “Integrated ecosystem assessment in the Barents Sea” (WGIBAR), we were using a subdivision of the Barents Sea into 15 subregions (polygons) (Fig. 1). We are using this subdivision (but with 13 polygons only due to lack of coverage in two northeastern polygons) to provide spatially resolved estimates of biomass of major ecosystem components, such as zooplankton, benthos, and fish, including 0-group.

In this communication, we provide an updated overview of the joint Norwegian-Russian 0-group investigations in the Barents Sea. We describe the procedures of sampling, analyses, and calculation of results, and discuss associated sources of error. One particular source of error with trawl sampling of small fish is the catchability: to what extent do the 0-group fish escape through the meshes of the trawl as function of fish length, what are the roles of herding, and how is low and variable catchability corrected for (Eriksen et al. 2009). We have used the TIBIA/WGIBAR subdivision to provide spatially resolved estimates of 0-group abundance of major species of fish collected in the 0-group survey. The new abundance estimates by TIBIA polygons are compared with the previous set of abundance indices as reported by Eriksen et al. (2009) and Eriksen and Prozorkevitch (2011). Eriksen and Prozorkevitch (2011) provided distribution maps of four species of fish (capelin, herring, cod and haddock) for each year from 1980 to 2008. Here we provide a new and updated set of distribution maps from 1980 to 2023 for the same 4 species as well as for polar cod and redfish ( Sebastes spp.) (included here in part 7. Spatial distribution). We consider the spatial and temporal coverage of the surveys and note years where incomplete coverage or timing could have influenced the results (Part 6. Survey area and coverage).

2.1 - From acoustics to trawl-based survey

The international 0-group survey in the Barents Sea shifted from an acoustic survey, where trawling was used to identify the species of 0-group fish in the acoustic layers, to a standardized trawl survey where acoustic records are used mainly to guide sampling (e.g. add extra steps in the vertical if acoustic records suggest that 0-group fish are distributed below 60 m depth) (Dragesund et al. 2008). A study performed in autumn 1963 on abundance and distribution of 0-group fish from acoustic records in the Barents Sea, suggested that it would be feasible to carry out an 0-group survey in autumn based on acoustic methodology (Dragesund and Olsen 1965). At this time, it was known that 0-group fish were abundant in the surface layers of the Barents Sea, stemming from spawning at ‘up-stream’ spawning grounds further south. An echo integrator had also been constructed, which facilitated the treatment of the acoustic recordings (Dragesund et al. 2008). Based on the initial investigation in 1963, and follow-up studies in 1964, it was decided to start a joint international 0-group survey in autumn 1965. The results and experiences from the first four years of the survey (1965-1968) were reported as an ICES publication in 1970 (Dragesund 1970).

The feasibility of an acoustic survey of 0-group fish in the Barents Sea was at the time considered positively, being an early and inspirational case of the general development of fisheries acoustics, where abundance of fish is estimated from acoustic records combined with trawl catches to help identify the acoustic scatterers and allocate the acoustic signals among them (Dragesund et al. 2008). However, it became apparent that use of the acoustic method for 0-group fish was a challenge due to the commonly mixed occurrences of the different species as well as abundant presence of other scatterers such as krill and jellyfishes, as well as 1-group capelin. This led to a shift in emphasis from acoustics to trawling as the basis for the survey.

2.2 - Standardized trawling procedure

The “Harstad” trawl is designed to capture small fish and has been the standard equipment since around 1980 for the 0-group fish survey, the capelin survey, and later the ecosystem survey (Anon. 1980, Eriksen and Gjøsæter 2013). In the first years of the survey, pelagic trawl hauls were taken frequently, usually no more than 40 nautical miles (nm) apart, targeting acoustic scattering layers to help identify and quantify the contribution by 0-group fish (Dragesund et al. 2008). In addition, some trawl hauls in the surface layer were also taken in areas where there were no clear acoustic records of 0-group fish. Based on advice from ICES, a new trawling procedure was introduced in 1980. This has since been the standard trawling procedure where the trawl is operated in steps with the headline at 0 m, 20 m and 40 m. With a nominal trawl opening of 20x20 m, this provides an integrated sample from the upper 60 m of the water column. The trawling procedure prescribes a towing speed of 3 knots and a tow distance of 0.5 nm for each depth interval (Fig. 2). Additional tows with the headline at 60 and 80 m, and with distance of 0.5 nm, were made when dense concentration was recorded deeper than 60 m on the echo-sounder.

Standardization has been an important aspect of the joint 0-group survey in the Barents Sea since its beginning in 1965. The same echo sounders were used on Norwegian and Soviet/Russian vessels in the early years, and inter-ship acoustic calibrations were carried out by comparing results from the same areas (Dragesund et al. 2008). The survey has been a large-scale, multi-ship operation with 3-6 vessels taking part annually. The vessels used in the first years were built as side-trawlers, being gradually replaced between 1970 and 1979 with larger stern-trawlers, better equipped and capable of operating larger trawls (Dragesund et al. 2008). From 1980 all participating vessels have used the same small-meshed sampling ‘standard’ trawl – the ‘Harstad’ trawl. This trawl is constructed with seven panels, with mesh size (un-stretched) decreasing from 100 mm in the first (front) panel to 30 mm in the last panel and 8 mm in a codend (Godø et al. 1993). While the trawl is considered standard and has been used on both Norwegian and Russian vessels, there have been adaptations and differences in rigging due to the Norwegian vessels initiated towing at the surface and the Russian vessels initiated towing at depth.

2.3 - Sample processing and analyses

When the trawl comes on deck, the trawl is shaken well, to allow for fish adhering to the trawl meshes to fall back into the trawl cod end or to the deck. This is to ensure that the calculated biomass and numbers of individuals are as accurate as possible, and to avoid fish from earlier hauls contaminating later samples. The problem of fish being trapped between trawl meshes is greatest at stations with a lot of 0-group capelin. The part of the catch that falls to the deck, usually in poor condition, is collected and processed separately. The sample from the deck is identified to species and weighed per species. The weight of the deck sample is added to the rest of the sample on a species basis to give the total sample weight for each species.

Catch processing in the fish laboratory starts with all jellyfish and larger fish (such as lumpfish) being sorted out to make the rest easier to handle. Jellyfish and larger fish are weighed separately. Sometimes it is necessary to remove excess of water so that the sample weight is affected as little as possible by the water. In the case of large catches, a sub-sample is taken. When sub-sampling, a conversion factor is used to calculate the total weight of all groups in the catch. A factor is calculated as the total weight divided by the sub-sample weight. The samples from the trawl are processed immediately after the catch is removed from the trawl. 0-group fish of different species, as well as other components of the catch (e.g. krill and pelagically distributed small non-commercial fishes), are sorted into groups that are weighed separately. The total weight of the catch is determined as the sum weight of the components. The extra variance introduced by subsampling has not been studied formally but is believed to be low compared to the high variance associated with the trawl samples of 0-group fish.

The 0-group fish are determined to the species level, while some of the small non-commercial species (families Agonidae, Stichaeidae, Cottidae and Myctophidae) could be determined to genus or family level (due to taxonomic difficulties, available expertise, and time constraint). Before 2014, 100 individuals of each species/group (if available) were weighed and separately length measured (to nearest mm on Norwegian and 0.5 mm on Russian vessels). The length sample weight and total catch weight are used to calculate the total number of fish caught. From 2014, the number of fish that were length measured was reduced to 30 individuals (based on statistical considerations described in Pennington and Helle, 2013).

Various ways of calculating abundance indices have been used during the history of the survey. In the early years of the survey, from 1965, the echo abundance was subjectively evaluated from the paper recordings (echograms) on a scale from 0 to 4 (0 - no recording, 1 - very scattered, 2 - scattered, 3 - dense, 4 - very dense). This information was then used during the first 6 years (1965-1970) to classify year-class strength as poor, average, or strong by expert judgement (Dragesund et al. 2008).

3.1 - Area index

The acoustic information was subsequently used to construct a quantitative (or semi-quantitative) abundance index, the so-called area index (Haug and Nakken 1977). Maps of distribution of various 0-group species had been prepared for the annual reports based on the 0 - 4 scale visual grading of paper echograms, guided by results on the 0-group fish counts in the supporting trawl hauls. Classification of the acoustic records was done for every nautical mile sailed along survey lines, with three density grades used to plot the results onto maps: absent, scattered, and dense (Dragesund et al. 2008). Haug and Nakken (1977) established empirical relationships between trawl catches and the 4 density grades (very scattered, scattered, dense, very dense). They noted some inconsistencies in the grading between vessels and years, and established criteria in terms of number of fishes per haul to help standardize the distinction between scattered and dense records of 0-group fish of four species (cod, capelin, redfish, and polar cod).

Haug and Nakken (1977) used the criteria to draw new distribution maps for the four species of 0-group fish for the years 1965-1972. The area index was calculated as the sum of the integrated area on the map with low abundance (scattered), plus the area with high abundance (dense) multiplied by factor 10. This factor was an approximation based on the empirical data (Haug and Nakken 1977). The area index was calculated for six species (cod, capelin, haddock, redfish, polar cod, and long rough dab) for the years 1965-1972. Average index values were used to reclassify year-class strength in each year in this (relatively short) period as average, poor, or strong (Haug and Nakken 1977).

The area index was calculated in subsequent years as one of two methods (the other was the logarithmic index; see below) used to produce time series from the 0-group survey (Dragesund et al. 2008). It became apparent that the area index had shown an increasing trend from 1965 until the early 1990s. Nakken and Raknes (1996) provided a correction to the area index time series by assuming that capture efficiency had increased proportional to the size of the trawls (trawl opening (“mouth”) area) used in the survey. They used the arithmetic mean trawl opening for the survey participating vessels (and trawls) each year, which they considered a rough approximation since differences in geography and catches among the vessels were not taken into account (which would have required much work). The correction represented more than a doubling of the area index values between 1970 and 1984 (Nakken and Raknes 1996). Nakken and Raknes (1996) also attempted an alternative method for correction, using the trend in the sum of index values for cod, haddock and redfish as an expression for the trend in overall capture efficiency. However, this depended strongly on an increasing trend for redfish, and it was uncertain how much of this increase was due to increased capture efficiency.

The corrected area index time series was updated annually and reported in the annual report from the 0-group survey to ICES. Nakken and Raknes (1996) provided corrections for cod, haddock, and redfish. Subsequently, similar corrections were made for Greenland halibut, long rough dab, and polar cod. The area index for herring was calculated by Toresen (1985) for the period 1965-1984. Dragesund et al. (2008) provided a graphical representation (in their Fig. 6.6, page 127) of the area indices for the 1965-2000 period for 7 species of 0-group fish (cod, haddock, herring, redfish, capelin, and polar cod,) (based on ICES 2003).

3.2 - Logarithmic index

The logarithmic index was developed by Randa (1984). The catch in numbers of 0-group fish at each station was log-transformed (natural logarithm, ln), and mean densities (catch rates per nautical mile) were calculated for 17 strata (geographical areas) of the 0-group survey area in the Barents Sea. The overall abundance index for a species was then calculated as the area-weighted mean logarithmic abundance, adjusted for the proportion of hauls with no catch. The method is based on the log-normal theory, and it allows confidence intervals to be calculated based on normal theory (Randa 1984). Randa (1982) showed that log-transformation normalized the catch data for 0-group cod (for the 1965-1979 period).

Randa (1984) took into account the different trawls used in the early years of the joint 0-group survey by estimating ‘relative fishing power’ (relative to R/V “G.O. Sars”, 1971-1979) for each of the participating vessels.

The logarithmic index was calculated by Randa (1984) for cod and haddock, and by Toresen (1985) for herring. These indices were updated and included in the annual reports to ICES from the joint 0-group survey.

The logarithmic index was further developed as one of two alternative indices by Dingsør (2005; the other was an arithmetic index based on stratified sample mean; see below), which he called the ‘Pennington estimator’ (Pennington 1996). While the 0-group data largely follow a log-normal distribution, they usually have many low values close to zero which may bias log-normal-based estimators. A cut level for low values (set at 20 % of the average abundance density in each stratum) was used to reduce the bias from low values and achieve better fit to log-normal distribution for the remaining values above the cut level (Folmer and Pennington 2000). Dingsør (2005) calculated time series of the logarithmic ‘Pennington estimator’ (with standard errors) for cod, haddock, capelin, redfish, and herring for the years 1980-2002. The index was calculated both with and without correction for capture efficiency (see section ‘Capture efficiency’) for cod and haddock. Dingsør (2005) compared the ‘Pennington estimator’ index with the old area index and the previous logarithmic index. He found similar trends but also some discrepancies, notably for some of the species in the 1980s (see Figs 4 and 5 in Dingsør 2005).

Dingsør (2005) recommended using the ‘Pennington estimator’ as the most appropriate method and new standard for presenting 0-group abundance indices in the Barents Sea. However, the arithmetic abundance index based on the ‘stratified sample mean’ method turned out to be the preferred index for routine use. With the start of the joint ecosystem survey (where the 0-group survey became an integral part) in 2004, the arithmetic (total abundance) index was used, and the logarithmic index was no longer calculated after 2004.

3.3 - Total abundance indices

At the transition to the joint ecosystem survey in 2004, a new abundance index was developed by Gjert E. Dingsør and Dmitry Prozorkevich and used for the 0-group results from the survey in 2004 (Anon. 2005, Dingsør 2005). The index was based on a stratified sample mean estimator, reflecting the mean areal density of 0-group fish in the survey area. The density of fish in length groups (number of fish per nm² ) was calculated for each trawl station, and mean density was calculated for each of 23 strata of the total survey area of the Barents Sea (Fig. 3; Dingsør (2005) used a division into only 4 larger strata). The stratified sample mean estimator of abundance was then calculated as the overall mean density of 0-group fish, by weighting the strata means by the proportion of the survey area in each stratum. The area covered with survey stations within each stratum was determined using GIS software.

The 23 0-group strata were combined into larger areas (north-western, northern, western, central, eastern and coastal; Fig. 3) used in Eriksen at al. 2009, 2011, 2012, and 2014. Later, in the project TIBIA, the Barents Sea was divided into 15 subareas (polygons, see Fig. 1). The division is based on topography and oceanography and is a modification (with some subdivision) of the system used by Eriksen et al. (2017) in a summary analysis of distribution of pelagic biomasses in the Barents Sea. At the ICES WGIBAR meeting in 2018 (ICES WGIBAR 2018), the division of the Barents Sea into 15 polygons was presented and adopted for use in reporting status and changes in the ecosystem.

The stratified sample mean estimator was expressed as a total abundance index by using the total area covered in the survey (sum of polygon mean density of fish, per nm², multiplied by polygons coverage in nm² ). The total abundance index was calculated both without and with length correction for low capture efficiency for small fish (see section ‘Capture efficiency’). These two sets of indices (corrected and non-corrected) were calculated back to 1980 for capelin, cod, haddock, herring, saithe, and polar cod, as were uncorrected values for redfish, Greenland halibut, and long rough dab (Tables 2.2 and 2.3 in Anon. 2005). The new total abundance index is calculated with variance and confidence intervals based on the variation in 0-group abundance among sampling stations. At the time it was agreed that the new total abundance index without correction would be the ‘official’ one, while the corrected index was ‘additional’.

Dingsør (2005) showed that the stratified sample mean estimator corresponded closely to the log-normal based ‘Pennington estimator’, with both showing similar temporal patterns from 1980 to 2002 (for cod, haddock, capelin, herring and redfish; see his Table 2).

The total abundance index was used for 0-group data for the next years of the ecosystem survey with some adjustments of the time series (in 2005 and 2007). The former logarithmic index was discontinued in 2005, while the old area index former reported in parallel to the new set of indices (total abundance, corrected and uncorrected) until (and including) 2007 when it also was discontinued.

An ‘overhaul’ of the total abundance index was done in 2009. It had become apparent that there were many mistakes and errors in the data (e.g., punching errors when data were transferred from paper sampling sheets to the computer), and inconsistencies between the data held in data bases of the two institutions conducting the surveys (IMR and PINRO). A major effort was therefore made over a three-year period to check the quality of the data, using cruise logbooks and original data records dating back to 1980.

New sets of total abundance indices based on the quality assured data were calculated and reported by Eriksen et al. (2009). This work included indices corrected for capture efficiency and uncorrected indices for cod, haddock, capelin, herring, saithe, and polar cod, and uncorrected indices for redfish, Greenland halibut, and long rough dab, from 1980 onwards (see Table 2 in Eriksen et al. 2009). The corrections were from slight to substantial in some cases (species and years). However, the broad temporal patterns and trends in 0-group year-class strength did not change much, reflecting that the amplitude of changes in abundance was generally much larger than the corrections (Eriksen et al. 2009). The estimation was carried out in SAS software and the indices of fish abundance for the 0-group are presented in part 9.1.

Eriksen et al. (2009) showed that the revised set of total abundance indices were positively correlated with the old area index for cod, haddock, capelin, and herring (r = 0.80-0.89). The abundance indices were also positively correlated with estimated abundances of the year classes as 1-group for capelin (r = 0.81-0.82), and 3-year old for haddock (r = 0.43-0.49).

Abundance and biomass estimates were calculated by different software during the last four decades: SAS (for the new 23 fisheries subareas, 1980-2017, 0-group strata and WGIBAR polygons ) and MatLab (for the new 15 WGIBAR- polygons ( for the period between 1980 and 2018, ICES WGIBAR 2018) and R (for the new 15 WGIBAR-subareas (2003-202 3 ). Due to software upgrading (which led to challenges with script running in SAS) and personal resource limitation (MatLab), it was decided to develop R-scripts (R core Team, 2023) for estimation of abundance and biomass indices. Two data sets (abundance and biomass indices calculated by R and SAS) were analyzed for similarities and were found to be highly significantly correlated (for capelin r=0.95, cod r=0.99, haddock r=0.94, herring r=0.98 and polar cod r=0.94).

During development of R scripts for abundance and biomass estimation, some errors in the IMR database were detected, that most likely occurred when all historical data were converted from an old to the new "Biotic" format. Apparently, some algorithms failed, which created duplicate rows of existing fish observations and recalculated total weight or abundance. A new quality check was carried out on the data in the new data format, which was corrected back to 2004. The older data (1980-2003) in the IMR database. have not been checked and corrected, and it is uncertain how many errors there are in this part. We note that the data compiled and used in this report were extracted from the database at an earlier stage and are not affected by these errors.

The last “official” updated time series of the abundance and biomass of the 0-group fish are reported in the BESS report 2023 (available at https://www.hi.no/hi/nettrapporter/imr-pinro-en-2024-2) and in Part 9.4 of this report.

Small juvenile fish, especially herring, pass through the meshes of the first panels of the Harstad trawl. This gives a low capture efficiency of the trawl when catch is referenced to the mouth opening of the trawl (Godø et al. 1993). The effect is inevitable due to the low maximum swimming speed of small 0-group fish relative to the mesh size and speed of the trawl. This was clearly demonstrated in experiments in the early 1990s, comparing catches of 0-group fish in the standard trawl with catches obtained with a specially designed 0-group trawl with finer meshes (Godø et al. 1993, Hylen et al. 1995).

The experimental trawl was smaller with mouth opening of 30 m² (compared to 300 m² for the standard trawl for a specific configuration of 20 m x 15 m), and mesh size decreased from 200 mm in the front panel to 10 mm in the cod end (Godø et al. 1993, Valdemarsen and Misund 1995). Experiments comparing the standard trawl and the experimental trawl were done in the Barents Sea in August 1991 (Godø et al. 1993), and during the 0-group survey in August/September 1992 and 1993 (Hylen et al. 1995). Both studies gave consistent results, with sampling efficiency (comparing density of 0-group fish in numbers per nm² ) around 3-4 times higher for the experimental trawl compared to the standard trawl for 0-group cod and haddock. Furthermore, there was a clear size selection, where juveniles smaller than 5 cm were captured to very low extent with the standard trawl (Godø et al. 1993). The capture efficiency was strongly size-dependent, increasing from around 10 % for 5 cm long juveniles to nearly 100 % for 10-cm long fish for the standard trawl relative to the experimental trawl (Hylen et al. 1995). For even larger juveniles (>10 cm), there were evidence that they were more effectively captured with the standard trawl, suggesting that they were either herded into the larger trawl or having some avoidance of the smaller experimental trawl (Godø et al. 1993, Hylen et al. 1995).

In addition to a size effect, Hylen et al. (1995) found indication of a considerable effect of density of 0-group fish on capture efficiency. Using acoustic recordings as reference, they found a clear and significant positive effect of fish density (as reflected in trawl catches) on capture efficiency (trawl catch relative to acoustically recorded density). Hylen et al. (1995) explained this relationship by density-dependent herding, with increasing degree of herding (either in front of or inside the trawl) with increasing density of fish.

Mamylov (1999) developed a theoretic model of capture efficiency by trawl. The model assumed that the lowest capture efficiency for small fish (4.5 cm and 12.5cm) was equal to the ratio between the cross-sectional area of the cod-end and the mouth opening of the trawl, which he set at 0.1, corresponding to a maximum correction factor of 10. He assumed the capture efficiency of large 0-group fish was equal to 1, i.e. all fish that passed the mouth opening were collected in the cod-end. The equation is Keff = 31.177*exp (-0.2708*L) and illustrated graphically in Figure 4.

Later, PINRO carried out several investigations, and of 1205 analysed trawl catches, 131 trawl catches were selected in which mainly one species was present (Mamylov 2004, Prozorkevich 2010, 2012). The trawl catches in terms of numbers and size of 0-group fish were converted (using target strength relationship) and expressed in units of acoustic backscattering. The acoustic data were scrutinized, and selected portions of the data were regressed against the trawl data expressed in the same units. The equations give very high factors for fish smaller than 4 cm (because of linear extrapolation), and therefore the maximum Keff (gadoids=8, herring =30 and capelin =4) was used for these small fish. The results from these experiments were close to the theoretical model, but they varied between species.

The correction curve for herring is very different, being much steeper than the lines for gadoids and capelin shown in Figure 4. For juvenile herring <6 cm long, the correction factor is higher than 10 (30 at 4 cm length), which is a theoretical maximum. For juvenile herring >10 cm, the correction factor is <1, corresponding to capture efficiency >1 (>100 %). This would imply active herding by doors and bridles in front of the trawl. While this cannot be ruled out, the very low capture efficiency in the low end, and the high capture efficiency in the high end, suggest that the steepness of the herring curve may be an artefact.

Hylen et al. (1995) provided similar empirical relationships for capture efficiency and correction factors for cod and haddock, using the experimental trawl as a reference for the catches obtained with the standard trawl. The relations from Hylen et al. (1995) have been plotted in Figure 5 using equations (2 and 3) from Dingsør (2005). The lines for cod and haddock are curvilinear on this log-scale plot because the equation is of a different form (declines exponentially to 1 rather than to zero). Apart from this, the line for cod from Hylen et al. (1995) is very similar to the Mamylov line. The haddock line is also close to the Mamylov line for fish in the size range from 8.5 to 13 cm. The haddock line swings upwards at low fish length, to values over 10 for fish <7 cm; again, this is possibly an artefact due to large variation in the underlying data (see Hylen et al. 1995, their tables 3 and 6).

The correction factors for gadoids, capelin and herring in Fig. 4 were used to correct the total abundance indices from the 0-group survey by Dingsør (2005) and Eriksen et al. (2009). Corrections were done for cod, haddock, saithe, and polar cod using the equation for gadoids, and for capelin and herring with their respective equations. The time series of abundance of 0-group fish of redfish, Greenland halibut, and long rough dab were not corrected, and uncorrected indices were used by Eriksen et al. (2009). The corrected abundance time series were used by Eriksen et al. (2011, 2017) where abundance was converted to biomass of 0-group fish.

In 2013-2016, several experiments were performed to study escapement of 0-group fish through the trawl panels and clogging of 0-group fish (BESS reports for 2013-2016, available at https://www.hi.no/hi/nettrapporter?y=2024&query=&serie=imr-pinro&fast_serie= ) with the aim to develop a new 0-group fish trawl. The trawl is designed to obtain constant trawl geometry independent of warp length and to obtain reduced clogging and escapement compared to the standard Harstad trawl. Unfortunately, the newly developed 0-group trawl with fine inner nets and constant opening was too heavy to be towed by the old Russian vessel. It was therefore decided that, for the time being, the Harstad trawl would be used as the standard trawl on all vessels participating in the BESS.

The timing and general design of the 0-group fish survey is to allow sampling of the 0-group part of populations of the different species while they still are in the upper pelagic zone. The early studies that used acoustic recordings, showed that the 0-group fishes were generally distributed in the upper 60 m water layer in early autumn, where they are feeding on zooplankton. This observation was the basis for the standard trawling procedure with three steps covering the 0-60 m depth interval (Fig. 2). The procedure is also to include one or two additional deeper steps (to 80 or 100 m) if the acoustic records show deeper distribution of 0-group fish.

There is little information in the literature about when cod change from pelagic life-stage to a demersal life-stage in the Barents Sea. Several studies from other areas have shown that there is no clear relationship between fish age (in days) and fish length (in mm), and that fish of similar length settle at different times (Hussy et al. 2003; Anon. 2009). Boitsov et al. (1996) found that the transition (settlement) is a rather long process that occurs in September-October in the Spitsbergen area and in October-November in the southern Barents Sea. The settlement of cod and their food items occurs gradually and it is likely to be connected with a convection mixing of water layers and deepening of the thermocline layer (Ozhigin et al. 1999). It is assumed that haddock follows a similar pattern to cod, with the transition occurring gradually during the autumn (Dingsør 2005; Anon. 2006, 2009).

When the 0-group survey became a part of the ecosystem survey (in 2004), bottom trawl samples were also taken. Some 0-group cod were collected by the bottom trawl indicating most likely cod settlement, although ‘contamination’ by 0-group cod from the water column when the bottom trawl was retrieved may also have contributed to the catch. The data indicated varied settlement pattern between years and areas. Prozorkevich and Eriksen (2013) examined 0-group cod distribution based on pelagic and bottom trawl for the years 2005-2012 (Figure 6). They found that numbers of cod taken by demersal trawl were generally low, varying between 0.2 and 1.1%, suggesting that the settled part of the 0-group of cod population is too small to influence 0-group abundance indices markedly. The study suggested that there was no strong relationship between fish settlement and year class strength. However, during some of the most recent years, 0-group fish, notably cod, haddock and capelin, were found to be abundant in the 100-150 m depth layer possibly reflecting early descent from the upper pelagic layer.

0-group fish of the different commercial species, taken together, occupy much of the area of the Barents Sea. Capelin and cod are most widely distributed, haddock and redfish are distributed mainly in the western and central areas, herring in the southern, central and western areas, while polar cod is distributed in the eastern and northern Barents Sea (see maps in Part 7).

The survey area has included the western, southern, and central Barents Sea during the whole survey period. The survey has been operated with 4-6 research vessels each year (Table 1). The vessels have covered different parts of the surveyed area, and cruise lines with sampling stations have been planned so that sampling effort is spread out more or less evenly over the survey area. One reason for this is the aim to monitor distribution and abundance of 0-group fish of several species that have different distribution patterns. The 0-group investigations have also been integrated with other survey elements, into what was called multi-species surveys from the late 1980s, and ecosystem survey from 2004 (Eriksen et al. 2018). Due to the many different purposes of the cruises, a stratified sampling design with higher effort in core areas of 0-group distribution and lower effort elsewhere, has not been used. The distance between trawl stations was about 30 miles until 1994 and 35 miles thereafter (Eriksen et al. 2018).

The survey has generally been run from south to north in the Barents Sea; that is, the research vessels have started in south and worked their way northwards. This is a broad pattern and there are many exceptions in specific years. Maps with cruise lines and station positions for the different research vessels are included in annual cruise reports that are available electronically (Table 2). The cruise lines are generally placed either in S-N or W-E directions, although zig-zag or more irregular patterns have also sometimes been used to obtain a good coverage of the survey area within the limits of time and ship availability.

A change in survey lines was made in the mid-1990s. From 1980 and up to 1994 (and also in 1997), the S-N survey lines followed longitudes and the E-W lines followed latitudes. From 1995 onwards (except 1997), the survey lines were placed equidistant (35 nm apart). The grid was oriented true North along the 30^o E longitude, while it deviated in NW direction in the western Barents Sea, and in NE direction in the eastern Barents Sea. A consequence of this was an opening of the sampling grid in the northern end, compared to when the lines followed longitudes.

The surveyed area has expanded northward in concert with reduction of sea ice in the Barents Sea. This can be seen from the maps with station locations in Part 7. A summary of the northern boundary of the survey area in three sectors is illustrated in Fig. 7.

In the Svalbard (Spitsbergen Archipelago) sector, the survey area has extended up along the west side of Spitsbergen to around 80-81^oN. Up to 2004, the survey extended north to 80-80.5^oN, while from 2006 it extended north to 81^oN or beyond (Fig. 7). The northwestern corner of Spitsbergen lies just south of 80^oN. With the northward extension from 2006, there was also an eastward extension to cover the waters north of Svalbard (Spitsbergen Archipelago) , east to 20-35^oE. In three of the years, the waters west of Spitsbergen was either not sampled (2016) or only partially sampled (north to 78 ^o N; 1999 and 2005).

In a sector through the central Barents Sea, east of Svalbard (Spitsbergen Archipelago) and east to about 38^oE, the sampling extended north to 76-77 ^oN in the years up to 2002 (except for two years, 1989 and 1991), while from 2004 the survey area extended north to 78 ^oN or beyond (Fig. 7). The northward shift reflects a change to less sea ice and more open water in the northern Barents Sea, while the large variability in recent year reflects variable ice conditions. A similar northward extension is seen for the area east of 38^o E, but with considerable variation among years reflecting variable sea ice conditions as well as vessel availability (Fig. 7).

The survey is semi-synoptic since it takes about 3-4 weeks, or in some cases longer, to complete the survey of 0-group distribution. The 0-group survey typically started in mid-August (10-20 August) and ended in early September (5-15 September). This was the case during the 1980s and 90s when the 0-group investigations were done as a separate cruise, or as the first part of a combined multispecies cruise. This pattern with a main part of sampling in the second half of August and the first part of September has continued after 2004 when the 0-group survey became part of the ecosystem survey, although there has been an extension of sampling later in September as the survey has extended northward (described in the following).

In this section, maps of spatial distribution of 0-group density of six species are shown, based on log transformed abundance per station (colored). The species are: Atlantic capelin Mallotus villosus , Atlantic cod Gadus morhua , haddock Melanogrammus aeglefinus , Atlantic herring Clupea harengus , polar cod Boreogadus saida , and redfish (Sebastes spp.).

Species abundance at stations have been estimated, based on species catches at station, by standard methods (Eriksen et al. 2009), taking into account the opening of trawl, vessels speed, towing distance, and number of depths layers covered (see section 2.3). The abundances have been corrected for size-dependent catch efficiency for all species except redfish Sebastes spp. (section 4). Abundances are given as areal density of 0-group, expressed as number of individuals per square nautical miles.

Abundance is shown by color, where light yellow indicates highest value, and dark blue lowest value. The scale is log10, with steps of one log10 unit, corresponding to factor 10. It ranges from >100 million individuals per (nautical miles)² to <100 individuals per (nautical miles)² . Stations are shown by red dots.

0-group capelin are generally widespread in the Barents Sea (except 1985-86 and 1992-1995) and occurrence area has varied from 116 to 1130 thousand km². Distribution of 0-group herring was widest in 1983, 1992, 2016 and in 2022-2023, and it seems that larger occupation area was not related to occurrence of strong year classes. 0-group cod are generally widely distributed and abundant year classes seem to be observed on larger area. Haddock were widespread in 2004, 2008, 2017 and 2023, and varied from 62 to 630 thousand km². Redfish and polar cod 0-group distributions are generally more restricted than for cod, capelin, haddock, and herring. The widest distribution was observed in 1982-83 (515 thousand km², redfish) and 1999 and 2005 (560 thousand km² , polar cod).

7.1 - Capelin

7.2 - Cod

7.3 - Haddock

7.4 - Herring 

7.5 - Polar cod

7.6 - Redfish

Abundance and biomass estimates were calculated by different software during the last for decades: SAS (for the 23 strata, see Fig. 3, 1980-2017), MatLab (for the new 15 TIBIA/WGIBAR- polygons (see Fig. 1, 1980- 2018, WGIBAR 2018) and R (for the 15 WGIBAR- polygons (2003-2023).

8.1 - Indices calculated in SAS