Approach | Reference point | Value | Basis |
|---|---|---|---|
MSY approach | MSY Btrigger | 154 094 | Bpa |
FMSY | 0.112 | Leads to long-term MSY, based on stochastic simulations (EqSim) | |
Precautionary approach | Blim | 110 893 | Bloss. Lowest SSB (1994) |
Bpa | 154 094 | Blim x e1.645 * 0.2 | |
Flim | 0.1672 | Fishing mortality that in stochastic equilibrium will result in median SSB at Blim | |
Fpa | 0.114 | Fp05, maximum F at which the probability of SSB falling below Blim is <5% |
Key signals
Survey biomass increased steadily from the early 2000s, peaking around 2014–2016, followed by a decline across all surveys in recent years.
Small redfish (<30 cm) were rarely observed in surveys from 2009 to 2020. However, since 2021, surveys show a marked increase in small individuals, suggesting signs of recent recruitment.
Length distributions from surveys and fisheries show a progressive shift toward larger fish over time, reflecting an aging population and a long period (2009–2020) with low recruitment.
Spawning stock biomass (SSB) peaked in 2016; has declined but remains high.
Fishing mortality (F) remained below FMSY since 2009, but increased in 2023 and 2024.
Stock description and management units
Golden redfish (Sebastes norvegicus) in ICES Subareas 5 and 14 (Greenland Sea, Icelandic Waters, and Faroes Ecoregions) have been treated as a single management unit. Catches from ICES Subarea 6 have traditionally been included in this report, and this practice has been maintained. However, data from Subarea 6 are not used in the analytical stock assessment.
Fishery
Landings
Total landings of golden redfish declined by over 70% between 1982 and 1994, falling from 130,429 t to 43,515 t (Figure 1). From 1995 to 2016, annual landings fluctuated between 33,451 t and 59,698 t, peaking in 2016. Since then, landings have decreased. In 2024, total landings were 42,279 t — an increase of approximately 6,300 t compared to 2023. Around 90% of the catch was taken in Icelandic waters (ICES Division 5.a) in 2024.
In Icelandic waters, landings decreased from 97,899 t in 1982 to 38,669 t in 1994 (Figure 1). Between 1995 and 2016, landings ranged from 31,686 t to 54,041 t, with the highest in 2016. Since then, annual catches have declined, reaching 38,355 t in 2024 — 6,163 t (19%) more than in 2023.
Between 90% and 95% of the golden redfish catch in Icelandic waters is taken by bottom trawlers specifically targeting the species. The remainder is caught as bycatch in gillnet, longline, and shrimp fisheries. In 2024, as in previous years, the majority of catches were made along the continental shelf southwest, west, and northwest of Iceland (Figure 2). However, a growing proportion of the catch is now being taken along the northwestern shelf, accompanied by a reduction in fishing effort in the southern and southwestern areas (Figure 3). Most catches occur at depths between 100 and 400 meters (Figure 4).
The number of vessels responsible for 95% of the landings has steadily declined (Figure 5), from over 110 in 1995 to around 50 in 2024.
In Faroese waters, landings declined from 9,194 t in 1985 to less than 200 t by 2016 (Figure 1). After a temporary increase during 2017–2020 to an average of 1,250 t, landings dropped sharply again in 2021–2024, averaging just 185 t per year. Most catches in Faroese waters are taken by pair and single trawlers (vessels >1,000 HP).
In East Greenland waters, landings peaked at 30,962 t in 1982 before declining sharply to 2,117 t by 1985 (Figure 1). During 1985–1994, annual catches ranged from 687 t to 4,255 t. From 1995 to 2009, there was little or no directed fishery for golden redfish, with landings under 200 t annually, mostly as bycatch in the shrimp fishery. In 2010, landings rose to 1,650 t, driven by increased S. mentella targeting. From 2010 to 2024, annual landings ranged from 1,000 t to 5,442 t. In 2024, landings totalled 3,573 t — 500 t (16%) more than in 2023.
Discard
Comparison of sea and port samples from the Icelandic discard sampling program does not indicate significant discarding due to high grading (Pálsson et al. 2010). Substantial discarding of small redfish occurred in the deep-water shrimp fishery between 1986 and 1992, prior to the introduction of mandatory sorting grids. Since the implementation of sorting grids and a decline in the abundance of small redfish in the area, discards have been considered negligible.
Currently, discarding of redfish species in the shrimp fishery in ICES Division 14.b is also regarded as insignificant.
Biological data from the commercial fishery
The table below shows sampling of golden redfish from the catches by ICES divisions in 2024.
| Area | Nation | Gear | Landings (t) | Samples | No. length measured | No. Age read |
|---|---|---|---|---|---|---|
| 5.a | Iceland | Bottom trawl | 38,355 | 97 | 15,480 | 903 |
| 5.b | Faroe Islands | Bottom trawl | 250 | 392 | ||
| 14 | Greenland | Bottom trawl | 3573 | 2 | 131 |
Sampling from the commercial catches in Icelandic waters is considered adequate. The program appears to cover both the spatial and seasonal distribution of the fishery effectively (Figure 6, Figure 7). In 2020, sampling effort — particularly onboard sampling — was substantially reduced due to the COVID-19 pandemic. Nevertheless, the available data are considered sufficiently representative of fishing activities and are not expected to have significantly impacted the stock assessment.
Landings by length and age
Length distributions from the Icelandic commercial trawler fleet during 1976–2024 show that most fish caught range between 30 and 45 cm in length, with modal lengths typically between 35 and 40 cm (Figure 8). Over the past decade, the length distribution has shifted to the right, becoming narrower, with a reduced proportion of fish <35 cm. The mean length has increased by nearly 5 cm during this period.
Catch-at-age data from the Icelandic fishery in Division 5.a reflect the dominance of strong year classes over time. The 1985 year class dominated catches from 1995 to 2002 (Figure 9), while the 1990 cohort contributed 25–30% of the total catch by weight between 2003 and 2007. The 1996–1999 cohorts were dominant in 2007–2010 but are now gradually declining. In 2023, the 2004–2009 cohorts (ages 14–19) were the most prominent in the catch. A notable decline in 7–10-year-old fish in recent catches suggests reduced recruitment in recent years, consistent with trends observed in surveys from both East Greenland and Icelandic waters.
Length distributions from the German commercial fleet in East Greenland indicate similar trends to those observed in Icelandic waters (Figure 10).
In Faroese commercial catches (2001–2020 and 2023–2024), fish are generally larger, with average lengths exceeding 40 cm and modal lengths between 45 and 50 cm (Figure 11).
Catch per unit effort
The unstandardized CPUE index from the Icelandic bottom trawl fleet operating in Division 5.a increased sharply from 2006, reaching the highest levels in the time series during 2017–2019 (Figure 12). Although CPUE has declined somewhat since then, it remains relatively high. Data for 2022 are not available.
Fishing effort directed at golden redfish has steadily decreased since 1986 and is currently at the lowest level recorded in the time series (Figure 12). While CPUE estimates derived from logbook data are not used in the stock assessment, the logbooks provide valuable information on fishing effort as well as the spatial and temporal distribution of the fishery.
CPUE data from other areas are not available.
Scientific data
This section describes stock development based on various surveys conducted annually on the continental shelves and slopes of ICES Subareas 5 and 14. A detailed description of the survey design and the methodology used to calculate the survey indices for golden redfish is provided in the Stock Annex (reg-5614_SA). The calculation of survey indices accounts for the species’ length-dependent diel vertical migration.
Division 5.a (Icelandic waters ecoregion)
Information on the abundance and biological characteristics of golden redfish in Division 5.a is derived from two surveys: the Icelandic spring groundfish survey (IS-SMB) and the Icelandic autumn survey (IS-SMH). The IS-SMB has been conducted annually in March since 1985, while the IS-SMH has been carried out each October since 1996, with the exception of 2011, when the autumn survey was not conducted.
The total biomass of golden redfish observed in the spring survey declined from 1988 to a record low in 1995 (Figure 13). From 2000 to 2016, biomass increased steadily, reaching the highest level in the time series. Since then, it has remained relatively high, although with some fluctuations and a decline over the past three years. The biomass index from the autumn survey followed a similar pattern up to 2019 (Figure 14), but declined sharply during 2020–2022. It increased again in 2023, before falling in 2024 to levels comparable to those observed during the earlier decline (Figure 13).
Length-disaggregated indices from the spring survey reveal strong recruitment signals in the 4–11 cm group in 1987 and 1991–1992, corresponding to the 1985 and 1990 cohorts, respectively (Figure 13). These cohorts contributed to the biomass increase observed between 1995 and 2005, as they recruited into the fishable stock roughly a decade later (Figure 15). Between 1999 and 2008, the abundance of small redfish was lower than during 1986–1990, with the highest values recorded in 2000–2003. From 2009 to 2020, very few small redfish (4–11 cm) were observed. However, from 2021 to 2025, the index increased to levels similar to those seen around the year 2000, with a corresponding rise in fish in the 12–29 cm length group (Figure 13).
Over time, the length distributions in both the spring and autumn surveys have shifted to the right and become more narrowly peaked, indicating an aging population. From 2006 to 2022, the abundance of fish smaller than 30 cm was at the lowest level observed in the time series (Figure 13, Figure 15, ?@fig-surveyldistsmh). Since 2023, the number of individuals below 30 cm has increased, suggesting improved recruitment, although overall abundance in this size group remains relatively low. Despite the shift toward larger modal lengths, the mean length has decreased in recent years, reflecting the presence of more small individuals in the population.
The marked increase in survey indices from 2005 onward reflects the recruitment of year classes from 1996 to 2007 (Figure 16). The 1996–2002 cohorts are now gradually declining, while fish from the 2003–2008 year classes currently dominate the stock. Age-disaggregated indices suggest that the 2009–2019 cohorts are comparatively small.
Most golden redfish are caught southwest, west, and northwest of Iceland (Figure 17) at depths of 100–400 m (Figure 18). In general, golden redfish are now more commonly found further north than in earlier years. Additionally, less golden redfish are caught at 100–200 m depth in the spring groundfish survey (SMB) compared to the beginning of the time series.
Division 5.b (Faroes ecoregion)
In Division 5.b, biomass indices for golden redfish are available from the Faroes spring groundfish survey (1994–2025) and the Faroes summer groundfish survey (1996–2024). Both surveys indicate a declining trend in biomass between 1996 and 2000, followed by consistently low levels thereafter (Figure 19). The length distributions are centered around 40–45 cm, and fish smaller than 35 cm are rarely observed (Figure 20, ?@fig-surveyldistq3far). On average, the redfish caught in these surveys are larger than those sampled in Icelandic waters and East Greenland surveys.
Subarea 14 (Greenland Sea ecoregion)
Information on the abundance and biological parameters of golden redfish in Subarea 14 is available from two surveys: the German Groundfish Survey and the Greenland Shallow Water Survey. Only data from the German survey are currently used in the assessment.
The German Groundfish Survey was conducted annually in the autumn from 1982 to 2017 and again in 2019–2020, covering shelf areas and continental slopes off East Greenland. It was not conducted in 2018 and in 2021–2024. Abundance and biomass indices for golden redfish (>17 cm) are shown in Figure 21. Following a period of severe stock depletion in the early 1990s, survey estimates increased significantly from 2003 to 2016. Biomass reached its highest levels in 2014–2016 but declined from 2017 to 2020 to values similar to those observed around 2006. It should be noted that the coefficients of variation (CV) for these indices are high, with the increase driven by a few very large hauls. The biomass of pre-fishery recruits (17–30 cm) declined during 2010–2020 compared to the previous five-year period, and very few individuals in this size range were observed from 2017 to 2020.
The Greenland Shallow Water Survey, conducted from 2008 to 2024 (excluding 2017–2019 and 2021), covers shelf areas off East Greenland down to 600 m depth. Index values have been highly variable throughout the time series (Figure 23). From very low levels in 2008–2010, the indices increased to their highest values in 2011–2016, driven by the appearance of small redfish (<30 cm). By 2016, these fish had grown to 25–40 cm, with a mode around 30 cm (Figure 24). Both abundance and biomass declined again in 2020–2024 to levels similar to those observed in 2008–2010 (Figure 24).
Abundance indices for redfish <18 cm (not classified to species) from the German survey show that juveniles were abundant in 1993 and 1995–1998 (Figure 25). In contrast, the index was very low during 2008–2016. The Greenland Shallow Water Survey also recorded low juvenile abundance (<18 cm, not classified to species) in 2013–2016, but observed a notable increase in 2022–2024 — the highest in over a decade. In both surveys, juveniles were only classified to the genus Sebastes spp., as species-level identification is difficult due to morphological similarities at small sizes. The recent increase in juvenile abundance in the Greenland Shallow Water Survey suggests a potential improvement in recruitment of either S. mentella, S. norvegicus, or both.
Analytical assessment
The stock was benchmarked in February 2023 (WKBNORTH 2023; ICES 2023), resulting in a change in the assessment method. The current assessment is based on the SAM model (Nielsen and Berg 2017), and updated reference points were adopted. Development of the previous Gadget model was discontinued, as a sufficiently long time series of age data is now available to support an age-based assessment.
Survey indices
Survey indices for golden redfish are calculated using a design-based method (Cochran, 1977) across five surveys conducted in the Greenland, Icelandic, and Faroes ecoregions. For input into the SAM model, two length-disaggregated indices were constructed to cover the full range of the stock:
- Spring survey index
Icelandic spring survey: 1985–2025.
German Autumn Survey: 1984–2020, with the index year shifted by one year (𝑦 + 1). For the missing year 2018, the average of 2017 and 2019 was used; for 2021–2024, the 2020 value was applied.
Faroese Spring Survey: 1994–2025. For the years 1985–1993, the average of the 1994–1999 period was used.
- Autumn survey index:
Icelandic Autumn Survey: 1996–2024. For the missing year 2011, the average of 2010 and 2012 was used.
German Autumn Survey (East Greenland): 1996–2020 from East Greenland.he 2018 index was replaced by the average of 2017 and 2019; for 2021–2024, the 2020 value was used.
Faroese Summer Survey: 1996–2024.
Figure 26 shows the two combined survey indices, disaggregated by area. The overall index is primarily driven by the Icelandic survey data.
Stock weights
Although golden redfish rarely exceed 60 cm in length or 2 kg in weight in surveys and commercial catches, their growth is highly variable from year to year. This variability results in a wide range of ages for fish around 30 cm in length. Consequently, age–length keys are also highly variable. This is believed to reflect true variation in growth rather than ageing error, as ageing consistency is generally considered good, albeit based on anecdotal evidence.
Despite these temporal differences in growth, the length–weight relationship remains highly stable, suggesting that condition does not vary substantially over time.
Weight-at-length data are available from both survey and commercial sources (Figure 27, Figure 28). Stock weights were estimated as mean weight-at-age, based on the combined spring survey, by converting lengths to weights using a power function fitted to fish with paired length and weight data from both surveys and commercial samples. For each year, stock weights were calculated as the mean expected weight based on that year’s observed length distribution.
For years prior to 1985, when survey data were unavailable, commercial catch weight data — available from 1966 — were used. In the rare instances where weight-at-age data were missing, values were supplemented using the alternative source. To reduce interannual variability, stock weights were smoothed using a two-year moving average including the current and previous year.
Maturity
Maturity at length is relatively stable across years and regions; therefore, a fixed maturity ogive is applied to length distributions and subsequently averaged within age groups after applying the age–length key (ALK). To ensure consistency with previous assessments and facilitate comparison between modelling frameworks, the same functional form as used in the former Gadget model has been retained:
\[ P = \frac{1}{1 + \exp(-0.3122 \cdot (\text{length} + 1.5 - 33.54))} \]
This ogive was updated by fitting it to maturity-at-length data pooled across all years, using data collected during the Icelandic spring survey. Although changes to the maturity ogive do not influence model estimation, they do affect the calculation of spawning stock biomass (SSB) and therefore the estimation of reference points. All reference points presented in this assessment are based on the updated maturity ogive.
To reduce year-to-year variation, maturity at age was smoothed. For individuals younger than age 15, maturity was calculated as the average of the current and previous year. For ages 15 and older, a four-year average (current year plus the three previous years) was used. The resulting maturity-at-age estimates are shown in Figure 29.
Natural mortality
In the previous Gadget model, natural mortality (M) was fixed at 0.05 for most age groups, with a higher value of 0.1 applied to the plus group. The same approach is used in the current model: all profile likelihoods incorporate a plus group with natural mortality set to 0.1.
Assessment
The SAM runs from 1966 onward and tracks ages 6 through 25+, treating age 25 as a plus group. Observations in SAM are modelled as arising from a multivariate normal distribution, with expected values derived from the population dynamics model. The framework supports flexible treatment of residual patterns by allowing age-specific parameterization of observation residual variances and correlations across all data sets. Users can also define age groupings for survey catchabilities, specify associated power functions, and set process variances for the residuals in log-abundance (log(N)) and log-fishing mortality (log(F)).
Development of this SAM model began with refinement of the age–length key (ALK) and selection of an appropriate age structure, with an emphasis on preserving correlations among consecutive cohort observations in catch-at-age and survey index data. The youngest ages observed in the catch data (ages 5 and 6) were excluded from the model due to high levels of noise. As a result, the model begins at age 6, which corresponds to the earliest age at which golden redfish are consistently observed in survey data.
Data and model settings
Below is a brief description of the data used in the model and the model settings is given.
The simulation period is from 1966 to 2024.
Two survey indices for the whole area used.
Spring survey length data from 1985–2024. As little age data are available for the spring survey, it was inputted as a single total biomass series.
Autumn survey length and age data from 1996–2023.
Age ranges in the model spanned ages 6–25+.
- Although age data range to 60, individual ages detected can be sparse by year in the range 25–60.
Age-length keys (ALKs) for the surveys were created and applied within regions (east versus west) to account for regional growth differences from autumn survey data.
- The east ALK was applied to length data from Faroese surveys and the west ALK was applied to length data from Greenlandic surveys.
ALKs generated from commercial samples were applied within biannual time periods (January-June and July-December, but not by region) to catch length distributions.
All ALKs were created using 2 cm length bins from 6–60 cm, with longer bins at lengths 0–6, 61–70, and 70+.
Catch at age and total landings are available from 1966, but only those from 1995 on-wards are used due to available age data.
An ALK generated by pooling data from the years 1995–2003 was applied to length distribution data in 1966 and 1972.
Annual ALKs were created from 1995 onwards to account for time-variable growth. These ALKs are time-specific (biannual, January-June and July-December) and applied to the approximate amount of catch from the corresponding period. This was done to account for differences in growth patterns between sampling times.
Total catch-at-age over sectors is used in the tuning.
Only Icelandic commercial length distribution data was used.
- These total catches at ages were scaled according to total landings across all countries and areas fished within the stock.
Recruitment was set at age 6.
Natural mortality (M) was set to 0.05, except for the oldest age (25+) which was set to 0.1.
Results of the assessment model
The population dynamics of golden redfish estimated by this model reveal a clearly defined period of variable recruitment between 1990 and 2013. Notably, relatively high recruitment during 2000–2013 coincides with increases in both SSB and catches after 2010 (Figure 30). Since 2014, however, recruitment has declined sharply and entered a prolonged low phase. It remains uncertain whether this trend reflects a shift in stock productivity or a natural, extended low in a highly autocorrelated recruitment time series.
SSB and total biomass peaked in 2016 but have declined since then, although they remain at relatively high levels (Figure 30). Fishing mortality declined steadily from the early 1990s to 2022 but has increased over the past two years (Figure 30).
The spawning stock biomass over the past decade, as estimated in this model, is higher than in the previous Gadget assessment. This difference is primarily attributed to improved accounting for variable growth: the model better incorporates the presence of older individuals in the stock, which increases the estimated number of older spawners. Additionally, the faster growth of smaller individuals enhances their contribution to the spawning biomass.
Caution is advised when interpreting trends prior to the availability of age data in 1996, as limited data constrain the reliability of model estimates during that earlier period.
Retrospective analysis
The analytical retrospective pattern, based on a five-year peel, is shown in Figure 31. The most recent assessment run shows a slight downward revision of spawning stock biomass (SSB) compared to the 2024 assessment, primarily due to lower biomass estimates from the 2024 autumn survey and the 2025 spring survey.
The table below presents the Mohn’s rho values for SSB, fishing mortality (F), and recruitment over this five-year retrospective period:
| Variable | Value |
|---|---|
| Fbar | 0.074 |
| SSB | -0.013 |
| Rec. | -0.630 |
The Mohn’s rho values for average fishing mortality (F̄) and SSB are 7.4% and 1.3%, respectively. In contrast, the Mohn’s rho for recruitment is notably higher at 63%, likely reflecting greater uncertainty due to low survey selectivity at the youngest modelled age (age 6). Despite this, all Mohn’s rho values fall within the acceptable range recommended by Carvalho et al. (2017), which suggests values below 0.2.
These results indicate that the model demonstrates relatively good retrospective consistency for SSB and F̄, supporting the stability of the assessment for these key indicators. The higher rho for recruitment highlights the known challenges in estimating early life stages, but does not substantially undermine the overall reliability of the assessment.
Diagnostics
Fits to the survey numbers-at-age indices and catch-at-age data are shown in Figure 32 and Figure 33, with fits to the spring survey index shown in Figure 34. The fit to total catch and landings data is presented in Figure 35. Among these, the model fits the catch and spring survey data most closely. Fits to the autumn survey series are somewhat noisier, but still follow the overall pattern. While fits to the landings data are more variable, model performance has improved in more recent years, particularly for catch-at-age data.
Neither observation nor process residuals display any clear trends, as shown in Figure 36 and Figure 37.
An overview of estimated model parameters is provided in Figure 38. Parameters with overlapping estimates across ages within a data source were grouped, and only those showing appreciable differences were retained separately.
Leave-one-out analysis
Figure 40 presents a comparison between the full model estimates and those from a sensitivity run in which the landings data were omitted from the observation likelihood. The exclusion of landings data had a limited effect on estimates of SSB and recruitment, which remained broadly consistent with the full model. However, estimates of fishing mortality (F) were more sensitive to the omission, showing greater variability and wider uncertainty bounds. This suggests that landings data play a more critical role in informing fishing mortality trends than in shaping biomass or recruitment estimates.
In contrast, when either the spring or autumn survey data were excluded from the model, it failed to converge. This underscores the essential role of both survey series in anchoring model estimates and ensuring convergence, particularly for population abundance and recruitment at age.
Reference points
During the 2023 Benchmark meeting, reference points were updated (Table 1). In accordance with ICES technical guidelines, MSY Btrigger was set equal to Bpa in simulations where the ICES advice rule was applied. Under this rule, a constant target fishing mortality (F) is applied when SSB is above Btrigger; when SSB falls below Btrigger, the target F is scaled down proportionally by the ratio SSB/Btrigger.
The fishing mortality that produces maximum sustainable yield (FMSY) is estimated to be 0.112. The precautionary fishing mortality threshold (Fp05) — defined as the maximum F that results in less than a 5% probability of SSB falling below Blim when the advice rule is followed — is greater than 0.112. Therefore, Fp05 does not constrain the estimate of FMSY.
State of the stock
Results from the SAM assessment model indicate that fishing mortality has remained low and below FMSY since 2009 (Figure 30) although F has increased in 2023 and 2024. While total biomass and SSB have been declining since 2016, they continue to remain at relatively high levels (Figure 30). Survey data from Iceland and East Greenland suggest that recruitment was poor for cohorts from approximately 2009 to 2019. However, surveys conducted from 2021 to 2025 in both regions show signs of increased abundance of small golden redfish (<12 cm). Although the reliability of these surveys as predictors of recruitment is uncertain, recruitment in the coming years is still expected to remain low.
Short term forecast
Short-term projections were conducted using the standard procedure in SAM using the forecast function. Three-year averages were applied for stock weights, catch weights, and maturity-at-age. These projections form the basis for deriving management advice. Given that recruitment has remained consistently below historical levels for the past eight years, the stock is projected using the mean recruitment over the previous five years, continuing the approach used in recent assessments (Table 2). Catches for 2025 were fixed at the level of the TAC set for that year.
The outcomes of the short-term forecast are presented in Table 3. The results indicate that, under fishing consistent with the ICES MSY approach, SSB is expected to decline slightly but remain well above the MSY Btrigger reference point. Specifically, the projected SSB in 2026 under FMSY is estimated to be [insert value] tonnes, with a projected catch of [insert value] tonnes in 2026. Under a no-catch scenario, SSB would increase modestly to [insert value], while fishing at a higher F (e.g., status quo F or Fp05) results in a lower SSB but still above the trigger point.
Variable | Value | Notes |
|---|---|---|
Fages9-19 (2025) | 0.132 | From the forecast for 2025, based on the assumed catch in 2025; in tonnes |
SSB (2026) | 277 646 | Projected from the assessment; in tonnes |
Recruitment age 6 (2025) | 42 905 | From the assessment; in thousands |
Recruitment age 6 (2026) | 67 314 | Average of the last five cohorts in 2021–2025; in thousands |
Catch (2025) | 46 911 | Sum of expected landings from (2025); in tonnes. |
Basis | Catch (2026) | Fishing mortality ages 9-19 (2026) | SSB (2027) | % SSB change1) | Advice change2) |
|---|---|---|---|---|---|
MSY approach | 41 345 | 0.112 | 265 015 | -5 | -12 |
1) SSB in 2027 relative to SSB in 2026 | |||||
2) Advice value for 2026 relative to advice value for 2025 relative to advice value for (41286 t) | |||||
Uncertainties in assessment and forecast
Significant changes in growth have been observed in golden redfish in recent years, affecting both older and younger age classes. While density dependence is a possible driver of these changes, concurrent ecosystem shifts observed in other species around Iceland suggest that broader environmental factors may also be involved. If growth continues to shift as expected during the anticipated stock decline over the next 5–10 years, it may become possible to model growth as a cohort or annual effect. This could improve the accuracy of short-term forecasts and help align realized harvest rates with those expected under the ICES advice rule.
These recent changes in growth have also altered our perception of SSB, raising questions about the implications of a changing age structure on recruitment. Understanding whether shifts in the composition of the spawning stock affect recruitment dynamics will be important for future assessments.
It remains unclear whether survey selectivity patterns follow a logistic function with age or are more dome-shaped, as both configurations produce similar fits to the data but yield different estimates of total stock biomass. Additionally, recent growth changes appear to coincide with shifts in commercial selectivity—likely driven by spatial changes in fishing effort. This suggests that further research into whether density-dependent growth is spatially explicit could provide valuable insights into both stock dynamics and fishery behavior.
To support effective management, continued monitoring of growth, age structure, and spatial fishing patterns will be essential for detecting and responding to changes in stock productivity.
Comparison with previous assessment and forecast
From 2014 to 2022, the Gadget model (Globally applicable Area Disaggregated General Ecosystem Toolbox) was used for the assessment of golden redfish. However, several challenges emerged with this framework, prompting the need for a benchmark revision.
First, length-based survey indices across different length ranges show internal inconsistencies. Specifically, fitting the model to indices of smaller golden redfish requires disregarding patterns in the larger size classes, and vice versa. Second, these inconsistencies are reflected in the length distribution data, which show narrow distributions with limited evidence of recent recruitment and a lack of larger-sized individuals, despite the species’ high longevity. Third, growth patterns appear to vary slightly by region, but length-at-age data are highly variable and indicate a trend toward faster growth in recent years — i.e., larger fish at younger ages. This may reflect density-dependent somatic growth.
In contrast, age-based models produce more stable results than the Gadget model when regional and temporal variation in growth is accounted for by applying region- and time-specific age–length keys (ALKs) during the generation of total catch and survey inputs.
Basis for advice
The ICES MSY approach applied in this assessment follows the framework agreed upon during the WKBNORTH meeting (ICES, 2023).
Management consideration
In 2009, a targeted redfish fishery was initiated in Subarea 14, with annual catches ranging from 6,000 to 8,500 tonnes between 2010 and 2024. Catches peaked in 2015 and were lowest in 2018. Although the fishery does not differentiate between redfish species, survey data indicate that golden redfish accounted for an estimated 1,000 to 2,700 tonnes annually during 2010–2015, increasing to 3,000 to 5,400 tonnes in 2016–2024.
Subarea 14 serves as an important nursery area for the entire stock. Protective measures for juveniles in this area—such as the use of sorting grids in the shrimp fishery—should therefore be maintained.
Currently, no formal trilateral agreement exists among the coastal states of Greenland, Iceland, and the Faroe Islands for the management of golden redfish. However, in July 2023, Iceland and Greenland reached a bilateral agreement to manage the golden redfish fishery based on the ICES management plan implemented in 2023. The strategy aims to maintain an exploitation rate that aligns with the precautionary approach and achieves maximum sustainable yield (MSY) in the long term. Effective from the beginning of 2024, the agreement allocates 91% of the total allowable catch (TAC) to Iceland and 11% to Greenland, with an additional 300 tonnes allocated annually to other areas before the national shares are applied.
In Iceland and Greenland, the fishery is regulated by TACs, while in the Faroe Islands it is managed through effort limitations. These regulatory approaches have historically resulted in total catches exceeding the TACs recommended by ICES.
Since 2009, surveys across the stock area have consistently shown very low abundance of young redfish (<30 cm). SSB and harvestable biomass increased from 1995 to 2015, driven by the recruitment of several strong year classes. However, biomass has declined since then. The persistent absence of clear signs of incoming cohorts raises concerns about the future productivity of the stock.
Regulations and their effects
In the late 1980s, Iceland introduced a sorting grid with a bar spacing of 22 mm in the shrimp fishery north of Iceland to reduce bycatch of juvenile fish, including golden redfish. This measure was intended to protect juveniles of several species and was implemented as part of broader bycatch reduction efforts. Following the disappearance of the large golden redfish year classes from the shrimp fishing grounds in the early 1990s, observer reports indicated that small redfish had become negligible in the Icelandic shrimp fishery. While questions remain about the effectiveness of sorting grids in areas with high redfish abundance, this is not currently a concern in Icelandic waters, where both the abundance of juvenile redfish and shrimp fishing activity are low.
There is no minimum landing size for golden redfish in Division 5.a. However, if more than 20% of a catch observed onboard is below 33 cm, a temporary closure of a small area may be enforced. In addition, a large area west and southwest of Iceland is permanently closed to protect juvenile golden redfish.
In Division 5.b, there are currently no regulations specific to the management of golden redfish.
Since 2002, the use of sorting grids has been mandatory in the shrimp fishery in Subarea 14 to reduce bycatch of juvenile redfish.
Management in Icelandic waters
Ministry of Industries (MI) in Iceland is responsible for the management of Icelandic fisheries, including the golden redfish fishery, and for implementing fisheries legislation within the Icelandic Exclusive Economic Zone (EEZ). The Ministry issues annual regulations for commercial fishing, covering the fishing year from 1 September to 31 August. These regulations include the allocation of total allowable catches (TACs) for stocks subject to catch limitations.
Golden redfish and Icelandic slope beaked redfish (S. mentella) have been within in the Individual Transferable Quota (ITQ) system since its inception. However, until the 2010/2011 fishing year, a joint quota was allocated for both species, and fishers were not required to report species-specific catches. Although the Marine and Freshwater Research Institute (MFRI) has provided separate stock advice for the two species since 1994, species-specific quotas were only implemented beginning with the fishing year starting on 1 September 2010.
Figure 41 shows the net transfer of quota to or from the golden redfish allocation by fishing year. Between the 2014/2015 and 2022/2023 fishing years, actual catches exceeded the set TAC by 2–24%, with the largest overages occurring during the 2020/2021 to 2022/2023 period. These implementation overruns were primarily due to features of the management system that allowed inter-annual quota transfers and the conversion of TAC between species — a mechanism referred to as species transformation. However, as of the 2023/2024 fishing year, species transformation is no longer permitted for golden redfish, addressing one of the key drivers of past overharvest.
This regulatory change is expected to improve adherence to TAC limits in future years and reduce the risk of exceeding catch recommendations, thereby strengthening the effectiveness of the management system
References
Carvalho, F., Punt, A. E., Chang, Y. J., Maunder, M. N., & Piner, K. R. 2017. Can diagnostic tests help identify model misspecification in integrated stock assessments? Fisheries Research, 192, 28–40.
Cochran, W.G. 1977. Sampling techniques. John Wiley & Sons.
ICES. 2023. Benchmark workshop on Greenland halibut and redfish stocks (WKBNORTH). ICES Scientific Reports. 5:33. 408 pp. https://doi.org/10.17895/ices.pub.22304638
Nielsen, A. and Berg, C. W. 2014. Estimation of time-varying selectivity in stock assessments using state–space models. Fisheries Research, 158: 96–101. https://doi.org/10.1016/j.fishres.2014.01.014
Pálsson, Ó., Björnsson, H., Björnsson, E., Jóhannesson, G. and Ottesen Þ. 2010. Discards in demersal Icelandic fisheries 2009. Marine Research in Iceland 154.