National Electrofishing Programme for Scotland (NEPS) 2023: status of juvenile Atlantic salmon and brown trout populations

Results

Capture Probability (P)

The final capture probability model for salmon and trout was:

logit P ~ Species + Lifestage + Pass + Species:Lifestage + Lifestage:Pass + Organisation_Team + s(Year) + Altitude + Species:Altitude + s(UCA:Lifestage) + Gradient + Species:Gradient + s(DoY:Lifestage) + s(HA)

where s() denotes smoothed responses and : indicates an interaction term.

The final capture model was generally consistent with previous years (Malcolm et al., 2023). Capture probability was higher for trout than salmon (for a given lifestage), for parr than fry and in the first pass than in subsequent passes (Fig. 5A).

Capture probability increased over time (across years). However, in contrast to previous analyses (Malcolm et al., 2023), the trend appeared to plateau in recent years (Fig. 5B). Within years, capture probability varied with DoY with a different pattern of change for each lifestage. Fry exhibited a strong modal response, while parr exhibited a more linear positive response (Fig. 5D).

Figure 5. The effects of Species : Lifestage : Pass, (A) Year (B), Organisation - Team (C) Day of the Year : Lifestage (D), Altitude : Species (E) Gradient : Species (F) Upstream Catchment Area : Lifestage (G), and Hydrometric Area (H) on capture probability. Where effects differed between Species or Lifestage they are plotted separately for salmon (black dotted), trout (orange dot dash), fry (blue solid), parr (green dashed). All effects are scaled to the mean fitted first pass capture probability. Approximate 95% pointwise confidence intervals are shown as shaded areas or vertical lines. Short vertical lines (a rug) indicates the distribution of the data on the x-axis. Only Organisation Teams contributing to NEPS in 2023 are shown.

Capture probability also varied spatially with Altitude, Gradient and UCA (Fig. 5E, F, G). Altitude had negative linear effects which varied with species, whereas the non-linear negative effect of UCA varied by lifestage. Altitude had stronger negative effects for trout than salmon. Gradient had a strong negative effect on the capture probability of salmon, but no effect for trout. UCA showed non-linear negative effects, which levelled off at lower UCA values, particularly for fry.

Consistent with previous analyses (Millar et al., 2016; Malcolm et al., 2019; Malcolm et al., 2023) there was substantial variability in capture probability between organisations and teams (Fig. 5C). As Organisations and Teams typically work within fixed geographic areas it was difficult to disentangle HA effects (Fig. 5H) from Organisations and Teams. Consequently, the combined effects of both HA and Organisation-Team are also illustrated in Figure 6 for those Organisations – Teams contributing to NEPS 2023.

Figure 6. Combined partial effect of Organisation - Team and Hydrometric Areas on capture probability. All effects are scaled to the mean fitted first pass capture probability. Approximate 95% pointwise confidence intervals are shown as vertical lines. Only Organisation - Teams contributing to NEPS in 2023 are shown.

Site-wise estimates of abundance and status from NEPS 2023

Estimates of salmon and trout densities at individual sites are shown in Figures 7 and 8 respectively. Comparison to the relevant benchmark (Malcolm et al., 2019a; Malcolm et al., in prep) provides an assessment of performance, based on the landscape habitat quality and thus average expected number of juveniles under healthy conditions. Figures 7 and 8 also show two measures of performance against the benchmark; percentage of benchmark (C and D) and absolute difference from benchmark (E and F).

There was considerable spatial heterogeneity in patterns of salmon abundance. In general, fry and parr densities were highest in northern rivers and those around the Moray Firth (Fig. 7 A, B). The Tweed also exhibited some high fry densities, but generally lower parr densities. Densities were generally lowest in the northeast corner (north of the Dee), central belt and south-west with many sites containing no salmon. Overall, there was a substantial number of sites below the benchmark. Only rivers in the north and around the Moray Firth contained substantial proportions of well performing sites.

Figure 7. Site-wise maps showing spatial variability in observed salmon densities (A, B) together with their percentage (C, D), and absolute (E, F) performance against benchmark. Panels A, C and E show the results for fry. Panels B, D and F show results for parr. Black points (panels A and B) indicate sites where no fish of the relevant lifestage were caught.

Figure 8. Site-wise maps showing spatial variability in observed trout densities (A, B) together with their percentage (C, D), and absolute (E, F) performance against benchmark. Panels A, C and E show the results for fry. Panels B, D and F show results for parr. Black points (panels A and B) indicate sites where no fish of the relevant lifestage were caught.

Trout fry and parr densities were generally greatest in the north-east (Fig. 8 A, B). The Tweed also had high abundances of trout fry, but not parr. Across the country as a whole, comparisons to the benchmark indicate generally poor performance, with many sites failing to achieve the benchmark. The Tweed (fry), Spey and Caithness areas appeared to be outliers in this context.

National assessments of abundance and status from NEPS 2023

At the national scale, salmon fry, salmon parr and trout parr density estimates (and associated upper confidence limits) were all below the benchmark (Fig. 9). This results in a Grade 3 assessment, indicating with a reasonable degree of confidence that, at the national scale, salmon (fry and parr) and trout parr stocks are performing less well than expected based on the benchmark models. For trout fry the density estimate was below the benchmark. However, it exceeded 50% of the benchmark and the upper confidence limit included the benchmark. This results in a Grade 2 status for trout fry at the national scale, reflecting lower confidence that trout fry are meeting the benchmark.

Figure 9. Mean density estimates of salmon and trout, fry and parr for the NEPS 2023 survey. Black circles and error bars indicate mean density estimates and 2-sided 90% confidence intervals. Green squares indicate the target benchmark.

Strata and regional assessments of abundance and status from NEPS 2023

A strength of the NEPS 2023 survey design is that it is possible to analyse and present the results at a wide range of spatial scales. However, this flexibility can also be challenging as different organisations want to see the results presented at different scales. For the purposes of this report, the analysis is illustrated at two spatial scales. Strata represent the basic underlying sampling and reporting unit so these results are presented first. However, in some cases these strata contain few samples or relate only to small sub-catchments. Consequently, the results are also presented at the larger regional scales.

There was substantial spatial variability in density and performance of NEPS 2023 strata against the salmon benchmark (Fig. 10). The general picture was of poor performance, with some strata reporting very low values that were well below the benchmark. Only the Beauly, Helmsdale, Northern and OuterHebrides_NHarris_SAC strata significantly exceeded the benchmark for both salmon fry and parr (i.e. lower 95% one sided confidence interval was above the benchmark).

Spatial variability in densities compared to the benchmark were reflected in subsequent grades (Fig. 11, Appendix 2). All Grade 1 strata for salmon fry were north of the central belt. All Grade 1 strata for salmon parr were in the Moray Firth (Spey) or northwards. Only a few strata achieved Grade 1 for both fry and parr. Aside from the Tay, overall Grade 1 strata were in the north (Beauly, Caithness_BerriLang_SAC, Conon, Helmsdale, KyleSutherland, Northern, OuterHebrides_NHarris_SAC, OuterHebrides, SkyeLochalsh, WestSutherland, WestSutherland_Laxford,) and Moray Firth (Spey).

Figure 10. Salmon density estimates (black circles), with associated 2-sided 90% confidence intervals for each of the strata included in NEPS 2023. Rows denote lifestage. Green squares indicate the benchmark (expected densities) against which estimated densities can be compared to determine performance grade. Numbers at the base of the plot show the number of samples in the strata.

Figure 11. NEPS 2023 strata grades for salmon fry, parr and overall gradings (fry and parr for the species).

Substantial variability in abundance and performance was also apparent when reporting to NEPS regions (Fig. 12). The point estimates for many regions were below the benchmark and only the Beauly and Brora_Helmsdale significantly exceeded the benchmark for both salmon fry and parr.

This was reflected in the regional grades (Fig. 13). All Grade 1 regions for salmon fry were north of the central belt and all Grade 1 regions for salmon parr were north of the Tay. Aside from the Spey, overall Grade 1 regions were in the north (Beauly, Brora_Helmsdale, Caithness, Conon, Kyle_Sutherland, Northern, Outer_Hebrides Skye_WesterRoss, West_Sutherland),

Figure 12. Salmon density estimates (black circles), with associated 2-sided 90% confidence intervals for each of the NEPS regions in 2023. Rows denote lifestage. Green squares indicate the benchmark (expected densities) against which estimated densities can be compared to determine performance grade. Numbers at the base of the plot show the number of samples in the region.

Figure 13. Regional salmon fry, parr and overall grades (fry and parr) in 2023.

In common with salmon, there was substantial variability in the abundance and performance of trout between strata (Fig. 14). Only two strata (Caithness_Thurso_SAC and WestSutherland_Laxford) significantly exceeded the benchmark for both fry and parr densities, although some strata significantly exceeded the benchmark for a single lifestage. For example, the Tweed significantly exceeded the benchmark for fry and Caithness, Lomond_Endrick_SAC and WestSutherland_Laxford for parr. There were also several strata which had very low densities well below the benchmark for one or both lifestages.

The performance of trout was generally poor (Fig. 15). There were many Grade 3 strata, especially for trout fry. Grade 1 strata for fry were only found in the north, Spey and Tweed. Grade 1 strata for parr were confined to the north and north-west, Spey and Lomond_Endrick_SAC. Strata with an overall Grade 1 were in the north (predominantly Caithness) and the Spey.

Figure 14. Trout density estimates (black circles), with associated 2-sided 90% confidence intervals for each of the strata included in NEPS 2023. Rows denote species and lifestage. Green squares indicate the benchmark (expected densities) against which estimated densities can be compared to determine performance grade. Numbers at the base of the plot show the number of samples in the strata. Note that in NEPS 2023 some strata have been further separated to allow SAC reporting.

Figure 15. NEPS 2023 strata grades for trout fry, parr and overall.

Variability in the abundance and performance of trout was also apparent at regional scales (Fig. 16) but the general picture was of poor performance. Only Caithness exceeded the benchmark (including confidence bounds) for both fry and parr densities and only the Tweed for fry. Densities were below the benchmark for one or both of the lifestages in a number of regions.

Generally poor performance across the country was reflected in the regional grades with a large number of Grade 3 regions (Fig. 17). In the case of trout fry, only 4 regions were Grade 1 (Caithness, Northern, Spey and Tweed) with 18 Grade 3 regions. Grade 1 strata for parr were confined to the north (Caithness, Beauly, Kyle_Sutherland and Spey) with Galloway a notable outlier. Regions that achieved an overall Grade 1 for trout were in the north of the country (Beauly, Caithness, Spey).

Figure 16. Trout density estimates (black circles), with associated 2-sided 90% confidence intervals for each of the NEPS regions in 2023. Rows denote lifestage. Green squares indicate the benchmark (expected densities) against which estimated densities can be compared to determine performance grade. Numbers at the base of the plot show the number of samples in the region.

Figure 17. NEPS 2023 regional grades for trout fry, parr and overall.

Inter-annual variability in densities and status

There have been substantial changes in sample frames between NEPS surveys (see methods section above), including the addition of larger rivers with higher benchmark densities. To allow inter-annual comparisons, sites in 5th order rivers or above barriers were removed and strata aggregated to the original NEPS regions before producing density estimates and benchmarks at regional and national scales.

National density estimates and benchmarks are shown in Figure 18. The national density estimate of salmon fry in 2023 was lower than in 2018 and 2021 but higher than in 2019 (Fig. 18). The national density estimates for salmon parr declined year on year and, by 2021, were substantially below the benchmark. However, the 2021 and 2023 estimates were broadly comparable, given the confidence intervals around them.

The national density estimate for trout fry in 2023 suggested a slight improvement on 2019 and 2021 and was broadly consistent with 2018. The upper confidence intervals for 2018 and 2023 included the benchmark, but those for 2019 and 2021 did not. The national density of trout parr in 2023 was substantially below the benchmark but broadly comparable with the densities in 2019 and 2021.

Substantial spatio-temporal variability in abundance across species and lifestages (Appendix 3) was reflected in Grades across regions, species and lifestages over the four NEPS survey years (Figs. 19 and 20, Appendix 4). The predominant pattern is one of decline, although some region, species and lifestage combinations appear stable in their grades across years.

Figure 18. National comparison of juvenile trout and salmon, fry and parr densities between years. Black circles and error bars indicate mean density estimates and 90% confidence limits. Green squares indicate the national benchmark. Data from NEPS 2021 and 2023 were post-stratified to provide a sample frame that was broadly consistent with 2018-19 NEPS surveys. Small differences in the benchmark between years reflect changes to the sample frame that could not be resolved.

Overall Grade 1 salmon regions in 2023 were constrained to the north and south-west (Galloway), with the latter being an atypical instance of improving status (although Galloway is only Grade 2 where all NEPS 2023 data are included). Only six regions had an overall Grade 1 status in all four survey years (Caithness, Beauly, Brora_Helmsdale, Northern, OuterHebrides, WesterRoss_Skye). There were more overall Grade 3 salmon regions in 2023 than in any other survey year. Nine regions were Grade 3 in 2023, compared to 7, 8 and 6 in 2018, 2019 and 2021 respectively. Only the Don and Ugie in the north-east of Scotland had an overall Grade 3 status for salmon in all four survey years indicating that these regions were in a consistently poor state.

The number of overall Grade 1 trout regions declined substantially from 9 in 2018 to 2 in 2023 (with 3 in 2019 and 2021). Only the Caithness and Spey regions had an overall Grade 1 status in all survey years. There were five regions with an overall Grade 3 status in all survey years (Aryshire, Esk, Forth, Lochaber, Ugie).

Figure 19. Inter-annual variability in the status (Grades) of NEPS regions for salmon fry, parr and overall. Data from NEPS 2021 and 2023 were post-stratified to allow comparison across years with common regions and sample frames.

Figure 20. Inter-annual variability in the status (Grades) of NEPS regions for trout fry, parr and overall. Data from NEPS 2021 and 2023 were post-stratified to allow comparison across years with common regions and sample frames.

Inter-annual variability in densities between Strahler river orders

Consistent with benchmark models, salmon densities increased with river order while trout densities tended to decrease (Fig. 21). These patterns are generally consistent across sampling years, i.e. patterns of abundance between river orders are preserved as overall abundance changes. In the case of salmon, there is evidence of a plateau in the mean densities between river order 4 and 5, which provides support for capping upstream catchment areas >250km2 in benchmark density predictions.

Figure 21. Relationships between Strahler river order and mean estimated densities for salmon and trout, fry and parr obtained by post-stratifying the NEPS data. Coloured filled points and error bars indicate different survey years. Green unfilled points indicate the benchmark for each river order in each survey year. Small differences to the benchmark between years reflect changes to sample frame that could not be resolved. Numbers at the base of the plot show the number of samples in the river order. Note that plots contain all data, including strata above barriers on the Forth, Spey and Shin that will reduce mean density estimates for salmon in 2021 and 2023 relative to 2018 and 2019 surveys.

Long-term trends in juvenile abundance and relationships to rod catch

Saturated habitat consistent with Grade 1 status would be expected to be characterised by relatively stable juvenile densities, independent of changing stock levels. This was tested for three Rivers with long-term juvenile time series and contrasting NEPS Grade profiles.

The final spatio-temporal model for the Tweed was:

Count ~ LS + s(RDS) + Altitude + UCA + LS:UCA + offset

When CohortYear was added to the final model, its effect was significant (p < 0.05) and non-linear, initially increasing during the 2000’s, before declining after ca. 2010 (Fig. 22a).

The final spatio-temporal model for the Aberdeenshire Dee was:

Count ~ LS + RDS + Altitude + UCA + Gradient + DoY + LS:Altitude + LS:Gradient + LS:DoY + s(CohortYear) + offset

The non-linear CohortYear temporal trend (p < 0.001) was broadly consistent with that seen for the Tweed: flat in the early years with a decline from ca. 2010 onwards (Fig. 22c).

The final spatio-temporal model for the Spey was:

Count ~ LS + s(RDS) + Altitude + LS: s(UCA) + s(DoY) + %Conifer + offset

There was no effect of CohortYear in the final model and, when added, this was relatively flat, linear and non-significant (p > 0.05) indicating no trend (Fig. 22e).

In the second series of models, where lagged rod catch was an additional candidate explanatory variable, the final model for the Tweed was:

Count ~ LS + s(RDS) + Altitude + UCA + LS:UCA + Rod + offset

This model had the same spatial terms as the original Tweed model, but also included a positive linear effect of rod catch (Rod). When CohortYear was added, its effect was linear and negative, but non-significant (p > 0.05).

The final model for the Dee was:

Count ~ LS + RDS + Altitude + UCA + Gradient + DoY + LS:Altitude + LS:Gradient + LS:DoY + CohortYear + Rod + offset

The spatial effects were again consistent with the original Dee model. However, the model also included a positive linear effect of Rod and a negative linear effect of CohortYear, both of which were significant (p < 0.01).

When data from the Marine Directorate Girnock and Baddoch sites were excluded from the Dee model selection process, there were fewer spatial covariates (RDS, Altitude, UCA), no interaction terms between LS and spatial covariates, and no effect of Rod. However, there remained a highly significant negative linear effect of CohortYear (p < 0.001).

The final model for the Spey was unchanged from the original Spey model with no evidence of a significant CohortYear trend or Rod effect.

These results suggest that the River Spey which was predominantly Grade 1, has relatively constant juvenile densities consistent with saturated habitat. In contrast, the Rivers Dee and Tweed have declining densities that are responsive to adult stock levels as indicated by relationships with rod catch (a proxy for adult abundance). This would be inconsistent with NEPS Grade 1 status.

Figure 22. Modelled temporal trends in salmon fry densities (number per metre squared) in the rivers Tweed (a), Dee (c) and Spey (e), together with rod catches (salmon and grilse, retained and released) for the same rivers. Catches are positively lagged by one year (+1) to match associated fry production. Note that rod catches are plotted on independent scales.

Assessing the status of Special Areas of Conservation (SACs) for salmon

There was a general pattern of poor performance across the SACs (Fig. 23, Table 1). None of the pointwise density estimates for SACs exceeded the benchmark for fry and parr in all years (i.e. > 100%, Fig. 23). Relatively low sample numbers in some SACs resulted in high uncertainty in mean abundance estimates (illustrated by wide confidence limits).

Only four of the SACs were considered to be in Favourable Condition for juveniles (Table 1). None of the SACs received an overall NEPS Grade 1 assessment in every year, with only two (Spey, Naver and Mallart Water) receiving a Grade 1 assessment in three of the four years (Table 1). However, the Naver and Mallart Water was also the only SAC to show an overall “Declining” trend in abundance, with the remaining SACs showing “No Change”. The Moriston received a Grade 3 assessment in two of the four years but was also the only SAC showing a “Recovering” trend (for fry).

Figure 23. Estimated mean densities as a percentage of the benchmark for salmon fry and parr in Special Areas of Conservation (SACs). The grey dashed line denotes 100%; densities below the line are less than the benchmark and those above the line exceed the benchmark. Data are shown for SACs with at least 5 samples in each NEPS sampling year.

* NC denotes “No Change”, R denotes “Recovering” and D denotes “Declining” trends

Relationships between juvenile densities, rod catches and NEPS grades

Across the four NEPS surveys, there were generally positive relationships between rod catch and estimates of mean juvenile salmon densities at the national scale (Fig. 24). An exception is the parr mean density estimate for 2023 which should relate to the 2021 rod catch (assuming parr are primarily in their second year in freshwater). However, this might be because the 2021 fishing season was heavily impacted by the coronavirus pandemic stay-at-home orders and restrictions on national and international travel. Surprisingly, there is no evidence of a similar effect on the 2021 fry data (arising from the 2020 spawner year) where rod catches might have been similarly affected by the pandemic. Trout fry densities are suggestive of a non-linear positive relationship with rod catch, consistent with a stock-recruitment relationship with density dependant effects. In contrast, there is no clear relationship between rod catch and trout parr densities.

Figure 24. Relationship between salmon and sea trout rod catch (retained and released) and national estimates of mean juvenile density of salmon and trout (in Strahler order 2-4 rivers). Colours indicate the NEPS sampling year and symbols the lifestage. Numbers that accompany data points indicate the spawner years expected to be associated with juvenile production (assuming parr are primarily 2 years old, i.e. 1+ parr).

Positive non-linear relationships between rod catch (scaled for wetted habitat) and juvenile density were also evident at regional scales for both salmon and trout (Fig. 25). These relationships were broadly consistent with a stock-recruitment relationship involving density dependence at higher stock levels. In general, and ignoring data limitations, Grade 1 was associated with regions nearing carrying capacity, particularly for salmon. There were some notable outliers. For example, Kyle_Sutherland, Nairn_Findhorn_Lossie and Deveron had low salmon fry densities given their observed rod catches. Furthermore, in the case of trout, the Ugie had both high rod catches (catch / km2 of accessible river wetted area) and relatively high juvenile densities but only achieved a NEPS Grade 3 status.

Figure 25. Relationship between salmon and trout rod catch density (catch / km2 of accessible river wetted area) and regional estimates of mean juvenile density (in Strahler order 2-4 rivers). Symbols indicate lagged spawner years assuming parr are predominantly in their second year in river (1+ parr). Colours indicate NEPS Grades. Black lines show a fitted Ricker stock-recruitment relationship.

Variability in the age structure of salmon parr

Many sites were characterised by multiple parr age classes (i.e. with fish older than 1+), but there was substantial spatial variability across the country (Fig. 26). Higher percentages of >1+ parr were observed in the north, Moray Firth and in the upper reaches of large eastern catchments (e.g. Spey, Tay). Low proportions of older parr were typically observed at lower altitudes and in more nutrient rich areas although no formal analysis was undertaken at this time. Sites in the Tweed and Ugie had no parr older than 1 year old.

Figure 26. Map showing the percentage of older salmon parr (>1+) present within sites. Grey squares indicate that all parr were 1+. Only sites where parr were caught are plotted.

Relationship between fish prevalence and mean density

There was a positive sigmoidal relationship between regional mean fish density and occupancy (percentage of river length where fish were present), for each species and lifestage (Fig. 27). There was substantial uncertainty in both the mean abundance and prevalence estimates (reflected in the wide error bars) resulting in substantial noise in the relationships. The relationships appear stronger for salmon than trout, but there were no clear differences between years. Much of the variability in prevalence occurred at intermediate levels of abundance.

Figure 27. Estimated occupancy (proportion of river length where fish of a given species and lifestage are present) plotted against the corresponding mean density estimates for each region, in each year, with associated regression lines. Error bars indicate the two-sided 90% confidence intervals for the proportion of river length (vertical lines) and mean density (horizontal lines). The density axis has been capped below 0.005 n m-2 for both lifestages and above 1.75 and 0.45 n m-2 for fry and parr respectively. Panels denote the species and lifestage, colours and symbols denote the NEPS sampling year.

Spatial variability in water quality: assessing pressures on salmonids

Spatial variability in water chemistry was broadly consistent across survey years (Figs. 28 - 30). Ammonia and nitrite concentrations were generally low across the country, with the occasional very high values that varied in location between years (Fig. 28). There were more consistent spatial patterns for nitrate and total soluble nitrogen, with the highest concentrations in lowland agriculture areas (north-east and south-west) and urban areas in the central belt (Fig. 28).

Phosphate and total soluble phosphorous exhibited similar spatial patterns to nitrate and total soluble nitrogen, with higher concentrations in the north-east, central belt and south-west (Fig. 29). In contrast, dissolved organic carbon concentrations were highest in the north and south-west (Fig. 29) reflecting variability in soil types and land-use. Variability in geology, groundwater contributions and river residence times also drive strong and consistent spatial patterns in silica, with the highest concentrations in the north-east and lowest concentrations in the north-west (Fig. 29).

Spatial variability in underlying geology, soils and land use was also reflected in pH measurements (Fig. 30). The highest pH values were in the east and the lowest in the north, north-west and south-west. Potassium concentrations were consistent with total soluble nitrogen and total soluble phosphorous, with the highest concentrations in the central belt, north-east and south-west. Chloride concentrations were generally highest in coastal areas (highest on the north and east coasts) and lowest in inland locations. Sulphate concentrations were highest around urban areas (central belt, east).

Figure 28. Spatial variability in Nitrogen and Nitrogen compounds between NEPS surveys. Columns relate to NEPS survey years (2018, 2019, 2021, 2023), rows relate to chemical determinands. All concentrations are in parts per million (ppm)

Figure 29. Spatial variability in Total Phosphorous, Phosphate, Dissolved Organic Carbon and Silica between NEPS surveys. Columns relate to NEPS survey years (2018, 2019, 2021, 2023), rows relate to chemical determinands. All concentrations are in parts per million (PPM)

Figure 30. Spatial variability in pH, Potassium, Chloride and Sulphate between NEPS surveys. Columns relate to NEPS survey years (2018, 2019, 2021, 2023), rows relate to chemical determinands. All concentrations are in parts per million (ppm) except for pH.