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

Discussion

Salmon and sea trout populations are in long-term decline across much of their native range. In Scotland, the protection of wild salmonids, their habitats, and the ecosystem services that they provide, including fisheries, is central to a wide range of national strategies including the Blue Economy Vision for Scotland, Scottish Biodiversity Strategy, Scottish Wild Salmon Strategy and the Vision for Sustainable Aquaculture. NEPS provides a robust quantitative, catch independent and fully scalable (site to national) assessment of the status of Scotland’s salmonid populations. When combined with information on pressures, NEPS provides the evidence base required to support the conservation of salmonids by identifying under-performing populations and informing management actions that appropriately address population bottlenecks and environmental pressures.

There is an increasing imperative to harmonise data collection and reporting across the public sector with a desire to “collect once – use many times”. Electrofishing data are collected for a wide range of purposes including local fisheries management, Site Condition Monitoring (Habitats Directive), Water Framework Directive classification, and status reporting by National Park authorities. In most of these cases, there is a requirement to assess both status and trends in salmonid populations. By collecting data in a robust framework that includes appropriate data collection methods, statistical survey design, and modelling, it is possible to meet multiple objectives with a single survey.

NEPS 2023 employed a new survey design that improved characterisation of the sample frame (i.e. the rivers that can be sampled by wading and electrofishing) and created new strata for SACs and index monitoring sites. Post-stratification and aggregation of strata among surveys allowed comparison of assessments across years, although this greatly increases the complexity of analyses.

Consistent with previous years, a capture probability model was used to correct and harmonise density data collected across a wide range of habitats, regions and organisations. Temporal trends in capture probability and the introduction of new electrofishing teams emphasised the need for continued collection of multi-pass data and regular updates to the capture probability models used in juvenile assessment.

The NEPS 2023 survey showed that at the national scale, salmon fry and parr densities were both significantly below the benchmark indicating poor overall status. While salmon fry densities varied substantially across years, estimates of parr density have consistently declined across the four NEPS surveys since 2018. Although there was considerable spatial and temporal variability in the status of NEPS regions across surveys, the number of overall Grade 3 regions in 2023 (9 of 27) was the highest so far.

Incorporation of a new trout benchmark allowed assessment of the status of trout for the first time. At the national scale, densities of trout parr declined sharply between 2018 and subsequent surveys. Although 2023 was a little better than 2019 or 2021, parr densities remained substantially and significantly below the benchmark. Fry densities also dropped between 2018 and 2019, but have since improved. The number of overall Grade 3 regions for trout increased from 12 in 2018 to 13 in 2023. Only two regions remained Grade 1 for trout (Caithness and Spey), declining from nine in 2018.

A range of approaches were used to assess whether NEPS Grades were broadly consistent with expectations in terms of population processes and reference points. Decadal trends in juvenile salmon densities, and relationships between rod catch (as a proxy of adult abundance) and juvenile densities were examined for selected rivers indicative of different prevailing NEPS Grades. NEPS juvenile densities were also plotted against rod catch at a regional scale and fitted with an indicative Ricker stock-recruitment curve. Taken together these data and analyses suggest that the NEPS Grade 1 is broadly consistent with saturated freshwater habitats.

Where mean salmon densities were low, salmon were absent from an increasing proportion of the river network. This increase in relative inter-site differences in abundance potentially creates challenges for obtaining precise estimates of mean abundance. Fish prevalence could be an alternative and complementary metric of population health, but would require new benchmarks for assessment purposes.

These issues are discussed in more detail below.

Juvenile assessment within the wider evidence landscape

Evidence based freshwater fisheries management requires a reliable understanding of the processes regulating fish populations, the status (health) of populations across a range of spatial scales, the environmental pressures acting on them (including relative importance and effect size) and the efficacy of management actions. Different data sources and analysis frameworks have varying strengths, weaknesses and capabilities in providing this evidence base. Consequently, most countries maintain some combination of geographically widespread adult and juvenile data collection and assessments, combined with a small number of intensively studied index monitoring sites where population characteristics are measured across multiple lifestages. In some cases, these data are integrated in complex modelling and assessment frameworks (e.g. Kuikka et al., 2014; ICES, 2024), In others, the data are used independently and for particular purposes, but provide complementary information.

Information on adult numbers is typically obtained from commercial or recreational fisheries. Adult exploitation data are frequently collected and reported in national statistics. They benefit from low cost of data acquisition (as a by-product of fisheries), broad geographic coverage, and individual river resolution. A central challenge in estimating reliable population sizes from these data is that exploitation rates can be highly variable between rivers and over time (Gregory et al., 2023). This is typically addressed by modelling spatio-temporal variability in exploitation rates from a small number of catchments containing fish counters (which are assumed to provide an accurate counts of adults) and then predicting across rivers without counters (Gregory et al., 2023). The resulting estimates of adult numbers can then be incorporated into assessment methods for application in national (e.g. Conservation Regulations) and international management contexts (e.g. ICES), primarily to regulate fisheries.

Juvenile electrofishing data can provide accurate, catch independent estimates of the numbers of juvenile fish in a section of river, provided they are collected and analysed using appropriate methods (Malcolm et al, 2023; Alexandre et al., 2025). By sampling multiple sites, it is possible to scale abundance estimates to any spatial scale of interest (Isaak et al., 2016; Glover et al., 2018, Malcolm et al., 2023). Except in a small number of locations with counters or traps, juvenile methods are currently the only approach available for assessing the status of salmonid populations at sub-catchment scales. When combined with suitable genetic data and tools this could provide a basis for assessing the status of different stocks, for example early or late running salmon (e.g. Cauwelier et al., 2024). In the context of brown trout, a recent review by members of the ICES sea trout working group (Alexandre et al., 2025) identified juvenile methods as the most promising approach for assessment. When combined with information on environmental pressures (e.g. water quality, genetic introgression, habitat), which can be collected at the time of sampling (Malcolm et al., 2023), juvenile data provide the evidence base necessary to prioritise (and potentially evaluate) management actions. Because juvenile data are catch independent, they can be particularly useful where fisheries are closed to protect stocks from over exploitation (White et al., 2023) or where there is a risk of increased mortality under high temperature and low flow extremes (Van Leeuwen et al., 2023). However, juvenile data can be logistically challenging and costly to obtain, requiring collection by experienced fisheries biologists within a carefully considered statistical and analytical framework such as NEPS. Consequently, it is challenging to provide adequate coverage of all salmon rivers on a regular basis, especially in countries or regions where there are many small rivers.

Index monitoring sites typically collect information on fish numbers, sizes and ages at multiple lifestages. These detailed data allow biologists to understand population processes, bottlenecks and compensatory processes in a changing environment (Soulsby et al., 2024). They are also used to construct stock-recruitment relationships for the critical freshwater lifestages where population regulation (density dependence) occurs, free from the noisy and trending density independent marine phase (Gurney et al., 2010). These stock-recruitment data are essential to parameterise and/or validate assessment methods and to interpret broader scale monitoring data (Alexandre et al., 2025). Wider process understanding is essential to understand the factors limiting and controlling populations and the likely efficacy of different management options in different contexts.

A combination of data and modelling approaches is therefore required to understand and manage salmonid populations, depending on the spatial scales of interest and objectives. Regular and carefully managed large-scale programmes of juvenile monitoring have an important and central role in this context that is complementary to other data sources and cannot be addressed by other data or methods.

The NEPS 2023 Survey Design

The NEPS 2023 survey design was intended to build on the experience of previous survey designs by 1. improving the spatial representation of the target population (i.e. sample frame) to reduce oversample requirements and increase the representation of larger (> Strahler order 4) rivers that can be sampled by wading and electrofishing 2. incorporating local reporting requirements where these could be supported with additional resource 3. allowing for flexible resource allocation to improve characterisation and classification of SAC rivers should additional resources become available in future years 4. integrating detailed and intensive monitoring efforts from index monitoring sites into the wider NEPS monitoring framework 5. ensuring adequate sample and oversample availability to allow surveys to operate over an extended duration (initially a period of 9 surveys).

The new survey design was successful in achieving these objectives. However, changes to the sample frame and the additional complexity of new strata created challenges for data analysis. This was particularly true in regions where the sample frame had changed substantially between years (e.g. to include larger fifth order rivers) and there was a need to compare abundance estimates across surveys. In these circumstances it was necessary to post-stratify the later surveys to exclude new rivers that had been added, and to combine strata to match NEPS regions originally defined as strata in 2018/19. This added substantially to the coding and effort required for reliable reporting. The additional work will need to be repeated each time there is a new survey and a requirement to compare with previous years. Major benefits could therefore be accrued from maintaining a consistent survey design over multiple years, providing that all the reporting requirements were specified and understood from the outset.

Although the new survey design allowed greater flexibility in the allocation of resources to SAC’s, there was no additional resource available at the time. Consequently, sample densities for 2023 were largely based on previous years and scaled for river length where regions were divided into multiple strata. This resulted in too few samples to obtain reliable assessments of status in some of the smaller SACs.

The integration of index monitoring programmes into the NEPS survey had several benefits. The most immediate was reduced reporting requirements, with a single analysis providing data for multiple programmes of work. However, the detailed monitoring data also supplement observations from the core NEPS survey increasing confidence in broader catchment or regional assessments. Furthermore, integration of index sites allows the juvenile assessment data to be interpreted in the wider context of adult, emigrant and stock-recruitment datasets provided by the index sites. This provides a basis for independent validation of assessment benchmarks or opportunities for scaling the juvenile benchmark relative to other commonly applied fisheries management targets or limits e.g. maximum sustainable yield (MSY)

Although significant efforts were made to identify and accommodate the reporting requirements of different partner organisations at the time of development, new monitoring and reporting requirements have since emerged. The most demanding of these relates to the developing Sea Lice Risk Assessment Framework (SLRAF) and associated regulatory framework being implemented by SEPA. This would require an extension of the sample frame to include the Northern Isles (Orkney and Shetland) that could be accommodated by a stand-alone survey or a new national survey design. Other recently identified end users of juvenile assessment data include the Cairngorms National Park (CNP) which requires robust data on the status of Atlantic salmon within the park area to support the Cairngorms Nature Index (CNI). The CNI will be used to report on the status of biodiversity within the park to the Scottish Government and assess ecological change within park boundaries. While the reporting requirements of CNP could be met from the NEPS 2023 survey, additional work would be required to post stratify data and report at relevant spatial scales.

Despite the value of the NEPS programme for multiple requirements and organisations, funding a full national programme remains challenging. In 2018 the programme was funded primarily by the Scottish Government, Marine Scotland (the predecessor to the Marine Directorate) with contributions from Nature Scot and SEPA. In 2019 and 2021 the data collection programme was funded by Marine Scotland and Crown Estate Scotland. In 2023, the data collection programme was funded by the Marine Directorate alone. In all years, project management, data analysis and reporting have been funded by the Marine Directorate. In the absence of an identified single central source of funding for future surveys there will be a requirement for a partnership approach to data collection, with different organisations funding different components of work either directly or in kind. A carefully designed national survey design, similar to NEPS 2023 but including the Northern Isles, could meet the monitoring and reporting needs of all stakeholders identified to date. However, future analyses of trends are likely to be challenging in the absence of national coverage and regular sampling.

Capture Probability (P)

Capture probability models can harmonise data collected across different habitats and regions, by different teams of people using different equipment, and ensure unbiased and relatively precise estimates of abundance. Failure to account for variability in capture probability can result in spatial (Millar et al., 2016) and temporal (Dauphin et al., 2018; Glover et al., 2019) biases. The capture probability model from NEPS 2023 is broadly consistent with those reported from previous surveys (e.g. Malcolm et al., 2019, 2020 and 2023). Trout were more catchable than salmon, parr more catchable than fry, and fish on the first pass more catchable than subsequent passes, potentially reflecting a combination of different behaviours and fish size. Capture probability also varied substantially between habitats, with lower Altitudes, higher UCA and Gradient being associated with poorer capture probability.

Previous national analyses (Malcolm et al., 2020 and 2023) have shown that capture probability varies over time, with a positive linear effect of year. The most recent model suggests a non-linear positive effect that has started to plateau in recent years. This indicates that further improvements in capture probability are unlikely to be sustained over the longer term, especially where these are driven by factors such as improved equipment or training. However, there could still be increases due to factors such as increasing fish size at age in response to reduced competition or increasing temperatures (Gurney et al., 2008). Nevertheless, appropriate accounting for trends in capture probability remains essential for assessing trends in density and for subsequent assessments of status (Glover et al., 2019; Dauphin et al., 2018) and this requires multi-pass data to be collected each year.

There remain substantial differences in capture probability (and associated uncertainty) between organisations and teams. Uncertainty was generally greater where fewer multi-pass electrofishing data were available for a particular Organisation-Team or where multiple teams were deployed by the same organisation without suitable harmonisation of procedures, effort and equipment. This can be a particular problem where new Organisation-Teams contribute to NEPS and suggests that new teams should fish a greater proportion of sites using multi-pass protocols to reduce uncertainty in their estimate of capture probability. As the NEPS datasets grow, it may become possible to separate the effects of individual teams within organisations in a particular year using anode operator as an additional predictor variable. However, there is insufficient multi-pass data to do this at present and the approach followed here is therefore pragmatic.

Interpretation of NEPS Benchmarks and Grades

The NEPS juvenile salmon assessment benchmark was published in 2019 using multi-pass electrofishing data collected across Scotland between 1997 and 2015 (Malcolm et al., 2019). The benchmark was intended to represent an achievable target, indicative of the expected abundance of fish for a particular habitat and location, assuming adequate spawners to saturate freshwater habitat and the absence of substantial anthropogenic pressures. The benchmark was derived from a habitat-abundance model that excluded regional predictors (thought to represent an anthropogenic pressure gradient in the vicinity of the central belt) and the negative effects of commercial conifer. However, it was not possible to independently validate or calibrate the benchmark using stock-recruitment data, as only two locations in Scotland (the Girnock and Baddoch on Aberdeenshire River Dee) had the multi-lifestage datasets and range of stock levels necessary to derive such relationships.

Despite a lack of suitable index monitoring sites to validate the NEPS benchmark and Grades, other data sources and approaches can be explored. Through a strength of evidence approach it is possible to consider whether NEPS Grade 1 assessments are broadly indicative of saturated freshwater habitats (i.e. carrying capacity or maximum production). Where freshwater habitats are saturated, juvenile densities would be expected to be independent of changing stock levels (spawner numbers) or their proxies (e.g. rod catch). Depending on when population regulation occurs, this could apply to fry (0+), parr (>0+) or both. For example, recent work at the Girnock Burn index monitoring site suggests that fry production is density independent (increases linearly with ova deposition), while parr production exhibits strong density dependence at high stock levels, where recruitment becomes independent of stock (Glover et al., 2019; Soulsby et al., 2024).

At the national scale, salmon fry densities appear to increase linearly with rod catch. Salmon parr densities were less clearly related to rod catch, but these relationships were obscured by the 2021 spawner year which was heavily affected by covid restrictions. There was also evidence of a stock-recruitment relationship between sea trout rod catches and trout fry densities. At the regional level, there is evidence of stock-recruitment relationships between spawner density (as indicated by rod catch density) and juvenile densities. In these relationships, Grade 1 appears to be broadly consistent with maximum juvenile production, although the relationships have substantial noise. This is unsurprising given the variation in exploitation rate that likely exists between rivers and over time (Gregory et al., 2023), and the differences in carrying capacity between freshwater habitats (Malcolm et al., 2019) that are ignored with this approach. Nevertheless, the available data suggest that the NEPS benchmark may be close to, or slightly below habitat saturation (or Rmax).

An alternative approach for validating the NEPS benchmark involves exploring temporal trends in juvenile abundance, and relationships between rod catch and juvenile abundance, for rivers with contrasting NEPS Grades and Grade profiles (changes over time). The River Spey was consistently Grade 1, or close to Grade 1 across NEPS survey years. The available juvenile data suggest no significant trends and no relationship with rod catch. This is consistent with a saturated freshwater habitat. In contrast, juvenile densities in the Tweed and Dee both declined significantly over time. Both catchments also exhibited positive relationships with rod catch. This suggests the Tweed and Dee are no longer saturated in terms of freshwater production (at least in some NEPS years) and is consistent with a NEPS Grade 2 or 3.

It has been suggested that low and declining parr densities in some regions could reflect changes in age at emigration, with increasing proportions of parr migrating before the NEPS summer census. This could be particularly problematic where rapid growth results in some juveniles migrating as 1+ parr, since 1+ parr are always a large proportion of the total parr production. However, it is hard to address this issue in detail without time-series observations of juvenile and emigrant numbers, sizes and ages (Gurney et al., 2008; Soulsby et al., 2024). In the case of the Tweed and Dee there is strong evidence that juvenile densities have declined over time. There is also evidence that some of the reductions in juvenile abundance can be explained by declining adult abundance (as indicated by rod catch). However, detailed observations from the Girnock and Baddoch index monitoring sites suggest that salmon are growing more rapidly and emigrating earlier in some circumstances (Gurney et al., 2008; Soulsby et al., 2024). This could represent a compensatory response to reduced competition for resources, or a change resulting from increasing temperatures, or both. In the former case, the NEPS benchmark would remain appropriate, as higher ova deposition would result in higher competition, lower growth, a wider range of parr ages and thus more robust populations. The latter could indicate shifting baselines and thus a benchmark which may no longer be attainable under a rapidly shifting climate. However, there is currently insufficient evidence to suggest that the NEPS benchmarks are too stringent or increasingly unattainable relative to the historic situation.

Status of Scotland’s salmon and trout populations

At the national scale, salmon fry and parr densities were significantly below the benchmark in 2023, with estimates of parr densities continuing a decline over time. The number of regions with an overall Grade 3 status also increased to its highest level across NEPS surveys, although the spatial pattern of grades varied considerably across years. This decline in status is broadly consistent with recent rod catch data and suggests that Scotland’s salmon populations are no longer fully occupying available freshwater habitat, with consequences for future smolt production and adult returns.

In general, the worst performing regions in 2023 were in the central belt and north-east, a pattern consistent with previous NEPS surveys. These regions are affected by a range of pressures including urbanisation, agriculture and legacy effects of historic land uses. These effects may depress juvenile production over and above the common effects of poor marine survival. The best performing regions continue to be found largely in the north and the Moray Firth (Spey). Rivers in these regions flow through some of the wildest lands in Scotland (Carver et al., 2012, Strus and Carver, 2024) and are likely to be impacted by fewer and less severe pressures than elsewhere.

In some circumstances, assessments of salmon status were affected by the spatial scales of analysis, with the performance of individual strata varying within wider regions. To some extent this will reflect genuine spatial variability in performance of difference rivers or stocks (e.g. early running versus later running fish). However, it could also reflect the effects of smaller sample sizes at finer spatial scales and thus greater uncertainty in assessments. There are therefore clear trade-offs in terms of reporting scales, sample sizes and confidence of assessment. These issues are not unique to juvenile assessments. For example, catch data can be limited for smaller rivers and affected by highly variable exploitation rates (e.g. Gregory et al., 2023), and adult assessments of individual stocks could also be challenging, requiring assumptions on run timing, spatial structure of stocks and the carrying capacity of habitats in different parts of the river.

This report provides the first assessment of juvenile trout. This suggests that at the national scale juvenile populations in 2023 were not in good status. This was particularly so for parr (Grade 3) where densities declined substantially between 2018 and subsequent surveys. At the regional level only the Spey, Ness (depending on the sample frame used for analysis) and Caithness remained at Grade 1 overall in 2023, although analysis at the finer strata scale also suggested healthy trout populations in the River Laxford, within the wider West Sutherland region. When regional assessments were compared across years the overwhelming impression was of declining status between 2018 and subsequent years with a large reduction in the number of Grade 1 regions. This rapid reduction in performance against benchmark is broadly consistent with sea trout rod catches at the national scale. Sea trout rod catches, reported since 1952, have declined over much of this period (Scottish Government, 2024, Adams et al., 2022). NEPS surveys have thus only been undertaken at relatively low population levels. However, the 2016 and 2017 rod catches, that would influence fry and parr production in 2018 were higher than the subsequent years. The observation of coincident changes in sea trout catches and juvenile abundance is consistent with studies elsewhere that suggest that anadromous trout can dominate juvenile trout production in accessible rivers (Goodwin et al., 2016; Rohtla et al., 2017).

Spatial distribution and abundance of juveniles in declining salmonid populations

Rod catches of salmon and sea trout (as a proxy of abundance) suggest widespread and long-term declines in adult abundance (Scottish Government, 2024). This is supported by catch independent data from counters (Thorley et al., 2005), traps (Soulsby et al., 2024) and electrofishing (Glover et al., 2019; this report). At high levels of adult abundance (and thus ova deposition) freshwater production of juveniles and emigrants will be largely independent of changes in adult abundance due to density dependent competition in freshwater where population regulation occurs (Gurney et al., 2010; Glover et al., 2020; Soulsby et al., 2024). As populations decline, ova deposition can be insufficient to fully stock freshwater environments and juvenile production will decline as a result, leading to negative population feedbacks (Soulsby et al., 2024). Reductions in juvenile production could manifest themselves in a number of different ways including changes in prevalence, species distribution or patterns of abundance.

Positive relationships between population abundance and occupancy have been observed for a number of species including plants, butterflies, fish and birds (Gaston et al., 2000). In this study positive relationships were observed between mean regional abundance and occupancy, potentially reflecting spatial variability in habitat quality between regions, and changes in species abundance between survey years. The relationships were strongest at intermediate population densities but offered little information at lower or higher densities. It has been suggested that prevalence could provide an alternative metric of population health that is potentially easier to characterise than abundance (Gasten et al., 2000). The current study suggests that prevalence is a useful supplementary metric of population health, but doesn’t offer the full range of population characterisation offered by abundance data. Furthermore, because occupancy is positively related to population abundance some habitats (sections of river) will naturally have lower levels of occupancy than others. Appropriate interpretation of occupancy data would therefore require a new benchmark model to be developed indicating the probability of fish presence for particular habitats given “healthy” fish populations.

Reductions in population size can also alter spatial patterns of abundance. For example, density dependant habitat utilisation could result in only the high-quality habitats being occupied, with expansion into lower quality habitats only when local abundance compromised fitness in high quality areas. Such a process results in disproportionate changes in density across the network (MacCall, 1990; Gibson et al., 2008; Avgar et al., 2020). For salmon and trout this could be manifested in remaining populations concentrating in high quality habitat areas at the centre of the range (larger rivers at lower altitudes in the case of salmon) and contracting from less favourable habitat (i.e. small rivers at higher altitudes).

At present there is insufficient information from NEPS to determine how spatial patterns of abundance are affected by population decline, particularly given the many regional differences in performance. However, proportional relationships in fish density between Strahler River Orders appeared broadly consistent across different stock levels (years), at least at the national scale. Furthermore, there was strong evidence that fish prevalence declined with mean abundance (both between regions and over years). This suggests that relative inter-site differences in density will increase, especially in low density regions of the network. It also suggests that increasing areas of the river network (lower abundance areas) will no longer support detectable fish densities (i.e. < 1 fish / 100m2). It remains unclear whether densities will decline in proportion to habitat quality (as indicated by benchmark densities), or whether marginal areas of the network will decline more rapidly, consistent with density dependent habitat use, or the basin theory (MacCall et al., 1990). However, in either case the result may appear as a contraction in species range.

Understanding the processes by which freshwater habitat utilisation responds to changing population strength has important consequences for survey design and for assessment, including the interpretation of time-series data collected from ad-hoc sites. However, there are few locations with detailed population monitoring, long-time series and a large range of stock levels necessary to explore these mechanisms (e.g. Glover et al., 2018). Future work should therefore focus on understanding processes at index monitoring sites such as the Girnock and Baddoch and look to extrapolate these findings more widely.

Future work

Although the NEPS sampling and assessment framework is now well established, there remain several areas where NEPS could offer greater value or where further research and development could increase confidence in assessments.

The current report explored the potential of presence – absence data (also known as prevalence or occupancy) to complement existing abundance data for assessing the status of juvenile salmonids. Prevalence was positively correlated with abundance. However, this relationship reflected both natural variability in habitat quality between regions and changes in abundance between surveys. Interpretation of prevalence data would thus require a new benchmark model that predicted regional patterns of prevalence under healthy (benchmark) conditions. This would likely vary between rivers and regions in a similar way to existing abundance benchmarks. Given the limited additional information provided by the prevalence models over existing benchmark models, this is unlikely to be an immediate priority for further work.

NEPS was initially developed to provide an assessment of the status of juvenile salmon. In the current report this was extended to include trout. However, data are collected on all sampled species. This strategy was deliberately chosen to allow future developments and maximise the value of data collection for multiple purposes. An obvious development would be to extend the assessments to include European eel. Scotland has historically relied on data from a few index monitoring sites to assess the status of eels and to meet international reporting requirements (Anon, 2024). In the future these requirements could be met through an extension of the NEPS analysis, or the use of NEPS electrofishing data within existing modelling frameworks (Briand et al., 2022; Hohne et al., 2023). The latter could be facilitated through greater use of common, open access spatial rivers datasets as discussed by Alexandre et al (2024) in the case of sea trout.

When the NEPS programme was established in 2018, the primary aim was to provide an unbiased, spatially balanced sample of juvenile Atlantic salmon that could be used to assess the status and trends in fish populations, complimenting adult-based assessment methods being developed to support Salmon Conservation Regulations. Since then, there has been increasing interest in using the NEPS electrofishing data to assess the status of sites within river systems and the pressures affecting them. In some cases, this has led to the over interpretation of data from individual sites. When the NEPS salmon benchmark was developed, the focus was on assessing regional performance from many sites using a scalable benchmark that could be obtained from spatial data alone (Malcolm et al., 2019). This approach is valuable where there is a requirement to scale benchmarks across spatial scales, or where historical habitat data were not collected to common standards at the time of sampling (Alexandre et al., 2024). However, these models also retain substantial unexplained inter-site variability since they do not explicitly incorporate site-specific habitat characteristics such as substrate size, depths, velocities or water quality.

Since 2018 the NEPS programme has collected data on substrate, flow type and water quality to a common standard creating a powerful new dataset. Provisional analyses of these data indicate that the inclusion of site-specific data can improve models of abundance and identify relationships between fish densities and pressure gradients (e.g. high nutrient concentrations). However, the resulting models can also be hard to interpret. For example, salmon densities vary strongly with electrical conductivity (a measure of dissolved material in a water sample) and this in turn varies naturally with factors such as geology. However, it also varies with anthropogenic factors such as land use and discharges from domestic and industrial sources. It can therefore be challenging to set appropriate reference values for particular sites and rivers from which to obtain benchmark predictions. Similar challenges exist where local channel morphology has been artificially altered. It is for these reasons that many previous models of habitat potential have focussed on large scale spatial patterns rather than local characteristics (Firman et al., 2011). Future work is therefore required to consider approaches for incorporating a greater range of local habitat variables in site-wise assessment tools and for setting appropriate reference values in heterogeneous landscapes. Such models could complement existing regional (scalable) benchmarks and provide a strong basis for managing pressures where there are clear anthropogenic gradients.

The NEPS survey design permits the collection of unbiased, spatially balanced data on fish abundance, habitat and the pressures affecting freshwater fish populations. The breeding of farmed Atlantic salmon with wild salmon populations and resulting introduction of non-native genetic material (introgression) is considered to be a substantial pressure on wild salmon in many countries (Forseth et al., 2017). In Scotland, there were limited data on which to assess the potential impacts of introgression until NEPS. Initial analyses focussed on site-wise classifications of impact (Gilbey et al., 2021). Current work (Gilbey et al., in prep.) improves on these earlier efforts by integrating the NEPS GRTS survey design with site-wise estimates of introgression to estimate mean levels of introgression across regions. Future work should integrate estimates of introgression and abundance to estimate the proportion of freshwater salmon production affected by introgression. This would represent a significant step forward in assessing spatial variability in introgression pressures.

Current NEPS analyses and reporting have focussed on regional reporting of status, although the status of individual rivers has also been reported through online applications (scotland.shinyapps.io/sg-national-electrofishing-programme-scotland/). Recent development of genetic tools has facilitated an improved understanding of the distribution of stocks within rivers (Cauwelier et al., 2017; 2024a; 2024b). In the future it may be possible to integrate knowledge of stock distribution and juvenile abundance to assess the status of different run-timing groups within rivers using juvenile approaches. However, a lack of baseline data on historic salmon stock distributions could make such objectives challenging, especially where the habitat is changing rapidly under climate change with varying effects for different stock components.

NEPS and other national monitoring and modelling projects such as the Scotland River Temperature Monitoring Network (SRTMN) depend heavily on spatial data, in particular the Centre for Ecology and Hydrology (CEH) Digital River Network (DRN). Significant work was undertaken by Marine Directorate staff to correct and improve this product for internal use over a period of many years. This involved ensuring that the network was topologically correct (without breaks and “flowing” in the right direction) and adding spatial covariates to line features (e.g. estimates of altitude, upstream catchment area etc.). Model predictions have also been added to the DRN. In the case of NEPS this includes the salmon and trout, fry and parr assessment benchmarks. When NEPS was first developed (ca. 2016) the CEH DRN was the best data product for Scotland’s rivers. However, use of this commercial product prevents outputs being made open access. More recently, new open access datasets have become available that could resolve some of these problems and improve international collaboration (Alexandre et al., 2025). However, some of these are insufficiently detailed (e.g. do not include smaller second Strahler order rivers), have not been topologically corrected (fine for mapping, but not modelling) or lack the necessary covariates that Marine Directorate staff have added to the CEH DRN. Additionally, Marine Directorate staff have developed substantial amounts of R code to work with the internally corrected version of the dataset. In the future it would be desirable to migrate to an open access data product thereby making modelling outputs and products available publicly and open access. However, this is a major undertaking that would involve substantial staff resource and include major rewrites of existing code and refitting of benchmark models, as different spatial datasets result in different covariate values.