Background & Summary

Infestation of salmonids by marine ectoparasitic copepods, commonly referred to as “sea lice”, remains a significant challenge to sustainable Atlantic salmon aquaculture1. In the case of farmed hosts, significant levels of infestation can lead to reduced growth, poor feed conversion efficiency, increased stress and in extreme cases, mortality2. In locations where significant stocks of wild salmonids are present, a clear understanding of sea lice infestation patterns in both wild and farmed salmon populations is of paramount importance due to potential spill-over effects, which has been identified as a conservation issue in a number of countries3,4,5,6,7,8.

These challenges have been identified over the past two decades in British Columbia and a wide range of studies have explored both the patterns of infestation9,10, the potential linkages between farmed and wild populations11,12,13,14, and the likely effects of these interactions15,16. It is not the purpose of this paper to engage with this broader set of debates, other than to note that in a number of these studies general claims are made about the situation for the whole of the BC coast based on rather limited datasets, either in terms of spatial coverage or temporal range. Papers that cover a wider spatial and/or temporal range, illustrate the range of variation that is typically found when considering infestation patterns at these larger scales9,17.

In some cases, data reported within these studies have been made available, or make reference to open access sites (e.g. from government reports18 or from NGO groups19). However, due to the variety of groups involved in sea lice monitoring, as well as changes in practice over time even within the same group, there tend to be differences in the formats of these data. In the case of farmed data, for example, the sea lice data on the official Fisheries and Oceans Canada (DFO) site noted above consisted of monthly mean values when reporting started in 2010, but switched to weekly means in 2013; over time there was also a shift from farm-level averages to pen-level data. Similar differences can be observed over time in the data reported on the websites maintained by the various aquaculture operators in BC. In the case of data reported from observations on wild Pacific salmonids the differences are even more marked. For example, in the early data reported here, limited details exist relating to sea lice species, particularly for parasites at an early stage of development. Where fish are observed in the field, as opposed to those which were lethally sampled and sent to a laboratory for identification of sea lice infestation, there are inevitable limits to the level of detail that it is possible to provide in terms of the sea louse species and/or stages that may be present on a fish.

One of the key objectives when creating this dataset was to find data formats that allowed for maximal temporal and spatial comparison while maintaining as much detail as was possible from the original datasets. It is our view that the debate around salmon farming policy in British Columbia has sometimes used small-scale or anecdotal findings to make overly generalised statements about the sector. The long-term dataset reported here can be used to support a clearer and more comprehensive understanding of sea lice infestation patterns along the BC coast over the past two decades. In addition, recent government policy has already led to a significant reduction in the number of salmon farms in at least two areas covered by this dataset. Exploring the impact of such policy decisions represents another important use case for these data.

Methods

Geographic extent

The data come from locations along the coast of BC, Canada. The responsible regulatory body, Fisheries and Oceans Canada (DFO), has historically used nine ‘fish health’ sub-zones to divide up this coastal region20, based on two larger salmonid fish health zones (Zone 2: Vancouver Island, and Zone 3: Mainland Coast). Within Zone 2 only two sub-zones (2.3 and 2.4) have active Atlantic salmon fish farms, while in Zone 3 all five sub-zones (3.1 to 3.5) have active farms. These seven sub-zones are shown in Fig. 1 and present a summary of the locations from which the data in this dataset were collected. The open triangles represent the location of Atlantic salmon farms from which data have been included, while a solid circle represents each location at which observations of sea lice on wild Pacific juveniles were made. Information panels are provided for each sub-zone, indicating the total number of records and temporal extent of data from farm and wild sea lice monitoring activities that have been included in this dataset. As can be seen from these summary panels, the temporal coverage of data from salmon farms has remained relatively consistent over the past two decades, while sampling from wild populations has varied over space and time. Those familiar with the most recent changes in the aquaculture sector in BC will not be surprised to see declines since 2021/2022 in the number of “farm events” (sea louse sampling events) being reported from the Discovery Islands (3.2), as well as the Broughton Archipelago (3.3), as a consequence of the reduced number of Atlantic salmon farms operating in those areas.

Fig. 1
figure 1

Map of the BC coastal area showing the location of farms sites (open triangles) and wild sampling sites (closed circles) from which data were obtained over the period 2001 to 2023. The panels for each fish health sub-zone illustrate the proportion of data from that zone obtained in each year from 2004 onwards from both farmed and wild sources. (The numbers of farm sampling events and wild fish observed are shown in the upper and lower panels for each sub-zone).

Full size image

Data sampled on wild populations

The dataset relating to wild observations is based on samples taken during the out-migration of wild juvenile fish from rivers to the Pacific Ocean stretching from March to July each year, but with the majority of observations (well over 90%) taking place in April, May and June (see Table 6). As can be seen from Table 1a, by far the most commonly observed wild Pacific species were juvenile chum (Oncorhynchus kata) and pink (O. gorbuscha) salmon, which between them account for around 93% of all fish sampled. Three-spined sticklebacks (Gasterosteus aculeatus) account for a further 3% of the observations, while each of the other Pacific salmon species typically represent little more than 1% of the samples (with the exception of 2003, when an anomalously high proportion of coho, chinook and “other” species were present in the dataset, due to a different sampling protocol followed by DFO in that year).

Table 1 a Total number of sampling events and wild fish assessed for sea lice infestation on the BC coast by host species in each of the sampling years. b Total number of wild fish assessed for sea lice infestation on the BC coast as represented within the various programmes active over the range of sampling years.
Full size table

It can also be seen from Table 1a that the number of fish sampled varied over the years, ranging from over 6,500 to almost 40,000; the exception being 2001 and 2002 where many fewer fish (primarily pink salmon) were observed, from just one of the sampling programmes. In general, these variations are due to the fact that different monitoring programmes existed over this time period. More information relating to each of the programmes listed in Table 1b is provided in the notes that accompany the data set, but these broadly fell into one of two main categories: ‘non-lethal’, where the fish are observed in-situ at the sampling site, and ‘lethal’, where the fish are retained at the sampling site and sent to a laboratory for assessment. Details on the specific protocols can be found both for non-lethal sampling16,21 and for lethal sampling9,15,22. The sampling programme with the longest temporal extent is that carried out by the Salmon Coastal Station (“SCS” in Table 1b), though this is limited to three sites in a single zone (3.3: Broughton Archipelago). The research programme co-ordinated by Dr. Martin Krkošek (“MK” in Table 1b) between 2003 and 2009, contributed a significant number of samples but was again limited to the Broughton. A more limited number of wild fish were sampled by the Cedar Coast Field Station (“CC” in Table 1b) in the Clayoquot Sound (sub-zone 2.3) between 2018 and 2021. All of these programmes largely adopted a similar ‘non-lethal’ protocol. Most of the remaining programmes adopted the ‘lethal’ protocol, with sea lice identification taking place in a laboratory, and were sponsored by local first nations (“Kit”), the Canadian government (“DFO”), the BC aquaculture sector (“MERP”, “MBC”, “Pacif”), or through co-operation among these groups (“BAMP”). The final programme integrated into this coast-wide dataset was run by the Hakai Institute (“Hak”) and adopted differing protocols (both ‘non-lethal’ and ‘lethal’) in different studies over time. Details relating to each of these programmes / protocols can be found in Table S1.

Field sampling took place weekly, biweekly, or monthly, depending on the programme, with the specific observation date and geographical coordinates of every sampling event being recorded. Each fish observed was measured for fork length (mm) and, in the case of the majority of those that used a ‘lethal’ protocol, weight (g) was also recorded. The numbers of fish observed during each sampling event varied according to location and timing within the season. Indeed, as can be deduced from Table 1a, that just over 25% of all attempts to sample wild juveniles resulted in no suitable fish (i.e. those in the target groups of Pacific salmonids, three-spined sticklebacks or herring) being caught. It can also be seen from the distribution plot shown in Fig. 2 that a fair proportion of sampling events resulted in only a few fish being assessed. In terms of the maximum number of fish assessed, under the non-lethal collection protocols, up to around 100 fish were assessed, as can once again be seen in Fig. 2. In early studies using the lethal protocol up to 30 fish from the most commonly occurring species (chum or pink) were assessed, though in many cases only one of these two species was present, which accounts for the peaks seen at 30 and 60 fish in Fig. 2. In later studies using lethal sampling, to avoid unnecessary death of fish, the numbers of a single species that were retained in some regions was reduced to 25 and then to 15, and once again the associated ‘peaks’ associated with these values can be seen in Fig. 2. In total well over 1 million wild fish were sampled across these wild sampling programmes, but the maximum levels put in place for active assessment noted above meant that only roughly 25% of these fish were assessed for sea louse infestation.

Fig. 2
figure 2

Frequency plot indicating the number of fish assessed in the 12,465 sampling events where fish were caught. (In almost 4,500 sampling events no fish were caught/assessed. There were also around 15 events where more than 115 fish were included in the sample – see code and output in Fig2_fish_per_sample.ipynb for details).

Full size image

The differing field protocols also led to differences in how sea lice infestation data were recorded. In the case of laboratory-based assessment (under the ‘lethal’ protocol), typically the developmental stage (copepodite, chalimus, pre-adult, or adult), species (Caligus clemensi, Lepeophtheirus salmonis, or not identified to species), as well as the sex (for motile stages) of any sea louse that was observed were recorded. However, those who are aware of the recent history of the L. salmonis species will know that only two chalimus stages are now recognised23, so even where stage data had been noted, records prior to 2014 which indicated “Chalimus 1” or “Chalimus 2” were updated to “Chalimus 1”, and those that had been “Chalimus 3” or “Chalimus 4” to “Chalimus 2”. In the case of the ‘non-lethal’ protocol, sea lice were categorised to broad development stage, but typically the chalimus lice stages were unidentified to species. Irrespective of the specific protocols adopted, we have attempted to maintain as much detail as was available, with the inclusion of codes to allow for categorisation into broader ‘common groupings’ where comparisons were being made across multiple protocols.

Data collected from salmon farms

The second major component of this dataset relates to information obtained as part of routine sea lice monitoring from just under 100 Atlantic salmon farms along the BC coast that have been in operation at various points over the past two decades. Not all of these farms are currently operating and, for some sites, data only exist for a few years; summary details based on each DFO fish health sub-zone can be seen in the information panels of Fig. 1. At the farm level, the main data recorded are the sea lice infestation levels at a given time on that farm. A number of pens, typically 2 to 3, are sampled each week and sea lice infestation data are recorded at the pen level. Typically, 20 fish are taken from each pen, resulting in estimates based on between 40 to 60 fish each week, or around 160 to 240 fish each month. At each pen-level event, fish are extracted into a tote where anaesthesia is typically applied to facilitate handling and examination without causing undue stress. In the dataset provided here, a mean monthly abundance value was calculated, based on the total sea lice counts divided by the number of sampled fish, for each month that a farm was in operation. Counts by developmental stage and sea lice species were recorded. All copepodite and chalimus are grouped as “chalimus”. Only motile sea lice were identified to species (C. clemensi or L. salmonis), while sex was also recorded for motile L. salmonis.

The size of each farm (in terms of farmed fish inventory) relative to all those operating within a given zone for a given month is captured, which allows a properly weighted average to be estimated from all farms within a zone. Even this does not allow for analyses that wish to come up with some estimate of the total numbers of sea lice from a particular farm. However, commercial sensitivity and legal restrictions mean that detailed inventories cannot be disclosed at the individual farm level. What has been done instead is that an average ‘load’ value is reported for each zone/month. This is the median ‘load’ value of estimated total sea lice at a specific stage from all sites in a given zone; where these totals are in turn based on the mean monthly abundance multiplied by the (known, but un-reported in this dataset) estimated farmed salmon inventory on that site in a given month.

The figures shown in Table 2 illustrate how these data could be used to review sea lice trends seen on salmon farms over the two decades under consideration. For each zone, an annual value is shown which represents the weighted mean L. salmonis motile infestation level based on data from all farms in that zone over the period from March to June (the ‘sensitive period’ defined within DFO regulations, to coincide with the main period of wild juvenile salmon outmigration). The number of farms from which data have been taken and the total number of monthly counts used in these estimations are also shown at the foot of the table.

Table 2 Weighted mean L. salmonis motile infestation abundance levels across all salmon farms by fish health zone, based on the wild juvenile salmon outmigration period (March–June) each year. (Entries that were unusually elevated are indicated with an *).
Full size table

In classical statistics the definition of an ‘outlier’ is any point lying more that 3 standard deviations from the mean, and such points are often removed from further analyses. We do not believe that excluding such values can be justified, but it can be useful in identifying ‘unusual’ data points such as the four cells (shown with an * and in bold). In addition, a slightly broader definition of ‘unusual’ could be taken to include any point lying more than 1.5 standard deviations away from the mean; the cells which met this criterion are also shown with an *. It seems clear that something unusual in terms of sea lice infestation on farms appears to have occurred in 2015, with all zones other than 2.3 and 3.1 (the most southerly zones) exhibiting unusually high levels of infestation. This was noted in a more fine-grained analysis, based on DFO published data over the period 2011 to 201624, and in that analysis was largely attributed to the so-called ‘warm blob’25, that it is estimated had its most significant impact in the ocean around Vancouver Island in 2015. It is interesting to note that there appear to be no other years over these two decades in which a similar widespread increase was observed, though given the ocean-wide impact that the ‘blob’ is thought to have exerted, it is not obvious why two zones would appear not to have been similarly affected in 2015.

In addition, a set of data curated by DFO from sea lice counts provided to them by aquaculture operators, as the regulator in BC, has also been included. These data are typically recorded and submitted to DFO at weekly intervals and when aggregated over year by zone, will give estimates similar to those shown in Table 2, though without the relative weighting for differing numbers of fish on the various sites involved in estimating each mean value. (See Table S2 for details.)

Data Records

The dataset is available at https://doi.org/10.6084/m9.figshare.2807810026, with this section being the primary source of information on the availability and content of the being described. There are five files associated with these data, two relating to sea lice observations on wild fish and three relating to sampling for sea lice abundance on Atlantic salmon farms. Table 3a describes the data fields associated with each wild sampling event, which includes the date and location of each observation and the observation programme (“Source”) under which each event was carried out. The file consists of 16,920 rows and 10 columns, where each row represents a field sampling event. Table 3b describes the data fields associated with the sea lice observations made on each wild fish. In addition to the 12 fields that are used to record sea lice presence according to various species and stages of life cycle development, the host species and physical characteristics are noted, as well as a reference field (“Event_ID”) to link each wild fish record to the field event during which it was sampled. This file contains a total of 376,764 rows, with each row representing a single fish.

Table 3 a Description of data fields associated with each wild sampling event [all_wild_sample_events]. b Description of data fields associated with each wild fish that was assessed for sea louse infestation [all_wild_fish_lice].
Full size table

Table 4a describes the data fields associated with each Atlantic salmon farm from which sea lice observations have been included. This includes the farm’s name, location and the aquaculture company responsible for that site. The file consists of 96 rows and 7 columns, with each row representing a different farm. Table 4b describes the data fields associated with the monthly mean abundance values for sea lice levels observed on farmed fish. The relevant farmed site, sampling year and month are noted, together with four columns providing mean sea louse abundance values and the number of sampled fish from which these means were generated, as well as a field indicating the ‘weighting’ that should be given to this monthly value when estimating zonal averages, based on the proportional number of fish present on that farm compared to the whole zone. There are 10,159 rows, each one representing a single monthly farm record. Table 4c describes a set of data fields that are derived from the farm-based sea lice abundance values and attempt to give a sense of the overall zonal sea louse ‘load’. In this case, each of the 1,527 rows contains information on the zone, sampling year and month, with the four sea lice columns representing the median total load of each species/stage recorded, where each farm’s load is estimated based on the monthly mean abundance on that farm multiplied by the estimated number of fish present on that farm during the month under consideration.

Table 4 a Description of data fields associated with each Atlantic salmon farm operating on the BC coast [industry_farm_details]. b Description of data fields associated with the mean monthly sea lice abundance estimates reported from each farm [industry_farm_abundance]. (An additional file covering similar data, but typically at a weekly level and published by DFO since 2011, has also been included. See Table S2). c Description of data fields associated with the estimated median monthly sea lice ‘load’ associated with each DFO fish health zone [industry_zone_loads_median].
Full size table

The summary statistics associated with key continuous variables from these various datasets are presented in Table 5. The dataset is provided as five separate files in CSV format and is published at the following figshare data repository (https://doi.org/10.6084/m9.figshare.28078100)26. The code to process these data sets can be found at – https://github.com/modailmara/BCSalmonData.

Table 5 Descriptive summaries of key continuous variables in these datasets (NA = missing value).
Full size table

Technical Validation

To explore the validity of the data, a summary of the physical characteristics of the wild fish sampled across the various zones and through the season is presented in Table 6. There is a clear progression in size over the sampled months with a gradual increase from March to May, before a more rapid increase in June (which extended into July, though the sample sizes are much more limited from this final month so mean values are not shown). The mean weight values recorded are in line with those shown in Jones and Hargreaves27, though the time periods reported in that paper cross monthly boundaries. It can also be seen that, as might be expected, when both Pacific salmon species were sampled concurrently, the chum tended on average to be a little longer/heavier than the pink salmon; again, the limited sample sizes later in the season make this less obvious in some zones.

Table 6 Breakdown of numbers of wild chum and pink salmon by DFO zone, with a summary of their mean sizes over the months March to June.
Full size table

Note that just over 60% of samples recorded the weight of the fish. However, for some measures of risk, sea lice density (measured in lice/g) is required, so it may be useful to estimate the weight for those fish that have only length data. The association between length and weight is subject to a degree of variability, but it has been demonstrated21 that it is possible to build relationships that provide reasonable estimates of fish weight for samples where length is present.

Usage Notes

With data from over 350,000 individual wild salmonid hosts and sea lice counts from an estimated 2 M farmed fish, over a two decade period, it may be tempting to assume that all the questions associated with sea lice interactions between farmed and wild populations can be answered using these data. This would be a mistake. As has been noted, the spatial coverage across the BC coast is fairly sporadic as far as wild sampling activities are concerned. With the exception of the Broughton Archipelago (Zone 3.3) and to a more limited extent the Central Coast area (Klemtu, Zone 3.5), other zones do not have extensive temporal coverage.

Even where data do exist, care must be taken to ensure that differences in, say, sampling protocol are adequately factored into any analyses that are carried out. It was noted that two very different sampling protocols have been used when estimating the sea louse infestation levels on wild fish in this dataset. As an illustration of the issues that may arise as a result, a summary of the estimated sea louse prevalence (proportion of fish with any sea louse infestation) and intensity (the mean number of sea lice on each infested fish) on wild chum and pink is shown in Table 7 for 2008 and 2009 on fish sampled under the ‘non-lethal’ and ‘lethal’ sampling protocols respectively. As the ‘non-lethal’ protocol was only used in Zone 3.3, the summarised data relate only to fish from that zone, with 2008 and 2009 being selected as there were relatively large sample sizes for all the months being compared.

Table 7 Estimates of sea louse prevalence and intensity on wild chum and pink in 2008 (N = 30,240) and 2009 (N = 26,371) for fish sampled under ‘non-lethal’ and ‘lethal’ sampling protocols respectively in sub-zone 3.3.
Full size table

As might be expected there are differences in the estimates generated under these differing protocols. In addition to the fact that the method of sampling fish and counting sea lice differed, the sampling events do not entirely overlap in terms of either the exact time in each month at which they were taken, nor in terms of spatial extent. What is clear is that similar patterns can be seen under both protocols. For example, infestation levels tend to rise over the course of the season. It can also be seen, particularly in 2008, that the prevalence levels estimated using the lethal method were around half those estimated using the non-lethal approach. However, when considering intensity, the levels reported under the lethal approach were slightly higher. This pattern appears to be repeated in 2009 with the exception of May where the intensity estimate for the non-lethal sampling is slightly higher, though in this month the prevalence difference reported from the lethal samples is also substantially lower. Indeed, there appears to be a strong correlation between these two metrics, in that as the overall prevalence estimates in the lethal sampling rise so the proportional difference in levels of intensity also increase.

The data shown in Table 7, have aggregated all sea lice into a single number, without reference to the stage or species that might be involved. This is partly due to the fact that different levels of granularity of description exist across assessment protocols and that these have in some cases changed over the two decades. Using the simple “any louse” categorisation allows the analyses to disregard these subtleties when exploring initial trends. In the case of wild sampling, the vast majority of sea lice observed tend to be in the early (copepodid or chalimus) stages and as such the absence of detailed lice stage data within certain protocols is less problematic.

It is not the purpose of this paper to delve into the details of such limitations or to attempt to explain the mechanisms that might cause differences in their interpretation. It is rather to illustrate that common patterns do exist, but also to make it clear that reported sea lice infestation metrics will tend to differ according to the protocols being used. Much more detailed analyses are required to generate comprehensive explanatory narratives and one of the intentions in making this dataset available is to enable such explorations to be carried out.