State of Salmon: Technical Documentation
Pacific Salmon Foundation
Version 1.0 (2024-Aug-21)
About the Report
The State of Salmon Report 2024 is a broad-scale evaluation of the state of all six species of Pacific salmon found in British Columbia and the Yukon, Canada. This flagship publication from the Pacific Salmon Foundation takes a data-driven approach to summarizing the status and trends in regional abundance for each species of salmon in nine major salmon-bearing regions in western Canada. The report also identifies key factors that are influencing salmon abundance and identifies ways in which society can contribute to salmon recovery and resilience.
Why PSF?
The Pacific Salmon Foundation, founded in 1987, is a non-profit environmental organization dedicated to making transformational change for the benefit of Pacific salmon at every scale – from the smallest streams to the open ocean. We work with First Nations, Government, ENGO partners, industry, and all salmon supporters to achieve our vision of healthy, sustainable, and naturally diverse populations of Pacific salmon for the benefit of ecosystems and people for generations to come. By bringing together diverse representatives, we embolden big ideas and support long-term positive changes for salmon. We operate at the nexus of science and action, and this is where the State of Salmon Report fits in: raising public awareness and political will for salmon to influence evidence-based decision making.
Our approach to assessing the State of Salmon is based in Western science and offers a data-driven perspective on broad-scale status and trends. For many regions and species, the scientific record is relatively short and may not adequately represent changes in abundance that have undoubtedly occurred over centuries of colonization, settlement, and human development. However, these data represent a type of information that can be relatively easily compiled, analysed, and compared across broad spatial scales. We encourage readers to seek out additional sources of information about salmon in their area, in particular from local First Nations who often have deep intergenerational knowledge and relationships with salmon.
Technical Documentation
This Technical Documentation augments the high-level description of Methods on the State of Salmon site and provides detailed information on the data sources and methods used to assess the State of Salmon. The suggested citation for this Technical Documentation is:
Pacific Salmon Foundation. 2024. The State of Salmon Report 2024: Technical Documentation. Version 1.0. Available online at https://salmonwatersheds.github.io/state-of-salmon/ [accessed dd-mm-yyyy].
More Information
Code and raw data are available in the state-of-salmon GitHub repository. Compiled data and output summary statistics on status and trends for each region and species are availabel for download in the Salmon Watershed Program’s Data Library.
Questions regarding data sources, analysis, and code can be directed to Stephanie Peacock (speacock@psf.ca).
Methods
General Approach
We report spawner abundance for each of nine regions that represent all major Pacific salmon-bearing watersheds in Canada: Yukon, Transboundary, Haida Gwaii, Nass, Skeena, Central Coast, Vancouver Island & Mainland Inlets, Fraser, and Columbia. These regions are also used to organize data in the Pacific Salmon Explorer. There are a relatively small number of Pacific salmon that spawn in the MacKenzie River basin in Arctic Canada that are currently not considered here.
We separate five species of Pacific salmon: Chinook, chum, coho, pink, and sockeye. We also report status of steelhead trout for regions where spawner data are are available and can reasonably be expanded to yield an estimate of regional abundance (Nass, Skeena, interior Fraser, and Columbia). When assessing biological status, pink salmon are often separated into even- and odd-year lineages due to their consistent 2-year life cycle. However, for the general overviews provided in State of Salmon we consider generational averages, which take the running average of even- and odd-year lineages for pink salmon. This approach of using generation running averages also smooths over dominant years for sockeye salmon, for which many populations display cyclic dominance. Shifts in dominance between even- and odd-year pink populations or declines in sub-dominant years of sockeye salmon are considered in a more nuanced way in our reporting when discussing how changes in abundance have been reflected in the diversity and distribution of each species within the region.
For each of these species, where data are available, we construct an index of spawner abundance at the regional scale. We focus on spawner abundance because these data are more readily accessible and easily summarised at different spatial scales. Spawners represent the abundance of salmon available to reproduce and contribute to future generations, as well as to meet cultural and ecological needs within watersheds, and thus provides a measure of status relevant to communities and ecosystems. We recognize that commercial catch has historically been a substantial portion of salmon that return to the coast, and that ignoring declines in catch will underestimate the declines in overall salmon abundance. Therefore, we include information on total abundance (i.e. spawners plus catch) for species and regions with reliable data.
We have strived to base our assessments on the best available data for each region and species. For example, stocks that are governed by international treaties may be monitored by the Pacific Salmon Commission (PSC), and tend to have reliable time series of abundance available at regional scales. These data sources are outlined in the Region-Specific Data section. For species and regions where aggregate abundance is not reported by the PSC or DFO at the scale needed, we adapted our approach to make the best use of available data. In most cases, this meant expanding spawner abundance from stream-level estimates to get a regional scale index of spawner abundance using two types of expansion factors (English et al. 2016). This expansion process is described in the Expansion Factors section.
Region-Specific Data Sources
Here we describe specific data sources for the abundance of salmon and steelhead at the regional scale.
Yukon
The Canadian portion of the Yukon River is home to Chinook, chum, and
coho salmon. Border escapement and total abundance (i.e., run size) of
Canadian-origin Chinook and fall chum salmon in the mainstem Yukon River
are counted at the Eagle Sonar station on the Yukon/Alaska border, and
are available from the Yukon
River Panel in their Joint Technical Committee (JTC) Reports.
Specifically, we used Chinook
RR Spawning escapement estimate and
RR Canadian origin total run size estimate from Appendix
B11 of JTC (2024). Chum salmon
spawner data were from Appendix B16 and total abundance from Appendix
B20 of JTC (2024).
There are also Chinook, chum, and coho salmon in the Canadian portion of the Porcupine River, which joins the mainstem Yukon River in Alaska. Data on escapement to the Porcupine River are more patchy and not currently included here.
Transboundary
For Chinook, coho, and sockeye, we assessed regional abundance in the
Transboundary using estimates of border escapement provided in the PSC’s
Joint
Transboundary Technical Committee Reports. The regional spawner and
total abundances of Chinook was calculated as the sum of escapement and
total return, respectively, to the Stikine, Alsek, and Taku watersheds,
available from Table B2 of CTC (2023) and provided to
PSF staff (E. Hertz) in Excel format following a data request to PSC’s
Chinook Technical Committee. Sockeye spawner and total abundances were
the sum escapement and Terminal Run,
respectively, from the Stikine River [Appendix B21; Transboundary Technical Committee (2022)] and the
Taku River [Appendix D15; Transboundary Technical
Committee (2022)]. Coho
spawner and total abundances are available for the Taku River only,
taken from Appendix D20 of Transboundary
Technical Committee (2022). The
most recent TTC report (Transboundary Technical
Committee 2022) includes data through 2021, and updated
Appendices with data through 2023 were provided to PSF staff on request
to Aaron Foos (DFO).
Pink and chum salmon are less extensively monitored in the Transboundary, with ongoing escapement available only from the Canyon Island fish wheel on the Taku River. We used the index of escapement from the Canyon Island fish wheel as an index of regional spawner abundance for pink and chum (Transboundary Technical Committee 2022). There are some historical data for pink salmon spawner abundance in the Nakina River of the Taku watershed, but this location has not reported data since 1998 and thus we did not include this stream survey in our index of regional abundance.
Steelhead trout have also been enumerated at the Canyon Island fish wheel since 1987, though recent estimates have been patchy and the timing of the fish wheel likely misses a substantial portion of the Taku summer run steelhead. Thus, we do not use the Canyon Island counts as an index of steelhead abundance.
We recognize that our approach in the Transboundary lacks information from many un-monitored watersheds, in particular the smaller watersheds of the Chilkat, Unuk, and Whiting Rivers. In the absence of better monitoring, we choose to report available data as a proxy for regional abundance, and note that the contribution of these smaller watersheds to total regional abundance for each species is likely small.
Haida Gwaii
For Haida Gwaii, we followed the Expansion Factors approach described below, with the exception of Chinook salmon for which the only data are from a single enumeration project on the Yakoun River. We show Yakoun River spawner abundance up to 2006, but no estimates were available for the most recent generation making the current status for Haida Gwaii Chinook “Unknown”. For other species, we compared our expansion approach to estimates of escapement to Area 1 reported in the TCNB (2023) (Appendix 30) but found the Area 1 estimates were much lower (except for Chinook, for which Appendix 30 matched the Yakoun River estimates exactly).
Nass
Chinook spawner and total abundances were taken from the PSC’s
Chinook Technical Committee (CTC) data sets and are reported on in
CTC (2024). Specifically,
Nass River (Area 3) escapement and terminal run are provided in Table B3
(fields Esc and t.run respectively).
Chum and pink salmon were expanded from available spawner surveys, as described below in Expansion Factors. Note that the designation of indicator streams for the Nass region was based on English et al. (2018) rather than the indicator designation in NuSEDS.
Nass coho spawner and total abundances were reconstructed by English et al. (2023) for
each of the three coho Conservation Units in the Nass region for
1992-2022. We summed the total escapement (TE) and total
abundance (Total Run) across these three Conservation Units
to yield the regional aggregate abundance.
Nass sockeye spawner and total abundances were derived from the
Northern Boundary Sockeye Run Reconstruction (NBSRR) Model that
estimates total escapement and total return for the Skeena and Nass
Rivers (English et al.
2004; English et
al. 2017). Specifically, we used fields TE
(Total Escapement) and Total run for Area 3 sockeye from
the 2022 update to the North and Central Coast (NCC) Salmon Database
Version 2, maintained by LGL Ltd. with support from PSF (English et al. 2016; Challenger et al.
2018; English et al.
2018).
The index of Nass steelhead spawner and total abundances are for the Nass Summer CU, developed in collaboration with the Nisga’a Fish and Wildlife Department and LGL Ltd. (English et al. 2023). There is another steelhead CU in the Nass region - Nass Winter - which is not well monitored and therefore not included in our index of spawner abundance.
Skeena
Chinook spawners were taken from the PSC’s
Chinook Technical Committee (CTC) data sets and are reported on in
CTC (2024). Specifically,
Skeena River (Area 4) escapement is provided in Table B3 (field
GSI esc). There is no total abundance estimate for the
Skeena River provided in the CTC data sets.
For Skeena chum salmon, we used escapement to Area 4 reported in Table 32 of TCNB (2023) rather than expanding escapement from spawner surveys because surveys of indicator stream have been increasingly patchy over the last 30 years, making expansions uncertain.
Skeena coho and pink salmon were expanded from available spawner surveys, as described below in Expansion Factors. Note that the designation of indicator streams for the Skeena region was based on English et al. (2018) rather than the indicator designation in NuSEDS.
Skeena sockeye spawner and total abundances were derived from the
Northern Boundary Sockeye Run Reconstruction (NBSRR) Model that
estimates total escapement and total return for the Skeena and Nass
Rivers (English et al.
2004; English et
al. 2017). Specifically, we used fields TE
(Total Escapement) and Total run for Area 4 sockeye from
the 2022 update to the North and Central Coast (NCC) Salmon Database
Version 2, maintained by LGL Ltd. with support from PSF (English et al. 2016; Challenger et al.
2018; English et al.
2018).
The index of Skeena steelhead spawner abundance is derived from estimated escapement of Skeena Summer steelhead at the Tyee Test Fishery (1956 - present), provided by the Province via email. As for other regions, these estimates may not capture winter-run steelhead, for which data are not available.
Central Coast
Estimates of spawner abundance for all salmon species on the Central Coast were expanded from available spawner surveys, as described below in Expansion Factors. Note that the designation of indicator streams for the Central Coast region was based on English et al. (2018) rather than the indicator designation in NuSEDS.
There are two spawner surveys for Steelhead trout on the Central Coast that we report in the Pacific Salmon Explorer. However, these surveys have been patchy through time with the most recent estimates from 2016. Therefore, spawner and total abundances for Central Coast steelhead was considered unknown.
Vancouver Island & Mainland Inlets
Estimates of spawner abundance for all salmon species and steelhead in Vancouver Island & Mainland Inlets were expanded from available spawner surveys, as described below in Expansion Factors. We began the time series of spawner abundance for this region in 1953 because earlier years had high expansion factor values indicating lack of monitoring on key indicator streams. We are exploring the potential use of well-monitored Chinook indicator stocks to inform an index of spawner and total abundances for Chinook (CTC 2023).
Fraser
Escapement of Fraser Chinook is provided in the PSC’s Chinook Technical Committee (CTC) data sets and are reported on in CTC (2024). We took the sum of escapement to indicator stocks in Table B6 (all Spring/Summer, Harrison, Lower Shuswap, Nicola, Lower Chilcotin, and Chilko) as an index of spawner abundance. Catch of Fraser Chinook is provided in Table A14 but cannot be directly added to indicator escapement to yield total return. PSF is pursuing Chinook run reconstruction output for the Fraser to inform total abundance.
Fraser chum catch and escapement are reported by PSC’s Chum Technical Committee (TTCHUM 2023), however the most recently published report only contains estimates for 2010-2019 and the data were not otherwise made available to us. As such, we estimated spawner abundance from available spawner surveys in NuSEDS as described below in Expansion Factors. We note that, although most NuSEDS data was updated to include surveys in 2022, for Fraser chum the most recent data was from 2020.
For interior Fraser coho CUs, spawner and total abundances were
provided by DFO on data request (pers. comm. Marissa Glavas, Data
Manager, Fraser River Stock Assessment). The Fraser coho estimates of
spawner and total abundances are based on Interior Fraser coho that
spawn upstream of Hells Gate in BC (including five CUs: Fraser Canyon,
Interior/Middle Fraser, Lower Thomposn, South Thompson, and North
Thompson). We report total abundance as the sum of
Total Pre-Fishery Abundance to all CUs and spawner
abundance as the sum of Total Return (i.e., final spawner
estimate plus fish removed from the system by DFO Salmon Enhancement
Program or First Nations) to all CUs. We recognize that using these
estimates ignores coastal coho populations, for which monitoring has
been patchy through time.
Data on spawner and total abundances of pink salmon and sockeye salmon in the Fraser region are provided by the PSC and accessed through the Fraser Panel Annual Report: Data Application (Pacific Salmon Commission 2024). Spawner abundance for Fraser sockeye aggregate was not yet available for 2023, and 2023 total abundance is considered preliminary. We note that Fraser River pink salmon are only counted in the dominant, odd-year run.
Steelhead trout in the Fraser are monitored by the Province in at least 10 different streams, but these data are not readily available. Relatively reliable estimates of steelhead spawner abundance are available at the CU-level for interior Fraser steelhead from the Thompson Summer CU (monitored at the Thompson River) and Mid Fraser Summer CU (primarily monitored at the Chilcotin River; data in the Pacific Salmon Explorer). We used the sum of CU-level spawner abundance for these two CUs as an index of Fraser steelhead abundance. We note that this approach does not include more coastal populations, such as the Lower Fraser Summer steelhead monitored in the Coquihalla River or Boundary Bay Winter steelhead, which may not have declined to the same extent over the past decade. However, a lack of publicly accessible data on coastal Fraser steelhead has limited our ability to include these CUs in our index of abundance.
Columbia
We used CU-level estimates of spawner abundance (run reconstructions) sourced from DFO (Ogden, pers. comm.) for Chinook and Stockwell and Hyatt (2003) and subsequent updates for sockeye.
There is no monitoring of steelhead trout in the Canadian portion of the Columbia region, but the Okanagan Nation Alliance does enumerate steelhead in akskwəkwant (Inkaneep Creek) and estimate a Canadian portion of steelhead spawning abundance. These data can be found in associated report (e.g. OBMEP (2022)) and are available in the Pacific Salmon Explorer as the CU-level spawner abundance for the Mid Columbia Summer CU.
Expansion Factors
For species and regions that lacked reliable data on spawner and total abundances at the appropriate scale, we estimated regional-scale spawner abundance from stream-level surveys. We started with spawner survey data shown in the Pacific Salmon Explorer. Spawner surveys were each assigned to one of the nine regions we considered based on their geographic location. We note that this is slightly different from how spawner survey data are organized in the Pacific Salmon Explorer, where data are organized by Conservation Units (CUs) that may span regional boundaries (e.g. for pink salmon that have relatively geographically large CUs). In cases of trans-regional CUs, the spawner surveys appear in both regions in the Pacific Salmon Explorer, whereas here we assign spawner surveys to the region in which they fall geographically, regardless of the CU boundary.
Spawner survey data are largely derived from river-level estimates in
DFO’s New
Salmon Escapement Database System (NuSEDS), but are cleaned up to
address issues of, for example, inconsistent naming of streams through
time or duplicate data. The spawner survey abundance is equal to the
MAX_ESTIMATE in NUSEDS for each year and river population,
calculated as the maximum of all fields containing spawner abundance
data (e.g. natural adult spawners, natural jack spawners, total
broodstock removals). Each of these river populations has been
designated as an indicator stream or
non-indicator. Indicator streams are observed more
consistently in recent decades, tend to have higher spawner abundance,
and tend to be monitored using more intensive methods that provide
greater accuracy [EnglishEtAl2016NorthCentralCoast]. For further
information on the compilation of spawner survey data, see the Pacific
Salmon Explorer Technical Report.
Expansion Factor 1, \(F_{1,y/d}\), expands the observed spawner abundances in indicator streams to account for indicator streams that are not monitored in a given year. It is calculated for each year \(y\) of the spawner time series, and relies on a decadal contribution of each indicator stream to the total escapement to all indicator streams, \(P_{d,i}\) in decade \(d\) [EnglishEtAl2016NorthCentralCoast]. The calculation of this decadal contribution requires at least one estimate from each indicator stream for the decade. If a decade does not contain sufficient information (i.e. one or more indicator streams are not monitored at all in a decade), then a reference decade is used to calculate \(P_{d,i}\). This reference decade is chosen to be: (1) the closest decade (historical or future) with sufficient information, or failing (1), (2) the 20-year period from 1980-1999 (Challenger et al. 2018).
For each decade (or reference decade if insufficient information) \(d\), the average number of spawners returning to indicator stream \(i\) is calculated as:
\[\bar{S}_{d,i} = \sum_{y = 1}^{Y_{d,i}} \frac{\hat{S}_{y/d, i}}{Y_{d,i}} \] where \(Y_{d,i}\) is the number of years for which spawner estimates are available within decade \(d\) for stream \(i\). From the average number of spawners for all indicator streams, the decadal proportional contribution of each indicator stream is calculated as:
\[P_{d,i} = \frac{\bar{S}_{d,i}}{\sum_{i=1}^{I} \bar{S}_{d,i}}\] where \(I\) is the total number of indicator streams.
Expansion Factor 1 is then calculated for each year within the decade \(y/d\) based on the decadal contributions and which streams were monitored or not in a given year:
\[F_{1,y/d}=\left( \sum_{i=1}^I P_{d,i} w_{y/d,i} \right)\] where \(w_{y/d,i}\) is 1 if stream \(i\) was monitored in year \(y\) and 0 if stream \(i\) was not monitored in year \(y\). Expansion Factor 1 is then multiplied by the sum of the observed spawners in all indicator stream to yield the expanded estimate of spawner abundances in all indicator streams in the region:
\[S'_{y} = F_{1,y/d} \sum_{i=1}^I \hat{S}_{y,i}\]
Expansion Factor 2 \(F_{2,d}\) expands the spawner abundance to all indicator streams, \(S'_{y}\), to account for non-indcator streams. Unlike Expansion Factor 1, this is calculated for each decade (rather than each year) and then applied to all years within a decade. Like Expansion Factor 1, there needs to be sufficient information within the given decade in order to calculate \(F_{2,d}\), or else a reference decade is chosen. See English et al. (2016) for detailed on how reference decades are chosen in that case.
Expansion Factor 2 is calculated as:
\[F_{2,d} = \frac{\sum_{i = 1}^I \bar{S}_{d,i} + \sum_{j = 1}^{J} \bar{S}_{d,j}}{\sum_{i = 1}^I \bar{S}_{d,i}}\] where \(\bar{S}_{d,i}\) and \(\bar{S}_{d,j}\) are the deacdal average number of spawners in indicator and non-indicator streams, respectively, calculated above. \(J\) is the total number of non-indicator streams. The adjusted total number of spawners in both indicator and non-indicator streams is then calculated as: \[ S''_{y} = F_{2,d} S'_{y} \]
Note that when expanding spawner abundance for spawner-recruit analysis, a third expansion factor is applied to account for streams that are never monitored and for observer (in)efficiency (Peacock et al. 2020). We did not apply this third expansion factor because it is highly undertain and we are interested in relative changes in abundance through time, so we do not require to expand to absolute abundance.
Smoothing
We smoothed time series of spawner and total abundances using a right-aligned running geometric mean over the length of a generation. This reduces the influence of dominant years and produces an index that is less sensitive to stochastic interannual variability that is common in salmon population dynamics. The generation length was based on the dominant life-history type for each species in a particular region. The smoothed spawner abundance in year \(y\) given a generation length \(g\) was calculated as: \[ \bar{S_{y}} = \left( \prod_{t = y-g+1}^y {S''_t} \right)^{1/g} \]
We smoothed abundance from the first year to the most recent year of
raw abundance data. If there were intermediate years with missing data,
the smoothed abundance was still calculated using the available years
(i.e. ignoring the missing data, with the exponent \(1/g\) adjusted so that \(g\) reflected the number of years with data
in the generation). If all years in a generation were missing data, than
the smoothed abundance was NA.
When plotting, we show the smoothed abundance relative to the long-term historical average, so that species that have vastly different abundances within a region can be plotted on the same y-axis for comparison.
Quantifying change
We summarize the time series of smoothed spawner or total abundance at the regional scale using three different metrics:
- Current status relative to historical average;
- Long-term trend over the entire time series; and
- Short-term trend over the most recent three generations.
The currents status provides information on how the most recent spawner abundance compares to past values, while the two trend metrics provide information on the average direction of change and are less sensitive to the magnitude of current abundance.
The current status relative to historical average is calculated as: \[ (\bar{S_{y}} - \bf{S}) / \bf{S} \] where \(y\) is the most recent year for which the index of spawner abundance could be calculated and \(\bf{S}\) is the average smoothed spawner abundance over the entire time series.
The long-term trend is calculated following the recommendations of D’Eon-Eggertson et al. (2015), who found that the correct identification of declines in salmon population abundance may be most reliable when considering the entire time series, and applying regression-based estimates of change calculated from log-transformed and smoothed spawner abundances. As such, we fit a simple linear model to the time series of \(\log ( \bar{S}_{y} )\) over \(y\). The resulting slope, \(m\), gives an average estimate of annual change:
\[ m = \log \left( \frac{\bar{S}_{y}}{\bar{S}_{y-1}} \right)\].
With some rearranging, we can calculate the average annual percent change as: \[ \frac{\bar{S}_{y} - \bar{S}_{y-1}}{\bar{S}_{y-1}} = \frac{ e^m \bar{S}_{y-1} - \bar{S}_{y-1}}{\bar{S}_{y-1}} = e^m - 1\]. If the slope \(m\) is not significantly different from zero (\(p \geq 0.05\)), then the trend is classified as “stable” regardless of the magnitude of the estimated slope. If the slope is significantly different from zero (\(p < 0.05\)), then the trend is classified as increasing if \(m > 0\) or decreasing if \(m < 0\).
The short-term trend is calculated as described above, but only using the time series over the most recent three generations.
Above: Simulated time series of spawner abundance (grey line), smoothed with a running geometric mean over the generation length (4 years; black line). The three metrics are illustrated: Current status relative to historical average (black); Long-term trend over the entire time series (green); and Short-term trend over the most recent three generations (clay). The shaded regions of the regression lines indicate the 95% prediction intervals. In this example, the short-term trend is not statistically significant (p > 0.05), and so the trend is characterized as ‘stable’ despite the negative slope.
Additional Considerations
The standardized approach to quantifying change described above was applied across regions and species. We then considered two additional attributes prior to reporting on the State of Salmon:
If the species and region had no data in the most recent generation, the current status was “Unknown” even if historical abundance information was available. This was the case for Haida Gwaii Chinook, for which there were no reliable estimates of spawners since 2006.
If the time series of abundance (spawner or total abundance) contained fewer than 20 years of data, we could not reliably establish a historical baseline and the current status was therefore “Unknown”. This applied to Columbia Chinook and Columbia steelhead, both of which have been monitored since 2006.
