05-conclusions.rmd

# Discussion

Although the SMMS provides the longest continuous bottom trawl time series for groundfish off the coast of WCVI, several changes to the survey mean that the data are less reliable before 2003. 
The most important change to the sorting protocol with species-level (or lowest taxonomic group possible) identification starting in 2003.
These have increased the number of species identified and would affect measures of biodiversity.
Some species groups, such as the flatfishes and rockfishes, did not appear to have been affected by the sorting protocol changes. 
Meanwhile, species observations for skates, eelpouts, and sculpins are only available or reliable after 2003. 
An additional uncertainty is the net change that occurred in 2006, which did not have sufficient calibration tows to include in index-standardization models for groundfish or to critically assess the effect on groundfish catchability or gear selectivity. 

When we compared spatiotemporal model-based indices of relative population biomass with traditional design-based indices, we found that the trends in each species were similar but the design-based index typically had more variable mean annual estimates and larger uncertainties compared with the modelled index.
Spatiotemporal models, which account for spatial correlation, have an advantage over design-based indices as they can be applied to both fixed-station designs [e.g., @webster2020] and to random sampling designs. 
They can be used when sampling locations vary over time, which occurs in the SMMS survey, where there is inter-annual variability in sampling locations over time. 
This suggests that a model-based approach to index standardization is better suited for calculating indices for groundfish than a design-based approach for the SMMS survey. 
Repeated surveys at fixed-stations could result in localised depletion of groundfish species, particularly species that have long lifespans and high site fidelity.
The spatiotemporal models in this report do not account for possible local depletion, which would negatively bias estimates of abundance.
This bias would stronger in recent years, as trawl overlap in 10-year spans has increased to ~50\% in a 10-year time period since 2009. 
However, prior to 2009, the consistency of trawl overlap was low (< 30\%). 

It was not possible to rigorously examine the uncertainties associated with the switch to comprehensive species sorting and identification that occurred in 2003 because the best dataset for comparison was the SYN WCVI survey, which only began in 2003.
Comparison with the SYN WCVI, particularly when using the same survey grid, showed comparable trends (e.g., correlations $\ge$ 0.5) across most species.

The comparison with commercial trawl CPUE in area 3CD---the only dataset that spans the 2003 change in sorting protocol---did not show strong correlations with survey index trends, with only 10/21 species having  $\ge$ 0.5 correlation between time series. 
However, due to the differences in spatial coverage of the SMMS and CPUE 3CD and the well-known possibility that the commercial CPUE is not necessarily proportional to abundance [e.g., @harley2001], it is difficult to draw clear conclusions.
In addition, the commercial CPUE 3CD index was estimated with a GLMM without explicit spatial structure compared to the spatiotemporal model-based SMMS index.

The information captured in the SMMS and SYN WCVI surveys was largely similar, but with potentially greater ability to detect recruitment from the SMMS.
When comparing the SMMS and the SYN WCVI, encounter rates, catch densities, and range of lengths and ages were similar for most species.
Of note, Eulachon, Flathead Sole, and Blackbelly Eelpout were encountered at least twice as frequently and at higher densities in the SMMS survey than in the SYN WCVI survey.
However, the SMMS may have provided an earlier signal of recruitment for several rockfish species given its smaller mesh than the SYN WCVI and annual sampling. 
There was some evidence of recruitment events in the SMMS data that were either not observed in the SYN WCVI survey or were not observed until later (Rougheye/Blackspotted Rockfish Complex, Darkblotched Rockfish).
The annual nature of the SMMS survey, compared to the biennial SYN WCVI, means that there is the potential to detect surprise events early (e.g., small Bocaccio and Lingcod captured in 2017) as well as to provide a more continuous data source to stock assessment models.

In conclusion, the analyses presented in this report suggest that the SMMS index should be considered as an index of relative biomass in BC groundfish stock assessment.
Analyses such as those included here can support decisions to include or exclude the SMMS index from assessments for particular stocks.
The SMMS indices correlated well with the SYN WCVI in overlapping years from 2003 onwards for many groundfish species.
Evidence for whether the SMMS is a reliable index for most groundfish species is less strong pre-2003, which is unfortunate since a primary use case of the SMMS would be providing a fisheries independent population index that begins before 2003.
The SMMS had lower correlation with commercial CPUE extending back to 1996 than the SYN WCVI from 2003 onwards.
It is unclear whether this is due to fisheries dependent vs. independent data sources, different standardization methods, different spatial domains, or changes to the SMMS survey from 2003 onwards in catch sorting and sampling.
Regardless of whether the SMMS is used for years pre-2003, it may be useful to groundfish assessments because it provides an annual index instead of a biennial index, and it tends to sample the same or smaller sized groundfish than the SYN WCVI survey.
As of 2024, there are plans to revise the SMMS survey design.
If the survey design is revised, future review of the impacts on groundfish indices may be required.

# Acknowledgements

We thank 
S. Acheson,
K. Fong,
D.R. Haggarty,
D.T. Rutherford, and
M.R. Wyeth,
for discussions and reviews of earlier drafts that substantially improved this report.
This document was produced with [csasdown](https://github.com/pbs-assess/csasdown/). 
Code to reproduce this report is available at <https://github.com/pbs-assess/gf-smms-techreport/>.