gfiphc: An R package for extracting and analysing groundfish data from the IPHC longline survey in BC

The International Pacific Halibut Commission (IPHC) conducts an annual stock assessment longline survey in waters from California to Alaska, including British Columbia (BC) waters. The survey's main goal is to provide data on Pacific Halibut (Hippoglosus stenolepis) for stock assessment purposes. However, data are also recorded on other species caught, making it the longest ongoing groundfish survey in BC waters and hence a valuable source of data for many species.

The gfiphc package contains functions for:

• extracting count data from the GFBio (GroundFish Biology) database, housed at the Pacific Biological Station, Nanaimo, BC, Canada
• plotting maps (and movies) showing locations of stations each year
• deriving relative biomass index trends from the IPHC survey for non-halibut groundfish species along the full BC coast, taking into account the changing survey methodologies and using consistent published methods
• input into the groundfish synopsis report by Anderson et al. (2019) for over 100 species of BC groundfish; the functions allow the report to be updated every few years, with the first update including all data up to 2021
• generating trends for a group of species (e.g. "skates combined")
• deriving relative biomass index trends for a specified area within BC waters, again taking into account the changing survey methodologies.

The latter example is being used to generate predators of Haida Gwaii Pacific Herring as information for a Case Study in DFO's Ecoystem Approach to Fisheries Management initiative. The survey data in general has not been used to its full potential, and we hope this package aids finding new applications. For future assessments, giphc provides the data and calculated indices. An example application would be future skate assessments. The last Big Skate and Longnose Skate assessments used IPHC survey data only from 2003-2011; with gfiphc data (and built-in methods) such data are now available back to 1995 and to 2021, potentially tripling the length of the time series used to assess these species.

Users outside of PBS can request someone at PBS to extract the data (try Andrew Edwards first - he has the data extracted and saved from running the All species vignette below) and then the analyses within the package will work. The set-level information (not species specific) is built into the package, as is some species-specific data.

A quick look

To see the available data for your particular species of interest, look at the saved All species vignette, and scroll down or search for the species (the order is the same as in Anderson et al., 2019). If there are no time series shown, or there are but they have lots of zeros, then you can conclude that the survey does not really catch your species of interest (unless you want to look at occasional observations). If the data do look useful then you'll likely want install and understand the package.

If you are interested in species that may be occasionally caught but are not listed in the All species vignette then install the package and run check_iphc_spp_name(ignore_obvious=FALSE) to see if they appear. If so it means there are some data available but not fully incorporated into gfiphc yet - please make a GitHub Issue and I'll look into it.

Vignettes

The vignettes are the best place to start, as they should cover many analyses users would want to do. The links here go to the saved .html versions on GitHub. The raw Rmarkdown code is available in the vignettes folder and should be used as a template.

• Single species - shows example data that are saved in the package, and for extracting and generating the longest index possible for a single species for the whole BC coast.
• Restricted area - for extracting and generating the longest index possible for a restricted area of interest.
• Herring predators - extraction of multiple species, and groups of species, for a restricted area, and analysed and saved as inputs for the Haida Gwaii Pacific Herring Case Study.
• All species - extraction of 113 species, to test updating of the groundfish synopsis report (Anderson et al. 2019).

Installation

In R you can install the latest version of the package with:

devtools::install_github("pbs-assess/gfiphc")

Methods to generate indices of abundance

Methods are described in detail in Appendix G of the groundfish synopsis report (Anderson et al. 2019). The methods build on those developed for assessments of Redbanded Rockfsh (Edwards et al. 2017) and Yelloweye Rockfsh (Yamanaka et al. 2018) . The Redbanded assessment was the first to develop an abundance index from the IPHC survey that went back to 1995, and included data up to 2012. For the Yelloweye assessment the methods were extended to demonstrate that the index based on waters north of Vancouver Island could be considered representative of the coastwide population. The synopsis report includes preliminary investigations into hook competition.

Updates since the 2019 synopsis report include the following, and are mostly noted in the updated report with 2021 data:

• splitting the original functions out of gfplot and into their own package gfiphc.
• vignettes and better documentation.
• Series A-D automatically exclude the expanded grid stations for 2018 (these were included in the original synopsis report).
• 2019 data are now in GFBio and automatically extracted in the queries here.
• 2020 and 2021 data are included in gfiphc since only the first 20 hooks of each skate were enumerated (and previous years in GFBio had all hooks enumerated).
• 2020 data included also used an expanded grid, defined slightly differently to the expanded grid in 2018. We have stuck with the 2018 definition of which stations are not part of the standard grid. See iphc-2020-data.pdf, and it's associated .Rmd if desired, for details. Also see that for instructions on extracting future years' data from the new IPHC data website. See the IPHC website for full details on survey protocols -- note that fishing in 2020 was mostly in July and August, whereas it is usually May to August.
• 2021 included six extra new stations that had never been fished before; these are excluded in iphc-2021-data.pdf.
• functionality to analyse data from a user-specified region (see Restricted Area vignette below).
• the boostrapping now uses 10,000 bootstrap samples instead of 1000, as it made a difference for some species.

Note that the analyses still exclude hook competition, but we are working on that (using outputs from gfiphc).

Updates since the report that had 2021 data are:

• 2022 data included. The survey had two extra new stations that had never been fished before; these are excluded in iphc-2022-data.pdf.

Differing data-collection protocols

A particular issue is that the data resolution is not the same across the years. This is explained in Table G.1 in the synopsis report. An updated version of that table is given here:

Year	Hooks enumerated	Data resolution	Location of data	WCVI?
1995	All	Set-by-set	gfiphc package	N
1996	All	Set-by-set	gfiphc package	N
1997-1998	First 20 of each skate	Set-by-set	gfiphc package	N
1999	First 20 of each skate	Set-by-set	gfiphc package	Y
2000	First 20 of each skate	Set-by-set	gfiphc package	N
2001-2002	First 20 of each skate	Set-by-set	gfiphc package	Y
2003-2011	All	Hook-by-hook	DFO database GFBio	Y
2012	All (bait experiment)	Hook-by-hook	DFO database GFBio	Y
2013	First 20 of each skate	Set-by-set	gfiphc package	Y
2014-2017	All	Hook-by-hook	DFO database GFBio	Y
2018	All (+ expansions stns)	Hook-by-hook	DFO database GFBio	Y
2019	All	Hook-by-hook	DFO database GFBio	Y
2020	First 20 of each skate	Set-by-set	gfiphc package	N
2021	First 20 of each skate	Set-by-set	gfiphc package	Y (reduced)
2022	First 20 of each skate	Set-by-set	gfiphc package	Y (reduced)

In 2018 there were extra expansion stations surveyed (see the vignettes), and in 2020 only the first 20 hooks were enumerated. For 2020-2022, the data were downloaded from the IPHC website and included in the package (see below). For 2021 and 2022 the waters off the WCVI were surveyed, but only a subsample of stations were surveyed. This does not affect the Series A, B, and AB calculations, but would affect the Series D and CD calculations (since we assume 2021 still surveys the full coast), and will slightly affect the determination of whether or not Series AB (usually the longest series that can be constructed) can be considered representative of the full coast. Newer methods are being developed that will deal with such subsampling and so it has not been further dealt with here.

Also for 2021 there were six new stations that have never been surveyed before. We declare these as non-standard in gfiphc for constructing the various Series, since they do not appear in previous years. Five stations are along the Alaskan border, extending further north than ever before, and the sixth is just outside the Scott Islands Rockfish Conservation Area off the northwest tip of Vancouver Island. See iphc-2021-data.pdf for maps and further details. For 2022 there were two new stations on the coast of northern tip of Vancouver Island, which we also declared as non-standard; see iphc-2022-data.pdf for maps and further details.

The structure of the data in GFBio is described by Cooke and Olsen (2020), but gfiphc's R functions and built-in SQL queries mean that gfiphc users do not need to have knowledge of GFBio.

Summary of Series that can be constructed

Due to the differing data-collection protocols, we developed different ways to obtain as long a time series as possible (see aforementioned Appendix G). We first defined Series A, B, C, and D in Table G.2 (reproduced and updated below), combined and compared them as possible to define Series AB for the north of West Coast of Vancouver Island (WCVI), and tested whether Series AB could be considered as representing the full coast (see Appendix G for full details). Here we also look at a user-defined restricted area of interest, defining Series E to be based on the first 20 hooks from each skate, and Series F to be based on all hooks from each skate (see the the vignette). Series A-F are defined as:

	Only north of WCVI	Full coast	Restricted area of interest
First 20 hooks from each skate	A (26)	D (22)	E
All hooks from each skate	B (18)	C (16)	F

with numbers in parentheses indicating the number of years for which data for each Series are available (updated for 2021). For E and F this depends on the particular area of interest.

For any spatiotemporal analyses, note that there the IPHC assign a "don't use for spatiotemporal models" code for each set, so look into that if necessary.

Instructions for importing and checking new data into gfiphc each year and re-running the vignettes

For 2020-2022, only the first 20 hooks were evaluated from each skate. So, like for 2013, the data have been included in this package. For future years, copy the code data-raw/iphc-2022-data.Rmd (and rename for a new year) and follow the instructions for downloading the data from the IPHC website, checking the data, and saving it formatted for this package. The results iphc-2020-data.pdf, iphc-2021-data.pdf, and iphc-2022-data.pdf are also available for easier reading (only commit a final version of future .pdf's, not as you are working on it).

So, for 2021 I wrote this, and used for 2022:

1. Adapt data-raw/iphc-2021-data.Rmd as just described.

2. Create/adapt new functions to include the data in the calculations. For data in GFbio this should be automatic, else see what was done just after commit a5ecd2a for 2020, or (easier) see these notes updated in 2021:

3. Further files in R/ that need checking (search through for most recent year and update; I checked all R/ files in 2021 and these are the ones; this could be automated but it's actually good to look through the code each year:

   • gfiphc.R
   • iphc-data.R
   • plot-iphc.R

4. Check that the lines for new year data_names_XXXX in check_iphc_spp_name() in iphc-data.R got updated.

5. Re-run (line by line to check) data-raw/sets-skates-hooks-yelloweye.R to save some more new data into package.

6. Go through and rerun the vignettes (first copying old results into vignettes-results-XXXX-data/, XXXX is previous last year of data, to keep if needed), adapting any called functions or printed results as necessary to include latest year of data, making sure the latest year is included and the maps make sense. Need install(build_vignettes = FALSE) first as they use the library version.

7. May have to repeat this process as necessary. For example, in 2020 I saved the data, but then kept examining it in data-raw/iphc-2020-data.Rmd, realising needed to change a standard designation in setDataExpansion, so then had to rerun all of these steps to update all .rda files. In 2021 I hadn't originally noticed the six extra stations (I thought they were doing a subset, not expanding a a bit), so had to go back to step 1 and expand the default northern extent of all maps and assign the new stations as non-standard.

8. Re-run data-raw/iphc-****-data.Rmd again to use all the finalised saved data, and commit its .pdf file.

9. Update the two tables above of what data are available each year and what Series can be made, checking with vignettes. Add notes as necessary.

10. Update version number (1.2.0 has included the 2022 data) and email colleagues.

Citation

If you use this package please cite it as it is helpful for maintaining it:

Edwards, A.M., S.C. Anderson, E.A. Keppel and C.J. Grandin (2023). gfiphc: Data Extraction and Analysis for Groundfish Data from the IPHC Longline Survey in BC. R package version 1.2.0. Available from https://github.com/pbs-assess/gfiphc.

Check citation("gfiphc") for the latest version number and for the bibtex entry (though you have to manually add the year and the website given above).

Issues, problems

Please report any problems as a GitHub Issue, using a minimal working example if possible (and please check the closed issues first).