In pbs-assess/gfiphc: Data Extraction and Analysis for Groundfish Data from the IPHC Longline Survey in BC

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  fig.width = 6,
  fig.height = 6
  )
# install(build_vignettes = FALSE)  #to update locally when developing, to run:
# rmarkdown::render("data_for_one_species.Rmd")

This vignette shows how to extract and analyse the data for a single species. Since the extraction from the GFBio database only works for users on the Pacific Biological Station network that have permission to access the database, an example data set for Yelloweye Rockfish is included in the package.

Quick version -- there is a wrapper function, iphc_get_calc_plot_full(sp) that does all the extraction, calculations, and plotting for a given species. This vignette shows the details to aid understanding. So for a quick look to see if the data look potentially useful, just do, for example, iphc_get_calc_plot_full("bocaccio"). If the resulting series looks useful then repeat this vignette for your species of interest to understand the results.

Even quicker version for a quick look at results -- all the most recent data are now downloaded, analysed and the results shown (though not saved anywhere due to size). See the All Species link on the main README.

library(gfiphc)
library(dplyr)    # To print tibble properly

Data for species of interest

Define common species name:

sp <- "yelloweye rockfish"
sp_short_name <- "Yelloweye"  # short name for legends

It is worth checking that your species of interest is not in this list of IPHC species names that are in the IPHC data but not automatically extracted by this package (because the synopsis report did not look at it, or it is a general term such as Eelpout):

check_iphc_spp_name()

If you want to look for one of these species, then the common name and IPHC name need to be added to inst/extdata/iphc-spp-names.csv. There are also some non-groundfish species names, such as crabs, starfish and a Steller Sea Lion (!) that get automatically ignored in check_iphc_spp_name() -- type check_iphc_spp_name to see the function and the embedded list if you need to check.

Someone at PBS has to run the next chunk of code to extract data from the GFBio database and send the external collaborator the resulting .rds file, which will have the format species-name.rds, e.g. longnose-skate.rds. As an example, the data for Yelloweye Rockfish have been extracted and included as data in the package. Hence this chunk is not evaluated for this vignette:

# This chunk will only work within PBS
cache_pbs_data_iphc(sp)
                                       # That creates species-name. [This is
                                       #  what gfsynopsis::get_data_iphc() calls,
                                       #  via gfsynopsis::get_data() in
                                       #  report/make.R of gfsynopsis
                                       #  repo.]. Note the argument can be a
                                       #  vector species names to save multiple species.

Once an external person has the species-name.rds file, the two lines in the next chunk can be appropriately commented/uncommented to load the data. For this vignette we will use the Yelloweye Rockfish data included in the package:

# Comment this out to analyse non-yelloweye data:
sp_set_counts <- yelloweye_rockfish

# Uncomment this to analyse non-yelloweye data:
# sp_set_counts <- readRDS(paste0(gsub(" ", "-", sp), ".rds"))

Details for all skates and stations, and hooks returned with bait

Other data sets are already built into the gfiphc package, the main ones are given here for reference.

Locations and effective skate values of stations for the 1995 survey:

setData1995

Counts of each species at each station in 1995:

countData1995

Catches at each station from 1996 to 2002:

data1996to2002

Station details for 2013 survey:

setData2013

Station details for 2020 survey:

setData2020

Station details for 2021 survey:

setData2021

Station details for 2022 survey:

setData2022

Station details regarding expansion stations from 2018 (and some later years):

setDataExpansion

The following data sets are extracted from GFBio but extracted at PBS and then included in the package (using data-raw/sets-skates-hooks-yelloweye.R), which will be run each year to update the data objects.

Station details from GFBio for years not mentioned above:

sets_other_years

Skate-level details from GFBio (though such data are not available for years for which data are only available at the set-by-set level):

skates_other_years

Counts of hooks returned with bait on them, for each set, for all years (note it is a list containing one tibble):

hooks_with_bait

Counts of Yelloweye Rockfish on each set for all years (note it is a list containing one tibble):

yelloweye_rockfish

See ?<dataset> for details of each, and data(package = "gfiphc") for other data sets. Notation such as E_it20 matches the write-up in the Groundfish Synopsis report.

The formats are different to each other due to the data that are available. For example, there are no hooksObserved values available for 1995 and 2013, which will complicate consideration of hook competition for those years. setDataExpansion is needed to identify stations in the expanded grid (that were not fished in previous years).

Look at the data for the species of interest, which combines all the data for the species (built-in data sets plus data extracted from GFBio):

sp_set_counts
                  # For each set, the calculatable catch rates for that
                  #  species, plus lat and lon and whether the set is usable.
                  # Includes the basic information for each set, with further
                  #  details in the objects mentioned above.
                  # See ?get_all_iphc_set_counts for full details.

tail(sp_set_counts$set_counts)

summary(sp_set_counts$set_counts)

Maps of stations

For a single year, and whether or not the survey caught the species of interest in that year:

plot_iphc_map(sp_set_counts$set_counts,
              sp_short_name = sp_short_name,
              years = 2008)

For the most recent year of data (see below for discussion):

plot_iphc_map(sp_set_counts$set_counts,
              sp_short_name = sp_short_name,
              years = max(sp_set_counts$set_counts$year))

For 2022 (and 2021) note that there is sporadic coverage off the west coast of Vancouver Island, with some of the usual stations being surveyed. In previous years it was all or nothing. This will affect the results of Series C and D (see README), and so whether or not the resulting Series AB can be considered representative for the whole coast. Series AB does not include the west coast of Vancouver Island. For 2022 there are also two stations (six in 2021) that were never previously fished and so we call these non-standard -- see brief mention in the README and iphc-2021-data.pdf and iphc-2022-data.pdf for full details.

Unusable stations are determined by the IPHC and any data are excluded from our calculations. The horizontal line is the cut-off for calculations for different Series (see the synopsis Research Document). Note that the sp_short_name argument is only for the legend, the first argument needs to contain the data for the species of interest.

To see a movie of the station locations through time (a panel plot is shown shortly):

.

For reference, here is the code to build the movie, but it is commented out since it caused Travis (and now presumably GitHub Actions) to fail. Don't worry about that. If you do want to make a different movie, run this vignette code with the next commented line uncommented (you need to install the gifski R package and the Gifski program), right-click on the animation in the html viewer and save it with the above filename (updated as appropriate).

# ```r
for(i in unique(sp_set_counts$set_counts$year)){
  plot_iphc_map(sp_set_counts$set_counts,
                sp_short_name = sp_short_name,
                years = i)
}

A panel plot of all years is:

years_to_plot <- unique(sp_set_counts$set_counts$year)
par(mfrow = c(ceiling(length(years_to_plot) / 3),
              3))

for(i in years_to_plot){
  plot_iphc_map(sp_set_counts$set_counts,
                sp_short_name = sp_short_name,
                years = i,
                mar_val =   c(1.8, 2, 2.0, 0.5),
                include_legend = (i == years_to_plot[1]) # only first panel
                )
}

For 2021, as mentioned above, there are six that were never fished before and we define as non-standard: five in the far north and one off the northwest tip of Vancouver Island (the latter would have been in a Rockfish Conservation Area according to the grid pattern). And for 2022 there are two close to the shore of northern Vancouver Island.

To see the stations without reference to any species (these are the 2008 stations, so think would be just the standard ones, can check code to confirm):

plot_iphc_map(sp_set_counts$set_counts,
              sp = NULL,
              years = 2008)

noting that you still need sp_set_counts$set_counts (for an arbitrary species) as an argument.

Can also use hooks_with_bait$set_counts to see stations that came back with no bait (the legend will need adapting though; empty circles are no hook with bait returned):

plot_iphc_map(hooks_with_bait$set_counts,
              sp = "Hooks with bait",
              years = 2008)

Expanded survey grid in 2018, 2020-2022

In the above movie and panel plot you see that the spatial coverage of sets changed in 2018, 2020-2022 due to the expanded survey grid that year that involved sets at extra stations:

sets_each_year <- sp_set_counts$set_counts %>%
  group_by(year) %>%
  summarise(total = n())

plot(sets_each_year$year, sets_each_year$total, type = "o",
     xlab = "Year", ylab = "Sets each year", ylim = c(0, max(sets_each_year$total)))

The extra stations are identified in gfiphc as not standard (standard is N), with Y being the standard stations (see iphc-2020-data.pdf, iphc-2021-data.pdf, and iphc-2022-data.pdf for explanation of why we changed the values for some stations from the IPHC raw data -- the IPHC called some of the new ones standard, even though they hadn't been used before). Removing the non-standard stations brings the number of sets in 2018 and 2020-2022 into line with other years, and so this should be done for most analyses:

sets_each_year_standard <- sp_set_counts$set_counts %>%
  filter(standard == "Y") %>%
  group_by(year) %>%
  summarise(total = n())

plot(sets_each_year_standard$year, sets_each_year_standard$total, type = "o",
     xlab = "Year", ylab = "Sets each year", ylim = c(0, max(sets_each_year$total)))

Except in recent years we have less stations after doing this, and 2022 is the lowest in the time series -- spatiotemporal methods could be used to include the new non-standard stations (though also check the above .pdfs for mention of the original column that says that some stations should not be used for spatio-temporal analyses).

Note that the expansion (non-standard) stations only appear in recent years:

filter(sp_set_counts$set_counts, standard == "N") %>%
  select(year) %>%
  unique()

So we use filter(..., standard == "Y", usable == "Y") for upcoming analyses (to also include only the usable stations).

Calculate longest time series possible for the full coast

In Appendix G of the synopsis report we described analyses to calculate as long a time series as possible by combining data from years when only the first 20 hooks were enumerated for each skate (Series A) with data from years when all hooks could be enumerated from each skate (Series B); there are overlapping years when all hooks were enumerated and the data are available at the hook-by-hook level. The resulting Series AB applies only to the area north of WCVI, but was tested to see if it was representative of the whole coast by also calculating Series C and D.

The function calc_iphc_full_res() does all the calculations in one step and saves all the output. See ?calc_iphc_full_res for full details of the output, but basically $ser_longest is the longest time series that can be made for the species, with $full_coast indicating whether it can be considered representative of the full coast. Series A, B, C, and D are also given, along with geometric means and the test results comparing scaled series.

series_ABCD_full <- calc_iphc_full_res(sp_set_counts$set_counts)
series_ABCD_full

Now plot Series A and B and show their scaled versions and the resulting Series AB (bottom right), which is the longest series available and is considered applicable to the whole coast if series_ABCD_full$full_coast is TRUE:

plot_IPHC_ser_four_panels_ABCD(series_ABCD_full,
                               sp = sp)
series_ABCD_full$type
series_ABCD_full$full_coast

Recall there is a wrapper function, iphc_get_calc_plot_full(sp), mentioned at the start of this vignette.

Note that the code may not yet fully account for all possibilities for the different Series, particularly for rarer species with, by definition, low count numbers. So do check that the output is sensible. The code may calculate a longest series, but it may not be ecologically useful. And a shorter series based on all years for which all hooks were counted (Series B) may be more useful than a longer one that uses only the first 20 hooks (Series A). In such cases maybe a Series BA could be implemented (use original Series B and then rescale Series A).

pbs-assess/gfiphc documentation built on July 4, 2023, 1:13 p.m.