In seananderson/gfplot: Data Extraction and Plotting for DFO PBS Groundfish Stocks

Adapting tidy_cpue_index for use with hook and line data in addition to trawl.

Extracting raw data

trawl <- gfdata::get_cpue_index()
hl <- gfdata::get_cpue_index(gear = "hook and line")


dat <- trawl
dat2 <- hl

• Changed min_positive_tows to min_positive_fe for consistency throughout the code. Applied this throughout gfplot, gfsynopsis and cpue_reports. Not sure if there are other packages to adapt.
• Separated out filters that apply to both trawl and hl from gear-type-specific filters
• Changed "hours fished" to "effort"
• Moved creation of alt_year to within the if use_alt_year statement to save computation time
• Removed/comemnted out code which was either redundant or did nothing

Testing function

tidy <- tidy_cpue_index(trawl, "yelloweye rockfish")
tidy2 <- tidy_cpue_index(hl, "yelloweye rockfish")

Set arguments in order to step through function:

species_common = "yelloweye rockfish"
year_range = c(1996, as.numeric(format(Sys.Date(), "%Y")) - 1)
alt_year_start_date = "04-01"
use_alt_year = FALSE
lat_range = c(48, Inf)
min_positive_fe = 100
min_positive_trips = 5
min_yrs_with_trips = 5
area_grep_pattern = "^3C|^3D|^5A|^5B|^5C|^5D|^5E"
lat_band_width = 0.1
depth_band_width = 25
clean_bins = TRUE
depth_bin_quantiles = c(0.001, 0.999)
min_bin_prop = 0.001
lat_bin_quantiles = c(0, 1)
gear = c("bottom trawl", "hook and line")

glimpse(tidy2)

How many rows of data are lost for hook and line data for those removed due to NA's? Total rows:

nrow(dat)

1982888

Count of NA's for various columns that may be filtered on:

dat_short <- dat %>% 
  select(vessel_name, vessel_registration_number, fe_start_date, fe_end_date, best_depth, latitude, longitude)
dat_short <- dat_short[!duplicated(dat_short, drop = FALSE)]
# colSums(is.na(dat_short))

vessel_name vessel_registration_number fe_start_date fe_end_date best_depth latitude longitude 8857 4497 495052 495050 198928 196644 196670

We lose about 10% of records (note that these numbers are species catch records, not individual fe's - there are about 230,000 unique fe_id's) by eliminating records with no lat/long or depth, 25% of records by filtering out those with no fe start or end date, 0.25% for those with no vessel reg numer.

Filtering out records without fe dates has been moved to just trawl data filtering in the function since it is dependent on dates for effort calculation.

Hooks per fe for hook and line effort

Longline gear specs are different in GFFOS than GFBio, so code from IPHC data extraction for effort cannot be used. Many hook data are missing, on the order of 350,000 out of 500,000 records in the gear specs table, out of approximately 2,000,000 records

How many hl fe's?

181,819 hl fe's in merged catch table 2008+ with associated longline spec's (and so potential number of hooks per set) 62,545 of these fe's have no number of hooks per set

Note these numbers change as data are updated in database Can we use average hooks/skate and average skates/set or skate length and hook spacing by vessel to estimate this? Sean will look at modelling missing values with vessel, depth, fishery, gear subtype, locality, lat, long...

How clean are the data?

What is a reasonable range for: -# skates/set - usually < 20, some very large numbers 100+ (perhaps these are meant to be numbers of hooks rather than skates) -# hooks/skate -skate length -hook spacing

hooks <- get_commercial_hooks_per_fe() hooks_sum <- hooks %>% summarise(min_sk = min(skates_obs, na.rm = TRUE), mean_sk = mean(skates_obs, na.rm = TRUE), max_sk = max(skates_obs, na.rm = TRUE)) hist(hooks$skates_obs) hooks2 = hooks %>% filter(skates_obs > 100) hist(hooks2$skates_obs) hist(hooks$ft_per_hook, breaks = seq(0,400,10), xlim = c(0,100)) hist(hooks$hooks_per_skate, breaks = seq(0, 16000, 100), xlim = c(0,3000)) hooks3 <- hooks %>% filter(!fishery_sector %in% c('HALIBUT', 'HALIBUT AND SABLEFISH')) hist(hooks3$hooks_per_skate, breaks = seq(0, 22000, 5), xlim = c(0, 2000))

For modelling the hooks per skate: Filter out skate_obs over 20. Filter out ft_per_hook over 20 Filter out less than 1200 on skate_ft Filter out over 300 hooks per skate

Can probably clean up the data with: - skate_obs > ~150 actually hooks/set - skate_obs btw ~20 & 150 not useful - ft_per_hook over 20 = ? - skate_ft < 1200 @~600 and 900 probably metric and could possibly be converted; others not sure can conclude reasoning and fix - hooks per skate > 300 unreasonable, investigate whether reporting overall hooks for the set

seananderson/gfplot documentation built on April 5, 2024, 6:29 a.m.