Stock assessment data simulation

Updated on March 14, 2022

EwE scenario

We used the output data from EwE AMO_PCP scenario for generating single species stock assessment input data. The AMO_PCP scenario incorporates Atlantic Multidecadal Oscillation and precipitation information in the ecosystem modeling work and it creates a strong response.

IP4EBFM::create_fishery() is used to simulate fishery data

  • Input data
    • EwE annual catch-at-age and weight-at-age data
    • BAM annual Coefficient of variation (CV) of catch and sample size data.
  • Output data
    • Ecosystem operating model total catch in biomass, total catch in abundance, catch-at-age in abundance, catch-at-age in biomass, CV of catch, sample size, and weight-at-age data
    • Observed total catch in biomass and catch-at-age in proportion for one fleet
    • Units information
# Create fishery ----------------------------------------------------------

fishery_sample_num <- cbind(
  menhadenSA_output$t.series$acomp.cRs.n[which(menhadenSA_output$t.series$year %in% years)],
  menhadenSA_output$t.series$acomp.cBs.n[which(menhadenSA_output$t.series$year %in% years)],
  menhadenSA_output$t.$acomp.cBn.n[which(menhadenSA_output$t.series$year %in% years)],
  menhadenSA_output$t.series$acomp.cRn.n[which(menhadenSA_output$t.series$year %in% years)]
)
fishery_sample_num[fishery_sample_num==-99999] <- 0

fishery <- create_fishery(
  file_path = file.path(ewe_output_path, "catch_annual.csv"),
  skip_nrows = 8,
  species = 4:10,
  species_labels = paste0("age", ages),
  ewe_years = 0:32,
  data_years = years,
  fleet_num = 1,
  selectivity = NULL,
  CV = rep(0.05, length(years)),
  sample_num = apply(fishery_sample_num, 1, sum),
  waa_path = file.path(ewe_output_path, "weight_annual.csv")
)

Catch biomass and catch-at-age in proportion over time

IP4EBFM::create_survey() is used to simulate survey data

  • Input data
    • EwE monthly biomass-at-age and weight-at-age data
    • BAM survey number, time, selectivity, catchability, CV, sample size data
  • Output data
    • Ecosystem operating model biomass-at-age, abundance-at-age, abundance index, CV of surveys, sample size, weight-at-age, and length-at-age data
    • Observed abundance indices, length composition in proportion/in number using different approaches
    • Units information
  • Survey information
    • survey1 survey: Oct from 1990 - 2017; length composition data from 1990 - 2017
    • survey2 survey: April from 1985 - 2017; length composition data from 2013 - 2017
    • survey3 survey: April from 1990 - 2017; no length composition data
    • survey4 survey: June from 1985 - 2017; no length composition data
# Create survey ----------------------------------------------------------

# selectivity settings
survey_num <- 4

survey_name <- c("survey1", "survey2", "survey3", "survey4")

# set up survey time
# Need to check Table 26 from BAM assessment: Length cutoffs used to distinguish age-0 from age-1+ Atlantic menhaden at different regions.

survey1_year <- 1990:2017 # Adult Index (survey1): age 1+ fish; September - January; Time of the year when menhaden were most abundant in this region

survey2_year <- 1985:2017 # Adult Index (survey2): age 1+ fish; March - May

survey3_year <- 1990:2017 # Adult Index (survey3): age 1+ fish: April - July

survey4_year <- 1985:2017 # YOY Index (survey4): covered all months, many surveys starts at July

survey_time <- list(
  survey1 = data.frame(
    year = survey1_year,
    month = rep(10, length(survey1_year)) # Oct 15
  ),
  survey2 = data.frame(
    year = survey2_year,
    month = rep(4, length(survey2_year)) # April 15
  ),
  survey3 = data.frame(
    year = survey3_year,
    month = rep(4, length(survey3_year)) # April 15
  ),
  survey4 = data.frame(
    year = survey4_year,
    month = rep(6, length(survey4_year)) # June 1
  )
)

# set up survey selectivity
survey1_sel <- IP4EBFM::logistic(
  pattern = "simple_logistic",
  x = ages,
  slope_asc = 2.2,
  location_asc = 3.0
)

survey2_sel <- IP4EBFM::logistic(
  pattern = "double_logistic",
  x = ages,
  slope_asc = 4.3,
  location_asc = 2.3,
  slope_desc = 3.5,
  location_desc = 2.3
)

survey3_sel <- IP4EBFM::logistic(
  pattern = "double_logistic",
  x = ages,
  slope_asc = 7.0,
  location_asc = 0.3,
  slope_desc = 7.0,
  location_desc = 2.0
)

survey_selectivity <- list(
  survey1 = as.data.frame(
    matrix(rep(survey1_sel, times = length(years)), ncol = length(ages), byrow = TRUE),
    row.names = years
  ),
  survey2 = as.data.frame(
    matrix(rep(survey2_sel, times = length(years)), ncol = length(ages), byrow = TRUE),
    row.names = years,
  ),
  survey3 = as.data.frame(
    matrix(rep(survey3_sel, times = length(years)), ncol = length(ages), byrow = TRUE),
    row.names = years,
  ),
  survey4 = as.data.frame(
    matrix(rep(c(1, rep(0, 6)), times = length(years)), ncol = length(ages), byrow = TRUE),
    row.names = years,
  )
)

survey_selectivity <- lapply(survey_selectivity, setNames, paste("age", ages))

# set up catchability
yr_catchability_change_survey4 <- 1986

survey_catchability <- list(
  survey1 = menhadenSA_output$t.series$q.nad[which(menhadenSA_output$t.series$year %in% years)],
  survey2 = menhadenSA_output$t.series$q.mad[which(menhadenSA_output$t.series$year %in% years)],
  survey3 = menhadenSA_output$t.series$q.sad[which(menhadenSA_output$t.series$year %in% years)],
  survey4 = c(menhadenSA_output$t.series$q.jai[which(menhadenSA_output$t.series$year %in% c(years[1]:yr_catchability_change_survey4))], menhadenSA_output$t.series$q2.jai[which(menhadenSA_output$t.series$year %in% c((yr_catchability_change_survey4 + 1):tail(years, n = 1)))])
)

survey_catchability <- lapply(survey_catchability, setNames, years)

# set up CV
survey_CV <- list(
  survey1 = menhadenSA_output$t.series$cv.U.nad[which(menhadenSA_output$t.series$year %in% years)],
  survey2 = menhadenSA_output$t.series$cv.U.mad[which(menhadenSA_output$t.series$year %in% years)],
  survey3 = menhadenSA_output$t.series$cv.U.sad[which(menhadenSA_output$t.series$year %in% years)],
  survey4 = menhadenSA_output$t.series$cv.U.jai[which(menhadenSA_output$t.series$year %in% years)]
)
survey_CV <- lapply(survey_CV, setNames, years)

# set up sample number
# survey_sample_num <- list(
#   survey1 = menhadenSA_output$t.series$lcomp.nad.nfish[which(menhadenSA_output$t.series$year %in% years)],
#   survey2 = menhadenSA_output$t.series$lcomp.mad.nfish[which(menhadenSA_output$t.series$year %in% years)],
#   survey3 = rep(NA, length = length(years)),
#   survey4 = rep(NA, length = length(years))
# )

survey_sample_num <- list(
    survey1 = rep(800, length= length(years)),
    survey2 = rep(800, length= length(years)),
    survey3 = rep(800, length= length(years)),
    survey4 = rep(800, length= length(years))
)
survey_sample_num <- lapply(survey_sample_num, setNames, years)

for (i in 1:length(survey_sample_num)){
  survey_sample_num[[i]][survey_sample_num[[i]] == -99999] <- NA
}

# set up age-length population structure
length_bin <- seq(10.0, 400, 10)/10 # in cm
mid_length_bin <- seq(10.5, 405, 10)/10 # in cm
nbin <- length(length_bin)
bin_width <- 1

length_CV <- list(
  survey1 = 0.12,
  survey2 = 0.17,
  survey3 = NA,
  survey4 = NA
)

# Create survey
survey <- IP4EBFM::create_survey(
  file_path = file.path(ewe_output_path, "biomass_monthly.csv"),
  skip_nrows = 8,
  species = 4:10,
  species_labels = paste0("age", ages),
  years = years,
  survey_num = survey_num,
  survey_time = survey_time,
  selectivity = survey_selectivity,
  catchability = survey_catchability,
  CV = survey_CV,
  sample_num = survey_sample_num,
  waa_path = file.path(ewe_output_path, "weight_monthly.csv"),
  length_bin = length_bin,
  mid_length_bin = mid_length_bin,
  nbin = nbin,
  bin_width = bin_width,
  length_CV = length_CV
)

Survey selectivity patterns

Survey index over time

Survey length composition over time

  • SS3 approach

  • BAM approach

Key findings

  • Tried two approaches to calculate age-length probability matrix
    • use SS3 approach using pnorm in R
    • rewrite ADMB function in R based on BAM code
    • the two approaches produce almost identical results, so will only use SS3 approach for further analysis
  • Observed a very high proportion of fish in length bin 10.5 cm
    • need to check selectivity-at-age 0 and perhaps modify selectivity pattern for survey survey1
    • expand length bins. According to EwE outputs, mean length-at-age 0 is around 6 cm, which is below the current first length bin 10.5 cm.

Modify selectivity pattern for survey survey1 and update length bins

  • Change slope_asc from 2.2 to 3.0:
    • selectivity-at-age from original scenario: 0.003, 0.030, 0.240, 1.00, 0.400, 0.027, 0.001
    • updated selectivity-at-age: 0.0003, 0.0060, 0.1147, 1.0000, 0.4251, 0.0268, 0.0014
  • Change the first length bin from 10 cm to 1 cm and the last length bin from 40.5 cm to 46.5 cm
survey1_sel <- IP4EBFM::logistic(
  pattern = "simple_logistic",
  x = ages,
  slope_asc = 3.0,
  location_asc = 3.0
)

survey1 = as.data.frame(
    matrix(rep(survey1_sel, times = length(years)), ncol = length(ages), byrow = TRUE),
    row.names = years
  )
colnames(survey1) <- paste0("ages", ages)

survey_selectivity$survey1 <- survey1

# set up age-length population structure
length_bin <- seq(10, 500, 10)/10 # in cm
mid_length_bin <- seq(15, 505, 10)/10 # in cm
nbin <- length(length_bin)
bin_width <- 1

# Create survey
survey <- IP4EBFM::create_survey(
  file_path = file.path(ewe_output_path, "biomass_monthly.csv"),
  skip_nrows = 8,
  species = 4:10,
  species_labels = paste0("age", ages),
  years = years,
  survey_num = survey_num,
  survey_time = survey_time,
  selectivity = survey_selectivity,
  catchability = survey_catchability,
  CV = survey_CV,
  sample_num = survey_sample_num,
  waa_path = file.path(ewe_output_path, "weight_monthly.csv"),
  length_bin = length_bin,
  mid_length_bin = mid_length_bin,
  nbin = nbin,
  bin_width = bin_width,
  length_CV = length_CV
)

Updated survey length composition over time

IP4EBFM::create_biodata() is used to prepare biological data for a stock assessment 

biodata <- create_biodata(nsex=1, narea=1, ages=ages, years=years,
                          length_bin=length_bin, mid_length_bin=mid_length_bin,
                          nbin=nbin, bin_width=bin_width, length_CV=length_CV,
                          annual_weight_path=file.path(ewe_output_path, "weight_annual.csv"),
                          monthly_weight_path=file.path(ewe_output_path, "weight_monthly.csv"),
                          species=4:10,
                          species_labels=paste0("age", ages),
                          skip_nrows=8,
                          lw_a=0.01, lw_b=3,
                          k=0.331,
                          t0 = -0.1,
                          winf = 0.237,
                          maturity_at_age=c(0.0, 0.1, 0.5, 0.9, 1.0, 1.0, 1.0), # From both BAM and EwE
                          natural_mortality_at_age=c(1.76, 1.31, 1.03, 0.9, 0.81, 0.76, 0.72) # From both BAM and EwE
                          )

Key findings

  • EwE Von Bertalanffy Growth model uses a specialized equation. Carrying capacity parameter k from the model needs to be carefully compared before using the value. The default values used in the EwE weight-at-age calculation are different compared to the BAM input values, so the weight-at-age matrices would be different from different models.