Herring index from stomachs

Data updated through 2022

1 Summary

This index uses herring observed in predator stomachs to evaluate potential herring trends for the 2025 RT assessment.

Methods are based on the forage index for Bluefish (Gaichas et al. 2023) and the State of the Ecosystem forage index. These indices were inspired by Ng et al. (2021), which specifically looked at herring in stomachs of predators, but for each predator separately.

Will discuss a good starting point with the team, but for now start at 1973 and use data through 2022.

2 Workflow

Operational updates to the forage index submitted to the State of the Ecosystem report are in the forageindex github repository: https://github.com/NOAA-EDAB/forageindex

Data input files are in the folder fhdat and were processed with the script VASTherringWG_ProcessInputDat.R in that folder: https://github.com/NOAA-EDAB/forageindex/blob/main/fhdat/VASTherringWG_ProcessInputDat.R

# Streamlined version of CreateVASTInputs.Rmd for operational updates to forage index indicators
# February 2024
#   This one uses only herring in stomachs, otherwise the same as SOE forage index
#   To be presented to the 2025 herring RT WG

library(tidyverse)
library(here)
library(dendextend)

# Load NEFSC stomach data received from Brian Smith

# object is called `allfh`
load(here("fhdat/allfh.Rdata"))

#object is called allfh21
load(here("fhdat/allfh21.Rdata"))

#object is called allfh22
load(here("fhdat/allfh22.Rdata"))

# bind all NEFSC stomach datasets
allfh <- allfh %>%
  dplyr::bind_rows(allfh21) |>
  dplyr::bind_rows(allfh22)

###############################################################################
# read predator similarity info to generate predator list
# Input NEFSC food habits overlap matrix:

dietoverlap <- read_csv(here("fhdat/tgmat.2022-02-15.csv"))

# use dendextend functions to get list
d_dietoverlap <- dist(dietoverlap)

guilds <- hclust(d_dietoverlap, method = "complete")

dend <- as.dendrogram(guilds)

dend <- color_branches(dend, k=6) # Brian uses 6 categories

labels(dend) <- paste(as.character(names(dietoverlap[-1]))[order.dendrogram(dend)],
                      "(",labels(dend),")", 
                      sep = "")

pisccomplete <- partition_leaves(dend)[[
  which_node(dend, c("Bluefish..S(37)", "Bluefish..M(36)", "Bluefish..L(35)"))
]]


# Filter NEFSC food habits data with predator list
pisccompletedf <- data.frame("COMNAME" = toupper(str_remove(pisccomplete, "\\..*")),
                             "SizeCat" = str_remove(str_extract(pisccomplete, "\\..*[:upper:]+"), "\\.."),
                             "feedguild" = "pisccomplete")

fh.nefsc.pisc.pisccomplete <- allfh %>%
  #filter(pynam != "EMPTY") %>%
  left_join(pisccompletedf, by = c("pdcomnam" = "COMNAME",
                                   "sizecat" = "SizeCat")) %>%
  filter(!is.na(feedguild))


##############################################################################
# Get prey list from NEFSC and NEAMAP
# Irrelevant! The prey list is now Atlantic herring and unid clupeids.

# preycount  <- fh.nefsc.pisc.pisccomplete %>%
#   #group_by(year, season, pdcomnam, pynam) %>%
#   group_by(pdcomnam, pynam) %>%
#   summarise(count = n()) %>%
#   #arrange(desc(count))
#   pivot_wider(names_from = pdcomnam, values_from = count)
# 
# 
# gencomlist <- allfh %>%
#   select(pynam, pycomnam2, gencom2) %>%
#   distinct()
# 
# NEFSCblueprey <- preycount %>%
#   #filter(BLUEFISH > 9) %>%
#   filter(!pynam %in% c("EMPTY", "BLOWN",
#                        "FISH", "OSTEICHTHYES",
#                        "ANIMAL REMAINS",
#                        "FISH SCALES")) %>%
#   #filter(!str_detect(pynam, "SHRIMP|CRAB")) %>%
#   left_join(gencomlist) %>%
#   filter(!gencom2 %in% c("ARTHROPODA", "ANNELIDA",
#                          "CNIDARIA", "UROCHORDATA",
#                          "ECHINODERMATA", "WORMS",
#                          "BRACHIOPODA", "COMB JELLIES",
#                          "BRYOZOA", "SPONGES",
#                          "MISCELLANEOUS", "OTHER")) %>%
#   arrange(desc(BLUEFISH))
# 
# # March 2023, formally add NEAMAP to prey decisions
# NEAMAPblueprey <- read.csv(here("fhdat/Full Prey List_Common Names.csv")) %>%
#   #filter(BLUEFISH > 9) %>%
#   filter(!SCIENTIFIC.NAME %in% c("Actinopterygii", "fish scales",
#                                  "Decapoda (megalope)", 
#                                  "unidentified material",
#                                  "Plantae",
#                                  "unidentified animal"))
# 
# NEAMAPprey <- NEAMAPblueprey %>%
#   dplyr::select(COMMON.NAME, SCIENTIFIC.NAME, BLUEFISH) %>%
#   dplyr::filter(!is.na(BLUEFISH)) %>%
#   dplyr::mutate(pynam2 = tolower(SCIENTIFIC.NAME),
#                 pynam2 = stringr::str_replace(pynam2, "spp.", "sp")) %>%
#   dplyr::rename(NEAMAP = BLUEFISH)
# 
# 
# NEFSCprey <- NEFSCblueprey %>%
#   dplyr::select(pycomnam2, pynam, BLUEFISH) %>%
#   dplyr::filter(!is.na(BLUEFISH)) %>%
#   dplyr::mutate(pynam2 = tolower(pynam)) %>%
#   dplyr::rename(NEFSC = BLUEFISH)
# 
# # new criteria March 2023, >20 observations NEAMAP+NEFSC, but keep mackerel
# # removes the flatfish order (too broad) and unid Urophycis previously in NEAMAP
# blueprey <- NEFSCprey %>% 
#   dplyr::full_join(NEAMAPprey) %>%
#   dplyr::mutate(NEAMAP = ifelse(is.na(NEAMAP), 0, NEAMAP),
#                 NEFSC = ifelse(is.na(NEFSC), 0, NEFSC),
#                 total = NEFSC + NEAMAP,
#                 PREY = ifelse(is.na(SCIENTIFIC.NAME), pynam, SCIENTIFIC.NAME),
#                 COMMON = ifelse(is.na(COMMON.NAME), pycomnam2, COMMON.NAME),
#                 pynam = ifelse(is.na(pynam), toupper(pynam2), pynam)) %>%
#   dplyr::arrange(desc(total)) %>%
#   dplyr::filter(total>20 | pynam=="SCOMBER SCOMBRUS") %>% # >20 leaves out mackerel
#   dplyr::mutate(COMMON = case_when(pynam=="ILLEX SP" ~ "Shortfin squids",
#                                    pynam2=="teuthida" ~ "Unidentified squids",
#                                    TRUE ~ COMMON)) %>%
#   dplyr::mutate(PREY = stringr::str_to_sentence(PREY),
#                 COMMON = stringr::str_to_sentence(COMMON))


fh.nefsc.pisc.pisccomplete.herring <- fh.nefsc.pisc.pisccomplete %>%
  mutate(herring = case_when(pynam %in% c("CLUPEA HARENGUS",
                                          "CLUPEIDAE",
                                          "CLUPEA HARENGUS SCALES",
                                          "CLUPEIDAE SCALES",
                                          "CLUPEA HARENGUS LARVAE",
                                          "CLUPEIDAE LARVAE")  ~ "herring",
                              TRUE ~ "othprey"))

###############################################################################
# Make the NEFSC dataset aggregating prey based on prey list

herringpyall_stn <- fh.nefsc.pisc.pisccomplete.herring %>%
  #create id linking cruise6_station
  #create season_ng spring and fall Spring=Jan-May, Fall=June-Dec
  mutate(id = paste0(cruise6, "_", station),
         year = as.numeric(year),
         month = as.numeric(month),
         season_ng = case_when(month <= 6 ~ "SPRING",
                               month >= 7 ~ "FALL",
                               TRUE ~ as.character(NA))
  ) %>%
  dplyr::select(year, season_ng, id, stratum,
                pynam, pyamtw, pywgti, pyvoli, herring, 
                pdcomnam, pdid, pdlen, pdsvol, pdswgt, 
                beglat, beglon, declat, declon, 
                bottemp, surftemp, setdepth) %>%
  group_by(id) %>%
  #mean blueprey g per stomach per tow: sum all blueprey g/n stomachs in tow
  mutate(herringwt = case_when(herring == "herring" ~ pyamtw,
                              TRUE ~ 0.0),
         herringpynam = case_when(herring == "herring" ~ pynam,
                               TRUE ~ NA_character_)) 

# Optional: save at prey disaggregated stage for paper
#saveRDS(bluepyall_stn, here("fhdat/bluepyall_stn.rds"))

# Now get station data in one line
stndat <- herringpyall_stn %>%
  dplyr::select(year, season_ng, id, 
                beglat, beglon, declat, declon, 
                bottemp, surftemp, setdepth) %>%
  distinct()

#pisc stomachs in tow count pdid for each pred and sum
piscstom <- herringpyall_stn %>%
  group_by(id, pdcomnam) %>%
  summarise(nstompd = n_distinct(pdid)) %>%
  group_by(id) %>%
  summarise(nstomtot = sum(nstompd))

#mean and var pred length per tow
pisclen <- herringpyall_stn %>%
  summarise(meanpisclen = mean(pdlen),
            varpisclen = var(pdlen))

# Aggregated prey at station level with predator covariates
herringpyagg_stn <- herringpyall_stn %>%
  summarise(sumherringwt = sum(herringwt),
            nherringsp = n_distinct(herringpynam, na.rm = T),
            npreysp = n_distinct(pynam),
            npiscsp = n_distinct(pdcomnam)) %>%
  left_join(piscstom) %>%
  mutate(meanherringwt = sumherringwt/nstomtot) %>%
  left_join(pisclen) %>%
  left_join(stndat)

# save at same stage as before, writing over old file
#saveRDS(bluepyagg_stn, here("fhdat/bluepyagg_stn.rds"))

# current dataset, fix declon, add vessel, rename NEFSC
#nefsc_bluepyagg_stn <- readRDS(here("fhdat/bluepyagg_stn.rds")) %>%
nefsc_herringpyagg_stn <- herringpyagg_stn %>%
  mutate(declon = -declon,
         vessel = case_when(year<2009 ~ "AL",
                            year>=2009 ~ "HB", 
                            TRUE ~ as.character(NA)))

##############################################################################
# Add NEAMAP to make full aggregated stomach dataset

# Read in NEAMAP updated input from Jim Gartland, reformat with same names
neamap_herringagg_stn <- read_csv(here("fhdat/NEAMAP_Mean stomach weights_Bluefish Prey_wWQ_AH Only.csv")) %>%  
  mutate(vessel = "NEAMAP") %>%
  rename(id = station,
         sumherringwt = sumbluepreywt,
         nherringsp = nbfpreyspp,
         #npreysp = ,
         npiscsp = npiscspp,
         nstomtot = nstomtot, 
         meanherringwt = meanbluepreywt,
         meanpisclen = meanpisclen.simple, 
         #varpisclen = ,
         season_ng = season,
         declat  = lat,
         declon = lon,
         bottemp = bWT,
         #surftemp = , 
         setdepth = depthm) 


# combine NEAMAP and NEFSC
herringagg_stn_all <-  nefsc_herringpyagg_stn %>%
  bind_rows(neamap_herringagg_stn) 

# Save before SST integration step
#saveRDS(bluepyagg_stn_all, here("fhdat/bluepyagg_stn_all.rds"))

###############################################################################
# Add SST into NEAMAP and reintegrate into full dataset

# Read back in if needed for SST
#bluepyagg_stn_all <- readRDS(here("fhdat/bluepyagg_stn_all.rds"))

NEFSCstations <- allfh %>%
  dplyr::mutate(id = paste0(cruise6, "_", station),
                year = as.numeric(year),
                month = as.numeric(month),
                day = as.numeric(day),
                declon = -declon) %>%
  dplyr::select(id, year, month, day, declat, declon) %>%
  dplyr::distinct()

# Need NEAMAP SST update! This is the old file
NEAMAPstationSST <- read.csv(here("fhdat/NEAMAP SST_2007_2022_Nov2023.csv"))

NEAMAPstations <- NEAMAPstationSST %>%
  dplyr::mutate(id = station,
                year = as.numeric(year),
                month = as.numeric(month),
                day = as.numeric(day)) %>%
  dplyr::select(id, year, month, day) %>%
  dplyr::distinct()

# remake diethauls
diethauls <- herringagg_stn_all %>%
  dplyr::select(id, declat, declon)

NEFSCstations <- dplyr::select(NEFSCstations, c(-declat, -declon))

Allstations <- bind_rows(NEFSCstations, NEAMAPstations)

#station id, lat lon, year month day
diethauls <- left_join(diethauls, Allstations)

#add year month day to diet data
herringagg_stn_all <- left_join(herringagg_stn_all, diethauls)

# add NEAMAP SST to surftemp field
NEAMAPidSST <- NEAMAPstationSST %>%
  mutate(id = station) %>%
  dplyr::select(id, SST)

herringagg_stn_all <- left_join(herringagg_stn_all, NEAMAPidSST, by="id") %>%
  mutate(surftemp = coalesce(surftemp, SST)) %>%
  dplyr::select(-SST)

# save merged dataset with day, month, and NEAMAP surftemp, same name
#saveRDS(bluepyagg_stn_all, here("fhdat/bluepyagg_stn_all.rds"))

###############################################################################
#Now match stations to OISST

#make an SST dataframe for 2022! Add to saved sst_data in data-raw/gridded

library(sf)
library(raster)
library(terra)
library(nngeo)

# already have all OISST data as dataframes on this repo

# # Bastille function from https://github.com/kimberly-bastille/ecopull/blob/main/R/utils.R
# 
# nc_to_raster <- function(nc,
#                          varname,
#                          extent = c(0, 360, -90, 90),
#                          crop = raster::extent(280, 300, 30, 50),
#                          show_images = FALSE) {
#   
#   message("Reading .nc as brick...")
#   
#   r <- raster::brick(nc, varname = varname)
#   
#   message("Setting CRS...")
#   raster::crs(r) <- "+proj=longlat +lat_1=35 +lat_2=45 +lat_0=40 +lon_0=-77 +x_0=0 +y_0=0 +datum=NAD83 +no_defs +ellps=GRS80 +towgs84=0,0,0"
#   
#   # not sure if this is necessary?
#   raster::extent(r) <- raster::extent(extent)
#   
#   if(show_images){
#     par(mfrow = c(1,2))
#     raster::plot(r, 1, sub = "Full dataset")
#   }
#   
#   message("Cropping data...")
#   ne_data <- raster::crop(r, crop)
#   #ne_data <- raster::rotate(ne_data) add here for future pulls
#   
#   if(show_images){
#     raster::plot(ne_data, 1, sub = "Cropped dataset")
#     par(mfrow = c(1,1))
#   }
#   
#   message("Done!")
#   
#   return(ne_data)
# }
# 
# # function to convert to dataframe based on
# # https://towardsdatascience.com/transforming-spatial-data-to-tabular-data-in-r-4dab139f311f
# 
# raster_to_sstdf <- function(brick,
#                             rotate=TRUE){
#   
#   if(rotate) brick_r <- raster::rotate(brick)
#   brick_r <- raster::crop(brick_r, raster::extent(-77,-65,35,45))
#   sstdf <- as.data.frame(raster::rasterToPoints(brick_r, spatial = TRUE))
#   sstdf <- sstdf %>%
#     dplyr::rename(Lon = x,
#                   Lat = y) %>%
#     tidyr::pivot_longer(cols = starts_with("X"),
#                         names_to = c("year", "month", "day"),
#                         names_prefix = "X",
#                         names_sep = "\\.",
#                         values_to = "sst",
#     )
#   return(sstdf)
# }
# 
# # pull the OISST data as raster brick, modified from 
# # https://github.com/kimberly-bastille/ecopull/blob/main/.github/workflows/pull_satellite_data.yml
# 
# varname <- "sst"
# 
# # 1985-2021 previously pulled, processed and stored. add 2022.
# # add 1981-1984 to extend back in time. No OISST before 1981.
# # 1981 is only Sept-Dec so don't use
# 
# years <- #1982:1984 # 2022
# for(i in years) {
#   name <- paste0(i, ".nc")
#   dir.create(here::here("data-raw","gridded", "sst_data"), recursive = TRUE)
#   filename <- here::here("data-raw","gridded", "sst_data", paste0("test_", i, ".grd"))
#   url <- paste0("https://downloads.psl.noaa.gov/Datasets/noaa.oisst.v2.highres/sst.day.mean.", i, ".nc")
#   download.file(url, destfile = name)
#   
#   text <- knitr::knit_expand(text = "test_{{year}} <- nc_to_raster(nc = name, varname = varname)
#                                      raster::writeRaster(test_{{year}}, filename = filename, overwrite=TRUE)",
#                              year = i)
#   print(text)
#   try(eval(parse(text = text)))
#   unlink(name) # remove nc file to save space
#   print(paste("finished",i))
# }
# 
# 
# # convert raster to dataframe
# #years <- 2022
# for(i in years) {
#   name <- get(paste0("test_",i))
#   filename <- here::here("data-raw","gridded", "sst_data", paste0("sst", i, ".rds"))
#   text <- knitr::knit_expand(text = "sst{{year}} <- raster_to_sstdf(brick = name)
#                                      saveRDS(sst{{year}}, filename)",
#                              year = i)
#   print(text)
#   try(eval(parse(text = text)))
#}
#read in diet data with month day fields

#bluepyagg_stn_all <- readRDS(here("fhdat/bluepyagg_stn_all.rds"))

stations <- herringagg_stn_all %>%
  dplyr::mutate(day = str_pad(day, 2, pad='0'),
                month = str_pad(month, 2, pad='0'),
                yrmody = as.numeric(paste0(year, month, day))) %>%
  dplyr::select(id, declon, declat, year, yrmody) %>%
  na.omit() %>%
  sf::st_as_sf(coords=c("declon","declat"), crs=4326, remove=FALSE) 



#list of SST dataframes
SSTdfs <- list.files(here("data-raw/gridded/sst_data/"), pattern = "*.rds")

dietstn_OISST <- tibble()


for(df in SSTdfs){
  sstdf <- readRDS(paste0(here("data-raw/gridded/sst_data/", df)))
  
  # keep only bluefish dates in SST year
  stationsyr <- stations %>%
    filter(year == unique(sstdf$year))
  
  # keep only sst days in bluefish dataset
  sstdf_survdays <- sstdf %>%
    dplyr::mutate(yrmody = as.numeric(paste0(year, month, day)) )%>%
    dplyr::filter(yrmody %in% unique(stationsyr$yrmody)) %>%
    dplyr::mutate(year = as.numeric(year),
                  month = as.numeric(month),
                  day = as.numeric(day),
                  declon = Lon,
                  declat = Lat) %>%
    dplyr::select(-Lon, -Lat) %>%
    sf::st_as_sf(coords=c("declon","declat"), crs=4326, remove=FALSE)  
  
  # now join by nearest neighbor and date
  
  #https://stackoverflow.com/questions/71959927/spatial-join-two-data-frames-by-nearest-feature-and-date-in-r      
  
  yrdietOISST <- do.call('rbind', lapply(split(stationsyr, 1:nrow(stationsyr)), function(x) {
    sf::st_join(x, sstdf_survdays[sstdf_survdays$yrmody == unique(x$yrmody),],
                #join = st_nearest_feature
                join = st_nn, k = 1, progress = FALSE
    )
  }))
  
  #   #datatable solution--works but doesnt seem faster?
  #    df1 <- data.table(stationsyr)
  #   
  #  .nearest_samedate <- function(x) {
  #    st_join(st_as_sf(x), sstdf_survdays[sstdf_survdays$yrmody == unique(x$yrmody),], join = st_nearest_feature)
  #  }
  # # 
  #  yrdietOISST <- df1[, .nearest_samedate(.SD), by = list(1:nrow(df1))]
  
  dietstn_OISST <- rbind(dietstn_OISST, yrdietOISST)
  
}

#saveRDS(dietstn_OISST, here("data-raw/dietstn_OISST.rds"))

# Now join with OISST dataset

#bluepyagg_stn_all <- readRDS(here("fhdat/bluepyagg_stn_all.rds"))
#dietstn_OISST <- readRDS(here("data-raw/dietstn_OISST.rds"))


dietstn_OISST_merge <- dietstn_OISST %>%
  dplyr::rename(declon = declon.x,
                declat = declat.x,
                year = year.x,
                oisst = sst) %>%
  dplyr::select(id, oisst) %>%
  sf::st_drop_geometry()

herringagg_stn_all_OISST <- left_join(herringagg_stn_all, dietstn_OISST_merge)

saveRDS(herringagg_stn_all_OISST, here("fhdat/herringagg_stn_all_OISST_1982-2022.rds"))

VAST models were run using the script VASTunivariate_seasonalherringindex.R in the folder VASTscripts: https://github.com/NOAA-EDAB/forageindex/blob/main/VASTscripts/VASTunivariate_seasonalherringindex.R

# This is the exact VAST code used in Gaichas et al 2023, but with data years 
# extended from 1973-2022
# VAST attempt 2 univariate model as a script
# modified from https://github.com/James-Thorson-NOAA/VAST/wiki/Index-standardization

# Load packages
library(here)
library(dplyr)
library(VAST)

#Read in data, separate spring and fall, and rename columns for VAST:

# this dataset created in SSTmethods.Rmd

herringagg_stn <- readRDS(here::here("fhdat/herringagg_stn_all_OISST_1982-2022.rds"))

# make SST column that uses surftemp unless missing or 0
# there are 3 surftemp 0 values in the dataset, all with oisst > 15
herringagg_stn <- herringagg_stn %>%
  dplyr::mutate(sstfill = ifelse((is.na(surftemp)|surftemp==0), oisst, surftemp))

#herringagg_stn <- bluepyagg_pred_stn

# filter to assessment years at Tony's suggestion

# code Vessel as AL=0, HB=1, NEAMAP=2

herringagg_stn_fall <- herringagg_stn %>%
  #ungroup() %>%
  filter(season_ng == "FALL") |>
        #,year > 1984) %>%
  mutate(AreaSwept_km2 = 1, #Elizabeth's code
         #declon = -declon already done before neamap merge
         Vessel = as.numeric(as.factor(vessel))-1
         ) %>% 
  dplyr::select(Catch_g = meanherringwt, #use bluepywt for individuals input in example
         Year = year,
         Vessel,
         AreaSwept_km2,
         Lat = declat,
         Lon = declon,
         meanpisclen,
         npiscsp,
         #bottemp, #this leaves out many stations for NEFSC
         #surftemp, #this leaves out many stations for NEFSC
         #oisst, #leaves out everything before 1982
         sstfill
         ) %>%
  na.omit() %>%
  as.data.frame()

herringagg_stn_spring <- herringagg_stn %>%
  filter(season_ng == "SPRING") |>
        #,year > 1984) %>%
  mutate(AreaSwept_km2 =1, #Elizabeth's code
         #declon = -declon already done before neamap merge
         Vessel = as.numeric(as.factor(vessel))-1
         ) %>% 
  dplyr::select(Catch_g = meanherringwt,
         Year = year,
         Vessel,
         AreaSwept_km2,
         Lat = declat,
         Lon = declon,
         meanpisclen,
         npiscsp,
         #bottemp, #this leaves out many stations for NEFSC
         #surftemp, #this leaves out many stations for NEFSC
         #oisst, #leaves out everything before 1982
         sstfill
         ) %>%
  na.omit() %>%
  as.data.frame()


# Make settings (turning off bias.correct to save time for example)
# NEFSC strata limits https://github.com/James-Thorson-NOAA/VAST/issues/302

# use only MAB, GB, GOM, SS EPUs 
# leave out south of Cape Hatteras at Elizabeth's suggestion
# could also leave out SS?
# CHECK if these EPUs match what we use in SOE

MAB <- c(1010:1080, 1100:1120, 1600:1750, 3010:3450, 3470, 3500, 3510)
GB  <- c(1090, 1130:1210, 1230, 1250, 3460, 3480, 3490, 3520:3550)
GOM <- c(1220, 1240, 1260:1290, 1360:1400, 3560:3830)
SS  <- c(1300:1352, 3840:3990)

coast3nmbuffst <- readRDS(here::here("spatialdat/coast3nmbuffst.rds"))

# Mid Atlantic
MAB2 <- coast3nmbuffst %>% 
  dplyr::filter(stratum_number %in% MAB) %>%
  dplyr::select(stratum_number2) %>%
  dplyr::distinct()

# Georges Bank EPU
GB2 <- coast3nmbuffst %>% 
  dplyr::filter(stratum_number %in% GB) %>%
  dplyr::select(stratum_number2) %>%
  dplyr::distinct()

# gulf of maine EPU (for SOE)
GOM2 <- coast3nmbuffst %>%
  dplyr::filter(stratum_number %in% GOM) %>%
  dplyr::select(stratum_number2) %>%
  dplyr::distinct()

# scotian shelf EPU (for SOE)
SS2 <- coast3nmbuffst %>%
  dplyr::filter(stratum_number %in% SS) %>%
  dplyr::select(stratum_number2) %>%
  dplyr::distinct()


# needed to cover the whole northwest atlantic grid
allother2 <- coast3nmbuffst %>%
  dplyr::filter(!stratum_number %in% c(MAB, GB, GOM, SS)) %>%
  dplyr::select(stratum_number2) %>%
  dplyr::distinct()

# all epus
allEPU2 <- coast3nmbuffst %>%
  dplyr::filter(stratum_number %in% c(MAB, GB, GOM, SS)) %>%
  dplyr::select(stratum_number2) %>%
  dplyr::distinct()

# # configs same FieldConfig as below but formatted differently
# FieldConfig <- c(
#   "Omega1"   = 0,   # number of spatial variation factors (0, 1, AR1)
#   "Epsilon1" = 0,   # number of spatio-temporal factors
#   "Omega2"   = 0, 
#   "Epsilon2" = 0
# ) 

# default configs, not really specified anyway
FieldConfig = matrix( "IID", ncol=2, nrow=3, 
                      dimnames=list(c("Omega","Epsilon","Beta"),c("Component_1","Component_2")))


RhoConfig <- c(
  "Beta1" = 0,      # temporal structure on years (intercepts) 
  "Beta2" = 0, 
  "Epsilon1" = 0,   # temporal structure on spatio-temporal variation
  "Epsilon2" = 0
) 
# 0 off (fixed effects)
# 1 independent
# 2 random walk
# 3 constant among years (fixed effect)
# 4 AR1

OverdispersionConfig    <- c("eta1"=0, "eta2"=0)
# eta1 = vessel effects on prey encounter rate
# eta2 = vessel effects on prey weight

strata.limits <- as.list(c("AllEPU" = allEPU2, 
                           "MAB" = MAB2,
                           "GB" = GB2,
                           "GOM" = GOM2,
                           "allother" = allother2))

settings = make_settings( n_x = 500, 
                          Region = "northwest_atlantic",
                          Version = "VAST_v14_0_1", #needed to prevent error from newer dev version number
                          #strata.limits = list('All_areas' = 1:1e5), full area
                          strata.limits = strata.limits,
                          purpose = "index2", 
                          bias.correct = TRUE,
                          #use_anisotropy = FALSE,
                          #fine_scale = FALSE,
                          #FieldConfig = FieldConfig,
                          #RhoConfig = RhoConfig,
                          OverdispersionConfig = OverdispersionConfig
                          )


New_Extrapolation_List <- readRDS(here::here("spatialdat/CustomExtrapolationList.rds"))

# select dataset and set directory for output

#########################################################
# Run model fall

season <- c("fall_500_lennosst_ALLsplit_biascorrect")

working_dir <- here::here(sprintf("herringpyindex/allagg_%s/", season))

if(!dir.exists(working_dir)) {
  dir.create(working_dir)
}

# subset for faster testing
#herringagg_stn_fall <- herringagg_stn_fall %>% filter(Year<1990)                       

fit <- fit_model(
  settings = settings, 
  extrapolation_list = New_Extrapolation_List,
  Lat_i = herringagg_stn_fall$Lat, 
  Lon_i = herringagg_stn_fall$Lon, 
  t_i = herringagg_stn_fall$Year, 
  b_i = as_units(herringagg_stn_fall[,'Catch_g'], 'g'),
  a_i = rep(1, nrow(herringagg_stn_fall)),
  v_i = herringagg_stn_fall$Vessel,
  Q_ik = as.matrix(herringagg_stn_fall[,c("npiscsp", 
                                         "meanpisclen", 
                                         "sstfill"
                                         )]),
  #Use_REML = TRUE,
  working_dir = paste0(working_dir, "/"))

saveRDS(fit, file = paste0(working_dir, "/fit.rds"))

# Plot results
plot( fit,
      working_dir = paste0(working_dir, "/"))

######################################################
# Run model spring

season <- c("spring_500_lennosst_ALLsplit_biascorrect")

working_dir <- here::here(sprintf("herringpyindex/allagg_%s/", season))

if(!dir.exists(working_dir)) {
  dir.create(working_dir)
}                         
  
# subset for faster testing
#herringagg_stn_spring <- herringagg_stn_spring %>% filter(Year<1990) 

fit <- fit_model( settings = settings,  
                 extrapolation_list = New_Extrapolation_List,
                 Lat_i = herringagg_stn_spring[,'Lat'], 
                 Lon_i = herringagg_stn_spring[,'Lon'], 
                 t_i = herringagg_stn_spring[,'Year'], 
                 b_i = as_units(herringagg_stn_spring[,'Catch_g'], 'g'), 
                 a_i = rep(1, nrow(herringagg_stn_spring)),
                 v_i = herringagg_stn_spring$Vessel,
                 Q_ik = as.matrix(herringagg_stn_spring[,c("npiscsp", 
                                                          "meanpisclen", 
                                                          "sstfill"
                                                          )]),
                # Use_REML = TRUE,
                 working_dir = paste0(working_dir, "/"))

saveRDS(fit, file = paste0(working_dir, "/fit.rds"))

# Plot results
plot( fit,
      working_dir = paste0(working_dir, "/")) 

3 Results

3.1 How many piscivore locations had herring?

Not that many. Seems a spring model will run anyway using the forage index configuration but a fall model will not.

herringagg_stn <- readRDS(here::here("fhdat/herringagg_stn_all_OISST_1982-2022.rds"))

# make SST column that uses surftemp unless missing or 0
# there are 3 surftemp 0 values in the dataset, all with oisst > 15
herringagg_stn <- herringagg_stn %>%
  dplyr::mutate(sstfill = ifelse((is.na(surftemp)|surftemp==0), oisst, surftemp))

#herringagg_stn <- bluepyagg_pred_stn

# filter to assessment years at Tony's suggestion

# code Vessel as AL=0, HB=1, NEAMAP=2

herringagg_stn_all <- herringagg_stn %>%
  #ungroup() %>%
  #filter(season_ng == "FALL") |>
  #,year > 1984) %>%
  mutate(AreaSwept_km2 = 1, #Elizabeth's code
         #declon = -declon already done before neamap merge
         Vessel = as.numeric(as.factor(vessel))-1
  ) %>% 
  dplyr::select(Catch_g = meanherringwt, #use bluepywt for individuals input in example
                Year = year,
                Vessel,
                AreaSwept_km2,
                Lat = declat,
                Lon = declon,
                meanpisclen,
                npiscsp,
                #bottemp, #this leaves out many stations for NEFSC
                #surftemp, #this leaves out many stations for NEFSC
                #oisst, #leaves out everything before 1982
                sstfill,
                season_ng
  ) %>%
  na.omit() %>%
  as.data.frame()

nonzeroherring <- herringagg_stn_all |>
  dplyr::filter(Catch_g > 0) |> 
  dplyr::group_by(season_ng, Year) |> 
  dplyr::summarise(n = dplyr::n()) |>
  tidyr::pivot_wider(names_from = "season_ng", names_prefix = "Herring", values_from = "n")

tows <- herringagg_stn_all |>
  dplyr::group_by(season_ng, Year) |> 
  dplyr::summarise(n = dplyr::n()) |>
  tidyr::pivot_wider(names_from = "season_ng", names_prefix = "Total", values_from = "n")

pisctows <- dplyr::left_join(tows, nonzeroherring) |>
  dplyr::select(Year, TotalSPRING, HerringSPRING, TotalFALL, HerringFALL)

tab1 <- flextable::flextable(pisctows) |>
  flextable::set_caption("Number of piscivore stomach locations by season and year (TotalSPRING and TotalFALL), and number of piscivore stomach locations having nonzero herring prey (HerringSPRING and HerringFALL).") 

flextable::colformat_double(tab1, big.mark = "", digits = 0)

3.2 Fall model using forage index setup: failed

About the same number of positive observations in spring and fall.

List of estimated fixed and random effects:
    Coefficient_name Number_of_coefficients   Type
1           beta1_ft                     48  Fixed
2           beta2_ft                     48  Fixed
3       L_epsilon1_z                      1  Fixed
4       L_epsilon2_z                      1  Fixed
5         L_omega1_z                      1  Fixed
6         L_omega2_z                      1  Fixed
7          lambda1_k                      3  Fixed
8          lambda2_k                      3  Fixed
9         ln_H_input                      2  Fixed
10         logkappa1                      1  Fixed
11         logkappa2                      1  Fixed
12         logSigmaM                      1  Fixed
13 Epsiloninput1_sff                  27750 Random
14 Epsiloninput2_sff                  27750 Random
15    Omegainput1_sf                    555 Random
16    Omegainput2_sf                    555 Random

### Checking model at initial values
All fixed effects have a nonzero gradient

... After 167 iterations...

The following parameters appear to be approaching zero:
          Param starting_value Lower           MLE Upper final_gradient
55 L_epsilon1_z              1  -Inf -8.171138e-07   Inf  -7.579271e-05
Please turn off factor-model variance parameters `L_` that are approaching zero and re-run the model

Error: Please change model structure to avoid problems with parameter estimates and then re-try; see details in `?check_fit`
In addition: Warning message:
In nlminb(start = startpar, objective = fn, gradient = gr, control = nlminb.control,  :
  NA/NaN function evaluation

Trying spring model

3.3 Spring results using forage index setup: ran

Check diagnostics.

Abundance index:

SOEinputs <- function(infile, season, outfile) {
  
  splitoutput <- read.csv(infile)
  
  # warning, hardcoded. obviously
  stratlook <- data.frame(Stratum = c("Stratum_1",
                                      "Stratum_2",
                                      "Stratum_3",
                                      "Stratum_4",
                                      "Stratum_5",
                                      "Stratum_6",
                                      "Stratum_7",
                                      "Stratum_8",
                                      "Stratum_9",
                                      "Stratum_10",
                                      "Stratum_11",
                                      "Stratum_12",
                                      "Stratum_13",
                                      "Stratum_14",
                                      "Stratum_15"),
                          Region  = c("AllEPU", 
                                      "MABGB", 
                                      "MABGBstate", 
                                      "MABGBfed", 
                                      "MAB",
                                      "GB",
                                      "GOM",
                                      "bfall",
                                      "bfin",
                                      "bfoff",
                                      "MABGBalbinshore",
                                      "MABGBothoffshore",
                                      "albbfin",
                                      "albbfall",
                                      "allother"))
  
  forageindex <- splitoutput %>%
    left_join(stratlook) %>%
    dplyr::select(Time, 
                  EPU = Region, 
                  "Forage Fish Biomass Estimate" = Estimate, 
                  "Forage Fish Biomass Estimate SE" = Std..Error.for.Estimate) %>%
    tidyr::pivot_longer(c("Forage Fish Biomass Estimate", "Forage Fish Biomass Estimate SE"), 
                        names_to = "Var", values_to = "Value") %>%
    dplyr::filter(EPU %in% c("MAB", "GB", "GOM", "AllEPU")) %>%
    dplyr::mutate(Units = "relative grams per stomach") %>%
    dplyr::select(Time, Var, Value, EPU, Units)
  
  forageindex$Var <- stringr::str_c(season, forageindex$Var, sep = " ")
  
  #readr::write_csv(forageindex, outfile)
  saveRDS(forageindex, outfile)
  
} 


# make data files

SOEinputs(infile = here::here("herringpyindex/allagg_spring_500_lennosst_ALLsplit_biascorrect/Index.csv"),
          season = "Spring", 
           outfile = here::here("herringpyindex/springherringindex.rds"))


forageindex <- readRDS(here::here("herringpyindex/springherringindex.rds"))

# test plot
foragewide <- forageindex %>%
  pivot_wider(names_from = Var, values_from = Value)


ggplot(foragewide, aes(x=Time, y=`Spring Forage Fish Biomass Estimate`, colour = EPU)) +
  geom_errorbar(aes(ymin=`Spring Forage Fish Biomass Estimate`+`Spring Forage Fish Biomass Estimate SE`,
                    ymax=`Spring Forage Fish Biomass Estimate`-`Spring Forage Fish Biomass Estimate SE`))+
  geom_point()+
  geom_line()

Log density pattern:

knitr::include_graphics(here::here("herringpyindex/allagg_spring_500_lennosst_ALLsplit_biascorrect/ln_density-predicted.png"))

Spatial residuals:

knitr::include_graphics(here::here("herringpyindex/allagg_spring_500_lennosst_ALLsplit_biascorrect/quantile_residuals_on_map.png"))

How much data in spring?

knitr::include_graphics(here::here("herringpyindex/allagg_spring_500_lennosst_ALLsplit_biascorrect/Data_by_year.png"))

3.4 Model selection

Same initial as forage index to see if spatial RE, spatio-temporal RE, and aniso make sense.

Comparisons of AIC are presented for both the first (spatial and spatio-temporal random effects, Table Sb3.2) and second (catchability covariates, Table Sb3.3) steps of model selection.

# from each output folder in pyindex, 
outdir <- here::here("herringpyindex/modsel1")
moddirs <- list.dirs(outdir) 
moddirs <- moddirs[-1]
# keep folder name
modnames <- list.dirs(outdir, full.names = FALSE)


# function to apply extracting info
getmodinfo <- function(d.name){
  # read settings
  modpath <- stringr::str_split(d.name, "/", simplify = TRUE)
  modname <- modpath[length(modpath)]
  
  settings <- read.table(file.path(d.name, "settings.txt"), comment.char = "",
    fill = TRUE, header = FALSE)
  
  n_x <- as.numeric(as.character(settings[(which(settings[,1]=="$n_x")+1),2]))
  grid_size_km <- as.numeric(as.character(settings[(which(settings[,1]=="$grid_size_km")+1),2]))
  max_cells <- as.numeric(as.character(settings[(which(settings[,1]=="$max_cells")+1),2]))
  use_anisotropy <- as.character(settings[(which(settings[,1]=="$use_anisotropy")+1),2])
  fine_scale <- as.character(settings[(which(settings[,1]=="$fine_scale")+1),2])
  bias.correct <- as.character(settings[(which(settings[,1]=="$bias.correct")+1),2])
  
  #FieldConfig
  if(settings[(which(settings[,1]=="$FieldConfig")+1),1]=="Component_1"){
    omega1 <- as.character(settings[(which(settings[,1]=="$FieldConfig")+2),2])
    omega2 <- as.character(settings[(which(settings[,1]=="$FieldConfig")+3),1])
    epsilon1 <- as.character(settings[(which(settings[,1]=="$FieldConfig")+4),2])
    epsilon2 <- as.character(settings[(which(settings[,1]=="$FieldConfig")+5),1])
    beta1 <- as.character(settings[(which(settings[,1]=="$FieldConfig")+6),2])
    beta2 <- as.character(settings[(which(settings[,1]=="$FieldConfig")+7),1])
  }
  
  if(settings[(which(settings[,1]=="$FieldConfig")+1),1]=="Omega1"){
    omega1 <- as.character(settings[(which(settings[,1]=="$FieldConfig")+3),1])
    omega2 <- as.character(settings[(which(settings[,1]=="$FieldConfig")+4),1])
    epsilon1 <- as.character(settings[(which(settings[,1]=="$FieldConfig")+3),2])
    epsilon2 <- as.character(settings[(which(settings[,1]=="$FieldConfig")+4),2])
    beta1 <- "IID"
    beta2 <- "IID"
  }
  
  
  #RhoConfig
  rho_beta1 <- as.numeric(as.character(settings[(which(settings[,1]=="$RhoConfig")+3),1]))
  rho_beta2 <- as.numeric(as.character(settings[(which(settings[,1]=="$RhoConfig")+3),2]))
  rho_epsilon1 <- as.numeric(as.character(settings[(which(settings[,1]=="$RhoConfig")+4),1]))
  rho_epsilon2 <- as.numeric(as.character(settings[(which(settings[,1]=="$RhoConfig")+4),2]))
  
  # read parameter estimates, object is called parameter_Estimates
  if(file.exists(file.path(d.name, "parameter_estimates.RData"))) {
    load(file.path(d.name, "parameter_estimates.RData"))
    
    AIC <- parameter_estimates$AIC[1]  
    converged <- parameter_estimates$Convergence_check[1]
    fixedcoeff <- unname(parameter_estimates$number_of_coefficients[2])
    randomcoeff <- unname(parameter_estimates$number_of_coefficients[3])
    
  }else if(file.exists(file.path(d.name, "parameter_estimates.txt"))){
    
    reptext <- readLines(file.path(d.name, "parameter_estimates.txt"))
    
    AIC <- as.double(reptext[grep(reptext, pattern = "AIC")+1])
    converged <- reptext[grep(reptext, pattern = "Convergence_check")+1]
    fixedcoeff <- as.integer(stringr::str_split(reptext[grep(reptext, pattern = "number_of_coefficients")+2], 
                                     boundary("word"))[[1]][2])
    randomcoeff <- as.integer(stringr::str_split(reptext[grep(reptext, pattern = "number_of_coefficients")+2], 
                                     boundary("word"))[[1]][3])
    
  }else{
    
    AIC <- NA_real_
    converged <- NA_character_
    fixedcoeff <- NA_integer_
    randomcoeff <- NA_integer_
  }
  
  
  #index <- read.csv(file.path(d.name, "Index.csv"))
  
  
  # return model attributes as a dataframe
  out <- data.frame(modname = modname,
                    n_x = n_x,
                    grid_size_km = grid_size_km,
                    max_cells = max_cells,
                    use_anisotropy = use_anisotropy,
                    fine_scale =  fine_scale,
                    bias.correct = bias.correct,
                    omega1 = omega1,
                    omega2 = omega2,
                    epsilon1 = epsilon1,
                    epsilon2 = epsilon2,
                    beta1 = beta1,
                    beta2 = beta2,
                    rho_epsilon1 = rho_epsilon1,
                    rho_epsilon2 = rho_epsilon2,
                    rho_beta1 = rho_beta1,
                    rho_beta2 = rho_beta2,
                    AIC = AIC,
                    converged = converged,
                    fixedcoeff = fixedcoeff,
                    randomcoeff = randomcoeff#,
                    #index = index
  )
    
    return(out)

}


modcompare <- purrr::map_dfr(moddirs, getmodinfo)

Models compared using REML are identified by model name (“modname” in Table Sb3.2) which always includes all prey aggregated, season (“all” for annual models of months 1-12, “fall” for models of months 7-12, and “spring” for models of months 1-6), number of knots (500 for all models), and which fixed and random spatial and spatio-temporal effects were included in which linear predictor (1 or 2). The names for model options and associated VAST model settings are:

Model selection 1 (spatial, spatio-temporal effects, no covariates) options (following Ng et al. (2021)) and naming: * All models set Use_REML = TRUE in fit_model specifications.
* Modeled effects, model name suffix, and VAST settings by model:

1. “_alleffectson” = Spatial and spatio-temporal random effects plus anisotropy in both linear predictors; FieldConfig default (all IID)
2. “_noaniso” = Spatial and spatio-temporal random effects in both linear predictors with anisotopy turned off; FieldConfig default (all IID) and use_anisotopy = FALSE
3. “_noomeps2” = Spatial and spatio-temporal random effects plus anisotropy only in linear predictor 1; FieldConfig = 0 for Omega2, Epsilon2
4. “_noomeps2_noaniso” = Spatial and spatio-temporal random effects only in linear predictor 1 with anisotopy turned off; FieldConfig = 0 for Omega2, Epsilon2 and use_anisotopy = FALSE
5. “_noomeps2_noeps1” = Spatial random effects plus anisotropy only in linear predictor 1; FieldConfig = 0 for Omega2, Epsilon2, Epsilon1
6. “_noomeps2_noeps1_noaniso” = Spatial random effects only in linear predictor 1 with anisotopy turned off; FieldConfig = 0 for Omega2, Epsilon2, Epsilon1 and use_anisotopy = FALSE
7. “_noomeps12” = Anisotropy, but no spatial or spatio-temporal random effects in either linear predictor; FieldConfig = 0 for both Omega, Epsilon
8. “_noomeps12_noaniso” = No spatial or spatio-temporal random effects in either linear predictor; FieldConfig = 0 for both Omega, Epsilon and use_anisotopy = FALSE

modselect <- modcompare %>%
  dplyr::mutate(season = case_when(str_detect(modname, "_fall_") ~ "Fall",
                            str_detect(modname, "spring") ~ "Spring",
                            str_detect(modname, "_all_") ~ "Annual",
                            TRUE ~ as.character(NA))) %>%
  dplyr::mutate(converged2 = case_when(str_detect(converged, "no evidence") ~ "likely",
                                str_detect(converged, "is likely not") ~ "unlikely",
                                TRUE ~ as.character(NA))) %>%
  dplyr::mutate(modname = str_extract(modname, '(?<=allagg_).*')) |>
  dplyr::group_by(season) %>%
  dplyr::mutate(deltaAIC = AIC-min(AIC, na.rm = TRUE)) %>%
  dplyr::select(modname, season, deltaAIC, fixedcoeff,
         randomcoeff, use_anisotropy, 
         omega1, omega2, epsilon1, epsilon2, 
         beta1, beta2, AIC, converged2) %>%
  dplyr::arrange(season, AIC)

# DT::datatable(modselect, rownames = FALSE, 
#               options= list(pageLength = 25, scrollX = TRUE),
#               caption = "Comparison of delta AIC values using Restricted Maxiumum Likelihood (REML) for alternative fixed and random effects model structures. See text for model descriptions.")

flextable::flextable(modselect %>%
                       dplyr::select(-c(use_anisotropy, 
         omega1, omega2, epsilon1, epsilon2, 
         beta1, beta2))) %>%
  flextable::set_header_labels(modname = "Model name",
                               season = "Season",
                               deltaAIC = "dAIC",
                               fixedcoeff = "N fixed",
                               randomcoeff = "N random",
                               converged2 = "Convergence") |>
  flextable::set_caption("Comparison of delta AIC (dAIC) values using Restricted Maxiumum Likelihood (REML) for alternative fixed and random effects model structures, with number of fixed (N fixed) and random (N random) coefficients. See text for model descriptions associated with each model name.") %>%
  flextable::fontsize(size = 9, part = "all") %>%
  flextable::colformat_double(digits = 2) |>
  flextable::set_table_properties(layout = "autofit", width = 1)

The annual model without anisotopy and with spatial and spatio-temporal random effects only in linear predictor 1 errored out; all other models completed and converged.

Spatial and spatiotemporal random effects were supported for both linear predictors for annual, fall, and spring models. For annual and spring models, estimating anisotropy (directional correlation) was supported as well, but the dAIC for the fall model suggests that estimating anisotriopy is only slightly more supported than not estimating it.

For now lets estimate spatial and spatiotemporal random effects with anisotropy for all models to test the catchability covariates.

For the second step of model selection with different combinations of vessel effects and or catchability covariates, “modname” in Table Sb3.3 follows a similar pattern as above, with all prey aggregated (“allagg” for all models), season (“all” for annual models of months 1-12, “fall” for models of months 7-12, and “spring” for models of months 1-6), number of knots (500 for all models), and which vessel effects or covariates were included. The names for model options and associated VAST model settings are:

Model selection 2 (covariates) options, FieldConfig default (all IID), with anisotropy:

1. “_base” = No vessel overdispersion or length/number covariates
   
2. “_len” = Predator mean length covariate
3. “_num” = Number of predator species covariate
4. “_lennum” = Predator mean length and number of predator species covariates
5. “_sst” = Combined in situ and OISST covariate
6. “_lensst” = Predator mean length and SST covariates
7. “_numsst” = Number of predator species and SST covariates
8. “_lennumsst” = Predator mean length, number of predator species, and SST covariates
9. “_eta10” = Overdispersion (vessel effect) in first linear predictor (prey encounter)
10. “_eta11” = Overdispersion (vessel effect) in both linear predictors (prey encounter and weight)

Strangely, the fall models all errored out with the same error as above, parameter going to 0. Even the same configuration that worked using REML above.

# from each output folder in pyindex, 
outdir <- here("herringpyindex/modsel2")
moddirs <- list.dirs(outdir) 
moddirs <- moddirs[-1]
# keep folder name
modnames <- list.dirs(outdir, full.names = FALSE)

modcompare2 <- purrr::map_dfr(moddirs, getmodinfo)

modselect2 <- modcompare2 %>%
  dplyr::mutate(season = case_when(str_detect(modname, "_fall_") ~ "Fall",
                            str_detect(modname, "spring") ~ "Spring",
                            str_detect(modname, "_all_") ~ "Annual",
                            TRUE ~ as.character(NA))) %>%
  dplyr::mutate(converged2 = case_when(str_detect(converged, "no evidence") ~ "likely",
                                str_detect(converged, "is likely not") ~ "unlikely",
                                TRUE ~ as.character(NA))) %>%
  dplyr::group_by(season) %>%
  dplyr::mutate(deltaAIC = AIC-min(AIC)) %>%
  dplyr::select(modname, season, deltaAIC, fixedcoeff,
         randomcoeff, use_anisotropy, 
         omega1, omega2, epsilon1, epsilon2, 
         beta1, beta2, AIC, converged2) %>%
  dplyr::arrange(season, AIC)

# DT::datatable(modselect2, rownames = FALSE, 
#               options= list(pageLength = 50, scrollX = TRUE),
#               caption = "Comparison of delta AIC values for alternative vessel effects and catchability covariates. See text for model descriptions.")

flextable::flextable(modselect2%>%
                       dplyr::select(-c(use_anisotropy, 
         omega1, omega2, epsilon1, epsilon2, 
         beta1, beta2))) %>%
  flextable::set_header_labels(modname = "Model name",
                               season = "Season",
                               deltaAIC = "dAIC",
                               fixedcoeff = "N fixed coefficients",
                               randomcoeff = "N random coefficients",
                               converged2 = "Convergence") |>
  flextable::set_caption("Comparison of delta AIC (dAIC) values for alternative vessel effects and catchability covariates. See text for model descriptions associated with each model name.") %>%
  flextable::fontsize(size = 9, part = "all") %>%
  flextable::colformat_double(digits = 2) |>
  flextable::set_table_properties(layout = "autofit", width = .9)

For the annual and spring models, the same set of covariates from the full forage model were best supported by the data: number of predator species at a station, mean length of predators at a station, and SST at the station.

So it looks like the spring model above, which is bias corrected, is the preferred model based on model selection.

Need to sort out what is going on with fall.

References

Gaichas, Sarah K., James Gartland, Brian E. Smith, Anthony D. Wood, Elizabeth L. Ng, Michael Celestino, Katie Drew, Abigail S. Tyrell, and James T. Thorson. 2023. “Assessing Small Pelagic Fish Trends in Space and Time Using Piscivore Stomach Contents.” Canadian Journal of Fisheries and Aquatic Sciences, October, cjfas-2023-0093. https://doi.org/10.1139/cjfas-2023-0093.

Ng, Elizabeth L, Jonathan J Deroba, Timothy E Essington, Arnaud Grüss, Brian E Smith, and James T Thorson. 2021. “Predator Stomach Contents Can Provide Accurate Indices of Prey Biomass.” ICES Journal of Marine Science 78 (3): 1146–59. https://doi.org/10.1093/icesjms/fsab026.