Tidy Tuesday: Bird Sightings at Sea

tidytuesday
R
birds
ecology
ocean
Which Southern Ocean seabirds chase ships, and which ones just fly on by? A look at three decades of pelagic bird observations from research vessels in the Australasian seas.
Author

Sean Thimons

Published

April 14, 2026

Preface

From TidyTuesday repository.

Between 1969 and 1990, naturalists aboard New Zealand and Australian research vessels recorded every seabird they could spot from the deck, along with the ship’s position, weather, sea state, and what the birds were doing. The result is a remarkable archive of pelagic seabird behavior in the Southern Ocean — albatrosses, petrels, shearwaters, and gulls — across more than 12,000 ship records and 49,000 individual sightings. Each row tells you not just what was seen, but whether it was following the ship, sitting on the water, flying past, or feeding.

The four tables in this week’s release split the work neatly: ships carries the position/time/weather context, birds carries the sightings keyed back to each ship record, and the two small lookup tables encode the Beaufort wind scale and World Meteorological Organization sea-state class.

Loading necessary packages

My handy booster pack that allows me to install (if needed) and load my usual and favorite packages, as well as some helpful functions.

Code 
{
  if (!requireNamespace("pak", quietly = TRUE)) {
    install.packages("pak", repos = "https://cloud.r-project.org")
  }

  install_booster_pack <- function(package, load = TRUE) {
    for (pkg in package) {
      if (!requireNamespace(pkg, quietly = TRUE)) {
        pak::pkg_install(pkg)
      }
      if (load) {
        library(pkg, character.only = TRUE)
      }
    }
  }

  core_packages <- c(
    'fs',
    'here',
    'janitor',
    'rio',
    'tidyverse',
    'skimr',
    'paletteer',
    'tidytuesdayR'
  )

  optional_packages <- c('ggtext', 'ggrepel', 'scales', 'lubridate')

  install_booster_pack(package = unique(c(core_packages, optional_packages)), load = TRUE)
  rm(install_booster_pack, core_packages, optional_packages)

  `%ni%` <- Negate(`%in%`)

  geometric_mean <- function(x) {
    exp(mean(log(x[x > 0]), na.rm = TRUE))
  }

  my_skim <- skim_with(
    numeric = sfl(
      n = length,
      min = ~ min(.x, na.rm = T),
      p25 = ~ stats::quantile(., probs = .25, na.rm = TRUE, names = FALSE),
      med = ~ median(.x, na.rm = T),
      p75 = ~ stats::quantile(., probs = .75, na.rm = TRUE, names = FALSE),
      max = ~ max(.x, na.rm = T),
      mean = ~ mean(.x, na.rm = T),
      geo_mean = ~ geometric_mean(.x),
      sd = ~ stats::sd(., na.rm = TRUE),
      hist = ~ inline_hist(., 5)
    ),
    append = FALSE
  )
}

Load raw data from package

raw <- tidytuesdayR::tt_load('2026-04-14')

ships          <- raw$ships
birds          <- raw$birds
beaufort_scale <- raw$beaufort_scale
sea_states     <- raw$sea_states

Exploratory Data Analysis

The my_skim() function is a modified version of skimr::skim() that returns the number of missing data points (cells as NA) as well as the inverse (e.g.: number of rows that are not NA), the count, minimum, 25%, median, 75%, max, mean, geometric mean, and standard deviation. It also generates a little ASCII histogram. Neat!

Ships

ships %>%
  select(-record_id, -observer) %>%
  my_skim()
Data summary
Name Piped data
Number of rows 12310
Number of columns 19
_______________________
Column type frequency:
character 6
Date 1
difftime 1
numeric 11
________________________
Group variables None

Variable type: character

skim_variable n_missing complete_rate min max empty n_unique whitespace
hemisphere 11 1.00 1 1 0 2 0
activity 0 1.00 7 18 0 7 0
cloud_cover 4787 0.61 5 16 0 3 0
precipitation 4803 0.61 3 15 0 8 0
census_method 0 1.00 4 7 0 2 0
season 0 1.00 6 6 0 4 0

Variable type: Date

skim_variable n_missing complete_rate min max median n_unique
date 0 1 1969-07-31 1990-12-21 1984-02-17 1435

Variable type: difftime

skim_variable n_missing complete_rate min max median n_unique
time 2 1 0 secs 84600 secs 45600 secs 315

Variable type: numeric

skim_variable n_missing complete_rate n min p25 med p75 max mean geo_mean sd hist
latitude 10 1.00 12310 -69.0 -40.6 -38.27 -35.50 -19.00 -39.22 NaN 5.89 ▁▁▆▇▁
longitude 11 1.00 12310 50.0 152.0 163.12 172.92 179.98 159.93 158.72 17.42 ▁▁▁▃▇
speed 4478 0.64 12310 0.0 10.5 14.50 16.00 22.00 12.71 13.12 5.18 ▂▁▅▇▂
direction 11055 0.10 12310 0.0 90.0 165.00 270.00 358.00 169.35 128.89 102.47 ▅▇▅▅▅
wind_speed_class 4688 0.62 12310 0.0 2.0 4.00 5.00 12.00 3.73 3.40 1.83 ▅▇▆▁▁
wind_direction 11018 0.10 12310 0.0 68.0 203.00 270.00 350.00 174.71 157.63 108.50 ▇▃▃▇▆
air_temperature 11274 0.08 12310 0.5 13.4 15.50 17.80 24.60 15.28 14.52 3.83 ▁▁▅▇▂
pressure 11540 0.06 12310 991.0 1010.0 1016.00 1021.00 1040.00 1014.84 1014.80 8.85 ▁▃▇▅▁
sea_state_class 4751 0.61 12310 1.0 3.0 4.00 4.00 6.00 3.56 3.33 1.10 ▂▇▆▃▁
sea_surface_temperature 11625 0.06 12310 -1.0 12.8 15.10 17.00 27.00 14.72 14.08 4.08 ▁▁▇▆▁
depth 12180 0.01 12310 60.0 162.5 290.00 507.50 2200.00 380.85 283.99 352.05 ▇▂▁▁▁ 
ships %>%
  count(season, name = "n_ship_records") %>%
  arrange(desc(n_ship_records))
# A tibble: 4 × 2
  season n_ship_records
  <chr>           <int>
1 spring           3585
2 summer           3477
3 autumn           2794
4 winter           2454
ships %>%
  summarise(
    first_voyage = min(date, na.rm = TRUE),
    last_voyage  = max(date, na.rm = TRUE),
    n_records    = n(),
    lat_min      = min(latitude,  na.rm = TRUE),
    lat_max      = max(latitude,  na.rm = TRUE),
    lon_min      = min(longitude, na.rm = TRUE),
    lon_max      = max(longitude, na.rm = TRUE)
  )
# A tibble: 1 × 7
  first_voyage last_voyage n_records lat_min lat_max lon_min lon_max
  <date>       <date>          <int>   <dbl>   <dbl>   <dbl>   <dbl>
1 1969-07-31   1990-12-21      12310     -69     -19      50    180.

The fleet works almost exclusively in the eastern hemisphere (hemisphere == "E"), wedged between roughly 19°S and 69°S — that’s the swath from sub-tropical Tasman waters down through the Roaring Forties, Furious Fifties, and into the Antarctic Convergence. Records span late July 1969 through December 1990 and are evenly spread across all four seasons, with a slight lean toward spring and summer when the southern weather permits more deck time.

Birds

birds %>%
  select(species_common_name, age, plumage_phase, count,
         following_ship, flying_past, sitting_on_water, feeding) %>%
  my_skim()
Data summary
Name Piped data
Number of rows 49019
Number of columns 8
_______________________
Column type frequency:
character 3
logical 4
numeric 1
________________________
Group variables None

Variable type: character

skim_variable n_missing complete_rate min max empty n_unique whitespace
species_common_name 691 0.99 9 56 0 321 0
age 38837 0.21 5 8 0 4 0
plumage_phase 48938 0.00 4 12 0 4 0

Variable type: logical

skim_variable n_missing complete_rate mean count
following_ship 21132 0.57 0.42 FAL: 16068, TRU: 11819
flying_past 21128 0.57 0.42 FAL: 16163, TRU: 11728
sitting_on_water 21122 0.57 0.12 FAL: 24427, TRU: 3470
feeding 21095 0.57 0.13 FAL: 24320, TRU: 3604

Variable type: numeric

skim_variable n_missing complete_rate n min p25 med p75 max mean geo_mean sd hist
count 2699 0.94 49019 1 1 2 4 99999 41.93 2.46 1283.41 ▇▁▁▁▁ 
birds %>%
  count(species_common_name, sort = TRUE) %>%
  head(10)
# A tibble: 10 × 2
   species_common_name                      n
   <chr>                                <int>
 1 Wandering albatross sensu lato PL2    3104
 2 Cape petrel                           2936
 3 Wandering albatross sensu lato PL3    2885
 4 Grey-faced petrel                     2597
 5 Giant petrel sensu lato               2094
 6 Prion (unidentified)                  2094
 7 Wandering albatross sensu lato PL1    1818
 8 Black-browed albatross sensu lato AD  1805
 9 Wandering albatross sensu lato PL4    1475
10 Australasian gannet AD                1366
birds %>%
  summarise(
    pct_following = mean(following_ship, na.rm = TRUE),
    pct_flying    = mean(flying_past,    na.rm = TRUE),
    pct_sitting   = mean(sitting_on_water, na.rm = TRUE),
    pct_feeding   = mean(feeding,        na.rm = TRUE)
  )
# A tibble: 1 × 4
  pct_following pct_flying pct_sitting pct_feeding
          <dbl>      <dbl>       <dbl>       <dbl>
1         0.424      0.420       0.124       0.129

The bird table is dominated by tubenoses — albatrosses, petrels, shearwaters, and prions. The single most-sighted “species” is the wandering albatross sensu lato in plumage phase 2, with over 3,000 records. (The “sensu lato” tag and PL1–PL5 codes flag that field observers were grouping look-alike sub-taxa and ageing them by their plumage transition, since wandering albatrosses take roughly a decade to reach definitive adult white.)

Roughly 42% of behaviorally-scored sightings record a bird actively following the ship — an extraordinary fraction, but consistent with what mariners have reported since the age of sail: certain seabirds, especially albatrosses, treat a vessel as a slow-moving feeding opportunity (offal, prop wash, disturbed plankton). The behavior is not uniform, though — and that variation is the story.

Which seabirds follow ships?

The “ship-follower” trade-off

Following a vessel costs energy — sustained flight, sometimes for hours — but pays out in the form of galley scraps, churned-up prey, and easy soaring on the ship’s wake-edge updrafts. Species that evolved to scavenge over vast empty ocean (albatrosses, giant petrels) take that bet readily. Species that pluck prey from the water surface in fast dives or that follow specific schooling fish (shearwaters, gannets) tend to ignore the ship entirely.

I want to rank species by the fraction of individual birds observed actively following the ship — not just the fraction of sightings, because flocks of dozens are common and one record can represent very different numbers of birds. To avoid the noise of rarely-seen species, I’ll restrict to species with at least 250 records on file, and collapse the wandering-albatross plumage codes into a single species group.

bird_ship <- birds %>%
  inner_join(ships, by = "record_id") %>%
  filter(!is.na(species_common_name)) %>%
  mutate(
    species_grp = species_common_name %>%
      str_replace(" sensu lato.*$", "") %>%
      str_replace("\\s+(AD|JUV|SUB.*|PL\\d+).*$", "") %>%
      str_trim()
  )

follow_summary <- bird_ship %>%
  group_by(species_grp) %>%
  summarise(
    n_records          = n(),
    total_birds        = sum(count, na.rm = TRUE),
    birds_following    = sum(n_following_ship, na.rm = TRUE),
    birds_flying_past  = sum(n_flying_past,    na.rm = TRUE),
    birds_sitting      = sum(n_sitting_on_water, na.rm = TRUE),
    birds_feeding      = sum(n_feeding,        na.rm = TRUE),
    .groups = "drop"
  ) %>%
  filter(n_records >= 250, total_birds > 0) %>%
  mutate(
    pct_following = birds_following   / total_birds,
    pct_flying    = birds_flying_past / total_birds,
    pct_sitting   = birds_sitting     / total_birds
  ) %>%
  arrange(desc(pct_following))

cat(sprintf("follow_summary: %d species after filtering\n", nrow(follow_summary)))
follow_summary: 35 species after filtering
stopifnot("follow_summary is empty — check joins/filters" = nrow(follow_summary) > 0)

follow_summary %>%
  select(species_grp, n_records, total_birds, pct_following, pct_flying, pct_sitting) %>%
  print(n = Inf)
# A tibble: 35 × 6
   species_grp        n_records total_birds pct_following pct_flying pct_sitting
   <chr>                  <int>       <dbl>         <dbl>      <dbl>       <dbl>
 1 White-chinned pet…       489        1246     0.474        0.104       0.0241 
 2 Wandering albatro…     11293       19905     0.468        0.00769     0.0121 
 3 Black petrel             403        1799     0.314        0.0506      0.0634 
 4 Black-browed alba…      3609        9057     0.308        0.0194      0.0153 
 5 Light-mantled soo…       259         679     0.239        0.0707      0.00295
 6 Cape petrel             2936       25790     0.183        0.00578     0.0238 
 7 Giant petrel            2168        5533     0.169        0.00886     0.0134 
 8 Wilson's storm pe…       257         654     0.122        0.243       0.00153
 9 Pomarine skua            308         980     0.0459       0.140       0.00510
10 Flesh-footed shea…       849        9824     0.0457       0.104       0.0932 
11 Yellow-nosed moll…       684        1438     0.0376       0.153       0.146  
12 Westland / White-…       371        1334     0.0337       0.114       0.124  
13 White-headed petr…       558        1403     0.0321       0.383       0.271  
14 Red-billed gull          754       11323     0.0273       0.0185      0.0646 
15 Buller's mollymawk       312         778     0.0231       0.239       0.0578 
16 Westland petrel          399        2662     0.0214       0.193       0.246  
17 Great-winged / Gr…       369        1238     0.0210       0.269       0.0121 
18 Southern black-ba…      1413       10281     0.0190       0.00953     0.128  
19 Grey-faced petrel       2597       10368     0.0129       0.314       0.0909 
20 Black-winged petr…       453        1192     0.00671      0.375       0.0143 
21 Shy / White-cappe…       383        2693     0.00668      0.105       0.0319 
22 Australasian gann…      1676       27037     0.00544      0.0818      0.0947 
23 Shy / white-cappe…      1534        5576     0.00538      0.0959      0.174  
24 Gould's petrel           350         775     0.00387      0.488       0.0271 
25 Salvin's mollymawk       275         857     0.00350      0.104       0.0968 
26 Crested tern             402         947     0.00211      0.470       0.00106
27 Prion (unidentifi…      2094      326499     0.000861     0.0643      0.0177 
28 Wedge-tailed shea…       472        8422     0.000594     0.330       0.0425 
29 Buller's shearwat…      1306       64750     0.000479     0.243       0.220  
30 Diving-petrel (un…       307       11683     0.000257     0.154       0.0876 
31 White-fronted tern       382        5424     0.000184     0.0400      0      
32 Sooty shearwater        1364      174329     0.0000975    0.135       0.0158 
33 Fluttering shearw…       507      113333     0.0000265    0.0216      0.151  
34 Short-tailed shea…       924      982553     0.0000142    0.441       0.0388 
35 Little shearwater        264       17428     0            0.216       0.0266

The ranking is striking. The wandering albatross, southern giant petrel, and cape petrel are the devoted shadows: half or more of every individual recorded was actively chasing the ship. At the other end, shearwaters and gannets — fast, prey-specialized fliers — barely register the vessel at all. The black-browed albatross sits in the middle, behaviorally as well as taxonomically; it’ll follow when convenient but doesn’t make a career of it.

Hero plot: the ship-follower spectrum

plot_df <- follow_summary %>%
  slice_max(n_records, n = 15) %>%
  mutate(
    species_label = species_grp %>%
      str_replace_all("\\s+/\\s+", " / ") %>%
      str_wrap(width = 32),
    species_label = fct_reorder(species_label, pct_following)
  )

hokusai <- as.character(paletteer::paletteer_d("MetBrewer::Hokusai1"))
ink_blue   <- hokusai[7]   # deep ocean
crest_teal <- hokusai[6]   # foam green
warm_mid   <- hokusai[5]   # sandy
warm_hi    <- hokusai[2]   # rust
warm_max   <- hokusai[1]   # dark wave-shadow

p <- ggplot(plot_df, aes(x = pct_following, y = species_label)) +
  geom_segment(
    aes(x = 0, xend = pct_following, yend = species_label, color = pct_following),
    linewidth = 1.2,
    lineend = "round"
  ) +
  geom_point(aes(size = total_birds, color = pct_following)) +
  geom_text(
    aes(label = scales::percent(pct_following, accuracy = 1)),
    hjust = -0.4, size = 3.5, color = "grey20", fontface = "bold"
  ) +
  scale_x_continuous(
    labels = scales::percent_format(accuracy = 1),
    limits = c(0, max(plot_df$pct_following) * 1.20),
    expand = expansion(mult = c(0, 0.02))
  ) +
  scale_size_continuous(
    range = c(3, 11),
    labels = scales::comma_format(),
    name = "Total birds\nobserved"
  ) +
  scale_color_gradientn(
    colors = c(ink_blue, crest_teal, warm_mid, warm_hi, warm_max),
    guide = "none"
  ) +
  labs(
    title    = "Which seabirds bother to chase a ship?",
    subtitle = "Share of individual birds recorded actively following the vessel,\nby species (Southern Ocean research voyages, 1969–1990).",
    x        = NULL,
    y        = NULL,
    caption  = "Source: TidyTuesday 2026-04-14 · Bird Sightings at Sea  |  Top 15 most-sighted species, plumage variants merged."
  ) +
  theme_minimal(base_size = 12) +
  theme(
    plot.title = element_text(face = "bold", size = 17, color = ink_blue),
    plot.subtitle = element_text(size = 11.5, color = "grey30", margin = margin(b = 14)),
    plot.caption  = element_text(color = "grey45", size = 9, margin = margin(t = 14)),
    panel.grid.major.y = element_blank(),
    panel.grid.minor   = element_blank(),
    panel.grid.major.x = element_line(color = "grey88", linewidth = 0.3),
    axis.text.y = element_text(size = 10, color = "grey15", lineheight = 0.95),
    axis.text.x = element_text(color = "grey40"),
    legend.position = "bottom",
    legend.title = element_text(size = 9, color = "grey30"),
    legend.text  = element_text(size = 9, color = "grey30"),
    plot.background  = element_rect(fill = "white", color = NA),
    panel.background = element_rect(fill = "white", color = NA),
    plot.margin = margin(18, 24, 14, 14)
  )

p

Final thoughts and takeaways

If you’ve ever stood on the stern rail of a ship south of New Zealand and watched a wandering albatross trail behind for the entire afternoon — never seeming to flap, just carving the air off the wake — this dataset will feel familiar. The number puts a percentage on the intuition: the wandering albatross spends most of its visible life behind boats, and so do the giant petrels and cape petrels that earn the old sailor’s name “Cape pigeon.” These are the obligate scavengers of the southern seas, and a ship is, to them, a feeding event in slow motion.

The shearwaters and the Australasian gannet sit at the opposite end of the same scale and are doing something different for a living: high-speed transit and plunge-diving on shoaling fish. A research vessel isn’t a buffet to them, it’s an obstacle to detour around.

A few caveats worth naming:

  • The behavioral flags are observer-coded from the deck, and missingness is real — only about 57% of sightings carry a behavior code at all. The denominator I used (total_birds) includes birds whose behavior wasn’t scored, so the percentages are conservative: the true rate of following-when-observed is higher.
  • “Following the ship” is partly a function of how slow you’re going. About 94% of the ship records are coded steaming, sailing, but speed varies; a stationary research vessel deploying gear is more attractive than a freighter at 18 knots.
  • The dataset is historic (1969–1990). Several of the species in the top rank — especially the wandering and black-browed albatrosses — have suffered severe population declines from longline fishing bycatch in the decades since. The behavior that made them visible to these naturalists is the very same behavior that makes them vulnerable to baited hooks set astern.

The data tells one clean story and gestures at a darker one: the seabirds most willing to keep us company at sea are also the ones most exposed to the way we now fish it.