Skip to contents

Decadal change in ES(50) per EOV

Source: vignettes/articles/decadal.Rmd
decadal.Rmd

occ_h3 carries the year of each record, so any indicator can be computed per period and cell. This article does that per decade for each Essential Ocean Variable and maps where ES(50) changed between two decades — Figure 5 of the OBIS → H3 → EOV manuscript. It is the store-backed counterpart of vignette("temporal_subsets"), which does the same on the shipped occ_* samples in memory.

Two rules keep the comparison honest. ES(50) needs at least 50 records in a cell-decade to be reliable, so the “reliable cells” count is reported alongside the raw record count. And change is only computed where both decades are reliable (calc_period_change()), so a map of Δ ES(50) never mixes “fewer species” with “fewer surveys”.

library(obisindicators)
library(dplyr)
library(ggplot2)

con <- obis_store_connect()   # the store named by OBIS_H3_DUCKDB

Precomputed. The store is not available where this documentation is built, so the chunks below were run locally by data-raw/precompute_articles.R and their output committed. This render used obis_h3_global_v20260728.duckdb, the global store.

Indicators per decade and cell

A coarser resolution (H3 2) than the maps in vignette("eov"), because splitting records across seven decades thins each cell (see vignette("scaling")).

starts <- seq(1960L, 2020L, by = 10L)
pd <- bind_rows(lapply(EOV_ORDER, function(e) {
  d <- calc_period_indicators(con, res = RES_TIME, eov = e, starts = starts, esn = 50L)
  if (nrow(d)) cbind(eov = e, d) else NULL
}))
summ <- pd |>
  group_by(eov, period) |>
  summarize(cells = n(), cells_reliable = sum(n >= 50), records = sum(as.numeric(n)),
            median_es = median(es[n >= 50], na.rm = TRUE), .groups = "drop")
knitr::kable(summ |> mutate(median_es = round(median_es, 2)),
             caption = sprintf("Per-decade coverage and ES(50) by EOV, H3 resolution %d", RES_TIME))
Per-decade coverage and ES(50) by EOV, H3 resolution 2
eov period cells cells_reliable records median_es
fish 1960 2621 569 873166 31.75
fish 1970 2048 699 1729628 30.91
fish 1980 1855 647 4493283 25.25
fish 1990 1983 770 7673014 22.14
fish 2000 2405 1075 11180145 22.14
fish 2010 1727 732 14529812 20.05
fish 2020 611 369 5919572 15.48
hardCorals 1960 429 39 9178 19.37
hardCorals 1970 395 73 16975 21.54
hardCorals 1980 406 100 25819 28.00
hardCorals 1990 452 135 108106 22.07
hardCorals 2000 531 194 285517 14.44
hardCorals 2010 438 177 350381 8.88
hardCorals 2020 149 38 29324 9.45
mangroves 1960 18 4 474 15.81
mangroves 1970 21 8 1330 14.79
mangroves 1980 28 9 1234 18.67
mangroves 1990 47 12 8180 8.46
mangroves 2000 42 24 41486 2.75
mangroves 2010 42 16 34009 3.67
mangroves 2020 21 2 1516 2.68
marineMammals 1960 353 41 31254 3.14
marineMammals 1970 456 69 40500 4.97
marineMammals 1980 794 174 79937 6.56
marineMammals 1990 1309 462 420446 2.95
marineMammals 2000 1992 1010 1539634 2.00
marineMammals 2010 1874 923 1949809 2.00
marineMammals 2020 841 236 566727 3.50
seaTurtles 1960 44 5 1215 1.00
seaTurtles 1970 68 7 2278 4.00
seaTurtles 1980 98 7 2373 2.00
seaTurtles 1990 631 120 26627 1.00
seaTurtles 2000 2023 252 69536 1.80
seaTurtles 2010 1497 348 264713 1.00
seaTurtles 2020 203 25 72599 1.08
seabirds 1960 447 87 105855 11.49
seabirds 1970 796 226 261923 12.33
seabirds 1980 1166 465 1921326 13.00
seabirds 1990 1890 646 2292579 9.75
seabirds 2000 3815 1509 2557880 5.90
seabirds 2010 4277 1852 8322502 7.68
seabirds 2020 2950 1222 7696206 18.55
seagrasses 1960 23 2 659 2.50
seagrasses 1970 52 3 703 7.87
seagrasses 1980 89 18 4380 6.94
seagrasses 1990 111 25 30889 4.35
seagrasses 2000 192 66 80068 3.80
seagrasses 2010 326 94 141857 3.88
seagrasses 2020 309 83 51186 3.23 
long <- bind_rows(
  transmute(summ, eov, period, metric = "cells reliable (n ≥ 50)", value = cells_reliable),
  transmute(summ, eov, period, metric = "records", value = records),
  transmute(summ, eov, period, metric = "median ES(50)", value = median_es))
p <- ggplot(long, aes(x = period, y = value, color = eov)) +
  geom_line() + geom_point(size = 1.5) +
  facet_wrap(~ metric, scales = "free_y") +
  labs(x = "decade", y = NULL, color = NULL, title = "Per-decade coverage and ES(50) by EOV",
       caption = store_label) +
  theme_minimal(base_size = 11) + theme(legend.position = "bottom")
p
plot of chunk fig5a

plot of chunk fig5a

Records climb steeply into the 2000s and 2010s for every EOV — that is observation effort, not biodiversity — which is exactly why the change maps below gate on reliability in both decades rather than comparing raw counts.

Change maps, 1980s → 2010s

calc_period_change() returns the per-cell difference for cells reliable in both decades, plus a coverage summary (cells reliable only in one decade or the other are reported, not mapped).

for (e in unique(pd$eov)) {
  ch <- calc_period_change(filter(pd, eov == e), from = 1980L, to = 2010L, esn = 50L)
  if (nrow(ch$cells) < 3) next
  cat("\n\n### ", obis_eov_label(e), " — cells reliable in both: ", ch$coverage$both,
      " (only 1980s: ", ch$coverage$only_from, "; only 2010s: ", ch$coverage$only_to, ")\n\n", sep = "")
  p <- gmap_cells(ch$cells, "delta", label = "Δ ES(50)") +
    labs(title = paste0(obis_eov_label(e, max_taxa = 1L), ": ES(50) 2010s − 1980s"),
         caption = store_label)
  save_fig(p, paste0("fig5b_change_", e), ch$cells, width = 9, height = 5)
  print(p)
}

Fish (Agnatha, Chondrichthyes, Osteichthyes) — cells reliable in both: 387 (only 1980s: 260; only 2010s: 345)

Hard corals (order Scleractinia) — cells reliable in both: 43 (only 1980s: 57; only 2010s: 134)

plot of chunk fig5b

plot of chunk fig5b

Mangroves (19 seed taxa: family/genus) — cells reliable in both: 4 (only 1980s: 5; only 2010s: 12)

plot of chunk fig5b

plot of chunk fig5b

Marine mammals (7 seed taxa: infraorder/order/species) — cells reliable in both: 141 (only 1980s: 33; only 2010s: 782)

plot of chunk fig5b

plot of chunk fig5b

Seabirds (class Aves) — cells reliable in both: 408 (only 1980s: 57; only 2010s: 1444)

plot of chunk fig5b

plot of chunk fig5b

Seagrasses (order Alismatales) — cells reliable in both: 13 (only 1980s: 5; only 2010s: 81)

plot of chunk fig5b

plot of chunk fig5b

Sea turtles (superfamily Chelonioidea) — cells reliable in both: 6 (only 1980s: 1; only 2010s: 342)

plot of chunk fig5b

plot of chunk fig5b

plot of chunk fig5b

plot of chunk fig5b

EOVs with fewer than three cells reliable in both decades are skipped — on a regional demo store that is most of them; the global store fills in the rest.

DBI::dbDisconnect(con, shutdown = TRUE)