Skip to contents

GEBCO seafloor bathymetry over the CalCOFI area

Source: R/transect.R
cc_bathy.Rd

A terra::SpatRaster of GEBCO 2025 sub-ice bathymetry cropped to the CalCOFI grid, 15 arc-second (~390 m x 460 m at 33 degN), as positive-down depth in metres with land clamped to 0. Downloaded once and cached, so the first call in a session costs a 4.3 MB fetch and later ones cost nothing.

Usage

cc_bathy(
  path = NULL,
  cache_dir = NULL,
  refresh = FALSE,
  remote = FALSE,
  extent = c("calcofi", "full")
)

Source

GEBCO Compilation Group (2025) GEBCO 2025 Grid, https://www.gebco.net/data_and_products/gridded_bathymetry_data/.

Arguments

path

explicit raster to load; defaults to the option / env var above, then to the cached download.

cache_dir

where to keep the download. Defaults to rappdirs::user_cache_dir("calcofi4r").

refresh

re-download even if cached (default FALSE).

remote

read the published object in place over /vsicurl/ instead of downloading it — GDAL then fetches only the blocks a query touches, so sampling a handful of points costs no download at all. Nothing is cached; the default stays download-and-cache (offline, deterministic).

extent

"calcofi" (default) for the CalCOFI crop (lon −165 → −100 × lat 15 → 56 — every released bottle / PIC / CUFES / dungeness / DIC / euphausiid position; ~129 MB), or "full" for the whole GEBCO source tile (lon −180 → −90 × lat 0 → 90; ~430 MB, so pair it with remote = TRUE). The full tile is published as raw GEBCO elevation; this function converts it on read to the same positive-down, land-0 depth_m convention as the crop, so callers never see the difference.

Value

A terra::SpatRaster, single layer depth_m.

Details

The sign and land convention are baked into the published raster rather than applied on read: a caller who sees 12.4 is 12.4 m under water, and 0 is land. Nothing downstream has to remember to negate an elevation.

Using a local file instead

Set options(calcofi4r.bathy = "/path/to.tif") (or the CALCOFI_BATHY environment variable) to bypass the download — how an app that already ships its own crop, or a machine with no network, keeps working.

The cache refreshes itself

When a cached copy exists, one cheap HEAD compares its size to the published object's and re-downloads on a mismatch (once per session per object) — so when the crop is re-cut nobody has to know to pass refresh = TRUE. Offline, the HEAD fails quietly and the cache is used as-is.