Served over OPeNDAP from the HYCOM THREDDS server at the Naval Research Laboratory. Unlike the Copernicus and FVCOM sources, HYCOM publishes one dataset per year rather than one aggregation, so a request spanning years opens several.
Value
a named list, one entry per archive, each with url, layout,
start, end, step_hours, kind, label, and reference
One reanalysis, then a chain of operational experiments
GLBv53X is the GOFS 3.1 reanalysis: one internally consistent run over
1994–2015, which is what makes it the default. Everything after it is
operational output — the model as it was running at the time, reassigned
an experiment number whenever it changed — so the later archives are short,
they overlap each other, and they are not a reanalysis.
Together they reach the present:
GLBv53X | 1994-01-01 to 2015-12-31 | reanalysis |
GLBv563 | 2014-07-01 to 2016-09-30 | operational |
GLBv572 | 2016-05-01 to 2017-02-01 | operational |
GLBv928 | 2017-02-01 to 2017-06-01 | operational |
GLBv577 | 2017-06-01 to 2017-10-01 | operational |
GLBv929 | 2017-10-01 to 2018-03-20 | operational |
GLBv930 | 2018-01-01 to 2020-02-19 | operational |
GLBy930 | 2018-12-04 to 2024-09-05 | operational, finer grid |
Two seams to know about
They overlap. GLBv563 starts eighteen months before GLBv53X ends, and
several of the 2017 experiments abut or overlap. Nothing here picks between
them, because which one to prefer where they overlap is a judgement: the
reanalysis is the more consistent, the operational run the more recent.
The seam that matters is the run, not the grid. Crossing from GLBv53X
into an operational archive means changing from a reanalysis to the model as
it was running at the time, which is a genuine discontinuity in how the
values were produced.
The grid difference is smaller than it looks. GLBv0.08 carries 3251
latitudes and GLBy0.08 carries 4251, but both span -80 to 90 and both are
spaced 0.04 degrees through the middle latitudes — GLBv0.08 stretches
toward the poles where GLBy0.08 stays uniform. Between 40 and 45 N the two
hold the same 126 latitudes, identically, so on a mid-latitude shelf the cells
do correspond and a series spanning the seam is on one grid. They diverge
at high latitude, where a polar study would be comparing different cells.
The longitude convention differs too — GLBv0.08 runs -180 to 180 and
GLBy0.08 runs 0 to 360 — but that is handled internally. Give
bounding_box negative west for either.
accessHYCOM() reads one archive per call and names the others when a
request falls outside the one asked for, rather than stitching them silently.
Examples
names(hycom_archives())
#> [1] "GLBv53X" "GLBv563" "GLBv572" "GLBv928" "GLBv577" "GLBv929" "GLBv930"
#> [8] "GLBy930"
hycom_archives()$GLBv53X$end
#> [1] "2015-12-31"
# Which archives cover a given day
hycom_covering(as.Date("2019-06-15"))
#> [1] "GLBv930" "GLBy930"