Samples cd's daily ERA5-Land air-temperature cube at a set of points and
returns one row per point, day and variable. The cube holds daily mean,
maximum and minimum 2 m temperature (°C) for British Columbia from 1950,
on local days (UTC−8), on the same 0.1° grid as the monthly layers.
It is built from ERA5-Land hourly data by scripts/backfill_edh_daily.py
and published to S3 as one Cloud-Optimized GeoTIFF per variable and year
(<source>/<variable>_daily_<year>.tif, one band per day). It is not in
the STAC catalog read by cd_catalog().
Arguments
- points
Point locations: an
sfor terra::SpatVector of points in any CRS, carrying theidcolumn, or a data frame with longitude/latitude columns (WGS84) named bycoords.- from, to
First and last day to return, as
Dateor"YYYY-MM-DD".- variables
Any of
"tmean","tmax","tmin".- id
Name of the column in
pointsthat identifies each point. Its values must be unique and not missing.- coords
For a data frame
points, the longitude and latitude column names.- source
Directory or base URL holding the cube files. Defaults to the published cube on S3; override with
options(cd.daily_url = ...).- cache
Logical. If
TRUE(default), each remote file is downloaded once throughcd_cache_fetch()and read locally: fastest for many points and for repeat calls (about 9–17 MB per variable and year). IfFALSE, only the 16 × 16-cell tiles holding the points are read over the network, which suits a few points in one call.
Value
A tibble with one row per point, variable and day:
- id
The point's identifier, from the
idcolumn.- date
Local day (
Date).- variable
"tmean","tmax"or"tmin".- value
Air temperature, °C.
- cell
ERA5-Land cell number on the cube's grid;
NAfor a point outside the cube.- cell_x, cell_y
Longitude and latitude of the cell centre;
NAfor a point outside the cube.- cell_moved
TRUEwhen the point's own cell had no data and a neighbour was used.- checksum
The checksum the cube's manifest gives for the file the row was read from, kept only when the file is shown to be that build (see Provenance); otherwise
NA.- cd_sha
Git commit of the cd code that built that file, from the file's own tags;
NAfor an untagged file.
Rows are ordered by point (input order), variable, then date.
Day boundary
A local day runs from 08:00 UTC to 07:00 UTC the next day: Pacific
Standard Time, with no daylight saving. A UTC day would run from one
afternoon peak (22:00–00:00 UTC) to the next, so a hot afternoon would
count toward two days and the daily maximum would read high, a bias that
does not cancel in absolute thresholds such as degree-days. The fixed
offset is an hour off for the Mountain Standard Time corner of eastern
BC. The monthly tmax/tmin layers in the catalog average these same
local days (cd#37).
Cell choice
Each point takes the ERA5-Land cell containing it. ERA5-Land covers land
only, so a point on the coast or a large lake can fall in a cell with no
data. Such a point moves to the nearest cell with data among the eight
around it, measured geodesically from the point to each cell centre
(ties go to the lower cell number), and its rows carry
cell_moved = TRUE. If none of the eight has data, the point keeps its
own cell, its values are NA, and a warning names it. Points sharing a
cell share a cell value, so identical series are easy to spot. There is
no elevation adjustment between the cell and the point.
A point outside the cube (British Columbia's box, 47.95–60.05° N,
140.05–113.95° W) gets NA values and NA cell columns, and a warning
names it with the cube's extent; the other points are unaffected.
Provenance
Each file of the cube is one variable and year, and checksum and
cd_sha say which build of it a row came from. cd_sha is read from the
file itself. checksum comes from <source>/manifest.json, read once per
call (through the cache when cache = TRUE, so it follows
cd.cache_revalidate like the files), and is kept only when the file's
run-time tag matches the manifest's
processing:datetime for it, so a stale cached copy gives NA and, for
an entry carrying both a checksum and a run time, a warning rather than
another build's checksum. An entry with no
processing:datetime cannot be checked and gives NA silently, as does
a file the manifest does not list. A local source with no manifest (such
as the bundled example) gives NA silently. A manifest that cannot be
read or is not the published shape (files, an object keyed by file name)
gives NA and a warning; with cache = TRUE, a cached manifest is used
instead when the remote one cannot be reached. A cached manifest can be
older than a file downloaded after it (with cd.cache_revalidate = FALSE):
a republished year then warns, and a year it does not list yet is NA
silently. cd_cache_fetch("<source>/manifest.json", refresh = TRUE)
replaces it. A checksum identifies a build, not values.
Examples
# One year (2002) of a 5 x 5-cell crop of the cube ships with the package.
src <- system.file("extdata", "example_daily", package = "cd")
stations <- data.frame(
id = c("upper", "lower"),
lon = c(-126.38, -126.07),
lat = c(54.62, 54.31)
)
d <- cd_extract_daily(stations, "2002-06-01", "2002-08-31", source = src)
head(d)
#> # A tibble: 6 × 10
#> id date variable value cell cell_x cell_y cell_moved checksum cd_sha
#> <chr> <date> <chr> <dbl> <int> <dbl> <dbl> <lgl> <chr> <chr>
#> 1 upper 2002-06-01 tmean 4.34 6 -126. 54.6 FALSE NA NA
#> 2 upper 2002-06-02 tmean 6.54 6 -126. 54.6 FALSE NA NA
#> 3 upper 2002-06-03 tmean 6.73 6 -126. 54.6 FALSE NA NA
#> 4 upper 2002-06-04 tmean 5.89 6 -126. 54.6 FALSE NA NA
#> 5 upper 2002-06-05 tmean 4.14 6 -126. 54.6 FALSE NA NA
#> 6 upper 2002-06-06 tmean 3.44 6 -126. 54.6 FALSE NA NA
# Growing degree-days above 5 °C over the summer, per station
tm <- d[d$variable == "tmean", ]
tapply(pmax(tm$value - 5, 0), tm$id, sum)
#> lower upper
#> 629.7610 574.8422
if (FALSE) { # \dontrun{
# The published cube, 2002 to 2025
cd_extract_daily(stations, "2002-01-01", "2025-12-31")
} # }