Skip to contents

Standardise a table of locally measured meteorology for bias correction

Source: R/prepare_obs_met.R
prepare_obs_met.Rd

Takes a user-supplied observation table (lake buoy, shore weather station, council / NIWA / CliFlo export, logger dump) and returns a clean data frame with a Date column and AEME MET_* columns in AEME units, resampled to the resolution you intend to bias-correct at. The output is designed to line up with extract_era5_hourly_met() output so the two can be joined on Date in fit_met_bias_correction().

Usage

prepare_obs_met(
  obs,
  col_map = NULL,
  datetime_col = NULL,
  date_format = NULL,
  tz = NULL,
  resample = c("none", "hour", "day"),
  interval = c("ending", "beginning"),
  derive = TRUE,
  convert_units = TRUE,
  wind_height = NULL,
  wind_z0 = 2e-04,
  station = NA_character_,
  verbose = TRUE
)

Arguments

obs

data frame, or path to a .csv / .rds file.

col_map

optional named character vector mapping AEME names to the columns in obs, e.g. c(MET_tmpair = "AirTemp_C", MET_wndspd = "WindSpd", MET_radswd = "SolarRad"). Columns not listed are dropped. If NULL, columns already named MET_* are kept and the rest are matched with guess_met_vars().

datetime_col

name of the timestamp column; auto-detected when NULL (first column named or containing date/time).

date_format

optional strptime format string for the timestamp column. When NULL (default) a set of common formats is tried; if the strings contain a time (a :), only date-time formats are considered so that a stray time-less row does not silently collapse the whole column to dates.

tz

time zone of the observation timestamps. Default "UTC" (matches the extract_era5_hourly_met() default). Set this to the zone your logger actually recorded in - e.g. "Etc/GMT-12" for fixed NZST, or "Pacific/Auckland" if it recorded civil time with DST.

resample

"none" (keep native step, de-duplicated), "hour" (mean per hour; rain/snow summed) or "day" (daily mean; rain/snow summed; Date returned as class Date).

interval

which hour a sub-hourly record belongs to when resample = "hour". "ending" (default) puts a record at 00:15 into the hour labelled 01:00, matching the accumulated-flux convention of ERA5 and extract_era5_hourly_met(), so the two line up when joined; "beginning" labels it 00:00. With "ending" the first bin of a record that starts exactly on the hour holds a single observation. Ignored for resample = "day", which always uses calendar days.

derive

if TRUE (default) fill MET_wndspd/MET_wnddir from u/v (or vice versa) and MET_tmpdew<->MET_humrel from air temperature, where the inputs are present.

convert_units

if TRUE (default) auto-detect and fix common unit mistakes (K->degC, hPa->Pa, RH 0-1 -> %, km/h->m/s, mm->... ).

wind_height

measurement height of the wind sensor, m. When given and not already 10, the wind columns are rescaled to 10 m (the ERA5 / lake-model convention) with met_wind_at_height() - a buoy anemometer at 2 m needs wind_height = 2. NULL (default) leaves wind untouched.

wind_z0

roughness length (m) for that adjustment; default 2e-4 (open water). See wind_at_height().

station

optional label stored on the result ("buoy", "shore", station id) - carried into the bias-correction metadata.

verbose

print what was matched / converted / resampled.

Value

data frame, Date first, then the matched MET_* columns. Attributes: tz, resolution ("subhourly"/"hourly"/"daily"), station, n_obs.

Details

Standard names / units (see met_vars()): MET_tmpair degC, MET_tmpdew degC, MET_humrel %, MET_wndspd m/s, MET_wnddir deg (FROM), MET_wnduvu/MET_wnduvv m/s, MET_radswd and MET_radlwd W/m2, MET_pprain/MET_ppsnow mm (per timestep), MET_prsttn/MET_prmslp Pa.

See also

fit_met_bias_correction(), apply_met_bias_correction()