thin_archive averages a processed depth record into blocks of
interval_s seconds, e.g. from 1 second to 15 second sampling. Use it
on the archive passed to create_wavelet() to reduce the memory and time
the wavelet analysis needs for long, high-frequency records.
Arguments
- archive
Data frame of processed time series depth data from
import_tag_data().- interval_s
Numerical. New sampling interval in seconds. Must be a multiple of the current sampling interval and divide a day evenly, e.g. 5, 10, 15, 30 or 60.
- verbose
Logical. If TRUE, progress messages are shown. Defaults to FALSE.
Value
The thinned archive, with the same columns and 'time_zone'
attribute. Returned unchanged if interval_s equals the current
sampling interval.
Details
Each block of interval_s seconds becomes one record, timed at the
start of the block, with the mean depth (and mean temperature, if present)
of the block. Averaging, rather than keeping every nth record, stops
movements faster than the new interval appearing as false longer cycles
(aliasing).
Thinning only removes cycles shorter than about twice the new interval. As
a rule of thumb, keep at least 20 records per cycle of the wavelet's lower
period, i.e. interval_s <= lower_period_mins * 60 / 20: up to 15
seconds for a 5 minute lower period.
Keep using the full record for create_depth_stats(), as averaging slightly
reduces maximum depths. Tags clustered together should all be thinned to the
same interval.
Examples
# Set file path
filepath <- system.file("extdata", package = "FishDiveR")
# Load archive_days, sampled every 60 seconds
archive_days <- readRDS(file.path(filepath, "data/archive_days.rds"))
# Thin to 5 minute sampling
archive_5min <- thin_archive(archive_days, interval_s = 300, verbose = TRUE)
#> Thinned from 60 to 300 second sampling: 14,400 to 2,880 records