specReBin: Spectral Re-Binner

In asgr/ProSpect: Professional Spectral Analysis Package

Spectral Re-Binner

Description

It is often useful to rebin spectra to either log binning or linear binning. This function does just that. Also allows an entire SSP library to be rebinned (useful is using a very high resolution SSP for broad-band work etc).

Usage

specReBin(wave, flux, wavegrid = NULL, bin = NULL, binfunc = median, interp = 'approx',
  logbin = TRUE, rough = FALSE, ...)

speclibReBin(speclib, wavegrid = NULL, bin = NULL, binfunc = median, interp = 'approx',
  logbin = TRUE, rough = FALSE, cores = 1L, check = FALSE, ...)

Arguments

wave	Numeric vector; the input spectral wavelength in any linear units. For convenience, if a two column matrix or data frame is supplied then the first column is taken to be wave and the second is taken to be lum.
flux	Numeric vector; the input flux density in any linear units.
speclib	List; the spectral library (e.g. BC03hr).
wavegrid	Numeric vector; the target spectral wavelength grid in any linear units (must be same as wave). If this is missing then it will be computed automatically using bin. See also binfunc.
bin	Numeric scalar; the bin size to use in either liner units (same as wave) or log units (i.e. dex). If not provided this will be automatically computed. If this is provided and wavegrid is missing, then internally wavegrid is computed to be seq(floor(wave), ceiling(wave), by = bin). See also binfunc.
binfunc	Function; if bin is not provived this function will be run on either the linear or logged wavelength (as per logbin) to choose a reasonable constant bin size.
interp	Character scalar; The interpolator to use. Either 'approx' (see approx) or 'spline' (see spline). The former is better behaved with discrete spectrum features, but the latter might work better when the spectrum is sufficiently smooth.
logbin	Logical; should the automatic wavelength grid be evenly log spaced? If TRUE it will be evenly log spaced, if FALSE it will be evenly linearly spaced. Say you have a wavelength 10 and 20, a linear bin would be at 15 and a log bin would be at 14.1.
rough	Logical; should rough interpolation be used to rebin the spectrum (TRUE) or should full pixel integration be done (FALSE, default). As of ProSpect v1.4.1 rough = FALSE is just as fast (if not faster) than the coarse interpolation method. For spectral fitting you certainly might want the more accurate integration, but it can cause discontinuities when moving from very coarse model data to quite fine resolution observed data.
cores	Integer scalar; the number of cores to use.
check	Logical; should checks be run on the spectral library being returned? Uses speclib_check. If checks fail the SSP is still returned, and can then be checked and potentially fixed by the user.
...	Extra arguments to pass to either approx or spline as per interp.

Details

It is often useful to rebin spectra to either log binning or linear binning. This function does just that.

Note you need to be careful the resolution is high enough to not lose flux. This function just interpolates the flux density, it does not integrate over bins and reditribute flux. To do a more accurate job you would rebin at very high resolution and then integrate that spectrum down into coarser bins.

Value

specReBin Data.frame; first column is wavelength [wave] and second column is observed flux density [flux].

speclibReBin List; the rebinnned SSP spectral library.

Author(s)

Aaron Robotham

See Also

approx, spline, addspec

Examples

library(magicaxis)
tempflux = ProSpectSED(sparse=1, emission=TRUE, stellpop = "BC03hr")$FinalFlux
tempflux = tempflux[tempflux$wave>2e3 & tempflux$wave<2e4,]

magplot(tempflux, type='l', xlim=c(7e3,8e3))
lines(specReBin(tempflux, rough=TRUE), col='red') #works pretty well automatically
lines(specReBin(tempflux, bin=1e-3, rough=TRUE), col='green') #misses lines
lines(specReBin(tempflux, bin=1e-5, rough=TRUE), col='blue') #very close to the original

# We can also compare rough and accuracte integration schemes:

magplot(tempflux, type='l', xlim=c(7e3,8e3))
lines(specReBin(tempflux, bin=1e-3, rough=TRUE), col='red') #rough and fast
lines(specReBin(tempflux, bin=1e-3, rough=FALSE), col='blue') #conserves flux

# And now a whole spectral library:

data(BC03hr)
data(BC03lr)
speclib_rebin = speclibReBin(BC03hr, BC03lr$Wave, rough=FALSE)

# Note the ProSpect library re-bin seems to be more accurate than the official BC03 one!

magplot(BC03hr$Wave, BC03hr$Zspec[[1]][100,], type='l', log='xy', xlim=c(1e3,1e4))
lines(BC03lr$Wave, BC03lr$Zspec[[1]][100,], col='red')
lines(speclib_rebin$Wave, speclib_rebin$Zspec[[1]][100,], col='blue')

asgr/ProSpect documentation built on Feb. 21, 2025, 1:43 a.m.