TES deconvolution cookbook

From DavinciWiki
Jump to: navigation, search

Description

<<<<UNDER CONSTRUCTION>>>>

This procedure illustrates the standard method to deconvolve TES spectra.

Procedure

(1) Find TES Spectra in JMARS

  • Open TES layer, select Add -> Pre-defined Templates -> Spectra: Warm Surface, High Quality, Clear Atmosphere
  • Replace Atmosphere Adjusted Emissivity with Emissivity. In Fields, select atm_adj_emissivity, select Delete. Next to Fields, select Add -> Emissivity.
  • A note on JMARS atmosphere-adjusted emissivity spectra:
These spectra are derived by deconvolving the spectra using a standard set of surface
and atmosphere endmembers. This works well for a first pass, but may remove surface
features since many mineral endmembers are not in this standard library. So, while you
could use this data set for more detailed work, it is highly recommend to redo the
atmosphere removal with your full mineral library.
  • Limit spacecraft motion. This includes high-gain antenna and solar panel motion. Under Selects, select Add, then:
hga_motion = 1 (min) to 1 (max)
pnl_motion = 1 (min) to 2 (max)
  • Limit spacecraft ocks. 1583 is the start of the science phase, same as orbit 0. For ocks above 7000, the TES detectors sometimes exhibit an instrumental error (the "glitch") that results in a reversal of the dust absorption near 1100 wavenumbers, creating a broad peak with random direction (peak or trough) and amplitude in the spectra centered around 1000 wn, and extending between 900-1100 wn. This is most likely caused by internal motion within the spacecraft affecting the interferometer. The glitch can be somewhat compensated for by averaging many icks from the same ock, but this isn't recommended for most applications. Bolometer/albedo data is not affected by the glitch.
ock = 1583 (min) to 7000 (max).
  • Change target_temp (can be lowered for greater coverage):
target_temp (min) = 260
  • Delete best_fit_opacities, replace with tot_ice and tot_dust. Recommended limits (can be adjusted to higher values for more coverage):
tot_ice = (min) 0.00 (max) 0.04
tot_dust = (min) 0.00 (max) 0.15
  • Hit Submit (and save as template if you don't want to redo this in the future!)


(2) Download emissivity spectra

  • Select spectra in map, highlight in spectra dialog. Save as text file.


(3) Load spectra and spectral libraries into Davinci

  • Use split_ock to load spectra
dv> tes = split_ock("file.txt")
  • View using pplot
dv> pplot(tes.ock6002avgcat,xaxis = lib.xaxis,x1 = 1300,x2 = 200)
  • Load spectral library
dv> lib = load("speclib.hdf")
  • Load atmospheric library
dv> atm = load("atm.hdf")
  • Published spectral libraries (e.g., Rogers et al, 2007) can be reconstructed from the speclib website (speclib.asu.edu).
  • Mineral library selection guidelines:
- Make sure the library captures the mineralogic diversity expected
- Make sure the library captures crystal chemical differences (make sure that all
wavelength regions are being represented)
- Pay close attention to spectral quality using the quality flags for each mineral,
and also by looking at each spectrum for particle size effects
  • Some notes on spectral contrast:
- The endmembers in the library have varying degrees of spectral contrast, mostly due
to grain size
- In general, avoid using hand samples because they have such high spectral contrast
- If you have to use a hand sample, try to scale the spectrum accordingly
(e.g., by 0.4 to 0.5)
- What to do with a surface spectrum with low spectral contrast? This shouldn't be a
problem in general, but be advised that this might increase the noise


(4) Remove atmospheric component

  • Use spectral mixture analysis (SMA) to do deconvolution:
dv> unmix1 = sma(tes.ock6002avgcat,lib,atm,wave1 = 232,wave2 = 1302,nn = 1,surface = 1)
  • View using pplot:
dv> pplot({unmix1.rematm,unmix1.modsur},xaxis = lib.xaxis,x1 = 1300,x2 = 200)


(5) Deconvolve surface spectrum

  • Use spectral mixture analysis (SMA) to do deconvolution:
dv> unmix2 = sma(unmix1.rematm,lib,wave1 = 232,wave2 = 1302,nn = 1,exclude = 43)
  • View quick summary of results, including errors (RMS):
dv> summary_sma(unmix2,1,1,group = 1)
  • Some notes on interpreting/using RMS error from spectral mixture analysis (SMA):
- Use RMS error to make sure your mineral library is in the right ballpark
- DO NOT only look at RMS error, but also check the fit to the spectrum to make sure
specific features are being correctly modeled
- Note that the spectra need to be normalized for spectral contrast if you want to
consider the RMS error as an absolute (to compare between regions)


DavinciWiki Mini-Nav Bar

Contents


Contact Developers

  • davinci-dev [AT] mars.asu.edu

All other topics

  • See navigation on the left

Personal tools