I believe that we hoped to get the stellar atmosphere modellers to extend the wavelength coverage. Is that still the plan behind this ticket?

The short-term solution is to fit the continuum at <1.2um  and extrapolate it out to 1.26um, and that is the plan of this ticket.  I am still thinking about asking the modeller to extend the model.  He asked me for a line list.  I iterated a bit with the GA team and I realized that there are few strong lines between 1.20 and 1.26um.  So, we may just ask him to extend without a line list.  We can mask any obvious lines between 1.20 and 1.26 in observed spectra, so that they do not affect flux calibration vectors.

This figure is an example of the extrapolation. 

(p7000_g+4.0_m0.0_t01_z+0.50_a+0.00.AMBRE_Extp.png)
p7000_g+4.0_m0.0_t01_z+0.50_a+0.00.AMBRE_Extp.png
The extrapolated spectrum from 12000 A to 13000 A was generated from the best fit exponential function. The fitting was done in 9800-11800 A. Before fitting, the spectrum was smoothed by a Gaussian kernel with 2.4 A FWHM (the PFS resolution in NIR), and the Paschen absorptions were masked. 

The generated spectrum (the smoothed spectrum + extrapolation) is saved in a csv file. 

Takuji Yamashita Masayuki Tanaka are you waiting now for feedback from others? If so, set the status of this ticket to 'In Review' and assign a reviewer (eg rhl).

Following the DRP telecon 2019-04-05, it was agreed to close this ticket based on the work done.

This ticket has been closed but I make a note of some revisions to reduce a file size.
The AMBRE stellar models were processed with extrapolateAMBRE.py in a branch “tickets/ /PIPE2D-364” of the git repository “drp_fstar”. A processed spectrum contains flux density from ~300 nm to 1300 nm at a step of 0.01 nm. FITS headers contain CRPIX1, CRVAL, CDELT for the wavelength solution, parameters of continuum fitting, and parameters of the original model. A size of a file is 0.4MB. In total, the size is ~2GB for models of Teff = 6000-8000 K.
 
The procedure of the code is as follows. The asterisk denotes the revisions.
1*. binning a spectrum from 0.001 nm/pix to 0.01 nm/pix.
2. smoothing the spectrum with Gaussian kernel with a FWHM of 0.24 nm (= the resolution in NIR of PFS).
3. mask Paschen absorptions. FWHM to be mask is set to be 30 nm.
4. fitting the spectrum (wavelength vs. F_lambda) at 980-1180 nm using an exponential function .
5. extrapolation between ~ 1200 nm and 1300 nm using the fitting result.
6*. save the processed spectrum (binned spectrum + extrapolation) in .fits with headers. The units is nm and nJy.
 
The step 2 will be revised after the line spread function is provided.