LRS2Multi is a post-processing and extraction toolkit designed to operate on the reduced, flux-calibrated outputs from Panacea. It provides higher-level combination, sky subtraction refinement, source extraction, and coaddition for the LRS2 integral-field spectrograph on the Hobby–Eberly Telescope (HET).

• Overview
• Key Features
• Installation and Setup
• Reduction Workflow
• Algorithm Summary
• Data Products
• Developer Pointers
• Citation

LRS2Multi extends Panacea’s reductions by combining multiple exposures, modeling residual backgrounds, and extracting science-ready 1D spectra.
Where Panacea works from CCD frames to flux-calibrated fibers, LRS2Multi works from fibers to final spectra — aligning, scaling, and coadding exposures across the four LRS2 channels (UV, Orange, Red, Far-Red).

• Seamless integration with Panacea outputs
  Ingests calibrated 2D fiber spectra and metadata directly from Panacea.

• Residual background and sky subtraction
  Refines Panacea’s sky model with local annular background fitting, per-fiber residual correction, PCA residual fitting, and/or polynomial modeling.

• Aperture and PSF-weighted source extraction
  Performs circular or custom extraction apertures, with adaptive weighting for partially resolved sources.

• Exposure alignment and coaddition
  Aligns wavelength grids, scales exposures using mean/median values of common wavelength overlap regions, and coadds using inverse-variance weighting.

• Quality-assurance products
  Generates diagnostic plots and intermediate FITS extensions to validate flux scaling and sky subtraction.

ssh [email protected]
cd $HOME
ln -s $WORK work-ls6
module load python3
pip3 install astropy seaborn specutils scikit-learn --user
cd /work/XXXX/USERNAME/ls6
git clone https://github.com/grzeimann/LRS2Multi.git

Then, go to the visualization portal in a browser: https://vis.tacc.utexas.edu/jobs/

You can sign in and proceed to the utilities page by clinking the tab at the bottom or visiting: https://vis.tacc.utexas.edu/jobs/utils/

On the Utilities page, it the three bottom buttons to link Python3 and your directories for your work.

Then go back to the jobs page which should look like this:

Request a job as shown in the attached image (just click submit when you pull up the same left hand settings). After a small wait time, a new screen will show up and you will click connect. Sometimes there are not enough nodes initially and you have to wait a bit longer. After you connect, you should be in your work directory, which will allow you to navigate to LRS2Multi/notebooks. There will be a file called example_reduction.ipynb. Open that notebook and follow the instructions to get started.

This example demonstrates a complete reduction workflow using LRS2Multi and LRS2Object.
It converts Panacea-reduced exposures into a single, flux-calibrated 1D spectrum with telluric correction,
sky subtraction, astrometry, extraction, and diagnostic plots.

Requirements:

• Panacea multi-extension FITS products on disk
• Python 3 environment with astropy, numpy, matplotlib, and seaborn
• lrs2multi and lrs2object available on your PYTHONPATH

import sys, glob, os.path as op
sys.path.append("..")  # import local lrs2multi modules

import numpy as np
import seaborn as sns
import matplotlib.pyplot as plt
from astropy.io import fits
from astropy.table import Table
from datetime import datetime, timedelta

from lrs2multi import LRS2Multi
from lrs2object import LRS2Object

%matplotlib inline

sns.set_context('talk')
sns.set_style('ticks')
plt.rcParams['font.serif'] = ['P052']
plt.rcParams['font.family'] = 'serif'

def_wave = np.arange(3650., 10500., 0.7)  # common wavelength grid

Two helper functions, get_scifiles_from_folder() and get_scifiles_from_folder_from_pos(),
collect relevant Panacea output FITS files from your observing directories.
They filter by object name, exposure time, observing date, and LRS2 channel pairing.

folders = ['/work/03946/hetdex/maverick/LRS2/UT24-1-015']

exclude_folder = [
  'multi_20230822_0000009_exp01', 'multi_20201110_0000013_exp01',
  'multi_20230727_0000010_exp01', 'multi_20250418_0000009_exp01',
  'multi_20250423_0000008_exp01', 'multi_20250429_0000007_exp01',
  'multi_20250912_0000014_exp01', 'multi_20250822_0000014_exp01'
]

objectname = 'OQ208'

filenames, _ = get_scifiles_from_folder(
    folders,
    object_name=objectname,
    exposure_min=0,
    exclude_folder=exclude_folder,
    keep_sky=False
)
if len(filenames) == 0:
    raise SystemExit(f"No files found for {objectname}")

Each file set includes the paired channels:

• LRS2-B: uv and orange
• LRS2-R: red and farred

Load telluric transmission curves and a response correction derived from standard stars.

telcor = fits.open('lrs2_avg_telluric.fits')[0].data[2]

response_all = fits.open('response_correction.fits')[0].data[1]
wsel = ((def_wave>9000) & (def_wave<10050)) | (def_wave<3800) | (def_wave>10200)
response = np.interp(def_wave, def_wave[~wsel], response_all[~wsel])

These curves update the 2019 baseline response and model median telluric absorption.

Create the LRS2Object instance to control data reduction.

LRS2 = LRS2Object(
    filenames,
    detwave=6600.,              # detection wavelength in Å
    wave_window=100.,           # full detection window
    red_detect_channel='red',   # channel for detection on red side
    blue_detect_channel='orange',
    ignore_mask=False
)

Tip: choose detwave between 6450–6950 Å when combining LRS2-B and LRS2-R.

The sky subtraction step removes background emission from each exposure.
This is the most tunable component of the reduction.

LRS2.subtract_sky(
    func=np.nansum,
    local=False,
    pca=False,
    correct_ftf_from_skylines=False,
    sky_radius=4.0,
    obj_radius=1.5,
    inner_sky_radius=2.5,
    outer_sky_radius=5.0,
    sky_annulus=True,
    polymodel=False,
    polyorder=5,
    ncomp=25,
    bins=25,
    peakthresh=2.,
    pca_iter=3,
    elliptical=False,
    a_radius=3.0,
    b_radius=5.5,
    rotation_angle=-50.
)

Tuning options:

• local=True for crowded or extended sources.
• Adjust sky_radius, inner_sky_radius, outer_sky_radius, and obj_radius to define background regions.
• pca=True enables principal component sky modeling for complex residuals.

Compute fiber-to-sky coordinates, extract spectra, and standardize resolution.

LRS2.get_astrometry()
LRS2.extract_spectrum(radius=2.5, use_aperture=True)
LRS2.smooth_resolution(redkernel=0.1, bluekernel=0.1)
LRS2.rectify(def_wave)

You can optionally set manual extractions (set_manual_extraction) or supply PSF-based extractions
with Gaussian models for better point-source control. If use_aperture=False, then the source is modeled through an automated pipeline and extracted using optimal weights built from the 2D Gaussian model.

Normalize, compute S/N, coadd, and apply calibration.

LRS2.normalize(detwave=6600., wave_window=100., func=np.nanmedian)
LRS2.calculate_sn()
LRS2.combine_spectra()
LRS2.spec1D = LRS2.spec1D * response
LRS2.write_combined_spectrum(telcor=telcor)

Output FITS files include flux, error, wavelength, and telluric-corrected columns.

LRS2.setup_plotting()
for key in LRS2.sides:
    for L in LRS2.sides[key]:
        if L.channel in ['orange','red']:
            L.plot_image(
                radius=2.5, func=np.nanmedian, attr='skysub',
                quick_skysub=False, sky_radius=4.0, wave_window=100.,
                inner_sky_radius=2.5, outer_sky_radius=5.0,
                sky_annulus=True, elliptical=False,
                a_radius=3.0, b_radius=5.5, rotation_angle=-50.
            )
LRS2.fig.savefig(f'{objectname}_image_plot.png', dpi=150)

This creates a collapsed image with apertures and sky annuli visualized for QA.

You can create and merge spatially rectified cubes.

LRS2.make_cube(def_wave, ran=[-7., 7., -7., 7.])
LRS2.combine_cubes()
LRS2.spec3D = LRS2.spec3D * response / telcor
LRS2.write_cube(outname=f"{objectname}_combined_cube.fits")

Produce multi-panel plots of the extracted and combined spectra.

wran = [[3650, 10450], [3800, 4150], [6120, 6620], [7000, 7300], [9600, 10000]]
fig, ax = plt.subplots(5, 1, figsize=(20, 10))

for i, wr in enumerate(wran):
    wave = LRS2.spec1D.spectral_axis.value
    flux = LRS2.spec1D.flux.value
    sel = (wave > wr[0]) & (wave < wr[1])
    ax[i].step(wave[sel], flux[sel]/1e-17, lw=0.5, color='firebrick')

    for key in LRS2.sides:
        for L in LRS2.sides[key]:
            s = (L.spec1D.spectral_axis.value > wr[0]) & (L.spec1D.spectral_axis.value < wr[1])
            ax[i].plot(L.spec1D.spectral_axis.value[s],
                       L.spec1D.flux.value[s]/1e-17 * response[s],
                       lw=0.5, color='grey')

fig.savefig(f'{objectname}_spectrum_plot.png', dpi=150)

graph TD
    A[Panacea multi_*.fits] --> B[Initialize LRS2Object]
    B --> C[Subtract Sky]
    C --> D[Astrometry & Extraction]
    D --> E[Rectify to Common Grid]
    E --> F[Normalize + S/N Weighting]
    F --> G[Combine Spectra + Apply Calibrations]
    G --> H[Write Final Spectrum + Plots]
    style A fill:#d3e8ff,stroke:#003366,stroke-width:1px
    style H fill:#d3ffd3,stroke:#006600,stroke-width:1px

This example notebook provides a comprehensive, reproducible workflow for LRS2 data combination and calibration.
Users typically adjust only:

• Target selection parameters (folders, objectname)
• Sky subtraction geometry (local, sky_radius, obj_radius)
• Extraction aperture (radius)
• Detection wavelength (detwave)

The rest of the process—extraction, sky subtraction, coaddition, and calibration—is handled automatically by LRS2Multi and LRS2Object.