Glossary#

This glossary defines terminology used in spectroscopic data reduction, with emphasis on long-slit spectroscopy and the specreduce package.

1D Spectrum#: Flux (or flux density) as a function of wavelength, frequency, or pixel position. This is the primary output of spectroscopic reduction, extracted from a 2D spectrum. A 1D spectrum may or may not be wavelength calibrated or flux calibrated. In specreduce and specutils, 1D spectra are represented by the specutils.Spectrum class.
2D Spectrum#: A two-dimensional image from a CCD or similar detector containing dispersed light from one or more sources. One axis corresponds to wavelength (the dispersion axis) and the other to spatial position along the slit (the cross-dispersion axis). Also called a spectral image. A 2D spectrum is the input to the extraction process that produces a 1D spectrum , and is the starting point for reduction in both long-slit spectroscopy and fiber-fed spectroscopy.
Absorption Line#: A feature in a spectrum where flux falls below the local continuum level at a characteristic wavelength, produced when atoms or molecules absorb photons at specific energies. Absorption lines in stellar spectra reveal the composition and physical conditions of stellar atmospheres, while interstellar and telluric absorption lines trace intervening material along the line of sight.
Airmass#: The path length of light through Earth’s atmosphere, expressed as a ratio relative to the path length at zenith. In the plane-parallel approximation, airmass = sec(z), where z is the zenith angle. At zenith, airmass = 1.0; at 60\(^\circ\) from zenith, airmass \(\approx\) 2.0. Airmass affects the amount of atmospheric extinction and is used in flux calibration.
Aperture#: A defined region on a 2D spectral image from which flux is extracted. In long-slit spectroscopy, the aperture is typically centered on the trace and has a specified aperture width. The aperture defines which pixels contribute to the extracted 1D spectrum.
Aperture Width#: The spatial extent (in pixels) over which flux is summed during extraction. For boxcar extraction, all pixels within the aperture width receive equal weight. The optimal width balances including most of the source flux while minimizing background noise.
Arc Lamp#: A calibration light source that produces emission lines at known wavelengths, used for wavelength calibration. Common arc lamps include helium (He), argon (Ar), neon (Ne), and mercury-argon (HgAr) combinations. The resulting calibration exposure is called an arc spectrum.
Arc Spectrum#: A calibration exposure taken with an arc lamp, showing discrete emission lines at known wavelengths. Arc spectra are used to establish the wavelength solution that maps pixel positions to wavelengths.
Atmospheric Extinction#: The wavelength-dependent absorption and scattering of light by Earth’s atmosphere. Atmospheric extinction increases at shorter wavelengths and with higher airmass. Correction for atmospheric extinction is part of flux calibration.
Background#: Signal in a spectral image that does not originate from the science target, including sky emission, scattered light, detector dark current, and other instrumental artifacts. Background subtraction removes this contaminating signal before or during extraction. In specreduce, the Background class handles background estimation and subtraction.
Background Aperture#: The spatial region(s) used to measure the background level. In two-sided background subtraction, apertures are placed on both sides of the trace; in one-sided background subtraction, a single aperture is used on one side.
Barycentric Correction#: A velocity correction that accounts for Earth’s orbital motion around the Solar System barycenter (center of mass). Applying barycentric correction converts observed wavelengths to the reference frame where the Solar System barycenter is at rest, enabling precise radial velocity measurements. See also heliocentric correction.
Bias Frame#: A zero-second exposure that captures the electronic offset (bias level) of the detector. Bias subtraction is typically performed during preprocessing to remove this fixed offset from science exposures.
Binning#: Division of an axis into discrete sections. In trace fitting, binning along the dispersion axis allows independent measurement of the trace position in each bin, which are then fit with a smooth function. In detector readout, binning combines adjacent pixels to reduce read noise at the cost of spatial or spectral resolution.
Boxcar Extraction#: A simple extraction method that sums all pixel values within the aperture with equal weights. Boxcar extraction is straightforward but does not account for the spatial profile shape or pixel-to-pixel variations in noise. In specreduce, this is implemented by the BoxcarExtract class. Compare with optimal extraction.
Calibrated 2D Image#: A 2D spectrum that has been processed to have a wavelength calibration applied (so pixel positions correspond to known wavelengths) but has not been resampled or rectified. The original pixel values are preserved.
Catalog Lines#: Reference wavelengths of emission lines from standard line lists, used to match against detected lines in an arc spectrum during wavelength calibration. Common catalogs include HeI, ArI, NeI, and HgI lines.
CCDData#: An Astropy data class (astropy.nddata.CCDData) for representing CCD images with associated uncertainty, mask, and metadata. CCDData is a subclass of NDData and is commonly used for spectroscopic images in specreduce.
Centroid#: The flux-weighted center position of a feature, such as an emission line or the spatial profile of a source. In trace fitting, centroiding is one method for determining the trace position at each dispersion axis bin.
Classification#: The process of identifying the type of astronomical object a spectrum represents (e.g., star, galaxy, quasar). Classification may be performed automatically through model fitting or manually through visual inspection.
Coadding#: Combining multiple spectra of the same object into a single spectrum, typically to improve signal-to-noise ratio or to merge spectra covering different wavelength ranges. Unlike stacking, coadding specifically refers to combining spectra of the same source.
Cross-Dispersion Axis#: The axis perpendicular to the dispersion axis in a 2D spectrum. In long-slit spectroscopy, the cross-dispersion axis corresponds to spatial position along the slit. Also called the spatial axis.
Dark Frame#: An exposure taken with the detector shutter closed, capturing the thermal signal (dark current) accumulated during an exposure. Dark subtraction during preprocessing removes this signal from science exposures.
Data Cube#: A three-dimensional data structure with two spatial dimensions and one spectral dimension. Data cubes are the standard output format for IFU observations, where each spatial pixel (spaxel) contains a complete spectrum. Also called a spectral data cube or hyperspectral cube.
Dispersion#: The separation of light by wavelength, or quantitatively, the change in wavelength per pixel (\(d\lambda/d\text{pixel}\)) along the dispersion axis. A spectrograph with higher dispersion spreads the spectrum over more pixels, providing higher spectral resolution. Not to be confused with resolving power.
Dispersion Axis#: The axis along which wavelength varies in a 2D spectrum. Light at different wavelengths falls on different positions along this axis.
Emission Line#: A feature in a spectrum where flux exceeds the local continuum level at a characteristic wavelength, produced when atoms or molecules emit photons at specific energies. Emission lines range from spectrally unresolved features in arc lamps to broad features in active galactic nuclei. Emission lines from arc lamps are used for wavelength calibration; emission lines from astronomical sources provide information about their physical conditions, composition, and kinematics.
Extinction Curve#: The wavelength-dependent function describing atmospheric extinction. Standard extinction curves (e.g., for specific observatories) provide the extinction in magnitudes per unit airmass as a function of wavelength.
Extraction#: The process of converting a 2D spectrum into a 1D spectrum by summing or averaging flux along the cross-dispersion axis within a defined aperture. In specreduce, the two main extraction methods are boxcar extraction and optimal extraction, performed by the BoxcarExtract and HorneExtract classes, respectively.
Fiber-Fed Spectroscopy#: A spectroscopic technique in which optical fibers relay light from the telescope focal plane to the spectrograph. Fibers decouple the spectrograph from the telescope, allowing it to be mounted in a stable environment, and enable flexible positioning of inputs across the focal plane. Fiber-fed designs are used in MOS instruments and IFU instruments, where each fiber produces a 2D spectrum on the detector.
FITS#: Flexible Image Transport System. A common file format for astronomical data, capable of storing images, tables, and metadata (headers). Most spectroscopic data, including raw and reduced spectra, are stored in FITS format.
Flat Field#: A calibration exposure of a uniformly illuminated source (e.g., dome flat, twilight flat, or lamp flat) used to measure pixel-to-pixel sensitivity variations. Flat-field correction during preprocessing divides science exposures by a normalized flat to remove these variations.
Flux#: In the strict physical sense, energy per unit time per unit area (\(\text{W/m}^2\)). In spectroscopy, “flux” is often used informally to refer to flux density or simply to the dependent variable (brightness) of a spectrum. The specutils Spectrum class uses flux as the attribute name for the spectral data values.
Flux Calibration#: The process of converting spectral data from instrumental units (counts, ADU, electrons) to physical flux density units (e.g., \(\text{erg/s/cm}^2\text{/\AA}\) or \(\text{W/m}^2\text{/nm}\)). Flux calibration requires observations of standard stars with known spectra and corrections for atmospheric extinction.
Flux Density#: Flux per unit wavelength or frequency interval, the standard physical quantity for spectroscopic measurements. Common units include \(\text{erg/s/cm}^2\text{/\AA}\), \(\text{W/m}^2\text{/nm}\), and Jy (jansky). See also flux.
FWHM#: Full Width at Half Maximum. A measure of the width of a peak or line profile, defined as the width at which the intensity drops (for emission lines) or rises (for absorption lines) to half its extreme value relative to the local continuum. FWHM is commonly used to characterize spectral resolution and the spatial profile of sources.
Gaussian Profile#: A bell-shaped curve (normal distribution) often used to model spectral lines and spatial profiles. In trace fitting, fitting a Gaussian to the cross-dispersion profile is one method for determining the trace position.
GWCS#: Generalized World Coordinate System. A Python package providing a framework for coordinate transformations, including non-linear mappings. In specreduce, GWCS is used to represent wavelength solutions produced by wavelength calibration.
Heliocentric Correction#: A velocity correction that accounts for Earth’s motion relative to the Sun. Applying heliocentric correction converts observed wavelengths to the reference frame where the Sun is at rest. See also barycentric correction, which provides higher precision.
IFU#: Integral Field Unit. An instrument or observing mode that obtains spectra simultaneously over a contiguous two-dimensional field of view, producing a data cube. IFUs use techniques such as fiber bundles, image slicers, or lenslet arrays to sample the field. Also called integral field spectroscopy (IFS).
Interpolated Profile#: In optimal extraction, an empirical spatial profile constructed by sampling the actual profile at multiple wavelength bins and interpolating between them. This accounts for wavelength-dependent variations in the profile shape. In specreduce’s HorneExtract, this is enabled with the interpolated_profile option.
Inverse Variance Weighting#: A statistical method for combining measurements that weights each value by the inverse of its variance (\(1/\sigma^2\)), giving more weight to higher-precision measurements. Inverse variance weighting is used in optimal extraction to maximize signal-to-noise ratio.
Line Matching#: The process of associating detected emission lines in an arc spectrum with known wavelengths from catalog lines. Line matching is a key step in wavelength calibration, establishing the correspondence between pixel positions and wavelengths.
Long-Slit Spectroscopy#: A spectroscopic observing technique using an elongated slit (typically a few arcseconds wide and tens of arcseconds to arcminutes long) to obtain spectra of extended objects or to capture spatial information along one dimension. The resulting 2D spectrum contains spectral information along one axis and spatial information along the other.
Mask#: A boolean array indicating which pixels in an image should be excluded from analysis (True = masked/bad, False = good). Masks flag cosmic rays, bad pixels, saturated pixels, and other defects. Specreduce provides various mask treatment options for handling masked pixels.
Mask Treatment#: The strategy for handling masked or non-finite pixels during processing. Options in specreduce include: apply (use mask as-is), ignore (discard existing mask), propagate (extend mask along cross-dispersion), zero_fill (replace masked values with 0), nan_fill (replace with NaN), and others. See the Mask Treatment documentation for details.
MOS#: Multi-Object Spectroscopy. A technique for obtaining spectra of multiple objects simultaneously using multiple slits, fiber positioners, or other multiplexing methods. Each object produces its own 2D spectrum on the detector.
NDData#: N-dimensional data. An Astropy base class (astropy.nddata.NDData) for representing data with associated uncertainty, mask, unit, and WCS. Specreduce accepts NDData objects as input and uses them internally.
One-Sided Background#: A background estimation method using a single aperture on one side of the trace. This is useful when the background on one side is contaminated (e.g., by a nearby source). In specreduce, use specreduce.background.Background.one_sided() to create a one-sided background.
Optimal Extraction#: An extraction method (also called Horne extraction, [Horne1986]) that weights pixels according to the spatial profile and their uncertainties, maximizing the signal-to-noise ratio of the extracted spectrum. Optimal extraction accounts for the varying contribution of different pixels and can reject cosmic rays. In specreduce, this is implemented by HorneExtract.
Order Sorting Filter#: An optical filter used to block unwanted orders of diffraction (e.g., preventing 2nd-order blue light from overlapping with 1st-order red light). Failure to use or account for these filters results in systematic errors during flux calibration.
Peak Method#: The algorithm used to locate the position of a spectral trace within each bin along the dispersion axis. Options in specreduce’s FitTrace include max (pixel with maximum value), gaussian (fit Gaussian and use center), and centroid (flux-weighted center).
Pipeline#: An automated sequence of processing steps that transforms raw data into reduced, calibrated data products. A spectroscopic reduction pipeline typically includes preprocessing, extraction, wavelength calibration, and flux calibration. Specreduce provides the building blocks for constructing such pipelines.
Pixel Scale#: The physical or wavelength interval corresponding to one pixel. In the spectral direction, pixel scale refers to the dispersion (wavelength per pixel). In the spatial direction, it refers to the angular size per pixel (arcseconds/pixel).
Preprocessing#: Initial processing steps applied to raw CCD images before extraction, including bias subtraction, dark subtraction, flat field correction, and bad pixel masking. Also called instrument signature removal. In the Astropy ecosystem, preprocessing is typically handled by the ccdproc package.
Rectified Spectrum#: A spectrum that has been resampled so that the dispersion axis is aligned with image rows or columns, with constant wavelength spacing. Rectification involves resampling. A rectified 2D spectrum has one axis that is purely spectral.
Redshift#: The fractional change in wavelength of spectral features due to the motion of the source (Doppler shift) or the expansion of the universe (cosmological redshift), defined as \(z = (\lambda_\text{observed} - \lambda_\text{rest}) / \lambda_\text{rest}\).
Reduction#: The process of converting raw observational data into calibrated, science-ready spectra. For spectroscopy, reduction typically includes preprocessing, extraction, background subtraction, wavelength calibration, and optionally flux calibration. The term comes from “reducing” the complexity of the data, though in practice the data volume often increases.
Reference Pixel#: An anchor point for polynomial wavelength solutions, typically near the center of the detector. The wavelength solution coefficients are defined relative to this reference position.
Resampling#: The process of binning spectral data onto a new wavelength or pixel grid. Resampling is needed when combining spectra with different wavelength solutions or when creating rectified spectra. Resampling should be performed with care to preserve flux and properly propagate uncertainties.
Resolving Power#: The ability of a spectrograph to distinguish closely spaced spectral features, defined as \(R = \lambda/\Delta\lambda\), where \(\Delta\lambda\) is the minimum resolvable wavelength difference (related to FWHM). Higher resolving power means finer spectral detail can be resolved. Not to be confused with dispersion.
Row-Stacked Spectra#: A collection of 1D spectra stored as a 2D array with one spectrum per row, sharing a common spectral axis. This format is used by specutils Spectrum when representing multiple spectra.
Sensitivity Function#: The wavelength-dependent response of the instrument, relating observed counts to true flux. The sensitivity function accounts for telescope throughput, instrument efficiency, and detector quantum efficiency. It is derived during flux calibration using standard star observations.
Sigma Clipping#: An iterative outlier rejection method that excludes values more than a specified number of standard deviations (sigma) from the mean or median. In specreduce, sigma clipping can be applied during background estimation to reject cosmic rays and other outliers.
Sky#: Emission from Earth’s atmosphere, including airglow lines, scattered moonlight, and light pollution. Sky emission must be subtracted from spectra to isolate the astronomical signal. See sky subtraction.
Sky Subtraction#: The removal of additive sky emission from spectra. In long-slit spectroscopy, sky is typically measured from regions of the slit away from the target and subtracted from the object spectrum. Sky subtraction is closely related to background subtraction.
Slit#: The narrow aperture at the focal plane of a spectrograph that defines the region of the sky being observed. In long-slit spectroscopy, the slit is elongated in one direction to capture spatial information. The slit width affects spectral resolution; the slit length determines the spatial coverage.
Slit Curvature#: The geometric distortion where a straight slit appears curved on the 2D spectrum. This occurs because the off-axis rays from the slit meet the grating at different angles. This must be corrected during wavelength calibration or tilt correction.
Spatial Axis#: See cross-dispersion axis.
Spatial Profile#: The distribution of flux across the slit in the cross-dispersion axis direction at a given wavelength. The spatial profile is typically determined by the seeing conditions and telescope optics, often approximated as a Gaussian. In optimal extraction, accurate modeling of the spatial profile is essential for maximizing signal-to-noise.
Spaxel#: A spatial pixel in a data cube, corresponding to a single spatial element of an IFU observation. Each spaxel contains a complete spectrum spanning the spectral axis of the data cube. The spatial sampling of a spaxel is determined by the IFU design and may not be square.
Spectral Resolution#: The ability to distinguish spectral features at nearby wavelengths, often characterized by the FWHM of unresolved lines. Can be expressed as the resolution element \(\Delta\lambda\) or as resolving power \(R = \lambda/\Delta\lambda\). Higher spectral resolution reveals finer detail in spectra.
Spectrum#: The distribution of light intensity as a function of wavelength or frequency. In specreduce and specutils, Spectrum specifically refers to the specutils class representing spectral data with flux, spectral axis, and optional uncertainty and mask.
Stacking#: Combining multiple spectra or images, typically to increase signal-to-noise ratio. In some contexts, stacking refers specifically to combining data without alignment (e.g., NumPy array stacking), while coadding implies proper combination of related spectra.
Standard Star#: A star with a well-calibrated spectrophotometric spectrum, used as a reference for flux calibration. Observations of standard stars under the same conditions as science targets allow determination of the sensitivity function. Spectrophotometric standards include Vega, BD+17 4708, and various white dwarfs.
Telluric Correction#: The removal of absorption features caused by molecules in Earth’s atmosphere (telluric features), such as \(\text{O}_2\), \(\text{H}_2\text{O}\), and \(\text{CO}_2\) bands. Telluric correction addresses the multiplicative effect of atmospheric absorption, distinct from the additive sky emission. Methods include division by telluric standard star spectra or model fitting.
Tilt Correction#: The process of correcting geometric distortions in a 2D spectrum where lines of constant wavelength are not aligned with detector columns (or rows). These distortions arise from slit curvature and optical aberrations in the spectrograph, causing emission lines to appear tilted or curved across the cross-dispersion axis. Tilt correction fits a 2D polynomial transformation between tilt-corrected and detector coordinate spaces using arc spectra. In specreduce, this is handled by TiltCorrection (calibration) and TiltSolution (transformation and resampling).
Trace#: The path of a spectrum across the 2D spectrum image, defining where the spectrum falls on the detector as a function of wavelength. Due to optical distortions, the trace may not be a straight line. In specreduce, trace classes include FlatTrace (constant position), ArrayTrace (arbitrary positions), and FitTrace (fit to image features).
Trace Fitting#: The process of determining the trace position as a function of wavelength by measuring the spectrum’s location in bins along the dispersion axis and fitting a smooth function (typically a polynomial) to these positions. In specreduce, FitTrace performs trace fitting with configurable peak method and polynomial order.
Two-Sided Background#: A background estimation method using apertures on both sides of the trace, symmetrically placed at equal distances (separation) from the center. The background values from both sides are averaged. In specreduce, use specreduce.background.Background.two_sided() to create a two-sided background.
Uncertainty#: A measure of the error or noise associated with each data value. Uncertainties are typically expressed as standard deviations or variances and should be propagated through all reduction steps. Specreduce supports uncertainty propagation in background subtraction and extraction.
Variance#: The square of the standard deviation, representing the expected squared deviation from the mean. Variances are used in inverse variance weighting and are propagated through calculations according to error propagation rules.
Visual Inspection#: Human examination of spectra or data products to verify quality, identify features, or make scientific judgments. Visual inspection remains important for assessing reduction quality and identifying issues that automated methods may miss.
Wavelength Calibration#: The process of determining the relationship between pixel position and wavelength, establishing a wavelength solution. Wavelength calibration typically uses arc spectra with known emission lines. In specreduce, this is handled by the WavelengthCalibration1D class, which produces a GWCS-based WCS.
Wavelength Solution#: A mathematical function (typically a polynomial) that maps pixel positions to wavelengths. The wavelength solution is determined through wavelength calibration and is applied to the data as a WCS.
WCS#: World Coordinate System. A standard for describing the mapping between pixel coordinates and physical coordinates such as wavelength, sky position, or other quantities. For spectra, the WCS maps pixel numbers to wavelengths. The FITS WCS standard and GWCS are common implementations. See the Astropy WCS documentation for details.
Window#: A restricted region used for analysis, such as a spatial window for limiting trace fitting to a subset of the cross-dispersion axis (useful when multiple objects are present), or a spectral window limiting analysis to a wavelength range.
Workflow#: A complete sequence of processing and analysis steps, often broader than a pipeline, potentially including data organization, job orchestration, quality control, and archiving in addition to the core reduction steps.

References#

[Horne1986]

Horne, K. (1986). “An optimal extraction algorithm for CCD spectroscopy.” Publications of the Astronomical Society of the Pacific, 98, 609.