Mitigating Resampling Artifacts for the JWST IFU Spectrometers with Adaptive Trace Modeling
David R. Law, Melanie Clarke · 2026
Read Mitigating Resampling Artifacts for the JWST IFU Spectrometers with Adaptive Trace Modeling by David R. Law, Melanie Clarke (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2603-26916.txt.
Introduction
Law and Clarke tackle resampling artifacts in the JWST integral-field spectrometers. The instruments undersample the telescope's sharp point spread function. Resampling into regular data cubes then creates large oscillating spectral artifacts. Their fix models the observed spectral traces with cubic basis splines. Interpolating onto a finer grid before cube building largely removes the artifacts. The method also reduces bias from clusters of bad pixels. Tests span NIRSpec and MIRI MRS data, from point sources to protostars with rich molecular bands.
Catalog record
| CREATOR | David R. Law, Melanie Clarke |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 51233 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2603.26916
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Mitigating Resampling Artifacts for the JWST IFU Spectrometers with Adaptive Trace Modeling David R. Law, Melanie Clarke Abstract The integral-field unit (IFU) spectrometers on board the James Webb Space Telescope (JWST) undersample the nearly diffraction-limited point spread function provided by the telescope optics. This undersampling produces large oscillating spectral artifacts when the data is resampled into regularly-gridded data cubes, which poses a significant challenge for many scientific analyses. We describe here a generalized technique to use cubic basis spline models to interpolate the observed spectral traces onto a higher-resolution grid prior to data cube rectification, which largely eliminates these artifacts in addition to helping reduce biases in point source spectra from clusters of bad pixels. We demonstrate the utility of this adaptive resampling technique for a variety of JWST NIRSpec and MIRI MRS observations ranging from isolated point sources to embedded AGN, crowded stellar fields with diffuse emission, and protostars with rich molecular bands. I Introduction According to the Nyquist-Shannon theorem, the sampling frequency of any signal must be at least twice as fast as the highest-frequency component of that signal in order to avoid aliasing power into lower frequency components during any resampling of the original signal. One of the direct implications of this for standard astronomical observations is that we must sample the point spread function (PSF) with at least two detector pixels per FWHM in order to avoid introducing artifacts when resampling data from the native pixel grid as is commonly done when combining dithered observations. This requirement can be particularly challenging for space-based instruments that achieve near diffraction-limited performance, and the effects of undersampling have been well documented for the Hubble, Spitzer, and James Webb space telescopes. Anderson 2016 for instance noted the impact of undersampling on reconstruction of the point-spread function for HST/WFC3 imaging data, which can complicate efforts to perform outlier detection using resampled data. In the spectral domain artifacts can be even more obvious, as spectra dispersed on a grid of detector pixels are typically tilted with respect to that grid, resulting in pixel samplings that differ as a function of wavelength. Dressel et al. 2007 noted this effect for HST/STIS, while Smith et al. 2007 described its impacts on efforts to create spectral maps from the Spitzer telescope’s Infrared Spectrograph. More recently, Law et al. 2023 described how undersampling can produce resampling artifacts in spectral data cubes for the slicer-type NIRSpec ( Böker et al. 2022 ) and MIRI ( Wright et al. 2023 ) integral-field units aboard the James Webb Space Telescope. As demonstrated by Law et al. 2023 , these resampling artifacts manifest as low-frequency oscillations in the spectra of drizzled data cubes, with the oscillation frequency being wavelength-dependent and tied to the tilt angle of the dispersed trace with respect to the detector pixel grid. These artifacts are strongest in the spectra of individual cube spaxels near unresolved point sources, are partially mitigated by dithering, and average out when extracting spectra from apertures comparable in size to the observational PSF. Immediate solutions to the problem are thus to either perform scientific analyses in the native detector pixel space, or to extract spectra from resampled data cubes using sufficiently large extraction apertures. Neither solution is ideal. Rectified and coadded data cubes, for instance, are much more convenient and easy to work with than the often-complex formats of the native detector-level data. At the same time, crowded astronomical scenes often require analysis of extremely small areas of sky in order to minimize source confusion. Indeed, a common method of IFU data analysis is to apply spectral fitting techniques to the spectra of each cube spaxel, and thus build up 2d maps of line emission, kinematics, or a host of other physical properties ( Westfall et al. 2019 , e.g.,) . Various groups have thus developed techniques to correct for these artifacts in JWST data. As part of the GA-NIFS survey for instance, Perna et al. 2023 corrected for resampling artifacts via sinusoidal modeling of spectra from the resampled cubes. Dumont et al. 2025 likewise developed the wicked package using Fourier analysis to model and remove regular sinusoidal features from affected cube spaxels. Most recently, Shajib 2025 developed the software package raccoon using sinusoidal chirp functions to correct the data cubes with template spectra derived from small apertures. All three implementations can be effective for specific science cases, but intrinsically rely on measuring and removing the artifacts after they have already been introduced into the data. As such, all require extensive customization to ensure that genuine spectral features do not bias the correction routines. Here, we present a turn-key method that uses adaptive PSF-modeling of spectral traces in the native detector space to avoid creating such artifacts in the IFU data cubes in the first place, thereby allowing it to handle arbitrary scenes and source spectra with little to no tuning. We describe our algorithm in § II , and evaluate its performance for point sources observed with both the MIRI MRS and NIRSpec IFU in § III . We discuss the application of this technique to a variety of astronomical data sets in § IV and Appendix A . We summarize our results in § V . II Adaptive Trace Model (ATM) Method As outlined in § I , the general problem of mitigating artifacts that arise when resampling undersampled data is unsolvable as the necessary information is irretrievably lost. However, in our specific case it is possible to bring to bear additional information provided by our knowledge of astronomical optics and the physical properties of the instruments. In particular, for well-calibrated instruments we know that the PSF should change only slowly with wavelength, and that unresolved point sources should have centroids that remain fixed in celestial coordinates at all wavelengths. These two assumptions allow us to take advantage of the very pixel boundary crossings that produce the sinusoidal artifacts in the first place, and are sufficiently powerful that they can largely eliminate the resampling problem in most JWST observations. In Figure 1 we show a NIRSpec IFU observation of an unresolved point source obtained by JWST program ID (PID) 3399 (PI: M. Perrin). In the left-hand panel we show the calibrated (‘ \ast\_ cal.fits’) image from the NRS1 detector. Each of the horizontal stripes corresponds to one of the 30 IFU slices, arranged in the order of their across-slice position \beta . The spectrum of each slice is dispersed horizontally with increasing wavelength to the right; within a given slice each column thus corresponds to a different wavelength. While the FWHM of this point source is comparable to the width of a single slice, the star is bright enough that the wings of the PSF illuminate multiple slices out to a distance of about 0.5 arcsec. Zooming in on the dispersed trace of the bright PSF core (Figure 1 , right-hand panel) we show that each stripe is about 30 pixels tall in the along-slice spatial direction \alpha . Each slice can thus be thought of as a dispersed long-slit spectrum; where a slice covers the PSF core the normalized spatial profile is narrow, and where a slice covers the PSF wings we observe Airy rings and other structures characteristic