Effects of Jump Detection and Ramp Fitting Algorithms on NIRISS/SOSS Exoplanet Time-Series Observations

TL;DR

This study compares jump detection and ramp fitting algorithms in JWST NIRISS/SOSS exoplanet data reduction, showing that a likelihood-based method improves spectral accuracy and reduces residual scatter compared to the default pipeline.

Contribution

It introduces a likelihood-based ramp fitting method and evaluates its advantages over the default approach for JWST exoplanet time-series data.

Findings

Likelihood method reduces residual scatter by 12-18%.

Differences in transmission spectra are of order 10's to 100's ppm.

Group-level 1/f noise correction is essential for accurate results.

Abstract

Jump detection and ramp fitting are fundamental steps when elevating JWST data products from up-the-ramp measurements to integration level count rate images. Occurring at an early stage of the overall data reduction and calibration framework, these steps, and any biases introduced by them, have the potential to significantly impact end-stage scientific measurements. Here we explore the differential impacts of jump detection and ramp fitting for the current default JWST pipeline implementation versus a newly developed, likelihood-based method, in the context of NIRISS/SOSS exoplanet time-series observations. Across both on-sky data and a suite of simulated data, we find that the two methods show differences in end product transmission spectra of order 10's to 100's parts-per-million. For the simulated data in particular, we show that the likelihood method offers a 12-18% improvement in the residual scatter and maximum deviations from the simulated ground truth. Finally, by repeating our simulations across various noise prescriptions, we corroborate the need for group-level 1/f noise corrections and re-emphasize the default implementation of such procedures as a key priority for the JWST pipeline with respect to near-infrared time-series observing modes.