Speaker
Description
In this talk we present improvements to the STScI JWST pipeline for extremly deep exposures. Although we focus on NIRSpec observations with the micro-shutter array (MSA) for the GARDEN Survey, most of the results are also directly relevant for other instruments. GARDEN's extremely deep and highly dithered exposures of faint targets, namely intermediate redshift galaxies, are an ideal data set to optimize the data reduction process. In this paper we discuss the updates we have made to the JWST data reduction pipeline that increase the signal-to-noise of the output science products. Overall, we updated twelve calibration reference files and developed nine new software enhancements to the JWST data reduction pipeline that significantly improved the outcome. The various enhancements have several themes in common. The first theme is that most calibration reference files are based on algorithms that did not fully represent the underlying physics of the detectors. This is primarily due to not carefully masking bad pixels and not using contemporaneous calibration observations. Another theme is that the current pipeline did not have any masking of samples that are affected by persistence from cosmic rays and snowballs. We also developed a method to measure and correct for thermal variations that cause large scale imprints in the images. The final theme is residual amplifier effect from bias jumps, 1/f noise, and even/odd row biases. To measure the improvement of our corrections we combined a set of 5-point dithers in an unilluminated region of the detectors to form a pseudo-mosaic for both a JWST Archive reduction and our GARDEN reduction. When we compared the standard deviations of the unilluminated regions with the two reductions, we see factor of sixty decrease in the standard deviation of the rate measurements.