Galaxy Nurseries: Crowdsourced Analysis of Slitless Spectroscopic Data

Abstract

The Galaxy Nurseries project was designed to enable crowdsourced analysis of slitless spectroscopic data by volunteers using the Zooniverse online interface. The data set was obtained by the WFC3 Infrared Spectroscopic Parallel (WISP) Survey collaboration (Atek et al. 2010) and comprises NIR grism (G102 and G141) and direct imaging of 432 fields. The scientific goals of WISP Survey require reliable identification of emission lines (e.g., Atek et al. 2014; Masters et al. 2014). Spectral contamination by overlapping signals from multiple sources or diffraction orders as well as residual artifacts of the data reduction can significantly complicate the automatic detection and identification of emission lines. Visual verification of automatically detected features has proved essential to obtain a pure sample of emission lines. In Galaxy Nurseries verification of putative emission lines was delegated to citizen-scientist volunteers. Data were presented as "subject images" using the fixed format illustrated in the top panels of Figure 1. For each target, volunteers were provided with the two-dimensional spectrum (B), the corresponding one-dimensional extraction (A) and its direct image (C). Volunteers were instructed to evaluate only the feature identified by the green crosshairs and to decide whether it was a genuine emission line or more likely an artifact, providing a Boolean Valued (i.e., "Real" or "Spurious") label. Following its launch, Galaxy Nurseries was completed in only 40 days, gathering 414,360 classifications from 3003 volunteers for 27,333 putative emission lines. At least 15 classifications were obtained for each subject image. For reference, it took approximately 4.5 months for the full sample of lines to be visually inspected by two members of the WISP Survey Science Team (WSST).