Title: DES Science Portal: I - Computing Photometric Redshifts

Abstract: We present the Dark Energy Survey (DES) Science Portal, an integrated web-based data interface designed to facilitate scientific analysis. We demonstrate how the Portal can provide a reliable environment to access complete data sets, provide validation algorithms and metrics in the case of multiple methods and training configurations, and maintain the provenance between the different steps of a complex calculation, while ensuring reproducibility of the results. We use the estimation of DES photometric redshifts (photo-$z$s) as an example. A significant challenge facing photometric surveys for cosmological purposes, such as DES, is the need to produce reliable redshift estimates. The choice between competing algorithms and configurations and the maintenance of an up-to-date spectroscopic database to build training sets, for example, are complex tasks when dealing with large amounts of data that are regularly updated and constantly growing. We show how the DES Science Portal can be used to train and validate several photo-$z$ algorithms using the DES first year (Y1A1) data. The photo-$z$s estimated in the Portal are used to feed the creation of catalogs for scientific workflows. While the DES collaboration is still developing techniques to obtain precise photo-$z$s, having a structured framework like the one presented here is critical for the systematic vetting of DES algorithmic improvements and the consistent production of photo-$z$s in future DES releases.