Realizing the Full Potential of Large Surveys
Twenty years ago the first digital surveys of the sky were in progress, eventually releasing optical and near-infrared catalogs of hundreds of millions of stars across the sky. Next-generation photometric and spectroscopic surveys are now underway, detecting ten to a hundred times as many stars and galaxies as these first surveys. Achieving the cosmological goals of these surveys requires more precise analysis than currently realized. In this talk, I will discuss three areas in which I have tried to enhance the accuracy obtainable in large surveys. First, I will present my work to simultaneously model the stars and galaxies in the WISE and DECaPS surveys---each of which contains billions of objects---to address the impact of blending, an increasingly prominent concern. Second, I will present work characterizing the throughput of surveys like PS1 and DECaLS, where we currently obtain 0.7% repeatability, but a factor of a few improvement is possible through improved modeling of the system. Finally, I will present improved maps of dust and its properties, mitigating a primary systematic uncertainty in astronomical surveys. These three areas---modeling, calibration, and astrophysical systematics---present major challenges to next-generation surveys, but likewise provide opportunities for dramatic improvements in the near future.