From data to meaning -- the challenge of inferring the past from astronomical data

Dr. Louis Abramson (Carnegie)
Tuesday, August 24, 2021 - 11:00am

The theory of galaxy evolution should predict low-redshift galaxy properties from high-redshift physical quantities. Inferred star formation histories (SFHs) have yielded novel insights and are attractive empirical tests. While obtaining them is standard, situating them in any past context is not. To find out if it’s possible, I tried to match mock z=0 data to the SFHs that generated them as they appeared 3 Gyr earlier. Under unrealistically good conditions, only 13% of inferences contained both a correctly identified progenitor and the true SFH. This fraction is independent of specific star formation rate, dust content, or metallicity. Worse, false-positives abound, and can have SFHs that differ markedly from those of the correct progenitors. Informatic constraints play a part in this unfortunate outcome, but the fact that progenitor candidates must be identified via a common terminus while real galaxy cohorts share only a common origin provides a higher knowledge floor, ultimately placing SFH-based investigations in the same epistemological bind as those based on observed quantities. Progress will come from abstracting away from the measurement/inferential level to (1) identify the SFH prior that describes real galaxies, (2) associate observed environmental or kinematic trends with bulk SFH properties, or (3) quantify the constraining power of the data. While not the same as saying “these galaxies did this," all three investigations could lay the groundwork for a truly meaningful confrontation of data and theory.

Talk Type: