Simulations Find Our Accounting of Dust-Obscured Star Formation May Be Incomplete

The bulk of the star-formation rate density peak at cosmic noon was obscured by dust. How accurately we can assess the role of dust obscured star-formation is affected by inherent biases in our empirical methods -- both those that rely on direct dust emission and those that rely on the inferred dust attenuation of starlight. We use a library of hydrodynamic simulations with radiative transfer to explore these biases. We find that for IR luminous galaxies that are in rapidly quenching systems (e.g. post-coalescence) standard luminosity-to-SFR relations can strongly overestimate the true SFRs. We propose using the $L_{IR}/L_{1.6}$ color to both help identify such systems and provide more accurate SFRs. Conversely, we find that the diagnostic UVJ plot misidentifies a subset of dusty star-forming galaxies. This is due to variability in the effective attenuation curves including being much grayer in the optical-to-near-IR regime than the Calzetti starburst law. This is in agreement with recent observations of IR-selected galaxies at cosmic noon. Our results support the view that we need a panchromatic approach from the rest-frame UV through the IR and SED modeling that includes realistic SFHs and allows for variable attenuation curves if we want to fully account for dust obscured star-formation across the epochs of greatest galaxy build-up.