Revisiting the cosmic star formation history: Caution on the uncertainties in dust correction and star formation rate conversion

The cosmic star formation rate density (CSFRD) has been observationally investigated out to redshift z ≃ 10. However, most of the theoretical models for galaxy formation underpredict the CSFRD at z ≳ 1. Since the theoretical models reproduce the observed luminosity functions (LFs), luminosity densities (LDs), and stellar mass density at each redshift, this inconsistency does not simply imply that theoretical models should incorporate some missing unknown physical processes in galaxy formation. Here, we examine the cause of this inconsistency at UV wavelengths by using a mock catalog of galaxies generated by a semi-analytic model of galaxy formation. We find that this inconsistency is due to two observational uncertainties: the dust obscuration correction and the conversion from UV luminosity to star formation rate (SFR). The methods for correction of obscuration and SFR conversion used in observational studies result in the overestimation of the CSFRD by ∼0.1-0.3 dex and ∼0.1-0.2 dex, respectively, compared to the results obtained directly from our mock catalog. We present new empirical calibrations for dust attenuation and conversion from observed UV LFs and LDs into the CSFRD.