2026ApJ..1000...70S

The effect of supernovae (SNe) on star formation in the interstellar medium (ISM) depends sensitively on where SNe explode with respect to ISM clouds. Observationally, SN ISM environments characterized by spatially resolved gas maps can empirically guide the placement of SNe in subgrid models, but unfortunately such measurements remain scarce, as SNe are rare and often distant. Here we demonstrate a new approach-mapping the ISM around evolved massive stars that are soon to explode. These provide a substantially larger sample of “explosion sites” (than just historical SNe) in nearby galaxies that have high-resolution atomic and molecular ISM maps from the Jansky Very Large Array and Atacama Large Millimeter/submillimeter Array. We demonstrate this technique in the well-resolved Local Group spiral M33 by analyzing the 50 pc scale projected ISM densities around red supergiants (RSGs; 8-30 M_⊙ _stars) Wolf-Rayet stars (W-Rs; >30M_⊙ _stars), and supernova remnants. We find a mass-dependent correlation between stars and gas clouds, with at least 45% of W-Rs and up to 77% of RSGs having no detectable H₂ at their pixel locations. In the sample with H₂ detections, we find that more-massive younger progenitors are coincident with denser gas. We show that the density distributions for stars >15 M_⊙ _are statistically distinct from random alignment of stars and gas in M33. Our work provides the first observationally derived estimate of the fraction of the SN-producing stellar population correlated with ISM density peaks. We demonstrate how this can be compared with galaxy simulations, and advocate similar comparisons to the community for constraining subgrid models.