We use local Cartesian simulations with a vertical gravitational potential to study how supernova ( SN ) feedback in stratified galactic discs drives turbulence and launches galactic winds .
Our analysis includes three disc models with gas surface densities ranging from Milky Way-like galaxies to gas-rich ultra-luminous infrared galaxies ( ULIRGs ) , and two different SN driving schemes ( random and correlated with local gas density ) .
In order to isolate the physics of SN feedback , we do not include additional feedback processes .
We find that , in these local box calculations , SN feedback excites relatively low mass-weighted gas turbulent velocity dispersions \approx 3 - 7 km s ^ { -1 } and low wind mass loading factors \eta \lesssim 1 in all the cases we study .
The low turbulent velocities and wind mass loading factors predicted by our local box calculations are significantly below those suggested by observations of gas-rich and rapidly star-forming galaxies ; they are also in tension with global simulations of disc galaxies regulated by stellar feedback .
Using a combination of numerical tests and analytic arguments , we argue that local Cartesian boxes can not predict the properties of galactic winds because they do not capture the correct global geometry and gravitational potential of galaxies .
The wind mass loading factors are in fact not well-defined in local simulations because they decline significantly with increasing box height .
More physically realistic calculations ( e.g. , including a global galactic potential and disc rotation ) will likely be needed to fully understand disc turbulence and galactic outflows , even for the idealized case of feedback by SNe alone .