1998A&A...332..429T

We have developed an automated package of object recognition software to detect and accurately characterize expanding HI supershells in nearby spiral galaxies. Our method avoids the ambiguity associated with visually identifying complex, expanding structures in spectral line datacubes. In this pilot study, an analytic expression for the time evolution of a supershell is used to derive the approximate structure of a parameterized shell expanding inside a plane-stratified gaseous disk. The resulting model is projected spatially and in velocity to determine an ``observable'' signature for the expanding structure. We then cross-correlate an observed HI datacube with an ensemble of these 3-dimensional filter kernels, which specify the predicted appearance of shells having different sizes and expansion velocities. Processed datacubes often contain correlation peaks of statistical significance, regions in which observed emission matches the appearance of a model supershell. To demonstrate our automated package, we describe properties of 50 high quality supershells in NGC 2403, comprising 2.1x10⁸M_☉ of HI and at least 2.2x10⁵⁴ ergs of kinetic energy. We are analyzing 20 other galaxies with our object recognition software, looking for variation in supershell characteristics as a function of Hubble type and star formation rate.