There are several reasons why HYSPLIT is not useful for close-range prediction. First, the turnaround time for a custom HYSPLIT run is long enough that a cloud of hazardous material could travel miles from the release site before the output becomes available. Second, HYSPLIT is a Lagrangian puff model designed for applications in wind fields that vary in space and time, while close to the source, plumes generally follow the typical straight-line transport pattern assumed by CAMEO/ALOHA.
HYSPLIT evaluations with field data have not focused on near-field scenarios. In contrast, CAMEO/ALOHA is designed for close-range applications and has been optimized for quick turnaround.
Hope this helps