Here’s some more interesting work from Joel Anderson: an examination of recent hazardous liquid spill volumes, from the PHMSA reporting databases.
Analyses like these not only help us with basic understanding of pipeline releases, but also provide useful input into risk assessments. In many cases, our estimates of location-specific loss potential are based on plausible scenarios taken from history. This analyses shows us that history.
Furthermore, since we usually need more than P90+ (worst case) scenarios, the distributions of historical ‘behavior’ shown here help us produce P50 or any other conservatism-dictated values.