Aggregate odds I think
I would guess it's something like:
Odds of coming down over a populated area e.g. 1 in 10
x Odds of hitting someone if a piece comes down in a populated area, e.g. if it comes down over a square kilometre with 5,000 inhabitants and we treat a piece entering the same square metre as a person a 'hit', then each piece has a 5,000 in 1,000,000 (1 in 200) chance of hitting.
So that's each piece got a 1/10 chance of having a 1/200 chance = 1/10 * 1/200 = 1/2000. (And with 30 pieces that's 30 1 in 2000 chances.)
Of course this example is just made up numbers, but then gain so is their estimate - if they don't know when it's going to come back to Earth (2 days+ of uncertainty) then they don't know where it's going to come down, hence the population of the theoretical impact area is uncertain at best. think of two scenarios for touchdown, one over the ocean, another over a capital city - wildly different odds of striking a person.
The important thing is that if they calculate this risk the same way every time, they can tell which impact scenarios are (relatively) more or less dangerous even if the numbers they are using turn out to be wrong.