Lets take two steps back, since cvg and peter are both right.
For a given location, the worst 6 hour long storm you're likely to see in 10 years will be of some depth (in mm in your case). That's the "10 year 6 hour rainfall depth."
For the same location, the worst 12 hour long storm you're likely to see in 10 years will be of some other depth (in mm in your case). That's the "10 year 12 hour rainfall depth."
For the same location, the worst 24 hour long storm you're likely to see in 10 years will be of some other depth (in mm in your case). That's the "10 year 24 hour rainfall depth."
Follow so far?
Your municipality should have given you a duration when they said "10 year," because "10 year" by itself doesn't give you enough information to know the rainfall depth.
If you were purely using the Rational Method to determine a peak flow rate, and didn't care at all about volume, then you could make the additional assumption that the worst intensity you were likely to see corresponded with a storm short enough to exactly match the Tc of your basin. Then you'd go to your IDF relationship, read the intensity off based on the Tc, and use that in Q=CIA to determine peak flow. That's all those IDF curves are "supposed" to be used for. For peak flow rates, not for volume calculations.
However.
You can back figure what the total depth of rainfall is for a given duration using your IDF curves, because of how they were developed. If your IDF curve says your 6 hour intensity is 8.1 mm/hr, then you can multiply 8.1 mm/hr * 6 hrs = 48.6 mm of water that fell in that "10 year 6 hour" storm. So that can give you the rainfall depth you're looking for.
But you're not done yet. Two more pieces to the puzzle.
First, that's a depth, not a volume, so to figure out the volume of rainfall that falls over the watershed, multiply by the watershed area, and watch your units. You'll end up with a volume that's in mm/hectare or something (I'm American, so I get inches per acre). Then multiply by the conversion factor to turn it into cubic meters.
But that's still rainfall, not runoff. You have to figure some of it gets stuck in birdbaths and leaks into the ground and is carried away by gremlins, which is all rolled into your "runoff coefficient." All a runoff coefficient does is give you a ratio of runoff to rainfall. So for a C=0.7, that just means 70% of the rainfall turns to runoff, the rest gets carried away by gremlins.
So you multiply the rainfall volume by that coefficient and you get your total volume of runoff.
That should be enough math to size your dry well. What Peter's getting into with critical storm analysis is the sort of thing the Florida DOT makes you do, where you run a bunch of trial storm durations through a watershed model to figure out which one floods the worst, and use that as your design storm. I'm not sure that really applies for rainfall retention systems with no outflow, and if it does, I don't know how it would. Peter alludes to that above as well, by saying you'd eventually settle on an infinite duration event.
The safest thing to do is ask another engineer in the area what he does. A lot of times this stuff is just decided on by convention, and the old farts reviewing it don't even think so much about the science behind it. Alternately, just pick a storm you want to use and clearly call that out on the plans and calculations that you used a "10 year SIX HOUR storm," (or whatever) and if they approve it, then they've agreed with your duration assumption.
Hydrology, Drainage Analysis, Flood Studies, and Complex Stormwater Litigation for Atlanta and the South East -
