How you model this depends on how you expect the system to behave. A common approach is to model the pipe and stone as providing storage for a single underground "pond". This assumes that the perforations are sufficient to allow the water level to equalize between the pipe and the stone, and that the final discharge is determined by some other "outlet" control, such as exfiltration into the surrounding soil, or perhaps an overflow pipe.
For details on modeling this scenario see
Here is the basic pond setup in brief:
Storage volume#1 = outer excavation with stone backfill and appropriate voids setting.
Storage volume#2 = horizontal cylinder (the perforated pipe) of appropriate length and/or quantity. Set the "Embedding" option to "Volume#1". This places the pipe inside the stone bed.
Now save the node and open the report window. Scroll down to the storage table, which will shown the detailed calculation for storage contributed by the stone and pipe.
If you want to perform a complete hydrograph analysis, you can add the appropriate outlet device, such as exfiltration and perhaps an overflow device. For a conservative design, you may want to include the overflow device only, since the actual exfiltration can often be very hard to predict. For details see
This discussion assumes that the pipe perforations are sufficient to create a "level pool", so that the actual perforations do not appear in the model. If the perforations are a significant restriction, such that the levels are different, the model becomes more complex. This may require the use of two coupled ponds: The first pond being the pipe volume with orifice outlets, and the second pond being the surrounding stone with exfiltration and/or overflow outlet. This will also require a tailwater-sensitive routing procedure, since the rising water level in the stone will constitute a "tailwater" for the orifice flow through the pipe perforations.
But remember, the two-pond approach is recommended ONLY if the perforations are small in relation to the inflow, and the "level pool" assumption does not hold. Otherwise, model it all as a single pond - MUCH easier.