Where does smoothing stop working?

Scheduling spreads the same trucks across the day. A matched return load removes a truck visit altogether: the import is collected by a truck already inside the port instead of a second truck driving in empty. That is why it is the only input here that can pull an over-capacity day back under the line without adding lanes.

A deterministic fluid-queue model at 15-minute resolution: trucks/day follow a modeled daily traffic shape (low overnight, two daytime peaks); within each hour they either cluster at :00 ("Today") or spread evenly ("LCB ONE"). Capacity = lanes × throughput/lane, held constant through the day. Return loads: matches/day = trucks/day × empty-leaving share × match rate; each match removes one gate visit and two road trips, and the gate service time of a combined drop-and-pickup is assumed equal to a single visit. This is an illustrative planning model, built to make the concept and the breaking point adjustable in front of stakeholders — it is not a calibrated traffic study. Before committing budget to lane-widening or gate counts, validate against a real microsimulation (e.g. SUMO or VISSIM) using PAT's actual road geometry, measured gate service times, and observed arrival data.