Background

Maritime traffic is increasingly volatile. Policy and demand shifts weaken extrapolation from the past. Governments and firms need tools that capture behaviour in volatile environments. Agent-based models can do this, but they require detailed data and explicit decision rules. TIDES originates from collaboration between the Supply Chain Intelligence Institute Austria, TU Delft, and the Complexity Science Hub Vienna. It fits routing behaviour from observed movements and simulates flows on a capacity-constrained network. A tariff case from a recent policy brief illustrates this method [1].

Model

Ships are agents with fixed identities and travel histories. The environment is a network of ports and sea lanes with service capacities and bottlenecks. Routing follows a second-order Markov process: the next port depends on the last two ports. We estimate parameters from AIS trajectories for 10,000 ships and 1,315 ports from 2019 to 2024[2, 3]. The simulation uses the same 10,000 ship agents and is event-driven. When capacity binds, queues and rerouting emerge. The second-order Markov design keeps routing numerically tractable while maintaining realistic behaviour such as service loops and transshipment. Empirical transitions reduce the need for hand-coded rules.

Interventions

We run two counterfactuals. A shock phase stops direct sailings between China and the United States. In this phase, we hold China’s total export demand fixed and reallocate former United States-bound demand to other destinations in proportion to baseline capacity shares. The rebound phase raises China to United States routing to 150% of baseline to reflect catch-up restocking and longer cycle times observed in past congestion episodes. The model is linear, so partial shocks scale proportionally.

Shock phase.

Declines concentrate in the United States and Mexico. Long Beach and Los Angeles decrease by about 17 ships per day, a 63.20% drop from baseline. Oakland decreases by about 4 per day (49.73%) and Tacoma by about 2 per day (45.94%). In China, Ningbo decreases by about 2 per day (3.36%), while most other large Chinese ports decrease only slightly. By region, arrivals change by: United States −14.8%, China including Hong Kong and Macau −2.35%, Vietnam −0.41%, Indonesia −0.28%, European Union +1.86%, South America +4.71%, Great Britain +2.26%, Turkey +1.31%, and Korea +1.23%. See Figure 1.

Rebound phase.

With China to United States routing at 150% of baseline, increases concentrate again at the same United States gateways. Long Beach and Los Angeles rise by about 19 ships per day (72.47%), Oakland by about 5 per day (60.91%), and Tacoma by about 3 per day (55.98%). At the national level, United States port activity rises by 18.87%. Offsetting reductions appear elsewhere: China including Hong Kong and Macau −0.38%, European Union −0.81%, Korea −1.43%, and Japan −1.12%. South America increases by +0.29% as it benefits from second-order effects of shared vessel rotations, where services that call at United States ports also include South American stops. See Figure 2.

Operational reading.

In the shock phase, capacity redistributes toward non-United States destinations in line with the demand rule. In the rebound phase, traffic concentrates into a few West Coast nodes, which raises the risk of larger queues and dwell times. Authorities can help during peaks by monitoring port call volumes, vessel turnaround times, cargo dwell times, and freight rate volatility, and by applying short-term measures such as diversions, higher terminal utilisation, and incentives to reduce dwell.

Limitations.

Limitations of the analysis are tied to the assumptions. The shock phase presumes a full halt of direct United States–China voyages and constant demand from other regions for Chinese exports. The rebound assumes a uniform 150% increase in flows from China to the United States. Results should therefore be interpreted not as forecasts of trade reallocation, but as indicators of which routes are likely to face over- or under-utilisation under these scenarios. Whether such shifts would materialise, and how price effects feed into this, are interesting future research avenues.