MODELING AND SIMULATING GLOBAL TRADE GEOGRAPHIES

Deep Frontiers Platform

The Deep Frontiers Platform brings together a computable representation of navigable space, route computation, network configurations, harmonized trade data, and tools for empirical comparison within a single scientific environment.

Ongoing research
This page presents the platform’s principles, capabilities, and selected results. Detailed methods for graph construction, parameterization, port allocation, and validation will be documented in forthcoming scientific publications.

Platform at a glance

Spatial & computational core

The navigable network

The platform’s spatial core represents major oceanic continuities, regional seas, straits, port access, and selected navigable inland-waterway continuities within a single network.

The graph represents navigable space itself, independently of the ports and routes computed within it. The 847 ports form a reference set of terminals used for the current global-scale computations. They are not a limit of the model: other terminals can be defined according to the territories, infrastructures, or questions being studied.

For each network configuration, Deep Frontiers can compute the complete set of directed routes between the selected terminals.

NAVIGABLE SPACE

A continuous representation of accessible space across multiple scales.

CONTROLLED TOPOLOGY

The mesh preserves coastlines, islands, archipelagos, and narrow passages; no edge crosses land.

INTEGRATED CONSTRAINTS

The network incorporates bathymetric information, depth constraints, and regulatory navigation areas.

24Mnodes

1.4B+edges

847reference ports

716,562computed routesall directed relations among the 847 reference ports

The graph is computational. Routes are outputs.

For each network configuration, under a specific set of constraints and cost functions, Deep Frontiers computes an optimized route between two ports. Routes are not stored as fixed geometries: they are generated from the port-to-port relation being examined and the configuration applied to the network.

Network configurationPort-to-port relationComputed routeDeep Frontiers therefore provides a simulation space: changing the properties of the graph means constructing a new possible geography of maritime circulation.

Maritime connectivity across the Gulf Caribbean Panama system — **Gulf–Caribbean–Panama system**Gulf–Caribbean–Panama system — computed maritime connectivity at regional scale.

Route concentration around the Strait of Malacca — **Strait of Malacca and Southeast Asia**Strait of Malacca and Southeast Asia — route concentration and navigable interfaces.

Configurations & experiments

The platform can simulate and compare multiple network configurations. Accessibility, constraints, and cost functions can be modified independently or combined within the same experiment.

ACCESSIBILITY

Open, close, or change access through an area or passage.

CONSTRAINTS

Incorporate physical, environmental, or regulatory conditions.

COST FUNCTIONS

Control route costs globally or locally: weight criteria, favor or penalize particular route geometries or transitions, and incorporate spatially varying costs.

REFERENCE STATE

The graph can adopt different reference states depending on the date or period considered. These states may be seasonal, monthly, or daily.

SCENARIOS

Alternative configurations designed to modify one or more network parameters and analyze the resulting flow geographies.

Arctic — September sea-ice probability along a computed routeArctic — example of a configuration based on the climatological probability of September sea-ice presence, estimated from more than 15 years of daily observations. The computed route is evaluated along its path against the field used to configure this network state.Sea-ice observations: U.S. National Ice Center (USNIC).
Climatological processing and route computation: Deep Frontiers.

Network accessibility can expand or contract.

Environmental information can therefore become a property of the computational space on which routes are generated.

Global trade & maritime flows

Deep Frontiers connects its computational network to international merchandise trade. Harmonized trade data make it possible to analyze commodity flows across countries, periods, and network configurations. With historical coverage extending back to 1964, the platform supports both long-run analysis and counterfactual simulation.

This makes it possible to examine the geography of flows, their evolution over time, and how their exposure and spatial organization change across network configurations.

TRADE DATA

Deep Frontiers integrates international merchandise trade data from UN Comtrade. Successive SITC revisions are reconciled to make commodity categories comparable over time. This provides a consistent long-run framework for analyzing changes in global trade since 1964.

HISTORICAL DEPTH

More than six decades of harmonized trade data provide a consistent basis for analyzing long-term changes in global trade.

FROM OBSERVATION TO SIMULATION

Trade relationships and quantities can be held constant while the network configuration changes. Routes are then recomputed and trade is remapped onto the new network configuration. This makes it possible to isolate and measure the modeled effects of changes in accessibility, constraints, or cost functions on the geography of global trade.

COMMODITY CATEGORIES

Trade can be analyzed at different levels of commodity aggregation, from broad sectors to more specific categories.

MARITIME GATEWAYS

Country-level trade is connected to the maritime network through a system of gateways. These gateways provide the spatial interface between reported international trade and port-to-port relations.

FLOW ANALYSIS

Once trade has been mapped onto the network, Deep Frontiers can quantify route exposure, detours, additional distance, transport work (tonne-miles), chokepoint dependence, and the concentration of impacts across trade relations.

Explore the case study →

Observation & computation

AIS density provides an independent empirical reference for examining the spatial consistency of computed routes with major observed maritime corridors.

Map of the Mediterranean Basin overlaying AIS shipping density and Deep Frontiers computed routes. — **Mediterranean Basin — computed routes and AIS density**The displayed links connect successive port calls identified in published liner-service rotations. Their geometry is entirely recomputed on the Deep Frontiers graph.

AIS DENSITY

AIS observations reveal the major corridors actually used by vessels.

PUBLISHED SERVICE ROTATIONS

Some comparisons use relations between successive port calls in published liner-service rotations.

SPATIAL CONSISTENCY CHECKS

Computed routes are compared with observed corridors and geographically constrained passages.

High-performance computing

Since 2025, the construction and refinement of the Deep Frontiers global graph have relied on sustained, iterative parallel-computation campaigns on local resources. The resulting graph versions are then deployed at scale on the Jean Zay supercomputer (GENCI/IDRIS), where parallel computing produces port-to-port route matrices and performs the recomputations required for experiments.

SEVERAL THOUSAND HOURSof local parallel computation since 2025

Jean Zay — GENCI/IDRIS500,000 CPU hours allocated