Modelling container transhipment throughput and analysing dynamics during significant external events: A case study of the Port of Busan and the China-to-U.S. trade route

Abstract

This study examines the crucial role of transhipment hub ports (THPs) in global container maritime logistics (GCML), highlighting the significance of effective and resilient THP operations. While accurate forecasting of container throughput is vital for efficient port management, predicting transhipment volumes remains particularly challenging due to their inherent complexity, external sensitivity, and dynamic behaviour. Consequently, this domain has experienced notable research gaps, especially under conditions of geopolitical and pandemic-related disruptions.
These gaps were revealed by the profound impacts of the U.S.-China Trade War (UCTW) and the COVID-19 pandemic, which destabilised established logistics practices, heightened uncertainties within GCML, and severely disrupted maritime trade flows. In response, this study pioneers a modelling framework for transhipment volumes at a specific THP — the Port of Busan — focusing on volumes from the China-to-U.S. trade route (TVCU), thereby addressing the need for robust and interpretable models.
This research is structured around three core objectives: (1) to investigate the evolution and functional characteristics of THPs within the GCML system, (2) to develop predictive and descriptive models for transhipment volumes based on industry-specific variables, and (3) to analyse dynamic behaviour of transhipment volumes through integrated literature review and data-driven modelling. To achieve these objectives, three modelling methods are employed: Multiple Regression Analysis (MRA), ARIMAX, and Explainable Boosting Machine (EBM). These methods are used to address research gaps and comprehensively examine TVCU dynamics.
The models are built using monthly data spanning 89 months (January 2016 – May 2023), incorporating critical industry-specific variables such as U.S. imports from China, container ship voyage duration, and metrics on shipping capacity and port congestion. While MRA and ARIMAX capture linear relationships and offer intuitive insights, EBM demonstrates superior predictive performance, achieving a mean absolute percentage error (MAPE) of 3.1% and an R-squared of 0.946 during the model fitting period. The validation period (May 2023–June 2024) yields a MAPE of 7.7% and an R-squared of 0.670. The drop in validation accuracy is primarily attributed to unforeseen geopolitical disruptions — particularly the Houthi rebel attacks near the Red Sea — which ultimately caused container rerouting from Chinese ports to Busan, resulting in TVCU spikes.
The findings from the models’ interpretation underscore the influence of external disruptions — notably the UCTW and the pandemic — in shaping transhipment volumes at a THP. EBM’s interaction analysis reveals substantially amplified pairwise variable effects on TVCU, with increases of up to 42% during the UCTW and 104% during the pandemic, highlighting the volatility and sensitivity of THPs to global disruptions. EBM’s transparent, glass-box modelling capability further enables stakeholders — including governments, port authorities, and terminal operators — to understand underlying transhipment dynamics and adopt more proactive, data-informed strategies for THP management.
In conclusion, this study recommends extending the modelling approach to other THPs and trade routes, while incorporating emerging factors such as port automation and Arctic maritime corridors. The findings contribute both theoretical and practical insights toward enhancing THP resilience, improving resource allocation, and strengthening uncertainty management within GCML framework.

More information

Identifiers

ORCID for John Preston: orcid.org/0000-0002-6866-049X

ORCID for Ioannis Kaparias: orcid.org/0000-0002-8857-1865

Catalogue record

Export record