Purpose of the paper

This paper examines whether combining port-day context indicators from the IMF PortWatch with voyage-level AIS trajectory features improves binary prediction of arrival delay risk for inbound vessels approaching the Port of Los Angeles and Long Beach. The study is motivated by the operational need for early risk identification in maritime logistics, where decision-makers often benefit more from reliable triage signals than from precise, minute-level delay estimates. Instead of forecasting exact delay duration, the paper focuses on a decision-oriented risk classification task that estimates the probability that an inbound voyage will experience an unusually high delay relative to a reproducible baseline ETA.

Design/methodology/approach

The paper develops a reproducible open-data fusion pipeline that integrates heterogeneous sources with varying temporal and analytical scales. Daily NOAA MarineCadastre AIS files for March to April 2025 are downloaded, filtered to a Los Angeles and Long Beach analysis region, and transformed into inbound voyage instances using a geofence-based approach with outer entry and inner arrival definitions. A train-anchored binary delay risk label is then constructed from a simple physics-based baseline ETA at the outer entry and a high-delay threshold estimated from training data only, which reduces leakage and supports deployment-like evaluation. Voyage-level AIS features summarize early approach behavior within a bounded post-entry window and include kinematic statistics, slow-speed fractions, maneuvering variability, stop-and-go proxies, anchorage-dwell proxies, and local traffic density signals. Daily PortWatch activity and trade estimate indicators are joined by the date of outer entry to provide day-level operational context. Transparent logistic regression models are estimated in three variants, namely AIS only, PortWatch only, and fused AIS plus PortWatch, and evaluated on a strict temporal holdout split using ROC AUC, PR AUC, and Brier score, supplemented by bootstrap confidence intervals, paired bootstrap model comparisons, calibration diagnostics, threshold analysis, rolling temporal backtests, and feature ablations.

Findings

The final dataset contains 1,291 inbound voyage instances, of which 1,143 are assigned to the training set and 148 to the test set, under a strict temporal split. Delay risk prevalence increases from 25.0 percent in training to 36.5 percent in testing, indicating temporal label shift and underscoring the importance of calibration and threshold policy. During the held-out test period, the fused model achieves the strongest discrimination, with ROC AUC of 0.771 and PR AUC of 0.649, while the AIS-only model performs closely, with ROC AUC of 0.753 and PR AUC of 0.624. The PortWatch only model is substantially weaker, with an ROC AUC of 0.542 and a PR AUC of 0.394. For probabilistic accuracy on this split, AIS only and fused are both clearly stronger than PortWatch only, with AIS only showing a slightly lower Brier score than fused. Paired bootstrap comparisons indicate that fusion clearly improves over PortWatch-only, while gains over AIS-only remain modest and statistically uncertain at the present horizon. Rolling temporal backtests preserve the same ranking pattern on average, supporting the interpretation that AIS features provide the dominant predictive signal for voyage-level delay risk, while PortWatch contributes contextual but secondary information.

Value/Originality

The paper presents a reproducible, deployment-oriented methodology for maritime risk analytics that links voyage-level AIS microfeatures with port-day contextual indicators from an open, macro-style monitoring platform. Its originality lies not only in the fusion design, but also in the evaluation philosophy, which extends beyond point performance reporting to include uncertainty quantification, calibration assessment, threshold-dependent operating behavior, temporal robustness checks, and targeted PortWatch ablations. This combination makes the study useful both as an empirical result and as a methodological template for future multi-source logistics risk prediction research.

Research limitations/implications

The analysis is limited to one major port complex and a relatively short observation window from March to April 2025. The binary delay label is derived from a simplified baseline ETA rather than proprietary schedules, so the predicted outcome should be interpreted as excess travel time relative to an idealized motion baseline and not as a direct measure of terminal waiting time. In addition, PortWatch predictors are day level aggregates and may face a scale mismatch when explaining voyage specific outcomes within the same day. Future research should extend the study period, test multiple ports and corridors, incorporate additional exogenous signals such as weather, labor, or network spillover indicators, and compare more expressive models while preserving reproducibility and operational interpretability.

Practical implications

For logistics risk management and port-related monitoring workflows, the results suggest that voyage-level AIS-derived features should serve as the core of early-warning and triage systems for inbound arrival-delay risk. Port day indicators can add useful context, especially for day-level baseline pressure, but should be treated as complementary signals and monitored under temporal shift. In practice, model deployment should not rely solely on a default probability cutoff. Instead, operators should explicitly tune thresholds to workload and miss cost trade-offs, and recalibrate predicted probabilities before using them for direct risk communication, escalation, or automated alerting.

Purpose

Although smart-port technologies are becoming essential to the operations of modern seaports, organizational capabilities, regulatory pressures, environmental expectations, and growing cybersecurity concerns all influence investment decisions. The purpose of this study is to explain why port decision-makers want to invest in smart-port technologies by looking at how attitudes and intentions are influenced by regulatory pressure, environmental concerns, and IT infrastructure readiness. It also looks at how perceived security risk could reduce the impact of attitudes on intentions.

Theoretical Framework

The Theory of Planned Behavior (TPB) serves as the foundation for the model. The following are listed as external antecedents of attitude toward smart-port investment: regulatory pressure, environmental concern, and IT infrastructure readiness. It is anticipated that regulatory pressure and IT readiness may both directly and indirectly affect intention through attitude, while environmental concern will only affect intention through attitude. The proximal determinant of intention is represented as attitude, and the relationship between attitude and intention is moderated by perceived security risk.

Methodology

Port authorities, terminal operators, and logistics service providers are among the important decision-makers in Taiwanese port operations who will be provided a structured questionnaire. In order to determine the relative significance of environmental, regulatory, and technological drivers as well as to estimate direct, mediated, and moderated relationships, data will be analyzed using Partial Least Squares Structural Equation Modeling (PLS-SEM).

Expected Findings

It is anticipated that the research will demonstrate that regulatory pressure and IT infrastructure readiness have both direct and attitude-mediated effects on investment intention, whereas environmental concern primarily shapes a favorable attitude toward smart-port investment. Additionally, the positive effect of attitude on intention is expected to be significantly reduced by perceived security risk, emphasizing cybersecurity as a significant barrier even in the presence of favorable environmental and regulatory conditions.

Originality / Value

This study incorporates technological, institutional, and environmental factors into a logical behavioral framework for smart-port investment by modeling both full and partial mediation at the same time as a cybersecurity-related moderation. With an emphasis on Taiwanese ports, it provides policymakers and port managers with new ideas on how to create cybersecurity, infrastructure, and regulatory plans that efficiently facilitate smart-port digitalization.

Purpose of this paper

Ports and container terminals are critical nodes in global logistics systems, yet they are increasingly exposed to climate-related risks that threaten operational continuity, infrastructure reliability, and supply chain performance. Climate change has intensified hazards such as flooding, sea-level rise, heatwaves, extreme rainfall, and strong winds, particularly in coastal and estuarine environments. These risks directly affect terminal operations, asset lifecycles, safety conditions, and multimodal connectivity, while also generating cascading disruptions across wider logistics networks and ecosystems. Despite growing awareness of these challenges, climate resilience in port logistics is often addressed through reactive disruption management rather than integrated, data-driven operational strategies.

The purpose of this paper is to examine how climate resilience can be operationalised within container port logistics through the deployment of digital technologies, structured risk assessment, and real-time decision-support systems. The paper focuses on European ports implemented at the container terminals, which serves as a demonstrator for embedding climate risk management into day-to-day terminal operations. Specifically, the paper aims to (i) identify the main climate-related risks affecting container terminal infrastructure and operations in Ports, (ii) analyse how digital platforms, sensor networks, and operational data are integrated to support early warning, impact assessment, and adaptive response, and (iii) assess the implications of these measures for logistics resilience and supply chain continuity.

By addressing these aims, the paper responds to the need for empirical studies that move beyond conceptual discussions normative recommendations of resilience and provide practical insights into how ports can adapt to climate uncertainty. The pilot ports offer a timely and relevant case, combining ongoing terminal modernisation with climate-resilience deployment in a complex, urban, and multimodal logistics context.

Design/methodology/approach

This paper adopts a qualitative, case-based research design with a systems-oriented perspective to analyse the deployment of climate-resilience measures within the ports. The study is grounded in the SAFRAN risk assessment methodology, which provides an evidence-based approach for identifying climate hazards, assessing asset exposure and vulnerability, and analysing potential cascading failures across interconnected systems. This methodology is particularly appropriate for port logistics environments, where infrastructure, operations, and external conditions are tightly coupled.

The empirical material is drawn from the design and early-stage implementation of risks measures at the Yilport Liscont container terminal. Data sources include climate risk assessments, port infrastructure analyses, and technical descriptions of the digital architecture supporting the pilot. The analysis focuses on how different digital components are integrated, including the Capacity Calculation Module for operational forecasting, the NextPort digital platform for dashboards and decision support, and the APL and MARLO systems for governance and sensor management. These internal systems are connected to terminal operational data as well as external information sources such as weather forecasts, tidal data, traffic conditions, and AIS data.

The methodological approach emphasises the operational lifecycle of climate-related disruptions in port logistics. Use cases are structured around six sequential stages: event detection, early warning, risk and impact estimation, contingency planning, response and adaptation, and monitoring and feedback. This process-oriented perspective enables the study to capture how resilience measures support logistics decision-making before, during, and after disruptive events. The theoretical scope of the paper lies at the intersection of logistics resilience, risk management, and digitalisation in port and terminal operations.

Findings

The analysis of European ports provides several key findings regarding the operationalization of climate resilience in container port logistics. First, climate-related risks at ports terminals are highly multidimensional. Flooding, storm surge, and intense rainfall threaten quay structures, drainage systems, and yard accessibility, while heatwaves and extreme temperatures affect crane performance, electrical equipment, and energy systems. Strong winds further disrupt navigation and cargo handling activities. These hazards do not operate in isolation but interact across infrastructure, operations, and multimodal transport connections, thus increasing the likelihood of cascading disruptions.

Second, the findings highlight the central role of digital integration in enhancing resilience. The combination of real-time sensors, terminal operational data, and external environmental information significantly improves situational awareness and early risk detection. The NextPort platform and Capacity Calculation Module enable dynamic assessment of terminal capacity under different climate scenarios, enabling operators to anticipate disruptions and adjust operational plans proactively. This represents a shift from static contingency planning to adaptive, data-driven decision-making in terminal logistics.

Third, the structured use-case framework adopted in the ports supports a systematic response to climate events. Linking early warning systems to predefined contingency and response actions reduces uncertainty and response time during disruptions. However, the findings also show that resilience outcomes depend not only on technology but on organisational coordination, data interoperability, and governance arrangements among terminal operators, port authorities, and technology providers. Without these ecosystem components, the potential benefits of digital tools remain constrained.

Value

This paper provides original value by offering empirical evidence on how climate resilience can be embedded into container terminal logistics through an integrated digital and operational framework. Unlike much of the existing literature, which remains largely conceptual or policy-oriented, the study demonstrates how resilience strategies are translated into concrete operational processes at the terminal level. The ports illustrates a scalable and transferable model for combining risk assessment, real-time monitoring, and decision support in port logistics. The value of the paper extends to multiple audiences. For logistics researchers, it advances understanding of resilience as an operational capability supported by digitalisation. For practitioners, it provides a practical reference for designing and implementing climate-resilient terminal operations. For policymakers and port authorities, the findings highlight the importance of coordinated, data-driven approaches to climate adaptation in critical logistics infrastructure.

Research limitations/implications

The study has several limitations. First, it focuses on the design and early deployment stages of the ports, meaning that long-term operational performance outcomes and quantitative resilience metrics are not yet fully observable. Second, as a single case study, the findings may not be directly generalisable to all port contexts, particularly those with different governance structures or climatic conditions. Future research should include longitudinal evaluations and comparative studies across multiple ports to assess scalability, transferability, and long-term effectiveness.

Practical implications

The findings suggest that climate resilience should be integrated into core terminal management and investment decisions rather than treated as a standalone environmental concern. Port authorities and terminal operators should prioritise interoperable digital platforms, sensor-driven monitoring, and structured response workflows to enhance preparedness and recovery. The ports shows that proactive climate risk management can strengthen operational reliability, reduce disruption impacts, and improve the resilience of logistics networks often dependent on container terminal performance.