Spatial risk mapping of maritime accidents in the Baltic Sea: a Bayesian hierarchical approach using AIS and environmental data

Welcome

Welcome to the online supplementary repository for my manuscript. This interactive research compendium has been created to ensure full transparency, reproducibility, and methodological rigor in accordance with open science practices.

Here, you will find the complete analytical workflow, including R code, model specifications, and extended Bayesian diagnostics that could not be included in the main text due to formatting and length constraints.

Abstract

This study presents a data-driven Bayesian hierarchical framework to model maritime accident risk in the Baltic Sea. By integrating accident records, AIS traffic density, and sea surface temperature (SST) into a zero-inflated negative binomial (ZINB) model, I explicitly treat traffic as an exposure offset. Results indicate that while traffic density primarily drives accident frequency, SST lacks a credible independent effect, suggesting the resilience of modern fleets to thermal conditions. However, a substantial increasing trend in baseline risk (approx. 18.5 percent per year) was identified, alongside distinct high-risk hotspots in the Danish Straits and the South-Western Baltic. The findings demonstrate that ignoring the exposure offset leads to confounding bias. The proposed framework effectively isolates residual spatial risk, providing a robust tool for marine spatial planning and dynamic resource allocation.

Reproducibility statement

All models were estimated using Hamiltonian Monte Carlo (HMC) sampling via the Stan probabilistic programming language, interfaced through the brms package in R.

Note to Reviewers: The R code used to generate the models and plots is embedded within this document. You can view the specific code chunks by clicking the “Show the code” buttons scattered throughout the pages.