AIS-based estimation of maritime fuel consumption and emissions: methods, validation and practical applicability

Abstract

The Automatic Identification System (AIS) has become a fundamental source for reconstructing maritime activity and developing fuel consumption and emission estimates with high spatial and temporal resolution. However, AIS does not directly provide power, fuel flow, or emissions, and therefore the results depend on the incorporation of technical characteristics, operational parameters, meteorological information, or actual fuel consumption records. This review analyses 24 open-access studies published between 2021 and 2026 that effectively use AIS data at some stage of the estimation procedure. The studies are classified according to their methodological approach and application context, including port, regional, and national inventories, ship- or voyage-level fuel consumption models, assessment of operational measures, temporal prediction, and comparison of methodologies. The difficulties associated with trajectory preprocessing, the availability of complementary data, model transferability, operational deployment, and the strategies used to validate the results are also examined. \textit{Bottom-up} energy-based approaches are the most widely used and allow large fleets to be processed using average parameters, although they provide limited vessel-specific technical characterisation. Physics-based and data-driven models can represent the behaviour of individual ships in greater detail, but they require technical, meteorological, and operational information that is often private or commercial. Validation evidence is heterogeneous: numerous studies verify AIS activity or compare their results with other models, whereas validation using operational fuel consumption is less common and direct emission measurements are exceptional. Therefore, there is no universally superior method. The selection should be made according to the objective, scale, available data, and required accuracy, while always distinguishing between activity resolution and the physical accuracy of the estimates.