Blue Hub Research Areas
Maritime Data, Data Fusion & Tracking
Vessel positioning data and static information
Information on the spatial distribution and identity of vessels at sea can be grouped into the following categories:
- Self-reporting positioning data, often referred to as “cooperative” or simply "reporting" data, are transmitted by the vessel in the vessel proximity (Automatic Identification System – AIS – for collision avoidance) or to competent authorities (Long Range Identification and Tracking – LRIT – for security and safety, Vessel Monitoring System – VMS – for fisheries monitoring). The technical specifications of such systems are regulated at national or international level and the relevant tracking capabilities are characterised by refresh rate, quantity and types of vessels under regulation (vessel coverage) and, sometimes, by spatial coverage and transmission delay. The information content of self-reporting positioning data, besides the state vector and other kinematic information, may also include voyage related data and characteristics of the ship.
- Observation-based positioning data are collected by active or passive sensors providing detection capabilities that vary depending on specific parameters (e.g. resolution, spatial coverage, update rate, latency), conditions (e.g. meteorological or physical oceonographic data) and target vessel properties (e.g. size and orientation). Observation-based data include space-based (Earth Observation - EO) Synthetic Aperture Radar (SAR), EO - Optical Images, data from coastal/mobile ground based radars or mounted on Maritime Patrol Aircrafts (MPA) or Remotely Piloted Aircrafts (RPAs). Additional sources of information for wide area surveillance that are topics of research are also HF radars, airborne systems, passive radar applications and radiolocation of emissions.
- Information Registries and Databases: vessel registries and databases contain information linking a ship identity to details about its structure, construction, appearance, history, management, safety/security inspections, etc. This type of information is static, meaning that it can be thought of as not changing or slowly changing in time with respect to the vessel current position and movement. Vessel registries and database therefore provide complementary data to positioning information. One good example is results of Port State Control inspections that point to vessel deficiencies. Such information inside registries, can be used for better maritime situational awareness.
There is no ideal sensor or technology that does not present performance limitations in any of the indicators and that could be used as a unique source of information for all surveillance applications in the maritime domain. Moreover, there are specific applications that cannot be properly addressed even by using all realistically available data. This is the case, for instance, for irregular migration, where small boats need to be detected and persistently tracked in the open seas, calling for innovative sensors and technologies. Nevertheless, it is clear that through integration and fusion it is possible to overcome the limitations of single self-reporting and observation-based systems. Different sensors and technologies present by their very nature different sampling time, data latency (i.e. the time lag between the data collection and the time thedata is made available), errors and uncertainties.
The fusion of the data is a necessary step in order to understand the actual position of vessels at sea. Increasing the number of sensors and technologies used, together with the relevant performance and limitations information, results in a more complete picture of what is happening at sea. Storing time series of pictures over time allows for the identification of consistent behaviours and paths that are followed collectively by vessels. Such data-driven knowledge, which represents the basis for many applications in a wide spectrum of domains, can be extracted everywhere in the world and is only limited by data access capabilities.
Knowledge Discovery in Databases refers to:
The first level of knowledge discovery in vessel movement data is represented by the visualisation of historical data. The video below shows the main traffic routes followed at global scale by ships flying the flag of States contributing to the EU Long Range Identification and Tracking (LRIT) Cooperative Data Centre (CDC): all EU Member States, Iceland, Norway, and Overseas Territories of EU Member States.
The video covers a period of one month, and shows the potential of the LRIT data for performing statistical and other analysis of maritime transport routes of vessels flying EU LRIT CDC flags. The dataset provides also first insights into maritime traffic: the knowledge of maritime traffic followed by merchant ships can be extracted by plotting density maps. It is worth noting that high vessel densities correspond to highly congested areas, often regulated though specific ship routing systems that separate the traffic to reduce the risk of collisions or grounding.
By using data mining and other track processing techniques, it is possible to decompose maritime traffic densities into a set of routes identified by properties such as distribution of speed along the route, travel time, shiptype, size, draught, etc. This is the basis for detecting low likelihood behaviours (anomalies) or predicting vessel positions. Once a vessel is seen to be following one of the individual known routes, it is possible to measure the level of alignment to the route in real time and understand where the vessel is likely to be in the future.
The detection of anomalies in maritime traffic can be based on specific rules (e.g. maximum allowed speed or polygon intersection). This method assumes the knowledge of the behaviours that need to be detected. An alternative approach is to acquire the “normality” of vessel behaviours and patterns so that any deviation from such “normality” can be measured and highlighted. The resulting behaviours would be flagged as “unexpected”, with the results brought automatically to the attention of operational authorities.
The Automatic Identification System, originally designed for collision avoidance, is becoming a cornerstone of maritime situational awareness. The recent increase of terrestrial networks and satellite constellations of receivers is providing global tracking data that enable a wide spectrum of applications beyond collision avoidance. Nevertheless, AIS suffers from the lack of security measures that makes it prone to manipulations such as receiving positions that are unintentionally incorrect, jammed or deliberately falsified. AIS Radiolocation is an effective method to validate the AIS data: after collecting timestamped data provided by a network of stations a Time-Difference-of-Arrival algorithm is applied to locate the emissions with an accuracy that ranges from few hundreds of meters, to a few kilometers in the worst case.
By adding a filtering stage, it is possible to narrow down the estimated position of the ship to a few hundreds of meters in a few steps (without using in any way the position reported by the ship). The radiolocation methodology developed can be applied to legacy AIS base stations and does not require the installation of new hardware or sensors.
Predictive analysis of vessel positions can be performed using many different approaches, ranging from a simple linear model (usually valid for short propagation times of the order of minutes or in the absence of ship routing systems) to more complex context-based methodologies. The latter makes use of specific knowledge that is either coded in IMO or National Authorities guidelines, criteria and regulations (e.g. the traffic separation scheme in the North Adriatic sea) or extracted from historical tracking data.
Data driven approaches are based on the assumption that vessels are mostly compliant with such regulations so that by observing maritime traffic for a sufficient amount of time, one can extract the main routing elements. The propagation using historically-derived patterns have proven to outperform linear models.
Moreover, Bayesian methods for predicting vessel motion patterns (e.g. Knowledge based Particle Filter) further increase the quality of prediction (see Mazzarella F., Fenandez Arguedas V., Vespe M., ‘Knowledge-Based Vessel Position Prediction using Historical AIS Data’, 10th Workshop Sensor Data Fusion: Trends, Solutions, and Applications, Bonn, 2015).
Mapping Activities at Sea
The behavioural characterisation of vessel activities enables the understanding of collective maritime uses. For example, by clustering fishing activities it is possible to map fishing grounds and fishing efforts. Through a link to fishing registries such as the Community Fishing Fleet Register, it is also possible to isolate behaviours related to different fishing categories such as trawlers, purse seiners, etc. (ISSCFG). In particular, the speed histogram of trawlers shows three peaks: one around 0 knots, when the vessel is at rest; one around 3.5 knots, interpreted as engaged in fishing; and one around 10.5 knots, when the vessel is steaming.
When the track of the fishing vessel is plotted on a map and colour-coded by these speed intervals, it is verified that the transit legs contain the high speed points, whereas the low speed points around 3.5 knots cluster in the middle of the sea at the far ranges of the track (red), and are therefore likely to be the fishing grounds. By applying this approach to all fishing vessels in a specific area, it is possible to understand the fishing grounds and map the high intensity fishing areas.
A similar approach can be followed to isolate exploration, transport, offshore platform activity (by mapping tug or service/supply vessels) and other activities at sea. Mapping activities at sea is valuable information to improve Maritime Situational Awareness (MSA) as it provides a-priori context to potential unidentified targets. For example, if a vessel is detected to be located in an area where the primary activity is known to be fishing, the likelihood that the detected target is actually fishing increases. Moreover, the use of behavioural analysis can help in identifying the vessel type based only on its dynamics. This also further enhances maritime situational awareness enabling the verification of the declared ship type.
Policy Areas and Applications
Maritime data fusion & tracking, automatic anomaly detection and situation prediction are key to operational authorities in order to improve Maritime Situational Awareness and therefore enhance maritime safety and security, including border surveillance, irregular migration, and countering illegal activities at sea (Irregular Unregulated Unreported fishing, smuggling, pollution, etc.).
The mapping of activities at sea, including shipping, tourism, exploration, fishing, etc. and such monitoring and analysis over longer time scales is a fundamental tool for many policy areas. Such information is often scattered or difficult to access, calling for alternative methods to extract it. Historical vessel tracking data mining has recently proved a key tool for this purpose. As an example, data driven knowledge of maritime uses is very relevant to planning and optimising the use of maritime space for human activities at sea in a sustainable way (see Maritime Spatial Planning).
This applies not only to Europe, but also to remote unexploited areas such as the Arctic, where the receding ice as a consequence of climate change is opening new economic opportunities. In such areas, research is needed to promote a sustainable development of the area and better understand trends and volume of activities to plan infrastructures and services (click here for a full animation over two years). This is reflected by the "safe and secure maritime activities" policy response to the "sustainable development" priority area of the recent Joint Communication to the European Parliament and the Council An integrated European Union policy for the Arctic.
The analysis of transport routes was recently demonstrated as a valuable tool to understand the impact on transport of international policy initiatives to counter geopolitical issues. For example, due to countermeasures such as EU Naval Force Atalanta, the NATO Ocean Shield and industry applying Best Management Practices (BMPs), the presence of armed guards on board and international cooperation programmes, piracy incidents started to decline. Through the analysis of historical LRIT data it was possible to provide quantitative evidence for the decline of piracy.
Finally, shipping is a large and growing source of greenhouse gas emissions (GHG), and the EU is looking for a global approach to reduce them. It is harder to estimate the spatial distribution of reported NOx emissions at sea compared to land, as the sea offers less accurate proxies. By using vessel density maps, more accurate high resolution maps of emissions can be achieved in support of Integrated Sustainability Assessment/Modelling and Air quality models.