AIS Traffic Analysis for the Calibration of Risk Assessment Methods

Brian Calder and Kurt Schwehr
Center for Coastal and Ocean Mapping and NOAA/UNH Joint Hydrographic Center
Chase Ocean Engineering Lab, 24 Colovos Road, Durham NH 03820, USA.

Abstract

Assessing the risk to a vessel of transiting or anchoring in a given
area is a fundamental task for the user of hydrographic data.  A
sufficiently nuanced analysis of risk is also one way to present, to
the user, the degree of uncertainty in the data being presented,
providing a much better means to analyze and understand the
completeness, accuracy and validity of the navigational product for an
individual than current methods such as source or reliability
diagrams, or equivalents in electronic products.

Plausible methods for expressing the risk to a vessel in any given
area requires more information that can be provided from the
hydrographic databases typically held by Hydrographic Offices.  In
particular, much of the assessment of risk revolves around the
behaviors of the vessel or, if the assessment is intended to describe
the mean behavior within a geographic area (such as a harbor or
approach), the aggregate behavior of all of the traffic in the area.
Other issues such as preferred transit lanes, traffic control measures
and local climatic conditions are also important.  Unfortunately, this
vital information is typically either poorly understood or
unavailable.

We consider the potential for harvesting and aggregating Automatic
Identification System (AIS) messages within a geographic area as a
means of characterizing the traffic in the area and its behaviors.
Regular AIS messages include such information as the physical
dimensions of the ship, its draft, speed, heading with respect to
course, etc., and analysis of this data as a function of time can
illuminate more complex behaviors of the traffic such as preferred
routes, speed distributions, or simply mix of shipping types.  Through
analysis of the AIS traffic for a period of two months at three ports
(Hampton Roads, VA; Houston-Galveston, TX; and San Francisco-Oakland,
CA) we investigate the problems in verifying and translating the raw
AIS messages transmitted in practice from transiting ships, and how
these may be reduced to statistical distributions appropriate for
calibration of risk assessment models.  We consider in particular the
distribution of shipping and the presence of Class B transceivers,
which are currently being phased in within the US.  We outline the
time-series analysis methods used to reduce the data to appropriate
statistical models, and the implications this has for risk assessment
models and port management in the geographical area.