Measuring offshore tsunami currents using ship navigation records

Great earthquakes and tsunamis have caused significant loss of life and property in coastal communities (ITIC 2015). In the 2004 Sumatra-Andaman, Indonesia, earthquake and the 2011 Tohoku, Japan, earthquake, most casualties (200,000 for the 2004 disaster and 20,000 for the 2011 disaster) were caused by the flooding due to the tsunamis (Satake 2014). Coastal mega cities have to prepare to reduce these losses especially from the tsunamis. In addition to constructing proper embankments in advance of earthquake and tsunami occurrences (Tomita et al. 2012; Suppasri et al. 2013; Raby et al. 2015), immediate actions of individuals and groups after the earthquake/tsunami occurrences are critical for reducing significant, early-stage loss of life (Mas et al. 2013; Makinoshima et al. 2016).

Real-time forecasting of great earthquakes and tsunamis can be utilized for suitable decision-making for disaster mitigation. Real-time forecast systems should be robust and reliable. Current systems are based on real-time Earth monitoring that uses onshore and offshore observations. These observations include seismic waves (Allen et al. 2009; Kamigaichi 2009), crustal deformation (Sagiya 2004; Ruhl et al. 2017), and offshore sea level or the tsunami itself (Falck et al. 2010; Kawai et al. 2013; Rabinovich and Eblé 2015).

Rapid earthquake-source estimation systems using seismic-wave observations are popular. However, these methods estimate a point source that may be indirectly related to tsunamis, and are somehow difficult to accurately constrain the tsunami source dimension particularly for great earthquakes (Katsumata et al. 2013; Hoshiba and Ozaki 2014). Earthquake-source estimation using Global Positioning System/Global Navigation Satellite System (GPS/GNSS) observations of land deformation will suitably constrain the source dimension with moment magnitude of nearshore (< 100 km from the coast) earthquakes (Melgar and Bock 2015; Kawamoto et al. 2017). For greater earthquakes (M_w > 8.5), in addition to the fault slip, seafloor failure due to strong ground motion likely causes a certain portion of tsunami (Kawamura et al. 2014; Løvholt et al. 2015). It is better to use direct observations of offshore sea level to estimate the resultant tsunami size.

Real-time forecasting of great tsunamis based on offshore tsunami observations has been popular. The tsunami forecast system using the Deep-ocean Assessment and Reporting of Tsunamis (DART) buoys is well known (Wei et al. 2013; Tang et al. 2016). The Japan Meteorological Agency (JMA) developed a tsunami forecast system using cabled seafloor pressure and GPS-buoy observations (Tsushima et al. 2011, 2012). These forecast systems will be enhanced using real-time geodetic monitoring as well (Tsushima et al. 2014; Wei et al. 2014). Forecast methods using very dense seafloor observation networks have been also proposed independently (Maeda et al. 2015; Igarashi et al. 2016; Yamamoto et al. 2016). These forecast systems using offshore tsunami observations will be robust and reliable.

The observatories have to be maintained and sustained at least until the next devastating tsunami. Such an event may occur once typically during tens to hundreds of years at a certain place (McCaffrey 2008; Rong et al. 2014). In general, offshore observatories require higher maintenance costs than onshore observatories, and replacements will be necessary after endurance periods, typically several decades after installation (Hirata et al. 2009). Tsunami forecast methods especially measurement systems need sustainability as well as reliability (NRC 2011; Bernard and Titov 2015). It is desirable to propose and develop other sustainable methods of monitoring and forecasting of great tsunamis.

Tsunami as sea-level change could be detected by shipborne GPS. A 10 cm tsunami caused by the 2010 Maule, Chile, earthquake was detected by a vertical position change derived from high-precision GPS of a navigating ship (Foster et al. 2012). This indicates that navigating ship networks will serve as offshore tsunami observation networks (Inazu et al. 2016). However, this requires high-precision GPS facilities on most of the vessels to precisely measure their positions as well as GPS buoys (Terada et al. 2015). On the other hand, ship positions are currently aggregated as the Automatic Identification System (AIS) data under International Maritime Organization (IMO) regulations (IMO 2002; Earles et al. 2010). The ship positions in the AIS data do not include vertical position, and they are mostly measured by ordinary GPS with meter-order precision which is employed in ordinary car navigation. Therefore, if we employ ship GPS/AIS data to measure tsunami height, it is necessary to update current facilities so as to estimate the high-precision locations of almost all of the vessels and modify the regulations of the IMO.

During the 2011 Tohoku tsunami, a number of vessels in ports drifted significantly and collisions occurred due to strong tsunami currents (Suga et al. 2013; Suppasri et al. 2014; Matsuda and Tomita 2015). The behaviors of the ship horizontal drifts with collision in a port were examined using AIS data with tsunami simulation (Matsuda et al. 2012). The ship drifts should be affected by tsunami currents; however, the relationship between them still seems unclear due to a complicated situation in the port. On the other hand, AIS data were also available in offshore regions during the tsunami. It was reported that movement directions of several offshore ships deviated during the tsunami (Makino 2013; Liu et al. 2015). However, these reports merely showed the facts and hardly described any quantitative relationship between the deviations and the tsunami. There is still much room to be clarified regarding the expected relationship between ship drifts and tsunamis.

In the present study, we examine the AIS records of multiple ships navigating near the source region of the 2011 Tohoku tsunami. The relationship is investigated quantitatively between the ship horizontal movement or drift and the tsunami current. We describe the usefulness of the AIS data to measure offshore tsunami currents and to be applied to tsunami forecast.

We can browse almost real-time ship distributions derived from AIS data on web sites such as MarineTraffic (https://www.marinetraffic.com/). According to the IMO regulations, ships exceeding 300 gross tonnage and all passenger ships have to send their AIS information via very-high-frequency (VHF) radio transmission (Tetley and Calcutt 2001; IMO 2002). The AIS data include dynamic and static information. The dynamic information includes latitude/longitude (without vertical height), speed over ground (SOG), course over ground (COG), and ship heading (HDG). The static information includes draft, Maritime Mobile Service Identity (MMSI), ship type, and ship name. The AIS data from nearshore vessels (< ~ 100 km from coast) are received by coastal stations, and those from ships farther offshore can be also received by low-Earth-orbit satellites such as ORBCOMM (Carson-Jackson 2012). The number of ships recognized with AIS data is increasing year by year with increasing seaborne trade and number of satellites (Tournadre 2014; Willick 2014).

Once the tsunami velocity can be measured by the ship velocity component in the direction normal to the ship heading, it is worth carrying out inversion for real-time tsunami forecast to estimate source and tsunami height. Fuji et al. (2013) examined a tsunami inversion/forecast experiment using pseudo observation of tsunami current derived from coastal ocean radars (Hinata et al. 2011; Lipa et al. 2012; Benjamin et al. 2016). They employed the tsunami current instead of tsunami height for the inversion for the tsunami source. In the present study, we employ tsunami currents derived from the observational ship drifting (Fig. 5). Green’s functions are composed by tsunami current components in the HDG-normal directions of the respective moving vessel by calculating from unit sources. Referring to our previous study (Inazu et al. 2016), we allocate unit sources with the same size/spacing and carry out the inversion/forecast.

The inversion/forecast is carried out using the AIS data for 25 min after the earthquake occurrence (Fig. 7). The result can show mostly similar features in terms of both the source and the maximum tsunami height at the coast, compared to those from the validated source (Fig. 1). This indicates that ship velocity components in the HDG-normal direction of multiple ships derived from AIS records will be possibly useful for tsunami source estimation and forecast in addition to conventional tsunami height observations at fixed stations.

The AIS records of offshore navigating vessels were examined during the 2011 Tohoku tsunami. The AIS records from 13 vessels near the tsunami source showed significant deviation of COG from HDG during the tsunami passage.

We compared the observed ship velocity component to the simulated tsunami velocity in the direction normal to the ship heading and confirmed good agreement in terms of amplitude (~ 1 m/s) with short lag times (< ~ 2 min) between them. Using an equation of motion of an offshore movable floating body, we obtained an analytical solution and understood that the ships immediately respond to the great tsunami current with the equivalent velocity amplitude in the HDG-normal direction.

This study indicates AIS data in the current framework can be used as a tsunami current meter in one direction for each vessel. We also showed that the observed tsunami velocity derived from the AIS data of multiple vessels during the 2011 Tohoku tsunami could be useful for tsunami source estimation and forecast.

The current AIS data are already provided via almost real-time successive transmission. The number of ships recognized by the AIS has been increasing year by year (UNCTAD 2017). We hope to exploit this large amount of data and expect the use of AIS data will become a new method for measuring and forecasting great tsunamis. Similar concepts of diverting information from crowd-sourced agents to Earth monitoring recently have been also applied to grasp earthquakes (Kong et al. 2016), winds, and ocean currents (Miyazawa et al. 2015).

Suitable methods for forecast will need to be developed. Assimilation/inversion of tsunami wave/current fields using moving point observations will be a possible way (Fujii and Satake 2007; Mulia et al. 2017). Such methods will be adopted also for assimilating wave height fields derived from satellite altimetry (Hamlington et al. 2012) and GNSS reflectometry (Stosius et al. 2011; Clarizia et al. 2016).

Efforts to ensure a good quality of AIS data should be made so that sophisticated forecast methods will work well (Table 1 and Additional file 1: Figure S1). The minimum requirement is to obtain the right records of the HDG and the GNSS ship position. SOG and COG are derived from GNSS. Ship officers can make adjustments so that HDG and COG are consistent and there is no substantial bias between them during offshore autopilot navigation in calm weather/ocean conditions. As long as based on Eq. (1), tsunami current evaluation may be difficult when SOG is very slow. We may easily pick up and omit very-slow-SOG vessels (< ~ 0.5 m/s) when a great earthquake occurs. Further discussion will be necessary for automatically obtaining high-quality AIS data in real-time operation.

Although Japan already has dense offshore observatories for real-time tsunami monitoring and forecasting (Kawai et al. 2013; Kaneda et al. 2015; Yamamoto et al. 2016), most countries, in particular developing countries, will not be able to install and maintain similar dense offshore observatories mainly for economic reasons. Hopefully, the use of the AIS data will play a role in new tsunami monitoring/forecasting methods especially for developing countries such as those in Southeast Asia and South America. The density of vessels increases in coastal mega cities with highly economic activities. Tsunami disaster mitigation is especially important for such coastal mega cities due to their large populations. Maritime activities will continue as long as populations remain large. The use of the AIS data for tsunami current monitoring may be a sustainable way of tsunami disaster mitigation.