The impact of vessel speed reduction on port accidents
Introduction
Vessel speed reduction (VSR) programs have been implemented across the world to control the emission of pollutants. The rationale of the program is that reducing the vessel speed is expected to reduce fuel consumption, thereby decreasing emissions. For example, in 2001 the Port of Los Angeles designated reduced speed zones (RSZs) 20 nautical miles (nm) from the port to control ship emissions. Since then, more ports have participated in similar initiatives, including the Port of San Diego (POSD) and the Port of New York and New Jersey (PONY). On the other hand, VSR programs have been pursued in some areas to preserve endangered marine species. For instance, (Hazel et al., 2007) and (Vanderlaan and Taggart, 2007) provided empirical support that reduced vessel speed could reduce collision risks for mammals. (Hazel et al., 2007) demonstrated that turtles can flee twice the distance when the vessel speed decreases from 19 km/h to 4 km/h. An estimation by (Vanderlaan and Taggart, 2007) suggested that the probability of lethal injuries by ship collision for whales increased from 0.21 to 0.79 as vessel speed changes from 8.6 to 15 knots. Along this line, this paper is motivated by the hypothesis that if reduced vessel speed prevented accidents involving mammals, then it could also reduce vessel accidents. This is because, with lower vessel speed, navigators can enhance awareness and buy their time in responding to approaching ships, preventing precarious driving and reducing collision risks.
While some studies and our intuition tentatively suggest that vessel speed can affect accident rate, no studies formally tested whether there exists a significant relationship between the two. The most relevant studies on this issue can be divided into two groups. The first group consists of studies that adopted engineering approaches to investigate vessel speed and accidents. They developed simulation or programming models to describe the environment at a certain port and computes collision possibilities according to vessel speed. However, the studies investigated only a specific port at a time with many assumptions imposed in modeling. Moreover, they did not measure actual effects of vessel speed on accidents using the real accident data. The other group of the studies used econometric models to estimate casualty rates or damage costs of given vessel accidents. They mainly identified determinants that intensify accident damage, i.e. vessel and accident type. Nevertheless, they did not take vessel speed into consideration, and they only conducted ex-post analysis, i.e. they only considered accidents that already occurred. This way, it was difficult to examine accidents that have been kept from happening in certain environments.
To fill the gap in the literature, this paper examines the effects of vessel speed on damages, casualty and frequency of vessel accidents. Specifically, this paper investigates how the vessel speed affected 1) the accident damage and casualty level of given accidents and 2) accident frequency in a port level. For the latter, this study was the first to conduct that kind of analysis to the authors’ knowledge. U.S. Coast Guard (USCG) data were collected for the analysis because the U.S. was one of the most active implementers of VSR, and the data set contained a large number of accident profiles. Estimating the effect of the vessel speed was challenging because speed was not recorded in each accident profile. However, this issue was addressed by comparing accidents in RSZ ports with those in non-RSZ ports. With all other factors being equal, differences in accidents between RSZ and non-RSZ ports were found to be attributable to differences in the vessel speed. Fig. 1 shows locations of vessel accidents in USCG data. Vessel accidents occurred mainly in coastal waters and inland waterways, where the vessel collision risk is high. The analysis was confined to coastal ports since it allowed us to observe clear distinction of vessel accidents between RSZ and non-RSZ ports. The analysis necessitated to separate RSZ accidents from non-RSZ ones and to generate accident frequency data at the port level from USCG data. Although the ArcGIS program was used to collect detailed data at this level, this task required substantial work. The detailed procedure for extracting necessary information from USCG data is described later.
To estimate the effects, three econometric models were used to handle different characteristics of damage, casualties and frequency data. First, damage was estimated using the tobit regression model. This was because the accident data showed many cases of zero damages, which called for censored distribution for damage data. Second, the zero-inflated regression model (ZINB) was employed to measure casualty counts. Previously, most of count data has been analyzed by using ordinary negative binomial model (NB) when the count data had unequal mean and variance structure. This was necessary especially for accident data because the number of casualties varied widely across geographic locations and environmental conditions, making the variance much greater than the mean. Although NB had some merits, it implicitly assumed that all ship accidents could necessarily hurt human beings aboard ships. This could be problematic as all accidents would not involve human casualties aboard ships, i.e. some ship collisions or groundings would not inherently incur injuries or death. While these accidents could add to excessive zero counts in the data, they were not governed by negative binomial distribution in a stricter sense, so the estimates of NB could be biased. Addressing this problem was possible through ZINB, which assumed two separate probability distribution for zero accidents. One process predicts accidents that necessarily produce zero casualty (accidents that are “safe for human” by their nature), and the other incorporates accidents, which happen to be zero casualties in spite of their inherent high risk (accidents that are hazardous to human but casualties do not occur by chance). A specification test was used to show that ZINB was more proper than NB for our casualty data. Lastly, we measured accident frequency through the panel NB regression model. The records of frequency were panel data covering accidents at coastal ports from 1992 to 2011. In this case, the latent effects that are specific to an individual port had to be controlled to correctly specify the estimation model (Chin and Quddus, 2003).
The rest of this paper is organized as follows: Section 2 reviews previous studies on vessel safety which adopted engineering or econometric approach. Section 3 develops damage/casualty and frequency estimation models. Section 4 details the data collection procedure and source of data. Section 5 analyzes the results, and Section 6 concludes with suggestions for future research.
Section snippets
Literature review
Vessel speed has been considered as a crucial factor to analyze ship accidents in engineering literature. Brown (2002) developed a simplified collision model to identify how ship speed, collision angle, ship type, and ship displacement affect damages in ship collisions. Generating random collision scenario, the study found that ship speed could have significant effects on ship damages. Mou et al. (2010) estimated the effects of ship speed, size and course to the closed point of approach (CPA),
Accident damage and casualty estimation model
A number of studies proposed unique estimation models of accident damages and casualty. Damage and casualty estimation models developed in this paper largely follow the line of previous models. Nevertheless, since our focus is mainly on the speed reduction effects, only factors that are most relevant to vessel speed are considered. The models can be expressed as a following functional form,Di and Ci indicate the damage cost and the number of casualties,1
Data collection and processing
Variables in the damage/casualty estimation are summarized in Table 1. Variable names and their description are presented in the first and last column, respectively. As dependent variables, damage cost per vessel tonnage (DAMAGET) and the number of casualties (CASUALTY) were used. To eliminate price effects, DAMAGET was converted into constant values by the Producer Price Index from the U.S. Bureau of Labor Statistics. Casualty counts were obtained by adding the number of the injured and dead
Damage/casualty model estimates
Table 4 reports the estimates for damage and casualty models. The two rightmost columns show the incidence rate ratio (IRR), that is, a ratio change in based on a unit increase in explanatory variables. IRR was calculated by , where subscript k means the k-th variable. The estimates for independent variables other than speed variables are discussed first, followed by those for the latter. In lower side of the table, Vuong test statistic is reported, which was to determine whether
Conclusion
This paper contributes to the literature in two ways. First, the paper is the first to develop damage/casualty and frequency estimation model of vessel accidents incorporating ship speed. Second, the paper estimated accident frequency at a port level, which previous studies have not explored yet. Both models were complementarily used to explore the safety issue regarding vessel speed since the damage/casualty model alone could not explain the accident deterrent effect of speed reduction. The
Acknowledgement
We are grateful to anonymous referees for their constructive comments that significantly improved the earlier version of this paper.
References (24)
Collision scenarios and probabilistic collision damage
Mar. Struct.
(2002)- et al.
Applying the random effect negative binomial model to examine traffic accident occurrence at signalized intersections
Accid. Anal. Prev.
(2003) - et al.
Traffic simulation based ship collision probability modeling
Reliab. Eng. Syst. Saf.
(2011) The determinants of fishing vessel accident casualty
Accid. Anal. Prev.
(2014)- et al.
An application of data mining tools for the study of shipping safety in restricted waters
Saf. Sci.
(2011) - et al.
Study on collision avoidance in busy waterways by using AIS data
Ocean Eng.
(2010) - et al.
The effect of high fuel costs on liner service configuration in container shipping
J. Transp. Geogr.
(2009) - et al.
Ship collision risk assessment for the Singapore Strait
Accid. Anal. Prev.
(2011) - et al.
Modeling accident frequencies as zero-altered probability processes: an empirical inquiry
Accid. Anal. Prev.
(1997) - et al.
Analysis of accidents in Greek shipping during the pre-and post-ISM period
Mar. Policy
(2009)
Investigation of shipping accident injury casualty and mortality
Accid. Anal. Prev.
Determinants of injuries in passenger vessel accidents
Accid. Anal. Prev.
Cited by (15)
Dynamic risk analysis of allision in port areas using DBN based on HFACS-PV
2024, Ocean EngineeringSimultaneous tasks as a contributory factor to maritime accidents: A socio-cultural approach
2022, Asian Journal of Shipping and LogisticsA data-driven Bayesian Network model for oil spill occurrence prediction using tankship accidents
2022, Journal of Cleaner ProductionAnalysis of the injury-severity outcomes of maritime accidents using a zero-inflated ordered probit model
2022, Ocean EngineeringCitation Excerpt :Weng et al. (2018) developed a ZINB model based on the maximum likelihood regression tree to predict the mortality of shipping accidents and examine the factors influencing the loss of human life in shipping accidents. Chang and Park (2019) estimated the impact of vessel speed reduction on vessel damages and casualties by ZINB regression and panel negative binomial regression. Past research (Jiang et al., 2013) in statistical modelling of accident data has shown that the generation mechanism underpinning the injury-free and injury-prone accidents may differ.
Accidental release in the bunkering of LNG: Phenomenological aspects and safety zone
2022, Ocean EngineeringCitation Excerpt :The use of liquefied natural gas (LNG) has been significantly promoted by economic, environmental, political, and logistic factors (Chen et al., 2017)(Osorio-Tejada et al., 2017), raising new concerns on safety aspects due to the cryogenic conditions and interactions with industrial operators working in the port areas (Fy et al., 2016)(Chang and Park, 2019)(Aneziris et al., 2020).