The effect of changes in wind strength and wave heights on the safety of vessels in shipping.
This article investigates whether changes in oceanographic conditions can be filtered out tomeasure their effect of the overall safety level of ships. The article is based on a unique datasetof 3.2 million observations from 20,729 individual vessels for the time period 1979 to 2007 inthe North Atlantic and Arctic region. It combines ship particular information, ship safetyinspections, casualties, ship economic cycles and oceanographic data. Standard econometricmodels are used to measure whether the effect of significant wave height and wind strengthtowards the probability of casualty can be measured and tests whether it changed over the time period on hand since changes in oceanographic conditions have been confirmed in theliterature for the North Atlantic. The results show that the effect of wind strength andsignificant wave height can be measured towards the probability of casualty although there isno clear seasonal pattern while overall; the probability of casualty is influenced by seasonalitywith the winter month showing the highest probability of casualty. With respect to changesover time periods, significant wave height shows an increasing effect in January, March, Mayand October while wind strength show a decreasing effect over time, especially in January,March and May. The results for significant wave height might be relevant for the policy makersuch as the International Maritime Organization (IMO) in the context of developing goal basedstandards for ship constructions or revising common structural rules used for the design ofships
Furthermore, the IMO Code on Intact Stability [14] outlines design criteria with reference tosevere wind and rolling criterion. Section 2 provides a description of the dataset which was used for this article and the type of corrections that were made to the oceanographic variables while section 3 describes the variables and model combinations used to measure the effects of interest. Section 4 presents and discusses the results of the models and section 5 provides the conclusions and recommendations for the policy makers.
2. Description of dataset used and corrections made for oceanographic data
2.1. General description of data sources to construct dataset
In order to measure the effect of changes in oceanographic conditions on safety, variousdatasets need to be combined such as data which is used to determine the general risk profile ofa vessel (ship particulars, their changes over the time period, information on safety inspectionsand audits), ship economic conditions, casualties and oceanographic conditions.Most of the data on ship particulars, casualties and ship economic cycles comes from a datasetused by Bijwaard and Knapp [15] with the addition of oceanographic data. The time periodused for this article is 1979 to 2007 and the total dataset used for the analysis contains 3.2million observations for general cargo vessels, dry bulk carriers, tankers, container vessels, passenger vessels and other ship types (unknown ship types and fishing vessels since thesample for fishing vessels does not provide a good representation of the fishing fleet in generalare exclude). The combination of data is given in Table 1 where we identify each of the datasources:
Table 1: Combination of data and data sources used for the datasetData type Data Source
Ship particulars (eg ship type, tonnage, age,etc.) and their changes over time (eg. flag,classification societies)Lloyd’s Register Fairplay (LRF)Safety inspections and ISM audits
•
Various Port State Control Regimes
•
Industry Inspections from RightShip, ChemicalDistribution Institute (CDI) and the Oil CompaniesInternational Marine Forum (OCIMF)
•
Flag state inspections and ISM audits from various flagstates*)Casualty data
•
Lloyd’s Register Fairplay (LRF)
•
Lloyd’s Maritime Intelligence Unit (LMIU)
•
International Maritime Organization (IMO)Ship economic cycles Shipping Intelligence Network, Clarksons ResearchOceanographic data International Comprehensive Ocean-Atmosphere Data Set(ICOADS)
*) The flags states would like to remain anonymous
The oceanographic dataset is the basis for this analysis to be complemented with data fromBijwaard and Knapp [15] since it provides the oceanographic variables (wave, swell and winddata and vessel traffic information. The data comes from the International ComprehensiveOcean-Atmosphere Data Set (ICOADS) for the time period January 1979 to December 2007and contains approx. 44 million worldwide observations which were reduced to observations inthe North Atlantic.
ICOADS is the largest available compilation of surface meteorological observations fromVoluntary Observing Ships (VOS), complemented for recent decades by data from buoys andother automated platform types. The VOS observations form a baseline data source for theanalysis of marine climate stretching back over 200 years and the scheme is operated under theauspices of the Joint WMO/IOC Technical Commission for Oceanography and MarineMeteorology (JCOMM). A peak in the total number of VOS was reached in 1984/85, when7,700 ships worldwide were listed as participating in the VOS Scheme. However, the numberof ships declined to around 6,000 by mid-2005.The ICOADS dataset of 44 million observations was reclassified into regions based onlongitude and latitude and the North Atlantic region was selected for this article. The datasetwas then merged against ship particular data from LRF using the call sign to match thecorresponding IMO number (IMO is a unique ID) to be able to add ship particular data, theirchanges over the time period in question and inspection data. From the original 7.5 millionobservations of the North Atlantic region, 3.9 million could be identified by IMO number andexcluding fishing vessels and unknown ship types, 3.2 million observations of 20,729individual vessels form the basis for the analysis.The data on safety inspections include the results of port state control inspections from various port state control regimes and industry inspections. The industry inspections are called vettinginspections and are performed by the Chemical Distribution Institute (CDI) on chemicaltankers and oil tankers, by RightShip
5
primarily on dry bulk carriers and by the Oil CompaniesInternational Marine Forum (OCIMF) on oil tankers. The inspection system of OCIMF iscalled SIRE which is used in this article.For the dataset used for the models, the ICOADS dataset is merged with the casualty data forthe same regions and time period. The safety inspection data then complements the datasetwhere all observations six months prior to the event date (e.g. casualty or observation withoutcasualty) is taken into consideration. For safety inspection, emphasize is given on detentions
and the number of deficiencies found during an inspections besides the fact that an inspectionwas performed. Finally, the dataset also accounts for changes in ownership, Document ofCompliance Company (DoC), registration of the vessel and class withdrawals for a time periodof one year prior the event date. In this way, the effect of these variables is accounted for asinvestigated by Knapp and Franses [16]. In addition, ship economic data such as earnings(USD) is added to account for ship economic cycles and which can also influence the safetylevel of a ship according to Bijwaard and Knapp [15]. Earnings are based on data receivedfrom Clarksons and was deflated to account for inflation
.As a final step, casualties are classified into their seriousness according to definitions from theInternational Maritime Organization (IMO) which is
very serious, serious or less serious
or bycasualty category (e.g. collision, contact, fire, hull related failures, eg.). The definitions of theseriousness of casualties are given in MSC/Circ. 953, MEPC/Circ. 372 [17] and MSCResolution MSC.255(84) [18]
RightShip is an independent vetting inspection system located in Melbourne, Houston and London and performs inspectionson all ship types but primarily dry bulk carriers
A vessel is detained when it is found in severe violation of the international conventions and is only released after therectification of its deficiencies.
Historical monthly inflation rates can be obtained from http://www.inflationdata.com
It is worth noting that the IMO Maritime Safety Committee (MSC84) adopted MSC Resolution 255(84) on 16 May 2008where the definitions were slightly changed and no longer distinguish between serious and less serious casualties. Thedefinition for very serious casualty remains however unchanged. The reporting requirements will also change in the future.
Relevant for the analysis and the dependent variables used in the models in section 3 are thedistinction between casualties that are
weather related
or
not-weather related.
Weather relatedcasualties cover the following categories:
flooding, foundering, capsizing, hull related failures,wrecked, stranded and grounded
. Table 2 provides an overview of the total amount ofobservations of the final dataset, the amount of observations without casualties and the amountof casualties with their respective split up into being weather related or not.
Table 2: Observations per ship type and casualties
Observations CasualtiesShip types grand total no casualtiesweatherrelatednot weatherrelatedtotalcasualties
general cargo ships 1,055,906 1,048,906 2,303 4,697 7,000container vessels 883,076 882,157 258 661 919tankers 450,735 445,535 1,628 3,572 5,200dry bulk carriers 293,282 290,278 1,116 1,888 3,004other ship types (cargo) 492,757 492,271 142 344 486 passenger ships 106,904 105,950 292 662 954
Total 3,282,660 3,265,097 5,739 11,824 17,563
2.2. Correction for wind and wave data
For some of the oceanographic variables used in the analysis such as the wind and wave data
9
some corrections need to be applied before it can be used in order to correct for biases due tothe different measurement techniques. For the wind data, the method describes by Thomas
etall
[19] to homogenize the wind speeds is used. The data is divided into wind speed which wasderived by measurement and wind speed derived from visual estimation. For wind speedsderived from measurements, the correction formula given in equation 1 is used while for theadjustment of visual estimates, the correction formula given in equation 2 was applied – bothequations are based on Thomas
et all
[19].
W
cor
=
W
u
8.7403/ln(HOA/0.0016) (1)
32
0004.00221.01888.10161.0
uuucor
W W W W
+−+=
(2)
Wcor
of equation 1 or 2 is the corrected wind speed,
Wu
is the uncorrected wind speed and
HOA
is the height of the anemometer when given (the value 25 was used for missing values).For the wave and swell correction and in order to calculate the significant wave height (SWH),a correction formula given by Gulev and Grigorieva [20] is used and which is presented inequation 3 where
WH
represents the wave height and
SH
the swell height.)(
SH WH SWH
+=
(3)For the casualty dataset which was added to the overall dataset, the significant wave height andwind strength are derived from average daily values based on the total North Atlantic region(7.5 million observations) where we match the average daily value within a Marsden grid (10 x10 square). Some outliers (7,311 observations).for the wave data are identified an since theirvalues are unusual high and far above the normal range with a maximum of 20 meters (refer toHolliday et
all
[21]), they are excluded from the final dataset
9
The exact variables used are: wind speed, the wind speed indicator, the wave and swell height
Order an Essay Now & Get These Features For Free:
Turnitin Report
Formatting
Title Page
Citation
Outline
