METOCEAN
Laboratoire 
Comportement des 
Structures en Mer

A metocean case study

Wave climate bases

tmsi

COMKISS

up to other case studies

How satellite measurements of waves and wind can improve the climate data used for the design of transportation at sea.

Statement of the problem
State of the Art - Old and new
Recommendations
Links and references

These web pages are based on a report prepared by Cees Leenaars of Dockwise N.V., Soazig Louazel, of Dockwise N.V. / Ifremer, and Patrick Brugghe, of Breeman Engineering and Services b.v. within the COMKISS European project.

 

Statement of the problem

fig1_climbase.jpg (8697 octets)
Figure 1 - GWS climatological areas

 

The design of marine transport and offshore installation works requires the input of realistic wave and wind conditions, to be made out from some appropriate climate database. Since 1854, visual observations of commercial ships have been collected and compiled into databases. Recently, hindcasts on one hand, and satellite measurements on the other hand, have led to the development of new types of databases.

These databases are significantly different in presentation, and sometimes in contents. The study discusses the use of OWS (ship observations), GWS (empirically corrected ship observations), IMDSS (hindcast) and CLIOSat (satellite measurements) databases for design of voyages on the major shipping routes of the world, in order to shed some light on the choice of a best strategy for those who want to use them.

fig2_climbase.jpg (11966 octets)
Figure 2 - IMDSS global spectra wave model grip 

carte.jpg (105279 octets)
Figure 3 - ClioSat climatological areas 

 

General description

The most important parameters in transport design are:
Vessel characteristics:
Motion response
Stability
Strength
Route characteristics:
Environmental conditions (wave, swell & wind climate)
Extreme events (Tropical cyclones, etc.)
Existence of safe havens
Re-routing potential
Weather avoidance:
Tactical decisions to avoid forecasted weather
Swift decisions to avoid imminent adverse vessel behaviour

Depending on the design method these parameters need to be modelled more or less accurately, or not at all.

The comparison bears on 4 main categories of design methods:
Rules of Thumb

For many years, classification societies have used very simple formulae to determine accelerations on board, irrespective of the design sea conditions. These formulae make use of characteristic rolling and pitching extreme angles associated with the ship's natural periods to obtain accelerations at the specified locations on board the ship.

Design Wave Methods

The design wave method takes into account the wave climate, in the sense that a design wave is determined for the voyage. It can be the extreme value for the worst area, or the extreme value for a complete route by combining scatter diagrams, and have a return period of 1 or more (10) trips or years.

Voyage Acceleration Climate (VAC) scatter diagrams

The response climate method call Voyage Acceleration Climate (VAC) models the distribution of all sea states on a trip. A distribution is computed by weighting the distribution of each climatological area along the route by the time spent in that area. A complete response distribution is then computed, that is then used for the design with respect to reliability objectives or fatigue.

Monte Carlo Simulation (MCS) to step through the voyage history

In this method, a full historical database of wind and waves conditions measured -- or simulated -- and forecast is used. A date of departure is drawn at random in the history for the season of interest, and the voyage is simulated in 12 hours steps, including decisions of routing and bad weather avoidance. For each sea state, the response of the ship is calculated. Gathering the results for all random draws, a good estimate of the response distribution is obtained.

Typical applications

Typical applications of these design methods are the design of sea-fastenings for the cargo on a voyage, and of ships themselves, validations of the safety requirements for stability and/or maximum acceptable motions and accelerations, advance cost and duration estimations for trips.

State of the Art - Old and new

ADover-Gibraltar
BNew Orleans-Gibraltar
CGibraltar-Port Said
DSuez-Aden
EAden-Muscat
FMuscat-Colombo
GColombo-Singapore
HSingapore-Kohsiung
IKohsiung-Hiroshima
JDover-Stavanger
KHalifax-Newcastle
LHiroshima-Muscat
MHamburg-Muscat
NHiroshima-San Francisco

Present methodology

Visual observations of commercial ships have been archived for a century and a half, and since 1961, the collection is systematic according to a resolution of the WMO. The most well-known compilations of these observations are the OWS (Ocean Wave Statistics, Hogben & Lumb, 1967) and the more recent Global Wave Statistics (GWS, Hogben, Dawnka & Olliver) that takes advantage of the experience of detected biases in OWS to correct them. The main advantages of GWS / OWS are the duration of the collection period and its suitability to shipping applications, because it takes into account bad weather avoidance and it is well-documented for the major shipping routes. The main drawbacks are the lack of information outside the main routes, the poor accuracy for periods (whereas heights are well estimated by these experienced observers), the lack of wind information, and some deficiencies in seasonal variations modeling and in reporting extremes, good or bad.

Hincasts compute wave heights from historical wind databases. The codes for this simulation have reached a good level of maturity, but errors and uncertainties in the input wind fields are amplified by this process, and the quality of the results is thus often impaired by the lack of accuracy or of validation of the wind data, especially for regions where few observations and little experience are available, such as in the southern hemisphere. The main advantages of hindcasts are that they provide worldwide long-duration histories of data. The main drawbacks are that they are proprietary and costly, that they depend on the personal skills of the analysts who verified and corrected the wind fields, and that they have limited accuracy in extreme conditions. It should be noted that the availability of satellite scatterometer measurements of winds has significantly improved the reanalysis processes for the last decade.

 

route_n.jpg (10257 octets)
Route N : North Pacific

Recent improvements

Earth observation satellites offer now reliable characterizations of wind and sea conditions:
significant wave height and wind speed from altimeters (every second, i.e. 7 km, along the orbital track);
wind fields, with speed and direction, for many cells (typically 25 km) of wide swaths (500 to 1500 km);
directional spectra from Synthetic Aperture Radars in wave mode;
sea surface temperature from radiometers.

Such satellite information can be accessed either as raw data, or from added value resellers such as Satellite Observing Systems Ltd., Météomer or Oceanor. The corresponding products are:
SOS: WAVSAT climatologies
Météomer: CLIOSat climatologies and atlas
Oceanor: World Wave Atlas

With respect to conventional databases, satellite information brings in the advantages of better quality / accuracy, especially in areas where there are few reliable field measurements to calibrate hindcast models; of a more detailed characterization of sea conditions (directional spectra, sea surface temperature), and of better time-space coverage (no blank areas, such as West Africa for GWS; seasonal or monthly details, ...).

The drawbacks result from the difficulties in finding automatic qualification methods that can both process the huge amounts of available data, and yet eliminate intelligently effects of rainfall; from the non taking into account of decadal trends due to the short operation period, as of now, of satellites; and from storm undersampling in regions influenced by tropical cyclones of small size and duration. In addition, it is not possible to reconstruct histories that could be used in Monte-Carlo Simulations because of the sparsity of the time-space sampling.

 

Analysis of the example

VAC was tried out for GWS, for the IMDSS hindcast database of Oceanweather, and for the CLIOSat atlas, on the 14 major shipping routes for each of the 4 seasons.

GWS gives consistently higher results, by about 30%, than IMDSS and CLIOSat. These two latter databases are very similar in average. Anomalies such as tropical cyclones may affect them both (but sometimes in different ways), but their effect should diminish as more years of data become available.

Recommendations

Research actions

Provide "best estimates" of climatologies, blending the various sources of data.

Refine seasonal discretization to suit local phenomena such as monsoons.

User persuasion

Building statistics on a route from the presentation given in atlasses such as CLIOSat should be made a much more straightforward task.

There is still a need to increase confidence on the qualification and reliability assessment of satellite data.

Expected improvements

Continued collection of satellite data will automatically increase the reliability of climatological databases.

Links and references

Providers of Climatologies

SOS: WAVSAT climatologies
Météomer: CLIOSat climatologies and atlas
Oceanor: World Wave Atlas
OceanWeather: IMDSS Hindcast

 The COMKISS project:

Project objectives 

The main objectives of the project are: 

to demonstrate to major segments of the marine transport industry the benefits of integrating satellite-derived information on sea state such as wave height and direction. 
to raise awareness of the usefulness of satellite data in increasing the safety and overall efficiency of shipping operations by using the EWSE (at CEO) as the principal channel for communicating progress. 

The results should be of interest for enterprises such as ship certification, fast ship/coastal traffic, and transportation of unconventional loads.
 

Project partners 

Satellite Observing Systems , Godalming, UK (David Cotton, project manager)
Mathematical Statistics, Lund University, Sweden (Georg Lindgren, project
co-ordinator)
Bureau Veritas, Paris-La Defense, France (Guy Parmentier)
Dockwise, Meer, Belgium (Cees Leenaars)
IFREMER, Brest, France (Michel Olagnon)
OPTIMER, Brest, France (Raymond Nerzic), 
Corsica Ferries, Bastia, France
 

Project period 

September 1, 1998 - August 31, 2000

 

Mise jour 09/10/00
Webmaster
Copyright Ifremer
[ Welcome IFREMER | Metocean | Sommaire | Search | Contact us ]
L> L>