Skip to main content
European Commission logo print header



Executive Summary:

The target of this research project is to develop an operative forecasting system for dynamic and especially compressive ice for the maritime. The elements of this system include the description of compression – magnitude and direction – in terms of quantities that can be applied in ship operation.
The use of these forecasts enables the merchant vessels to select their route for safety while the assisting icebreakers can determine the waypoints for the merchant ship routes based on most efficient and safe passage of ships. Thus the development of an operative tool for forecasting the hazardous area of ice compression will improve the overall safety and efficiency of the winter navigation system.

Additionally, the forecast of ice compression is integrated to the existing IBNet system, which is management system of IB operations onboard Finnish, Swedish and Estonian icebreakers.

Moreover, framework enabling risk assessment of ice navigation for different ice class vessels is developed. This framework allows estimation of the effect of ice compression on ship performance, quantifying how the ice cover and its characteristics impede the ship navigation. Once the real risks are known, these can be taken into account for future ship designs. The risk analysis results can also be taken into account when the future ice rule requirements will be developed. The technology developed from Baltic Sea models could easily be also transferred to any relevant region of the world.

Finally, it is envisaged that the developed forecasting system to be implemented on board IBs as decision support tool for ice navigation, may have a large impact on the probability of ships getting stuck in ice and thus decrease the risks remarkably.
Project Context and Objectives:
Background / State of the art
The oil tanker traffic is strongly increasing at present in the Baltic. Winter conditions add the risks related to the navigation. Most of the largest damages occur in a moving ice cover. If ships get stuck in dynamic ice, large compressive forces can act on the ship hull and hull damage or even rupture can occur. Dynamic ice and compressive ice are the two most important hazards for ships navigating in ice.

In order to aid ship navigation and reduce risks, ships should get a warning for compressive regions. The basic ice charts are too general for use in ship operations. The present operational models provide daily forecast of the ice motion, ice thickness, concentration and ridging. However, there is still large gap between the needs of the tactical navigation and safety of the shipping and the products provided by the operational sea ice models.

The studies have highlighted the importance of the dynamic and especially compressive ice. Even if the importance of dynamic ice is known, the behaviour of ships in this kind of ice is not well known. The target of this research project is to develop information and methods that can be used in predicting the compressive ice and that can be used operationally. The present project contributes directly increasing the safety and efficiency of winter navigation.

1. Develop methods to predict the compression in ice operationally. This objective is related to sea ice dynamics models that have been developed and are further developed to be suited for the operational task.
2. Develop quantities that can be applied to describing the ship progress and loading in ice and that can be obtained from ice dynamics predictions.
3. Verify the accuracy and applicability of the ship-scale quantities obtained from mesoscale ice dynamics models.
4. Develop an operational procedure to describe ice compression to be broadcasted to shipping.
5. Apply the developed ice compression forecasting procedure to Arctic waters for oil exploration/production facilities.

Work plan
The project is divided into eight work packages out of which the first and the last one are dealing with project management (either management directly or reporting). The remaining six work packages can be divided broadly into four categories (observations, sea ice modelling, operative forecasts and applications).

The partners of the project are: Aalto University (AALTO), Alveus d.o.o (AS2CON), Arctic and Antarctic Research Institute (AARI), Finnish Meteorological Institute (FMI), Finnish Transport Agency (FTA), ILS Oy (ILS), Stena Rederi AB (STENA), Swedish Maritime Administration (SMA), Swedish Meteorological and Hydrological Institute (SMHI), Tallinn University of Technology (TUT) and Tallink.

The theoretical research work on the sea ice dynamics models is carried out by AARI, FMI, SMHI and TUT. Research on the quantities related to ships in dynamic and compressive ice is mainly carried out by AALTO, ILS and AS2CON. The operational forecasting of ice dynamics and compression in ice cover will be made by ice services participating (AARI, FMI and SMHI) in the project.

The applications of the project include the use of the operational forecasts of ice conditions as well as the use of the knowledge on ice dynamics and compression on development of risk control methods for winter navigation and ice management schemes. Here the role of the maritime authorities (FTA, SMA) and the ship owners (STENA, Tallink) is crucial. The administrations and ship owners give the scientists in the project a possibility to access ships that are operating in ice.

Overall, the partners in SAFEWIN complement each other well as the whole span of research from theoretical research via applied research to applications in practical operations is covered
Project Results:

This document presents the S&T achievements of SAFEWIN. These are list and discussed per each Work Package.

1.1 WP2 Observation campaign

Compressive situations in dynamic ice field are known to be difficult conditions for ship, but in the beginning of the project almost no measurements data existed about compressive situations. In Work Package 2, the aim was to execute observation campaigns to gather knowledge and information about the mechanical and physical ice properties and their seasonal changes, to conduct specific campaigns for measuring compressive forces and their forcing factors in the Baltic Sea and to analyse shipborne observation data and AIS data in order to assess influence of ice conditions and particularly compressive ice on the efficiency of motion and safety of icebreakers and merchant ships.

Results and implementation
Three observation campaigns were conducted in the winters 2009-10, 2010-11, 2011-12. The measurements and their results are reported in the Deliverables D2-3, D2-5 and D2-6. The results and their implementation are reported in Deliverable D4-3. Several different measurements and observations were made in the campaign and the following chapters will describe these actions and their implementation.

Flatjack stress measurements were carried out during the observation cruises on R/V Aranda. The objective of these measurements was to provide calibration, develop measurement practices and quantify maximum stresses. Consistent data was measured and interesting new correlations were observed from the measurement results.

The object of the measurements of compression in the ice field was to determine the pressure in the ice field for the validation of the forecast model and for the ice model tests to be conducted in the ice basin of Aalto University. The correlation between the equipment and FMI CRREL buoy was good and results reasonable. The measurements also showed that the pressure has a direction.

The main goal of the compressive ice strength measurements was to define the compressive strength of ice for the model tests to be conducted in the ice tank of Aalto University. The knowledge gained from the measurements can also be used to estimate the maximum bearing capacity of the ice in compressive situations.

A very extensive helicopterborne EM thickness mapping campaign was conducted in the beginning of March 2011. During longer flights a general understanding of ice thickness statistics was obtained. In spite of the long research tradition on Baltic ice there has not been a good understanding of the ice thickness distribution especially for ridged ice types. The data collected provided a convincing data for validating the ice model results and correcting the shortcomings of present ice information. Using the HEM data, extrapolated and updated with respect to drift with SAR imagery, the thickness variations along a ship route can be tracked down. The collected dataset has unique comprehensiveness and detail in this respect and has global general relevance.
During the RV Aranda ice camp close to Tankar, the thickness characteristics of the station were mapped. There were three motivations for this work: to get validation and sub-footprint data for the HEM data, interpretation of stress buoy and stress array data, intercomparison and methodology study using different EM-31 devices. The conducting of calibration exercises generates a cumulating understanding of optimal retrieval parameters which in a long run enables the use of EM in a more routine fashion.

The objective of the stress buoy measurements was to collect a continuous stress record for quantifying stress levels and for relating the stresses to wind forcing, ice cover geometry and ice kinematics. The stress sensor produced good and valuable data for two weeks. The data will be used to address the phenomenon of stress generation and propagation. This work has general global relevance as the Arctic campaigns have not had as comprehensive supporting datasets and relationships between ice cover stresses and kinematics is still poorly understood. On the other hand, the data will be directly used to validate the compression fields of HELMI ice models and related to the performance of the ships navigating at the close field. The objective of the GPS drifters was to determine the kinematic variables like convergence, divergence, shear and rotation which constitute the basic reference when the relationship between dynamical ice stresses and ice motions are studied. The buoys provide drift data that can be used to validate modeled ice drift and tune ice model parameters.

The coastal radars at Marjaniemi and Tankar provided a continuous high-resolution time series of radar images from the respective areas. One of the main motivations for installing the radar capture systems was to obtain detailed data on drift and deformation events. From the imagery, transferred real time to FMI, the opening of a coastal lead, ridging, compressing against fast ice and similar events can be immediately spotted. It is not unfeasible that working further from the kinematic algorithms it would be possible to develop methods that recognize compressive conditions basing on kinematic fields only. The SAR images, about one image per day throughout the season, provide a basic background reference for all ice research activities. The objectives of ship radar and shipborne EM measurements were similar as for the coastal 'observatories': to monitor ice kinematics and thickness and relate these with other geophysical variables and the performance of Aranda.

Ice drift and deformation parameters can be in situ monitored using drifter buoys which are fixed in the ice. Drifters well showed actual ice movements, drift parameters like speed and direction, but also other dynamic properties in drift ice, also response of drift ice to wind forcing. Together with forcing factors this dataset of drifters’ movement should give more information about creation and nature of ice compression events. Drifters were released both in the Gulf of Finland (2010) and in the Gulf of Riga, also Pärnu Bay (2010-1012). Drifter data were synthesized with ships and icebreaker operations data, both ice observations from board of the ships and movement of ships themselves (AIS) form basis for ice compression analysis in certain sea area.

With aim to experimentally study wind, ice and current interaction in local scale (which in fact create potential for ice compression), ADCP technology with bottom-track (BT) facility was applied. The latter ADCP configuration applied on the bottom mounted ADCP allowed to distinguish the presence of ice cover in the location as well as its main dynamic properties, like drift speed and pattern of movements. Both observed phenomena the presence of ice cover and the ice drift are the output parameters of ice forecast models and in this way the ADCP measurements serve the model calibration and validation purposes.

Ship hull vibrations were measured in order to give estimate to the magnitude of ice resistance experienced by the ship. The measurement data show that conditions of ice compression could vary much just for two following days and even in case of stable ice conditions (fast ice, same ice thickness) the wind direction as main driver vary the ice compression along the route.

The objective of the compression survey was to gather more knowledge about compressive situations; occurrence frequency, severity, ships behavior and duration of the situations. The compression reports were compared with AIS data. The AIS data correlated well with the reported compressions in general. As far as the ice compression forecast model has been compared with the AIS-data, the model does not correlate that well with the AIS-data along the route of a ship. The compressive survey and AIS data can be used also in the future for the validation of the forecasting model. In addition, combining knowledge from the model tests with the survey and AIS data can be used for the development of the risk model.

Meteorological observations and wind parameters first of all are needed to interpret the ice dynamics and ice compression process in particular. Ship-borne ice observations on basic ice parameters (thickness, ice drift, ridges) as well ice compression are of high value in order to understand ship performance in dynamic ice conditions. This experience got from the expedition helps to asses all reports about ice compression situations originated from other merchant vessels and icebreakers.

Innovative nature
The simple method for measuring the compression in the ice field used by Aalto had been used for similar measurements by Helsinki University of Technology (nowadays Aalto University) in the 90’s. However, the applicability of the method and instrumentation was demonstrated within the SAFEWIN project. Due to the good correlation of the simple method and the simultaneous measurements with CRRL buoy, it can be concluded that the method is applicable for compression measurements inside the ice field and can be used in the future when similar measurements are conducted in ice infested sea areas.

There are very few ice dynamics data of such temporal resolution as ADCP allow to obtain, that’s innovative. Despite these data are limited in space well drawn picture about ice dynamics and its high-resolution response to forcing factors could be obtained. Dataset of drifters movements give better spatial coverage and are also frequent in time giving Lagrangian ice movement properties spanning over several months. During the experiments drifters were most of the time on-line and ice movement data were used in real time in organisation of winter navigation off Kunda (Gulf of Finland) and in the mouth of the Pärnu Bay (Gulf of Riga) – this was innovative concerning these measurements.

Linking to other SAFEWIN results
The compression survey and AIS data were used to estimate the accuracy of the FMI’s forecasting model. The survey and AIS data were compared to hindcast results and forecasted compression after each observation campaign and the results of the comparisons were reported. Based on the comparison, the feasibility of the model was estimated.

Other drifter experiments took place in Bothnian bay, carried out by FMI. Comparison of ice dynamic properties from these different basins would contribute into general knowledge about ice dynamics in the Baltic Sea in general. FMI model HELMI as well ice dynamics model used by AARI could be validated using these ice movement data.

Implementation and exploitation
Datasets of ice movement from drifters and ADCP are available for further analysis, both for process analysis and ice dynamics model validations. Data cover the Gulf of Finland and Gulf of Riga during period 2009-2012.

1.2 WP3 Forecasting Methods

1.2.1 D3-1 Report on first ice modeling result

Problem addressed:
Compressive situations in Sea Ice can severely delay the time schedule of ships and, in the worst case, lead to ship damage. The project thus aimed to investigate and forecast compressive situations in Baltic Sea ice. The development of forecast methods has been crucial for the success of the project which aimed at operative readiness at the end of the project. The development task can be divided into two subproblems: 1) the enhancements and implementations of models so that compression is included as prognostic variable, and 2) validating exercises targeting the modeled compression fields and tuning of the model for optimal performance. The validating, on the other hand, typically proceeds in two steps. The compression is a dynamic variable determined by the strain fields (ice drift) and the rheological parameters linking the internal stresses with the rheology. The prerequisite for successful modeling of compression is that the model performs well for the strains. This can be studied from the ridged ice thickness generated by strain induced deformation, and by direct comparison with drift data from imagery and buoys. In the next phase, after tuning the model for best possible performance with respect to ice drift, the validation exercises targeting compression can be taken up.

FMI, SMHI and AARI implemented stresses as a new prognostic variable into their existing sea ice models. The models are have differences as concerns the ice cover variables, adopted rheology and compression descriptors, deformation physics and numerical implementation. The numerical tests and hindcasts conducted also targeted different aspects and time periods. FMI and SMHI used the same ice model code (HELMI) and described the compression with the force arising from ice internal friction; this is a descriptor of dynamic intensity in the ice and the unit is N/m2. AARI used a definition based on the ice stress tensor and the results are in kPa.

For FMI model HELMI hindcast simulations with compression description were conducted for several winters. Observations were made on the spatial and temporal distributions of the compression, on the variation of compression magnitude and on the areas typically generating high compression. In addition, the forecasted thickness fields were compared with helicopterborne electromagnetic (HEM) thickness soundings for two winters. These showed some discrepancies in the amount of deformed ice and suggested the need of further validation exercises with drift data.

SMHI conducted similar analysis on the stress distributions but made also 45-year hindcast simulations with BASIS data comparison. In addition to multicategory HELMI a single-category ice model was used (solution typical in large scale Arctic models) and results were compared. Climatological results on the occurrence of compression were generated.

AARI modeled pressure (compression) in the Gulf of Finland and conducted a model skill core analysis using their established procedures. Results on the spatial and temporal distribution, on pressure magnitude variation on seasonal variation, and on the locations of high pressure are were obtained.

Innovative nature:
The inclusion of compression variable to ice forecast models is a central innovation of SAFEWIN project. Compression forecasts are high on the navigator's wishlist not only in the Baltic but in all ice covered sea areas. As the approach is new, the hindcasting and validation exercises are very extensive. The hindcast results, covering tens of years, the extensive datasets from WP2, compression observations from the ships, and AIS-retrieved ship performance data create a very comprehensive set. Only a fraction of the possibilities offered by this set have been yet utilised by the project. The research on the themes is expected to continue in other projects that can directly proceed to central research questions as the arranged data is at hand.

Link with other SAFEWIN-results:
The modelling work takes a central position in the project and is linked with all other WP's. These are collecting data for model validation and development purposes (WP2), addressing the basic physics of ice cover stressess (WP2&4), linking the modeled compression fields to the ship experiences (WP4), operationalising the model developments (WP5), and developing all aspects of winter navigation information and decision making system.

Implementation and exploitation:
The results were developed further in other work packages. The end result of the research chain is the operative implementation of compression forecasting. This will be addressed more in other sections of this report.

1.2.2 D3-2 Ice model validation on local scale

Ice forecast models principally forecast ice drift fields and the reliability of these in general determine the quality of the forecast. If the drift fields are realistic, also the localisation and thickness of deformed ice types can be expected to become correctly modeled as ice deformation is driven by convergent ice movements. On the other hand, the compression fields, as described by the internal friction force, depend on both the thickness variation and state of ice drift. However, compression forecasting sets higher than normal requirements for the model performance. The usefulness of drift forecasts is not reduced by small discrepancies if only the state of drift/no-drift, the drift speed and drift direction are about right. In contrast the compression is a localised and transient phenomenon and the location of high compression areas is sensitive to small changes in wind and ice drift direction. This is also the observation of ice navigating ships experiencing compressive situations.

The uncertainty of ice forecasts, especially when compression is included, can have many sources. On fundamental level the continuum assumption may be unrealistic or the idealised rheology does not capture true physics behind the compression phenomenon, the description of thickness and thickness change may be oversimplified, the rheological parameters that describe aggregate properties of the pack cannot be fixed, and the forcing data, especially wind fields, can be unreliable. Various hindcasting exercises using reanalysed wind fields, that is, best possible forcing data, were conducted with HELMI model in order to study the performance of the model from basic model physics viewpoint.

The validation data consisted of GPS drifter data and compression reports from ships. The drifter data included two arrays with seven drifters in the Bay of Bothnia. Statistical comparisons made evident that the model predicted too high drift speeds. Also the onset of drift occurred too easily, that is, the ice was drifting in the model when it should have been immobile. The discrepancies were diagnosted to the parametrisation of aggregate ice cover strength. After a series of experiments the strength parameter was replaced by a new value that was 2.5 times higher than the previous. This value was adopted in the operative ice forecasts of FMI.

The modeled compressions were then compared with ship observations, reported as numerals from 1 to 4. Both the average and maximum compression in an area of certain size around the ship was used in the comparisons. For each numeral the variation of the model values was large but on the average the mean and median increased with the numeral. The ship observations were also put on the map together with the modeled compression fields. For days with largest numbers of observations the comparison was made for every hour. It was seen that the high observed values were often in areas of low modeled compression but close to the edge of high compression area. The locations of the high compression areas could also change considerably in an hour so that the accuracy of the observation report time stamp was crucial. The studies in general pointed out the need of thorough analysis of the spatial and temporal variation of compression fields that was then taken up.

Innovative nature
The validation exercised followed in part standard procedures. However, the drifter comparisons were crucial as they resulted to a significant readjustment of model parameters. Without this the hopes to get the forecasted compression fields about right would have been meagre, and the adjustment also improved the operative model performance in all respects. As the modelling of compression and validating the results was an activity that had not been done before, the validation exercises were first interpreted to indicate poor performance of the compression forecast. However, it was finally realised that compression fields behave very differently from other modelled ice variables. This suggested new lines of study that finally resulted in the adoption of ensemble forecasting, as described in WP5 section.

Linking to other SAFEWIN results
As described for D3-1 the modelling work links to all other work done in the project.

Implementation and exploitation
This deliverable was a link in a chain of development efforts that resulted to the adoption of ensemble forecasting and its operative implementation as described in WP5.

1.2.3 D3-3 Modeled statistics of ice compression situations

Several experiments focused on direct comparison between ice pressure values obtained from AARI model and from visual observations conducted in the Gulf of Finland in 2010 and 2011, and Kara Sea in 2009 showed that the AARI model, at least qualitatively, adequately simulates the real ice pressure events. More detailed description of this study is presented in D7.2.

The basic goals of the WP3.3 study were:
• to obtain the statistical correlations between ice pressure in various spatial-temporal scales
• to develop the methodology to estimate the extreme ice pressure of given probability in given spatial-temporal scale

Using the models of ice dynamics in the Baidara Bay (the Kara Sea) with various spatial scales and temporal steps, the standard deviation of ice pressure within the model cell area was analyzed.

The analysis showed:
• Standard deviation of ice pressure statistically depends on average pressure (correlation coefficients are about 0.5-0.6) i.e. the higher average pressure is the higher standard deviation
• Standard deviation of ice pressure statistically depends on ratio of scales, i.e. the higher ratio of scales is the higher standard deviation;
• The maximum standard deviation corresponds to temporal step about 6 hours, while the most detailed temporal resolution of the model – 1 hour – gives smoother picture of ice pressure.

On the basis of this analysis and using the Gumbel distribution, the algorithm of estimating the standard deviation of ice pressure within the model cell area and the extreme ice pressure of given probability in given spatial-temporal scale is proposed. It seems that the proposed algorithm is only one of possible ways to downscale the ice pressure field.

Undoubtedly, the proposed approach has several weak points. The basic ones of them are:
• Statistically the algorithm is based only on three cases of scale ratio: 1/4, 1/6.25 1/25. It is not sufficient
• Statistical correlations are not very close, the correlation coefficients and R2 indicators are predominantly about 0.5-0.7
• The empirical formulas correlating standard deviation and average ice pressure will not work well enough if average pressure is very low

Nevertheless, in general, the approach seems to be interesting and hopeful.

Linking to other SAFEWIN results
The measurement and observation data collected in the WP2 Observation Campaign were used when the simulated ice pressure fields were compared to the measurements. Furthermore, the conducted analysis showed satisfactory qualitative correspondence, i.e. the model results can be in general recognized as adequate. This study was continued in the study for D7.2 Quality of ice forecasts for ice dynamics forecast.

1.3 WP4 Compression in ship-scale
In this chapter, the description of the results and implementation are given from each Deliverable. The results and implementation of D4-3 are not described here as those were described together with the Work package 2 in the chapter above.

1.3.1 D4-1 Data available about compression

Background and results
Sea ice compression is a subject that is very little studied, and there are hardly any systematic observations. The long-term systematic observations about the influence of compression on ships from the Baltic Sea are totally missing. This lack of data is surprising as compression is one of the most dangerous situations for ships in first year ice fields. Often compression is used loosely to describe an ice field where the resistance is increased but this increased resistance is mostly due to locally thicker ice. Compression as meant here requires a converging ice cover where either stresses are present and/or ice is moving.

The purpose of the D4-1 Data available about compression was to gather information of compression and present the existing models to describe ice loads and ship motions in a compressive ice field. In the review, the compression was described from the ship point of view.

There are two categories in triggering the compression in the ice field, dynamic ice compression and static compression, which divide the character of the compression. The report discussed on how the ice is compacting and what the driving forces are. The wind is the most dominant driving force in the Baltic Sea, but several other driving forces exist – the sea currents are the most important of these. The Russian system to observe compression was also presented. The Russian system has a long history, and lot of data has been gathered from the Arctic. The second part of the report discussed the compression in a sea ice field, and the processes involved in compressive ice are presented. Added resistance, failure process of the ice cover and ridging forces are the most evident processes and also the most important ones.

In order to examine the research on ships in compressive ice, a state-of-the-art of the models and descriptions was written. The literature review consisted of three parts:
• The first part presents the study of ship hull damages, focused on damages caused by compressive sea ice field. Two incidents from winter 2003 were described, and the ice conditions from the incident time period were presented.
• In the second part the literature of ship operability in compressive sea ice field was presented. The compression index was derived and results from crushing process measurements of the level ice acting at the ship side presented. A project included in the INSROP (International Northern Sea Route Programme) developed a Compaction Factor which was also described.
• The third part includes discussion of the ice-ship interaction and results from estimations of ice cover strength parameters. The last part of the report ice dynamic model HELMI (HELsinki Multicategory Ice model) was presented. Ship channel closing model is also reviewed in the end of the report. The model is a preliminary one, which uses boundary conditions from HELMI. It is the first channel closing model ever done.

Linking to other SAFEWIN results
The state-of-art literature research about compressive ice field was conducted in D4-1. This literature research served as background information for all the work done within the project. Therefore it is in a sense linked to all the research conducted within the project.

1.3.2 D4-2 Description of initial downscaling

Ice dynamics models allow the expression of ice compression on a meso-scale grid both in hindcast and prognostic mode using a physical quantity, the internal stress in ice cover. This is an output parameter of dynamic ice models and is usually expressed as integrated over ice thickness and as averaged over the calculation grid of the model. Meso-scale ice stresses can’t be handled directly as ice compression on ship scale, which is how ships experience it in ice fields. Going further with dynamic modeling, in order to have ice compression on ship scale, the appropriate downscaling scheme is needed; however a good solution for such purpose, has not yet been found. Therefore the statistical downscaling approach was proposed.

Statistical downscaling (also called empirical downscaling) is a tool for downscaling the information from coarse spatial scales (geophysical or meso-scale) to finer scales. It may be applied as an alternative, or as a supplement, to dynamic downscaling (i.e. small scale - ship scale modeling). The underlying concept is that local physical parameters of ice are conditioned by meso-scale ice parameters and by local features (topography, distance to coast, etc.). At a specific location, therefore, links should exist between meso-scale and local ice conditions. Statistical downscaling consists of identifying empirical links between meso-scale patterns of ice elements (predictors) and local ice elements (the predictand). Common methods include linear methods such as canonical correlation analysis (CCA), singular value decomposition (SVD), and multiple linear regression analysis (MLR) and nonlinear techniques such as neural nets or fuzzy logic modeling.

From above proposed statistical methods we implemented the fuzzy modeling, which allow the combined use of both physical and non-physical parameters for analysis and ice compression estimates. In the fuzzy logic model approach, the predictor is called the affecting aspect and the predictand the affected aspect. Different meso-scale ice dynamics model outputs can be taken as affecting aspects in the fuzzy logic model as well as other parameters such as ship speed and heading, vicinity of boundary, water depth, ice floe size, channel closing speed, etc.

The advantage of using fuzzy logic is the possibilities more or less freely add the expert knowledge, both on ice conditions and ship performance in ice, into system on the same level as physical parameters. Even if the required exact relationships (links) are not fully established this fill the gap we have now trying to apply purely physical downscaling procedures.

Innovative nature
Meso-scale ice dynamics models give compression for a grid with scale not less than 1 nm, which is not sufficient resolution in order to evaluate forces acting on a ship hull. Therefore outputs of these models have to go through a downscaling procedure to yield compression at the ship scale. For this purpose, a conceptual model is proposed which takes into account (besides information at meso-scale grid points) sub-grid specific information and can incorporate information at the ship scale when it is available. The model is based on fuzzy logic modeling. The advantage of this approach and innovative point is that both the qualitative information and the information presented as physical quantities can be merged to give a qualitative estimate for compression hazard at a given location.

Linking to other SAFEWIN results
The statistical downscaling approach proposed (based on fuzzy logic model) is a tool for downscaling the information from coarse spatial scales (geophysical or meso-scale) to finer scales. Different meso-scale ice dynamics model outputs (e.g. HELMI) can be taken as affecting aspects in the fuzzy logic model as well as other parameters such as ship speed and heading, vicinity of boundary, water depth, ice floe size, channel closing speed, etc.

Implementation and exploitation
The above concept was the base to develop and implement the local scale fuzzy logic model to forecast the ice resistance of the ship voyage on certain fairway into the Pärnu port in the Gulf of Riga, the Baltic Sea.

1.3.3 D4-4 Model scale testing

The observation campaigns conducted within the project provided valuable information on the compressive situation; more specifically about ice dynamics and compression in the ice field. However, the campaigns didn’t provide the information about the additional ice loading and added resistance induced by the compression in the ice field for the ships. In order to gain a new knowledge on the added loading and resistance, model scale test series were conducted in the ice basin of Aalto University within the project.

The measurements showed that a static compression in the ice field does not increase the resistance nor the ice loads occurring on the ship model hull both at the bow shoulder and midship. The ice load measurements showed that ice loads on the model hull mainly occur at bow and bow shoulder area. The frequency and magnitude of the ice loads are significantly lower at midship when compared to the bow shoulder in level ice and static compressive cases.

Closing channel tests showed that dynamic compression has an impact on the resistance and on the frequency and magnitude of the ice loads occurring at the midship. The resistance is increasing as a function of the width of the channel in front of the model. After the channel had closed, the resistance stabilized. The situation could be considered as dynamic compression in level ice. The resistance was significantly higher in dynamic compression than in static compression or in level ice. The resistance increases in closing and closed channel when compression level increases from one to two, but remains at the same range in level two, three and four compression.

Comparison of the line loads in static and in dynamic compression showed great increase in frequency and in magnitude in dynamic situation. The measurement results indicate that the added resistance due to dynamic compression results from added line loads at midship area. The resistance increases when the magnitude and frequency of the line loads at midship increase. Furthermore, the magnitude and frequency of line loads at midship area depends on the speed of the ice sheet towards the model side.

Linking to other SAFEWIN results
The data obtained in the model scale tests were additionally used in two Master Thesis within the project when added resistance due to compression was modeled. First, Kaups’ modeled the added resistance due to compression using semi-empirical approach. In the approach, the added resistance is caused by the added contact area at the midship. The modeled added resistance represented the model test results fairly well. Külaots continued the Kaups’ work in his thesis. He discovered an additional component of resistance and presented the resistance in relatively simple formula. The developed formula for the total resistance gave good results when the results were compared to the model test results.

Innovative nature
New formulas were developed for estimating the total resistance of the ships in compressive ice field. The data can be used in the future to further improve the formulas.

Implementation and exploitation
The new formula developed for the total resistance could be implemented in the future into the trafficability models developed in the project. Furthermore, the formulka can be used when the ship’s performance are evaluated.

2.3.4 D4-5. Description of the operative forecast methods

Problem addressed
The general objective of WP4 was to study the ice compression as experienced by ice navigating ships. The ice forecast models describe the ice cover in a continuum approximation, which can be assumed to apply in scales where the true discrete nature of the ice cover does not show. Usually the safe scale limit L is assumed to be about ten times the characteristic size Lc of ice floes. One possibility to define Lc is to require that 50% from ice area is covered by floes smaller than Lc and the remaining 50% by floes larger than Lc . On the other hand, the ice model equations are solved numerically in discrete grid with a certain resolution. The model performance cannot be improved by increasing the resolution below L.
Ice forecast model HELMI has 1NM grid resolution and a compression forecasts gives one value for compression in a grid cell. On the other hand, a ship traversing the grid cell experiences variation in the compression so that route segments of low or nonexistent compression are followed by segments with high compression. The magnitude of the local compression peaks exceeds the grid cell value. If the added resistance for the ship progress can be linked with the ship scale compression magnitude, the effect of compression to ship traverse can be modeled if the variation of compression along the route in ship scale is at hand. The linking of the model scale compression with the ship scale compression variation is often called downscaling.
The linking of model and ship scales was one of the main objectives of WP4. Specifically, the objective was use this research develop such quantities, to be used in compression forecasts that would more directly relate to the actual ship performance. The deliverable D4-5 was intended to address these aspects while the actual exposition of forecast methods is given in extensive deliverables of WP5.

Compression forecasting was implemented in FMI as first as additional variables to previous forecasts. As the forecasts did not perform expectedly, the ice forecasting was then reimplemented as an ensemble approach. Several parallel runs with the ice model, using somewhat different parameters, are made and the results are expected in terms of probabilities. This was motivated by validation exercises that revealed the sensitivity of modeled compression fields to small parameter variations. Work related to the downscaling and subgridscale aspects were conducted in several WP:s. The ship compression observations were used to relate the statistics of experienced compression, to which affects subgridscale ice conditions and ship particulars, to the modeled compression. Field experiments on stress distribution were conducted and similar data from the Arctic analyzed. Using these data an extreme statistics approach of local ice stresses in space and time domain was constructed and applied to the analysis of stress propagation. However, this fundamental research did not produce results conclusive enough to be implemented to the operativce compression forecasts as a donwscaling module, which was the targeted final outcome.

Innovative nature
The connection between scales is the holy grail of ice cover dynamics and a notoriously difficult problem. A well-founded theory would provide deeper understanding of fundamental ice dynamics processes, method of assess the idealizations and rheologies of continuum ice models, and results applicable to ship performance. The advances attained point towards further progress possible with dedicated fundamental research. The potential applications are numerous, especially forecasting of ship speed and the possibility of functioning ice routing.

Link with other SAFEWIN-results:
The operative forecasting was a major objective of SAFEWIN and interacted with almost all other research. There were several different approaches to downscaling or subgridscale compression and ship research pursued in WP5 and other WP's, linked together by the common problematic, shared methodologies and common validation datasets. The HELMI hindcasts generated during the project were used extensively in ship performance assessments.

Implementation and exploitation:
The operative forecasting of compression using ensemble approach has been implemented in FMI. The research conducted on downscaling and subgridscale problematic, includind the exposition of different factors contributing to navigational difficulty, will be continued using the collected datasets and hindcasts.

1.4 WP5 Operative forecasting

1.4.1 D5-1 Presentation of the forecasting tool prototype

It was a major objective of the project to implement the modelling results of WP3 and other work packages into the operative practice. This concerns especially but not exclusively the compression forecasts. The operative production of the forecasts is one component here. The functioning delivery and end-used friendly presentation are as important. The 'tool' aspect also includes the idea that the forecasts could aid decision making, for example route selection, directly through suitable presentation, through additional operative products based on ice forecasts and supplementing them, or with an on board terminal with appropriate software and using environmental information from various sources. The tool aspect is mainly relevant for icebreakers and ships on regular route, especially ferries.

Operative compression information can also have other sources than model based forecasts. Real time information obtainable from AIS-retrieved ship performance data, from ship based instruments or from real time observations made by coastal radars can complement the modelling in important ways, especially when targeting areas with intense traffic and known to generate difficult compression, for example port entrances. Also, how a ship experiences compressive conditions depends also on factors peculiar to the ship (hull particulars, engine power, ice class), to the mode of navigation, and local ice conditions not captured by the 1 NM resolution of the forecast model. Various inference models using Bayesian inference or fuzzy logic are applicable here.

Two main lines of development were followed here. The FMI operative implementation relies on the enhancing of established information chain with compression information. On the other hand, the compression forecasting methods developed by TUT rely on real time observations and a fuzzy logic scheme that can assimilate data from different real time sources and knowledge bases.

The research and development results around HELMI ice model are implemented to the normal operative routine of FMI. The data delivery has two main channels: the forecast pages on FMI web site, and the IBNet information network for icebreakers which delivers the forecasts in a format viewable as layers in the on board terminal software IBPlott. There is also a commercial version of IBPlott, ViewIce, but has been in little use. The improving internet connections to ships on Baltic routes are also making ViewIce type solutions obsolete, as the information, including also ship specific solutions, can be accessed from the data provider's pages. The internet solution is also very flexible, as alternative prototype presentation methods can be used and end user comments collected easily. This work is ongoing and will continue as a part of normal development work of FMI operative services.

The existing operative structures are summarised and a wealth of possible solutions of using and further exploiting the compression information is presented in think-tank style. The presentation proceeds from the assumption that the forecasting in done in ensemble style where expected compression and probabilities to exceed given compression are the basic operative products. However, a large number of other choices are available. Different end user groups (icebreakers, VTS centres, merchant vessels, shipping companies, ferries, general audience) have specific needs that can be answered with moderate effort if the basic forecast production is up and running. Various parameters and parameter combinations are suggested, including derived parameters that more directly assess ship performance (navigational difficulty, risks). Here other solutions require the inclusion of other ice parameters (thickness, concentration). Several possibilities of presentation are given: animations, interactive maps, regional averages, graphs presenting time history, along route conditions. Tailored presentations are a possible solution for ferries and other ships with a fixed route and schedule; alternative presentation possibilities are given.

The system developed by TUT consists of two parts, first on-line measurement system based on understanding that different ice resistances in plane or ridged ice, presence of compressive ice etc. results with interaction between ice and ship hull causing ship hull vibrations of varying intensity - higher vibration intensity refers to more ship resistance in ice. Vibration of the ship hull is recorded in terms of 3D acceleration sensor which is tightly fixed to the ship hull. An instrument was built and installed on board Icebreaker EVA-316 (Estonian Maritime Administration) assisting ships and organising winter traffic in the Gulf of Riga and Pärnu Bay, the Baltic Sea. Measurement data were transferred in real time into FTP server of the Marine Systems Institute, Tallinn University of Technology, using GSM/GPRS protocol. From raw data the ship resistance index is calculated, based on vibration intensity of the ship hull, for each point of the voyage and special web based user interface built showing data in real time (

Comparison of acquired ship resistances in ice and on board observations on same ship, also satellite images showing ice conditions fit well with each other and severity of ship resistance in certain ice channels could be estimated. An attempt to relate the ship resistance in ice to relevant forcing parameters, wind direction first of all and also speed was successfully made. It was found that wind blowing perpendicular to ice channel in fast ice causes more resistance than wind parallel to channel, because of compressive forces in ice sheet. Ship hull vibration data showed adequately also ship manoeuvres in ice and relevant ship resistances. Recorded data of ship resistance in ice together with ship speed, course and ships momentary engine power data makes possible to assign a specific rank for the severity of ice conditions.

Second, system was complemented with forecast ability using fuzzy logic based relational schemes, using wind as forcing, ice conditions and knowledge based components as potential for ice resistance and compression. Ship resistance forecast system was realised during 2012/2013 winter in preoperational mode with open access web-based user interface. System present on-line data of ship resistance as well 48h ice resistance forecast forced by HIRLAM wind forecasts and information about ice conditions. The main area has been so far fairway into Pärnu Port, Gulf of Riga as operation area for IB EVA-316. Forecasts were validated with ship observations during icebreaking season and results allow to say that applied method of detection and forecast of ship resistance in ice channel is a useful tool for on-line monitoring the ice conditions along the ice channel, thus in situ where ships navigate in winter. Collected data during entire ice season allow analyse the navigation conditions and ship performance statistically afterward.

Innovative nature
In addition to the implementation of compression forecasts as an additional variable to the standard ice forecasts a large number of other possibilities were outlined in a think-tank style. These have a large innovation potential that can be exploited after the project completion. Some of the ideas are likely to be developed further in FMI but others are rather suitable for independent companies providing ship specific environmental data services. The ice forecast data will be made accessible following the FMI open data policy which creates the breeding ground for such activities. The results are also adaptable to other sea areas, especially Northern Sea Route and areas of active hydrocarbon exploitation or prospecting in the Arctic.

The TUT-developed multilayered on-line system reporting ship resistance in ice channels in real time (based on analyses of measured ship hull vibration data) and forecasting ship resistance over time span of 48h was built. The system integrates available data and expert knowledge operationally for certain ship routes and fairways. Fuzzy logic module provides to the system forecast ability of ship resistance in ice, on the same route over time span of 48h. In our knowledge this approach is innovative as there are no such systems reported earlier.

Linking to other SAFEWIN results
The readiness to operative forecasting of compression is one of the main objectives of the project. The FMI activities here use results from all other WP's and can also be utilised by all other WP's. This applies also to the future, as the established WP connections define lines of research that are likely to continue the work and exploit the results. For the TUT part of work, both measurement data and forecasts form an extensive dataset covering entire winter of ship resistances in ice in certain sea area with ship traffic, could be used for validation of ice compression forecast given by ice dynamics model developed in WP5.

Implementation and exploitation
FMI will implement operative forecasting as described in the next section for deliverable D5-2. This will comply to the standard styles of the FMI ice forecasts and IBPlott presentation layers. However, a large number of possible ways to utilise the forecasts has been outlined. Some of these may be seized by FMI but others provide market potential for SME's in environmental information business. There are especially promising potential opportunities in the Arctic, as forecasting hazards from difficult ice conditions is a bottleneck in navigation and hydrocarbon exploiting.

For TUT, the multilayered webbased user interface for forecast of ship resistance in ice was implemented in fairway into Pärnu Harbour, Gulf of Riga. Following layers were included: MODIS satellite data were used to give ice situation as background, disks of different colors showed graded measured ship resistance of IB EVA-316 (Estonian Maritime Administration), arrows expressed momentary wind speed and direction (violet measured and blue HIRLAM forecasted wind) and squares of different colors showed grading of ship resistance forecast hourly for 48h ahead (results of fuzzy model forecasts). Among other users of the system were merchant ships crews, Pärnu Harbour Master, other interested partners, total number of unique users of the system was 305 during 2012/2013 icebreaking season.

1.4.2 D5-2 Forecasting tool validation and customer feedback

The work addresses the operatively implemented methods of compression forecasting. For FMI this includes validation of the ensemble approach. One aspect of validation is to show that ensemble approach is superior to the ordinary deterministic forecasts. The basic validation is however against compression reports from the ship, as the forecasts should perform well enough to be useful for the end user and not become abandoned after first few trials. The compression reports contain reports collected during the project for development and validation purposes and presently reports accumulating trough the IBNet icebreaker information system to which new easy-to-use reporting functionalities were added. Another aspect is the comments of the end user on the usefulness of the forecasts and the user-friendliness of presentation methods. However, as the operative readiness with proper ensemble forecasting was attained only after the last project winter this kind of surveying was not conducted. On the other hand, the views of end users were taken into account in the presentation of the forecasts. Direct validation exercises were conducted for the alternative TUT methodology.

The FMI operative compression forecasts are produced in ensemble model runs. That is, several model runs are done using varying slightly the model parameters, initial values and forcing data. In supercomputing environment the ensemble members can be run parallelly and presently the computing resources allow 12 ensemble members.

That deterministic forecasting is not possible, or does not generate expected results, has been convincingly demonstrated. The compression fields of a deterministic run can change completely during a 6 minute timestep of the numerical solving. A counterpart to this is similar variation between ensemble members referring to the same time instant. The variation is due to the sensitivity of the compression fields to small changes and is in a part an artefact generated by the numerics. This does not prevent forecasting as the expectations and probabilities calculated from the ensemble are well behaving. In addition, the compression forecasts are time averaged over 3 hours, which generates forecasts that vary smoothly in time and space and are comparable to other model variables in this respect.

Several direct validation exercises with ship observations were also conducted in the project. These employed the hindcasted deterministic compression fields where the variation due to sensitivity was reduced trough regional averaging. According to the results also the deterministic forecasts can be used if the tolerance with respect to magnitude variation and localisation is generous.

The operative forecasts use compression numeral ranging from 0 to 4. Scalings between model units and numerals were studied, but it was apparent that the relationship must be statistical. There are also other factors affecting the interpretation of compressive situation by the ships, the ship particulars and engine power, operation mode, local ice conditions not resolved by the model. A method for direct conversion of model units to numeral using the observations was given.

TUT tested the tool developed during the SAFEWIN project, for on-line estimate of ice compression through ship resistance in ice. Measurement system (measuring ship hull vibration) was installed on board the icebreaker EVA316 (Estonian Maritime Administration) and on-line user interface was created as webpage showing both measured data and forecast of ship resistance in ice. Results available on-line via webpage, with 1 minute interval, were validated against visual observations on board the ship navigating along fairway into Pärnu Harbour, Gulf of Riga. System was complemented with forecast ability based on fuzzy logic relational scheme. The model output was the 48 h forecast of ice compression along the fairway relying on the HIRLAM wind forecast (48 hours) as the dynamic input and ice conditions from satellite images (MODIS, RADARSAT, etc.) as well from in-situ observations. Web-page was updated with ability to overlay both measurements and forecast with actual ice conditions from satellite images as well measured and modelled wind properties.

The modeled ice stresses were evaluated with the observations collected during SAFEWIN as well as with the help of ship speed observations originating from the Automatic Identification System for ships (AIS). Extremely helpful for model evaluation were additionally the numerous compression reports of voluntary ships.

The concept of ensemble forecasting into ice forecasting and in particular to forecast ice stress was introduced by FMI
• To focus on compression in an ice forecast and to investigate compressive situations in detail was a novel approach
• Regions with high risk for compression were identified
• To compare ice forecasts with ship speeds obtained by the AIS-system was novel as well and provided the opportunity to evaluate ice forecast specific to the needs of winter navigation.

The comparison with the AIS-data revealed that the ice stresses nicely summarize situations that impact the time schedule of ships and provides a bulk of the information that requires the analysis of several factors otherwise. Ensemble forecast include some measure of uncertainty, which is very relevant information for the end-users. The AIS-data comparison indicates that the new product might facilitate routing considerably.

Innovative nature
Compression forecasting is pioneered by the project and ensemble forecasting of ice conditions is not conducted elsewhere. The research, modelling exercises and validation studies have revealed the basic restrictions compression forecasting must meet. The results are expected to have high impact both in research and operative fields.

The system developed by TUT of on-line estimate and forecast of ship resistance in ice is innovative as in our knowledge no such exist elsewhere. Validation of measurements and forecast showed that there exists a general agreement on what is measured and observed on board of the ship. During the test period TUT tuned fuzzy logic forecast module in the system what improved the performance of the system considerably.

Linking to other SAFEWIN results
Operative forecasting of compressive ice conditions is a focus objective of the project and takes input from all other WP's. The results that emerged during the project and that will be accumulating in the operative practice constitute basic data for research conducted on ship ice resistance and winter navigation.

For TUT solution, both measurement data and forecasts form extensive dataset covering entire winter of ship resistances in ice in certain sea area with ship traffic, could be used for validation of ice compression forecast given by ice dynamics model developed in WP5 and further used in WP6 in risk assessment and risk control of ships navigating in ice.

Implementation and exploitation
The ensemble forecasting has been implemented for operative use in FMI. The methods are expected to become adopted by other ice information providers. Scientifically, research lines with high potential impact have been opened. Great potential is seen in the Arctic areas.

The TUT on-line system for on-line ship resistance estimate and it’s forecast is realised as open access webpage what allowed access of wide range of users which beside of usage statistics also contributed with remarks and recommendations about functionality and performance of the system.

1.5 WP6 Risk Control

1.5.1 D6-1Probabilities for ice movement for risk assessment

Probabilities for ice movements and occurrence of compressive ice, risk assessment of ships navigating in ice.

TUT performed statistical analysis of ship resistance data (with time interval 1 minute, covering almost 6 months of ice season) recorded on the board of icebreaker EVA316 in the Gulf of Riga with aim to have data for fuzzy logic relational scheme validation and calibration. In parallel the forcing data obtained from nearby meteorological stations and HIRLAM atmosphere model were analysed. According to the collected material the risk control could be as hazard identification by means of on-line warning created by fuzzy logic relational scheme which considers both the potential for hazard and latter triggering function(s).

Innovative nature
Valuation of hazards for winter ship traffic was analysed and forecast method implemented in terms of fuzzy logic model which considers in calibration phase both the measured physical parameters (like ice properties, ship resistance in ice etc.) as well ship captain’s expert knowledge about occurrence of ice compression and ship performance in various ice conditions. This last is innovative approach in order to increase the safety of navigation in compressive ice, as nature and occurrence of compressive ice is not well known yet, especially in ship scale.

Linking to other SAFEWIN results
The analysis of ship resistance in ice data could be used in the validation of the FMI's HELMI model as well other ice dynamic models, capable to create good general picture on risk of ice compression in entire sea area.

Implementation and exploitation
System was tested in collaboration with the crew of the icebreaker EVA316 and Pärnu Port administration, they found this information useful for winter navigation risk control purposes locally on certain fairway, in Pärnu bay this case.

1.5.2 D6-2 Monte Carlo simulation for determining risk in compressive ice

Background and results
The methodic to estimate exploitation reliability and risk of navigation in the ice (with and without icebreaker assistance) of modern and perspective vessels has been developed in the AARI. This methodic was developed as the computer model of estimation the risk of navigation. The model is based on the Monte Carlo method, allowed to determine permanent amount of accident situations during the significant time period. Parameters for each simulated route were determined with the help of random number generator and statistical distribution laws of these parameters.

Statistical distribution laws are based on results of field research and theoretical concepts. This method was tested and applied for ice conditions in the south-eastern Barents Sea (Pechora Sea). The data of navigation the vessels of ice class Arc4 (in Russian Maritime Register of Shipping, 2007) or IA (in Finnish-Swedish Ice Class Rules) in very difficult ice conditions in the Pechora Sea were used. The data were obtained with the help of special shipborne observations and air reconnaissance in April (1980-1989 and 1994-2002). As a result, probability of vessel damage in the level compressed ice is 0.33 but in the hummocked compressed ice it is increasing up to 0.76.

Because the Gulf of Finland is covered by close ice in winter: new, young, thin, medium, and thick first-year ice, we can easily apply the model for estimation of risk in compressive ice there. However we need to adjust the method to the Baltic Sea ice conditions and certain type of vessels, using in the Baltic Sea.

For preparing of the model external parameters the following information is needed: expedition and remote sensing data about standard navigation routes within close ice, length of the compressed ice zones, sea ice thickness, hummock concentration, amount and length of ridges, the ridges consolidated layers, relation between the ridge width and its height, probability of ice compression. These parameters must be determined for each week or decade during period of ice cover existence.

1.5.3 D6-3 Analysis of risk control options

In this deliverable risk control options for maritime traffic system operating in the Baltic Sea during wintertime are proposed. Special attention is put to the system operating in dynamic ice. This type of ice, in certain circumstances, may lead to besetting of ships followed by possible hull damage if the conditions remain for a longer period. Therefore the description of risk associated with shipping in dynamic ice should comprise the probability for a ship to get stuck and consequences of this event. In order to control the risk, those two components should be addressed.
In this report we present two ways of mitigating the risk associated with navigation in dynamic ice field. First, we demonstrate the use of the probabilistic rules and fuzzy-logic to determine the probability for a ship being beset, given a set of ice conditions. Second, we apply nonlinear FEM to obtain numerical model estimating ship resistance and ship speed in compressive ice field. Utilizing the same method – FEM – it is possible to develop a numerical model, estimating forces, which are exerted by compressive ice field on ship plating, and to estimate the resulting volume of hull deformation.

This deliverable provides risk control options for maritime traffic system operating in the Baltic Sea during wintertime. The presented ways of mitigation risk associated with navigation in dynamic ice can be summarized as follows:
1. probabilistic models and fuzzy logic based model which ultimately lead to the reduction of the probability for a ship to get beset in ice field;
2. a numerical model to evaluate ship resistance and ultimately ship speed in compressive ice,
3. a numerical model to estimate an extent and the volume of damage (hull deformation) that ship may experience as a result of ship-dynamic ice interactions.

The results, which are obtained with the use of the above methods, are validated with the available data from the model tests and field measurement and good agreement is found for the considered ship and the geographical area.

Innovative nature
The probabilistic models estimating ship performance in ice feature several novelties, first they predict the ship performance in a probabilistic fashion, with the use of full scale data, second the models consider the joint effect of various ice features on ship performance, finally the ice compression which is known to have significant negative correlation with ship’s speed has been taken into account.

The fuzzy logic model provides information about ice compression and ice resistance, using five-grade scoring system. The model operates both, in hindcast mode, giving ship scale ice compression hazard based on source information in mesoscale, as well as in forecast mode relying on forecast of atmospheric model.

The FEM based model determining ship resistance in ice and ice loads, is a novel way to look at the problem of ship-ice interaction and modeling. The material modeling is a simplified derivation from a novel approach developed in AALTO and NTNU Norway, to simulate the behavior of ice produced in the ice tank of Aalto University, where the model test were conducted. The simulation results are found to comply reasonably well with the experimental measurements, even though the level ice resistance is estimated conservatively.

Linking to other SAFEWIN results
The probabilistic models describing ship performance in ice have been developed with the use of HELMI model and its ability to forecast ice compression. The later is a development done in the course of SAFEWIN, described in D4-5, D5-1 and D5-2.

Fuzzy-logic model for determining ice compression is linked with other SAFEWIN development, modelling and forecasting ice compression with HELMI model, using ensemble approach for user products broadcasted to users through IBNET environment installed on icebreakers. Estimates of ship resistance in ice form valuable database covering winter navigation routes dense in time and space and serve well for validation and further developments of HELMI - IBNET ice compression forecast system, developed in SAFEWIN project.

The material modeling is a simplified derivation from a novel approach developed in AALTO and NTNU Norway, to simulate the behavior of ice produced in the ice tank of Aalto University. The results of model test conducted under WP4, as reported in D4-4, have been utilized to develop the numerical model presented here.

Implementation and exploitation
The probabilistic models estimating ship performance in ice can be used to determine the probability that a ship will attain a certain speed class or to specify the areas where she may get stuck in ice. This is especially important in the case where assistance of an icebreaker is not available immediately. Thereby, we expect this new approach to facilitate the optimal route selection problem for ice-infested waters where the ship performance is reflected by an objective function.

Expert system for monitoring and forecast of ship resistance in ice was implemented in fairway into Pärnu Port, Gulf of Riga, the Baltic Sea - -
System was tested in pre-operational mode with input from HIRLAM wind forecasts, but also with different information about ice condititions collected from satellite images and direct in situ observations by ship crews. We can conclude that the ship hull vibration measurements applied for detection of the ship resistance in ice channel is a useful tool for local on-line monitoring of ship resistance in ice and forecast of latter using fuzzy logic model works well for optimization of winter traffic along the fairways locally, making it more seamless. First responces from users point out also fuel economy up to 30-50%, which connected with possibility to avoid navigation in high compressive ice channels.

The simulation results of ship performance in ice and estimation of potential damage caused by compressive ice, using FEM, were found to comply reasonably well with the experimental measurements. These analyzes enable FEM modeling of previously mentioned experiments and thus allow large savings in time and money.

1.6 WP7 Ice Management

1.6.1 D7-1 Reporting uncertainties in ice forecasting

Background and content
The present sea ice forecasting products are based on single simulation of near future evolution of pack ice. This kind of forecast are called as a deterministic forecast. In general, forecasts are rather skillful to estimate changes of large scale pattern of sea ice drift, opening or closing of coastal flaw leads and compression regions. However, in a local scale, the forecast could differ rather much from the reality, in particular in timing of sudden changes.

Uncertainties of the forecast are due to the several factors.
• The numerical model used
o Physical processes which can't be explicitly resolved by the model are parameterized and usually methods of parameterization are based on limited field data.
o The numeric of the model introduces uncertainties.
• The initial sea ice conditions of the model simulations
• An actual meteorological forecast.

Understanding of the uncertainties is essential for all users utilizing the forecasts. This is particular important in a situation when course of development is very variable or forecast indicates occurrence of some extreme event. Uncertainty could also vary between the model parameters, but presently model products don't include any kind of indicator of forecast quality.

In a weather prediction, ensemble prediction systems (EPS) has become a standard method for a provision the probabilistic information, but oceanic EPS's are still under development. Ensemble forecasts system have been developed for an estimations of wind wave statistics, surface drift calculations and algae bloom predictions.

The deliverable D7-1 Reporting uncertainties in ice forecasting summarized sources of uncertainties (listed above) and present a road-map to include probabilistic information for sea ice forecasting products. A plan of ensemble sea ice forecasting system was presented in addition to the application to the ice management.

Innovative nature
As far as the writers know, the only EPS providing sea ice parameters is the SMHI PolarView prediction system ( and the MeteoFrance's seasonal Arctic Sea ice prediction system. Within the SAFEWIN project the EPS system was implemented in an operational ice forecasting model (FMI’s HELMI model), which is the first time that the EPS system has been implemented for ice forecasting model in the Baltic Sea.

Linking to other SAFEWIN results
The research conducted within D7-1 is directly linked to the work conducted in WP3 and WP5 as it discusses about the uncertainties of the forecasting models studied and developed within WP3 and WP5. Furthermore, the EPS system presented in D7-1 was implemented in to the operative forecasting model within WP5.

1.6.2 D7-2 Quality of ice forecasts for ice dynamics forecasting

Description of the results
The general considerations and approaches on ice pressure forecast assessment show that theoretically the modeled and observed values can be compared using the traditional statistical criteria like skill score or correlation coefficient. Correct comparisons require the modeled and observed data to be of the same scales and units.

Several examples of direct comparison between modeled (AARI model) and visually observed data demonstrated that, in general, the model gives the adequate qualitative picture of ice pressure, however, the quantitative correlation is not high (correlation coefficients are about 0.4-0.5). Probably, it means that either the model overestimates the pressure, i.e. the model contains a systematic error, or the observers underestimate the pressure.

More comprehensive study of correlation between modeled (FMI model) and visually observed data showed that the FMI model was consistent with the observations in the majority of cases (88%), i.e. the FMI model would need only a slight tweak. Most of the observed compressions were larger than the modeled values.

However, even if modeled (forecasted) ice pressure correlates well with direct visual observations or instrumental measurements, it is not sufficient for reliable quantitative assessment of forecast quality due to differences in scales, areas, units, temporal regularity.

Visual ship-borne observations should be predominantly used for general qualitative verification of the model, special instrumental measurements – for better “scientific” understanding of natural processes.

The alternative way to assess regular “everyday” ice dynamic forecasts is to use the actual fields of ice drift reconstructed from satellite (coastal radar) images, drifting buoys, etc.

If actual ice drift field is reconstructed, ice pressure field can be obtained with the help of strain rate tensor method. The reconstructed fields of ice pressure and drift should satisfy the general properties of ice dynamics. Such correspondence can be considered as a kind of qualitative criterion of adequateness of reconstructed fields.

Innovative nature
First, now, at the end of the project, one can judge with sure how adequate the ice pressure forecasts obtained from the FMI model which is the basic SAFEWIN prognostic tool. Now we know that the FMI model gives quantitatively adequate picture of expected ice pressure. This is confirmed by the detailed analysis of the FMI ice pressure forecasts vs the great amount of natural observations. We know that the FMI model needs only slight correction in some specific regions of the Baltic Sea.

Second, now we have the general understanding how to compare the visual observations and model results taking into account the differences in scales and units.

Third, now we understand in which direction the methodology of ice pressure forecasts evaluation may develop, namely: to reconstruct the actual ice drift fields (using the radar data) and, then, to reconstruct the actual ice pressure field (using the strain rate tensor method). This will allow obtaining the ice pressure field free of subjective factors typical for visual observations.

Link to other SAFEWIN results
The evaluation of ice pressure forecast has sense only if we have forecast (i.e. the dynamic model and corresponding technological realization). So, this is the link of WP7 with WP3 (Forecasting method) and with WP5 (Operative forecasting of ice compression).

From the other hand, ice pressure forecast can be assessed only if we have the actual picture of ice compression. It requires the vast flux of natural observations, and, hence, relatively simple and clear method to observe and report the ice pressure events. This is the link of WP7 with WP2 (Observation campaigns).

Implementation and exploitation
The Russian experience of hydrometeorological support of the Arctic shipping shows that the total expenses on ice information (satellite images, ice charts, ice forecasts, routing recommendations, telecommunications, overheads, taxes, etc.) per year are close to expenses on atomic icebreaker exploitation per 3-4 days. It means that usage of ice information for ice shipping support is profitable in any case. Undoubtedly, in the Baltic region one has the similar picture: though ice conditions in the Baltic are not so severe, but the navigation intensity makes ice information support extremely important.

1.6.3 D7-3 Icebreaker deployment philosophies for ice management

The approach of deliverable D7-3, Recommendations regarding icebreaker philosophies for a set of ice conditions, was to collect data about icebreaker operations in the Baltic Sea in compressive ice and the views of icebreaker masters on compressive ice situations: forecasting, operation and reporting. For the report Finnish, Swedish and Estonian icebreaker captains were interviewed and also observations were made onboard icebreakers. With the interviews and observations the methods of forecasting compression onboard were collected.

Forecasting compressive situations is crucial for the smooth assistance of merchant vessels. Icebreaker captains usually perform the forecasting according to their professional knowledge of ice and ice movements, with the help from satellite pictures and wind forecasts provided. Icebreakers have distinct operational models for compressive situations. Overall operations take into consideration the routing and icebreaker placement for the duration of a few days. Assistance is routed according to the forecasting to avoid compressive situations. In cases of individual vessels, the assistance of the icebreaker aims to the relieving of pressure, most often circling the stuck vessel or cutting it from the leeward side. Also travel directions are set against wind to avoid compression. Towing is a commonly used when compression is present.

The icebreakers report compressive situations via IBNet. The assessment of the compression is made by observing the situation according to the assessment criteria that is based on the Russian system for the assessment of ice compression severity. It became evident with the research that icebreaker crews and masters have intimate knowledge of local ice dynamics and ice behavior. They also are extremely proficient in forecasting the changing ice dynamics on their own, with wind forecasts and satellite images. Still, new forecasts and forecasting tools were welcome to the icebreaker captains because the more they have tools to work with, the more reliable their decisions can be.

The icebreaker crews/masters were positive in their attitudes towards new ice compression forecasting tools before the implementation. New tools are not seen as a threat to the professional pride of the job, but rather the crews are welcoming all new support tools for their practice. The profession of icebreaking is demanding and the forecasting of the changing ice situations can be difficult. Thus the icebreaker crews appreciate all new ways to gain information and forecasts to help their work. The icebreaker captains have accumulated their knowledge of ice dynamics with years of icebreaking experience and do not feel threatened by new additions. In cases of this forecasting tool and others, the decisionmaking still lays in the hands of the crew and they can use the tools just as supporting mechanism for their decisions.

This approach to ice compression was important to the overall goals of the project. There had been not much research done from the direction of icebreaker operations in especially compressive situations. These results can be used along with for example the analysis of ice compression cases reported by icebreakers. There is now more knowledge on the reporting of ice compression and this can be adapted with the compression case analysis. Also, the thoughts of icebreaker masters were used in the development of the forecasting tool. It is clear that ice compression as a physical phenomena has to be linked with the actual operation of vessels and icebreakers.

Potential Impact:

In this chapter, the main technology implementations during the project are presented.

1.1 Operative ice forecasting tool in IBNet
Figure 1 shows a schematic presentation of the main implementation of the operative forecasting tool, basing on the enhanced FMI operative ice model. The developments made during the project are presented with red font. The knowledge on compressive ice gained and research during the project have enabled to include the compression of ice to the forecasting models and to the operative forecasting tool. Compression forecast has been implemented to the IBNet system used by Estonian, Finnish and Swedish icebreakers. An essential part of the system is the permanent feedback loop created by the IBNet ice condition reports, as they make possible the continuous improving of the forecasts using amassed ship observations.

1.2 Local forecasting system for the Pärnu Bay
In addition to the IBNet implementation a local scale system was designed and implemented in the Pärnu Bay (the Gulf of Riga) to on-line monitor the ship resistance in ice and forecast conditions on winter fairways. The system consists of two parts, first on-line measurement system for recording and transmitting the ship resistances in ice and secondly, forecast system for ship resistance in ice. Acquired ship resistances in ice were compared with on-board observations on the same ship, also satellite images showing ice conditions fit well with each other and the severity of ship resistance in certain ice channels could be estimated.

Considering the presented concept of ice compression in a ship scale and observational campaign data (the Gulf of Riga) the fuzzy logic model for the specified ship route (fairway to Pärnu Harbour) was built up and implemented in winter 2012/2013. The implemented fuzzy logic model considers both the potential (ice thickness, ice compactness and ridged ice amount) and triggering measures (wind forcing in relation to the ship course) and forecasts the ice resistance for a given location. Free access web-based user interface was used to disseminate the results via The system was launched in pre-operational in 5th of February 2013, with daily forecasts of ship resistance on the fairway into Pärnu Harbour.

Dataset presented on the website includes an estimate of ship resistance in ice (5 grades), to date forecast covering fairway into Pärnu Harbour, as well forcing data, both measured and forecasted winds. Ice conditions are given by overlay of Pärnu Bay and Gulf of Riga with to date MODIS satellite image. The captain of EVA-316 was contacted by phone practically daily and validation information obtained such a way, as well first user reactions about the forecast system. For further finer calibration and validation of the model the ship hull vibration data collected during entire ice breaking season 2012/2013 navigation period would be used. The users of the system consisted of merchant ships crews, Pärnu Harbour Master, and other interested partners. The total number of unique users of the system was 305 during 2012/2013 icebreaking season.


In this chapter, the plans for possible technology implementations after the project are presented.

2.1 A risk model for vessels
The performance of a specific ship in compressive ice field can be evaluated with the model developed during the project. Using the performance analysis together with the forecasting tool, it is possible to estimate the probability of the ship having speed reduction or getting stuck in a specific route. This has not been implemented during the project, but the research challenges associated with this subject have been identified, and it is possible in the future to improve the estimation model and to proceed with the implementation of the evaluation in practice.

2.2 Ship and route specific forecasts
The compression forecasts provide opportunities of different kinds of tailored presentations and ship specific services. The routine forecasts are perhaps not optimal for ice navigators as these require projecting of the compression fields along the future route of the ships. For ships with regular route and schedule forecasts can be presented in terms of along route expected conditions. Such products could be added to the paid services offered to mariners by FMI.

2.3 Arctic model implementation
FMI has plans to implement ice-ocean forecast model to the Arctic, serving as a cornerstone of Arctic service providing. This will implement the SAFEWIN results and adopt ensemble modelling. User friendly, route and ship performance oriented presentations will be developed and specific need of drilling rigs addressed.

2.4 The operative ice information in Russia
At present in the Russian Arctic the ice information is disseminated among the ships and maritime authorities in a following way. The main manufacturer of the operative ice information (ice charts, ice forecasts, routing recommendations) in Russia is Arctic and Antarctic Research Institute (AARI). The main customers of ice information are the merchant ships (tankers, gas-carriers, dry cargo ships), sea oil-gas loading terminals, administrations of private shipping companies and Administration of the North Sea Route which is a department within the Ministry of Transport. The icebreaker captains (including the nuclear icebreakers) usually prefer raw satellite images because they are sure in wide possibilities of their icebreakers.

The AARI produces the ice information in several possible formats: texts, tables, simple raster graphic files, shape-files in accordance with SIGRID-3 standard, files in special formats compatible with special software (so-called “End User Terminals”, EUT) intended to visualize the ice information (both actual and prognostic) as independent layers along with other navigation information. The EUT files are based either on the SIGRID-3 standard or on the S-57 standard.

The EUT software – the most advanced and modern variant – is disseminated on the commercial base, that is why not all customers have it. So, the choice of the format depends on what software installed on the customer’s computer. In recent time the quantity of customers preferring to receive ice information regularly in modern formats is widening, but the process is going not so quickly.

The most widely used ways of communication are E-mail or various Internet variants (e.g. FTP protocol). The choice of communication method also depends on technical, financial and professional possibilities of the customer. Major part of ice information is disseminated on the commercial base, and only the generalized ice charts are in free Internet access.

Some customers buy ice information regularly, but some ones – episodically, some customers buy full set of information (image + actual ice chart + prognostic ice charts + routing recommendations), but some ones ignore forecasts and recommendations.
Some ships use the icebreaker escorting, but some ships navigate without it (most probably, due to financial reasons).

So, as may be seen, the actually existing situation looks like the chaotic fragments of regular ice information dissemination system. That is why it seems doubtful that the forecasting system developed in the framework of SAFEWIN project can be transferred mechanically to the Russian Arctic within nearest time, because, on the one hand, some embryos of such system already exist, and, on the other hand, existing official laws and non-official traditions are not ready for uniform systematic dissemination of ice (and speaking wider – hydrometeorological) information.

But, looking on this mosaic picture and assessing the trend of its development, and taking into account the expected development of the Arctic hydrocarbon fields, one can imagine that in not very far future the ice information support of the Arctic shipping would be more and more systematic. Possibly, it would be more or less similar to the system developed under SAFEWIN project. In such case the experience accumulated in the Baltic countries, including the SAFEWIN achievements will be very important and useful.


The mutual understanding of the phenomena among scientific community and operational personnel has made a huge leap forward. This enables the future research use SAFEWIN as a state-of-the-art. Tools to report, predict and validate predictions in a more systematic way have been developed. Testing and modelling in both micro and macro scale of the phenomena has made possible to further develop our understanding the true nature of it. The international community is expected to seize the methods and develop them further.

Three observation campaigns have provided a new knowledge on compression in the ice field and have been used in improving and validating the forecasting tools. The measurement results can further be used in the future for validating and improving the forecasting tools and for the ship design. Together with the downscaling approaches, the observation campaign results are and can be implemented in the creation of more detailed operational ice forecasting methods that can provide valuable navigational information for ships navigating in ice.

The model tests provided quality data about the added difficulties by compressive ice on the navigation of ships in compressive ice. The data has the application in providing more information about how compressive ice can affect a ship: how it can add the resistance and cause damage to the ship hull. This is especially important, as the effect of ice compression is not taken into consideration in the traditional models of calculating ship resistance in ice and ice-hull interaction. The collected data enabled to develop models, being able to quantify the effect that compressive ice has on winter navigation, both in the form of economical inconveniences caused by delays due to compressive ice and safety aspect in the form of hull damage.

The gained knowledge becomes even more important in the future as we see that due to environmental regulations (EEDI, Energy Efficiency Design Index) future merchant vessels will be less capable to independently proceed in ice, especially during compressive situations. The data and methods generated during the project can be used to predict this impact to the whole winter navigation system and for instance required icebreaking capacity, ice-warnings and ice-restrictions. Numerical data allows further development of holistic winter traffic models.


The results obtained in the course of the project have been successfully implemented into the winter navigation system, which is operational in the Baltic Sea. The new developments, which were implemented, received very good feedback from the end users (IB crew and Maritime Authorities).
Moreover, new areas for the exploitation of the results and their development have been identified.

List of Websites:

Pentti KUJALA (Prof.)
Jakub Montewka (PhD)

AALTO University
Department of Applied Mechanics
Espoo, Finland