The present invention relates to a method for controlling the ice conditions for a certain fixed operation point at sea where ice occurs. More specifically, the invention relates to a method for forecasting future ice drift in relation to such operation point.

During industrial activity at sea, such as deep ocean drill- ing from a drilling vessel, close attention must be paid to ice conditions at the operation point for the activity. Since the drill at the sea bed is fixed to the vessel at the surface, too high ice pressure on the vessel, or a collision with a major ice floe, may cause damage to equipment or, in worst case, personal injury or environmental damage.

Different ice conditions may constitute differently large risks during such and similar activities at sea, and since the ice usually drifts, the conditions may change relatively quickly. Therefore, icebreaking vessels are conventionally used for breaking approaching ice to manageably small ice floes before the ice drifts up to the operation point. What areas that are to be prioritized for breaking, at each point in time, by the available icebreaking fleet, with the purpose of minimizing the risk for unmanageably large ice floes at the operation point, is determined based upon actual and forecast values regarding the direction and velocity of the ice drift. However, the available information is in many such areas at sea insufficient. Statistical data regarding sea currents, tidal water, wind strength and direction, and so on, is combined with available satellite images of the ice conditions as well as weather forecasts and earlier experience in order to arrive at an assessment of the ice situation. Since the ice drift velocity in general is such that the ice drift during the next 48 hours or less is decisive for which ice that will finally reach the operation point, major uncertainties are typically present in conventional ice drift forecasts. As a consequence, larger areas than necessary have to be managed (that is, broken) by the icebreaking fleet in order to achieve acceptably low risk levels at the operation point, which in turn increases the need for equipment and personnel, leading to both higher cost and more severe environmental impact. For example, it is often difficult to foresee ice drift directional changes, especially at times when the ice drift velocity decreases before a coming directional change. At such occasions, it has conventionally been necessary to extend the ice managed area around the operation point using large margins.

Hence, it would be desirable to achieve a method for controlling the ice conditions at the operation point, requiring only smaller areas being ice managed during operation in order to achieve acceptably low risk levels at the operation point .

The present invention solves the above described problems.

Thus, the invention relates to a method for producing a forecast of future directions and velocities for ice drift in relation to a fixed operation point at sea, and is characte ^¬ rised in that a computer system is caused to calculate the ice drift forecast using an ice drift model according to which an ice drift path is calculated based upon the value of at least one first physical parameter selected from the wind catch factor; the angular difference between the wind direc ^¬ tion and the wind-induced drift component of the ice; local rotational motion of the ice caused by the motion dynamics of the ice itself; direction and velocity of surface water currents; and/or local rotational motion of surface water currents, the value of which first parameter affects the veloci- ty and direction over time of the ice drift; and the value of at least one second physical parameter selected from wind strength and direction; air pressure; tidal water; latitude; and/or direction and velocity of surface water currents, the value of which second parameter also affects the velocity and direction over time of the ice drift, in that the actual local ice drift is measured during a historical time period preceding the calculation of the forecast, in that the computer system is caused to determine a certain optimal value for the first parameter, which optimal value maximizes the conformance between the forecast and the actual ice drift during the historical time period when the forecast is calcu ^¬ lated based upon the said optimal value for the first parameter and upon a known value for the said second parameter, and in that the computer system is caused to use the said optimal value of the first parameter when calculating the forecast ice drift.

In the following, the invention will be described in detail, with reference to exemplifying embodiments of the invention and to the appended drawings, where:

Figure 1 is a first flow chart of a method according to the present invention;

Figure 2 a second flow chart of a method according to the present invention;

Figure 3 is a time chart illustrating a method according to the present invention;

Figure 4 is an overview diagram of an ice managed area at sea; Figure 5 is a schematic overview of a system for performing a method according to the present invention;

Figures 6a-6g are graphs showing actual and forecast ice drift during a certain time period for different parameter values; and

Figure 7 is a graph showing an actual ice drift path.

Figure 4 illustrates a typical situation during industrial operation at sea. A drilling vessel 10 is located at an oper ^¬ ation point which is ice managed. That the operation point is "ice managed" is to be interpreted so that the water at the operation point is controlled regarding its ice conditions so that the industrial operations at the operation point are not threatened by ice floes at or near the operation point. The ice managed area is delimited by the line 3, outside of which the water is at least partly covered with ice and inside of which the ice cover has been managed to successively smaller maximum floe sizes the further said floes are distanced from the operation point upstream in ice drift direction. The shape of the line 3 is determined by the historic ice drift and the past ice managing activity.

It is to be understood that instead of a point, a "fixed operation point" according to the invention may be comprised in a possibly curved operation line, such as for example along a pipeline on the sea bed. Along such a line there may be any number of fixed work points. Furthermore, operations may similarly be related to a set of fixed lines, such as for seismic measurements using dragging hydrophore equipment along predetermined parallel lines across the ice covered water. In the presentation herein, what is said in relation to an operation point in the form of a point is analogously applicable to an operation line or lines, where a certain area around the operation line or lines needs to be ice ma- naged with the same purpose of lowering the risk of a fatal collision with an ice floe or the like.

The predicted or forecast ice drift is shown using a line 1, and an area 2 (hatched) to be properly ice managed extends from the operation point, covering the ice path which according to a current ice drift forecast will later pass through the operation point due to ice drift. Since the ice drift forecast is associated with some uncertainty, the width 4 of the area 2, perpendicularly to the forecast ice path, is larger the longer the time before the respective point along the ice path will pass the operation point according to the forecast . Two icebreakers 20, 30 work together to manage the incoming ice in order to guarantee safe operations at the vessel 10 position or operation point. Two measurement stations 40, 50 are fixed to and drift along with the ice, and are arranged to measure their respective direction and velocity over time. Such stations may be in the form of ice buoys or the like, and are deployed using icebreaking vessels 20, 30, a helicopter or similarly.

Figure 5 is a high-level diagram showing the system setup, using the same reference numerals as in figure 1. Onboard the drilling vessel 10, there is a computer system 11 arranged to gather the data, perform forecast calculations etc., as described below. The computer system 11 is also arranged to display the forecasts produced by the method according to the present invention to a user and/or to feed the forecast data into the existing ice management planning and operation sys ^¬ tem installed on the drilling vessel or otherwise. The icebreakers 20, 30, as well as the measurement stations 40, 50, are all connected to the computer system 11 on the drilling vessel 10, whereby for example location information from the buoys 40, 50 may continuously be sent to the comput- er system 11. Furthermore, an external source 60 of information, such as a forecast supplier, is connected to the computer system 11. Another external source 70 of information, such as a satellite image supplier, is likewise connected to the computer system 11.

A measurement station 13 is arranged to continuously measure the current wind vector and other locally measurable data, such as air pressure and geographical location. A database 12 is arranged to store available data, such as delivered fore- casts and imagery, locally measured data, etc. The database 12 may be standalone or for example incorporated as a functional part of the computer system 11.

Communications between external suppliers 60, 70 and the computer system 11, but also between the computer system 11 and the operation internal components 20, 30, 40, 50, may in practice be facilitated via a wireless communication system 13, which may be conventional as such. It is also realized that the computer system 11 may also be installed at a loca- tion which is not on the drilling vessel 10.

Vessels 20, 30 also feature one respective computer 21, 22 each, with a respective information screen. The present invention is based upon the calculation of geographic ice drift forecast paths using a predetermined ice drift model, in which varying the values of certain predetermined input parameters yields different respective specific forecasts of the future ice drift path. As seen in figure 1, in a step 101, the method is commenced. A fixed operation point at sea, such as the stationary position of the drilling vessel 10, is selected. It is realized that the position of such a vessel can vary somewhat during operation, for example due to movements in the connection between the drilling vessel and the drilling site on the sea bed. However, the general position of the operation point should be stationary rather than moving during operation.

In a step 102, the current direction and velocity of the ice drift is measured during a certain initial time period. A preferred way to perform such measurement is by using at least two measurement stations 40, 50. The measurement stations are arranged to continuously or intermittently report their measurement result to a recipient, such as the computer system 11 or another system in contact with the computer system 11, for use in the calculations described below for ice drift forecasting. It is preferred that the average ice drift as reported by the at least two such measurement stations is used as the measured ice drift. This will minimize the impact of local ice floe rotation and other measurement noise .

The step 102 is also viewed in figure 3, the x axis of which displays time, in hours, from a certain start time (0 hours) up to about 36 hours after the start of an exemplary method according to the present invention. As seen in figure 3, the ice drift measurement, shown using a broken (short dashes) , horizontal line, is commenced at time 0, and continues uninterrupted from this time, feeding the measured ice drift values to the computer system 11. The purpose of the initial measuring period, which in the eKample shown in figure 3 extends from the left-most broken, vertical line to the neighboring broken, vertical line, is to establish a general picture of the current ice drift situa- tion in the local area around the operation point. In figure 3, the initial period is thus 6 hours long, but may be longer still .

This general picture is then, in a step 103, used to perform an initial estimate of the value for at least one first physical parameter selected from a first category of such first physical parameters. The initial estimate may be performed automatically by the computer system 11 or manually. Alternatively, the initial estimate may be performed in the same way as in step 107 as described below, using the available ice drift data collected during the step 102 and the currently available second category parameters as described below in connection to step 104. The expression "physical parameter" herein denotes a parameter describing some physical condition affecting the ice drift over time. Each such physical parameter may include one or several actual values, such as a velocity vector, including an absolute velocity and a direction, or a curvature radius.

The respective value (s) of parameters of the said first category are not directly, and preferably neither indirectly, available by measurement to the equipment used in the opera- tions.

The said initial estimate may also be partly based upon earlier experience and externally available data, and aims at obtaining a good input value for the continued, iterative forecasting method as described below.

Examples of parameters of the first category include:

• Wind catch factor: The long-term ice drift velocity resulting, all other things equal, as a consequence of a constant wind blowing over the ice. This parameter is affected by, inter alia, the ice surface structure. It is preferred that the initial estimation of the wind catch factor is such that the wind-induced ice drift is between about 1.0 and 2.5%, preferably about 1.7%, of the current or forecast wind velocity.

• Angular difference between the wind direction and the wind- induced drift component of the ice: Due to, for example, coriolis effects, the long-term ice drift direction will not, all other things equal, coincide with the direction of a constant wind, but will experience a certain non-constant angular difference. It is preferred that the initial estimation of the angular difference is between about 45 and 60 degrees, preferably about 50 degrees, off the current or forecast wind direction (the direction in which the air moves) in the northern hemisphere.

• Local rotational motion of the ice caused by coriolis forces ("inertial oscillations") : The moving ice cover will normally undergo a local rotation due to the coriolos force acting on the moving ice cover. In the northern hemisphere, the rotation is clockwise and its peripheral speed may be determined as about 12 hours divided by the sine of the latitude of the operation point. The radius of the rotation may vary, and must be estimated.

• Local rotational motion of the ice caused by the motion dynamics of the ice itself: The ice cover moves in a dynamic way, due to certain ice elasticity, local differences in ice coverage and quality, etc. This dynamic motion will in general give rise to a possible local rotation in the ice drift, which may be estimated using a rotation radius and peripheral speed.

• General direction and velocity of surface water currents: The surface currents affect the ice drift, but are often difficult or expensive to measure accurately.

• Local rotational motion of surface water currents: At many locations, the surface water current is not straight, but follows a curved path. The exact path is often not known, and local variations may also occur.

For example, the curvature of local rotation may be quanti ^¬ fied by a curvature radius value.

In the example of figure 3, the first category parameters used are wind-induced drift velocity and angle; angle differ ^¬ ence between wind and drift; dynamic ice rotation; and direction, velocity and local rotation of water currents. The first estimation of the first category parameters used is made at the end of the initial time period, thus after 6 hours (estimation points shown as rings) .

In a step 104, the current value for at least one second physical parameter, selected from a second category of physi ^¬ cal parameters, is measured, and/or a forecast over future values for the said at least one second physical parameter is obtained. The respective values are fed into the computer system 11.

As opposed to the first physical parameter category, the value (s) of a parameter of the said second category are known, in the sense that they are either directly or indi- rectly available by measurement, by external information channels and/or in the form of an externally provided forecast. Similarly to the first category, second category parameters also affect the velocity and direction over time of the ice drift. Examples include:

• Wind strength and direction: It is preferred that forecasts of local wind conditions, obtained from external forecasting services, are supplemented by local wind measurements. For example, systematic errors in the forecast may be corrected for using local wind statistics; or both the current wind and the forecast future wind may be used as separate second category parameters.

• Air pressure: The local air pressure may affect the ice cover by pressing down or pulling up, thus affecting the ice drift. Externally obtained forecasts are preferably supplemented by local measurements.

• Tidal water: Tidal water may be forecast and will affect the ice drift. In general, in the northern hemisphere, the tidal motion is clockwise elliptical with a cycle of about 12 hours.

• Latitude: The latitude of the operation point is of course available by measurement, and affects the coriolis-induced part of the ice drift.

• Direction and velocity of surface water currents: Surface current measurement values may also be available not only by forecast, but also by direct measurement.

Thus, depending on which types of information are available by measurement or forecast, a certain physical parameter, such as surface water current velocity vector, may be either a first category parameter or a second category parameter, or a first category parameter regarding the current value and a second category parameter regarding the forecast values. In case updated forecast information is available only at less frequent intervals than a desired ice drift forecasting interval, it is preferred to interpolate the obtained forecast data to the desired ice drift forecast period.

In figure 3, second category parameters include the current wind velocity and direction, which is continuously measured locally (shown using horizontal broken line, short dashes) from the outset of the method. Furthermore, future wind ve- locity and direction; air pressure; and tidal water are used as second category parameters, the forecasts of which are obtained every 18 hours (forecast delivery times shown as squares) . It is realized that values for different forecast parameters may be obtained at different times and intervals. The values are again fed into the computer system 11. In the example of figure 3, the current wind velocity is gathered historically and used primarily to adjust the received wind forecast with respect to systematic errors in the said forecast due to local, systematic wind variations.

As is apparent from figure 3, the steps 102, 103 and 104 need not necessarily be performed in the order shown in figure 1.

In a step 105, a forecast for the future direction and veloc- ity of the local ice drift during at least an immediately following time period is calculated by the computer system 11, based upon at least the values for the used first category parameter (s) and the second category parameter (s) . The expression "immediately following time period" herein denotes a time period of certain length, commencing at the latest at the time for the issuing of the forecast for said period.

1 An exemplary such immediately following time period is shown in figure 3 as a dash dotted (one dash, one point), horizon ^¬ tal line running between 6 hours and 9 hours (as shown by vertical, broken lines) from the start of the method.

The forecast, however, may also cover a longer future time period than the immediately following time period. In figure 3, forecast time periods are shown in dash dotted lines (one dash, two dots) . The forecast period associated with the immediately following time period running between 6 and 9 hours starts from 6 hours and runs up until 30 hours, thus incorporating the following time period in question. This way, the ice drift can be forecast using the present invention for longer future time periods. It is preferred that the forecast time periods are at least 12 hours, more preferably at least 24 hours, most preferably at least 36 hours of dura ^¬ tion.

When the ice drift forecast is calculated, it is preferred that the local ice cover is treated as a stiff sheet which may be translated and rotated, but not sheared nor contracted or expanded. For each geographic point and for each of a series of future times during the forecast period, a forecast direction and speed of motion of this sheet is preferably calculated for each of wind induced drift; water stream induced drift and rotation; inertial oscillations; and ice dynamics induced rotation. Then, these motion vectors are added together for each geographical point and each future time resulting in a forecast ice drift of the said sheet. From here, as the forecast ice drift curve is selected a curve following the general forecast drift of the said sheet and which passes the operation point. Thus, it is preferred that at least two rotational motions of the ice sheet, each derivable from a respective first category parameter, are superposed in order to calculate the total forecast ice drift. Specifically, it is preferred that the first category parameters comprise at least two of the local rotational motion of the ice caused by the motion dynamics of the ice itself, local rotational motion of the ice caused by coriolis forces and local rotational motion of surface water currents, and that in this case the said rotational motions are superposed in the calculation of the total ice drift.

It is to be understood that "rotational motion" of the ice sheet in this context relates to the curvature of the local motion of the ice, and that such rotational motion needs not display a complete revolution.

In this case, where several rotational motions are superposed, it may, however, be so that the forecast total ice drift displays loops in which the ice drift direction performs several complete 360 degrees turns. Then, it is preferred that an additional first category parameter, when used in the calculation of the ice drift forecast, in the forecast ice drift path causes a relative decrease in the size of any such loops arising in the ice drift path over time, so that said loop size decrease in the forecast ice drift path is imparted to a higher extent for downstream such loops than for upstream such loops. In other words, such looping ice drift patterns are in fact dampened by the said additional first category parameter. The present inventors have discovered that this generally leads to more accurate ice drift forecasts .

Figure 7 shows an actual ice drift path 700 as measured during several days, from a starting point 700a to an end point 700b. Loops 701, 702, 703, 704, 705 appear as a consequence of added rotational motions. As is clear from this figure, the loop sizes tend to become smaller over time - the loop 701 is the largest one, and at 705, it is no more than a bump in the ice drift path. With the exception of loop 703, every consecutive loop is smaller than its predecessor. This is typically the case as long as there are no rapid shifts in for example wind direction. This behavior is best captured using a separate first category parameter.

The resolution as regards time and geography used for the forecast may vary depending on the actual purposes and condi ^¬ tions, but it is preferred that the ice drift is measured at least once a minute and that the time density of the dataset comprised in the forecast is also at least one forecast time per minute.

In a step 106, performed during said following time period, the actual direction and velocity of the ice drift is then measured, in a way which may be similar to the one described above in connection to step 102.

Then, in a step 107, which is performed after the end of said following time period, the forecast ice drift is compared, by the computer system 11, to the actual ice drift as measured in step 106 and possibly also earlier.

The comparison is carried out for a certain historical time period, comprising the immediately following time period during which step 106 was last conducted. This means that the historical time period may be as short as the immediately following time period, but may also extend backwards in time. This may be advantageous, since more data is then available for the calculations, which potentially leads to a better comparison. Contrary thereto, certain physical parameters may change over time in ways that make comparisons across too long time periods unreliable. In figure 3, these historical time periods are shown using broken lines (long dashes) . The historical time period associated with the immediately following time period running between 6 and 9 hours, as discussed above, runs from the start of the method, at time 0, to 9 hours, thus incorporating the immediately following time period in question.

In case the historical time period is the same as the imme ^¬ diately following time period, the forecast produced in step 105 may be used directly. If the historical time period starts earlier than the immediately following time period, a new, retroactive forecast needs to be calculated for the compariso .

According to a preferred embodiment, the computer system 11 varies the value of at least one of the, or each, used first category parameter and numerically calculates a conformance measure between the forecast and the actual ice drift during the historical time period in question using several values for the varying first parameter ( s) . Then, a respective optim ^¬ al value for each varied first category parameter is selected as the one that maximizes the said conformance measure.

According to a preferred embodiment, the complete ice drift path during the historical time period in question is consi ^¬ dered in the conformance measure, or at least a stepwise path built from the discreet ice drift measurement points available and time-wise corresponding forecast points.

In a way which is similar to the forecast for the immediately following time period as discussed above, the calculated retroactive forecast ice drift may comprise a dataset in turn comprising a respective direction value and a respective velocity value for the ice drift for each of a plurality of times during the historical time period. Analogously, the dataset may comprise a set of geographical positions with a corresponding set of time stamps. The measurement of the real direction and velocity of the ice drift will then result in a corresponding dataset as the one comprised in the retroactive forecast .

Based upon the said comparison, a certain respective optimal value for each one of the at least one first category parameters is determined by the computer system 11, which optimal value maximizes the said conformance measure between the forecast and the actual ice drift during the historical time period when the forecast is calculated based upon the said respective optimal value for the one or several first catego ^¬ ry parameters and upon the said one or several second catego ^¬ ry parameters .

The computer system 11 may preferably calculate the degree of conformance between said two datasets based upon the pairwise difference between respective measurement points with corres ^¬ ponding time stamps of the forecast and actual ice drift path, preferably also taking into account the relative dif ^¬ ferences between consecutive points in both respective paths.

The present inventors have found the calculations to be par ^¬ ticularly efficient, and the correlation to be particularly relevant, if calculated according to the following:

1) Store the forecast and actually measured ice drift paths as two respective time series of geographical points with individual timestamps. If necessary, relate the time scales of the two time series so that the time stamps of one series correspond to those of the other. This may involve adjusting one or both of the time series by interpolation in order for them to share the same set of time stamps. If necessary, one or both time series may also be stripped of points that have no corresponding point in the other time series, such as in the beginning and/or the end of one or both of the series.

2) Translate the points so that both paths start at the geographical position.

3) Then, calculate the correlation C as:

C = ^actual

^"act al

where L _actU ai is the total length of the measured, actual ice drift path, and AL is the total length of a path defining the difference between the measured and forecast ice drift paths, according to

^actual ~∑| P actual ] ^{a n <} ^ = ^ (^? f _orecmt ~ P actual ) ~ \P forecast ~ Paelua/ \ '

ί= where N is the total number of points in each respective used time series, p _a ' _mmi is the geographic position of the measured ice drift path at point i and p' _ftmaisl is the geographic position of the forecast ice drift path at point i.

The higher the correlation C, the better the forecast.

Hence, in the exemplary embodiment of figure 3, at time 9 hours, a retroactive forecast is calculated for the ice drift during the historical time period between 0 and 9 hours, calculated based upon the actual, known ice situation at 0 hours and upon the wind, air pressure and tidal water fore- casts (second category parameters) as obtained at time 0 and varying the values for wind-induced drift and angle difference, dynamic ice rotation and water current (first category parameters) . Each such retroactive forecast is compared to the actual ice drift as measured between hours 0 and 9.

A respective conformance measure of each pair of retroactive forecast and actual ice curve is calculated for each combination of first category parameters, and an optimal set of first category parameter values is obtained as the one maximizing the said conformance measure.

Hence, the said optimal set constitutes a set of updated first category parameter values which, if used in the same forecast model as used in step 105 above at time 0 hours, would have produced a maximally correct forecast of the ice drift during the period between 0 and 9 hours.

It may be advantageous to use known values of the second category parameter values at the beginning of the historical time period, in other words at time 0 in the present example. In some cases, this may for instance provide an automatic correction of any systematic errors in externally provided forecasts, via corresponding systematic changes in first category parameter values. However, according to an alternative, preferred embodiment, those second category parameters for which more recent or correct values exist at the time for the comparison are used instead of the known values from the beginning of the historical time period. If a second category parameter which is updated during the immediately following time period is used, such as would be the case if the current wind as shown in figure 3 would be used as a second category parameter, it is preferred to update the retroactive forecast using the most current values throughout the whole considered historical period.

Next, in a step 108, the set of first category parameters values estimated previously in step 103 are replaced with the corresponding optimal set of values obtained in step 107. In other words, said optimal values are used as the estimation of the value for the at least one first category parameter. Thereafter, in a step 109, the method is repeated either from step 104, if there are updated second category parameter data available, or, otherwise, from step 105. This way, an updated ice drift forecast can be obtained for the next following time period.

As is seen from figure 3, the next such following time period runs between 9 and 12 hours, having an associated historical time period running from 0 to 12 hours and an associated forecast time period running from 9 to 33 hours. No new fore- cast values are available for wind, air pressure and tidal water at 9 hours, so the method carries on using the values obtained at 0 hours. However, the current wind vector data is used to correct the wind forecast obtained at time 0. The current wind vector data could also be input directly into the forecast produced at 9 hours as a second category parame ^¬ ter .

Thereafter, consecutive three-hour following time periods appear, each with an associated historical time period and an associated forecast time period, all of which time periods occur with three-hour intervals. Historical time periods are 12 hours of duration or less, which is the case for the first two historical time periods which both start at time 0. In other words, the duration of each historical time period is at least the smaller of 12 hours and the total time since the step 102 was performed. Alternatively, the historical time periods are at least as long as the corresponding immediately following time period. Forecast time periods are each 24 hours in the present example. Preferably, each forecast time period is at least 12 hours of duration.

At 18 hours and 36 hours, new forecast data is available regarding wind, air pressure and tidal water. From this re- spective point and onwards, the updated data is used when producing forecasts.

The above described method is capable of producing a constantly updated forecast of the local ice drift situation on a relatively short time scale, and making use of the current ^¬ ly available data. By using the assumption of existence of certain semi-stable factors affecting the ice drift locally, such as ice wind catch behavior and local ice rotation due to water currents and ice dynamics, an accurate forecast can be produced once relevant values for these parameters have been established. Since the method is iterative, the forecasts may grow increasingly more precise as more local measurement data is available, and will thereafter adapt to changing local conditions as time goes by, so as to always produce a rela- tively reliable forecast.

The produced ice drift forecasts may then be displayed to a user of the system, such as an operation fleet manager, using a suitable display means, and/or be fed into an ice raanage- ment operation system which for example instructs the ice ^¬ breaker 20, 30 captains as to what areas to prioritize when managing the ice.

Γ It is preferred that each immediately following time period used in the step 105 is about 5 hours or less, preferably about 3 hours or less. This yields a reasonable good time resolution as regards forecast updates, and at the same time limits the resources required for forecast calculations. On the other hand, according to another preferred embodiment, each immediately following time period is only about 10 minutes or less, preferable about 5 minutes or less, most preferably about 2 minutes or less. This will require more data computing power, but will yield a more or less continuously updated forecast, always using the most recent information on current wind values, ice drift, etc. According to one preferred embodiment, a new immediately following period will commence, and step 107 thus be performed, each time a new local ice drift measurement value is received. It may also be possible for a user of the system to manually trigger the start of a new immediately following time period at any time, using the most currently known information to provide a forecast .

Table 1 and figures 6a-6g show an example of a series of ice drift forecast calculations according to the present invention. Figures 6a~6g display as x and y axes longitude and latitude, respectively.

During a time period of 12 hours, the ice drift was measured near the operation point, in a way such as described above. The measured ice drift is depicted in each one of the graphs of figure 6a-6g as a broken line (short dots, "TL") , where time roughly increases to the left in the figures.

During the same time, the actual wind direction and velocity were measured at the operation point. A forecast was availa ^¬ ble regarding the wind direction and velocity during the

H following 38 hours at the operation point. The latitude and the longitude of the operation point were known in advance. These were thus the available second category parameters. A number of first category parameters were selected according to the above described, including angular difference between the wind direction and the wind-induced drift component of the ice; wind catch factor; direction, velocity and radius of rotation of surface water currents; and radius of rotation of inertial oscillations.

A forecast ice drift model was defined according to the fol ^¬ lowing :

Total ice drift = Wind drift induced ice drift + Water stream induced ice drift + Inertial oscillation induced ice drift, where

o Wind drift induced ice drift is calculated based upon the wind as measured during the first 10 hours and then as forecast during the next 38 hours, using the estimated wind angle difference and catch factor;

o Water stream induced ice drift is calculated based upon an imagined, constantly circularly rotating water stream having the estimated direction at the operation point and streaming with the estimated velocity and radius of rotation; and

o Inertial oscillation induced ice drift is calculated based upon the peripheral speed, in turn calculated from the latitude of the operation point, and the estimated radius of rotation.

After 10 hours of ice drift measurement, optimal values for the first category parameters were established. Table 1 shows a series of first category parameter candidates for each of the first category parameters used. A respective forecast was calculated using the above described model for each such set of first category parameters, and using the second category parameters as described above, running from the start of the past 10 hour period and up to the end of the following 38 hour period. The result is shown in figures 6a-6g, corresponding to the likewise named columns in Table 1. In each of the figures 6a-6g, the forecast ice drift is displayed using two lines, one representing past time {solid, <=T") and one representing future time (broken, long dots, ">T") .

Table 1

Figure 6a 6b 6c 6d 6e 6f 6g

Wind angle difference (°) 55 50 50 80 80 65 65

Wind catch factor {%) 2.00 1.50 1,50 1.50 1.55 1.55 1.00

Water stream direction

230 250 250 250 250 190 280

P)

Water stream velocity

0.01 0.01 0 , 01 0.01 0.01 0, 10 0.15 (kno s)

Water stream rotation

99 99 99 99 99 2 4 radius (nautical miles)

Inertial oscillation

0.35 0.35 0.20 0.20 0.20 0.20 0.15 radius (nautical miles) The initially selected parameters, according to figure 6a, were selected based upon the measured ice drift in combination with the experience of the forecasting personnel. The consecutive sets of parameters were obtained by iteratively modifying the previous set using a priori knowledge of how each parameter generally affects the ice drift. For example, the significant lowering of the water stream rotation radius between figures 6e and 6f yields a stronger clockwise rotation, which brings the forecast for the first 10 hours closer to the actually measured ice path. For each forecast, the curve for the past 10 hours was then compared to the actual ice drift during that same time as described above, and a correlation C was calculated for each of the said forecasts.

Alternatively, a brute force method could be used, wherein all first category parameters are varied within a certain respective interval, and forecasts are calculated based on all possible such combinations of first category parameters.

Then, all calculated correlations were compared across forecast candidates. The ice drift of figure 6g was the one closest correlated to the actually measured ice drift during the 10 hour time period. This is also visually clear from the figures 6a-6g. Therefore, the forecast for the coming 38 hours was calculated based upon the first category parameter values as used in figure 6g.

This procedure was then followed again after another 10 minutes of ice drift measurement, yielding an updated set of first category parameters for the next 38 hours after a total of 10 hours 10 minutes of continuous ice drift measurement, and so on every 10 minutes, as long as updated forecasts were desired. The updated set of first category parameters was obtained by brute force variation of the parameters as described above, but while varying the value of each parameter only within the vicinity of the previously estimated respective parameter value since the expected parameter value changes were typically not large because of the short elapsed time. Good results were obtained when only allowing each parameter to vary at the most 10% from its previous value. W

26

When no well correlated forecast curves were occasionally obtained, in other words when the best correlated forecast curve had a correlation C lower than a predetermined value, the first category parameters were allowed to instead vary s across a significantly broader respective interval before settling for an estimated set of parameters.

For each calculation of the first category parameters resulting in the best correlated forecast curve, the most recently w available wind measurement and forecast data was used.

According to a preferred embodiment, the forecasting method described above is used as a component in a method for managing the ice conditions at the above-described fixed operation 15 point at sea. Such a method is depicted in figure 2.

Thus, in a step 201, the method is started. Then, in a step 202, the fixed operation point is established at sea. With reference to figure 4, this may mean clearing or managing the 0 ice immediately surrounding the drilling vessel 10, which is then positioned at the operation point.

In a step 203, a local ice drift forecast 1 is calibrated during a certain calibration time period, by repeatedly mea- suring the ice drift and causing the computer system 11 to produce a repeatedly updated forecast of the future local ice drift according to the forecasting method 204 described above . 0 In a subsequent step 207, an operation time period follows, during which the industrial operation, such as drilling, is performed, and during which the ice drift is continued to be measured repeatedly, and during which the computer system 11 is caused to continue to produce a repeatedly updated fore- cast of the future ice drift using the method 204. In addition, during the step 207, the ice in the area 2 (see figure 4) is managed according to the above said, with the aim of keeping the risk of ice floe damage to the equipment at the operation point at or below a certain highest acceptable level .

According to a preferred embodiment, the calibration in step 203 is performed until the change in the estimated parameter values for the at least one first category parameter which is estimated in the ice drift forecasting method 204, between two consecutive iterations, is less than a predetermined value. In other words, the ice drift is continuously forecast according to the above until at least one of the used first category parameters only changes by an amount which is smaller than the predetermined value. Preferably, all used first category parameters must change by a respective predetermined value or less before step 203 is finished. Alternatively, a metric calculated based upon the used first category parame- ters, such as their weighted average, must change by a respective predetermined value or less. Thus, the calibration will continue until a relatively stable forecasting model is achieved before starting the operation, Thus, in a step 205 after one or each iteration of the above described forecasting method 204, it is checked whether the first category parameter is stable or not. If not, another calibration iteration step 203 is performed. If the first category parameter is indeed stable, it is preferably checked whether a predetermined shortest calibration time has elapsed since calibration 203 first started. If not, another calibration iteration step 203 is performed. Otherwise, the method jumps to the operation step 207. Steps 205 and 206 may be alternatively performed in the opposite order. It is especially preferred that the predetermined shortest calibration time period is at least 6 hours, in order to guarantee a stable forecasting model before commencing the industrial activity.

In a final step 208, the operation is terminated.

Above, preferred embodiments have been described. However, it is apparent to the skilled person than many modifications may be made to the described embodiments without departing from the idea of the invention.

Thus, the ice drift forecast produced by a method according to the present invention may be used not only for deep ocean drilling operations. For example, an oil spill may be moni ^¬ tored using the forecasting method in order to estimate the future path of spilt oil travelling with the ice cover. Thus, the invention shall not be limited to the described embodiments, but is variable within the scope of the enclosed claims .