JP4998039B2 - Observation data assimilation method - Google Patents

Description

  The present invention relates to an observation data assimilation method for assimilating observation data such as ocean temperature obtained using an XBT (Expendable Bathythermograph) or the like into prediction data calculated based on a numerical model or the like.

  In the prediction system in the meteorological and marine fields, the observation data obtained by observation is the most important. However, observation data is discrete data both spatially and temporally, especially in the marine field, and it is non-uniform data. Have difficulty.

  Therefore, various attempts have been made to make the best use of this observation data to accurately grasp the actual ocean condition, and now it is assimilated into prediction data calculated based on a numerical model etc. by interpolating the observation data. As a result, highly accurate spatially continuous corrected prediction data is generated.

  Various methods have been considered for the interpolation of this observation data. To date, it has been most widely used in the field of weather forecasting and ocean forecasting, and is an effective method in terms of accuracy and calculation cost. As one, the optimal interpolation method is known.

  The optimal interpolation method is a method for estimating values (temperature, salinity concentration, ocean current velocity, etc.) at each lattice point set in the vicinity of the observation data in consideration of measurement errors included in the observation data. In this optimal interpolation method, the value of each grid point is estimated so that the error between the true value at each grid point (which is unknown because it cannot be actually observed) and the estimated value at each grid point is minimized.

  In the conventional optimal interpolation method, vertical observation data, which is discrete values, is interpolated individually for four-dimensional observation data (space cubic, temporal primary) to align the vertical grid between the observation data, and then there is Focusing on the horizontal plane of the depth, the observation data is optimally interpolated in the temporal and horizontal directions. Hereinafter, such a method is referred to as a three-dimensional optimal interpolation method.

  Conventionally, the vertical direction of observation data is interpolated first, and then the procedure for optimal interpolation in the horizontal direction is adopted. The main reasons are that it is considered sufficient, that computational resources can be constrained, and that the vertical spatial scale is not taken into account because the vertical fluctuation characteristics of the ocean are not well understood.

  In the ocean, observations have shown that the horizontal seawater flow is much larger than the vertical seawater flow. When assimilating observation data with prediction data obtained from a numerical model covering a wide sea area, sufficient accuracy can be obtained by the above three-dimensional optimal interpolation method.

  However, when the water temperature profile is actually confirmed by XBT (Expendable Bathythermograph) or the like, it has been confirmed that changes in the mixed layer, thin cold water masses, etc. appear in the vertical direction as local fluctuations. For example, in underwater acoustic engineering, it is known by numerical calculation that this local fluctuation in water temperature in the vertical direction has a very large influence on the propagation of sound in seawater. Therefore, the water temperature estimation grid data (predicted data) generated by the conventional three-dimensional optimal interpolation method that does not take into account minute vertical fluctuations in the water temperature observation data emphasizes local vertical ocean fluctuations. When used in fields such as sound wave propagation, the accuracy becomes a problem.

Note that a technique for interpolating observation data and assimilating it into prediction data obtained by numerical calculation is also described in Patent Document 1, for example. Patent Document 1 discloses a technique for integrating the correction result of the forecast data manually by the operator into the forecast result of the numerical ocean model without destroying the physical characteristics of the ocean.
JP 2004-85394 A

  As described above, in the conventional observation data assimilation method using the optimal interpolation method, the vertical component is first interpolated for each observation data to adjust the depth, and then the optimal interpolation is performed in the horizontal and temporal directions. The procedure to be adopted is adopted.

  This is a characteristic of observation water temperature profile data that is representative of observation data at present. Many of the observation data use the singular point depth and the water temperature as the singular point where the water temperature changes greatly in the vertical direction. This is because the depth of this singular point varies in each observed water temperature profile.

  That is, in the conventional observation data interpolation method, the representative depth is first determined by interpolation in the vertical direction, and then the horizontal section at the representative depth is optimally interpolated. Since the singularity depth of the data is different and there is no relation between the representative depths, the observation water temperature profile data cannot be reproduced unless the representative depth is set to the same interval as the singularity. For this reason, there is a problem when emphasizing local water temperature fluctuations in the vertical direction, particularly in the ocean conditions where the vertical fluctuation is large, such as in summer.

  The present invention has been made to solve the above-described problems of the prior art, and an object thereof is to provide an observation data assimilation method and apparatus capable of obtaining highly accurate prediction data in the vertical direction. .

In order to achieve the above object, the observation data assimilation method of the present invention is an observation data assimilation method for associating observation data with data obtained from a numerical model,
Using the average standard deviation representing the magnitude of the average fluctuation of the observation data and the statistical data of the acquisition sea area of the observation data to determine the standard standard deviation used to determine the magnitude of the vertical data fluctuation,
When the average standard deviation is less than or equal to the reference standard deviation, the vertical direction of the observation data is first interpolated, and then interpolated using a three-dimensional optimal interpolation method that optimally interpolates in the horizontal and temporal directions. Run
When the average standard deviation is larger than the standard standard deviation, an interpolation process is performed using a four-dimensional optimal interpolation method that optimally interpolates in the horizontal direction, vertical direction, and time direction of the observation data,
It is a method of assimilating the data after the three-dimensional optimal interpolation method or the four-dimensional optimal interpolation method into data obtained from the numerical model.

On the other hand, the observation data assimilation device of the present invention is an observation data assimilation device that assimilate observation data into data obtained from a numerical model,
Using the average standard deviation representing the average fluctuation magnitude of the observation data and the statistical data of the sea area where the observation data was acquired, a standard standard deviation used to determine the magnitude of vertical data fluctuation is obtained, and the average standard If the deviation is less than or equal to the standard standard deviation, the vertical direction of the observation data is interpolated first, and then an interpolation process is performed using a three-dimensional optimal interpolation method that optimally interpolates in the horizontal and temporal directions. When the average standard deviation is larger than the reference standard deviation, an interpolation process is performed using a four-dimensional optimal interpolation method that optimally interpolates in the horizontal direction, vertical direction, and time direction of the observation data, A processing apparatus for assimilating the data after the three-dimensional optimal interpolation method or the four-dimensional optimal interpolation method into the data obtained from the numerical model;
A data storage device for storing the observation data, the statistical data, and data obtained from the numerical model;
Have

  In the method and configuration as described above, when it can be determined that the average standard deviation of the observation data is larger than the reference standard deviation and the vertical data fluctuation is large, the observation data is interpolated using the four-dimensional optimal interpolation method. Therefore, the singularity data of the observed water temperature profile can be used for interpolation as it is, and even in a marine environment with large fluctuations such as in summer, the fluctuation of the water temperature in the vertical direction of the ocean is displayed on the grid data. And more accurate prediction data can be obtained.

  In addition, if the average standard deviation is less than the standard standard deviation and it can be determined that the vertical data fluctuation is small, the observation data is interpolated using the three-dimensional optimal interpolation method, which suppresses the increase in calculation cost. Is done.

  According to the present invention, when it can be determined that the data fluctuation in the vertical direction is large, the observation data is interpolated using the four-dimensional optimal interpolation method, so that more accurate prediction data can be obtained. Further, when it can be determined that the data fluctuation in the vertical direction is small, the observation data is interpolated using the three-dimensional optimal interpolation method, so that an increase in calculation cost is suppressed.

  Next, the present invention will be described with reference to the drawings.

  FIG. 1 is a block diagram showing a configuration example of an observation data assimilation apparatus according to the present invention.

  The observation data assimilation apparatus of the present invention can be realized by a computer, for example. For example, as shown in FIG. 1, the computer includes a processing device 10 that executes predetermined processing according to a program, an input device 20 for inputting commands and information to the processing device 10, and processing of the processing device 10. And an output device 30 for monitoring the result.

  The processing device 10 stores a CPU 11, a main storage device 12 that temporarily stores information necessary for the processing of the CPU 11, and a program for causing the CPU 11 to execute processing of an observation data assimilation method of the present invention described later. Data for storing various data necessary for ocean prediction processing using the observation data assimilation device of the present invention, such as the recording medium 13, water temperature observation data, quality controlled data described later, water temperature climate data, water temperature prediction data, etc. A memory control interface unit 15 that controls data transfer between the storage device 14, the main storage device 12, the recording medium 13, and the data storage device 14, and an I / O interface unit that is an interface device between the input device 2 and the output device 3 16. The CPU 11 is connected to the memory control interface unit 15 and the I / O interface unit 16 via a bus 18. The processing device 1 may include a communication control device 17 that is an interface for transmitting / receiving information to / from a server or a database via a network.

  The processing device 1 executes processing of an observation data assimilation method described below according to a program recorded on the recording medium 13. The recording medium 13 may be a magnetic disk, a semiconductor memory, an optical disk, or other recording medium.

  Note that the observation data assimilation apparatus of the present invention is not limited to the configuration realized by the computer shown in FIG. 1, and is, for example, an LSI (Large Scale Integration) or DSP (Digital Signal Processor) composed of a logic circuit or the like. It may be realized by a semiconductor integrated circuit device such as a memory and a memory. Although FIG. 1 shows a configuration in which the data storage device 14 is provided in the processing device 10, the data storage device 14 may be provided independently of the processing device 10.

  In such a configuration, the observation data assimilation apparatus of the present invention generates quality-controlled data by removing artificial errors from the water temperature observation data measured by XBT, etc. Calculate the average standard deviation in the large mixed layer. Furthermore, the standard standard deviation used for judging the magnitude of the water temperature fluctuation in the vertical direction is determined using statistical data. And when the average standard deviation of observation data is larger than a standard standard deviation, it judges that the water temperature fluctuation | variation of a perpendicular direction is large, and performs the interpolation process using the 4-dimensional optimal interpolation method mentioned later. In addition, when it is determined that the fluctuation in the water temperature in the vertical direction is small, the interpolation process is executed using the above-described three-dimensional optimum interpolation method. The observation data processed by each interpolation method is assimilated by performing a nudging process on the prediction data obtained by numerical calculation, and is output as corrected prediction data.

  The processing procedure of the observation data assimilation method of the present invention will be described below with reference to the drawings.

  FIG. 2 is a flowchart showing the processing procedure of the observation data assimilation apparatus of the present invention.

  In FIG. 2, the processing apparatus 10 first improves the quality of the water temperature observation data 1 acquired using XBT, CTD (Conductivity Temperature Depth profiler), etc., in the quality control process (step A1). The result of the quality control process is stored in the data storage device 14 as quality controlled data. Here, improving the quality of the water temperature observation data refers to a process of removing from the water temperature observation data a systematic error (an error having a certain tendency not caused by chance, such as an artificial error) that occurs during measurement. For example, in the measurement of seawater temperature using XBT, the measurement depth of the water temperature is obtained from the elapsed time after dropping the probe using a predetermined conversion formula, and errors due to this conversion formula are also included in the system error. Since the optimal interpolation method is a statistical method, the error needs to be a coincidence error caused by chance. Therefore, it is important to remove systematic errors from the water temperature observation data.

  Next, the processing apparatus 10 determines the depth of the ocean to be processed in the standard deviation calculation process (step A2) based on the water temperature observation data (quality controlled data) after the quality control process. The depth of the ocean depends on the water temperature distribution in the vertical direction, the mixed layer is characterized by temperature changes from near the surface to a depth of about 200 m, the seasonal climbing layer where the water temperature falls greatly regardless of the season, the water temperature is It can be roughly classified into three types of deep layers of several thousand meters or less that hardly change. Here, a mixed layer whose water temperature fluctuates in a relatively small space unit according to a change in season is set as a processing target layer, and the following processing is performed on this mixed layer.

The processing apparatus 10 first determines the depth D ₀ from the sea surface of the mixed layer according to the following procedure.

The processing device 10 is configured to surround a water temperature observation data acquisition area within a minimum rectangular shape of the global statistical grid data WOA01 (World Ocean Database 2001) or GDEM (Generalized Digital Environmental Model). Extract water temperature statistical data. Here, the season of the water temperature statistical data is winter data in which the mixed layer is deepest. The processing apparatus 10 calculates the average value (statistical average data) of the total water temperature statistical data existing in this sea area for each depth grid. The value of each depth grid is obtained by vertical interpolation at 1 m intervals using the well-known Akima method. Then, the depth of sea temperatures from sea level drops once defined as D _0, is calculated on the basis of the D ₀ in the statistical average data.

Next, based on two points, a singular point closest to the surface layer of each observation profile and a singular point closest to the depth D ₀ , a linear equation of depth and water temperature is derived by linear interpolation. The straight line indicated by this linear equation is called the reference straight line of each observation profile. Here, when the number of observation profiles is N and the number of singular points of an observation profile i is M, the water temperature / depth coordinates at the j-th singular point from the surface layer of the observation profile i are (d _j ⁱ , t _j ⁱ ). It is represented by The water temperature t _j ^{i ′} at the depth d _j ⁱ on the reference straight line derived corresponding to the i-th observation profile is referred to as the reference water temperature at that depth.

The singular point deviation Δt _j ⁱ from the reference straight line at the j-th depth of the observation profile i can be expressed by the following equation (1) based on the reference water temperature.

  Here, the average value of the singular point deviations at all depths from the reference line in the observation profile i can be described by the following equation (2) except for the upper and lower two points used for the reference line.

  Thereby, the average water temperature deviation from the reference straight line in each profile is known. Furthermore, when this equation (2) is used, the standard deviation with respect to the water temperature deviation of the singular point from the reference straight line can be obtained by the following equation (3).

  If the standard deviation obtained in this way is further averaged over the entire observation profile, the average standard deviation σ at the singular point in the entire space of the observation profile becomes as shown in Equation (4).

  By this equation (4), the standard deviation with respect to the reference straight line in the region including the observation data is obtained. This average standard deviation σ represents the magnitude of the average fluctuation of the observation data, and the larger σ indicates that the vertical water temperature field changes more complicatedly.

  When obtaining the average standard deviation σ in the standard deviation calculation process, the processing device 10 compares the reference standard deviation σ threshold value with the average standard deviation σ in the interpolation processing method determination (step A3), and determines the reference standard deviation σ. The flag is set to “0” when the average standard deviation σ is smaller than the threshold, the flag is set to “1” when the average standard deviation σ is larger than the reference standard deviation σ threshold, and the three-dimensional when the flag is “0”. The process proceeds to an interpolation process (step A4). If the flag is “1”, the process proceeds to a four-dimensional interpolation process (step A5). The three-dimensional interpolation process is a two-dimensional interpolation process in which the vertical component is interpolated for each observation data described above to adjust the depth, and then optimally interpolated in the horizontal direction (+ time direction). Further, the four-dimensional interpolation process is a three-dimensional interpolation process that optimally interpolates in the horizontal direction and the vertical direction (+ time direction) without interpolating the vertical component of the observation data and aligning the depth.

The reference standard deviation σ threshold is the statistical data of the sea area set so as to be surrounded by the smallest rectangular shape with respect to the area where the water temperature observation data is acquired, and the statistical data for each season of the sea area in the same manner as the calculation of the average standard deviation σ. The average value and standard deviation are calculated for each depth based on all statistical data in the sea area up to the depth D _{0 of the} mixed layer, and the standard deviation of all layers is averaged.

  As described above, when the average standard deviation σ is small, it indicates that the spatiotemporal fluctuation of the observation data is small, and when the average standard deviation σ is large, the spatiotemporal fluctuation of the observation data is large. This means that the vertical fluctuation of the observation data is also large. The estimated value obtained by the three-dimensional optimal interpolation method is considered to have a large error when the spatiotemporal variation is large. Therefore, in the three-dimensional interpolation process (three-dimensional optimal interpolation method), values at each depth are estimated using only observation data interpolated in the same vertical direction as the estimated depth. On the other hand, in the above-described four-dimensional interpolation process (four-dimensional optimal interpolation method), values at each depth are estimated using also observation data (singular points) adjacent to each other in the vertical direction of the correlated estimated depth. By executing this four-dimensional interpolation process, if the standard deviation is large, that is, if the fluctuation in the vertical direction is also large, use the data of the singular point in the depth direction of the observation data to consider the change according to the depth of the water temperature. Therefore, the estimated value in the depth direction (vertical direction) can be calculated with high accuracy. Furthermore, in the conventional method using only the three-dimensional optimal interpolation method, it was necessary to interpolate the depth of each observation data in advance according to the estimated depth and prepare the data of the same representative depth. In the above four-dimensional interpolation processing, it is possible to estimate the depth value determined as it is, regardless of the singular point depth of the obtained observation data. However, the four-dimensional interpolation process increases the calculation cost because the number of dimensions is increased by one compared to the three-dimensional interpolation process. Therefore, an increase in calculation cost can be suppressed by executing three-dimensional interpolation processing for observation data with small vertical fluctuations.

  As described above, in this embodiment, the interpolation process is performed using the three-dimensional optimum interpolation method or the four-dimensional optimum interpolation method according to the value of the average standard deviation σ, and either the corrected water temperature data 8a or 8b is used. Output one.

  It should be noted that the parameters used in the optimal interpolation method (correlation coefficient and error variance, which will be described later) are known to vary depending on the sea area, and these parameters are determined by roughly dividing the sea area. The current situation is. Therefore, even when determining the type of interpolation processing according to the value of the standard deviation, only water temperature observation data in the sea area to be processed is used. This is because if a parameter corresponding to the sea area is ignored and a standard deviation of a wider area is obtained, very large fluctuations always appear.

  Next, the procedure of the four-dimensional interpolation process shown in FIG. 2 will be described with reference to FIG.

  FIG. 3 is a flowchart showing a processing procedure of the four-dimensional interpolation processing shown in FIG.

  In FIG. 3, the processing apparatus 10 first outputs anomaly data 12 from the difference between the quality-controlled data and the water temperature / climate data (anomaly conversion process: step B1).

  When the seawater temperature is estimated using the optimal interpolation method, it is necessary to first determine an initial value (first estimated value) that serves as a reference for the amount to be estimated. The processing apparatus 10 uses global statistical data such as the above-mentioned WOA01 and GDEM as water temperature and climate data which is data for this purpose.

  Next, the processing apparatus 10 selects the time, space scale, and error variance used in the four-dimensional optimal interpolation method as the interpolation parameter selection process (step B2).

In the optimal interpolation method, an autocorrelation coefficient including a spatiotemporal scale as a variable and an error variance value are used as parameters. The autocorrelation coefficient μ _{ij, mn} in the four-dimensional optimal interpolation method is defined by the following equation (5).

Here, the function H is a horizontal function that defines the horizontal correlation including the horizontal spatial scale C _xy using the horizontal latitude and longitude differences as variables. The function V is a vertical function that defines a vertical correlation including a vertical space scale C _z with a depth difference as a variable. The function T is a time function that defines the correlation in the time direction including the time scale C _t with the time t as a variable.

Since the autocorrelation coefficient μ _{ij, mn} in the four-dimensional optimal interpolation method is a value obtained by multiplying the autocorrelation coefficient in the three-dimensional optimal interpolation method by the vertical function V, the horizontal function included in the above equation (5) H and the time function T are the same as the three-dimensional optimal interpolation method, and a Gaussian type is used as a representative function thereof.

Many studies have already been made on the time and the horizontal spatial scale included in each of the above functions, and it is known that the horizontal spatial scale C _xy differs for each sea area. For this reason, a map of the horizontal spatial scale C _xy divided for each sea area is used according to the sea area of the observation data. For example, C _xy = 150 km, which is considered to exhibit characteristic fluctuations around Japan, is selected. Similarly, the time scale _Ct uses a map prepared in advance according to the sea area of the observation data. For example, C _t = 720 (hours) around Japan. However, care must be taken with the vertical spatial scale.

  The spatial scale in the vertical direction is very small (several tens of meters to several hundreds of meters), unlike the spatial scale in the horizontal direction (several km to several hundred km). In addition, as described above, as described above, the ocean has a mixed layer with a depth of about 200 m from the surface layer, a seasonal climatic layer in which the water temperature drops greatly regardless of the season, and almost no change in temperature. It is classified into three types of layers of several 1000 m or less close to 0 ° C.

For this reason, it is considered that there is little correlation between the water temperature changes between these layers, and in particular, there is no correlation between the mixed layer and the deep layer of 1000 m or less. Therefore, in the vertical spatial scale and the vertical correlation coefficient, it is necessary to determine the value and function form from the average mixed layer thickness of the observation data in consideration of this characteristic. Therefore, V˜exp [− (Δdep) ² / C _Z ] is used as the correlation coefficient in the vertical direction as a function having a sharper peak than in the horizontal direction. This correlation coefficient is similarly applied to all the three types of layers. Here, Δdep represents a depth difference between each singular point depth, and C _Z represents a vertical spatial scale. Thereby, the correlation of each layer is set to a smaller value. Further, the spatial scale C _Z in the vertical direction has a seasonal variation unlike the horizontal direction, and therefore a value corresponding to the season is selected.

As the value of the spatial scale C _Z , an average value of the thicknesses of the three layers is obtained in advance based on statistical data, and is used according to the season of the observation data. For example, in the winter mixed layer, C _Z = Z ₀ (Elevation of Mean Level above Chart Datum).

  Further, the error variance value λ is selected as a parameter, and this also selects a value common to the three-dimensional optimal interpolation method based on a map created in advance corresponding to the sea area, similarly to the horizontal autocorrelation coefficient. The above is the interpolation parameter selection process.

  Next, using these interpolation parameters, the anomaly data is estimated as grid data at a desired grid interval and depth in the four-dimensional optimal interpolation calculation process (step B3). The grid data of the anomaly thus output is output as water temperature grid estimation data in the water temperature conversion process (step B4) using the water temperature climate data.

  The processing apparatus 10 assimilates the water temperature lattice estimation data obtained by the four-dimensional optimal interpolation method by performing a known nudging process on the water temperature prediction data calculated by the numerical simulation (step B5), and the corrected water temperature lattice data. Output as. FIG. 3 shows a processing procedure in the case of assimilating an estimated value generated by the optimal interpolation method into a numerical ocean model.

  FIG. 4 is a schematic diagram showing the concept of the optimum interpolation method performed in the vertical direction.

  The vertical direction in FIG. 4 is the vertical direction, the horizontal direction is the horizontal direction (latitude and longitude directions), and the three vertical lines shown in FIG. 4 indicate the vertical distribution of the observation data. Observation data A and observation data B are data after interpolation of each data at the singular point S1, and show how the water temperature fluctuates. The thick center line indicates the vertical grid position of the estimated data estimated by the optimal interpolation method.

  A horizontal dotted line indicates an estimated depth in the optimal interpolation method, and a position where the dotted line and the observation data A and B intersect is an interpolated observation point S2 used in the three-dimensional interpolation method. On the other hand, the position where the same dotted line and the thick line intersect is assumed to be the estimated point S3 optimally interpolated by the vertical interpolation point of the singular point.

  When estimating the water temperature between the water temperature distributions as indicated by the observation data A and B, in the three-dimensional interpolation method, only the interpolated observation point S2 is interpolated in the horizontal direction (here, one estimated data is 2). Since there is no relation between the representative depths, the complete water temperature profile cannot be reproduced from the optimally interpolated estimated point S3 unless the representative depth is set finely.

  On the other hand, the four-dimensional interpolation method does not use the observation point S2 interpolated at the representative depth, but generates the estimated point S3 using a plurality of singular points S1 close to the representative depth. Can be interpolated by taking into account the water temperature changes.

  FIG. 5 is a diagram showing the concept of the three-dimensional optimum interpolation method and the four-dimensional optimum interpolation method. FIG. 5 is a diagram supplementing the difference between the two optimal interpolation methods shown in FIG. 3 as in FIG.

  The upper diagram of FIG. 5 is a diagram showing the concept of the three-dimensional interpolation method, and the lower diagram is a diagram showing the concept of the four-dimensional interpolation method. Further, the left side of FIG. 5 shows the concept of observation data, and the right side shows the concept of estimation data.

  The observation data shown in FIG. 5 corresponds to the water temperature observation data shown in FIG. 2, and the estimation data corresponds to the water temperature grid estimation data shown in FIG.

  The three-dimensional interpolation method interpolates in the horizontal direction based on observation data that has been aligned in the vertical direction by vertical interpolation in advance, while the four-dimensional interpolation method spatially interpolates the observation data as it is, Create interpolation correction data for the figure.

  In this embodiment, the observation data assimilation method of the present invention has been described by limiting to ocean water temperature data. However, the observation data assimilation method of the present invention is applicable to assimilation processing of other ocean data such as salinity data and flow velocity data. Can also be applied. In that case, since the spatio-temporal distribution differs depending on the data to be applied, it is necessary to appropriately select parameters such as the autocorrelation function and error distribution, as well as the criteria used to determine the appropriate use of the 3D optimal interpolation method and 4D optimal interpolation method. That's fine.

FIG. 1 is a block diagram showing a configuration example of an observation data assimilation apparatus according to the present invention. It is a flowchart which shows the process sequence of the observation data assimilation apparatus of this invention. It is a flowchart which shows the process sequence of the four-dimensional interpolation process shown in FIG. It is a schematic diagram which shows the concept of the optimal interpolation method implemented in a perpendicular direction. It is a figure which shows the concept of 3D optimal interpolation method and 4D optimal interpolation method

Explanation of symbols

10 processor 11 CPU
12 main storage device 13 recording medium 14 data storage device 15 memory control interface unit 16 I / O interface unit 17 communication control device 18 bus 20 input device 30 output device

Claims (10)

An observation data assimilation method for associating observation data with data obtained from a numerical model,
Using the average standard deviation representing the magnitude of the average fluctuation of the observation data and the statistical data of the acquisition sea area of the observation data to determine the standard standard deviation used to determine the magnitude of the vertical data fluctuation,
When the average standard deviation is less than or equal to the reference standard deviation, the vertical direction of the observation data is first interpolated, and then interpolated using a three-dimensional optimal interpolation method that optimally interpolates in the horizontal and temporal directions. Run
When the average standard deviation is larger than the standard standard deviation, an interpolation process is performed using a four-dimensional optimal interpolation method that optimally interpolates in the horizontal direction, vertical direction, and time direction of the observation data,
An observation data assimilation method that assimilates the data after the three-dimensional optimal interpolation method or the four-dimensional optimal interpolation method into data obtained from the numerical model.   In the four-dimensional optimal interpolation method, an autocorrelation coefficient and an error variance value including a spatio-temporal scale as a variable are determined in advance according to the sea area and time, and are correlated with data obtained from the numerical model. The observation data assimilation method according to claim 1, wherein a value at each depth is estimated using observation data adjacent in the vertical direction of the estimated depth. When the horizontal function is H (lon _i , lat _i , lon _j , lat _j ), the time direction function is T (t _i , t _j ), and the vertical function is V (dep _m , dep _n ). ,
The autocorrelation coefficient μ _{ij, mn} is

The observation data assimilation method according to claim 2, which is calculated using In the mean standard deviation,
The observation data assimilation method according to claim 1, wherein the average standard deviation of the mixed layer characterized by the water temperature fluctuation near the surface layer and divided by the water temperature fluctuation characteristics in the vertical direction is used.   The spatial map for each horizontal area of the horizontal spatial scale, time scale, and error variance is created in advance, and parameters used in the optimal interpolation method are determined for each observation data using the map. Observation data assimilation method described in the section. An observation data assimilation device that assimilate observation data into data obtained from a numerical model,
Using the average standard deviation representing the average fluctuation magnitude of the observation data and the statistical data of the sea area where the observation data was acquired, a standard standard deviation used to determine the magnitude of vertical data fluctuation is obtained, and the average standard If the deviation is less than or equal to the standard standard deviation, the vertical direction of the observation data is interpolated first, and then an interpolation process is performed using a three-dimensional optimal interpolation method that optimally interpolates in the horizontal and temporal directions. When the average standard deviation is larger than the reference standard deviation, an interpolation process is performed using a four-dimensional optimal interpolation method that optimally interpolates in the horizontal direction, vertical direction, and time direction of the observation data, A processing apparatus for assimilating the data after the three-dimensional optimal interpolation method or the four-dimensional optimal interpolation method into the data obtained from the numerical model;
A data storage device for storing the observation data, the statistical data, and data obtained from the numerical model;
Observation data assimilation device. The processor is
In the four-dimensional optimal interpolation method, an autocorrelation coefficient and an error variance value including a spatio-temporal scale as a variable are determined in advance according to the sea area and time, and are correlated with data obtained from the numerical model. The observation data assimilation apparatus according to claim 6, wherein a value at each depth is estimated using observation data adjacent in the vertical direction of the estimated depth. The processor is
When the horizontal function is H (lon _i , lat _i , lon _j , lat _j ), the time direction function is T (t _i , t _j ), and the vertical function is V (dep _m , dep _n ). ,
The autocorrelation coefficient μ _{ij, mn} is

The observation data assimilation apparatus according to claim 7, wherein the observation data assimilation apparatus is calculated using The processor is
In the mean standard deviation,
7. The observation data assimilation apparatus according to claim 6, wherein the average standard deviation of the mixed layer characterized by the water temperature fluctuation near the surface layer, which is divided by the water temperature fluctuation characteristics in the vertical direction, is used. The processor is
10. A spatial map for each sea area of a horizontal spatial scale, a time scale, and an error variance is created in advance, and parameters used in the optimum interpolation method are determined for each observation data using the map. Observation data assimilation device described in the section.

Images