Weather forecasting device, weather forecasting system, and weather forecasting method

Abstract

The weather forecasting device (1) includes an observation data acquisition unit (2) that acquires grid data obtained by dividing a predetermined area including a prediction target area into a plurality of grids, the grid data including meteorological data observed for each divided grid; a non-uniform grid construction unit (4) that subdivides a predetermined grid among the grids indicated by the grid data acquired by the observation data acquisition unit and merges the predetermined grid among the grids indicated by the grid data with a grid adjacent to the predetermined grid to generate non-uniform grid data including meteorological data for each grid, the non-uniform grid data being data indicating a non-uniform grid in which subdivided grids and merged grids are mixed; and a non-uniform grid prediction processing unit (5) that predicts meteorological data for the prediction target area included in the predetermined area using the non-uniform grid data generated by the non-uniform grid construction unit.

Description

Technical Field

[0001] The present disclosure relates to a weather prediction device, a weather prediction system, and a weather prediction method.

Background Art

[0002] Conventionally, a weather prediction device for predicting the weather in a prediction target area is known. In a weather prediction system using this weather prediction device, future weather conditions such as rain and wind in the prediction target area are predicted from weather data such as temperature, humidity, and atmospheric pressure observed in the prediction target area.

[0003] In order to capture local heavy rainfall and tornadoes, etc., information on grids (meshes) obtained by finely dividing the prediction target area is necessary. However, in order to perform long-term predictions, it is necessary to consider information on a relatively wide area around the grids. However, generally, when performing wide-area predictions using fine grids, a huge amount of data needs to be handled by a computer. Therefore, not only does the scale of the computer increase, but the calculation time also increases.

[0004] Therefore, in a conventional weather prediction system, a technique has been proposed to shorten the calculation time required for prediction by a method called a nesting method, in which first the overall weather situation is predicted, and then the weather situation at a specific local point is predicted.

[0005] For example, Patent Document 1 discloses a system for predicting how a diffused substance emitted from a point of interest diffuses due to local airflow elements. In this system, in order to predict the diffusion state of the diffused substance emitted from the point of interest, a local area A1 including the point of interest, a middle area A2 including the local area A1, and a large area A3 including the middle area A2 (A1 < A2 < A3) are set. Then, airflow calculations in the middle area A2 and the large area A3 wider than the local area A1 are performed, and the airflow in the local area A1 is predicted using the results, thereby reducing the processing load and shortening the processing time.

Prior Art Documents

Patent Documents

[0006] [Patent Document 1] JP 2008-89418 A Summary of the Invention [Problem to be solved by the invention]

[0007] In the system described in Patent Document 1 (hereinafter also referred to as the "conventional system"), evaluation points are set in advance in a grid pattern at predetermined intervals for each of the local area A1, the medium area A2, and the large area A3. In the conventional system, airflow data is obtained for each of the evaluation points set for each of the above areas, and the airflow data is used to perform diffusion calculations to predict the diffusion state of a diffused substance released from a point of interest.

[0008] However, in conventional systems, the evaluation points used to predict the diffusion status are fixedly positioned at predetermined intervals for each of the above-mentioned areas, which has the problem that the system is not necessarily configured sufficiently to improve the accuracy of the diffusion status prediction.

[0009] The present disclosure has been made to solve the above-mentioned problems, and aims to provide a weather forecasting device that can improve the accuracy of forecasting weather conditions while suppressing an increase in processing load compared to conventional devices. [Means for solving the problem]

[0010] The weather forecasting device of the present disclosure includes a first acquisition unit that acquires grid data obtained by dividing a specified area including a prediction target area into a plurality of grids, the grid data including weather data observed for each divided grid; a generation unit that subdivides a specified grid among the grids indicated by the grid data acquired by the first acquisition unit and merges the specified grid among the grids indicated by the grid data with a grid adjacent to the specified grid, thereby generating non-uniform grid data including weather data for each grid, the non-uniform grid data including weather data for each grid; and a prediction processing unit that predicts weather data for the prediction target area included in the specified area using the non-uniform grid data generated by the generation unit. Effect of the Invention

[0011] According to the present disclosure, it is possible to improve the accuracy of weather condition predictions while suppressing an increase in processing load compared to the conventional art. [Brief description of the drawings]

[0012] [Figure 1] 1 is a diagram showing an example of the configuration of a weather prediction system including a weather prediction device according to a first embodiment. [Diagram 2] 4 is a flowchart illustrating an example of the operation of the weather prediction device according to the first embodiment. [Diagram 3] FIG. 3A is a diagram showing an example of observation data in embodiment 1, FIG. 3B is a diagram showing an example of selection by a region selection unit in embodiment 1, and FIG. 3C is a diagram showing an example of a non-uniform grid in embodiment 1. [Figure 4] FIG. 4 is a diagram showing an example of adjacent grid determination data in the first embodiment. [Diagram 5] FIG. 4 is a diagram showing an example of an interpolation process performed by a non-uniform grid prediction processor in the first embodiment. [Figure 6] 6A and 6B are diagrams illustrating an example of a hardware configuration of the weather prediction device according to the first embodiment. [Figure 7]FIG. 11 is a diagram showing a configuration example of a weather prediction system including a weather prediction device according to a second embodiment. [Figure 8] 10 is a flowchart illustrating an example of the operation of the weather prediction device according to the second embodiment. [Figure 9] FIG. 11 is a diagram showing a configuration example of a weather prediction system including a weather prediction device according to a third embodiment. [Figure 10] 13 is a flowchart illustrating an example of the operation of the weather prediction device according to the third embodiment. [Figure 11] FIG. 11 is a diagram showing a configuration example of a weather prediction system including a weather prediction device according to a fourth embodiment. [Figure 12] 13 is a flowchart illustrating an example of the operation of the weather prediction device according to the fourth embodiment. [Figure 13] FIG. 13 is a diagram showing an example of selection by a region selection unit in the fourth embodiment. [Figure 14] FIG. 13 is a diagram showing a configuration example of a weather prediction system including a weather prediction device according to a fifth embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0013] Hereinafter, embodiments of the present disclosure will be described in detail with reference to the drawings. Embodiment 1

[0014] Fig. 1 is a diagram showing a configuration example of a weather prediction system 100 including a weather prediction device 1 according to embodiment 1. The weather prediction system 100 includes an observation device 10, an observation data recording unit 15, a division setting value data recording unit 16, and the weather prediction device 1, for example, as shown in Fig. 1.

[0015] <Observation Equipment 10> The observation device 10 is a device that observes weather conditions (weather data) in a predetermined area including an area that is the subject of weather forecasting (hereinafter also referred to as a "prediction target area"). The observation device 10 is installed, for example, at an arbitrary location within the predetermined area including the prediction target area, or at an arbitrary location in the vicinity of the predetermined area.

[0016] The observation device 10 observes the weather conditions in a specific area, for example, at a preset time, and generates observation data D1. The observation data D1 is, for example, grid data obtained by dividing the specific area into a plurality of grids. The grid data includes weather data observed for each grid.

[0017] For example, if the above-mentioned predetermined area is divided into a total of nine grids of three rows and three columns, and each grid is expressed as grid (i, j), the observation data D1 includes the following meteorological data for each grid: Grid(1,1)=(Sunny, North wind, Temperature 15℃, Humidity 26%, ...), Grid(1,2)=(sunny, east wind, temperature 17℃, humidity 30%, ...), Grid(1,3)=(Cloudy, West wind, Temperature 12℃, Humidity 45%, ...), Grid(2,1)=(Sunny, North wind, Temperature 16℃, Humidity 28%, ...), Grid(2,2)=(sunny, east wind, temperature 14℃, humidity 32%, ...), Grid(2,3)=(Cloudy, West wind, Temperature 13℃, Humidity 48%, ...), Grid(3,1)=(Sunny, North wind, Temperature 11℃, Humidity 25%, ...), Grid(3,2)=(sunny, east wind, temperature 13℃, humidity 33%, ...), It includes data such as grid (3,3) = (cloudy, westerly wind, temperature 10°C, humidity 50%, ...). The observation device 10 generates observation data D1 including meteorological data for each grid such as those described above, and causes the generated observation data D1 to be recorded in the observation data recording unit 15.

[0018] The observation data recording unit 15 is configured with a recording medium such as a hard disk drive (HDD) and a solid state drive (SDD), etc. The observation data recording unit 15 records the observation data D1 generated by the observation device 10.

[0019] The divided setting value data recording unit 16 is configured with a recording medium such as a hard disk drive (HDD) and a solid state drive (SDD), etc. The divided setting value data recording unit 16 records the divided setting value data D2.

[0020] The division setting value data D2 is a parameter used when the non-uniform grid construction unit 4, which will be described later, performs grid conversion. Specifically, the division setting value data D2 is data (hereinafter also referred to as "conversion target data") that specifies a grid that is to be subjected to grid conversion among a plurality of grids included in the above-mentioned predetermined region. Note that, as will be described in detail later, grid conversion refers to making a grid finer (subdividing) or making a grid coarser (integrating)

[0021] For example, the division setting value data D2 is data that specifies grids to be refined (divided) and grids to be coarsened (merged). Note that, in this case, "fining a grid" refers to subdividing an area inside a grid, and "coarsening a grid" refers to combining a grid with an adjacent grid to combine them into a single grid. In the following description, a grid to be refined is also referred to as a "grid to be subdivided," and a grid to be coarsened is also referred to as a "grid to be merged."

[0022] For example, in the case of the nine grids in total of three rows and three columns described above, in the division setting value data D2, grid (1,1), grid (1,3), and grid (2,1) are specified as grids to be subdivided, and grid (3,3) is specified as a grid to be integrated. The division setting value data D2 is generated in advance, for example, by a user of the weather forecasting system 100 (hereinafter also simply referred to as "user"), and recorded in the division setting value data recording unit 16.

[0023] Here, "subdividing (dividing) the area inside a grid" means to further subdivide a grid and perform subsequent processing to predict weather data in the area to be predicted with greater accuracy.

[0024] For example, for a grid including a prediction target area, the user designates this grid as a grid to be subdivided in the division setting value data D2. Also, if there is an area where meteorological elements (such as typhoons and yellow sand) that have a significant impact on weather forecasts in the prediction target area may occur, the user also designates this grid as a grid to be subdivided in the division setting value data D2 for grids including such areas (such as areas frequently hit by typhoons and desert areas where yellow sand is taken into consideration). In the following description, an area where meteorological elements that have a significant impact on weather forecasts in the prediction target area may occur is also referred to as an "affected area."

[0025] On the other hand, "coarsening (merging) the area inside a grid" means merging a grid with its adjacent grid and then performing subsequent processing, thereby performing prediction processing while suppressing increases in processing load and computational costs.

[0026] For example, if the specified area includes an area with little change in weather data and flat land, a highly accurate prediction based on a fine grid is not necessary for such an area, and a prediction based on a grid of a certain width may be sufficient. Therefore, for a grid including an area with little change in weather data or an area with little change over time in weather data, the user specifies this grid as a grid to be integrated in the division setting value data D2. Note that "an area with little change in weather data" or "an area with little change over time in weather data" refers to, for example, an area in which the amount of change in weather data in a predetermined time interval (e.g., 5 minutes) is equal to or less than a predetermined value. In the following description, such an area is also referred to as a "monotonic area."

[0027] <Weather forecasting device 1> As shown in FIG. 1, the weather forecasting device 1 includes an observation data acquisition unit 2 (first acquisition unit), a division setting value data acquisition unit 3 (second acquisition unit), a non-uniform grid construction unit 4 (generation unit), a non-uniform grid prediction processing unit 5 (prediction processing unit), and an adjacent grid determination data recording unit 6.

[0028] The observation data acquiring unit 2 acquires observation data D1 (grid data) from the observation data recording unit 15. The observation data acquiring unit 2 outputs the acquired observation data D1 to the non-uniform grid constructing unit 4.

[0029] The divided set value data acquisition unit 3 acquires divided set value data D2 from the divided set value data recording unit 16. The divided set value data acquisition unit 3 outputs the acquired divided set value data D2 to the non-uniform grid construction unit 4. Note that here, in the divided set value data D2, for example, a grid including a prediction target region and a grid including an affected region are designated as grids to be subdivided, and a grid including a monotonic region is designated as a grid to be integrated.

[0030] The non-uniform grid construction unit 4 acquires the observation data D1 from the observation data acquisition unit 2. The non-uniform grid construction unit 4 also acquires the division setting value data D2 from the division setting value data acquisition unit 3. The non-uniform grid construction unit 4 then subdivides a specific grid among the grids indicated by the observation data D1, and integrates the specific grid among the grids indicated by the observation data D1 with a grid adjacent to the specific grid, thereby constructing a grid (hereinafter also referred to as a "non-uniform grid") in which the subdivided grids and the integrated grids are mixed, and generates data (hereinafter also referred to as "non-uniform grid data") indicating the non-uniform grid. At this time, the non-uniform grid construction unit 4 selects a grid to be subdivided (grid to be subdivided) and a grid to be integrated (grid to be integrated) based on the division setting value data D2 acquired from the division setting value data acquisition unit 3, for example.

[0031] <Non-uniform grid construction part 4> The non-uniform grid constructing unit 4 includes a region selecting unit 41, a selected region grid converting unit 42, and a grid reconstructing unit 43, as shown in FIG.

[0032] The region selection unit 41 selects a subdivision target grid and an integration target grid from the grids indicated by the observation data D1 based on the division setting value data D2. The region selection unit 41 outputs data indicating the selection result to the selected region grid conversion unit 42.

[0033] The selected area grid conversion unit 42 obtains data indicating the selection result from the area selection unit 41, and performs grid conversion based on the obtained data. Specifically, the selected area grid conversion unit 42 temporarily extracts, for example, the subdivision target grid selected by the area selection unit 41 from the observation data D1, and subdivides the extracted subdivision target grid into a predetermined number of grids. The selected area grid conversion unit 42 also temporarily extracts, for example, the integration target grid selected by the area selection unit 41 from the observation data D1, and integrates the extracted integration target grid with an adjacent grid. The selected area grid conversion unit 42 outputs data indicating the result of the grid conversion (subdivision and integration) to the grid reconstruction unit 43.

[0034] The selected area grid conversion unit 42 also generates adjacent grid determination data D3 indicating the adjacency relationship between the grids in the observation data D1 (grid data) after the grid conversion. The selected area grid conversion unit 42 then causes the generated adjacent grid determination data D3 to be recorded in the adjacent grid determination data recording unit 6.

[0035] The adjacent grid determination data recording unit 6 is configured with a recording medium such as a hard disk drive (HDD) or a solid state drive (SDD), etc. The adjacent grid determination data recording unit 6 records the adjacent grid determination data D3.

[0036] The grid reconstruction unit 43 acquires data indicating the result of the grid conversion from the selected area grid conversion unit 42. Based on the acquired data, the grid reconstruction unit 43 constructs a grid (non-uniform grid) in which subdivided grids and integrated grids are mixed, and generates data (non-uniform grid data) indicating the non-uniform grid. The grid reconstruction unit 43 outputs the generated non-uniform grid data to the non-uniform grid prediction processing unit 5.

[0037] The non-uniform grid prediction processor 5 acquires non-uniform grid data from the grid reconstruction unit 43. The non-uniform grid prediction processor 5 predicts weather data for a prediction target area included in a predetermined area, using the acquired non-uniform grid data and adjacent grid determination data D3 recorded in the adjacent grid determination data recording unit 6. Then, the non-uniform grid prediction processor 5 outputs the predicted weather data for the prediction target area as a prediction result.

[0038] At this time, in the non-uniform grid indicated by the non-uniform grid data, the grid including the prediction target area is subdivided, and the grid including the monotonic area is integrated with the adjacent grid. Therefore, the non-uniform grid prediction processor 5 can compensate for the increase in the processing load in the subdivided grid by reducing the processing load in the integrated grid. As a result, the non-uniform grid prediction processor 5 can perform highly accurate weather prediction in the prediction target area while suppressing the increase in the processing load of the prediction process as a whole.

[0039] In addition, in the non-uniform grid, the grid including the affected area is also subdivided. This allows the non-uniform grid prediction processor 5 to accurately reflect the influence of meteorological elements occurring in the affected area on the prediction target area, thereby further improving the accuracy of the weather prediction in the prediction target area.

[0040] Next, an operation example of the weather forecasting device 1 according to the first embodiment will be described with reference to the flowchart shown in FIG.

[0041] For simplicity's sake, in the following explanation, it is assumed that the observation data acquisition unit 2 of the weather forecasting device 1 acquires observation data D1 in advance from the observation data recording unit 15 and outputs the acquired observation data D1 to the non-uniform grid construction unit 4. Similarly, it is assumed that the division setting value data acquisition unit 3 of the weather forecasting device 1 acquires division setting value data D2 in advance from the division setting value data recording unit 16 and outputs the acquired division setting value data D2 to the non-uniform grid construction unit 4.

[0042] In the following description, it is assumed that in the division setting value data D2, the grid including the prediction target region and the grid including the influence region are specified as grids to be subdivided, and the grid including the monotonic region is specified as grid to be integrated.

[0043] First, the region selection unit 41 of the non-uniform grid constructing unit 4 selects a subdivision target grid and an integration target grid from the grid indicated by the observation data D1 based on the division setting value data D2 (step ST01).

[0044] An example of processing by the non-uniform grid constructing unit 4 including the area selecting unit 41 is shown in Fig. 3. Fig. 3A shows an example of observation data D1 (grid data) acquired from the observation data recording unit 15. Note that, as described above, each grid shown in the observation data D1 shown in Fig. 3A contains meteorological data for each grid. In Fig. 3A, differences in this meteorological data are expressed by the shading of the grids.

[0045] The region selection unit 41 selects grids to be subdivided and grids to be integrated from the grids indicated by the observation data D1, as shown in Fig. 3B, for example, based on the division setting value data D2 acquired from the division setting value data recording unit 16. For example, in Fig. 3B, the grids surrounded by thick lines indicate grids selected as grids to be subdivided or grids to be integrated.

[0046] Next, the selected area grid conversion unit 42 performs grid conversion on the grid selected by the area selection unit 41 (step ST02). Specifically, the selected area grid conversion unit 42 subdivides the subdivision target grid selected by the area selection unit 41, and integrates the integration target grid selected by the area selection unit 41.

[0047] 3B, the selected area grid conversion unit 42 first extracts the subdivision target grid from the observation data D1, and then subdivides the extracted subdivision target grid into, for example, four grids. Note that the number of grids into which the selected area grid conversion unit 42 subdivides the subdivision target grid can be arbitrarily set by the user in, for example, the division setting value data D2.

[0048] Although not shown in FIG. 3B, the selected area grid conversion unit 42 may integrate the integration target grid in the same manner as when subdividing the subdivision target grid. For example, the selected area grid conversion unit 42 may temporarily extract the integration target grid and the grids (adjacent grids) adjacent to the integration target grid in the east-west-north-south (up-down-left-right) directions from the observation data D1, and combine the extracted grids into one grid. In addition to the above, the selected area grid conversion unit 42 may temporarily extract the grids adjacent to the integration target grid in the diagonal direction from the observation data D1, and combine the extracted grids into one grid. Note that the user can arbitrarily set in the division setting value data D2 the extent to which the integration target grid is to be combined with the adjacent grids by the selected area grid conversion unit 42, including the above example.

[0049] After performing the grid conversion as described above, the selected area grid conversion unit 42 needs to supplement the weather data in the converted grid. In this case, for example, the selected area grid conversion unit 42 may instruct the observation data acquisition unit 2 to reacquire weather data in the converted grid from the observation device 10. Upon receiving this instruction, the observation data acquisition unit 2 reacquires weather data in the converted grid from the observation device 10 and outputs the reacquired weather data to the selected area grid conversion unit 42. Then, the selected area grid conversion unit 42 may supplement the weather data in the converted grid with this reacquired weather data.

[0050] Alternatively, the selected area grid conversion unit 42 may supplement the meteorological data in the converted grid by interpolating based on the meteorological data in the grids adjacent to the converted grid. After supplementing the meteorological data as described above, the selected area grid conversion unit 42 outputs data indicating the result of the grid conversion, including the supplemented meteorological data, to the grid reconstruction unit 43.

[0051] At this time, the selected area grid conversion unit 42 generates adjacent grid determination data D3 and records the generated adjacent grid determination data D3 in the adjacent grid determination data recording unit 6. As described above, the adjacent grid determination data D3 is data that indicates the adjacency relationship between each grid in the grid indicated by the observation data D1 (i.e., non-uniform grid data) after the selected area grid conversion unit 42 has performed grid conversion.

[0052] An example of the adjacent grid determination data D3 will be described with reference to Fig. 4. Fig. 4 shows an example of a non-uniform grid. For ease of explanation, the grids to be integrated are not shown in Fig. 4. Also, for ease of explanation, a number is assigned to each grid in Fig. 4 to identify each grid.

[0053] In the case of the example of the non-uniform grid shown in FIG. 4, the adjacent grid determination data D3 is, for example, Grids adjacent to grid 1 = (grid 61, grid 62, and grid 2), Grids adjacent to grid 2 = (grid 1, grid 71, grid 72, and grid 3), Grids adjacent to Grid 3 = (Grid 2, Grid 8, Grid 41, and Grid 43), Grids adjacent to grid 41 = (grid 3, grid 42, and grid 43), Grids adjacent to grid 42 = (grid 41, grid 44, and grid 51), Grids adjacent to grid 43 = (grid 3, grid 41, grid 44, and grid 91), Grids adjacent to grid 44 = (grid 42, grid 43, grid 53, and grid 92), Grids adjacent to grid 51 = (grid 42, grid 52, and grid 53), The grids adjacent to grid 52 are generated as follows: grid 51 and grid 54.

[0054] Next, the grid reconstruction unit 43 constructs a non-uniform grid in which a subdivided grid and a merged grid are mixed, for example, as shown in Fig. 3C, based on the data indicating the result of the grid conversion obtained from the selected area grid conversion unit 42, and generates non-uniform grid data indicating the non-uniform grid (step ST03). Note that in the example of Fig. 3C, the subdivided grid is subdivided into four grids. Also, in the example of Fig. 3C, the grid indicated by the blank space at the bottom right is the merged grid.

[0055] For example, as shown in Fig. 3C, the grid reconstructing unit 43 returns the grid divided by the selected area grid converting unit 42 to its original position in the observation data D1 and combines it with another grid. In this case, it is preferable that the grid reconstructing unit 43 stores in advance the original position of the divided grid in the observation data D1.

[0056] 3C illustrates an example in which the grid reconstructing unit 43 returns the subdivided grid to its original position in the observation data D1, but the same applies to the case in which the grid reconstructing unit 43 returns the integrated grid to its original position in the observation data D1. The grid reconstructing unit 43 outputs the non-uniform grid data generated as described above to the non-uniform grid prediction processing unit 5.

[0057] In the above example, the selected area grid conversion unit 42 temporarily extracts the subdivision target grids and integration target grids selected by the area selection unit 41 from the observation data D1, subdivides and integrates the extracted grids, and the grid reconstruction unit 43 returns the subdivided and integrated grids to their original positions in the observation data D1. However, the selected area grid conversion unit 42 is not limited to this, and may construct a non-uniform grid by subdividing the subdivision target grids and the like selected by the area selection unit 41 at their original positions without extracting them from the observation data D1, and generate non-uniform grid data. In this case, the grid reconstruction unit 43 may be omitted.

[0058] In the above example, the selected area grid conversion unit 42 generates the adjacent grid judgment data D3 and records it in the adjacent grid judgment data recording unit 6. However, this is not limited to the above example. For example, the grid reconstruction unit 43 may generate the adjacent grid judgment data D3 and record it in the adjacent grid judgment data recording unit 6.

[0059] Next, the non-uniform grid prediction processing unit 5 predicts weather data for the prediction target area included in the specified area using the non-uniform grid data acquired from the grid reconstruction unit 43 and the adjacent grid determination data D3 recorded in the adjacent grid determination data recording unit 6 (step ST04).

[0060] Here, the weather forecasting method by the non-uniform grid forecast processor 5 will be described in comparison with a conventional weather forecasting method using grid data.

[0061] For example, conventional weather forecasts are based on the premise of using data (hereinafter also referred to as "uniform grid data") indicating grids (hereinafter also referred to as "uniform grids") in which each grid is basically configured to be adjacent to the others on a one-to-one basis, and weather forecasts for the area to be forecasted are made by referencing the weather data for each grid in this uniform grid.

[0062] For example, in conventional weather forecasts, weather forecasts are made by repeating a calculation called a seven-point stencil calculation, which uses a certain grid in a uniform grid, four grids adjacent to the grid in the east-west-north (top-bottom-left-right) directions, and two grids adjacent to the grid in the height direction, for a total of seven grids. Specifically, for example, in conventional weather forecasts, the weather data (temperature, humidity, etc.) in each of these seven grids is multiplied by a predetermined weighting factor, and then predetermined calculations such as addition and multiplication are performed on each of these weather data. In conventional weather forecasts, this calculation is repeated in order so that all grids included in the grid data are subject to calculation, thereby predicting the weather data of the forecast target area.

[0063] On the other hand, in a non-uniform grid in which a subdivided grid and an integrated grid are mixed, the adjacent relationship between the grids is not necessarily one-to-one. For example, as shown in FIG. 4, as a result of grid conversion by the selected area grid conversion unit 42, a state may occur in which multiple grids are adjacent to one side of the integrated grid (for example, on the east side). In the following description, such a relationship in which multiple grids are adjacent to one side of a single grid is also referred to as a "non-uniform adjacent relationship." In this case, the non-uniform grid prediction processing unit 5 cannot perform processing using the conventional method as described above.

[0064] Therefore, the non-uniform grid prediction processor 5 identifies pairs of grids that have a non-uniform adjacent relationship by referring to the adjacent grid determination data D3 recorded in the adjacent grid determination data recording unit 6. Then, the non-uniform grid prediction processor 5 performs an interpolation process on the identified pairs of grids, thereby substantially eliminating the non-uniform adjacent relationship.

[0065] An example of the interpolation process by the non-uniform grid prediction processor 5 is shown in Fig. 5. For example, in Fig. 5, reference numeral 501 is an integrated grid (a check mark), and reference numerals 502 to 504 are grids (circles, triangles, and squares, respectively) adjacent to the left of grid 501. In this case, the size of the grids is grid 501>grid 502>grid 503=grid 504.

[0066] In this case, the non-uniform grid prediction processor 5 determines whether there are multiple grids adjacent to the grid 501 in the east-west, north-south (up-down, left-right) directions based on the adjacent grid determination data D3. For example, in the example of Fig. 5, three grids (grids 502 to 504) are adjacent to the west (left) side of the grid 501. In this case, the non-uniform grid prediction processor 5 determines that these grids 501 to 504 have a non-uniform adjacent relationship.

[0067] Then, as shown in FIG. 5, for example, the non-uniform grid prediction processor 5 first interpolates the grids 502 to 504 to their midpoint positions, and as a result, obtains a new grid 505 (x).

[0068] 5 is a graph that diagrammatically illustrates meteorological data (such as temperature, humidity, and atmospheric pressure) in grids 502 to 504. The "o", "△", and "□" on the horizontal axis of graph 506 correspond to the symbols "o", "△", and "□" attached to each grid, respectively, and the "x" on the horizontal axis of graph 506 corresponds to the new grid "x" obtained as a result of interpolation.

[0069] For example, if the weather data is temperature and the temperature in grid 504 marked with the symbol "□" is 20°C, the temperature in grid 503 marked with the symbol "△" is 19°C, and the temperature in grid 502 marked with the symbol "○" is 16°C, the non-uniform grid prediction processing unit 5 will interpolate these temperatures and determine the temperature in grid 505 marked with the symbol "X" to be 18°C.

[0070] As a result, the grid 501 marked with a check mark has only one adjacent grid 505 marked with a symbol "x" to the left. The non-uniform grid prediction processor 5 then performs this process sequentially for pairs of grids in a non-uniform adjacent relationship. This allows the non-uniform grid prediction processor 5 to perform calculations by the conventional method using the non-uniform grid data acquired from the grid reconstruction unit 43.

[0071] The above process will be explained based on the flowchart in Fig. 2. When each grid in the non-uniform grid is represented as (i, j), the non-uniform grid prediction processing unit 5 refers to each grid one by one while counting up i and j from 1 (step ST41). Note that i is a variable indicating the position of each grid in the east-west direction (left-right direction), and j is a variable indicating the position of each grid in the north-south direction (up-down direction).

[0072] Next, the non-uniform grid prediction processing unit 5 calculates the number of grids adjacent to the grid (i, j) in each of the north-south, east-west directions based on the adjacent grid determination data D3 (step ST42).

[0073] Then, the non-uniform grid prediction processor 5 judges whether or not there are a plurality of grids adjacent to the grid (i, j) in each of the north, south, east and west directions (step ST43). As a result, if it is judged that there are a plurality of grids adjacent to the grid (i, j) in each of the north, south, east and west directions (step ST43; YES), the non-uniform grid prediction processor 5 interpolates those grids by the interpolation process as described above, and converts them into one grid (step ST44). Then, the process returns to step ST41.

[0074] On the other hand, if it is determined that the number of grids adjacent to the grid (i, j) in each of the north, south, east and west directions is not plural (step ST43; NO), the process returns to step ST41.

[0075] Thereafter, the non-uniform-grid prediction processor 5 repeats the processes of steps ST41 to ST44 for all grids. After that, the process proceeds to step ST45, and the non-uniform-grid prediction processor 5 predicts weather data in the prediction target area using the non-uniform-grid data for which the processes of steps ST41 to ST44 have been completed (step ST45).

[0076] In the weather forecasting device 1, even when the non-uniform grid prediction processor 5 performs the above-mentioned interpolation process, there is a possibility that grid conversion will be performed again on the grid before the interpolation (for example, grids 502 to 504 shown in FIG. 5). Therefore, even when the non-uniform grid prediction processor 5 performs the above-mentioned interpolation process, it is preferable that the data related to the grid before the interpolation is left as it is, rather than being deleted from the observation data D1. Also, for the same reason as above, it is preferable that the non-uniform grid prediction processor 5 does not update the adjacent grid determination data D3 that was generated before the interpolation.

[0077] Furthermore, when performing the above-mentioned interpolation processing, the non-uniform grid prediction processor 5 may add newly obtained data (weather data, etc.) related to the grid 505 to the observation data D1 from the viewpoint of efficiency. Furthermore, in the above case, the non-uniform grid prediction processor 5 may add adjacent information between the grids after the interpolation (for example, information indicating that the grid 505 and the grid 501 shown in FIG. 5 are adjacent to each other) to the adjacent grid determination data D3 from the viewpoint of efficiency.

[0078] The non-uniform grid prediction processor 5 can predict weather data of the prediction target area using the non-uniform grid data by performing the above-mentioned processing. In particular, in the above example, in the non-uniform grid indicated by the non-uniform grid data, the grid including the prediction target area is subdivided, while the grid including the monotonic area is integrated with the adjacent grid. Therefore, the non-uniform grid prediction processor 5 can compensate for the increase in the processing load in the subdivided grid by reducing the processing load in the integrated grid. As a result, the non-uniform grid prediction processor 5 can perform weather prediction in the prediction target area with high accuracy while suppressing the increase in the processing load of the prediction process as a whole.

[0079] In the above example, the grid including the affected area is also subdivided in the non-uniform grid indicated by the non-uniform grid data. This allows the non-uniform grid prediction processor 5 to accurately reflect the influence of meteorological elements occurring in the affected area on the prediction target area, thereby further improving the accuracy of the weather prediction in the prediction target area.

[0080] Next, a hardware configuration example of the weather forecasting device 1 according to the first embodiment will be described with reference to Fig. 6. The functions of the observation data acquiring unit 2, the division setting value data acquiring unit 3, the non-uniform grid constructing unit 4 (the area selecting unit 41, the selected area grid converting unit 42, and the grid reconstructing unit 43), and the non-uniform grid prediction processing unit 5 in the weather forecasting device 1 are realized by processing circuits. The processing circuit may be dedicated hardware as shown in Fig. 6A, or may be a CPU (also called a Central Processing Unit, a central processing unit, a processing unit, an arithmetic unit, a microprocessor, a microcomputer, a processor, or a DSP (Digital Signal Processor)) 52 that executes a program stored in a memory 53 as shown in Fig. 6B.

[0081] When the processing circuit is a dedicated hardware, the processing circuit 51 corresponds to, for example, a single circuit, a composite circuit, a programmed processor, a parallel programmed processor, an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array), or a combination of these. The functions of each of the observation data acquisition unit 2, the division setting value data acquisition unit 3, the non-uniform grid construction unit 4, and the non-uniform grid prediction processing unit 5 may be realized by the processing circuit 51, or the functions of each unit may be realized by the processing circuit 51 collectively.

[0082] When the processing circuit is a CPU 52, the functions of the observation data acquisition unit 2, the division set value data acquisition unit 3, the non-uniform grid construction unit 4, and the non-uniform grid prediction processing unit 5 are realized by software, firmware, or a combination of software and firmware. The software and firmware are written as programs and stored in the memory 53. The processing circuit realizes the functions of each unit by reading and executing the programs recorded in the memory 53. That is, the weather forecasting device 1 includes a memory for storing a program that, when executed by the processing circuit, results in the execution of each step shown in FIG. 2, for example. It can also be said that these programs cause a computer to execute the procedures and methods of the observation data acquisition unit 2, the division set value data acquisition unit 3, the non-uniform grid construction unit 4, and the non-uniform grid prediction processing unit 5. Here, examples of memory 53 include non-volatile or volatile semiconductor memories such as RAM (Random Access Memory), ROM (Read Only Memory), flash memory, EPROM (Erasable Programmable ROM), EEPROM (Electrically EPROM), magnetic disks, flexible disks, optical disks, compact disks, mini disks, and DVDs (Digital Versatile Discs).

[0083] It is also possible to realize some of the functions of the observation data acquisition unit 2, the division setting value data acquisition unit 3, the non-uniform grid construction unit 4, and the non-uniform grid prediction processing unit 5 with dedicated hardware and some with software or firmware. For example, the functions of the observation data acquisition unit 2 can be realized by a processing circuit as dedicated hardware, and the functions of the division setting value data acquisition unit 3, the non-uniform grid construction unit 4, and the non-uniform grid prediction processing unit 5 can be realized by the processing circuit reading and executing a program stored in the memory 53.

[0084] Thus, the processing circuitry may implement the above-described functions through hardware, software, firmware, or a combination of these.

[0085] As described above, according to the first embodiment, the weather forecasting device 1 includes an observation data acquiring unit 2 that acquires grid data (observation data D1) obtained by dividing a predetermined region including a prediction target region into a plurality of grids and including meteorological data observed for each divided grid, a non-uniform grid constructing unit 4 that subdivides a predetermined grid among the grids indicated by the grid data acquired by the observation data acquiring unit 2, and merges the predetermined grid among the grids indicated by the grid data with a grid adjacent to the predetermined grid to generate non-uniform grid data including meteorological data for each grid, the non-uniform grid data being data indicating a non-uniform grid in which the subdivided grids and the merged grids are mixed, and a non-uniform grid prediction processing unit 5 that predicts meteorological data for the prediction target region included in the predetermined region using the non-uniform grid data generated by the non-uniform grid constructing unit 4. As a result, the weather forecasting device 1 according to the first embodiment can improve the prediction accuracy of meteorological conditions while suppressing an increase in processing load compared to conventional methods.

[0086] Furthermore, the non-uniform grid construction unit 4 generates adjacent grid determination data D3 indicating the adjacent relationship of each grid in the non-uniform grid, and the non-uniform grid prediction processing unit 5 identifies a pair of grids having a non-uniform adjacent relationship from the non-uniform grid indicated by the non-uniform grid data based on the adjacent grid determination data D3 generated by the non-uniform grid construction unit 4, and performs an interpolation process on the identified pair of grids before making a prediction. This allows the weather forecasting device 1 according to the first embodiment to accurately make weather predictions using the non-uniform grid data.

[0087] Furthermore, the non-uniform grid construction unit 4 subdivides grids that include a prediction target area among the grids indicated by the grid data, and merges grids that include areas where the weather data does not change much over time with grids adjacent to the grids. This allows the weather prediction device 1 according to the first embodiment to perform highly accurate weather prediction in the prediction target area while suppressing an increase in the processing load of the prediction process as a whole.

[0088] Furthermore, the non-uniform grid constructing unit 4 subdivides grids indicated by the grid data that include areas where meteorological elements that affect the weather forecast in the prediction target area may occur. This allows the weather forecasting device 1 according to the first embodiment to accurately reflect the effects of meteorological elements occurring in the above areas in the prediction target area, thereby further improving the accuracy of the weather forecast in the prediction target area.

[0089] The weather forecasting device 1 also includes a division set value data acquisition unit 3 that acquires conversion target data (division set value data) D2, which is data indicating a subdivision target grid to be subdivided and data indicating an integration target grid to be integrated with an adjacent grid, among the grids indicated by the grid data, and the non-uniform grid construction unit 4 performs subdivision and integration based on the conversion target data D2 acquired by the division set value data acquisition unit 3. This allows the weather forecasting device 1 according to the first embodiment to accurately perform grid subdivision and integration.

[0090] The non-uniform grid construction unit 4 also includes an area selection unit 41 that selects subdivision target grids and integration target grids from the grids indicated by the grid data based on the conversion target data acquired by the division set value data acquisition unit 3, and a selected area grid conversion unit 42 that generates non-uniform grid data by subdividing the subdivision target grids selected by the area selection unit 41 and integrating the integration target grids selected by the area selection unit 41 with grids adjacent to the grids. This allows the weather forecasting device 1 according to the first embodiment to generate non-uniform grid data by appropriately selecting the subdivision target grids and integration target grids.

[0091] Furthermore, the selected area grid conversion unit 42 supplements the weather data in the subdivided and integrated grid, and the non-uniform grid prediction processing unit 5 performs prediction using the non-uniform grid data after the weather data has been supplemented by the selected area grid conversion unit 42. This allows the weather prediction device 1 according to the first embodiment to perform highly accurate weather prediction in the prediction target area based on the supplemented weather data.

[0092] In addition, the selected area grid conversion unit 42 supplements the meteorological data in the subdivided and integrated grids by interpolating the meteorological data in the grids adjacent to each of the grids. This allows the weather forecasting device 1 according to the first embodiment to easily supplement the meteorological data in the subdivided and integrated grids.

[0093] In addition, the selected area grid conversion unit 42 instructs the observation data acquisition unit 2 to acquire weather data in the subdivided and integrated grids, thereby replenishing the weather data in each grid. This allows the weather forecasting device 1 according to the first embodiment to easily replenish the weather data in the subdivided and integrated grids.

[0094] Embodiment 2 In the first embodiment, a weather forecasting device that forecasts weather data of a prediction target area included in a predetermined area using non-uniform grid data in which a subdivided grid and an integrated grid are mixed is described. In the second embodiment, a weather forecasting device that forecasts weather data of a prediction target area included in a predetermined area when not only the predetermined area but also its surrounding areas are subject to observation of weather data is described.

[0095] Fig. 7 is a diagram showing a configuration example of a weather prediction system 100b including a weather prediction device 1b according to embodiment 2. The weather prediction system 100b according to embodiment 2 differs from the weather prediction system 100 according to embodiment 1 shown in Fig. 1 in that a plurality of observation devices 10 are installed and a position data recording unit 17 is added.

[0096] Also, the weather forecasting device 1b according to the second embodiment differs from the weather forecasting device 1 according to the first embodiment shown in Fig. 1 in that a position data acquiring unit 7 (third acquiring unit) is added, and the non-uniform grid constructor 4 is changed to a non-uniform grid constructor 4b, and the non-uniform grid prediction processor 5 is changed to a non-uniform grid prediction processor 5b. Note that each component of the non-uniform grid constructor 4b is represented by adding "b" to the end of the reference numeral.

[0097] The other configurations of the weather forecasting system 100b and the weather forecasting device 1b of embodiment 2 are the same as those of the weather forecasting system 100 and the weather forecasting device 1 of embodiment 1 shown in Figure 1, so the same symbols are used and their descriptions are omitted.

[0098] As in the first embodiment, the observation device 10 is a device that observes weather conditions (weather data) in a predetermined area including the area that is the subject of weather forecasting (prediction target area). In the second embodiment, a plurality of observation devices 10 are installed, and one of the observation devices 10 observes weather conditions in a predetermined area (hereinafter also referred to as a "reference area") including the prediction target area. The other observation devices 10 observe weather conditions in predetermined areas other than the reference area (hereinafter also referred to as a "surrounding area"). The reference area and the surrounding area are, for example, adjacent (contiguous) areas.

[0099] Each observation device 10 observes weather conditions in an observation target area, for example, at a preset time, and generates observation data D1. As in the first embodiment, the observation data D1 is grid data obtained by dividing the reference area or the surrounding area into a plurality of grids, and the grid data includes weather data observed for each grid. Each observation device 10 records the generated observation data D1 in the observation data recording unit 15.

[0100] The divided set value data recording unit 16 records the divided set value data D2. In the second embodiment, the divided set value data D2 is generated by, for example, a user for each of the reference region and the peripheral region, and is recorded in the divided set value data recording unit 16.

[0101] The position data recording unit 17 is configured with a recording medium such as a hard disk drive (HDD) and a solid state drive (SDD), etc. The position data recording unit 17 records the position data D4.

[0102] The position data D4 is data indicating the respective positions of the reference area and the peripheral area. For example, when the reference area and the peripheral area are each rectangularly partitioned, the position data D4 includes data indicating the latitude and longitude of the center point of the rectangularly partitioned area and the latitude and longitude of the four corner points of the rectangularly partitioned area. The position data D4 is generated in advance by the user and recorded in the position data recording unit 17. Note that the position data D4 shown here is merely an example, and the position data D4 may include data other than the above as long as it is data indicating the respective positions of the reference area and the peripheral area.

[0103] The position data acquiring unit 7 acquires the position data D4 from the position data recording unit 17. The position data acquiring unit 7 outputs the acquired position data D4 to the non-uniform grid constructing unit 4b.

[0104] The non-uniform grid construction unit 4b acquires observation data D1 for each of the reference region and the surrounding region from the observation data acquisition unit 2. The non-uniform grid construction unit 4b also acquires division setting value data D2 for each of the reference region and the surrounding region from the division setting value data acquisition unit 3. Then, for each of the observation data D1, the non-uniform grid construction unit 4b subdivides a specific grid among the grids indicated by the observation data D1 (grid data), and merges the specific grid among the grids indicated by the observation data D1 with a grid adjacent to the specific grid, thereby generating non-uniform grid data indicating a non-uniform grid in which the subdivided grids and the merged grids are mixed. At this time, the non-uniform grid construction unit 4 selects a grid to be subdivided and a grid to be merged in the corresponding region based on the division setting value data D2 for each of the reference region and the surrounding region acquired from the division setting value data acquisition unit 3. The non-uniform grid construction unit 4 also generates non-uniform grid data based on the position data D4 acquired from the position data acquisition unit 7.

[0105] The region selection unit 41b selects grids to be subdivided and grids to be integrated from grids indicated by the observation data D1 in the corresponding regions based on the division setting value data D2 for each of the reference region and the surrounding regions. The region selection unit 41b outputs data indicating the selection result to the selected region grid conversion unit 42b.

[0106] The selected area grid conversion unit 42b acquires data indicating the selection result from the area selection unit 41b. Based on the acquired data, the selected area grid conversion unit 42b performs grid conversion (subdivision and integration) for each of the reference area and the surrounding areas, similar to the first embodiment. The selected area grid conversion unit 42b outputs data indicating the result of the grid conversion to the grid reconstruction unit 43b.

[0107] The selected area grid conversion unit 42b also generates adjacent grid determination data D3 indicating the adjacency relationship of each grid in the observation data D1 (grid data) after the grid conversion for each of the reference area and the surrounding areas. The selected area grid conversion unit 42 also causes the generated adjacent grid determination data D3 for each of the reference area and the surrounding areas to be recorded in the adjacent grid determination data recording unit 6.

[0108] The grid reconstruction unit 43b acquires data indicating the result of the grid conversion from the selected area grid conversion unit 42b. Based on the acquired data, the grid reconstruction unit 43b constructs a grid (non-uniform grid) in which a subdivided grid and a combined grid are mixed for each of the reference area and the surrounding areas, and generates data (non-uniform grid data) indicating the non-uniform grid. In addition, the grid reconstruction unit 43b combines the non-uniform grid data generated for each of the reference area and the surrounding areas based on the position data D4, and generates non-uniform grid data (hereinafter, also referred to as "extended non-uniform grid data") related to the area in which the reference area and the surrounding areas are joined. The grid reconstruction unit 43b outputs the generated extended non-uniform grid data to the non-uniform grid prediction processing unit 5b.

[0109] The non-uniform grid prediction processor 5b acquires the extended non-uniform grid data from the grid reconstruction unit 43b, and predicts weather data for a prediction target area included in the reference area using the acquired extended non-uniform grid data and adjacent grid determination data D3 for each of the reference area and surrounding areas recorded in the adjacent grid determination data recording unit 6. Then, the non-uniform grid prediction processor 5b outputs the predicted weather data for the prediction target area as a prediction result.

[0110] Next, an example of the operation of the weather forecasting device 1b according to the second embodiment will be described with reference to a flowchart shown in FIG.

[0111] First, the region selection unit 41b of the non-uniform grid construction unit 4b selects a subdivision target grid and a integration target grid from the grids indicated by the observation data D1 in the corresponding regions based on the division setting value data D2 for each of the reference region and the surrounding regions (step ST11). Note that the region selection unit 41b may perform this selection in parallel for each of the reference region and the surrounding regions, as shown in the dotted line frame in FIG. 8.

[0112] Next, the selected area grid conversion unit 42b performs grid conversion for each of the reference area and the surrounding areas on the grid selected by the area selection unit 41b (step ST12). Note that the selected area grid conversion unit 42b may perform this grid conversion for each of the reference area and the surrounding areas in parallel, as shown in the dotted line frame in FIG. 8.

[0113] Next, the grid reconstruction unit 43b constructs a non-uniform grid for each of the reference region and the surrounding regions based on the data indicating the result of the grid conversion obtained from the selected region grid conversion unit 42b, and generates non-uniform grid data. The grid reconstruction unit 43b also combines the non-uniform grid data generated for each of the reference region and the surrounding regions based on the position data D4, and generates extended non-uniform grid data for the region in which the reference region and the surrounding regions are joined (step ST13).

[0114] Next, the non-uniform grid prediction processing unit 5b acquires extended non-uniform grid data from the grid reconstruction unit 43b, and predicts weather data for the prediction target area included in the reference area using the acquired extended non-uniform grid data and adjacent grid determination data D3 for each of the reference area and surrounding areas recorded in the adjacent grid determination data recording unit 6 (step ST14).

[0115] In this way, the weather forecasting device 1b according to the second embodiment can combine and use observation data obtained by the observation devices 10 that observe different regions. Therefore, the weather forecasting device 1b can perform highly accurate weather forecasting in the prediction target region while efficiently taking into account observation data observed in a wider range of regions.

[0116] As described above, according to the second embodiment, the observation data acquisition unit 2 acquires grid data D1 obtained by dividing a predetermined region including a prediction target region into a plurality of grids, and grid data D1 obtained by dividing a peripheral region other than the predetermined region into a plurality of grids, and the non-uniform grid construction unit 4b generates extended non-uniform grid data including weather data for each grid, which is data showing a non-uniform grid for a region obtained by joining the predetermined region and the peripheral regions, based on the grid data D1 for each of the predetermined region and the peripheral regions acquired by the observation data acquisition unit 2, and the non-uniform grid prediction processing unit 5b performs prediction using the extended non-uniform grid data generated by the non-uniform grid construction unit 4b. As a result, the weather prediction device 1b according to the second embodiment can perform highly accurate weather prediction in the prediction target region while efficiently taking into account observation data observed in a wider range of regions, in addition to the effects of the first embodiment.

[0117] Moreover, the weather prediction device 1b includes a position data acquisition unit 7 that acquires data D4 indicating the positions of the predetermined region and the surrounding region, and the non-uniform grid construction unit 4b generates extended non-uniform grid data based on the position data D4 acquired by the position data acquisition unit 7. As a result, the weather prediction device 1b according to the second embodiment can accurately generate extended non-uniform grid data in addition to the effects of the first embodiment.

[0118] Embodiment 3 In the first embodiment, a weather forecasting device that forecasts weather data of a forecast target area included in a predetermined area using non-uniform grid data in which a fine grid and an integrated grid are mixed has been described. In the third embodiment, a weather forecasting device that can determine whether or not a forecast result indicated by the forecasted weather data has a predetermined accuracy will be described.

[0119] FIG. 9 is a diagram showing a configuration example of a weather forecasting system 100c including a weather forecasting device 1c according to the third embodiment. The weather forecasting device 1c according to the third embodiment is different from the weather forecasting device 1 according to the first embodiment shown in FIG. 1 in that a determination unit 8 is added, and the non-uniform grid construction unit 4 is changed to a non-uniform grid construction unit 4c, and the non-uniform grid prediction processing unit 5 is changed to a non-uniform grid prediction processing unit 5c. Each component of the non-uniform grid construction unit 4c is represented by adding "c" to the end of the reference numeral. The other configurations of the weather forecasting system 100c and the weather forecasting device 1c according to the third embodiment are the same as those of the weather forecasting system 100 and the weather forecasting device 1 according to the first embodiment shown in FIG. 1, so that the same reference numerals are used and the description thereof is omitted.

[0120] The non-uniform grid prediction processor 5c outputs the predicted weather data of the prediction target area (hereinafter, also referred to as "prediction data") to the determiner 8 before outputting it to the outside.

[0121] The determination unit 8 acquires prediction data from the non-uniform grid prediction processing unit 5c. Then, based on the acquired prediction data, the determination unit 8 determines whether or not a prediction result indicated by the prediction data can ensure a predetermined accuracy.

[0122] The determination unit 8 performs the determination in this case, for example, as follows. For example, the determination unit 8 acquires prediction data indicating a prediction result at any future time (for example, 200 seconds later) from the non-uniform grid prediction processing unit 5. The determination unit 8 also acquires meteorological data of the prediction target area (hereinafter also referred to as "actual measurement data") actually observed by the observation device 10 at the above time (for example, 200 seconds later). In this case, the determination unit 8 may, for example, acquire the above actual measurement data directly from the observation device 10, or may instruct the observation data acquisition unit 2 to acquire the above actual measurement data from the observation device 10 and acquire the actual measurement data acquired by the observation data acquisition unit 2.

[0123] Then, the determination unit 8 calculates the matching rate between the predicted data and the actual data, and if the calculated matching rate is equal to or greater than the threshold, determines that the predicted result indicated by the predicted data has a predetermined accuracy. On the other hand, if the calculated matching rate is less than the threshold, the determination unit 8 determines that the predicted result indicated by the predicted data does not have a predetermined accuracy. Note that this threshold may be set in advance by, for example, a user and recorded in the division setting value data recording unit 16.

[0124] When the determination unit 8 determines that the prediction result indicated by the prediction data has a predetermined accuracy, it notifies the non-uniform grid prediction processor 5c to that effect. On the other hand, when the determination unit 8 determines that the prediction result indicated by the prediction data does not have a predetermined accuracy, it notifies the non-uniform grid prediction processor 5c to that effect and instructs the non-uniform grid construction unit 4c to reconstruct the non-uniform grid and regenerate the non-uniform grid data.

[0125] Upon receiving this instruction, the region selection unit 41c of the non-uniform grid construction unit 4c refers to the grids selected as grids to be subdivided in the previous prediction, and determines whether or not at least one of these grids can be further subdivided.

[0126] This determination is made, for example, as follows: For example, the user sets in advance the maximum number of times that one grid can be subdivided, and records data indicating the maximum number of times in the division set value data recording unit 16. Then, for a grid that was selected as a grid to be subdivided in the previous prediction, if the number of times that grid has been subdivided has not reached the maximum number of times, the area selecting unit 41c determines that the grid can be further subdivided.

[0127] Alternatively, the user may set in advance the minimum area that one grid can have, and record data indicating the minimum area in the division set value data recording unit 16. Then, when the area of ​​a grid that was selected as a grid to be subdivided in the previous prediction is larger than the minimum area, the region selecting unit 41c determines that the grid can be further subdivided.

[0128] When the region selection unit 41c determines that at least one of the grids selected as the subdivision target grids can be further subdivided, it outputs data indicating that to the selected region grid conversion unit 42c. On the other hand, when all of the grids selected as the subdivision target grids cannot be further subdivided, the region selection unit 41c notifies the non-uniform grid prediction processing unit 5c of that fact.

[0129] When the selected area grid conversion unit 42c acquires data indicating that at least one of the grids selected as grids to be subdivided can be further subdivided from the area selection unit 41c, the selected area grid conversion unit 42c further subdivides the grid in the same manner as in embodiment 1. The selected area grid conversion unit 42c outputs data indicating the subdivision result to the grid reconstruction unit 43c. In addition, the selected area grid conversion unit 42c generates adjacent grid determination data D3 indicating the adjacency relationship of each grid in the observation data D1 (grid data) after the subdivision, and causes the generated adjacent grid determination data D3 to be recorded in the adjacent grid determination data recording unit 6.

[0130] The grid reconstruction unit 43c constructs a non-uniform grid including a mixture of subdivided grids and integrated grids in the same manner as in the first embodiment, and generates non-uniform grid data indicating the non-uniform grid. The non-uniform grid prediction processing unit 5c then performs prediction again using the generated non-uniform grid data, and outputs the resulting prediction data to the determination unit 8. The determination unit 8 determines whether or not the prediction result indicated by the prediction data has a predetermined accuracy, in the same manner as described above.

[0131] When the non-uniform grid prediction processor 5c is notified by the determination unit 8 that the prediction result indicated by the prediction data has a predetermined accuracy, the non-uniform grid prediction processor 5c outputs the prediction result to the outside. When the non-uniform grid prediction processor 5c is notified by the region selection unit 41c that all of the grids selected as grids to be subdivided are grids that cannot be further subdivided, the non-uniform grid prediction processor 5c may output to the outside a message to that effect together with a message to the outside that the prediction result indicated by the prediction data does not have a predetermined accuracy.

[0132] In the weather forecasting device 1c, the output of forecast data from the non-uniform grid forecast processing unit 5c to the determination unit 8 and the determination by the determination unit 8 may be repeated. In this case, the number of repetitions may be preset, for example, by a user and recorded in the division setting value data recording unit 16.

[0133] Next, an example of the operation of the weather prediction device 1c according to the third embodiment will be described with reference to the flowchart shown in Fig. 10. Note that steps ST21 to ST23 in the flowchart shown in Fig. 10 are similar to steps ST01 to ST03 in the flowchart shown in Fig. 2, and therefore will not be described again.

[0134] In step ST24, the non-uniform grid prediction processing unit 5c predicts meteorological data in the prediction target area using the non-uniform grid data, and outputs the prediction data obtained by the prediction to the determination unit 8.

[0135] In step ST25, the determination unit 8 acquires prediction data from the non-uniform grid prediction processing unit 5 and determines whether or not the prediction result indicated by the acquired prediction data has a predetermined accuracy (step ST25). As a result, when the determination unit 8 determines that the prediction result indicated by the prediction data has a predetermined accuracy (step ST25; YES), the non-uniform grid prediction processing unit 5 outputs the prediction result indicated by the prediction data and ends the process. On the other hand, when the determination unit 8 determines that the prediction result indicated by the prediction data does not have a predetermined accuracy (step ST25; NO), it instructs the non-uniform grid construction unit 4c to reconstruct the non-uniform grid and regenerate the non-uniform grid data. After that, the process returns to step ST21, and the area selection unit 41c selects an area again (selects a subdivision target grid that can be further subdivided). Thereafter, the weather prediction device 1c repeats the processes of steps ST21 to ST25 until it is determined in step ST25 that the prediction result indicated by the prediction data has a predetermined accuracy.

[0136] In this way, when a prediction result with a predetermined accuracy is not obtained as a result of the prediction process using a non-uniform grid, the weather prediction device 1c according to the third embodiment performs the prediction process by reconstructing the non-uniform grid and regenerating the non-uniform grid data. This allows the weather prediction device 1c to avoid outputting a prediction result that does not have the predetermined accuracy, and to output a prediction result with higher accuracy.

[0137] As described above, according to the third embodiment, the weather forecasting device 1c includes a determination unit 8 that determines whether or not the forecast result indicated by the weather data predicted by the non-uniform grid forecast processing unit 5c has a predetermined accuracy. When the determination unit 8 determines that the forecast result does not have the predetermined accuracy, the determination unit 8 instructs the non-uniform grid construction unit 4c to regenerate the non-uniform grid data. When the non-uniform grid constructing unit 4c receives the instruction, it regenerates the non-uniform grid data by further subdividing the grids that can be further subdivided among the subdivided grids included in the non-uniform grid. In this way, the weather forecasting device 1c according to the third embodiment can avoid outputting a forecast result that does not have a predetermined accuracy, in addition to the effect of the first embodiment, and can output a forecast result with higher accuracy.

[0138] Furthermore, the determination unit 8 calculates a coincidence rate between the weather data indicating the prediction result at any time in the future predicted by the non-uniform grid prediction processing unit 5c and the weather data actually observed in the prediction target area at that time, and determines that the prediction result does not have a predetermined accuracy if the calculated coincidence rate is less than a threshold. As a result, the weather prediction device 1c according to the third embodiment can accurately determine whether or not the prediction result has a predetermined accuracy, in addition to the effect of the first embodiment.

[0139] Embodiment 4 In the first embodiment, a weather forecasting device that forecasts weather data of a prediction target area included in a predetermined area using non-uniform grid data in which a fine grid and a unified grid are mixed has been described. In the fourth embodiment, a weather forecasting device that can forecast rainfall and wind in a prediction target area with higher accuracy will be described.

[0140] Fig. 11 is a diagram showing a configuration example of a weather prediction system 100d including a weather prediction device 1d according to embodiment 4. The weather prediction device 1d according to embodiment 4 differs from the weather prediction device 1 according to embodiment 1 shown in Fig. 1 in that the region selection unit 41 is changed to a region selection unit 41d, and the non-uniform grid construction unit 4 is changed to a non-uniform grid construction unit 4d. The other configurations of the weather prediction system 100d and the weather prediction device 1d according to embodiment 4 are the same as those of the weather prediction system 100 and the weather prediction device 1 according to embodiment 1 shown in Fig. 1, so the same reference numerals are used and the description thereof is omitted.

[0141] Similar to the first embodiment, the region selection unit 41d selects subdivision target grids and integration target grids from among the grids indicated by the observation data D1 based on the division set value data D2.

[0142] After the above selection, the area selection unit 41d refers to the grids indicated by the observation data D1 one by one and determines whether the wind speed and humidity indicated by the weather data in each grid are equal to or greater than the threshold. The area selection unit 41d then additionally selects the grids whose wind speed and humidity are equal to or greater than the threshold as subdivision target grids. If the grids whose wind speed and humidity are equal to or greater than the threshold have already been selected as subdivision target grids based on the division setting value data D2, the area selection unit 41d maintains the selection.

[0143] On the other hand, the area selection unit 41d does not additionally select, as a subdivision target grid, a grid for which it has determined that the wind speed and humidity are not equal to or greater than the threshold. If the area selection unit 41d has already selected, as a subdivision target grid, a grid for which it has determined that the wind speed and humidity are not equal to or greater than the threshold based on the division setting value data D2, the area selection unit 41d may maintain the selection, or may cancel the selection in consideration of the processing load. After that, the area selection unit 41d outputs data indicating the selection result to the selected area grid conversion unit 42. The thresholds for the wind speed and humidity may be set in advance by, for example, a user and recorded in the division setting value data recording unit 16.

[0144] Next, an example of the operation of the weather predicting device 1d according to the fourth embodiment will be described with reference to a flowchart shown in FIG.

[0145] First, the region selection unit 41d of the non-uniform grid construction unit 4d selects a grid to be subdivided and a grid to be integrated (step ST31).

[0146] Specifically, the area selection unit 41d first selects grids to be subdivided and grids to be integrated from the grids indicated by the observation data D1 based on the division setting value data D2, for example as shown within the dotted frame in FIG. 12 (step ST311).

[0147] Next, after the selection in step ST311, the area selection unit 41d refers to the grids indicated by the observation data D1 one by one (step ST312). For example, as shown in Fig. 13, when each grid is represented as (i, j), the area selection unit 41d refers to each grid one by one while counting up i and j from 1. Note that i is a variable indicating the position of each grid in the east-west direction (left-right direction), and j is a variable indicating the position of each grid in the north-south direction (up-down direction).

[0148] Then, the area selection unit 41d acquires weather data (weather conditions) in the grid (i, j) (step ST313).

[0149] Next, the area selection unit 41d judges whether the wind speed and humidity indicated by the weather data in the grid (i, j) are equal to or greater than the threshold (step ST314). As a result, when the area selection unit 41d judges that the wind speed and humidity are equal to or greater than the threshold (step ST314; YES), the area selection unit 41d additionally selects the grid (i, j) as a grid to be subdivided (step ST315). On the other hand, when the area selection unit 41d judges that the wind speed and humidity are not equal to or greater than the threshold (step ST314; NO), the area selection unit 41d does not additionally select the grid (i, j) as a grid to be subdivided. After that, the process returns to step ST312, and the area selection unit 41d repeats steps ST312 to ST315 for all grids.

[0150] When step ST31 is completed, the process proceeds to step ST32. Steps ST32 to ST34 are similar to steps ST2 to ST4 in the flowchart shown in Fig. 2, and therefore will not be described again. The grids additionally selected as grids to be subdivided are subdivided by the selected area grid conversion unit 42b in step ST32, as shown in Fig. 13, for example, but the previous selection results are maintained for the other grids.

[0151] In this way, the weather forecasting device 1d according to the fourth embodiment additionally selects, as subdivision target grids, grids for which the wind speed and humidity are determined to be equal to or greater than the threshold value, in addition to the selection based on the division setting value data D2. This is based on the assumption that, for example, when predicting rainfall and wind in a prediction target area, the wind speed and humidity in particular have a large influence on the prediction.

[0152] For example, when focusing on rainfall, humidity, or more precisely, the meteorological element of the amount of water vapor mixed in the sky, strongly influences the presence or absence of precipitation and the amount of precipitation. As for wind, the wind direction, etc. can also be determined by the wind speed (wind strength). Therefore, in the weather forecasting device 1d, as described above, for grids whose wind speed and humidity are equal to or greater than the threshold, they are additionally selected as grids to be subdivided, in addition to the selection based on the division setting value data D2. This allows the weather forecasting device 1d to make more accurate predictions regarding rainfall and wind in the prediction target area.

[0153] In the above description, the area selection unit 41d selects the subdivision target grids and the integration target grids based on the division setting value data D2, and then refers to each grid one by one to select additional subdivision target grids. However, the area selection unit 41d is not limited to this, and may, for example, omit the selection of the subdivision target grids based on the division setting value data D2 and only select the subdivision target grids based on the wind speed and humidity.

[0154] In the above description, the area selection unit 41d additionally selects the subdivision target grid based on wind speed and humidity. However, the area selection unit 41d is not limited to this, and may use, for example, air pressure in addition to wind speed and humidity as a determination factor for the above selection. This is based on the fact that air pressure is generally said to be an important meteorological factor in weather prediction. For example, rainfall and wind speed are easily affected by air pressure, and depending on the relationship with air pressure, there is a possibility that typhoons and torrential rains may occur. Therefore, the area selection unit 41d can perform predictions regarding rainfall and wind in the prediction target area with even higher accuracy by using air pressure in addition to wind speed and humidity as a determination factor for the above selection.

[0155] As described above, according to the fourth embodiment, the non-uniform grid constructing unit 4d subdivides grids in which the wind speed and humidity indicated by the meteorological data in the grid are equal to or greater than a threshold value, among the grids indicated by the grid data D1 acquired by the observation data acquiring unit 2. As a result, the weather forecasting device 1d according to the fourth embodiment can make predictions regarding rainfall and wind in the prediction target area with higher accuracy, in addition to the effects of the first embodiment.

[0156] Embodiment 5. In the first embodiment, a weather prediction system including one weather prediction device has been described. In the fifth embodiment, a weather prediction system capable of speeding up prediction processing by parallel processing using two weather prediction devices will be described.

[0157] Fig. 14 is a diagram showing a configuration example of a weather prediction system 100e including a weather prediction device 1e (first weather prediction device) and a weather prediction device 1f (second weather prediction device) according to embodiment 5. The weather prediction system 100e according to embodiment 5 differs from the weather prediction system 100 according to embodiment 1 shown in Fig. 1 in that the weather prediction device 1 is changed to a weather prediction device 1e and a weather prediction device 1f is added.

[0158] The weather forecasting device 1e according to the fifth embodiment is the weather forecasting device 1 according to the first embodiment from which the selected area grid conversion unit 42 has been deleted, and the weather forecasting device 1f according to the fifth embodiment is equipped with the selected area grid conversion unit 42. The other configurations of the weather forecasting system 100e and the weather forecasting device 1e according to the fifth embodiment are the same as those of the weather forecasting system 100 and the weather forecasting device 1 according to the first embodiment shown in Fig. 1, so the same reference numerals are used and the description thereof will be omitted.

[0159] In the weather prediction system 100e, the weather prediction device 1e and the weather prediction device 1f are communicatively connected via a network (not shown) as shown in Fig. 14. In the weather prediction system 100e, the weather prediction device 1e is positioned as a main processing device that is responsible for main processing in the weather prediction processing, and the weather prediction device 1f is positioned as an external processing device that is located outside the main processing device and mainly performs grid conversion processing.

[0160] The weather prediction device 1e and the weather prediction device 1f may have the same level of computing performance, but in addition to this, for example, the weather prediction device 1f, which is an external processing device, may have higher computing performance than the weather prediction device 1e, which is a main processing device. For example, the weather prediction device 1e may have a CPU (Central Processing Unit), and the weather prediction device 1f may have a GPU (Graphics Processing Unit) or an FPGA (Field Programmable Gate Array).

[0161] In the weather prediction system 100e, it is assumed that the weather prediction device 1e transfers grid data to be converted to the weather prediction device 1f and manages the processing results by the weather prediction device 1f, and that the weather prediction device 1f mainly performs grid conversion on the grid data transferred from the weather prediction device 1e.

[0162] Next, an operation example of the weather prediction system 100e according to the fifth embodiment will be described. First, the area selection unit 41 of the weather prediction device 1e selects a subdivision target grid and an integration target grid from the grid indicated by the observation data D1 based on the division setting value data D2, as in the first embodiment. The area selection unit 41 transmits data indicating the selection result together with the observation data D1 to the selected area grid conversion unit 42 of the weather prediction device 1f via the network.

[0163] When the selected area grid conversion unit 42 of the weather prediction device 1f receives the data indicating the selection result and the observation data D1 transmitted from the area selection unit 41, the selected area grid conversion unit 42 performs grid conversion (subdivision and integration) based on the data indicating the selection result. Then, the selected area grid conversion unit 42 transmits data indicating the result of the grid conversion to the grid reconstruction unit 43 of the weather prediction device 1e via the network.

[0164] In addition, the selected area grid conversion unit 42 generates adjacent grid determination data D3 indicating the adjacency relationship of each grid in the observation data D1 (grid data) after the above-mentioned grid conversion, and records the generated adjacent grid determination data D3 in the adjacent grid determination data recording unit 6 of the weather prediction device 1e via a network.

[0165] When the grid reconstruction unit 43 of the weather prediction device 1e receives data indicating the result of the grid conversion from the selected area grid conversion unit 42, the grid reconstruction unit 43 constructs a non-uniform grid based on the received data and generates non-uniform grid data indicating the non-uniform grid. The grid reconstruction unit 43 outputs the generated non-uniform grid data to the non-uniform grid prediction processing unit 5.

[0166] When the non-uniform grid prediction processor 5 of the weather prediction device 1e acquires the non-uniform grid data from the grid reconstruction unit 43, it predicts weather data for a prediction target area included in a predetermined area using the acquired non-uniform grid data and adjacent grid determination data D3 recorded in the adjacent grid determination data recording unit 6. Then, the non-uniform grid prediction processor 5 outputs the predicted weather data for the prediction target area as a prediction result.

[0167] Here, the grid conversion process by the selected area grid conversion unit 42 of the weather forecasting device 1f is independent for each grid and can be processed in parallel. Therefore, the selected area grid conversion unit 42 performs the grid conversion process for each grid in parallel, and transmits data indicating the result to the grid reconstruction unit 43 of the weather forecasting device 1e as soon as the grid conversion process for one grid is completed. Also, each time the grid reconstruction unit 43 of the weather forecasting device 1e receives data indicating the result of the grid conversion from the selected area grid conversion unit 42, it returns the grid indicated by the data to its original position in the observation data D1 and combines it with other grids to construct a non-uniform grid. At this time, the grid conversion process by the selected area grid conversion unit 42 and the construction process of the non-uniform grid by the grid reconstruction unit 43 can be performed in parallel.

[0168] In this way, in the weather prediction system 100e according to the fifth embodiment, the processes required for constructing a non-uniform grid are shared between the weather prediction device 1e and the weather prediction device 1f. As a result, the weather prediction system 100e can perform, for example, grid conversion processing and non-uniform grid construction processing in parallel, and can speed up the prediction processing compared to the case where a single weather prediction device is used.

[0169] In this case, in the weather forecasting system 100e, the computational performance of the weather forecasting device 1f can be made superior to that of the weather forecasting device 1e, and the grid conversion processing, which has a larger processing load, can be performed by the weather forecasting device 1f, thereby further speeding up the grid conversion processing, and as a result, the prediction processing can be further speeded up.

[0170] As described above, according to the fifth embodiment, the weather prediction system 100e is a weather prediction system 100e including a weather prediction device 1e and a weather prediction device 1f connected to each other so as to be able to communicate with each other. The weather prediction device 1e includes an observation data acquisition unit 2 that acquires grid data D1 obtained by dividing a predetermined area including a prediction target area into a plurality of grids, the grid data D1 including meteorological data observed in each of the divided grids, an area selection unit 41 that selects a grid to be subjected to grid conversion from the grids indicated by the grid data and transmits data indicating the selection result to the weather prediction device 1f, and a region selection unit 42 that selects a grid to be subjected to grid conversion from the grids indicated by the grid data and transmits data indicating the selection result to the weather prediction device 1f. The weather forecasting device 1f includes a grid reconstructing unit 43 that generates non-uniform grid data indicating a grid in which a subdivided grid and a unified grid are mixed based on data indicating the result, and a non-uniform grid prediction processing unit 5 that predicts weather data of a prediction target area included in a predetermined area using the non-uniform grid data generated by the grid reconstructing unit 43. The weather forecasting device 1f includes a selected area grid converting unit 42 that performs grid conversion on the grid selected by the area selecting unit 41 and transmits data indicating the result of the grid conversion to the weather forecasting device 1e, and the process by the selected area grid converting unit 42 and the process by the grid reconstructing unit 43 are performed in parallel. As a result, the weather forecasting system 100e according to the fifth embodiment can improve the prediction accuracy of weather conditions while suppressing an increase in processing load compared to the conventional system, and can speed up the prediction process compared to the case of using one weather forecasting device.

[0171] In addition, the present disclosure allows free combination of the respective embodiments, modification of any of the components of each embodiment, or omission of any of the components of each embodiment. [Industrial Applicability]

[0172] INDUSTRIAL APPLICABILITY The present disclosure makes it possible to improve the accuracy of forecasting weather conditions while suppressing an increase in processing load compared to the prior art, and is suitable for use in weather forecasting devices, weather forecasting systems, and weather forecasting methods. [Explanation of symbols]

[0173] 1, 1b, 1c, 1d, 1e, 1f Weather forecasting device, 2 Observation data acquisition unit (first acquisition unit), 3 Division setting value data acquisition unit (second acquisition unit), 4, 4b, 4c, 4d Non-uniform grid construction unit (generation unit), 5, 5b, 5c Non-uniform grid prediction processing unit (prediction processing unit), 6 Adjacent grid determination data recording unit, 7 Position data acquisition unit (third acquisition unit), 8 Determination unit, 10 Observation device, 15 Observation data recording unit, 16 Division setting value data recording unit, 17 Position data recording unit, 41, 41b, 41c, 41d Area selection unit, 42, 42b, 42c Selected area grid conversion unit, 43, 43b, 43c Grid reconstruction unit, 51 Processing circuit, 52 CPU, 53 Memory, 100, 100b, 100c, 100d, 100e Weather forecasting system, 506 Graph, A1 local area, A2 medium area, A3 large area.

Claims

1. a first acquisition unit that acquires grid data obtained by dividing a predetermined area including a prediction target area into a plurality of grids, the grid data including meteorological data observed for each divided grid; a generating unit that subdivides a specific grid among the grids indicated by the grid data acquired by the first acquiring unit, and merges the specific grid among the grids indicated by the grid data with a grid adjacent to the specific grid, thereby generating non-uniform grid data including meteorological data for each grid, the non-uniform grid data being data indicating a non-uniform grid in which subdivided grids and merged grids are mixed; a prediction processing unit that predicts weather data of a prediction target area included in the predetermined area by using the non-uniform grid data generated by the generation unit; A weather forecasting device comprising:

2. The generation unit is generating adjacent grid determination data indicating an adjacent relationship between each grid in the non-uniform grid; The prediction processing unit: Based on the adjacent grid determination data generated by the generation unit, a pair of grids having a non-uniform adjacent relationship is identified from the non-uniform grid indicated by the non-uniform grid data, and the prediction is performed after performing an interpolation process on the identified pair of grids.

2. The weather forecasting device according to claim 1.

3. The generation unit is Among the grids indicated by the grid data, a grid including the forecast target area is subdivided, and a grid including an area in which the weather data shows little change over time is integrated with a grid adjacent to the subdivided grid.

3. The weather forecasting device according to claim 1 or 2.

4. The generation unit is Among the grids indicated by the grid data, a grid including an area where a meteorological element that affects the weather forecast in the forecast target area may occur is subdivided.

4. The weather forecasting device according to claim 3.

5. a second acquisition unit that acquires conversion target data, which is data indicating a subdivision target grid to be subdivided and data indicating an integration target grid to be integrated with an adjacent grid, from among the grids indicated by the grid data; The generation unit is The segmentation and integration are performed based on the conversion target data acquired by the second acquisition unit.

3. The weather forecasting device according to claim 1 or 2.

6. The generation unit is a region selection unit that selects the subdivision target grid and the integration target grid from the grid indicated by the grid data based on the conversion target data acquired by the second acquisition unit; a selected area grid conversion unit that generates the non-uniform grid data by subdividing a subdivision target grid selected by the area selection unit and integrating the integration target grid selected by the area selection unit with a grid adjacent to the integration target grid; 6. The weather forecasting device according to claim 5, further comprising:

7. The selected area grid conversion unit supplements the meteorological data in the grid that has been subdivided and integrated, The prediction processing unit performs the prediction using the non-uniform grid data after the weather data is supplemented by the selected area grid conversion unit.

7. The weather forecasting device according to claim 6.

8. The selected area grid conversion unit The meteorological data in the grids in which the subdivision and the integration have been performed is supplemented by interpolating the meteorological data in the grids adjacent to each of the grids.

8. The weather forecasting device according to claim 7.

9. The selected area grid conversion unit Instructing the first acquisition unit to acquire weather data for the grids in which the segmentation and integration have been performed, thereby replenishing the weather data for each of the grids.

8. The weather forecasting device according to claim 7.

10. The first acquisition unit is Obtaining grid data obtained by dividing a predetermined area including the prediction target area into a plurality of grids and grid data obtained by dividing a peripheral area other than the predetermined area into a plurality of grids; The generation unit is generating extended non-uniform grid data including meteorological data for each grid, the extended non-uniform grid data being data representing a non-uniform grid for an area obtained by connecting the specified area and the surrounding areas, based on the grid data for each of the specified area and the surrounding areas acquired by the first acquisition unit; The prediction processing unit: The prediction is performed using the extended non-uniform grid data generated by the generation unit.

2. The weather forecasting device according to claim 1.

11. a third acquisition unit that acquires data indicating positions of the predetermined region and the surrounding region; The generation unit is generating the extended non-uniform grid data based on the position data acquired by the third acquisition unit; 11. The weather forecasting device according to claim 10.

12. a determination unit that determines whether or not a prediction result indicated by the weather data predicted by the prediction processing unit has a predetermined accuracy; When the determination unit determines that the prediction result does not have a predetermined accuracy, the determination unit instructs the generation unit to regenerate non-uniform grid data; When the generation unit receives the instruction, the generation unit regenerates non-uniform grid data by subdividing grids that can be further subdivided among the subdivided grids included in the non-uniform grid.

2. The weather forecasting device according to claim 1.

13. The determination unit is A coincidence rate between weather data indicating a prediction result at an arbitrary time in the future predicted by the prediction processing unit and weather data actually observed in the prediction target area at the arbitrary time is calculated, and if the calculated coincidence rate is less than a threshold, it is determined that the prediction result does not have a predetermined accuracy.

13. The weather forecasting device according to claim 12.

14. The generation unit is Among the grids indicated by the grid data acquired by the first acquisition unit, a grid in which the wind speed and humidity indicated by the meteorological data in the grid are equal to or greater than a threshold value is subdivided.

2. The weather forecasting device according to claim 1.

15. A weather forecasting system including a first weather forecasting device and a second weather forecasting device communicatively connected to each other, The first weather forecasting device is a first acquisition unit that acquires grid data obtained by dividing a predetermined area including a prediction target area into a plurality of grids, the grid data including meteorological data observed in each divided grid; an area selection unit that selects a grid to be subjected to grid conversion from the grids indicated by the grid data, and transmits data indicating the selection result to the second weather forecasting device; a grid reconstruction unit that generates non-uniform grid data representing a grid including a mixture of a subdivided grid and a unified grid based on data representing a result of the grid conversion transmitted from the second weather forecasting device; a prediction processing unit that predicts weather data of a prediction target area included in the predetermined area by using the non-uniform grid data generated by the grid reconstruction unit; Equipped with The second weather forecasting device is a selected area grid conversion unit that performs grid conversion on the grid selected by the area selection unit and transmits data indicating a result of the grid conversion to the first weather forecasting device; A weather forecasting system, characterized in that the processing by the selected area grid conversion unit and the processing by the grid reconstruction unit are performed in parallel.

16. A weather forecasting method using a weather forecasting device, comprising: A step in which a first acquisition unit acquires grid data obtained by dividing a predetermined area including a prediction target area into a plurality of grids, the grid data including meteorological data observed for each divided grid; a generation unit generating non-uniform grid data including meteorological data for each grid, the non-uniform grid data being data indicating a non-uniform grid including a mixture of subdivided grids and integrated grids by subdividing a predetermined grid among the grids indicated by the grid data acquired by the first acquisition unit and integrating the predetermined grid among the grids indicated by the grid data with a grid adjacent to the predetermined grid; A prediction processing unit predicts weather data of a prediction target area included in the predetermined area by using the non-uniform grid data generated by the generation unit; A weather forecasting method comprising the steps of:

Images