Information processing device, information processing method, and program

Abstract

The present invention accurately performs long-term weather prediction including, for one month or longer into the future, extreme high temperatures and low temperatures, heavy rain, drought, the end of rainy season, the number of typhoons that will approach, etc.　An index acquisition unit 51 acquires, from an exponent computation device 2 resulting from a numerical forecasting model for calculating an index focusing on a weather phenomenon correlated with a prediction region and a distant region, the index of the distant region as prediction data (index), and provides the index to a weather prediction unit 52. The weather prediction unit 52 performs long-term weather prediction for the prediction region on the basis of the index (prediction data) for the distant region. The weather prediction unit 52 performs long-term weather prediction of the prediction region by using a prediction-type regression coefficient 61 obtained as a result of prescribed machine learning. The prediction-type regression coefficient 61 is generated or updated by a learning unit 54.

Description

Information processing device, information processing method, and program

[0001] The present invention relates to an information processing device, an information processing method, and a program.

[0002] There is a high demand in various fields for long-term weather forecasts of one month or more in advance (see, for example, Patent Document 1), and this is also disclosed in Patent Document 1, for example.

[0003] Japanese Patent Application Laid-Open No. 2022-191176

[0004] However, there are physical limitations to long-term weather forecasts, and mid- to high-latitude weather forecasts using numerical weather forecasting models have low accuracy, making it impossible to predict extreme high and low temperatures, heavy rainfall, and droughts more than one month in advance. Therefore, guidance that supports forecasting by correcting the grid point values ​​directly above the target area in a numerical weather forecasting model, such as that used by the Japan Meteorological Agency, has low accuracy for forecasts more than one month in advance. Furthermore, guidance that uses only the grid point values ​​directly above such numerical weather forecasting models is unable to predict the end of the rainy season, which has a significant impact on the economy, or the number of approaching typhoons.

[0005] The present invention has been made in consideration of these circumstances, and aims to accurately forecast long-term weather, including extremely high or low temperatures, heavy rain, and drought for more than one month in advance, as well as the end of the rainy season and the number of approaching typhoons.

[0006] In order to achieve the above object, an information processing device of one embodiment of the present invention comprises: an index acquisition means for acquiring an index of a remote area from a calculation means for calculating the index focusing on weather phenomena correlated between the prediction area and the remote area; and a weather forecasting means for making a long-term weather forecast in the prediction area based on the index of the remote area.

[0007] According to the present invention, it is possible to accurately predict long-term weather including extremely high or low temperatures, heavy rain, drought, the end of the rainy season, the number of approaching typhoons, and the like, for more than one month in advance.

[0008] 1 is a diagram showing an example of the configuration of an information processing system including a weather prediction device as an embodiment of the information processing device of the present invention. FIG. 2 is a block diagram showing an example of the hardware configuration of the weather prediction device in the information processing system of FIG. 1. FIG. 3 is a functional block diagram showing an example of the functional configuration of the weather prediction device of FIG. 2. FIG. 4 is a diagram showing a first specific example of a long-term weather forecast performed by a weather prediction device having the functional configuration of FIG. 3. FIG. 5 is a diagram showing a second specific example of a long-term weather forecast performed by a weather prediction device having the functional configuration of FIG. 3. FIG. 6 is a diagram showing a third specific example of a long-term weather forecast performed by a weather prediction device having the functional configuration of FIG. 3. FIG. 7 is a diagram showing a fourth specific example of a long-term weather forecast performed by a weather prediction device having the functional configuration of FIG. 3.

[0009] DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will now be described with reference to the accompanying drawings. Fig. 1 is a diagram showing an example of the configuration of an information processing system including a weather forecasting apparatus as an embodiment of an information processing apparatus of the present invention.

[0010] The information processing system shown in Figure 1 is composed of a weather forecasting device 1, an index calculation device 2 using a numerical forecasting model, and a weather observation data providing device 3 for the forecast target, all of which are connected to each other via a predetermined network such as the Internet.

[0011] The numerical forecast model-based index calculation device 2 is an information processing device that calculates an index focusing on meteorological phenomena that are correlated between the prediction region and the remote region. The weather forecasting device 1 is one embodiment of an information processing device to which the present invention is applied, and is an information processing device that performs long-term weather forecasts in the prediction region based on the index of the remote region calculated by the numerical forecast model-based index calculation device 2.

[0012] Here, a long-term weather forecast refers to a weather forecast for a predetermined period of one month or more in the future. In one embodiment, it may be two months or more in the future, and in another embodiment, it may be three months or more in the future. A prediction region is a region for which a long-term weather forecast is performed, and may include, for example, Japan, the United States, Europe, etc. A remote region is a region that is far enough away from the prediction region that the weather in the same time period is not related to the prediction region. For example, if the prediction region is Japan, the subtropical or tropical Indian Ocean or the vicinity of the Philippines would be a remote region. Also, for example, if the prediction region is the United States, the El Niño monitoring area would be a remote region.

[0013] Specifically, for example, the index calculation device 2 using a numerical forecast model calculates the index using the strength of convective activity in the Indian Ocean or near the Philippines. The weather forecasting device 1 predicts the sunshine hours in Japan for a predetermined month (e.g., July relative to March) that is one month or more ahead of a predetermined time point based on the index. For example, the index calculation device 2 using a numerical forecast model calculates the index using the change in sea surface temperature in an El Niño monitoring area based on the index. The weather forecasting device 1 predicts the precipitation amount in the United States for a predetermined month (e.g., July relative to March) that is one month or more ahead of a predetermined time point based on the index. A specific example of long-term weather forecasting for a forecasting area using such indices for a remote area will be described later with reference to Figures 4 to 7.

[0014] Here, the weather forecasting device 1 of this embodiment can perform long-term weather forecasting for a prediction area using a model obtained as a result of predetermined machine learning (for example, the regression coefficients 61 of the prediction formula in FIG. 3 , which will be described later). When a model is generated or updated by such predetermined machine learning, the weather observation data providing device 3 for the prediction target provides actual weather observation data of the prediction area as part of the learning data. Details of the predetermined machine learning will be described later in the explanation of the regression coefficients 61 of the prediction formula in FIG. 3 . Note that the regression coefficients of the prediction formula referred to here can be calculated by any method.

[0015] FIG. 2 is a block diagram showing an example of the hardware configuration of the weather forecasting device in the information processing system shown in FIG.

[0016] The CPU 11 executes various processes according to programs recorded in the ROM 12 or programs loaded from the storage unit 18 into the RAM 13. The RAM 13 also stores data and the like necessary for the CPU 11 to execute various processes.

[0017] The CPU 11, ROM 12, and RAM 13 are interconnected via a bus 14. An input / output interface 15 is also connected to this bus 14. An input unit 16, an output unit 17, a storage unit 18, a communication unit 19, and a drive 20 are connected to the input / output interface 15.

[0018] The input unit 16 is configured with, for example, a keyboard, and is used to input various types of information. The output unit 17 is configured with, for example, a display such as an LCD, a speaker, and the like, and outputs various types of information as images and sounds. The storage unit 18 is configured with, for example, a DRAM (Dynamic Random Access Memory), and stores various types of data. The communication unit 19 communicates with other devices (for example, the index calculation device 2 based on the numerical forecast model and the weather observation data providing device 3 for the forecast target in FIG. 1 ) via a network N including the Internet.

[0019] Removable media 30, such as a magnetic disk, optical disk, magneto-optical disk, or semiconductor memory, is appropriately attached to the drive 20. Programs read from the removable media 30 by the drive 20 are installed in the storage unit 18 as needed. The removable media 30 can also store various data stored in the storage unit 18 in the same way as the storage unit 18.

[0020] Although not shown, the index calculation device 2 based on the numerical forecast model and the forecast target meteorological observation data providing device 3 in Fig. 1 can also have a configuration basically similar to the hardware configuration shown in Fig. 2. Therefore, a description of the hardware configuration of the index calculation device 2 based on the numerical forecast model and the forecast target meteorological observation data providing device 3 will be omitted.

[0021] By cooperation of the various hardware and software constituting the information processing system of Figure 1, including the weather forecasting device 1 of Figure 2, it is possible to perform long-term weather forecasts in the prediction area based on indices that focus on meteorological phenomena that are correlated between the prediction area and remote areas.

[0022] FIG. 3 is a functional block diagram showing an example of the functional configuration of the weather forecasting device 1 of FIG. 2 in the information processing system of FIG.

[0023] 3, the CPU 11 of the weather forecasting device 1 functions as an index acquisition unit 51, a weather forecasting unit 52, a meteorological observation data acquisition unit 53, and a learning unit 54. Regression coefficients 61 of the forecast formula are stored in one area of ​​the memory unit 18.

[0024] The index acquisition unit 51 acquires the index of the remote region as forecast data (index) from the index calculation device 2 using a numerical forecast model that calculates indexes focusing on weather phenomena that are correlated between the forecast region and the remote region, and provides the index to the weather forecast unit 52. The weather forecast unit 52 performs long-term weather forecasts in the forecast region based on the index of the remote region (forecast data).

[0025] The weather forecasting unit 52 of this embodiment performs long-term weather forecasting for a forecast region using regression coefficients 61 of a forecasting formula obtained as a result of predetermined machine learning. The forecasting formula is expressed as Y=F(X, a), where Y is the meteorological observation data to be forecasted, X is an index of the numerical prediction model, and a is the regression coefficient. The regression coefficients 61 of this forecasting formula are generated or updated by the learning unit 54. Therefore, the functional configuration in the learning phase will be described below.

[0026] The index acquisition unit 51 acquires simulation data of indices of a remote region up to a predetermined period into the future based on a past numerical forecast model from the index calculation device 2 based on a numerical forecast model as learning data (indices), and provides this to the learning unit 54. Specifically, for example, the index acquisition unit 51 acquires simulation data of sea surface temperatures and convective activity in a remote region up to seven months into the future based on a past numerical forecast model from the index calculation device 2 based on a numerical forecast model as learning data (indices), and provides this to the learning unit 54.

[0027] The meteorological observation data acquisition unit 53 acquires actual meteorological observation data for the prediction area as learning data from the meteorological observation data providing device 3 that is the prediction target, and provides it to the learning unit 54 .

[0028] The learning unit 54 analyzes the relationship between the learning data (indicators) acquired by the index acquisition unit 51 and the learning data (actual weather observation data) acquired by the weather observation data acquisition unit 53 using a predetermined machine learning method, and generates or updates a regression coefficient 61 of the prediction formula as a relational equation that explains the results of long-term weather observations in the prediction area using indices in the remote area.

[0029] As a result, in the prediction and operation phase, the index acquisition unit 51 and the weather forecasting unit 52 function as described above. Specifically, for example, when the index acquisition unit 51 acquires forecasts of sea surface temperatures and convective activity in a remote region for up to seven months in advance based on a new numerical forecasting model as forecast data (indexes), it provides this to the weather forecasting unit 52. The weather forecasting unit 52 incorporates the indexes (forecast data) of the remote region into the regression coefficients 61 (relational equation) of the prediction formula and performs long-term weather forecasts in the prediction region.

[0030] Next, several examples will be described as specific examples of long-term weather forecasts made by a weather forecasting device having the functional configuration of Figure 3. Figures 4 to 7 are diagrams showing first to fourth specific examples of long-term weather forecasts made by a weather forecasting device having the functional configuration of Figure 3, respectively.

[0031] In FIG. 4 , the long-term weather forecast result 101 of the first example to which this embodiment is applied (hereinafter referred to as the "first example result 101 of this embodiment") shows the precipitation amount (% of the average) for July predicted as of January for each year (1993 to 2016) for the Pacific coast of western Japan as the forecast region. In the first example result 101 of this embodiment, the remote location indicators used are predictions of convective activity in the subtropical and tropical Indian Ocean and near the Philippines using a numerical forecast model. In the first example result 101 of this embodiment, the thick solid line indicates the observation (the actual meteorological observation data value). The thin solid line indicates the prediction (the long-term weather forecast result). The dashed line indicates the 70% prediction interval.

[0032] Figure 4 shows the results 102 of long-term weather forecasting using a conventional method (prediction using the precipitation amount of the grid point value directly above the numerical forecast model as an index) (hereinafter referred to as "conventional method results 102") for comparison with the first case result 101 of this embodiment.

[0033] It can be seen that while the conventional method result 102 was unable to predict extremely heavy or light rain six months in advance, the first case result 101 of this embodiment shows that there are more cases where extremely heavy or light rain can be predicted. It can also be seen that the first case result 101 of this embodiment makes it possible to grasp the expected range of error using the prediction interval.

[0034] In FIG. 5 , the long-term weather forecast result 103 of the second example to which this embodiment is applied (hereinafter referred to as the "second example result 103 of this embodiment") shows the predicted end date of the rainy season (difference from the average) as of January of each year (1993 to 2022) for the Kanto-Koshinetsu region of Japan. In the second example result 103 of this embodiment, the remote location indicators used are predictions of convective activity in the subtropical and tropical Indian Ocean and near the Philippines using a numerical forecast model. In the third example result 103 of this embodiment, the thick solid line indicates the observation (value of actual meteorological observation data). The thin solid line indicates the prediction (result of the long-term weather forecast). The dashed line indicates the 70% prediction interval.

[0035] Although not shown in Figure 5, the guidance of the conventional method (a method that uses grid point value indexes directly above the numerical prediction model) was unable to express the "timing" of the end of the rainy season, whereas, as shown in the second case result 103 of this embodiment, by applying the weather forecasting device 1 of this embodiment, it is possible to predict the timing of the end of the rainy season, etc.

[0036] In FIG. 6 , the long-term weather forecast result 104 for the third case in which this embodiment was applied (hereinafter referred to as the "third case result 104 of this embodiment") shows the number of typhoons predicted for September as of June for each year (1993 to 2023) for the prediction region covering Japan. The long-term weather forecast result 105 for the third case in which this embodiment was applied (hereinafter referred to as the "third case result 105 of this embodiment") shows the number of typhoons approaching the mainland in September as of June for each year (1993 to 2023) for the prediction region covering Japan. Here, the mainland refers to Hokkaido, Honshu, Shikoku, and Kyushu. In the third case results 104 and 105 of this embodiment, the remote location indicators used are predictions of convective activity in the subtropical and tropical Indian Ocean and near the Philippines using a numerical weather forecast model. In the third case results 104 and 105 of this embodiment, the thick solid lines indicate observations (actual meteorological observation data values). The thin solid line shows the forecast (result of long-term weather forecasting), and the dashed line shows the 70% prediction interval.

[0037] As shown in the third case results 104 and 105 of this embodiment, by applying the weather forecasting device 1 of this embodiment, it is possible to predict trends such as the number of typhoons that occur and the number of typhoons that approach, such as whether they will be high or low.

[0038] In FIG. 7 , actual results 106 (hereinafter referred to as "fourth case actual results 106") shown for comparison with the fourth case show the sunshine hours in July of each year (2019 to 2022) with Japan as the prediction region. Results 107 of the long-term weather forecast of the fourth case to which this embodiment is applied (hereinafter referred to as "fourth case result 107 of this embodiment") show the sunshine hours in July predicted as of April of each year (2019 to 2022) with Japan as the prediction region. In the fourth case result 107 of this embodiment, predictions of convective activity in the subtropical and tropical Indian Ocean and near the Philippines using a numerical weather forecast model are used as remote location indicators.

[0039] Comparing the fourth case result 106 with the fourth case result 107 of this embodiment, it can be seen that by applying the weather forecasting device 1 of this embodiment, it has become possible to roughly grasp regional trends of increased and decreased sunshine.

[0040] Although one embodiment of the present invention has been described above, the present invention is not limited to the above-described embodiment, and modifications, improvements, etc. within the scope of achieving the object of the present invention are considered to be included in the present invention.

[0041] 4 to 7, Japan is used as the prediction region, and convective activity in the subtropical or tropical Indian Ocean or near the Philippines is used as the indicator of the remote region, but this is not particularly limited. Specifically, as described above, the United States may be used as the prediction region, and sea surface temperatures in the El Niño monitoring area may be used as the indicator of the remote region.

[0042] For example, the hardware configuration of the weather forecasting device 1 shown in FIG. 2 is merely an example for achieving the object of the present invention, and is not particularly limited.

[0043] 3 is merely an example and is not particularly limited. That is, it is sufficient if the function capable of executing the above-described processing as a whole is provided, and the type of functional block used to realize this function is not particularly limited to the example in FIG.

[0044] Furthermore, the locations of the functional blocks and databases are not limited to those shown in Figure 3 and may be arbitrary. Specifically, for example, the learning unit 54, the index acquisition unit 51 for acquiring learning data, and the meteorological observation data acquisition unit 53 may be transferred to an information processing device dedicated to learning (not shown). Also, for example, both the function of the index calculation device 2 using a numerical forecast model and the weather forecasting unit 52 may be included in a single information processing device (not shown). Also, for example, the regression coefficients 61 of the prediction formula may be stored in a storage device of another information processing device (not shown).

[0045] The above-described series of processes can be executed by hardware or software, and each functional block can be configured by hardware alone, software alone, or a combination of both.

[0046] When a series of processes is executed by software, the programs constituting the software are installed onto a computer or the like from a network or a recording medium. The computer may be a computer incorporated into dedicated hardware. The computer may also be a computer capable of executing various functions by installing various programs, such as a server, a general-purpose smartphone, or a personal computer.

[0047] The recording medium containing such a program may be composed not only of a removable medium (not shown) that is distributed separately from the device main body in order to provide the program to the user, but also of a recording medium that is provided to the user in a state that is pre-installed in the device main body.

[0048] In this specification, the steps describing the program to be recorded on the recording medium include not only processes that are performed in chronological order, but also processes that are not necessarily performed in chronological order but are performed in parallel or individually.

[0049] To summarize the above, an information processing device to which the present invention is applied is sufficient as long as it has the following configuration, and can take various forms. That is, an information processing device to which the present invention is applied (e.g., the weather forecasting device 1 in FIG. 1 ) is sufficient as long as it includes: index acquisition means (e.g., the index acquisition unit 51 in FIG. 3 ) that acquires an index for a remote region (e.g., convective activity in the subtropical or tropical Indian Ocean or near the Philippines) from calculation means (e.g., the index calculation device 2 using a numerical forecasting model in FIG. 1 ) that calculates the index focusing on meteorological phenomena that correlate between the prediction region (e.g., Japan) and a remote region (e.g., the subtropical or tropical Indian Ocean or near the Philippines), and weather forecasting means (e.g., the weather forecasting unit 52 in FIG. 3 ) that performs a long-term weather forecast for the prediction region (e.g., the long-term weather forecasts shown in FIGS. 4 to 7 ) based on the index for the remote region.

[0050] This will enable accurate long-term weather forecasts, including extremely high or low temperatures, heavy rain, and droughts more than a month in advance, as well as the end of the rainy season and the number of approaching typhoons.

[0051] Here, for example, the indicator may be at least one of convective activity in the ocean of the remote region and sea surface temperature.

[0052] The information processing device may further include a learning means (e.g., the learning unit 54 in Figure 3) that generates a model (e.g., the regression coefficient 61 of the prediction formula in Figure 3) that explains the results of long-term weather observations of the prediction area using the index of the remote area by analyzing the relationship between simulation data of the index of the remote area up to a predetermined period afterwards based on a past numerical forecast model and actual weather observation data of the prediction area using a predetermined machine learning method; the index acquisition means acquires a prediction of the index of the remote area up to the predetermined period afterwards based on a new numerical forecast model; and the weather forecasting means inputs the acquired prediction of the index of the remote area into the model and uses the output of the model as the long-term weather forecast in the prediction area.

[0053] 1...Weather forecasting device, 2...Index calculation device using numerical forecasting model, 3...Device for providing weather observation data of forecast target, 11...CPU, 12...ROM, 13...RAM, 14...bus, 15...input / output interface, 16...input unit, 17...output unit, 18...storage unit, 19...communication unit, 20...drive, 30...removable media, 51...index acquisition unit, 52...weather forecasting unit, 53...weather observation data acquisition unit, 54...learning unit, 61...regression coefficient of forecasting formula

Claims

1. An information processing device comprising: an index acquisition means for acquiring an index of a remote area from a calculation means for calculating an index focusing on meteorological phenomena correlated between a prediction area and a remote area; and a weather forecasting means for performing a long-term weather forecast in the prediction area based on the index of the remote area.

2. The information processing device according to claim 1, wherein the indicator of the remote region is at least one of convective activity and sea surface temperature in the ocean of the remote region.

3. An information processing device as described in claim 1 or 2, further comprising a learning means for generating a model that explains the results of long-term weather observation of the prediction area using the index of the remote area by analyzing the relationship between simulation data of the index of the remote area for a predetermined period of time using a past numerical forecast model and actual weather observation data of the prediction area using a predetermined machine learning method, wherein the index acquisition means acquires a prediction of the index of the remote area for the predetermined period of time using a new numerical forecast model, and the weather forecasting means inputs the acquired prediction of the index of the remote area into the model and uses the output of the model as the long-term weather forecast in the prediction area.

4. An information processing method executed by an information processing device, comprising: an index acquisition step of acquiring an index for a remote area from a calculation means that calculates an index focusing on meteorological phenomena that are correlated between a prediction area and a remote area; and a weather forecasting step of making a long-term weather forecast for the prediction area based on the index for the remote area.

5. A program for causing a computer to execute a control process including: an index acquisition step for acquiring an index for a remote area from a calculation means for calculating an index focusing on meteorological phenomena correlated between a prediction area and a remote area; and a weather forecasting step for making a long-term weather forecast for the prediction area based on the index for the remote area.

Images