JP4908346B2 - Weather forecast data analysis apparatus and weather forecast data analysis method - Google Patents

Description

  The present invention relates to a weather prediction data analysis apparatus and a weather prediction data analysis method for analyzing prediction data output from a weather prediction system.

In a conventional weather prediction system, weather prediction is performed by calculating the atmospheric flow using observation data obtained by a weather radar or the like, or GPV (Grid Point Value) data provided by the Japan Meteorological Agency. Weather forecast information is not only familiar to people but also important information related to life and property, such as typhoons and torrential rains, so that the reliability of forecast data is improved. For example, a method has been proposed that allows the accuracy of prediction data to be continuously maintained even when the time interval for obtaining observation data is long (see, for example, Patent Document 1).
JP 2003-090888 A

  However, the weather prediction information is only a prediction, and has uncertainties accompanied by temporal and spatial deviations. However, the value of prediction information exists. For example, if the possibility of the occurrence of torrential rain is predicted in advance, the information is sufficiently valuable even if the possibility is low. In other words, the way of providing weather forecast information determines the value, and there is a problem in the providing method.

  The present invention has been made paying attention to the above circumstances, and its object is to provide a weather prediction data analysis apparatus and a weather prediction capable of providing an accurate index for judging various effects due to weather phenomena. It is to provide a data analysis method.

  In order to achieve the above object, a weather prediction data analysis apparatus according to the present invention includes weather observation data and acquisition means for acquiring weather prediction data calculated based on a weather prediction model based on the weather observation data; Determination means for determining whether or not a plurality of weather determination elements can be adapted to a specific weather phenomenon based on the acquired weather prediction data and weather observation data, and the weather acquired by the acquisition means for a certain period of time About the prediction data and the weather observation data, a totaling unit that totals the determination results of each weather determination element by the determination unit, and a dependency relationship between the plurality of weather determination elements based on the totaled results Creating means for creating a conditional probability table (CPT) that represents the conditional probability in the method; Probability calculation means for calculating the occurrence probability of the specific weather phenomenon with reference to a conditional probability table corresponding to the weather determination element for the determination result by the determination means when the prediction data and the weather observation data are acquired; And a notification means for notifying the probability information of the specific weather phenomenon when the calculated occurrence probability exceeds a threshold value.

  In the weather prediction data analysis apparatus having the above-described configuration, the weather prediction data and the weather observation data acquired for a certain period are respectively determined as to whether or not a plurality of weather determination elements can be applied to a specific weather phenomenon, and the result of the determination is calculated. A conditional probability table representing the conditional probability in the dependency relationship between each element is created. Then, when new weather forecast data and weather observation data are acquired, each element is determined for the acquired data, and the conditional probability table corresponding to the weather determination element is referred to based on the determination result. The probability of occurrence of a specific meteorological phenomenon is calculated. In this way, for example, it is possible to notify the probability of occurrence of a specific weather phenomenon such as snowfall that affects the means of transportation, so an accurate index for determining various effects due to the weather phenomenon Can be provided.

  Therefore, according to the present invention, it is possible to provide a weather prediction data analysis apparatus and a weather prediction data analysis method capable of providing an accurate index for judging various effects due to weather phenomena.

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.
FIG. 1 is a functional block diagram showing an embodiment of a weather forecast data analyzing apparatus according to the present invention. The meteorological prediction data analysis apparatus is connected to the Japan Meteorological Agency data server DS0, radar site servers DS1, DS1, and the like via a network NT.

  In FIG. 1, the weather prediction data analysis apparatus includes a communication interface 11 connected to the network NT, a communication processing unit 12, a wide area prediction data storage unit 13, an observation data storage unit 14, and a local weather prediction model calculation unit 15. And a local prediction data storage unit 16 for performing weather prediction using a local weather prediction model. Further, in order to calculate / provide the occurrence probability of the remarkable weather phenomenon from the prediction data stored in the local prediction data storage unit 16, the probability calculation unit 17, the conditional probability table creation unit 18, and the data determination / distribution unit 19 With.

  Observation data / prediction data as a basis for weather prediction is input from the Japan Meteorological Agency data server DS0 and radar site servers DS1, DS2 via the network NT. The input observation data is stored in the observation data storage unit 14. The wide-area prediction data storage unit 13 stores, for example, a Japan Meteorological Agency RSM (Regional Spectral Model) as wide-area weather prediction data. The data stored in the observation data storage unit 14 and the wide area prediction data storage unit 13 are selectively sent to the local weather prediction model calculation unit 15 in response to a request from the local weather prediction model calculation unit 15.

  The local weather prediction model is a method for predicting atmospheric pressure, wind, temperature, rainfall, etc., which is a weather field, by solving a fluid dynamics equation system by time integration. Currently, “non-hydrostatic meteorological models” that can be predicted in the range of about 100 m to 10 km of horizontal grid spacing have been developed. For example, in addition to CReSS (Cloud Resolving Storm Simulator), there are calculation codes such as Japan Meteorological Agency NHM (JMA NonHydrostatic Model), MM5 (PSU / NCAR mesoscale model). Each model takes into account a physical process on a horizontal scale of developed cumulonimbus clouds (about 10 km).

  In addition to atmospheric pressure, wind, temperature, rainfall, etc., which are generally distributed as forecast data, physical quantities calculated using the local weather model can also calculate snowfall.

  The local weather forecast model has a high resolution but requires a large-scale computer system as the calculation area widens.Therefore, although the resolution is lower than usual, a wide range of weather forecast data with a wide calculation area is the initial value. As a boundary value, weather forecast in a limited area is performed.

  The local weather prediction model calculation unit 15 is activated along with the distribution of the wide area weather prediction data, and executes the weather prediction calculation based on the data stored in the wide area prediction data storage unit 13 and the observation data storage unit 14. . For example, in the case of the Japan Meteorological Agency RSM, it is delivered every 12 hours. The prediction calculation result output from the local weather prediction model calculation unit 15 is stored in the local prediction data storage unit 16.

  In addition, the observation data storage unit 14 stores weather radar data distributed every 10 minutes, for example, with a short distribution interval. When new observation data is input, the local weather prediction model calculation unit 15 starts again, takes out the calculated value predicted by the local weather prediction model every 12 hours, for example, every hour, and 1 hour before that The rain and wind speed data obtained from the latest meteorological radar data of the minute are corrected by recalculating the local weather forecast model using a data assimilation method called “nudge method” to improve the prediction accuracy several hours ahead. A prediction calculation is performed by repeating the above processing, and a prediction value as a result is stored in the local prediction data storage unit 16, and the local prediction data is provided in a form such as a screen display.

  Further, the conditional probability table creation unit 18 creates a conditional probability table (CPT) that represents the conditional probability in the dependency relationship of a plurality of weather determination elements for a specific weather phenomenon. The probability calculation unit 17 refers to the conditional probability table based on the prediction data and the observation data stored in the wide area prediction data storage unit 13 and the observation data storage unit 14, and calculates the occurrence probability value of a specific weather phenomenon. To do. Details of the probability calculation process will be described later. The calculated probability is notified to the user only when the data determination / distribution unit 19 calculates a probability equal to or higher than a certain threshold. As a notification method, for example, mail delivery to a mobile phone terminal can be considered.

Next, the probability calculation process that is a feature of the present invention will be described.
For the probability calculation process, a Bayesian Network method is used. A Bayesian network is a method of counting the interaction of probabilities for a chain of uncertain events, and its application is spreading to cognitive science, data mining, etc. as a “computation model for handling uncertainty”. FIG. 2 shows an example of a Bayesian network used in the present invention. In the present embodiment, it is assumed that the probability of snowfall is obtained as an example of a specific weather phenomenon. A plurality of weather determination elements are represented by nodes 1 to 5, and a link extending between the nodes represents a direction of dependency. A conditional probability table is created for each link. The Bayesian network shown in FIG. 2 is merely an example, and other network configurations can be considered.

(Conditional probability table creation / update process)
First, the processing of the conditional probability table creation unit 18 will be described. FIG. 3 shows a procedure for creating a conditional probability table and a flow showing the contents in a DFD (Data Flow Diagram) format. FIG. 5 is a diagram illustrating an example of a determination result for each link. FIG. 6 is a diagram showing a conditional probability table created based on the aggregation results shown in FIG.

  In FIG. 3, the conditional probability table creation unit 18 is based on the meteorological observation data, the wide-area prediction data, and the local weather prediction data calculated by the local weather prediction model calculation unit 15. It is determined whether or not the determination element is compatible (step S3a). By this determination, for example, a node determination result table as shown in FIG. 4 is created. FIG. 4 shows the result of determining whether or not the weather determination elements of the nodes 1 to 5 are compatible (True / False) based on the data read every hour.

  Next, the conditional probability table creation unit 18 aggregates the determination results for each link based on the node determination result table (step S3b). FIG. 5 shows an example of determination results for each link. FIG. 5 shows a result of aggregation of the link between the node 1 and the node 2, and a to d represent the frequency at which a set of determination results of the node 1 and the node 2 occurs in a certain period.

  The conditional probability table creation unit 18 creates a conditional probability table for each link on the basis of the totaled determination totalization results (step S3c). FIG. 6 is a diagram showing a conditional probability table created based on the aggregation results shown in FIG. A conditional probability value in the dependency relation between the node 1 and the node 2 is calculated.

  The conditional probability table creation unit 18 uses the data accumulated daily in the wide area prediction data storage unit 13, the observation data storage unit 14, and the local prediction data storage unit 16. The conditional probability value is updated. By updating the conditional probability table based on the latest data, the probability of the probability value calculated by the probability calculation unit 17 can be improved.

(Probability calculation process)
The probability calculation unit 17 calculates the occurrence probability of a specific weather phenomenon using the conditional probability table created in this way. FIG. 7 shows a flow of the creation process of the probability calculation process and the contents thereof in a DFD (Data Flow Diagram) format.

  When new data is stored in the wide area prediction data storage unit 13, the observation data storage unit 14, or the local prediction data storage unit 16, the probability calculation unit 17 performs determination for each node based on this data. Probability calculation processing is performed based on the node determination result by this determination and the conditional probability table created by the conditional probability table creating unit 18 (step S7a).

A probability calculation method using a Bayesian network will be described with reference to FIG. Here, it is assumed that the probability of occurrence of snowfall at node 2 is obtained. The probability calculation unit 17 uses the data stored in the wide area prediction data storage unit 13, the observation data storage unit 14, or the local prediction data storage unit 16 for the weather determination elements of the nodes 1, 3, 4, and 5. Judgment of suitability is made respectively. For example, assume that node 1 = True, node 3 = True, node 4 = False, and node 5 = False. At this time, on the basis of the determination result of each node and the conditional probability table corresponding to that node, the probability P _2T that the node 2 is True can be calculated as shown in FIG. 8 by Bayes' theorem. The probability calculation unit 17 outputs the probability value obtained in this way.

  As described above, in the above-described embodiment, the conditional probability table creation unit 18 determines whether or not a plurality of weather determination elements can be applied to a specific weather phenomenon with respect to the weather prediction data and the weather observation data acquired in a certain period. A conditional probability table representing the conditional probabilities in the dependency relationship between each element is created based on the result of the determination. When new weather prediction data and weather observation data are acquired, the probability calculation unit 17 determines each element for the acquired data, and a conditional probability table corresponding to the weather determination element based on the determination result. The occurrence probability of a specific meteorological phenomenon is calculated with reference to FIG. In this way, for example, it is possible to notify the probability of occurrence of a specific weather phenomenon such as snowfall that affects the means of transportation, so an accurate index for determining various effects due to the weather phenomenon Can be provided. Furthermore, by updating the created conditional probability table with the latest observation data and prediction data, it is possible to calculate a probability value with a higher probability.

  Therefore, according to the above-described embodiment, it is possible to accurately provide information obtained from weather prediction data by eliminating uncertainty of the prediction data as much as possible. By providing this information, high-value information that supports judgment is provided to users who need to judge the risk of weather phenomena, such as evacuation guidance for residents, dam drainage operation, and operation of urban sewer pump stations. be able to.

The present invention is not limited to the above embodiment. In the above-described embodiment, the method of calculating the probability value of snowfall is described as an example. However, the present invention can be applied to the calculation of a probability value of an integrated rainfall amount that causes a sediment disaster or the like. The above-described determinations (1) to (5) are made by using numerical data and comparing with each threshold value. By changing the threshold value, it is possible to easily calculate probability values such as “snowfall” and “sleet”.

  The Bayesian network shown above is only an example, and it is of course possible to calculate probability values related to weather disaster prevention such as “strong wind” and “heavy rain” by constructing other Bayesian networks. However, it is the same in that an event that supports the event to be predicted is included in the node.

  That is, the present invention is not limited to the above-described embodiment as it is, and can be embodied by modifying the constituent elements without departing from the scope of the invention in the implementation stage. Further, various inventions can be formed by appropriately combining a plurality of constituent elements disclosed in the embodiment. For example, some components may be deleted from all the components shown in the embodiment. Furthermore, you may combine suitably the component covering different embodiment.

The functional block diagram which shows one Embodiment of the weather forecast data analysis apparatus which concerns on this invention. The figure which shows an example of the Bayesian network for calculating the occurrence probability of snowfall in the weather prediction data analysis apparatus shown in FIG. The figure which shows the procedure of the creation process of a conditional probability table | surface, and its content. The figure which shows an example of a node determination result table. The figure which shows an example of the determination total result for every link. The figure which shows an example of a conditional probability table | surface. The figure which shows the procedure of the probability calculation process, and its content. The figure explaining the probability calculation method using a Bayesian network.

Explanation of symbols

DESCRIPTION OF SYMBOLS 11 ... Communication interface, 12 ... Communication processing part, 13 ... Wide area prediction data storage part, 14 ... Observation data storage part, 15 ... Local weather prediction model calculating part, 16 ... Local prediction data storage part, 17 ... Probability calculation part , 18 ... Conditional probability table creation unit, 19 ... Data determination / distribution unit, NT ... Network, DS0 ... Japan Meteorological Agency data server, DS1, DS2 ... Radar site data server.

Claims (6)

An acquisition means for acquiring weather observation data;
A computing means for computing weather forecast data relating to a specific weather phenomenon based on a weather forecast model based on the weather observation data;
Based on the weather forecast data and meteorological observation data, and each determining means for determining whether the adaptation of the plurality of meteorological determination factors for the particular weather events,
For the weather forecast data and meteorological observation data obtained in the predetermined period, the collecting unit that aggregates the determination results of the weather determining element by the determining means, respectively,
Creating means for creating a conditional probability table (CPT) representing a conditional probability in a dependency relationship between elements for each of the plurality of weather determination elements based on the aggregated results;
When the weather forecast data and weather observation data are acquired, probability calculation for calculating the occurrence probability of the specific weather phenomenon with reference to the conditional probability table corresponding to the weather determination element for the determination result by the determination means Means,
Notification means for notifying the probability information of the specific weather phenomenon when the calculated occurrence probability exceeds a threshold ,
The weather prediction data analyzing apparatus , wherein the weather determination element is determined in consideration of temporal and spatial deviations caused by the weather prediction model .   The weather prediction data analysis apparatus according to claim 1, wherein the conditional probability table is created based on a Bayesian network configured by dependency relationships between the plurality of weather determination elements. Said generating means, when the weather forecast data and the meteorological data is obtained, the obtained weather forecast data and the value of the meteorological data and the created probability probabilities table for each conditional conditions based on The weather prediction data analysis apparatus according to claim 1, wherein the weather prediction data analysis apparatus is updated. Get the weather observation data,
Calculate weather forecast data related to a specific weather phenomenon based on the weather forecast model based on the weather observation data,
Based on the weather forecast data and meteorological observation data, a plurality of whether adaptation of meteorological determination element determines respectively for the particular weather events,
For the acquired weather forecast data and weather observation data for a certain period, the determination results of each of the determined weather determination elements are totaled,
A conditional probability table (CPT) representing a conditional probability in a dependency relation between elements for each of the plurality of weather determination elements based on the aggregated results is created,
When the weather forecast data and weather observation data are acquired, referring to a conditional probability table corresponding to the weather determination element for the determined result, the occurrence probability of the specific weather phenomenon is calculated,
Notifying the probability information of the specific weather phenomenon when the calculated occurrence probability exceeds a threshold ,
The weather prediction data analysis method , wherein the weather determination element is determined in consideration of a temporal and spatial shift caused by the weather prediction model .   5. The weather forecast data analysis method according to claim 4, wherein the conditional probability table is created based on a Bayesian network configured by dependency relationships between the plurality of weather determination elements. When the weather prediction data and the weather observation data are acquired, the conditional probability values of the prepared conditional probability tables are updated based on the acquired weather prediction data and weather observation data. The method for analyzing weather forecast data according to claim 4.

Images