Weather management system and method of weather management

MYPI2025003756A0Undetermined Publication Date: 2024-07-13KK TOSHIBA +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
MYPI2025003756
Authority / Receiving Office
MY · MY
Patent Type
Applications
Current Assignee / Owner
Priority Date
2023-03-16
Filing Date
2023-05-17
Publication Date
2024-07-13

AI Technical Summary

Technical Problem

Current weather radar systems provide estimates of meteorological phenomena that may differ from actual events due to limitations in observation equipment, particularly for rare events like hail and tornadoes, which are infrequently observed and difficult to deploy disdrometers for, leading to inaccuracies in analysis based on weather radar data.

Method used

An information management device that collects and compares weather-related posted information from web servers with weather radar observation data to provide true values of weather phenomena, using dual-polarization meteorological Doppler radar to distinguish precipitation types and intensities, and adjusts analysis criteria based on real-time feedback from posted information to improve accuracy.

Benefits of technology

Enhances the accuracy of weather phenomenon classification and prediction by integrating user-reported data with radar observations, reducing the likelihood of missed events and improving warning systems for hazardous weather conditions.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure PI2025003756_DRAWING_001
    Figure PI2025003756_DRAWING_001
  • Figure PI2025003756_DRAWING_002
    Figure PI2025003756_DRAWING_002
  • Figure PI2025003756_DRAWING_003
    Figure PI2025003756_DRAWING_003
Patent Text Reader

Abstract

To provide a weather management system used a weather phenomenon corresponding to an observation data (231) from a weather radar, the weather phenomenon acquired from weather-related posted information. A weather management system according to embodiments includes an analysis section (221) configured to classify observation data (231) from a weather radar according to a predetermined criterion and output a result of analysis based on the classification and a criterion control section (222) configured to change the criterion based on a weather phenomenon found in weather-related posted information, the weather-related posted information posted to a Web server (300) configured to receive posted information (131) from poster terminals (T) and matched the observation data (231) observed by a weather radar. Fig. 1
Need to check novelty before this filing date? Find Prior Art

Description

Information management device and weather management device

[0001] An embodiment of the present invention relates to a technology for comparing observation results based on observation data observed by a weather radar with actual weather phenomena and providing true values ​​(levels of accuracy) of weather phenomena corresponding to the observation data of the weather radar.

[0002] Weather radar emits radio waves (microwaves) and observes rain or snow within a specified range. It measures the distance to the rain or snow from the time it takes for the emitted radio waves to return, and can observe the intensity of the rain or snow from the strength of the returned radio waves.

[0003] In recent years, dual-polarization meteorological Doppler radar has been introduced, which uses radio waves that vibrate horizontally (horizontal polarization) and radio waves that vibrate vertically (vertical polarization), making it possible to more accurately distinguish the types of precipitation particles in clouds and estimate the intensity of precipitation.

[0004] Park, H. -S and AV Ryzhkov, DS Zrnic and K. -E Kim 2009: The Hydrometeor Classification Algorithm for the Polarimetric WSR-88D:Description and Application to an MCS. Wea. Forecasting., 24, 730-748 Hideaki Kagesawa (and five others), Title: "Proposal and Consideration of a Sensing System Using Twitter", Publication Year: July 2, 2014, URL: https: / / ipsj.ixsq.nii.ac.jp / ej / ?action=pages_view_main&active_action=repository_view_main_item_detail&item_id=104972&item_no=1&page_id=13&block_id=8

[0005] An information management device is provided that compares observation results based on observation data observed by weather radar with actual weather phenomena and provides true values ​​(correct levels) of weather phenomena corresponding to the observation data of the weather radar.

[0006] The information management device of the embodiment has an information collection unit that acquires weather-related posted information from multiple posted information posted to a web server that accepts posted information from poster terminals, an extraction unit that extracts the weather-related posted information that matches the observation area and observation time of observation data observed by weather radar, and a data management unit that outputs meteorological phenomena identified from the extracted weather-related posted information as meteorological phenomena corresponding to the observation data.

[0007] 1 is a diagram illustrating the functional blocks and network configuration of an information management device of a first embodiment. FIG. 1 is a diagram illustrating the generation process of true value information linking meteorological phenomena grasped from posted information as meteorological phenomena corresponding to observation data of the first embodiment. FIG. 2 is a diagram illustrating the process of extracting weather-related posted information from posted information using various evaluation information of the first embodiment. FIG. 3 is a diagram illustrating the processing flow executed by the information management device of the first embodiment. FIG. 4 is a diagram illustrating the functional blocks and network configuration of a weather management device of a second embodiment. FIG. 5 is a diagram illustrating the judgment criterion change processing of the second embodiment. FIG. 6 is a diagram illustrating the processing flow executed by the weather management device of the second embodiment. FIG. 7 is a diagram illustrating the functional blocks and network configuration of an information management device of a third embodiment. FIG. 8 is a diagram illustrating the first function of classifying weather-related posted information based on insurance payment history and the second function of classifying weather-related posted information using a classification model of the third embodiment. FIG. 9 is a diagram illustrating the processing flow of the first function executed by the classification control unit of the third embodiment. FIG. 10 is a diagram illustrating the processing flow executed by a weather management device linked to an information management device in the third embodiment. FIG. 11 is a diagram illustrating the functional blocks and network configuration of a weather management device of a fourth embodiment. FIG. 12 is a diagram illustrating the processing flow executed by the weather management device of the fourth embodiment. It is a figure which shows the processing flow which is executed by the weather management device of 4th embodiment. It is a figure which shows the processing flow which is executed by the weather management device of 4th embodiment. It is a figure which shows the functional block and the network configuration of the weather management system of 4th embodiment.

[0008] Hereinafter, embodiments will be described with reference to the drawings.

[0009] As mentioned above, weather radars such as dual-polarization meteorological Doppler radars make it possible to distinguish between different types of precipitation particles in clouds and estimate the intensity of precipitation more accurately. However, analysis results based on observational data from weather radars are merely "estimates" and may differ from the actual weather phenomena occurring at the site. Precipitation particles are classified into categories such as wet snow, dry snow, ice crystals, drizzle, rain, graupel, hail, and heavy rain. For example, in an observation area where observational data indicates that "wet snow fell in this observation area," the actual weather phenomenon occurring was "hail."

[0010] Therefore, the Japan Meteorological Agency is working to improve the accuracy of estimates based on observation data by combining weather radar observation data with data from observation equipment such as rain gauges owned by the Japan Meteorological Agency, the Ministry of Land, Infrastructure, Transport and Tourism, and local governments across the country, and using weather phenomena actually observed at or near the site.

[0011] However, while sensor devices such as rain gauges and snow gauges can determine whether the amount of rainfall or snow accumulation was high or low, they cannot determine whether it was drizzle, wet snow, dry snow, hail, or sleet. Disdrometers (precipitation particle size and velocity distribution measurement devices) can observe drizzle, rain, hail, and snow, but they are mostly used for research purposes and are not deployed at the Japan Meteorological Agency's AMeDAS observation stations nationwide. Hail and sleet, in particular, occur infrequently and are rarely observed on the ground, making the deployment of disdrometers unrealistic.

[0012] In this way, weather phenomena that cannot be observed or are difficult to observe using the observation equipment at AMeDAS observation stations are difficult to feed back into analyses based on weather radar observation data, making it difficult to improve the accuracy of analyses based on weather radar observation data.

[0013] In addition to weather phenomena such as hail and sleet, gusty winds such as tornadoes and whirlwinds also occur infrequently and are rarely observable on the ground. Although it is possible to observe the occurrence of cumulonimbus clouds based on weather radar observation data and estimate the possibility of tornadoes and other phenomena occurring, their low frequency and limited range of occurrence make them difficult to observe with observation equipment.

[0014] Therefore, we developed a new technology that compares observation results based on weather radar data with actual weather phenomena, and provides true values ​​(level of accuracy) of weather phenomena corresponding to the weather radar observation data. Specifically, we use weather-related posted information posted from the poster's terminal to a web server that accepts posted information.

[0015] In addition, in a weather management device that classifies and analyzes weather radar observation data based on predetermined criteria, the criteria can be changed based on the true value (correct level) of the weather phenomenon based on the posted information, for example, by lowering the criteria, thereby preventing the occurrence of a predetermined weather phenomenon from being overlooked.

[0016] 1 to 4 are diagrams for explaining a first embodiment of the present invention. Fig. 1 is a diagram showing the functional blocks and network configuration of an information management device 100 according to the present embodiment.

[0017] The information management device 100 is provided with information related to weather radar observation data from the weather management device 200 and with posted information posted by posters from the web server 300. The information management device 100 provides a function for extracting and generating true values ​​(correct labels) of meteorological phenomena identified from weather-related posted information posted by posters as meteorological phenomena corresponding to observation data observed by weather radar. In other words, the information management device 100 functions as a device for generating a dataset of observation data and correct labels corresponding to the observation data.

[0018] Weather radar, for example, emits radio waves (microwaves) to observe rain or snow. The distance to the rain or snow is measured from the time it takes for the emitted radio waves to return, and the intensity of the rain or snow is observed from the strength of the returned radio waves. Furthermore, as mentioned above, dual-polarized meteorological Doppler radar has been introduced, and by using radio waves that vibrate horizontally and vertically (horizontal polarization and vertical polarization), it is possible to distinguish the type of precipitation particles in clouds and estimate the intensity of precipitation.

[0019] Dual-polarized meteorological Doppler radar can estimate the shape of precipitation particles from the ratio of their amplitudes. The larger the precipitation particles, the more air resistance they experience and the flatter they become. This is observed using horizontally and vertically polarized waves, and the shape of the precipitation particles can be estimated from the amplitude ratio of the reflected waves. The intensity of rain can also be estimated from the phase difference. When radio waves travel through water, such as raindrops, their speed is slightly slower than when they travel through empty air. Taking advantage of the fact that the heavier the rain, the slower the speed of horizontally polarized waves, observations are made using horizontally and vertically polarized waves, and the intensity of rain can be estimated from the phase difference of the reflected waves.

[0020] In precipitation particle estimation based on amplitude ratios, categories such as wet snow, dry snow, ice crystals, drizzle, rain, graupel, hail, and heavy rain are predefined. For example, if the amplitude ratio is equal to or greater than a predetermined value, the precipitation particles are determined (categorized) as hail. The amplitude ratio is set as the determination criterion, and meteorological phenomena in the observation area are estimated based on observation data. Hail is defined as ice particles with a diameter of 5 mm or more that fall from cumulonimbus clouds, while ice particles with a diameter of less than 5 mm are defined as graupel.

[0021] The web server 300 accepts posted information from the poster terminal T and provides a website that publishes the accepted posted information on the web. Posters are users registered with the web server 300, and include an unspecified number of posters. As an example, information posted to a social networking service (SNS) such as Twitter (registered trademark) or Facebook (registered trademark) can be used. In this case, the web server 300 also stores posted information other than weather-related posted information.

[0022] As another example, information posted to a website that accepts weather-related information can be used. For example, a contributor may be a weather reporter or observer who posts daily weather information. In this case, only weather-related posted information is stored on the web server 300.

[0023] The information management device 100 connects to one or more websites 300 to acquire weather-related posted information. The poster terminal T is a mobile terminal such as a multi-function mobile phone such as a smartphone or a tablet computer, and is equipped with a data communication function via an IP (Internet Protocol) network or a mobile communication network, a calculation function (CPU, etc.), and a storage device (memory, auxiliary storage device, etc.).

[0024] The poster terminal T may also include a display control application such as a browser, a camera for capturing still images and / or moving images, a touch panel display input device, and a GPS device. As will be described later, the posted information posted from the poster terminal T includes the posted content as well as location information and time information acquired by the GPS device.

[0025] The weather forecasting device 400 provides a weather forecasting function using a weather forecasting model, such as the Japan Meteorological Agency's mesoscale model or the WRF mesoscale weather model, and provides forecast results for weather phenomena (snow, rain, hail, sleet, etc.) for a forecast area and forecast time. The weather forecasting device 400 (weather forecasting model) is a well-known technology, and detailed description thereof will be omitted.

[0026] As shown in Fig. 1, an information management device 100 is connected to weather management information 200, a website 300, and a weather forecasting device 400 via an IP network or a dedicated line. A communication device 110 controls data communication between each device. The information management device 100 includes a control device 120 and a storage device 130. The control device 120 includes an information acquisition unit 121, an extraction unit 122, and a data management unit 123.

[0027] FIG. 2 is a diagram for explaining the process of generating true value information in which meteorological phenomena identified from posted information are linked as meteorological phenomena corresponding to observation data in this embodiment.

[0028] The posted information posted to the web server 300 includes a posted article and / or a posted image, location information, time information, and poster information. A posted article is a posted text, and a posted image is an image taken with a camera provided in the poster terminal T, or an image taken with a camera separate from the poster terminal T. The location information is the current location of the poster (poster terminal T) and may include address information corresponding to the current location. The posted time is date and time information when the web server 300 received the posted information. The poster information is a user identifier such as a user ID registered with the web server 300.

[0029] A poster inputs a post article and a post image on a posting screen provided by web server 300 via poster terminal T, and transmits the posted information to web server 300. Web server 300 accepts the posted information transmitted from poster terminal T, and posts the posted information on a posting website provided by web server 300 for public viewing.

[0030] Although the above description has been given of an example in which the information management device 100 acquires posted information, i.e., collects posted information, from a web server 300 that accepts and publishes posted information, the present invention is not limited to this. For example, posted information may be acquired via a search server (search site). A search site can collect posted information related to search keys such as keywords from multiple different posting websites (multiple different websites 300). Thus, the information management device 100 can acquire posted information posted to the web server 300 from the website 300 or from a search site that can collect posted information from the website 300.

[0031] As shown in FIG. 2 , the information acquisition unit 121 acquires weather-related posted information from multiple posted information on the web server 300 based on the similarity of weather-related keywords or weather-related images. One or more pieces of weather-related posted information can be acquired. For example, posted information containing the keyword "hail" can be acquired as weather-related information, or posted information having posted images similar to a reference image containing "hail" can be acquired as weather-related posted information. While "hail" has been used as an example, weather-related posted information related to "graupel" can be similarly collected. Note that an image recognition AI model can also be used to extract posted information having posted images related to "hail." For example, a group of images containing "hail" can be input as training data in advance, and a training process can be performed to generate a "hail image recognition" AI model. The generated AI model can then be used to calculate the similarity of the posted image to "hail," and posted information having a similarity equal to or greater than a predetermined value can be acquired as weather-related information.

[0032] Next, the extraction unit 122 performs a matching process with the observation data of the weather radar, that is, extracts weather-related posted information that matches the observation area and observation time of the observation data observed by the weather radar.

[0033] For example, the extraction unit 122 matches the observation area and observation time determined to be hail in the analysis results of weather radar observation data, i.e., in the estimation of precipitation particles based on the amplitude ratio, with the location information and posting time of the weather-related posted information related to "hail." The extraction unit 122 determines whether the location information and posting time of the weather-related posted information related to "hail" fall within a predetermined range that is sufficiently close to the observation area and observation time, and extracts the weather-related posted information related to "hail" that is within the sufficiently close predetermined range.

[0034] The data management unit 123 associates the weather-related posted information that matches the observation data as a weather phenomenon corresponding to the observation data. In other words, the data management unit 123 outputs the weather phenomenon identified from the weather-related posted information extracted by the extraction unit 122 as a weather phenomenon corresponding to the observation data.

[0035] The data management unit 123 outputs weather phenomena identified from weather-related posted information as true values ​​(correct labels) of weather phenomena corresponding to the observation data, and can function as a data generation unit that generates a dataset including the observation data and the true values ​​(correct labels) of the weather phenomena. This dataset can be fed back to the weather management device 200 as supervised learning data and can be used to tune the analysis process of the weather radar observation data described above, the analysis algorithm, and the judgment criteria (thresholds) used in the analysis process. Note that as long as the true values ​​(correct labels) of the weather phenomena can be linked to the corresponding observation data, the observation data itself does not need to be stored as a dataset. For example, when feeding back information to the weather management device 200, the information management device 100 (data management unit 123) can generate and provide a dataset including the observation area, observation time, and the true values ​​(correct labels) of the output weather phenomena.

[0036] In the above example, in estimating precipitation particles from observation data, a mode of matching observation data determined to be hail with weather-related posted information related to "hail" has been described. However, for example, the observation area and observation time determined to be hail may be matched with the location information and posting time of weather-related posted information related to "hail," and "hail" may be linked as the true value (correct label) of the weather phenomenon to the observation data for the observation area and observation time determined to be hail. The same is also true vice versa.

[0037] In this way, the information management device 100 of this embodiment checks the weather phenomena estimated from the analysis results of the observation data using posted information that can be collected on the web, and generates the weather phenomena identified from the weather-related posted information as true values ​​(correct answer labels) of the weather phenomena corresponding to the observation data.

[0038] In particular, for meteorological phenomena such as hail and sleet, which occur infrequently and are rarely observed on the ground, it will be possible to cross-check the results of analysis of observation data, thereby improving the accuracy of analysis based on weather radar observation data.

[0039] FIG. 3 is a diagram for explaining the process of extracting weather-related posted information from posted information using various types of evaluation information according to this embodiment.

[0040] The information collection unit 121 extracts weather-related posted information from a plurality of pieces of posted information, and can evaluate the reliability of the posted information and extract weather-related posted information that has an evaluation above a certain standard. The information collection unit 121 can include first to fourth processing units.

[0041] As shown in Figure 3, the first processing unit 121A of the information collection unit 121 extracts candidate posted information from multiple posted information posted to the web server 300 based on the similarity of weather-related keywords or weather-related images (first processing).

[0042] The second processing unit 121B of the information collection unit 121 generates a poster evaluation value for the poster of the candidate posted information based on a predetermined evaluation criterion (second processing). The poster evaluation value can be calculated using evaluation parameters (evaluation criteria), such as the poster's number of past posts, the number of evaluations from other posters, and the number of times that past posted information was adopted as a true value (correct label). The poster's number of past posts and the number of evaluations from other posters can be acquired from the posting history managed by the web server 300. The information management device 100 can accumulate posters of weather-related posted information that were adopted as a true value (correct label). The number of times that past posted information was adopted as a true value (correct label) is stored in the storage device 130 as poster evaluation information 132.

[0043] The third processing unit 121C of the information collection unit 121 calculates the number of pieces of posted information by other posters that are similar to the candidate posted information extracted by the first processing unit 121A as the posting accuracy of the candidate posted information. Specifically, the third processing unit 121C extracts, from the multiple pieces of posted information posted to the web server 300, pieces of posted information by other posters that include location information and time information within a predetermined range close to the location information and time information of the candidate posted information and that have similar weather-related keywords or weather-related images. The third processing unit 121C generates a posting accuracy evaluation value for the candidate posted information based on the extracted similar posted information (third processing).

[0044] The fourth processing unit 121D of the information collecting unit 121 evaluates the information accuracy of the candidate posted information extracted by the first processing unit 121A using the forecast information generated by the weather forecasting device 400. Specifically, the fourth processing unit 121D generates a weather event evaluation value for the candidate posted information based on forecast information generated by a predetermined weather forecast model corresponding to the location information and time information of the candidate posted information (fourth processing). If the weather phenomenon in the location information and time information of the candidate posted information is similar to the weather information in the forecast information, the weather phenomenon evaluation value is calculated to be high, and if the weather phenomenon in the location information and time information of the candidate posted information differs from the weather information in the forecast information, the weather phenomenon evaluation value is calculated to be low.

[0045] The information collection unit 121 then extracts weather-related posted information from candidate posted information whose evaluations are equal to or greater than a predetermined value based on the poster evaluation value, the posting accuracy evaluation value, and the weather event evaluation value. For example, the information collection unit 121 can extract candidate posted information whose evaluation values ​​are equal to or greater than a predetermined value as weather-related posted information, or extract candidate posted information whose total evaluation values ​​is equal to or greater than a predetermined value as weather-related posted information. Furthermore, weights may be applied to the poster evaluation value, the posting accuracy evaluation value, and the weather event evaluation value. For example, the weights for the poster evaluation value and the posting accuracy evaluation value may be set high, and the weight for the weather event evaluation value may be set low, to calculate the total evaluation value. In this case, the weights may be set arbitrarily.

[0046] The poster evaluation value, the posting accuracy evaluation value, and the weather event evaluation value can be configured to be each independently or in any combination. The information collection unit 121 can be configured to evaluate candidate posted information using at least one of the poster evaluation value, the posting accuracy evaluation value, and the weather event evaluation value, or any combination thereof, and extract weather-related posted information.

[0047] The extraction process for weather-related posted information using the evaluation values ​​described above may not be applied depending on the characteristics of the posted information. For example, if the weather-related posted information is collected from a weather information posting web server 300 (website) that specializes in accepting weather information posts, the posted information may be treated as information for which the reliability of the poster and the accuracy of the information as posted information have been guaranteed in advance, and the information management device 100 may not select the posted information based on the evaluation values. In other words, if the web server 300 that accepts posted information selects posted information for which the reliability of the poster and the accuracy of the information have been guaranteed and makes it available on the web, the candidate posted information extracted by the first processing unit 121A may be extracted as weather-related posted information as is.

[0048] FIG. 4 is a diagram showing a processing flow executed by the information management device 100 of this embodiment.

[0049] The web server 300 stores posted information from contributors via contributor terminals T. At this time, posted information other than that related to weather is also included.

[0050] The information management device 100 extracts candidate posted information from a plurality of pieces of posted information posted to the web server 300 based on similarities between weather-related keywords or weather-related images (S101). For example, the posted information stored on the web server 300 is searched using keywords such as "hail" or "graupel" or sample images of "hail" or "graupel" as extraction keys. The extraction keys are arbitrarily set by the operator of the information management device 100.

[0051] The candidate posted information is posted information that matches either one of a weather-related keyword and a sample image, or posted information that matches both a weather-related keyword and a sample image.

[0052] Next, the information management device 100 performs the evaluation value calculation process described in FIG. 3 on the extracted candidate posted information (S102). The information management device 100 extracts weather-related posted information from one or more candidate posted information whose evaluation is equal to or greater than a predetermined value based on the calculated poster evaluation value, post accuracy evaluation value, and weather event evaluation value (S103). For example, the candidate posted information with the highest evaluation result based on the evaluation value can be extracted as the weather-related posted information.

[0053] The information management device 100 acquires observation data from the weather management device 200. The information management device 100 extracts weather-related posted information that matches the observation area and observation time of the observation data observed by the weather radar (S104). The information management device 100 then outputs the weather phenomenon identified from the extracted weather-related posted information as the true value (correct label) of the weather phenomenon corresponding to the observation data (S105). The true value (correct label) of the weather phenomenon based on the output posted information is stored in the storage device 130 (S106).

[0054] As described above, the information management device 100 can also function as a data generation unit that generates a data set including observation data and true values ​​(correct labels) of meteorological phenomena. The information management device 100 can provide the data set to the weather management device 200 and the weather forecasting device 400 at the appropriate timing, which can be useful for improving the accuracy of the analysis process of the weather management device 200 and the prediction process of the weather forecasting device 400.

[0055] In the processing steps of the information management device 100 described above, the matching process centered on step S104 includes two modes.

[0056] First, based on the location information and time information of the weather-related posted information extracted in step S103, the information management device 100 (control device 120) can be configured to obtain observation data including the relevant observation area and observation time from the weather management device 200, and then match the two in step S104.

[0057] Second, the weather management device 200 can be configured to obtain in advance observation data that observes specific weather phenomena such as "hail" or "graupel" as analysis results, and in the process of extracting weather-related posted information from steps S101 to S103, extract weather-related posted information that corresponds to the observation area and observation time of the observation data.

[0058] In both processes, weather-related posted information collected from the web server 300 is matched with observation data using location information and time information, and weather-related posted information that matches the observation area and observation time of the observation data observed by the weather radar is extracted, and this is output as the true value (correct label) of the weather phenomenon corresponding to the observation data.

[0059] Second Embodiment Figures 5 to 7 are diagrams for explaining a second embodiment. In this embodiment, a weather management device 200 analyzes observation data and provides a weather management function that outputs alerts based on the analysis results. At this time, true values ​​(correct labels) of meteorological phenomena for the observation data generated by the information management device 100 of the first embodiment are fed back in real time. The weather management device 200 of this embodiment cooperates with the information management device 100 of the first embodiment to form a weather management system.

[0060] 5 is a diagram showing the functional blocks and network configuration of the weather management device 200 of this embodiment. The weather management device 200 includes a communication device 210, a control device 220, and a storage device 230. The control device 220 includes an analysis unit 221, a judgment criteria control unit 222, an alert control unit 223, and a true value information acquisition unit 224.

[0061] The analysis unit 221 classifies the weather radar observation data based on predetermined criteria and outputs analysis results based on the classification. As described above, the analysis unit 221 can perform precipitation particle estimation processing (analysis processing) based on amplitude ratios, and categories (classifications) such as wet snow, dry snow, ice crystals, drizzle, rain, hail, sleet, and heavy rain are prepared in advance. The analysis unit 221 sets the amplitude ratio as the criteria and estimates meteorological phenomena in the observation area based on the observation data.

[0062] The alert control unit 223 outputs alert information based on the analysis results of the analysis unit 221. For example, the alert information can be output to an information distribution system to which multiple users are registered, or to a web server that publishes information on a website, or the alert information can be provided to a user (user terminal) registered in the weather management device 200.

[0063] In this embodiment, a judgment criteria control unit 222 is provided, and the judgment criteria are changed based on the meteorological phenomenon (true value) grasped from the weather-related posted information corresponding to the observation data output from the information management device 100 of the first embodiment.

[0064] FIG. 6 is a diagram for explaining the determination criterion change process of this embodiment.

[0065] The weather management device 200 performs precipitation particle estimation processing using weather radar observation data A. The observation data A was observed at 15:31 on January 5, 2023, in the observation area of ​​City A. The analysis unit 221 outputs a particle diameter of "3 mm" as the analysis result of the observation data A based on the judgment criteria. The analysis unit 221 outputs the particle diameter of "3 mm" and the meteorological phenomenon "hail" corresponding to the analysis result. The alert control unit 223 generates and outputs alert information such as, for example, "Hail may be falling in City A. Please be careful."

[0066] This observation data A is also provided to the information management device 100 in real time in parallel with the analysis process of the weather management device 200, and a process is performed to generate true values ​​(correct labels) of meteorological phenomena based on weather-related posted information for the observation area and observation time. The true values ​​(correct labels) are then linked to the observation data A and provided from the information management device 100 to the weather management device 200. The true value information acquisition unit 224 performs the process of providing the observation data A to the information management device 100 and the process of acquiring true values ​​(correct labels) from the information management device 100.

[0067] When the determination criteria control unit 222 receives the true value (correct label) of the observation data A, it performs a process of comparing the analysis result with the true value (correct label). If the comparison process determines that the meteorological phenomenon in the analysis result differs from the true value (correct label) of the observation data A, the determination criteria control unit 222 temporarily changes the determination criteria. In the example of FIG. 6 , the analysis result is "graupel" and the true value (correct label) is "hail", which are different. The determination criteria control unit 222 changes the determination criteria, assuming that the meteorological phenomenon based on the observation data A is the true value (correct label).

[0068] For example, the criteria for the precipitation particle estimation process may be such that if the amplitude ratio is equal to or greater than a predetermined value X but less than a predetermined value Y, it is determined to be "graupel," and if the amplitude ratio is equal to or greater than the predetermined value Y, it is determined to be "hail," and "graupel," etc., are output as the analysis result of the meteorological phenomenon. In the example of FIG. 6 , the criteria control unit 222 outputs "graupel" as the analysis result because the amplitude ratio of observation data A was equal to or greater than the predetermined value X but less than the predetermined value Y. However, since "hail" actually fell, the criteria control unit 222 changes the predetermined value Y, which is the boundary threshold between "graupel" and "hail," to a predetermined value Y1. The predetermined value Y1 is a value smaller than the predetermined value Y, and the threshold change is performed to expand the area in the observation data where "hail" is determined.

[0069] In this way, the threshold value is changed to temporarily narrow the area determined to be "graupel" and widen the area determined to be "hail." The changed threshold value is then applied to the analysis process of observation data B, which was observed after observation data A.

[0070] The observation data B was observed at 15:42 on January 5, 2023, in City B (next to City A). In this case, the observation data B is the same as the observation data A, and the analysis result using the previous criteria would be determined to be "graupel." However, based on the true value (correct label) of the observation data A, the analysis unit 221 outputs the meteorological phenomenon "hail" as the analysis result of the observation data B using the new criteria. The alert control unit 223 generates and outputs alert information such as, for example, "Hail may be falling in City B. Please be careful," or "Hail may soon fall in City B. Please be careful." Because weather radar can observe precipitation particles in the sky, it may be possible to output alert information several minutes to several tens of minutes before they hit the ground.

[0071] Here, the change in the judgment criteria can be applied to each observation area. For example, analysis processing can be performed on observation data C in an observation area far away from the observation area of ​​observation data A without changing the judgment criteria. On the other hand, the system can be configured to apply the changed judgment criteria to analysis processing on observation data B that is close to the observation area of ​​observation data A and was observed within a predetermined time from the observation time of observation data A.

[0072] Therefore, the judgment criteria can be changed for each observation area and observation time of the observation data corresponding to the true value provided by the information management device 100. A valid area and an expiration date are set for the changed value of the judgment criteria. That is, the observation area to which the changed value applies is set to an area within a predetermined range from the observation area of ​​the observation data A corresponding to the true value (correct label), and the expiration date is set to a predetermined time (e.g., 30 minutes) from the observation time of the observation data A. When the expiration date has passed, the changed value is initialized, and thereafter, analysis processing is performed using the judgment criteria value before the change (default value).

[0073] FIG. 7 is a diagram showing a processing flow executed by the weather management device 200 of this embodiment.

[0074] The weather radar performs continuous observations in time series at predetermined intervals and outputs the observation data. When the weather management device 200 receives observation data A from the weather radar, it performs an analysis process for the observation data A (S201). At this time, the weather management device 200 provides the observation data A to be analyzed to the information management device 100.

[0075] The weather management device 200 performs the above-described precipitation particle estimation process. For example, it classifies the precipitation particles based on predetermined criteria and outputs an analysis result based on the classification. The analysis result is stored in the storage device 230 (S202). The weather management device 200 outputs alert information based on the analysis result (S203).

[0076] The weather management device 200 receives true value information of the meteorological phenomenon for the observation data A from the information management device 100. Upon receiving the true value information from the information management device 100, the weather management device 200 starts a determination criterion change process (S204). The weather management device 200 compares the analysis result of the observation data A with the received true value (correct label) information, and if it determines that the meteorological phenomenon is the same, it does not change the determination criterion (NO in S205). On the other hand, if it determines that the meteorological phenomenon is different (YES in S205), it changes the determination criterion based on the true value (correct label) identified from the weather-related posted information corresponding to the observation data A (S206).

[0077] The weather management device 200 uses the determination reference value changed in step S206 to repeatedly analyze the subsequent weather radar observation data (from NO in S208 to S201).

[0078] At this time, as described above, when the criteria control unit 222 changes the criteria based on a weather phenomenon identified from weather-related posted information corresponding to observation data A, it sets an effective area and effective time to which the changed criteria will be applied based on the observation area and observation time of the observation data A (S207). Then, in step S201, the analysis unit 221 determines whether the observation area and observation time of the subsequent observation data B are within the effective range corresponding to the set effective area and effective time, and if it is determined that they are within the effective range, it applies the changed criteria to the analysis process of the subsequent observation data B.

[0079] In this embodiment, the true value (correct label) provided by the information management device 100 is applied in real time to the analysis of observation data in the weather management device 200, and is reflected in the analysis of subsequent observation data. In other words, if the analysis result of the observation data and the true value (correct label) based on the weather-related posted information differ from each other, the judgment criteria can be changed based on the true value (correct label), thereby making it possible to obtain more accurate analysis results of subsequent observation data.

[0080] Hail, in particular, is a natural disaster that causes significant damage. Even if the analysis results of observation data indicate that hail is falling rather than hail, there may be cases where hail is actually falling (in the posted information). Therefore, in the subsequent analysis process of the observation data, even similar observation data that was determined to be hail is more strictly determined to be hail. In other words, the frequency of hail determinations is increased, and the number of missed determinations is reduced. This configuration makes it easier to issue alerts for meteorological phenomena that cause significant damage as natural disasters, allowing for appropriate disaster prevention awareness.

[0081] In addition, the area and time range to which the change in the criteria applies are set based on the observation area and observation time of the observation data that triggered the change in the criteria. As described above, hail is a unique meteorological phenomenon that is observed infrequently and does not affect a wide area over a long period of time. Therefore, the change in the criteria is applied to observation data within a radius of, for example, 50 km and within two hours of the observation area of ​​the observation data that triggered the change in the criteria, making it easier to issue alerts for meteorological phenomena that could cause serious damage as natural disasters. This configuration makes it possible to appropriately limit the area for which a hail alert is issued while avoiding unnecessarily expanding the area for which a hail alert is issued, thereby reducing confusion caused by the alert.

[0082] 8 to 12 are diagrams for explaining a third embodiment. This embodiment differs from the first embodiment in that it includes a function for classifying weather-related posted information based on insurance payment history. Note that components similar to those in the first embodiment are denoted by the same reference numerals in the drawings, and descriptions thereof will be omitted.

[0083] 8 is a diagram showing the functional blocks and network configuration of the information management device 100 of this embodiment. The control device 120 of the information management device 100 further includes a classification control unit 124, and the storage device 130 stores classification-related information 134.

[0084] 9 is a diagram illustrating a first function of classifying weather-related posted information based on insurance payment history and a second function of classifying weather-related posted information using a classification model according to this embodiment. The first function generates training data from past weather-related posted information and generates a classification model using the generated training data. The second function uses the generated classification model to perform a classification determination process on weather-related posted information newly extracted by the extraction unit 122 and associates the result of the classification determination process with a meteorological phenomenon corresponding to the observation data output from the data management unit 123, i.e., a true value (correct label).

[0085] The insurance payment history is the history of actual insurance payments made for damages caused by weather phenomena such as hail, sleet, tornadoes, whirlwinds, etc., under insurance products offered by insurance companies. Insurance products include personal insurance products that cover natural disasters, such as fire insurance, flood insurance, hail insurance, and vehicle insurance, as well as insurance products that cover natural disasters for businesses.

[0086] The history of insurance payments is the record of insurance payments made by the insurance company based on claims from policyholders, after the insurance company has grasped the damage situation, and includes information such as the payment items, location of the damage, date and time of the damage, and the person to whom the payment is made.

[0087] The payment item is a meteorological phenomenon related to the occurrence of damage, and for example, "hail damage" is included as a payment item. The location of damage is the location where the payment target was damaged by hail (location information where the damage occurred), and is address information such as XX town, XX city, XX prefecture. The payment target is the asset that suffered damage, and is the insured object. For example, in the case of vehicle insurance, it would be the vehicle, and in the case of fire insurance, it would be the house. This insurance payment history information is provided by the insurance company. The insurance payment history information can be obtained from the insurance company (the insurance company's server or terminal) via a network, or can be obtained as physical list information and input into the information management device 100 by the operations manager. The insurance payment history information is stored in the storage device 130 as classification-related information 134.

[0088] In this embodiment, the insurance payment history information is used to classify weather-related posted information into a group of weather phenomena with a high risk of damage (a group of weather phenomena with a history of insurance payment) and other groups (a group of weather phenomena with no history of insurance payment). That is, among multiple pieces of weather-related posted information, weather-related posted information corresponding to the location and date and time of damage for which insurance payment history has occurred is extracted, and information indicating that the weather phenomenon has a high risk of damage is associated with it. This makes it possible to distinguish between weather phenomena with a high risk of damage based on insurance payment history (performance), even for the same weather phenomenon. For example, weather-related posted information belonging to the weather phenomenon "hail" can be classified into weather-related posted information belonging to the weather phenomenon "hail (high risk of damage)" and other weather-related posted information.

[0089] Therefore, the classification control unit 124 associates (labels) the classification information "high damage risk" with weather-related posted information corresponding to the location and date and time of damage for which there was an insurance payment history. The data management unit 123 can then perform control so that, when generating a true value (correct label), the classification information associated with the weather-related posted information used to generate the true value (correct label) is included. For example, the true value information 133 of meteorological events by observation data stored in the storage device 130 can be configured to include the classification information "high damage risk" in addition to the observation time, observation area, and meteorological phenomenon based on the weather-related posted information.

[0090] As mentioned above, hail is ice particles with a diameter of 5 mm or more, but a diameter of 10 mm is also considered hail. However, the larger the diameter, the greater the damage, and the smaller the diameter, the less damage there may be. Therefore, even though hail is the same meteorological phenomenon, the damage (loss) caused by the meteorological phenomenon varies. Therefore, in this embodiment, insurance payment history information (actual damage record) is used as evidence of actual damage caused by the meteorological phenomenon "hail."

[0091] Then, the weather-related posted information corresponding to the past insurance payment history (the weather-related posted information used to generate the true value (correct label)) is accumulated as training data. The training data is stored in the storage unit 130 as classification-related information 134.

[0092] For example, images and posted text of weather-related posted information corresponding to past insurance payment history can be accumulated as training data. The images are used as training data for hail images that caused actual damage, and the posted text is used as training data for text (keywords, etc.) related to hail that caused actual damage. For example, posted text containing keywords related to hail itself or keywords related to damage (loss) caused by hail, such as "it was 30 mm in diameter," "about the size of a golf ball," "it dented the hood of the car," and "it broke through the roof of the garage," is used as training data.

[0093] The accumulated learning data is used in the generation process (learning process) of a classification model that determines whether newly extracted weather-related posted information corresponds to a weather phenomenon with a high risk of damage. The classification model calculates the image similarity and the similarity of the posted text based on keywords, etc. from the images and / or posted text of the newly extracted weather-related posted information, and classifies weather-related posted information with a similarity equal to or greater than a predetermined value into the "weather phenomenon with a high risk of damage" group.

[0094] 10 is a diagram showing the processing flow of the first function executed by the classification control unit 124. The classification control unit 124 performs the following processing.

[0095] 1) A first process (S1001, S1002) of classifying past weather-related posted information extracted by the extraction unit 122 into a weather phenomenon group with a high risk of damage based on insurance payment history information including the location of the damage related to the insurance payment, the date and time the damage occurred, and the weather phenomenon related to the occurrence of the damage; 2) a second process (S1003) of storing the weather-related posted information classified into the weather phenomenon group with a high risk of damage as learning data in a specified memory area; 3) a third process (S1004) of generating a classification model using the learning data to classify weather-related posted information extracted by the extraction unit 122 in the future into a weather phenomenon group with a high risk of damage.

[0096] 11 is a diagram showing a processing flow executed by the information management device 100, and corresponds to the processing flow of the first embodiment shown in FIG. 11. Steps S104A and S104B are added to FIG. 11, and these steps correspond to the second function of the classification control unit 124.

[0097] The classification control unit 124 performs the following processes on the weather-related posted information extracted in step S104: 4) A fourth process (S104A) of performing a classification determination process using the generated classification model to determine whether the weather-related posted information extracted by the extraction unit 122 belongs to a meteorological phenomenon group with a high risk of damage, and 5) A fifth process (S104B) of associating the result of the classification determination process with the meteorological phenomenon (true value (correct label)) corresponding to the observation data output from the data management unit 123.

[0098] Then, the classification determination result by the classification model (identification information indicating that the meteorological phenomenon has a high risk of damage) is added to the true value (correct label) output from the data management unit 123. In other words, the true value (correct label) output from the data management unit 123 is configured to include information on whether or not the meteorological phenomenon belongs to a group with a high risk of damage (S105).

[0099] Next, a case will be described in which the true value (correct label) including the identification information indicating that the weather phenomenon has a high risk of damage in this embodiment is applied to the second embodiment.

[0100] First, in the second embodiment, as shown in Fig. 6, the judgment criteria can be changed using the true value (correct label) provided by the information management device 100. Here, even if the true value (correct label) provided by the information management device 100 of this embodiment includes identification information indicating that the weather phenomenon is a "weather phenomenon with a high risk of damage," the weather management device 200 can change the judgment criteria regardless of whether the weather phenomenon is a "weather phenomenon with a high risk of damage."

[0101] On the other hand, the weather management device 200 can also control the range of change in the judgment criteria to be different for true values ​​(correct labels) that belong to "weather phenomena with a high risk of damage" and true values ​​(correct labels) that do not belong to "weather phenomena with a high risk of damage." For example, when changing the judgment criteria based on true values ​​(correct labels) that belong to "weather phenomena with a high risk of damage" (S206), the judgment criteria can be changed to a change value P1, and when changing the judgment criteria based on true values ​​(correct labels) that belong to "weather phenomena with a high risk of damage," the judgment criteria can be changed to a change value P2 that is smaller than the change value P1 (P1>P2).

[0102] 12, the alert control unit 223 of the weather management device 200 determines whether the true value (correct label) provided by the information management device 100 belongs to a "weather phenomenon with a high risk of damage" (S2004). If it is determined that the true value belongs to a "weather phenomenon with a high risk of damage" (YES in S2004), the alert control unit 223 can control the alert output to output alert information (e.g., a warning alert) indicating that the weather phenomenon has been reported to have caused actual damage in the past (S2005). On the other hand, if it is determined that the true value does not belong to a "weather phenomenon with a high risk of damage" (NO in S2004), the alert control unit 223 can control the alert output to output normal alert information (S2006).

[0103] Therefore, even with the same alert output process, an alert can be output for cases where there is a high risk of damage and actual damage has occurred in the past, and for example, it is possible to more strongly recommend evacuation indoors in the event of actual hail falling or hail that is likely to fall in the future.

[0104] Furthermore, the output destination of the alert information can be controlled to be different depending on whether the weather phenomenon belongs to a "weather phenomenon with a high risk of damage" or not. As described above, in the second embodiment, the alert information can be output to an information distribution system to which multiple users register, or to a web server that publishes information on a website. In this case, the alert control unit 223 determines that the weather phenomenon belongs to a "weather phenomenon with a high risk of damage," for example, as being urgent, and provides the alert information directly to the terminals of users registered in the information distribution system. If the weather phenomenon does not belong to a "weather phenomenon with a high risk of damage," the alert information can be output to the information distribution system or web server, and each user's terminal can receive the alert information via the information distribution system or web server. Note that even if the weather phenomenon belongs to a "weather phenomenon with a high risk of damage," the output destination may not be changed, and alert information belonging to a "weather phenomenon with a high risk of damage" may be provided to the information distribution system or web server, and then provided from the information distribution system or web server to each user's terminal.

[0105] In this way, the information management device 100 of this embodiment classifies weather-related posted information used to generate true values ​​(correct labels) into a high-risk weather phenomenon group and other groups based on past insurance payment history. In other words, multiple weather-related posted information (true values ​​(correct labels)) are classified by risk level based on past insurance payment history. For example, even for true values ​​(correct labels) of the same weather phenomenon, "hail," different risk levels can be distinguished. This makes it possible to provide feedback data by risk level for analysis based on weather radar observation data, further improving analysis accuracy.

[0106] Furthermore, the information management device 100 of this embodiment accumulates weather-related posted information classified as a "weather phenomenon with a high risk of damage" group based on past insurance payment history as training data, and uses this training data to generate a classification model for determining "weather phenomena with a high risk of damage." As a result, when applied to the second embodiment, the weather management device 200 can provide different alert outputs depending on the classification result of the "weather phenomenon with a high risk of damage" added to the true value (correct label), and can control the change value of the determination criteria to vary depending on the risk level. This allows the weather management device 200 to appropriately issue disaster prevention alerts in real time.

[0107] For example, if large, highly dangerous hail falls, it could scratch the hoods of cars and damage crops. If this could be known in advance, even at the last moment, it could be possible to reduce the damage by taking actions such as moving cars under cover or covering crops with sheets. The same is true for weather phenomena other than hail; if the occurrence of a sudden gust of wind could be predicted in advance, it could be possible to prevent damage by suspending train services, closing highways, and ceasing work at height. While normal alert output is an effective way to encourage actions to prevent damage, high-risk alert output, which has previously caused actual damage, can be used as a means to strongly recommend actions to prevent damage.

[0108] Furthermore, from the perspective of insurance contracts, encouraging actions to prevent damage before it occurs also has the following merits: Insurance companies that offer insurance products that cover damage can notify policyholders of alert information and have them take action to reduce damage, which allows policyholders to prevent damage before it occurs and insurance companies to reduce insurance payments, providing benefits to both parties.

[0109] The first to third embodiments have been described above. However, the information management device 100 can provide true values ​​(correct labels) of meteorological phenomena other than hail and sleet, such as tornadoes and whirlwinds. In this case, the weather-related posted information is posted information related to hail, sleet, tornadoes, or whirlwinds, and the information collection unit 121 can acquire weather-related posted information containing tornadoes and whirlwinds as keywords and / or similar images from the web server 300. The extraction unit 122 then extracts weather-related posted information that matches the observation area and observation time of the observation data of tornadoes or whirlwinds observed by weather radar. The data management unit 123 can output the meteorological phenomenon of tornadoes or whirlwinds identified from the extracted weather-related posted information as the meteorological phenomenon corresponding to the observation data.

[0110] One method for observing tornadoes and whirlwinds is meteorological Doppler radar observation. In addition to determining the location and strength of precipitation, the Doppler effect of radio waves reflected from precipitation particles carried by the wind can be used to measure the wind components approaching and receding from the weather radar. This is called the Doppler velocity. If the approaching velocity is above a predetermined value and the receding velocity is above a predetermined value, it can be determined that a wind vortex (a rotation within a developed cumulonimbus cloud that causes a tornado; a mesocyclone) is occurring. For example, if the Doppler velocity is above a predetermined value, it can be determined that a tornado is occurring.

[0111] In this way, the weather management device 200 can acquire Doppler velocity observation data from the meteorological Doppler radar and perform analysis processing to determine whether tornadoes or whirlwinds are occurring.The weather management device 200 can then provide the Doppler velocity observation data to the information management device 100.

[0112] Furthermore, the weather forecasting device 400 in the first embodiment can use an EHI (Energy-Helicity Index) weather forecast model (prediction algorithm) for tornadoes and whirlwinds. In this case, the information collecting unit 121, as the fourth processing unit 121D, can generate a weather event evaluation value for the candidate posted information based on forecast information generated by the tornado or whirlwind weather forecast model corresponding to the location information and time information of the candidate posted information. The information collecting unit 121 can then extract weather-related posted information from the candidate posted information in consideration of the weather event evaluation value (extract weather-related posted information from candidate posted information whose evaluation value for the occurrence prediction of a tornado or whirlwind is equal to or greater than a predetermined value). Known methods other than EHI can also be applied to the weather forecast model for tornadoes and whirlwinds.

[0113] The weather management device 200 of the second embodiment can output alert information and change the criteria for tornadoes and whirlwinds. The criteria change process can change the Doppler velocity threshold for determining the occurrence of a tornado based on the true value (correct label) provided by the information management device 100. The function of the classification control unit 124 of the third embodiment can also be configured to target tornadoes, whirlwinds, and gusts of wind. In this case, the insurance payment history information includes tornadoes, whirlwinds, and gusts of wind as meteorological phenomena related to the occurrence of damage.

[0114] In addition, each of the devices 100, 200 in the first, second and third embodiments may be configured to target any one of hail, sleet, tornadoes and whirlwinds, or may be configured to target all of hail, sleet, tornadoes and whirlwinds, or any combination thereof.

[0115] 13 to 17 are diagrams for explaining a fourth embodiment. This embodiment is provided with a mechanism for improving the analysis function (weather detection function) of the weather management device 200 of the second embodiment described above by using weather-related posted information (true value information) provided by the information management device 100 of the first embodiment described above. The same components as those of the above two embodiments are denoted by the same reference numerals in the respective drawings, and their description will be omitted.

[0116] 13 is a diagram showing the functional blocks and network configuration of the weather management device 200A of this embodiment. The configuration is basically the same as that of FIG. 5, but the analysis unit 221 is configured as a weather detection unit 221A equipped with a weather detection model 221a. The weather detection model 221a performs weather detection processing to determine the type of precipitation particles in clouds above the ground based on the judgment criteria, and outputs the type of precipitation particles that will fall to the ground based on the results of the type determination processing.

[0117] That is, the weather detection model 221a (analysis unit 221) performs a type discrimination process (type estimation process) of precipitation particles based on the amplitude ratio observed by the weather radar, and classifies the precipitation particles into each type, such as wet snow, dry snow, ice crystals, drizzle, rain, hail, sleet, and heavy rain, depending on the magnitude of the amplitude ratio. At this time, a determination criterion is set for each type of precipitation particle, and the weather detection model 221a determines to which type determination criterion the amplitude ratio observed by the weather radar belongs, and determines the type of one or more precipitation particles present in the clouds above. Then, the weather detection model 221a outputs the type of precipitation particles that will fall to the ground based on the result of the type discrimination process of precipitation particles in the clouds above.

[0118] The information output from the weather detection unit 221A can be configured to include the type of precipitation particles falling to the ground (e.g., hail) determined by the weather detection model 221a, and the location and time based on the observation area and observation time of the observation data. The location can also be configured to output the corresponding region (e.g., XX town, XX city, XX prefecture) based on the latitude and longitude of the observation area based on map information.

[0119] It is also possible to configure the system to output information indicating the reliability of the detected meteorological phenomenon along with the location and time. The reliability will be described later.

[0120] The determination criteria control unit 222 is configured as a learning unit 222A of the weather detection model 221a. The learning unit 222A generates the weather detection model 221a and uses meteorological phenomena identified from weather-related posted information as learning data (correct labels) to learn the types of precipitation particles that have actually fallen to the ground in relation to the types of precipitation particles in the sky based on the observation data, and performs a learning process (tuning process) to change the determination criteria.

[0121] Furthermore, the weather management device 200A of this embodiment can be configured to include a setting unit 221B for setting priorities for multiple types of precipitation particles. Multiple different types of precipitation particles exist in clouds in the sky, and for example, from observation data, it is possible to determine the distribution of hail areas, sleet areas, snow areas, and rain areas in the clouds in the sky. In other words, multiple different types of precipitation particles exist in the clouds in the sky in three dimensions (latitude, longitude, and altitude).

[0122] On the other hand, the types of precipitation particles that actually fall to the ground must be identified in two dimensions (latitude and longitude). In this embodiment, the setting unit 221B controls the setting unit 221B so that a priority can be set for each type of precipitation particle. The weather detection model 221a then determines the types of precipitation particles in the sky present within the observation area of ​​the observation data, and when outputting the types of precipitation particles that will fall to the ground, outputs the types of precipitation particles that will fall to the ground in descending order of the set priority. For example, assume that the priorities are set in the order of hail > sleet > snow > rain. If two precipitation particles, "hail" and "sleet," are detected, the weather detection model 221a can output "hail," which has the highest priority, as the type of precipitation particles that will fall to the ground.

[0123] Although the output control using the priority by the setting unit 221B has been described, the present invention is not limited to this. For example, the weather detection model 221a may be controlled to output only "hail," which is a natural disaster that causes significant damage. In other words, the weather detection model 221a may be configured to output the type of precipitation particles that will fall to the ground when a specific weather phenomenon is detected. Furthermore, the priority setting may be changed at any timing. For example, during the weather detection process, snow may be given a higher priority than hail, and then, partway through, the priority may be switched to outputting "snow" as a precipitation particle preferentially.

[0124] As such, the weather management device 200A of this embodiment comprises a communication device 210, a control device 220, and a storage device 230, and the control device 220 comprises a weather detection unit 221A (corresponding to the analysis unit 221) equipped with a weather detection model 221a, a setting unit 221B, a learning unit 222A (corresponding to the judgment criteria control unit 222), an alert control unit 223, and a true value information acquisition unit 224.

[0125] Next, the weather detection model 221a and the learning unit 222A will be described. Fig. 14 is a diagram showing the processing flow executed by the weather management device 200A. Fig. 14 corresponds to Fig. 7 of the second embodiment, and is a flowchart adapted to the processing of this embodiment. Steps S201a to S207a correspond to steps S201 to S207 in Fig. 7.

[0126] The weather detection model 221a is a functional unit that sequentially receives observation data observed by a weather radar, and uses the observation data as input to determine and output the current weather phenomenon (type of precipitation particles falling to the ground) (S201a). The weather detection unit 221A stores the output weather detection results in the storage device 230 (S202a), and the alert control unit 223 performs alert output processing (203a). Note that in the example of FIG. 14, the setting unit 221B has previously set the priority (S202A).

[0127] The learning unit 222A performs a learning process using the learning data to generate the weather detection model 221a (S201A). The learning data is observation data that sets the weather-related posted information to a correct answer level, and is information that combines past observation data obtained by weather radar with the weather-related posted information corresponding to the observation data. In the example of FIG. 13, this corresponds to observation data 231 and true value information 233, and is stored in the storage device 230.

[0128] The learning unit 222A performs a first learning process (step S201A) to generate a weather detection model 221a using accumulated past learning data, and a second learning process to tune (adjust) the weather detection model 221a generated by the first learning process. The second learning process is a process of tuning (adjusting) the judgment criteria for classifying the observation data into multiple precipitation particle types, updating the weather detection model 221a, and generating individual weather detection models 221a with different judgment criteria through tuning, as will be described later.

[0129] The second learning process is divided into real-time processing (step S206a) that is executed in real time when weather-related posted information (true value information) is input, and batch processing (S210a) that suspends the second learning process even when weather-related posted information is input and executes the second learning process at a predetermined timing. The tuning methods include a first tuning method that executes only real-time processing, a second tuning method that uses both real-time processing and batch processing, and a third tuning method that executes only batch processing.

[0130] In the second tuning method, a second learning process is performed when weather-related posted information (true value information) is input during weather detection processing by the weather detection model 221a, and the judgment criteria are tuned in real time (S206a). Then, at any time when weather detection processing is not being performed, for example, during a maintenance period, the judgment criteria are retuned (re-learning process) using multiple pieces of observation data and weather-related posted information accumulated since the previous maintenance until the current maintenance, and the weather detection model 221a is updated (S210a). Figure 14 shows the processing flow of this second tuning method.

[0131] Although the first tuning method does not perform the second learning process by batch processing, like the second tuning method, the second learning process is performed during the weather detection process to perform tuning, so if the weather phenomenon identified from the observation data differs from the weather phenomenon identified from the weather-related posted information, the weather-related posted information can be used as the correct label and the judgment criteria can be changed in real time.As a result, the weather detection process can be performed on subsequent observation data using the tuned judgment criteria.

[0132] On the other hand, in the third tuning method, the weather detection model 221a is not updated in real time. Therefore, if the weather phenomenon identified from the observation data differs from the weather phenomenon identified from the weather-related posted information, the output result of the weather detection unit 221A for the subsequent observation data (the type of precipitation particles falling to the ground) does not reflect the weather-related posted information corresponding to the previous observation data as the correct label. Therefore, when the third tuning method is applied, as a temporary measure until the second learning process is performed, a replacement process is performed in which the type of precipitation particles output from the weather detection model 221a is replaced with the weather phenomenon (correct label) identified from the weather-related posted information corresponding to the previous observation data.

[0133] FIG. 15 is a diagram showing a processing flow according to the third tuning method. The weather detection unit 221A determines whether the type of precipitation particles output from the weather detection model 221a based on the observation data differs from the weather phenomenon identified from the weather-related posted information (S205a). If the weather detection unit 221A determines that the weather phenomena differ (YES in S205a), it temporarily stores the feature values ​​of the observation data and the weather phenomenon identified from the weather-related posted information (S2010) and turns on the replacement processing flag (S2011). Note that the feature values ​​are information representing the weather conditions identified from the observation data, such as the distribution range of each precipitation particle and the altitude at which the precipitation particles exist, within the clouds above. Note that the feature values ​​extracted from the observation data can be calculated using known methods and can be represented, for example, by a vector of parameters indicating the weather conditions.

[0134] Then, when the weather detection process is performed on the subsequent observation data and the type is determined, step S2012 checks whether the replacement process flag is ON. If the replacement process flag is ON, the process proceeds to step S2013, where the similarity between the feature values ​​of the subsequent observation data and the feature values ​​of the retained previous observation data is determined. The similarity determination is performed, for example, by determining whether the feature values ​​of the subsequent observation data are within a certain distance from the feature values ​​(weather state vector) of the retained previous observation data. If the feature values ​​(weather state vector) of the subsequent observation data are within a certain distance from the feature values ​​of the previous observation data, the subsequent observation data can be determined to be similar. This vector distance can be determined using known methods such as Euclidean distance, Manhattan distance, and Mahalanobis distance.

[0135] Then, in step S2014, if it is determined that the subsequent observation data is similar to the previous observation data, the weather detection unit 221A compares the correct label based on the previous weather-related posted information with the result of the weather detection process for the subsequent observation data. If the result of the weather detection process for the subsequent observation data differs from the correct label based on the previous weather-related posted information, the weather detection unit 221A outputs the type of precipitation particles, assuming that the correct label based on the previous weather-related posted information is correct. On the other hand, if it is determined that the subsequent observation data is not similar to the previous observation data, the weather detection unit 221A outputs the result of the weather detection process for the subsequent observation data.

[0136] By configuring in this manner, it is possible to output weather detection processing results that reflect the correct label based on the previous weather-related posted information in real time, even if the judgment criteria (weather detection model 221a) are not changed in real time.

[0137] Next, as explained in the second embodiment above, in accordance with the aspect in which different judgment criteria are held, in this embodiment as well, the tuned weather detection model 221a can be generated as a separate weather detection model from the pre-tuning weather detection model 221a. Figure 16 is a diagram showing a processing flow in which, in real-time processing of the second learning process, multiple weather detection models with different judgment criteria are generated, and a corresponding weather detection model is selected for input observation data to perform weather detection processing.

[0138] 16 , if the weather phenomenon identified from the observation data differs from the weather phenomenon identified from the weather-related posted information during the tuning process (YES in S205a), the learning unit 222A performs a learning process based on the currently used weather detection model 221a, using the weather-related posted information as the correct answer label, and individually generates a weather detection model 221a (corresponding to a first weather detection model) with modified judgment criteria (S2061a).The learning unit 222A then sets a valid area and a valid time for the individually generated weather detection model 221a based on the observation area and observation time of the observation data (S2071a).

[0139] Then, when subsequent observation data is input, the weather detection unit 221A selects a weather detection model 221a that matches the observation area and observation time of the subsequent observation data from multiple weather detection models 221a with different effective areas and effective times (S2001a), and controls the weather detection processing for the subsequent observation data to be performed using the selected weather detection model 221a (S201a).

[0140] In addition, in step S2001a, if there is no weather detection model 221a that matches the observation area and observation time of the subsequent observation data, the weather detection model 221A can be configured to select the weather detection model 221a (reference model) generated in step 201A or re-learned in step S210a and perform weather detection processing.

[0141] This configuration allows multiple weather detection models to be switched between for use. In other words, the weather detection model 221a generated in step S201A or re-learned in step S210a can be configured as a reference model for long-term use. Each weather detection model 221a individually generated by real-time processing in the second learning process can be applied as a model for short-term use, with at least one of different determination criteria, validity area, and expiration date. The weather detection unit 221A can then switch from a reference model for long-term use to a model for short-term use, or switch between multiple models for short-term use, depending on the observation data, thereby improving the accuracy of the weather detection process. When the validity period (expiration date) of the applied weather detection model 221a expires, the weather detection unit 221A can be controlled to switch to the reference model and perform weather detection processing.

[0142] Here, the reliability output from the weather detection unit 221A will be described. The weather detection model 221a uses weather-related posted information used as learning data. Therefore, the reliability of the weather detection model 221a can be calculated based on each evaluation value related to the weather-related posted information in the first embodiment. The reliability of the weather detection model 221a can then be configured to be output along with the results of the weather detection process. For example, the weather detection unit 221A can output information including the weather phenomenon "hail," the area "XX City, XX Prefecture," and a reliability of "80%."

[0143] Specifically, as described in the first embodiment, weather-related posted information can be evaluated using a poster evaluation value, a posting accuracy evaluation value, and a weather event evaluation value. Therefore, the information management device 100 provides the weather management device 200A with true value information and the evaluation values ​​of the weather-related posted information on which the true value information is based. The reliability of the weather detection model 221a can then be calculated using the sum of the evaluation values ​​of the weather-related posted information used as learning data, or any of the evaluation values. When performing learning processing using multiple different weather-related posted information, the average or median of the evaluation values ​​of the weather-related posted information can be calculated and used as the reliability of the weather detection model 221a. The reliability calculation method is arbitrary.

[0144] On the other hand, it is also possible to generate a highly reliable weather detection model 221a. That is, the evaluation value of weather-related posted information can be used as an index for selecting weather-related posted information to be used as training data. For example, weather-related posted information accumulated over a certain period of time can be sorted in order of reliability, and training can be performed in descending order of reliability. The training process can then be terminated when a desired result is obtained. For example, weather detection processing can be performed using other observation data that outputs a type different from the correct label as sample test data, verifying whether the output weather phenomenon type is the same as the correct label, and terminating the training when the verification result reaches a predetermined value or greater. Alternatively, the training process can be configured to use training data with an evaluation value greater than a certain value. This configuration can suppress training processing using low-reliability training data as weather-related posted information, thereby preventing overtraining of the weather detection model 221a.

[0145] The learning method of the weather detection model 221a, i.e., the method of tuning the judgment criteria by the learning unit 222A, can be performed by any one of the following means or a combination of two or more of the following means. In the following description, the type of precipitation particles obtained from the weather detection model 221a through the type discrimination process of the observation data is referred to as "A," and the type of precipitation particles (weather phenomenon) identified from the weather-related posted information is referred to as "C."

[0146] (A) From the past learning data learned up to now, observation data that falls within a certain range of similarity to the weather phenomena (weather conditions) extracted from the input observation data is extracted. Then, the correct answer label associated with the extracted observation data is replaced from "A" to "C." In other words, the correct answer label of the observation data that has a certain degree of similarity as learning data is replaced with "C," the type of precipitation particles identified from the weather-related posted information. Then, using the group of observation data whose correct answer label has been replaced with "C," as learning data, a re-learning process is performed to update the weather detection model 221a.

[0147] The similarity determination can be performed by applying the similarity determination process described above, which compares the features of the observed data. The threshold for determining whether the similarity is within a certain range can be dynamically changed. For example, if the output result of the updated weather detection model 221a is still precipitation particle type "A," the threshold for determining whether the similarity is within a certain range can be updated (increased), the distribution range to be used as training data can be increased, and the training process can be performed again to update the weather detection model.

[0148] (B) In the case of weather detection processing using a probability density function or the like, the probability of each precipitation particle type (classification) can be calculated based on the input observation data, and the type with the highest probability can be output. In this case, the learning unit 222A can increase the probability that the input observation data will be classified as type "C" or decrease the probability that the input observation data will be classified as type "A." The increase and decrease amounts of the probability may use preset fixed values, or the values ​​of the increase and decrease amounts may be dynamically changed.

[0149] When increasing or decreasing the probability, an offset may be uniformly added or subtracted, or the probability density function itself may be modified. When modifying the probability density function, the mean and variance of the probability density function may be modified, and the amount of modification may be preset or dynamically varied. For example, to increase the probability, the mean value of the probability density function of type "C" is modified so that it approaches the input value of the probability density function extracted from the input observation data. Conversely, to decrease the probability, the mean value of the probability density function of type "A" is modified so that it moves away from the input value of the probability density function extracted from the input observation data. The result based on this updated probability can be used as the final output. If the result based on the updated probability is still type "A," the same operation may be repeated until the desired result is obtained.

[0150] (C) As in (B) above, a probability density function or the like can be used, and a probability density function or the like can be provided for each feature (high-dimensional vector feature) of a meteorological phenomenon extracted from the input observation data. In this case, the probabilities of each feature can be multiplied to calculate the final probability for each type. At this time, the probability of each feature can be increased uniformly, or the amount of increase in probability can be dynamically assigned depending on the results based on the probability. For example, a probability density function can be provided for feature 1, feature 2, and feature 3, and the probability of each feature can be calculated for type "B." If the probability calculated from feature 1 is 0.99, the probability calculated from feature 2 is 0.91, and the probability calculated from feature 3 is 0.95, the probability of feature 2 is low, so the probability of feature 2 is significantly increased. The same method as in (B) above can be applied to increase or decrease the probability.

[0151] (D) When using the probability density function or the like described in (B) above, if multiple weather-related posts with high evaluation values ​​(reliability) are obtained, the probability density function may be updated using these multiple correct labels. For example, a Bayesian update method can be used, in which the probability density function is used as a prior distribution and a posterior distribution is calculated from reliable results. In this case, the probability density function for type "C" can be updated using this posterior distribution as the probability density function, and the result based on this updated probability can be used as the final output. If the result based on the updated probability still outputs type "A," the same operation can be repeated until the desired result is obtained.

[0152] In the above description, the weather management device 200 is linked to the separately configured information management device 100 as an example, but the present invention is not limited to this. For example, as shown in Fig. 17, it can be configured as a weather management device or weather management system that includes each functional unit and each piece of information of the information management device 100. In this case, it can be configured as a weather management system having the following configuration.

[0153] (1) The weather management system (weather management device 200A) has an information management unit that acquires weather-related posted information from multiple posted information items posted to a web server that accepts posted information from poster terminals and extracts the weather-related posted information that matches the observation area and observation time of observation data observed by weather radar; a weather detection unit that performs a type discrimination process on precipitation particles in clouds in the sky based on the observation data and is equipped with a weather detection model that outputs the type of precipitation particles that fall to the ground based on the results of the type discrimination process; and a learning unit that uses meteorological phenomena identified from the weather-related posted information as learning data, learns the type of precipitation particles that actually fall to the ground for the type of precipitation particles in the sky based on the observation data, and generates a weather detection model.

[0154] The type determination process of the weather management system described in (1) above can be configured as a process of classifying aerial weather phenomena based on observation data into corresponding types based on criteria set for each type of precipitation particle. The learning unit performs a learning process using weather phenomena identified from weather-related posted information corresponding to the observation data as correct labels, and tunes the criteria. Furthermore, the weather management system described in (1) above can have the functions of a weather detection unit 221A and a learning unit 222A, including the weather detection model 221a of this embodiment.

[0155] Furthermore, the replacement process (steps S201a to S208) described in FIG. 15 can be configured as a weather management device that reflects true value information through the replacement process without changing the judgment criteria, and can be configured as follows.

[0156] (3) A weather management device includes an analysis unit that classifies weather radar observation data based on predetermined criteria and outputs analysis results based on the classification; and an information acquisition unit that acquires weather-related posted information corresponding to the observation data output from the information management device described in claim 1, wherein the analysis unit performs a type discrimination process for precipitation particles in clouds in the observation data based on the criteria set for each type of precipitation particle, and outputs the type of precipitation particles falling to the ground based on the result of the type discrimination process, and the analysis unit compares the weather phenomenon identified from the weather-related posted information corresponding to the observation data with the output type of precipitation particles falling to the ground, and if it is determined that the weather phenomena are different, it retains the observation data and the weather-related posted information that are determined to be different weather phenomena, The analysis unit determines the similarity between the subsequent observation data and the retained observation data, and if it determines that there is similarity between the subsequent observation data and the retained observation data, it compares the weather phenomenon understood from the retained weather-related posted information with the type of precipitation particles falling to the ground output for the subsequent observation data, and if the weather phenomenon obtained from the subsequent observation data is different from the weather phenomenon understood from the retained weather-related posted information, it outputs the weather phenomenon understood from the retained weather-related posted information as the weather detection result for the subsequent observation data.

[0157] Furthermore, similar to the example shown in FIG. 17, the weather management device or weather management system having each functional unit and each piece of information of the information management device 100 can be configured to have the following components.

[0158] (4) The weather management system includes an information management unit that acquires weather-related posted information from multiple posted information items posted to a web server that accepts posted information from poster terminals and extracts the weather-related posted information that matches the observation area and observation time of observation data observed by weather radar; and a weather detection unit that performs weather detection processing that performs type discrimination processing on precipitation particles in clouds above based on the observation data and outputs the type of precipitation particles that will fall to the ground based on the results of the type discrimination processing. The weather detection unit then compares the weather phenomenon identified from the weather-related posted information corresponding to the observation data with the type of precipitation particles falling to the ground output by the weather detection process, and if it is determined that the weather phenomena are different, it retains the observation data and the weather-related posted information that are determined to be different weather phenomena.Furthermore, the weather detection unit determines the similarity between the subsequent observation data and the retained observation data, and if it is determined that there is similarity between the subsequent observation data and the retained observation data, it compares the weather phenomenon identified from the retained weather-related posted information with the type of precipitation particles falling to the ground output by the weather detection process using the subsequent observation data as input.If the weather phenomenon obtained from the subsequent observation data is different from the weather phenomenon identified from the retained weather-related posted information, it outputs the weather phenomenon identified from the retained weather-related posted information as the weather detection result for the subsequent observation data.

[0159] In this way, the weather management device or weather management system described above in (3) and (4) can have a system configuration aspect that does not assume a learning process or the like that reflects true value information based on weather-related posted information in the weather detection process itself. For example, an existing system that performs weather detection processing can be equipped with a replacement processing function that, while working in conjunction with an information management device 100 that manages weather-related posted information, replaces the type of precipitation particles output in the weather detection process with a weather phenomenon (correct label) identified from the weather-related posted information corresponding to the previous observation data. This configuration can improve the accuracy of weather detection without changing the weather detection process itself.

[0160] The alert output process of the alert control unit 223 may be configured to obtain a large amount of highly reliable weather-related posted information regarding meteorological phenomena. As described in the second embodiment, the alert control unit 223 can generate and output alert information such as, "There is a possibility of hail in City A. Please be careful." At this time, a message such as, "What kind of weather phenomena are those in the vicinity of City A experiencing? Please post." is inserted into the alert information. This configuration allows the information management device 100 to collect a large amount of weather-related posted information, and the weather management device 200A can obtain a large amount of learning data. Furthermore, when providing alert information to users (user terminals) registered in the weather management device 200, push notifications can be sent to the user terminals, allowing a large amount of weather-related posted information to be collected. In this way, by introducing a mechanism for encouraging posting in conjunction with the alert output process, posts regarding highly reliable (certain) weather phenomena can be obtained, improving the accuracy of the weather detection model 221a.

[0161] Furthermore, similar to the first to third embodiments, this embodiment can also perform weather detection processing that can detect weather phenomena such as tornadoes and whirlwinds.

[0162] As described above, each function constituting the above-mentioned information management device 100 and weather management device 200 can be realized by a program, and a computer program prepared in advance to realize each function is stored in an auxiliary memory device, and a control unit such as a CPU reads the program stored in the auxiliary memory device into a main memory device, and the control unit executes the program read into the main memory device, thereby operating the functions of each unit.

[0163] The program can also be provided to a computer in a state recorded on a computer-readable recording medium. Examples of computer-readable recording media include optical disks such as CD-ROMs, phase-change optical disks such as DVD-ROMs, magneto-optical disks such as MO (Magneto Optical) and MD (Mini Disk), magnetic disks such as floppy disks and removable hard disks, and memory cards such as CompactFlash (registered trademark), SmartMedia, SD memory cards, and memory sticks. Hardware devices such as integrated circuits (e.g., IC chips) specially designed and configured for the purposes of the present invention are also included as recording media.

[0164] Although the embodiments of the present invention have been described, these embodiments are presented as examples and are not intended to limit the scope of the invention. This novel embodiment can be embodied in various other forms, and various omissions, substitutions, and modifications can be made without departing from the spirit of the invention. These embodiments and their modifications are included within the scope and spirit of the invention, and are also included in the inventions and their equivalents as set forth in the claims.

[0165] 100 Information management device 110 Communication device 120 Control device 121 Information collection unit 121A First processing unit 121B Second processing unit 121C Third processing unit 121D Fourth processing unit 122 Extraction unit 123 Data management unit 124 Classification control unit 130 Storage device 131 Posted information 132 Poster evaluation information 133 True value information 134 Classification related information 200, 200A Weather management device 210 Communication device 220 Control device 221 Analysis unit 221A Weather detection unit 221a Weather detection model 221B Setting unit 222 Judgment criterion control unit 222B Learning unit 223 Alert control unit 224 True value information acquisition unit 230 Storage device 231 Observation data 232 Judgment criterion information 233 True value information 234 Setting information 300 Web server 400 Weather forecasting device T Terminal (contributor terminal)

Claims

1. An information management device comprising: an information collection unit that acquires weather-related posted information from multiple posted information items posted to a web server that accepts posted information from poster terminals; an extraction unit that extracts the weather-related posted information items that match the observation area and observation time of observation data observed by weather radar; and a data management unit that outputs meteorological phenomena identified from the extracted weather-related posted information items as meteorological phenomena corresponding to the observation data.

2. The posted information includes posted articles and / or posted images, location information, time information, and poster information, and the information collection unit performs a first process of extracting candidate posted information from multiple posted information posted to the web server based on similarities between weather-related keywords or weather-related images, and a second process of generating a poster evaluation value for the poster who posted the candidate posted information based on a predetermined evaluation criterion, and the information collection unit extracts the weather-related posted information from the candidate posted information of posters whose poster evaluation value is equal to or greater than a predetermined value. This is an information management device as described in claim 1.

3. The posted information includes posted articles and / or posted images, location information, time information, and poster information, and the information collection unit performs a first process of extracting candidate posted information from a plurality of posted information posted to the web server based on similarity of weather-related keywords or weather-related images, and a third process of extracting posted information from the plurality of posted information posted to the web server as similar posted information that includes location information and time information within a predetermined range close to the location information and time information of the candidate posted information and has similarity of the weather-related keywords or weather-related images, and generating a posting accuracy evaluation value for the candidate posted information based on the extracted similar posted information, characterized in that the information collection unit extracts the weather-related posted information from the candidate posted information whose posting accuracy evaluation value is equal to or greater than a predetermined value.

4. The posted information includes posted articles and / or posted images, location information, time information, and poster information, and the information collection unit performs a first process of extracting candidate posted information from a plurality of posted information posted to the web server based on similarities between weather-related keywords or weather-related images, and a fourth process of generating a weather event evaluation value for the candidate posted information based on forecast information generated by a predetermined weather forecast model corresponding to the location information and time information of the candidate posted information, and the information collection unit extracts the weather-related posted information from the candidate posted information whose weather event evaluation value is equal to or greater than a predetermined value.

5. The posted information includes posted articles and / or posted images, location information, time information, and poster information, and the information collection unit performs the following processes: a first process of extracting candidate posted information from the plurality of posted information posted to the web server based on similarity of weather-related keywords or weather-related images; a second process of generating a poster evaluation value for the poster of the candidate posted information based on predetermined evaluation criteria; a third process of extracting posted information from the plurality of posted information posted to the web server as similar posted information, the posted information including location information and time information within a predetermined range close to the location information and time information of the candidate posted information and having similarity of the weather-related keywords or weather-related images, and generating a posting accuracy evaluation value for the candidate posted information based on the extracted similar posted information; and a fourth process of generating a weather event evaluation value for the candidate posted information based on forecast information generated by a predetermined weather forecast model corresponding to the location information and time information of the candidate posted information. The information management device described in claim 1, characterized in that the information collection unit extracts the weather-related posted information from the candidate posted information whose evaluation is equal to or greater than a predetermined value based on the poster evaluation value, the posting accuracy evaluation value, and the weather event evaluation value.

6. The information management device according to claim 1, further comprising a classification control unit, wherein the classification control unit performs the following steps: a first process of classifying the weather-related posted information extracted by the extraction unit into a high-risk weather phenomenon group based on insurance payment history information including the location of the damage related to the insurance payment, the date and time the damage occurred, and the weather phenomenon related to the occurrence of the damage; a second process of storing the weather-related posted information classified into the high-risk weather phenomenon group as learning data in a predetermined memory area; a third process of using the learning data to generate a classification model for classifying the weather-related posted information extracted by the extraction unit into the high-risk weather phenomenon group; a fourth process of using the generated classification model to perform a classification determination process to determine whether the weather-related posted information extracted by the extraction unit belongs to the high-risk weather phenomenon group; and a fifth process of associating the result of the classification determination process with the weather phenomenon corresponding to the observation data output from the data management unit.

7. An information management device as described in any one of claims 1 to 6, characterized in that the weather-related posted information is posted information related to hail, sleet, tornadoes or whirlwinds, the extraction unit extracts the weather-related posted information that matches the observation area and observation time of observation data of hail, sleet, tornadoes or whirlwinds observed by weather radar, and the data management unit outputs the meteorological phenomenon of hail, sleet, tornadoes or whirlwinds identified from the extracted weather-related posted information as the meteorological phenomenon corresponding to the observation data.

8. An information management system comprising: an information collection unit that acquires weather-related posted information from multiple posted information items posted to a web server that accepts posted information from poster terminals; an extraction unit that extracts the weather-related posted information that matches the observation area and observation time of observation data observed by weather radar; and a data management unit that outputs meteorological phenomena identified from the extracted weather-related posted information as meteorological phenomena corresponding to the observation data.

9. A program executed by a computer, causing the computer to realize the following functions: a first function of acquiring weather-related posted information from multiple posted information posted to a web server that accepts posted information from poster terminals; a second function of extracting the weather-related posted information that matches the observation area and observation time of observation data observed by weather radar; and a third function of outputting meteorological phenomena identified from the extracted weather-related posted information as meteorological phenomena corresponding to the observation data.

10. A weather management device comprising: an analysis unit that classifies weather radar observation data based on predetermined criteria and outputs analysis results based on the classification; an alert control unit that outputs alert information based on the analysis results; and a criteria control unit that changes the criteria based on meteorological phenomena identified from the weather-related posted information corresponding to the observation data output from the information management device described in claim 1.

11. The weather management device described in claim 10, characterized in that the judgment criteria control unit compares the weather phenomenon understood from the weather-related posted information corresponding to the observation data output from the information management device described in claim 1 with the analysis results of the observation data by the analysis unit, and if it is determined that the weather phenomenon is the same, does not perform processing to change the judgment criteria, but if it is determined that the weather phenomenon is different, performs processing to change the judgment criteria based on the weather phenomenon understood from the weather-related posted information corresponding to the observation data.

12. The weather management device described in claim 10 or 11, characterized in that when the judgment criteria control unit changes the judgment criteria based on a meteorological phenomenon identified from the weather-related posted information corresponding to the observation data, it sets an effective area and effective time to which the changed judgment criteria will be applied based on the observation area and observation time of the observation data, and the analysis unit determines whether the observation area and observation time of the subsequent observation data are within the effective range corresponding to the effective area and the effective time, and if it is determined that they are within the effective range, it applies the changed judgment criteria to the analysis processing of the subsequent observation data.

13. The weather management device described in claim 10, characterized in that the analysis unit is a weather detection unit equipped with a weather detection model that performs a type discrimination process for precipitation particles in clouds in the sky in the observation data based on the judgment criteria and outputs the type of precipitation particles that will fall to the ground based on the results of the type discrimination process, the judgment criteria control unit functions as a learning unit for the weather detection model, and the learning unit uses meteorological phenomena identified from the weather-related posted information as learning data, learns the type of precipitation particles that actually fell to the ground in relation to the type of precipitation particles in the sky based on the observation data, and performs a learning process to change the judgment criteria.

14. The weather management device of claim 13 further has a setting unit that sets priorities for multiple types of precipitation particles, and the weather detection model determines the type of each precipitation particle in the air that exists within the observation area of ​​the observation data, and outputs the type of precipitation particles that will fall to the ground in order of highest priority.

15. The weather management device described in claim 13 or 14, characterized in that the type determination process of the analysis unit is a process of classifying meteorological phenomena in the sky based on the observation data into the corresponding type based on the determination criteria set for each type of precipitation particle, and the learning unit compares the meteorological phenomenon identified from the weather-related posted information corresponding to the observation data with the type of precipitation particles falling to the ground output by the weather detection model, and if it is determined that the meteorological phenomenon is different, performs the learning process based on the meteorological phenomenon identified from the weather-related posted information.

16. The weather management device described in claim 13 or 14, characterized in that the type discrimination process of the analysis unit is a process of classifying meteorological phenomena in the sky based on the observation data into the corresponding type based on the discrimination criteria set for each type of precipitation particle, the learning unit performs an update process of the weather detection model to change the discrimination criteria based on meteorological phenomena identified from the weather-related posted information corresponding to the observation data, and sets an effective area and effective time to which the updated weather discrimination model is applied based on the observation area and observation time of the observation data, and the weather detection unit determines whether the observation area and observation time of subsequent observation data are within the effective range corresponding to the effective area and the effective time, and if it is determined that they are within the effective range, controls to process the subsequent observation data using the updated weather detection model.

17. The weather management device described in claim 13 or 14, characterized in that the type discrimination process of the analysis unit is a process of classifying meteorological phenomena in the sky based on the observation data into the corresponding type based on the judgment criteria set for each type of precipitation particle, and the learning unit, when the weather-related post information corresponding to the observation data is input, generates a first weather detection model in which the judgment criteria have been changed based on the meteorological phenomenon understood from the weather-related post information corresponding to the observation data, separately from the weather detection model before the weather-related post information was input, and the learning unit sets an effective area and effective time for the separately generated first weather detection model based on the observation area and observation time of the observation data, and the weather detection unit selects the first weather detection model that matches the observation area and observation time of subsequent observation data from multiple first weather detection models with different effective areas and effective times, and controls to process the subsequent observation data using the selected first weather detection model.

18. The type determination process of the analysis unit is a process of classifying meteorological phenomena in the sky based on the observation data into the corresponding type based on the determination criteria set for each type of precipitation particles, and the weather detection unit compares the meteorological phenomenon identified from the weather-related posted information corresponding to the observation data with the type of precipitation particles falling to the ground output by the weather detection model, and if it is determined that the meteorological phenomena are different, temporarily stores the observation data and the weather-related posted information determined to be different meteorological phenomena, The weather detection unit determines the similarity between the subsequent observation data and the temporarily stored observation data until a learning process is performed using the weather phenomena identified from the temporarily stored weather-related posted information as learning data, and if it is determined that there is similarity between the subsequent observation data and the temporarily stored observation data, compares the weather phenomenon identified from the temporarily stored weather-related posted information with the type of precipitation particles falling to the ground output by the weather detection model using the subsequent observation data as input, and if the weather phenomenon obtained from the subsequent observation data is different from the weather phenomenon identified from the temporarily stored weather-related posted information, outputs the weather phenomenon identified from the temporarily stored weather-related posted information as the weather detection result for the subsequent observation data.

19. A weather management device comprising: an analysis unit that classifies weather radar observation data based on predetermined criteria and outputs analysis results based on the classification; and an information acquisition unit that acquires weather-related posted information corresponding to the observation data output from the information management device described in claim 1, wherein the analysis unit performs a type discrimination process for precipitation particles in clouds in the sky in the observation data based on the criteria set for each type of precipitation particle, and outputs the type of precipitation particles falling to the ground based on the result of the type discrimination process, and the analysis unit compares the weather phenomenon identified from the weather-related posted information corresponding to the observation data with the output type of precipitation particles falling to the ground, and if it is determined that the weather phenomena are different, it retains the observation data and the weather-related posted information that are determined to be different weather phenomena, The analysis unit determines the similarity between the subsequent observation data and the retained observation data, and if it is determined that there is similarity between the subsequent observation data and the retained observation data, compares the weather phenomenon identified from the retained weather-related posted information with the type of precipitation particles falling to the ground output for the subsequent observation data, and if the weather phenomenon obtained from the subsequent observation data differs from the weather phenomenon identified from the retained weather-related posted information, outputs the weather phenomenon identified from the retained weather-related posted information as the weather detection result for the subsequent observation data.