Information processing apparatus, radar apparatus, method, and storage medium

A method executed by processing circuitry comprising at least one of (i) one or more hardware processors executing instructions stored in a computer-readable memory, or (ii) dedicated hardware, the method comprising:

• acquiring (i) first weather data in a first time zone including a first radar rainfall amount value observed in the first time zone at a first observation point by a first weather radar, the first radar rainfall amount value calculated based on a reflected wave of a radio wave transmitted from the first weather radar, the first weather data including, in addition to the first radar rainfall amount value, at least one observation point data item acquired from the first weather radar, the at least one observation point data item including at least one of reception power, a reflectivity factor, a Doppler velocity, a Doppler spectrum width, a differential reflectivity, a correlation coefficient between polarizations, a differential phase, a specific differential phase, a flag value indicating a calculation method of the first radar rainfall amount value, or an integrated value of a rainfall attenuation amount at the first observation point, (ii) second weather data in the first time zone including a second radar rainfall amount value observed in the first time zone at a second observation point by a second weather radar different from the first weather radar, the second radar rainfall amount value calculated based on a reflected wave of a radio wave transmitted from the second weather radar, the second weather data including, in addition to the second radar rainfall amount value, at least one observation point data item acquired from the second weather radar, the at least one observation point data item acquired from the second weather radar including at least one of reception power, a reflectivity factor, a Doppler velocity, a Doppler spectrum width, a differential reflectivity, a correlation coefficient between polarizations, a differential phase, a specific differential phase, a flag value indicating a calculation method of the second radar rainfall amount value, or an integrated value of a rainfall attenuation amount at the second observation point, and (ii) ground rainfall amount data in the first time zone including a ground rainfall amount value measured in the first time zone in a ground rainfall meter;
• deriving, based on the first and second weather data in the first time zone, a plurality of first feature values related to accuracy of the first and second radar rainfall amount values included in the first and second weather data in the first time zone, the plurality of first feature values including at least the first and second radar rainfall amount values, a distance from the first and second weather radars to the first and second observation points, a flag value indicating a calculation method of the first and second radar rainfall amount values, an integrated value of the rainfall attenuation amounts, and a difference between the first and second radar rainfall amount values;
• generating first learning data sets, each based on the ground rainfall amount value included in the ground rainfall amount data in the first time zone and a corresponding combination of the plurality of first feature values;
• generating, based on the generated first learning data sets, a plurality of first learned models while varying combinations of the first feature values, the plurality of first learned models being trained to output a parameter for estimating a rainfall amount value corresponding to the ground rainfall amount value included in the ground rainfall amount data in the first time zone when the corresponding combination of first feature values is input;
• evaluating the generated first learned models using a part of the generated first learning data sets or evaluation data corresponding thereto by calculating an index relating to a correlation between an estimated rainfall amount value obtained using each of the first learned models and the ground rainfall amount value, and selecting a third feature value set comprising all or a subset of the first feature values, based on the calculated indices;
• generating a second learned model based on second learning data generated based on the ground rainfall amount value included in the ground rainfall amount data in the first time zone and the selected third feature value set;
• acquiring (i) first weather data in a second time zone which is after the first time zone, the first weather data in the second time zone including a first radar rainfall amount value observed in the second time zone at each of a plurality of first observation points by the first weather radar, and (ii) second weather data in the second time zone including a second radar rainfall amount value observed in the second time zone at each of a plurality of second observation points by the second weather radar;
• converting the first and second weather data in the second time zone at each of the plurality of first and second observation points into data corresponding to a coordinate system defined for composite rainfall amount data;
• deriving, based on the converted first and second weather data in the second time zone, a plurality of second feature values related to accuracy of the first and second radar rainfall amount values included in the first and second weather data in the second time zone, the plurality of second feature values corresponding to the first feature values constituting the selected third feature value set;
• inputting the plurality of second feature values into the second learned model to obtain a parameter output from the second learned model;
• and
• composing, based on the parameter output by the second learned model, the first and second radar rainfall amount values included in the first and second weather data in the second time zone to generate the composite rainfall amount data, the parameter controlling contributions of the first and second radar rainfall amounts in the second time zone in the generated composite rainfall amount data on a per-coordinate-point basis with respect to each coordinate point in the coordinate system defined for the composite rainfall amount data.

A non-transitory computer-readable storage medium having stored thereon a program which is executable by a computer of an information processing apparatus, the program comprising instructions capable of causing the computer to execute functions of:

• acquiring (i) first weather data in a first time zone including a first radar rainfall amount value observed in the first time zone at a first observation point by a first weather radar, the first radar rainfall amount value calculated based on a reflected wave of a radio wave transmitted from the first weather radar, the first weather data including, in addition to the first radar rainfall amount value, at least one observation point data item acquired from the first weather radar, the at least one observation point data item including at least one of reception power, a reflectivity factor, a Doppler velocity, a Doppler spectrum width, a differential reflectivity, a correlation coefficient between polarizations, a differential phase, a specific differential phase, a flag value indicating a calculation method of the first radar rainfall amount value, or an integrated value of a rainfall attenuation amount at the first observation point, (ii) second weather data in the first time zone including a second radar rainfall amount value observed in the first time zone at a second observation point by a second weather radar different from the first weather radar, the second radar rainfall amount value calculated based on a reflected wave of a radio wave transmitted from the second weather radar, the second weather data including, in addition to the second radar rainfall amount value, at least one observation point data item acquired from the second weather radar, the at least one observation point data item acquired from the second weather radar including at least one of reception power, a reflectivity factor, a Doppler velocity, a Doppler spectrum width, a differential reflectivity, a correlation coefficient between polarizations, a differential phase, a specific differential phase, a flag value indicating a calculation method of the second radar rainfall amount value, or an integrated value of a rainfall attenuation amount at the second observation point, and (iii) ground rainfall amount data in the first time zone including a ground rainfall amount value measured in the first time zone in a ground rainfall meter;
• deriving, based on the first and second weather data in the first time zone, a plurality of first feature values related to accuracy of the first and second radar rainfall amount values included in the first and second weather data in the first time zone, the plurality of first feature values including at least the first and second radar rainfall amount values, a distance from the first and second weather radars to the first and second observation points, a flag value indicating a calculation method of the first and second radar rainfall amount values, an integrated value of the rainfall attenuation amounts, and a difference between the first and second radar rainfall amount values;
• generating first learning data sets, each based on the ground rainfall amount value included in the ground rainfall amount data in the first time zone and a corresponding combination of the plurality of first feature values;
• generating, based on the generated first learning data sets, a plurality of first learned models while varying combinations of the first feature values, the plurality of first learned models being trained to output a parameter for estimating a rainfall amount value corresponding to the ground rainfall amount value included in the ground rainfall amount data in the first time zone when the corresponding combination of first feature values is input;
• evaluating the generated first learned models using a part of the generated first learning data sets or evaluation data corresponding thereto by calculating an index relating to a correlation between an estimated rainfall amount value obtained using each of the first learned models and the ground rainfall amount value, and selecting a third feature value set comprising all or a subset of the first feature values, based on the calculated indices;
• generating a second learned model based on second learning data generated based on the ground rainfall amount value included in the ground rainfall amount data in the first time zone and the selected third feature value set;
• acquiring (ii) first weather data in a second time zone which is after the first time zone, the first weather data in the second time zone including a first radar rainfall amount value observed in the second time zone at each of a plurality of first observation points by the first weather radar, and (ii) second weather data in the second time zone including a second radar rainfall amount value observed in the second time zone at each of a plurality of second observation points by the second weather radar;
• converting the first and second weather data in the second time zone at each of the plurality of first and second observation points into data corresponding to a coordinate system defined for composite rainfall amount data;
• deriving, based on the converted first and second weather data in the second time zone, a plurality of second feature values related to accuracy of the first and second radar rainfall amount values included in the first and second weather data in the second time zone, the plurality of second feature values corresponding to the first feature values constituting the selected third feature value set;
• inputting the plurality of second feature values into the second learned model to obtain a parameter output from the second learned model;
• composing, based on the parameter output by the second learned model, the first and second radar rainfall amount values included in the first and second weather data in the second time zone to generate the composite rainfall amount data, the parameter controlling contributions of the first and second radar rainfall amounts in the second time zone in the generated composite rainfall amount data on a per-coordinate-point basis with respect to each coordinate point in the coordinate system defined for the composite rainfall amount data.