Weather Radar Rainfall Composition Using Learned Feature Models

The patent applies parameter changes by introducing multiple feature values (10 or more) that characterize different radar systems and their operating conditions. These feature values include radar type, observation mode, and other system-specific parameters. By changing and standardizing these parameters across multiple radars, the system can accurately compose rainfall amount data from heterogeneous sources while maintaining high precision, thus resolving the contradiction between expanded observation range and maintained accuracy.

The patent introduces an intermediary learning model that is trained using ground rainfall amount data as reference. This learning model acts as a mediator between raw radar observations and composite rainfall amount data, automatically learning the optimal composition method. The model takes feature values from multiple radars and produces accurate composite data without requiring complex manual processing, thus resolving the contradiction between processing complexity and accuracy.

The patent applies preliminary action by pre-training a learning model using ground rainfall amount data before actual composition operations. During this preliminary training phase, the model learns the relationships between multiple feature values and accurate rainfall measurements. Once trained, the model can rapidly process new data without requiring complex real-time computations, thus achieving high reliability while maintaining efficient processing.