Precipitation field space downscaling method

By combining convolutional neural network and denoising diffusion probability model, the spatial descaling method of precipitation field is solved, and the limitations of traditional methods in capturing the fine structure and local changes of heavy precipitation are achieved, accurate reconstruction and noise removal of high-resolution precipitation fields are improved, and the accuracy and reliability of precipitation forecasts are improved.

CN119720071AActive Publication Date: 2025-03-28CHENGDU PLATEAU METEOROLOGICAL INST OF CHINA METEOROLOGICAL ADMINISTRATION
7 Cites 6 Cited by

Patent Information

Application Number
CN202411550232.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-11-01
Publication Date
2025-03-28
Estimated Expiration
2044-11-01

AI Technical Summary

Technical Problem

The traditional downscale method has significant limitations in capturing the fine structure and local changes of heavy precipitation, and it is difficult to accurately reflect the microdistribution characteristics and short-term change trends of heavy precipitation. In the application of precipitation spatial downscale, there are problems such as optimization instability and noise amplification.

Method used

A precipitation field spatial descaling method combining convolutional neural networks and denoising diffusion probability model is adopted. By obtaining high-resolution topographic information data and precipitation pairing data, normalizing the process, and inputting the pre-trained model to generate a high-resolution precipitation field. This model uses a frequency domain mixed loss function and multi-scale feature extraction fusion network to capture global information and high-frequency details of the precipitation field, and effectively remove noise through the denoising process.

Benefits of technology

Accurate reconstruction of high-resolution precipitation fields is achieved, the accuracy and reliability of precipitation field data distribution is improved, and the requirements for high-resolution precipitation spatial distribution are effectively met, and more refined precipitation forecasting and disaster prevention are supported.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

The invention discloses a precipitation field space downscaling method, and relates to the technical field of meteorological data processing, and the method comprises the steps: obtaining high-resolution topographic information data and precipitation pairing data composed of high-resolution and low-resolution precipitation data, and carrying out the normalization processing of the high-resolution topographic information data and the precipitation pairing data; inputting the normalized precipitation pairing data and the high-resolution terrain information data into a pre-trained precipitation field space downscaling model, and outputting to obtain a high-resolution precipitation field; wherein the precipitation field space downscaling model comprises a convolutional neural network and a de-noising diffusion probability model, and the de-noising diffusion probability model is constructed based on residual calculation, forward noise addition, a multi-scale feature extraction fusion network and reverse de-noising. According to the method, the high-resolution precipitation field can be reconstructed from the low-resolution precipitation field.
Need to check novelty before this filing date? Find Prior Art