Deep learning and dynamic downscaling fused small watershed future scene climate rapid prediction method

By integrating deep learning and dynamic downscaling methods, a standardized full-process technical system adapted to high-altitude and cold mountainous areas was constructed. This system solved the problems of accuracy, physical consistency, and computational efficiency in climate prediction for small watersheds in high-altitude glaciers, and enabled the efficient generation of high-resolution climate data, supporting geological and hydrological disaster risk assessment and engineering applications in high-altitude and cold mountainous areas.

CN122087740AActive Publication Date: 2026-05-26INST OF MOUNTAIN HAZARDS & ENVIRONMENT CHINESE ACADEMY OF SCI
5 Cites 0 Cited by

Patent Information

Application Number
CN202610552760.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-04-24
Publication Date
2026-05-26
Estimated Expiration
2046-04-24

AI Technical Summary

Technical Problem

Existing climate downscaling techniques cannot simultaneously balance prediction accuracy, physical consistency, and computational efficiency, and cannot meet the high-resolution climate prediction needs of small watersheds in high-altitude glaciers. In particular, they lack reliable and refined meteorological data support for geological and hydrological disaster risk assessment and engineering applications in high-altitude and cold mountainous areas.

Method used

By employing a method that integrates deep learning and dynamic downscaling, high-resolution labeled data is generated using the WRF-CPM convection-allowable scale regional climate model. A TA-UNet network model is constructed, and a terrain attention mechanism and a physical constraint loss function are introduced to train a downscaling surrogate model. Combined with bias correction, high-resolution future climate prediction data is generated.

Benefits of technology

It enables the rapid and stable generation of high-resolution future climate predictions for small watersheds in high-altitude glaciers, improves prediction accuracy and physical consistency, reduces computational costs and time, provides reliable meteorological data support, and is suitable for geological and hydrological disaster risk assessment and engineering construction in high-altitude and cold mountainous areas.

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

Abstract

The invention relates to the technical field of climate prediction and meteorological data downscaling, and provides a small watershed future scene climate rapid prediction method fusing deep learning and dynamic downscaling, which comprises the following steps: step 1, delimiting a target area and preprocessing input data; 2, performing dynamic downscaling simulation and labeling and generating reference data; step 3, constructing a training sample and training a TA-UNet downscaling proxy model; and 4, downscaling the future climate and generating a prediction data set. The method solves the core contradiction that prediction precision, physical consistency and calculation efficiency cannot be considered in the prior art, provides a high-mountain small-watershed future scene climate prediction scheme, constructs a standardized whole-process technical system adaptive to a high-cold mountain area scene, realizes rapid and stable generation of high-resolution future climate prediction data of the high-mountain glacier small watershed, and improves the prediction efficiency of the high-mountain glacier small-watershed future climate. And a reliable refined meteorological data support is provided for geological and hydrological disaster risk assessment and major engineering construction in the cold and cold mountainous area.
Need to check novelty before this filing date? Find Prior Art