Scholarlink

Abstract <p>This study explores the application of the AI-based approach for spatial downscaling of surface wind fields over the Barents and Kara Seas using deep artificial neural networks with skip connections. This method aims to improve spatial resolution while significantly reducing computational costs compared to nonhydrostatic modeling. Low-resolution input data are derived from the ERA5 global reanalysis, while high-resolution reference data are provided by Weather Research &amp; Forecasting (WRF) modeling. The results of AI-based downscaling are compared with the baseline bilinear interpolation. The proposed model enhances the mesoscale atmospheric structure compared to ERA5. It also achieves a 50x increase in computational efficiency compared to WRF modeling.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

AI-Based Spatial Downscaling of Surface Wind Fields over the Barents and Kara Seas

  • 期刊论文

  • 发表日期：

    2025年12月1日

查看全文
  • V. Yu. Rezvov,
  • M. A. Krinitskiy,
  • A. V. Gavrikov

摘要

Abstract

This study explores the application of the AI-based approach for spatial downscaling of surface wind fields over the Barents and Kara Seas using deep artificial neural networks with skip connections. This method aims to improve spatial resolution while significantly reducing computational costs compared to nonhydrostatic modeling. Low-resolution input data are derived from the ERA5 global reanalysis, while high-resolution reference data are provided by Weather Research & Forecasting (WRF) modeling. The results of AI-based downscaling are compared with the baseline bilinear interpolation. The proposed model enhances the mesoscale atmospheric structure compared to ERA5. It also achieves a 50x increase in computational efficiency compared to WRF modeling.