Scholarlink

<p>Effective identification and forecasting of Convective Initiation (CI) are crucial for the early warning of severe convective weather. Artificial intelligence (AI) has significant potential in CI forecasting and warning. This study presents a dataset, named CIDS, designed for AI models to identify and forecast CI. The dataset covers southeastern China and samples intense convective weather events. These events include short-duration heavy precipitation, thunderstorm winds, and hail occurring from March to September between 2018 and 2023. It provides feature data and labels at 10-minute intervals. Feature data includes 10 radar mosaic products (spatial resolution: 0.01° × 0.01°) and FY-4A satellite radiance (or reflectance) across nine spectral bands (the spatial resolution of the visible channel is 0.005° × 0.005°, the shortwave infrared and mid-wave infrared channels are 0.02° × 0.02°, and the other channels are 0.04° × 0.04°). An algorithm based on radar composite reflectance factors identifies CI, determines if a storm cell is incipient, and marks its location. By analyzing the spatial extent and intensity evolution of CI over the next 30 minutes, the algorithm assigns CI category labels. The CIDS comprises 136,728 samples, identifying 4,159,491 CIs, including 1,789,208 developing CIs, which refer to CIs with enhancements of area and average echo intensity over the next thirty minutes.</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

A dataset for machine learning model to convective initiation detection and nowcasting over southeastern China

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年2月28日

查看全文
  • Yujia Liu,
  • Anyuan Xiong,
  • Na Liu,
  • Yunying Li,
  • Zitong Chen

摘要

Effective identification and forecasting of Convective Initiation (CI) are crucial for the early warning of severe convective weather. Artificial intelligence (AI) has significant potential in CI forecasting and warning. This study presents a dataset, named CIDS, designed for AI models to identify and forecast CI. The dataset covers southeastern China and samples intense convective weather events. These events include short-duration heavy precipitation, thunderstorm winds, and hail occurring from March to September between 2018 and 2023. It provides feature data and labels at 10-minute intervals. Feature data includes 10 radar mosaic products (spatial resolution: 0.01° × 0.01°) and FY-4A satellite radiance (or reflectance) across nine spectral bands (the spatial resolution of the visible channel is 0.005° × 0.005°, the shortwave infrared and mid-wave infrared channels are 0.02° × 0.02°, and the other channels are 0.04° × 0.04°). An algorithm based on radar composite reflectance factors identifies CI, determines if a storm cell is incipient, and marks its location. By analyzing the spatial extent and intensity evolution of CI over the next 30 minutes, the algorithm assigns CI category labels. The CIDS comprises 136,728 samples, identifying 4,159,491 CIs, including 1,789,208 developing CIs, which refer to CIs with enhancements of area and average echo intensity over the next thirty minutes.