Title: Using machine learning to downscale coarse-resolution environmental variables for understanding the spatial frequency of convective storms Show Chinese title

Title: 使用机器学习来降尺度粗分辨率环境变量以理解对流风暴的空间频率

Abstract: Global climate models (GCMs), typically run at ~100-km resolution, capture large-scale environmental conditions but cannot resolve convection and cloud processes at kilometer scales. Convection-permitting models offer higher-resolution simulations that explicitly simulate convection but are computationally expensive and impractical for large ensemble runs. This study explores machine learning (ML) as a bridge between these approaches. We train simple, pixel-based neural networks to predict convective storm frequency from environmental variables produced by a regional convection-permitting model. The ML models achieve promising results, with structural similarity index measure (SSIM) values exceeding 0.8, capturing the diurnal cycle and orographic convection without explicit temporal or spatial coordinates as input. Model performance declines when fewer input features are used or specific regions are excluded, underscoring the role of diverse physical mechanisms in convective activity. These findings highlight ML potential as a computationally efficient tool for representing convection and as a means of scientific discovery, offering insights into convective processes. Unlike convolutional neural networks, which depend on spatial structure and grid size, the pixel-based model treats each grid point independently, enabling value-to-value prediction without spatial context. This design enhances adaptability to resolution changes and supports generalization to unseen environmental regimes, making it particularly suited for linking environmental conditions to convective features and for application across diverse model grids or climate scenarios. Show Chinese abstract

Abstract: 全球气候模型（GCMs），通常以约100公里的分辨率运行，可以捕捉大尺度环境条件，但无法在公里尺度上解析对流和云过程。 允许对流的模型提供更高分辨率的模拟，可以显式模拟对流，但计算成本高，对于大规模集合运行不切实际。 本研究探索了机器学习（ML）作为这两种方法之间的桥梁。 我们训练简单的基于像素的神经网络，从区域允许对流模型产生的环境变量预测对流风暴频率。 ML模型取得了有前景的结果，结构相似性指数测量（SSIM）值超过0.8，能够在没有显式时间或空间坐标作为输入的情况下捕捉昼夜循环和地形对流。 当使用的输入特征较少或排除特定区域时，模型性能会下降，这突显了多种物理机制在对流活动中的作用。 这些发现突显了ML作为计算效率高的工具在表示对流方面的潜力，以及作为科学发现手段的作用，提供了对对流过程的见解。 与依赖空间结构和网格大小的卷积神经网络不同，基于像素的模型独立处理每个网格点，实现了无需空间上下文的价值到价值的预测。 这种设计增强了对分辨率变化的适应性，并支持对未见过的环境条件进行泛化，使其特别适合于将环境条件与对流特征联系起来，并适用于各种模型网格或气候情景的应用。