Fig. 7: Network structure diagram. | npj Climate and Atmospheric Science

Fig. 7: Network structure diagram.

From: Toward skillful forecasting of super El Niño events using a diffusion-based westerly wind burst parameterization

Fig. 7

a, b, and c, respectively, exhibit the detailed modules of Down-/Up-Sampling block, ConvNeXt block, and Swin-Transformer block used in DDPM-based WWBs parameterization, where \(k,s,p\) and \(c=64\) represent the kernel size, stride, padding, and channel. Pachify represents an operation that cuts the gridded data into small patches. Inverse Pachify is the restoration process. \({MLP}\) represents a multi-layer perceptron. \({Window}-{MSA}\) represents the multi-head sliding-window self-attention layer.

Back to article page