search

GISWLH / CAS-Canglong

SST prediction model
MIT License
2 stars 0 forks source link

问题 #1

Open hahne0922 opened 6 days ago

hahne0922 commented 6 days ago
  1. 这套架构适不适用于陆地S2S每日的天气预报？
  2. 所选海洋和大气变量在网络中是否有分别考虑？
GISWLH commented 6 days ago

  1. 陆地S2S不确定性很大，一些关键变量的预报受海温的影响（如降水），因为当前海温效果不错，也许当前的结构可以直接用来S2S其它变量预报。根据Fuxi的工作，S2S需要模型结构中考虑不确定性，当前的模型没有。 Land S2S with large uncertainty, key features in land variables such as precipitation are influenced by SST. For the good skill in SST in current model, this current model may be straight for S2S forecast.

  2. 这个是很好的想法，可以把输入分类来分别嵌入，只需要单独调用patch3D/4D This is a great idea for we can separate the input into land and ocean, then we go to the patch3D module independently