jweyn / DLWP-CS

Deep learning models for global weather prediction on a cubed sphere
GNU General Public License v3.0
90 stars 40 forks source link

CNN input layer dimension #4

Closed Priya-Sharma07 closed 3 years ago

Priya-Sharma07 commented 3 years ago

Hello @jweyn could you please tell me why you considered exactly 48 points on each side of the cube faces, which corresponds to roughly 1.9-degree grid spacing in latitude and longitude in the center of the equatorial faces. what 48 indicates here?. it will be very helpful to understand ask1 CNN architecture.

jweyn commented 3 years ago

Hi @miniproject2018, sorry for the delay. The choice was fairly arbitrary, it seemed like a good balance of atmospheric resolution and data processing requirements to start there. Most recently we have been working with 64-point cubed sphere faces, about 1.4 degrees.

Priya-Sharma07 commented 3 years ago

okay, thank u.