Previously, I was updating the modified prognostic variables using the
conservation principals. Because the size of the time-step implicitly
effects this updated state, this led to problems when applying the the
neural network with a shorter time step than it was trained with.
The energy and moisture conservation laws can constrain the tendencies
directly.
Previously, I was updating the modified prognostic variables using the conservation principals. Because the size of the time-step implicitly effects this updated state, this led to problems when applying the the neural network with a shorter time step than it was trained with.
The energy and moisture conservation laws can constrain the tendencies directly.
Resolves #10