Explicit core

The Explicit core uses the POM separation between the depth-averaged external mode and the three-dimensional internal mode. External free-surface and depth-averaged velocity updates subcycle within each internal step. See Mode splitting and Time stepping for the formulation.

Implementation

The production class is POM2K_Core in src/neuralpom/cores/explicit/core.py. Its modules separate grid construction, initialization, momentum, tracers, baroclinic operators, diagnostics, checkpointing and the integration loop. core_factory.resolve_core_class("explicit") returns this class.

Training configuration

The Explicit templates select eager physical execution with the PyTorch backend. configs/examples/explicit_1h.json and explicit_1day.json pair the core with hourly and daily endpoint corrections. Reference data retain the Explicit generation settings; Data format describes their storage.