How to cache the lookup tables

The spectral (non-gray) radiation methods need the RRTMGP gas/cloud/aerosol lookup tables, which are parsed from NetCDF artifacts. Caching the tables avoids re-reading the NetCDF files within a session and allows running without NCDatasets in constrained environments.

Reuse within a session

lookup_tables returns a typed LookupBundle; build it once and pass it to every solver (and to every solve call):

using RRTMGP, NCDatasets

lookups = RRTMGP.lookup_tables(grid_params, method)
solver_a = RRTMGP.RRTMGPSolver(grid_params, method, params, bcs_lw, bcs_sw, as_a; lookups)
out = RRTMGP.solve(profile; lookups)

Cache across sessions

save_lookup_tables serializes a bundle to disk; load_lookup_tables restores it onto the device described by your grid parameters:

# once, in an environment with NCDatasets:
using RRTMGP, NCDatasets
lookups = RRTMGP.lookup_tables(grid_params, RRTMGP.ClearSkyRadiation(false))
RRTMGP.save_lookup_tables("clearsky_lookups.jls", lookups)

# later, e.g., in a classroom environment with only RRTMGP:
using RRTMGP
lookups = RRTMGP.load_lookup_tables("clearsky_lookups.jls", grid_params)
out = RRTMGP.solve(profile; lookups)

A cache, not an interchange format

The file uses Julia's Serialization stdlib and is tied to the Julia version and package layout that wrote it. Regenerate it when environments change; the NetCDF artifacts remain the source of truth.