Does the MJO survive in stock CAM5? — HyCAM vs CAM5 control (AMIP, 1990-2005)

Both runs are FAMIPC5 (AMIP: observed transient SST/GHG/aerosol), f19_g16, 1990-2005, identical history tapes; analysis 1991-2005 (boreal-winter Nov-Apr where noted). The only difference is the physics: HyCAM= CAM5 with the ftorch NN super+cloud emulator, CAM5= stock CAM5 (no NN), versus Obs= ERA-Interim / TRMM / NOAA.

Bottom line: stock CAM5 has a weak, non-propagating MJO — convection stalls over the Indian Ocean and its spectral power leaks into faster Kelvin waves. HyCAM restores a coherent eastward-propagating MJO with a 30-80 day, wavenumber-1 spectral peak, much closer to observations.

1. Wheeler-Kiladis normalized spectrum — OLR/FLUT (symmetric, 15S-15N)

👀 What to look for: The MJO is enhanced power at eastward wavenumber 1-3, period 30-80 days (just above the ‘30 days’ line, right of centre), distinct from the faster Kelvin band. HyCAM shows a concentrated low-frequency eastward peak; CAM5’s power leaks into the Kelvin band & westward — the dedicated MJO peak is weak.

HyCAM (NN)
CAM5 control (stock)
NOAA OLR (obs)

2. Lag-correlation propagation — boreal winter (Indian-Ocean precip base)

👀 What to look for: Eastward tilt of the precip correlation (top panel: lower-left→upper-right) = eastward MJO propagation across the Maritime Continent. HyCAM tilts eastward like obs; CAM5’s signal is near-vertical / stationary, locked near 60-90°E and fails to cross the Maritime Continent.

HyCAM (NN)
CAM5 control (stock)
ERA-Interim (obs)

3. Composite life cycle — Nov-Apr (8 MJO phases; OLR anom + 850-hPa wind)

👀 What to look for: A coherent eastward-marching convective dipole from the Indian Ocean (P1-3) into the Maritime Continent / W Pacific (P4-7). CAM5’s phases are weaker & less organized / less propagating.

HyCAM (NN)
CAM5 control (stock)
ERA-Interim (obs)

4. CLIVAR wavenumber-frequency — precip (boreal winter, 15S-15N)

👀 What to look for: A distinct eastward (positive) spectral peak at wavenumber 1, ~30-80 days is the MJO. HyCAM concentrates power there; CAM5 is flatter / more spread.

HyCAM (NN)
CAM5 control (stock)
TRMM (obs)

Energy / water conservation (context)

Both AMIP runs track the same observed seasonal water cycle (shared SST). HyCAM’s total energy drifts up slightly (NN energy source); CAM5 sits lower & flatter.

→ TE/TW figure (HyCAM vs CAM5 AMIP + F2000 refs)

Full per-model diagnostics: HyCAM overview  |  CAM5 overview  |  HyCAM variability  |  CAM5 variability. Generated by make_mjo_compare.py.