jgfs_marine_anlvar ecf (v17) Analysis - TerrenceMcGuinness-NOAA/global-workflow GitHub Wiki
jgfs_marine_anlvar.ecf (GFS v17)
Branch / tenant: dev/gfs.v17 (gw_v17)
Path: ecf/scripts/gfs/analysis/marine/jgfs_marine_analvar.ecf
Analysis date: 2026-06-11
Summary
jgfs_marine_anlvar.ecf runs the marine variational analysis solver — the core SOCA 3DVar/4DEnVar minimization that assimilates ocean/ice observations to produce the ocean and sea-ice analysis. This is the computational heart of the marine DA system. It calls JGLOBAL_MARINE_ANALYSIS_VARIATIONAL.
1. Resources & Configuration
| Field | Value |
|---|---|
| PBS resources | 8 nodes, 64 cpus/node = 512 MPI ranks, 500 GB/node, exclusive host, 30-min wall |
| Modules | load_modules.sh ufsda (JEDI/SOCA) |
| J-Job | JGLOBAL_MARINE_ANALYSIS_VARIATIONAL |
| Rocoto | marineanlvar.sh |
2. Execution chain
jgfs_marine_analvar.ecf → JGLOBAL_MARINE_ANALYSIS_VARIATIONAL
→ exglobal_marine_analysis_variational.py
→ SOCA variational solver (3DVar/4DEnVar)
→ Assimilates SST, SLA, T/S profiles, sea ice, etc.
→ Produces ocean/ice analysis increments
3. v17-specific notes
- Heavy resources: 8 nodes x 64 cpus = 512 ranks, 500 GB/node. One of the most resource-intensive DA tasks.
- Exclusive host — needs full node memory for ocean state.
- Part of marine DA chain:
anl_init→bmat_init→bmat→anlvar→anlfinal→anlchkpt. - CRITICAL: produces the ocean/ice analysis that feeds the coupled forecast.
- Python-based SOCA variational solver via
pygfs.
Sources
- On-disk v17:
ecf/scripts/gfs/analysis/marine/jgfs_marine_analvar.ecf.