jgfs_atmos_anal ecf (v17) Analysis - TerrenceMcGuinness-NOAA/global-workflow GitHub Wiki
jgfs_atmos_anal.ecf (GFS v17)
Branch / tenant: dev/gfs.v17 (gw_v17)
Path: ecf/scripts/gfs/analysis/atmos/jgfs_atmos_anal.ecf
Analysis date: 2026-06-11
Method: On-disk v17 worktree analysis.
Summary
jgfs_atmos_anal.ecf is the NCO production ecFlow task that runs the GSI atmospheric analysis — the core data assimilation step that produces the atmospheric initial conditions for the GFS forecast. It assimilates conventional observations, satellite radiances, GPS-RO, and other data to generate the 3D atmospheric analysis. It calls JGLOBAL_ATMOS_ANALYSIS.
1. The script itself
| Section | Detail |
|---|---|
| PBS resources | 100 nodes, 16 mpiprocs, 8 OMP threads, 128 cpus/node, exclusive host, 80-min wall |
| Modules | load_modules.sh gsi (GSI analysis stack) |
| J-Job | JGLOBAL_ATMOS_ANALYSIS |
| Rocoto breadcrumb | anal.sh |
2. Execution chain
jgfs_atmos_anal.ecf
+-- JGLOBAL_ATMOS_ANALYSIS
+-- exglobal_atmos_analysis.sh (or Python)
+-- GSI variational solver (gsi.x)
+-- Outer/inner loop minimization
+-- Produces atmospheric analysis increments
3. v17-specific notes
- Second largest job: 100 nodes x 128 cpus = 12,800 ranks. Only the forecast is larger.
- GSI-based: uses
gsimodules (not JEDI for atmos DA in v17 GFS). - Hybrid threading: 16 MPI x 8 OMP per node — optimal for GSI's parallelism.
- 80-min walltime — GSI outer/inner loop minimization is computationally intensive.
- Critical path: atmospheric analysis must complete before the forecast can run.
- Exclusive host placement for all 100 nodes.
4. Change-impact snapshot
| Symbol | Risk | Notes |
|---|---|---|
JGLOBAL_ATMOS_ANALYSIS |
CRITICAL | Produces atmospheric ICs — forecast depends on it |
Sources
- On-disk v17 worktree:
ecf/scripts/gfs/analysis/atmos/jgfs_atmos_anal.ecf.