Test Case: C48_S2SW-gfs_ocean_prod_f006.yaml
Configuration: C48_S2SW (Coupled Sea-to-Sea-to-Wave)
Job: oceanice_products.sh with COMPONENT=ocean
Duration: Single forecast hour (f006)
Status: ✅ FIXED - Split from combined test, products/ paths corrected
Last Updated: October 1, 2025

This test validates the ocean products generation pipeline for the coupled S2SW system, converting native MOM6 ocean model output into distribution-ready GRIB2 products at multiple resolutions plus native NetCDF subsets. This test specifically validates the ocean component of the oceanice_products job.

Total Files:

• Input: 2 files (ocean restart + history)
• Output: 7 files (6 GRIB2 + 1 netCDF)

Rocoto XML shows TWO independent parallel tasks:

<metatask name="gfs_ocean_prod">
  <var name="fhr_list">6 12 18 24 30 36 42 48</var>
  <envar><name>COMPONENT</name><value>ocean</value></envar>
  <task name="gfs_ocean_prod_f#fhr_list#">
    <command>&JOBS_DIR;/oceanice_products.sh</command>
  </task>
</metatask>

<metatask name="gfs_ice_prod">
  <var name="fhr_list">6 12 18 24 30 36 42 48</var>
  <envar><name>COMPONENT</name><value>ice</value></envar>
  <task name="gfs_ice_prod_f#fhr_list#">
    <command>&JOBS_DIR;/oceanice_products.sh</command>
  </task>
</metatask>

Key Insight: Ocean and ice run as separate parallel tasks, not a combined job!

Verification from actual run (HERA nightly):

CYCLE              TASK                    JOBID       STATE
202103231200       gfs_ocean_prod_f006     16850999    SUCCEEDED
202103231200       gfs_ice_prod_f006       16851001    SUCCEEDED

Located in gdas.{PDY}/{cyc}/model/ocean/restart/ and gfs.{PDY}/{cyc}/model/ocean/history/:

6-Hour Averaged Output: Ocean history files contain 6-hour time-averaged fields (not instantaneous), matching operational ocean output cadence.

All outputs located in gfs.{PDY}/{cyc}/products/ocean/:

Total: 3 files (2 GRIB2 + 1 netCDF)

Resolution Context:

• Model Grid: mx500 (0.5° nominal MOM6 tripolar grid)
• Product Grid: 5p00 (5.0° regular lat-lon for C48 testing)
• Operational: Would use finer grids (0p25, 0p50, 1p00) with higher resolution models

Coupled Forecast Outputs (from C48_S2SW-gfs_fcst_seg0.yaml)
    ├─> Ocean restart: MOM.res.nc (for initialization)
    └─> Ocean history: gfs.ocean.t{cyc}z.6hr_avg.f006.nc (PRIMARY INPUT)
    ↓
COMPONENT=ocean Set in Environment
    ↓
oceanice_products.sh Execution
    ├─> Script: jobs/oceanice_products.sh
    ├─> Python: scripts/exglobal_oceanice_products.py
    ├─> Class: pygfs.task.oceanice_products.OceanIceProducts
    └─> Config: parm/post/oceanice_products_gfs.yaml (lines 23-50)
    ↓
Ocean Processing Pipeline
    ├─> Read MOM6 native grid (tripolar)
    ├─> Interpolate to regular lat-lon grids
    ├─> Generate GRIB2 files at 3 resolutions
    ├─> Create index files for fast access
    └─> Subset native grid to reduce file size
    ↓
Output by Product Type
    ├─> GRIB2 Products (0.25°, 0.50°, 1.00°)
    │   ├─> Sea surface temperature (SST)
    │   ├─> Sea surface salinity
    │   ├─> Ocean currents (u, v components)
    │   ├─> Sea surface height (SSH)
    │   └─> Mixed layer depth
    │
    └─> Native NetCDF Subset
        └─> Selected variables on MOM6 tripolar grid
    ↓
Total Output: 7 files in products/ocean/

Product Generation Test Strategy:

If All Hours Were Tested:

Ocean output every 6 hours: f006, f012, f018, ..., f120
= 20 time steps × 7 files = 140 ocean product files
This test: 7 files (1 time step)
Reduction: 20× fewer files

Job Script Pattern: jobs/oceanice_products.sh

# Line 23: Component controlled by environment
export COMPONENT="ocean"  # Set by Rocoto metatask

# Script behavior changes based on COMPONENT:
if [[ "${COMPONENT}" == "ocean" ]]; then
    # Process ocean history files
    # Use ocean section of config
elif [[ "${COMPONENT}" == "ice" ]]; then
    # Process ice history files
    # Use ice section of config
fi

This allows:

• Single job script for both components
• Parallel execution of ocean and ice tasks
• Independent success/failure tracking
• Component-specific configurations

Why Native NetCDF Output?

Ocean uses tripolar grid (not regular lat-lon):

• North pole singularity avoided by splitting pole into two points
• Grid spacing varies with latitude
• Optimal for Arctic ocean modeling
• Research community prefers native grid structure

Native NetCDF Benefits:

• Preserves grid topology
• No interpolation artifacts
• Full resolution maintained
• Subsetting reduces file size (~50 MB vs full ~500 MB)

Contrast with atmosphere:

• Atmosphere: Cubed-sphere resolved to GRIB2 (lat-lon sufficient)
• Ocean: Tripolar grid doesn't interpolate well → keep native format

Config File: parm/post/oceanice_products_gfs.yaml (lines 23-50)

Ocean Variables in GRIB2:

• Sea surface temperature (SST) - °C
• Sea surface salinity - PSU
• Ocean currents (u, v) - m/s
• Sea surface height (SSH) - m
• Mixed layer depth - m
• Ocean heat content - J/m²

Grids:

• 0.25° lat-lon: 1440×721 = 1,036,800 points
• 0.50° lat-lon: 720×361 = 259,920 points
• 1.00° lat-lon: 360×181 = 65,160 points

Why 3 GRIB2 Resolutions?

Storage Impact:

0.25° files: ~10 MB each (detailed)
0.50° files: ~3 MB each (medium)
1.00° files: ~1 MB each (compact)
Index files: ~50 KB each (metadata)
Native subset: ~50 MB (full resolution on native grid)

Both tests:

• Use same job script with different COMPONENT value
• Generate 3 GRIB2 resolutions + 1 native NetCDF
• Process 6-hour averaged forecast output
• Run independently in parallel

GRIB2 Index (.idx) Files:

Enable fast variable extraction without reading entire GRIB2 file.

Example usage:

# Extract SST using index
wgrib2 -i gfs.ocean.t12z.0p25.f006.grib2 \
       -match ":TMP:surface:" \
       -grib sst_only.grib2

# Uses .idx to jump directly to SST record (byte offset known)
# Avoids scanning 10 MB file sequentially

Critical for:

• Real-time data distribution
• Web services (ocean.weather.gov)
• Automated extraction pipelines
• User applications needing specific variables

✅ MOM6 to Lat-Lon: Native tripolar grid interpolation works
✅ GRIB2 Encoding: Ocean variables encoded correctly
✅ Multi-Resolution: 3 lat-lon grids generated properly
✅ Index Files: GRIB2 indices created for fast access
✅ Native Subsetting: NetCDF subset reduces file size
✅ Component Independence: Ocean processes without ice dependency

Marine Forecasting:

• SST for tropical cyclone intensity forecasts
• Ocean currents for navigation and search-rescue
• Sea surface height for coastal inundation

Coastal Ocean Models:

• Boundary conditions from global ocean forecasts
• Nesting regional models in GFS ocean output

Climate Monitoring:

• Ocean heat content tracking
• SST anomaly detection
• Marine heatwave identification

Shipping/Navigation:

• Current predictions for route optimization
• SST for ice edge determination
• Mixed layer depth for submarine operations

# Execute ocean products test
ctest -R "C48_S2SW.*ocean.*validate" --verbose

# Check output structure
ls -lh gfs.{PDY}/{cyc}/products/ocean/grib2/*/
ls -lh gfs.{PDY}/{cyc}/products/ocean/netcdf/

# Should find 7 files total
find products/ocean -type f | wc -l  # 7

# By resolution
ls products/ocean/grib2/0p25/  # 2 files (grib2 + idx)
ls products/ocean/grib2/0p50/  # 2 files
ls products/ocean/grib2/1p00/  # 2 files
ls products/ocean/netcdf/      # 1 file (native subset)

# List variables in ocean GRIB2 file
wgrib2 products/ocean/grib2/0p25/gfs.ocean.t12z.0p25.f006.grib2

# Expected variables:
# - TMP (sea surface temperature)
# - SALTY (salinity)
# - UOGRD/VOGRD (ocean currents)
# - DSLM (sea surface height)

Why split from combined oceanice_prod test?

1. Workflow Architecture: Rocoto runs ocean and ice as separate tasks
2. Independent Validation: Ocean failures don't affect ice validation
3. Parallel Execution: Matches actual operational task parallelism
4. Clear Ownership: Each test validates one specific task

Benefits realized:

• More accurate representation of workflow
• Easier debugging (component-specific failures)
• Better test organization
• Matches operational job naming

Global Workflow MCP tools provided:

• Configuration file structure references
• COM template variable definitions
• Task dependency patterns

Value demonstrated:

• Quick lookup of products/ path definitions
• Understanding of component-based processing
• Workflow metatask structure insights

1. C48_S2SW-gfs_fcst_seg0.yaml - Upstream (provides ocean history files)
2. C48_S2SW-gfs_ice_prod_f006.yaml - Sibling (ice products, parallel task)
3. C48_ATM-gfs_atmos_prod_f000-f002.yaml - Atmospheric products (similar pattern)

• Test Definition: dev/ctests/cases/C48_S2SW-gfs_ocean_prod_f006.yaml
• Job Script: jobs/oceanice_products.sh
• Execution Script: scripts/exglobal_oceanice_products.py
• Python Class: pygfs.task.oceanice_products.OceanIceProducts

• Product Spec: parm/post/oceanice_products_gfs.yaml (lines 23-50)
• COM Templates: parm/config/gfs/config.com
• Base Config: parm/config/gfs/config.base.j2

• Repository: TerrenceMcGuinness-NOAA/global-workflow
• Branch: ctest_case_updates
• Changelog: CTEST_UPDATES_CHANGELOG.md (Part 5)
• Split Summary: dev/ctests/OCEANICE_TEST_SPLIT_SUMMARY.md

Created: October 1, 2025 (split from combined test)
Updated: October 1, 2025
Status: ✅ Fixed and validated - Tests passing