jgfs_marine_bmat.ecf (GFS v17)

Branch / tenant: dev/gfs.v17 (gw_v17) Path: ecf/scripts/gfs/analysis/marine/jgfs_marine_bmat.ecf Analysis date: 2026-06-11 Method: AWS Bedrock AgentCore agentcore-mcp-rag MCP cross-checked against the on-disk v17 worktree.

---

Summary

jgfs_marine_bmat.ecf is the NCO production ecFlow task that computes the marine background error covariance matrix (B-matrix) for the ocean/ice data assimilation. This is a computationally intensive task using ensemble statistics to characterize background uncertainty. It calls JGLOBAL_MARINE_BMAT.

---

1. The script itself

Section	Detail
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 stack)
J-Job	JGLOBAL_MARINE_BMAT
Rocoto breadcrumb	marinebmat.sh

---

2. Execution chain

jgfs_marine_bmat.ecf
  +-- JGLOBAL_MARINE_BMAT
       +-- exglobal_marine_bmat.py (Python/pygfs)
            +-- Computes ocean B-matrix from ensemble spread
            +-- Computes ice B-matrix
            +-- Uses SOCA/OOPS bump operators

---

3. Key environment variables

Variable	Meaning
COMIN_OCEAN_BMATRIX	Output: ocean B-matrix
COMIN_ICE_BMATRIX	Output: ice B-matrix

Configuration

Sources jjob_header.sh -e "marinebmat" -c "base marineanl marinebmat" (inferred from pattern).

---

4. v17-specific notes

• Heavy resources: 8 nodes x 64 cpus = 512 ranks with 500 GB per node. This is one of the most memory-intensive DA tasks.
• Exclusive host placement — full-node allocation.
• Python-based driver using pygfs and SOCA/OOPS B-matrix computation.
• JEDI/SOCA: loads ufsda modules.
• Part of marine DA chain: after anl_init / bmat_init, before anlvar.

---

5. Change-impact snapshot

Symbol	Risk	Notes
JGLOBAL_MARINE_BMAT	MEDIUM	Feeds B-matrix to marine variational solver

---

Sources

• On-disk v17 worktree: ecf/scripts/gfs/analysis/marine/jgfs_marine_bmat.ecf.
• MCP (tenant gw_v17): on-disk analysis.