fdb-to-grib

Extract fields from FDB and write them to grouped GRIB files, driven by a YAML config.

Requirements

• balfrin (FDB filesystem is only mounted there)
• uenv fdb/5.18:v3

Scripts

Script	Purpose
fdb_to_grib.py	Read config, fetch from FDB, write GRIB files
grib_to_config.py	Inspect an existing GRIB file and generate the matching config

---

fdb_to_grib.py

uenv run --view=<view> fdb/5.18:v3 -- python fdb_to_grib.py config.yaml

Model	--view
ICON-CH1-EPS, ICON-CH2-EPS	realtime
icon-rea-l-ch1	rea-l-ch1

Config file

The config is validated against config_schema.json at startup. In VS Code, add the following line at the top of your YAML file for inline validation and autocomplete (requires the YAML extension):

# yaml-language-server: $schema=./config_schema.json

Top-level keys

Key	Required	Type	Description
model	yes	string	Model name. Sets class, stream, expver automatically. One of ICON-CH1-EPS, ICON-CH2-EPS, icon-rea-l-ch1.
type	no	string	MARS type. Default: cf. Typical values: cf (control forecast), pf (perturbed forecast), an (analysis).
date_range	yes	object	start and end dates in YYYYMMDD format (inclusive).
times	yes	list of strings	Run times in HHMM format.
steps	yes	list or range	Global step specification applied to all blocks (unless a block overrides it).
blocks	yes	list of block objects	One block per levtype.
group_by	yes	list of strings	GRIB metadata keys used to split output files. Fields with identical values for all keys go into the same file.
output_dir	no	string	Root directory for output files. Default: ./output. Created if it does not exist.
output_template	no	string	Filename template using {key} placeholders for each group_by key. Slashes in the template create subdirectories. Default: {key1}_{key2}_....grib.

Step / level specification

Both steps and levels accept either an explicit list or a range dict:

# explicit list
steps: [0, 6, 12, 24]

# arithmetic range  →  produces "0/to/24/by/6" in the FDB request
steps:
  start: 0
  end: 24
  by: 6          # optional, default 1

# minute-based range  →  produces "0m/to/1440m/by/10m"
steps:
  start: 0
  end: 1440
  by: 10
  unit: m

Block keys

Key	Required	Type	Description
levtype	yes	string	Level type: sfc, ml, pl, dp, hl.
params	yes	list of int	paramId values to retrieve.
levels	for non-sfc	list or range	Levels to retrieve (levelist). Not used for levtype: sfc.
steps	no	list or range	Per-block step override. Useful when different levtypes use different step units.

---

Examples

Example 1 — REA-L-CH1: surface and model-level fields, grouped by date and step

# yaml-language-server: $schema=./config_schema.json
model: icon-rea-l-ch1
type: cf

date_range:
  start: "20250101"
  end: "20250103"

times:
  - "0000"

steps:
  start: 0
  end: 24
  by: 1

blocks:
  - levtype: sfc
    params:
      - 500011   # T_2M
      - 500017   # TD_2M
    steps:       # sfc instant vars use minute-based steps
      start: 0
      end: 1440
      by: 10
      unit: m

  - levtype: ml
    params:
      - 500014   # T
      - 500028   # U
      - 500030   # V
    levels:
      start: 1
      end: 80

group_by:
  - date
  - time
  - step

output_dir: ./output
output_template: "{date}_{time}_{step}.grib"

uenv run --view=rea-l-ch1 fdb/5.18:v3 -- python fdb_to_grib.py config.yaml

Output: one GRIB file per (date, time, step) combination, e.g. output/20250101_0000_0.grib.

---

Example 2 — ICON-CH1-EPS: one directory per run, one file per step

# yaml-language-server: $schema=./config_schema.json
model: ICON-CH1-EPS
type: cf

date_range:
  start: "20260402"
  end: "20260402"

times:
  - "0900"

steps:
  start: 0
  end: 33
  by: 1

blocks:
  - levtype: sfc
    params:
      - 500011   # T_2M
      - 500041   # TOT_PREC
      - 500046   # CLCT

  - levtype: ml
    params:
      - 500014   # T
      - 500001   # P
    levels:
      start: 1
      end: 81

group_by:
  - date
  - time
  - step

output_dir: ./output
output_template: "{date}{time}/forecast_{step}.grib"

uenv run --view=realtime fdb/5.18:v3 -- python fdb_to_grib.py config.yaml

Output: output/202604020900/forecast_0.grib, output/202604020900/forecast_1.grib, …

---

Example 3 — group by date only (all steps in one file per day)

group_by:
  - date

output_template: "{date}.grib"

All steps and levtypes for a given date are written into a single file.

---

grib_to_config.py

Inspects an existing GRIB file and generates the config needed to reconstruct it from FDB.

uenv run --view=<view> fdb/5.18:v3 -- python grib_to_config.py input.grib [output_config.yaml]

If output_config.yaml is omitted the config is printed to stdout.

Steps and levels are automatically expressed as ranges where possible. The group_by defaults to [date, time, step]; adjust manually if needed.

Example

uenv run --view=realtime fdb/5.18:v3 -- \
    python grib_to_config.py \
    /store_new/mch/msopr/osm/ICON-CH1-EPS/MEC_RING/2026040209/i1emec01090000.ctrl \
    config_out.yaml

Output:

model: ICON-CH1-EPS
type: cf
date_range:
  start: '20260402'
  end: '20260402'
times:
- 0900
steps:
- 33
blocks:
- levtype: ml
  params: [500001, 500014, 500028, 500030, 500032, 500035, 500100, 500101]
  levels:
    start: 1
    end: 81
- levtype: pl
  params: [500048, 500049, 500050]
  levels:
    start: 0
    end: 800
    by: 400
- levtype: sfc
  params: [500000, 500010, 500011, 500015, 500016, 500017, 500027, 500029,
           500041, 500045, 500046, 500164, 500480, 500481]
group_by: [date, time, step]
output_dir: ./output