diff --git a/workflow/rules/data.smk b/workflow/rules/data.smk
index 0696657..9892f8f 100644
--- a/workflow/rules/data.smk
+++ b/workflow/rules/data.smk
@@ -4,60 +4,6 @@ from pathlib import Path
 include: "common.smk"
 
 
-if "extract_cosmoe" in config.get("include-optional-rules", []):
-
-    rule extract_cosmoe:
-        input:
-            archive=Path("/archive/mch/msopr/osm/COSMO-E"),
-        output:
-            fcts=protected(
-                directory(Path("/store_new/mch/msopr/ml/COSMO-E/FCST{year}.zarr"))
-            ),
-        resources:
-            cpus_per_task=4,
-            runtime="24h",
-        params:
-            year_postfix=lambda wc: f"FCST{wc.year}",
-            steps="0/120/6",
-        log:
-            OUT_ROOT / "logs/extract-cosmoe-fcts-{year}.log",
-        shell:
-            """
-            python workflow/scripts/extract_baseline_fct.py \
-                --archive_dir {input.archive}/{params.year_postfix} \
-                --output_store {output.fcts} \
-                --steps {params.steps} \
-                    > {log} 2>&1
-            """
-
-
-if "extract_cosmo1e" in config.get("include-optional-rules", []):
-
-    rule extract_cosmo1e:
-        input:
-            archive=Path("/archive/mch/s83/osm/from_GPFS/COSMO-1E"),
-        output:
-            fcts=protected(
-                directory(Path("/store_new/mch/msopr/ml/COSMO-1E/FCST{year}.zarr"))
-            ),
-        resources:
-            cpus_per_task=4,
-            runtime="24h",
-        params:
-            year_postfix=lambda wc: f"FCST{wc.year}",
-            steps="0/33/1",
-        log:
-            OUT_ROOT / "logs/extract-cosmo1e-fcts-{year}.log",
-        shell:
-            """
-            python workflow/scripts/extract_baseline_fct.py \
-                --archive_dir {input.archive}/{params.year_postfix} \
-                --output_store {output.fcts} \
-                --steps {params.steps} \
-                    > {log} 2>&1
-            """
-
-
 rule download_obs_from_peakweather:
     localrule: True
     output:
diff --git a/workflow/scripts/extract_baseline.py b/workflow/scripts/extract_baseline.py
index f450b82..075ad1d 100644
--- a/workflow/scripts/extract_baseline.py
+++ b/workflow/scripts/extract_baseline.py
@@ -20,21 +20,45 @@
 )
 
 
+def get_input(root: Path) -> list[Path]:
+    """Get list of tarfiles or directories in root directory."""
+    input_files = sorted(root.glob("*.tar"))
+    if not input_files:
+        gribfiles = sorted(root.glob("*_*/grib/i?eff00000000_000"))
+        input_files = [f.parent.parent for f in gribfiles]
+    if not input_files:
+        raise ValueError(f"No files found in {root}.")
+    return input_files
+
+
+def get_reftime(file: Path) -> datetime:
+    if ".tar" in file.suffixes:
+        return reftime_from_tarfile(file)
+    else:
+        return reftime_from_directory(file)
+
+
+def reftime_from_directory(directory: Path) -> datetime:
+    """Extract reftime from directory name."""
+    dir_stem = directory.name.rsplit("_", 1)[0]
+    return datetime.strptime(dir_stem, "%y%m%d%H")
+
+
 def reftime_from_tarfile(tarfile: Path, suffix: str | None = None) -> datetime:
     """Extract reftime from tarfile name."""
     suffix = tarfile.stem[-4:] if suffix is None else suffix
     return datetime.strptime(tarfile.stem.removesuffix(suffix), "%y%m%d%H")
 
 
-def check_reftime_consistency(tarfiles: list[Path], delta_h: int = 12):
+def check_reftime_consistency(input: list[Path], delta_h: int = 12):
     """Check that all reftimes are available and every delta_h hours."""
 
     # note the lower case y in the format string, it's for 2-digit years
 
-    first_reftime = reftime_from_tarfile(tarfiles[0])
+    first_reftime = get_reftime(input[0])
     expected_reftime = first_reftime
-    for file in tarfiles:
-        reftime = reftime_from_tarfile(file)
+    for file in input:
+        reftime = get_reftime(file)
         if reftime != expected_reftime:
             raise ValueError(f"Expected reftime {expected_reftime} but got {reftime}.")
         expected_reftime += timedelta(hours=delta_h)
@@ -42,31 +66,47 @@ def check_reftime_consistency(tarfiles: list[Path], delta_h: int = 12):
 
 
 def extract(
-    tar: Path, lead_times: list[int], run_id: str, params: list[str]
+    file: Path, lead_times: list[int], run_id: str, params: list[str]
 ) -> xr.Dataset:
-    LOG.info(f"Extracting fields from {tar}.")
-    reftime = reftime_from_tarfile(tar)
-    if "COSMO-E" in tar.parts:
+    LOG.info(f"Extracting fields from {file}.")
+    reftime = reftime_from_tarfile(file)
+    if "COSMO-E" in file.parts:
         gribname = "ceffsurf"
-    elif "COSMO-1E" in tar.parts:
+    elif "COSMO-1E" in file.parts:
         gribname = "c1effsurf"
+    elif "ICON-CH1-EPS" in file.parts:
+        gribname = "i1eff"
+    elif "ICON-CH2-EPS" in file.parts:
+        gribname = "i2eff"
     else:
-        raise ValueError("Currently only COSMO-E and COSMO-1E are supported.")
-    tar_archive = tarfile.open(tar, mode="r:*")
+        raise ValueError("Currently only COSMO-E/1E and ICON-CH1/2-EPS are supported.")
     out = ekd.SimpleFieldList()
-    for lt in lead_times:
-        filename = f"{tar.stem}/grib/{gribname}{lt:03}_{run_id}"
-        LOG.info(f"Extracting {filename}.")
-        stream = tar_archive.extractfile(filename)
-
-        # LOG.info(f"Reading fields...")
-        streamfieldlist: StreamFieldList = ekd.from_source("stream", stream)
-        for field in streamfieldlist:
-            shortname = field.metadata("shortName")
-            if shortname in params:
-                out.append(field)
-        stream.close()
-    tar_archive.close()
+    if ".tar" in file.suffixes:
+        tar_archive = tarfile.open(file, mode="r:*")
+        for lt in lead_times:
+            filename = f"{file.stem}/grib/{gribname}{lt:03}_{run_id}"
+            LOG.info(f"Extracting {filename}.")
+            stream = tar_archive.extractfile(filename)
+
+            # LOG.info(f"Reading fields...")
+            streamfieldlist: StreamFieldList = ekd.from_source("stream", stream)
+            for field in streamfieldlist:
+                shortname = field.metadata("shortName")
+                if shortname in params:
+                    out.append(field)
+            stream.close()
+        tar_archive.close()
+    else:
+        for lt in lead_times:
+            lh = lt % 24
+            ld = lt // 24
+            filepath = file / "grib" / f"{gribname}{ld:02}{lh:02}0000_{run_id}"
+            LOG.info(f"Extracting {filepath}.")
+            fields = ekd.from_source("file", filepath)
+            for field in fields:
+                shortname = field.metadata("shortName")
+                if shortname in params:
+                    out.append(field)
 
     out = out.to_xarray(profile="grib")
     out = out.expand_dims(
@@ -95,16 +135,16 @@ def _parse_steps(steps: str) -> int:
 
 
 def main(cfg: ScriptConfig):
-    tarfiles = sorted(cfg.archive_dir.glob("*.tar"))
+    input = get_input(cfg.archive_dir)
     delta_h = 12
-    if "COSMO-1E" in tarfiles[0].parts:
+    if "COSMO-1E" in input[0].parts or "ICON-CH1-EPS" in input[0].parts:
         delta_h = 3
-    first_reftime, last_reftime = check_reftime_consistency(tarfiles, delta_h)
-    LOG.info(
-        f"Found {len(tarfiles)} tar archives from {first_reftime} to {last_reftime}."
-    )
+    if "ICON-CH2-EPS" in input[0].parts:
+        delta_h = 6
+    first_reftime, last_reftime = check_reftime_consistency(input, delta_h)
+    LOG.info(f"Found {len(input)} forecasts from {first_reftime} to {last_reftime}.")
 
-    reftimes = np.array([reftime_from_tarfile(f) for f in tarfiles], dtype="datetime64")
+    reftimes = np.array([get_reftime(f) for f in input], dtype="datetime64")
     missing = reftimes
     if not cfg.overwrite:  # only check dataset when we want to append as this is slow
         existing_reftimes = np.array([])
@@ -130,7 +170,7 @@ def main(cfg: ScriptConfig):
     _, indices, _ = np.intersect1d(reftimes, missing, return_indices=True)
 
     for i in indices:
-        file = tarfiles[i]
+        file = input[i]
         ds = extract(file, cfg.steps, cfg.run_id, cfg.params)
 
         LOG.info(f"Extracted: {ds}")
@@ -144,9 +184,12 @@ def main(cfg: ScriptConfig):
         zarr_encoding = {
             "forecast_reference_time": {"units": "nanoseconds since 1970-01-01"}
         }
+        cfg.output_store.parent.mkdir(parents=True, exist_ok=True)
         if i == 0:
+            LOG.info(f"Creating new zarr store at {cfg.output_store}.")
             ds.to_zarr(cfg.output_store, mode="w", encoding=zarr_encoding)
         else:
+            LOG.info(f"Appending to existing zarr store at {cfg.output_store}.")
             ds.to_zarr(cfg.output_store, mode="a", append_dim="forecast_reference_time")
 
 
@@ -186,13 +229,27 @@ def main(cfg: ScriptConfig):
 
 """
 Example usage:
-python workflow/scripts/extract_baseline_fct.py \
+
+To submit as a batch job on compute nodes
+sbatch --wrap "uv run python ..."
+
+python workflow/scripts/extract_baseline.py \
     --archive_dir /archive/mch/msopr/osm/COSMO-E/FCST20 \
     --output_store /store_new/mch/msopr/ml/COSMO-E/FCST20.zarr \
     --steps 0/120/6
 
-python workflow/scripts/extract_baseline_fct.py \
+python workflow/scripts/extract_baseline.py \
     --archive_dir /archive/mch/s83/osm/from_GPFS/COSMO-1E/FCST20 \
     --output_store /store_new/mch/msopr/ml/COSMO-1E/FCST20.zarr \
     --steps 0/33/1
+
+python workflow/scripts/extract_baseline.py \
+    --archive_dir /store_new/mch/msopr/osm/ICON-CH1-EPS/FCST24 \
+    --output_store /store_new/mch/msopr/ml/ICON-CH1-EPS/FCST24.zarr \
+    --steps 0/33/1
+
+python workflow/scripts/extract_baseline.py \
+    --archive_dir /store_new/mch/msopr/osm/ICON-CH1-EPS/FCST25 \
+    --output_store /store_new/mch/msopr/ml/ICON-CH1-EPS/FCST25.zarr \
+    --steps 0/33/1
 """