diff --git a/aurora/pipelines/process_mth5.py b/aurora/pipelines/process_mth5.py
index 85ee9040..aa496bcd 100644
--- a/aurora/pipelines/process_mth5.py
+++ b/aurora/pipelines/process_mth5.py
@@ -50,7 +50,6 @@
 )
 from aurora.transfer_function.TTFZ import TTFZ
 
-
 SUPPORTED_PROCESSINGS = [
     "legacy",
 ]
@@ -188,6 +187,13 @@ def process_mth5_legacy(
         tfk.apply_clock_zero(dec_level_config)
 
         stfts = get_spectrograms(tfk, i_dec_level, units=units)
+        if stfts is None:
+            logger.warning(
+                "Skipping transfer-function estimation for decimation level {} "
+                "because no complete STFT set is available",
+                i_dec_level,
+            )
+            continue
 
         local_merged_stft_obj, remote_merged_stft_obj = merge_stfts(stfts, tfk)
 
diff --git a/aurora/time_series/spectrogram_helpers.py b/aurora/time_series/spectrogram_helpers.py
index 3cbd0019..cf775917 100644
--- a/aurora/time_series/spectrogram_helpers.py
+++ b/aurora/time_series/spectrogram_helpers.py
@@ -391,7 +391,7 @@ def save_fourier_coefficients(
 
 def get_spectrograms(
     tfk: TransferFunctionKernel, i_dec_level: int, units: str = "MT"
-) -> dict:
+) -> dict | None:
     """
     Given a decimation level id, loads a dictianary of all spectragrams from information in tfk.
     TODO: Make this a method of TFK
@@ -409,9 +409,10 @@ def get_spectrograms(
 
     Returns
     -------
-    stfts: dict
+    stfts: dict or None
         The short time fourier transforms for the decimation level as a dictionary.
         Keys are "local" and "remote".  Values are lists, one (element) xr.Dataset per run
+        or None when a valid run cannot produce any STFT windows.
     """
 
     stfts = {}
@@ -438,13 +439,24 @@ def get_spectrograms(
         # Musgraves workaround for old MT data
         triage_run_id(row.run, run_obj)
 
-        stft_obj = make_stft_objects(
-            tfk.config,
-            i_dec_level,
-            run_obj,
-            run_xrds,
-            units,
-        )
+        try:
+            stft_obj = make_stft_objects(
+                tfk.config,
+                i_dec_level,
+                run_obj,
+                run_xrds,
+                units,
+            )
+        except ValueError as error:
+            logger.warning(
+                "Skipping decimation level {} because station {}, run {} "
+                "cannot produce STFT windows: {}",
+                i_dec_level,
+                row.station,
+                row.run,
+                error,
+            )
+            return None
         # TODO: Cast stft_obj to a Spectrogram here or in make_stft_objects
 
         # Pack FCs into h5
@@ -491,7 +503,7 @@ def run_ts_to_stft(
         run_xrds, dt=1.0 / decimation_obj.decimation.sample_rate
     )
     if not np.prod(windowed_obj.to_array().data.shape):
-        raise ValueError
+        raise ValueError("STFT windowing produced no samples")
 
     windowed_obj = WindowedTimeSeries.detrend(data=windowed_obj, detrend_type="linear")
     tapered_obj = WindowedTimeSeries.apply_taper(
diff --git a/tests/synthetic/test_processing_pytest.py b/tests/synthetic/test_processing_pytest.py
index ddf60308..df543a50 100644
--- a/tests/synthetic/test_processing_pytest.py
+++ b/tests/synthetic/test_processing_pytest.py
@@ -4,9 +4,14 @@
 """
 
 import logging
+from unittest.mock import Mock, patch
 
+import numpy as np
+import pandas as pd
 import pytest
+import xarray as xr
 
+from aurora.time_series.spectrogram_helpers import get_spectrograms
 from aurora.test_utils.synthetic.processing_helpers import (
     process_synthetic_1,
     process_synthetic_1r2,
@@ -21,6 +26,56 @@ def setup_logging():
     logging.getLogger("matplotlib.ticker").disabled = True
 
 
+def test_empty_stft_skips_decimation_without_saving_fourier_coefficients():
+    """A valid run that yields no windows skips its entire decimation level."""
+    run_object = Mock()
+    mth5_object = Mock()
+    mth5_object.from_reference.return_value = run_object
+    run_dataarray = xr.DataArray(
+        np.arange(4.0)[np.newaxis, :],
+        dims=["channel", "time"],
+        coords={
+            "channel": ["hx"],
+            "time": np.datetime64("2020-01-01T00:00:00")
+            + np.arange(4).astype("timedelta64[s]"),
+        },
+    )
+    tfk = Mock()
+    tfk.dataset_df = pd.DataFrame(
+        [
+            {
+                "mth5_obj": mth5_object,
+                "run_hdf5_reference": "run-reference",
+                "fc": False,
+                "run_dataarray": run_dataarray,
+                "run": "001",
+                "station": "test1",
+                "remote": False,
+            }
+        ]
+    )
+    tfk.is_valid_dataset.return_value = True
+
+    with (
+        patch("aurora.time_series.spectrogram_helpers.triage_run_id"),
+        patch(
+            "aurora.time_series.spectrogram_helpers.make_stft_objects",
+            side_effect=ValueError("STFT windowing produced no samples"),
+        ),
+        patch(
+            "aurora.time_series.spectrogram_helpers.save_fourier_coefficients"
+        ) as save_fourier_coefficients,
+        patch("aurora.time_series.spectrogram_helpers.logger.warning") as log_warning,
+    ):
+        stfts = get_spectrograms(tfk, i_dec_level=2)
+
+    assert stfts is None
+    save_fourier_coefficients.assert_not_called()
+    log_warning.assert_called_once()
+    assert "Skipping decimation level" in log_warning.call_args.args[0]
+    assert log_warning.call_args.args[1] == 2
+
+
 @pytest.mark.skip(
     reason="mt_metadata pydantic branch has issue with provenance.archive.comments.value being None"
 )