diff --git a/imap_processing/glows/l2/glows_l2_data.py b/imap_processing/glows/l2/glows_l2_data.py
index c91b797f9..46a91ad2e 100644
--- a/imap_processing/glows/l2/glows_l2_data.py
+++ b/imap_processing/glows/l2/glows_l2_data.py
@@ -370,10 +370,20 @@ def __init__(
         """
         active_flags = np.array(pipeline_settings.active_bad_time_flags, dtype=float)
 
+        # Apply sunrise/sunset offsets to extend the night region around
+        # is_night transitions before selecting good blocks.
+        flags = self.apply_is_night_offsets(
+            l1b_dataset["flags"].data,
+            is_night_idx=6,  # is_night is the 7th bad-time flag (0-indexed)
+            sunrise_offset=pipeline_settings.sunrise_offset,
+            sunset_offset=pipeline_settings.sunset_offset,
+        )
+        flags_da = xr.DataArray(flags, dims=l1b_dataset["flags"].dims)
+
         # Select the good blocks (i.e. epoch values) according to the flags. Drop any
         # bad blocks before processing.
         good_data = l1b_dataset.isel(
-            epoch=self.return_good_times(l1b_dataset["flags"], active_flags)
+            epoch=self.return_good_times(flags_da, active_flags)
         )
         # TODO: bad angle filter
         # TODO: filter bad bins out. Needs to happen here while everything is still
@@ -533,6 +543,97 @@ def return_good_times(flags: xr.DataArray, active_flags: NDArray) -> NDArray:
         good_times = np.where(np.all(flags[:, active_flags == 1] == 1, axis=1))[0]
         return good_times
 
+    @staticmethod
+    def apply_is_night_offsets(
+        flags: np.ndarray,
+        is_night_idx: int,
+        sunrise_offset: float,
+        sunset_offset: float,
+    ) -> np.ndarray:
+        """
+        Apply sunrise/sunset offsets to is_night transitions.
+
+        Per algorithm doc v4.4.7, Sec. 3.9.1, item 2 (raw is_night: 1=night, 0=day):
+
+        sunset_offset applies at both transitions:
+          >0: night shortens by N at each end (first N night epochs at sunset become
+              day; last N night epochs before sunrise become day)
+          <0: night extends by |N| at each end
+
+        sunrise_offset is an additional adjustment at sunrise (is_night 1->0) only:
+          >0: night extends N histograms past the raw sunrise transition
+          <0: night shortens by |N| before the raw sunrise transition
+
+        In the processed flags array: 0 = bad (night), 1 = good (day).
+
+        Parameters
+        ----------
+        flags : numpy.ndarray
+            Flags array with shape (n_epochs, FLAG_LENGTH), 0=bad, 1=good.
+        is_night_idx : int
+            Column index of the is_night flag in the flags array.
+        sunrise_offset : float
+            Additional histogram shift at the sunrise (is_night 1->0) transition.
+        sunset_offset : float
+            Histogram shift applied at both the sunset and sunrise transitions.
+
+        Returns
+        -------
+        numpy.ndarray
+            Copy of flags with the is_night column adjusted per the offsets.
+
+        Notes
+        -----
+        Algorithm doc v4.4.7, Sec. 3.9.1, item 2
+        is_night: 1 = daytime (good), 0 = night (bad)
+        """
+        flags_with_offsets = flags.copy()
+
+        # If sunrise_offset=0 and sunset_offset=0 then no corrections are needed
+        # relative to is_night transition set onboard.
+        if sunrise_offset == 0 and sunset_offset == 0:
+            return flags_with_offsets
+
+        is_night_col = flags[:, is_night_idx]
+        n = flags.shape[0]
+        diff = np.diff(is_night_col.astype(int))
+        sunset_index = np.where(diff == -1)[0]
+        sunrise_index = np.where(diff == 1)[0]
+
+        if sunrise_offset > 0:
+            # Night (flag = 0) extends by sunrise_offset relative
+            # to is_night 0 -> 1 transition.
+            for i in sunrise_index:
+                flags_with_offsets[
+                    i + 1 : min(n, i + 1 + int(sunrise_offset)), is_night_idx
+                ] = 0
+
+        if sunrise_offset < 0:
+            # Night (flag = 0) shortens by sunrise_offset relative
+            # to is_night 0 -> 1 transition.
+            for i in sunrise_index:
+                flags_with_offsets[
+                    max(0, i + 1 + int(sunrise_offset)) : i + 1, is_night_idx
+                ] = 1
+
+        if sunset_offset > 0:
+            # Night (flag = 0) shortens by sunset_offset relative
+            # to is_night 1 -> 0 transition.
+            for i in sunset_index:
+                flags_with_offsets[
+                    i + 1 : min(n, i + 1 + int(sunset_offset)), is_night_idx
+                ] = 1
+
+        if sunset_offset < 0:
+            # Night (flag = 0) extends by sunset_offset relative
+            # to is_night 1 -> 0 transition.
+            for i in sunset_index:
+                flags_with_offsets[
+                    max(0, i + 1 + int(sunset_offset)) : i + 1, is_night_idx
+                ] = 0
+
+        return flags_with_offsets
+
     def compute_position_angle(self) -> float:
         """
         Compute the position angle based on the instrument mounting.
diff --git a/imap_processing/tests/glows/test_glows_l2_data.py b/imap_processing/tests/glows/test_glows_l2_data.py
index cb7a71ac6..028c71e80 100644
--- a/imap_processing/tests/glows/test_glows_l2_data.py
+++ b/imap_processing/tests/glows/test_glows_l2_data.py
@@ -342,6 +342,38 @@ def test_filter_good_times():
     assert np.array_equal(good_times, expected_good_times)
 
 
+@pytest.mark.parametrize(
+    "sunrise_offset, sunset_offset, expected_is_night",
+    [
+        # sunset>0 shortens at sunset; sunrise>0 extends at sunrise
+        (1, 1, [1, 1, 1, 1, 0, 0, 0, 1]),
+        # sunset>0 shortens at sunset; sunrise<0 shortens at sunrise
+        (-1, 1, [1, 1, 1, 1, 0, 1, 1, 1]),
+        # sunset<0 extends at sunset; sunrise>0 extends at sunrise
+        (1, -1, [1, 1, 0, 0, 0, 0, 0, 1]),
+        # sunset<0 extends at sunset; sunrise<0 shortens at sunrise
+        (-1, -1, [1, 1, 0, 0, 0, 1, 1, 1]),
+        # zero offsets: no change
+        (0, 0, [1, 1, 1, 0, 0, 0, 1, 1]),
+    ],
+)
+def test_apply_is_night_offsets(sunrise_offset, sunset_offset, expected_is_night):
+    """Test apply_is_night_offsets function."""
+
+    # Setup: epochs 0-2 day, 3-5 night, 6-7 day (processed flags: 0=night, 1=day).
+    flags = np.ones((8, 17), dtype=float)
+    flags[3:6, 6] = 0  # epochs 3-5 are night
+
+    result = HistogramL2.apply_is_night_offsets(
+        flags,
+        is_night_idx=6,
+        sunrise_offset=sunrise_offset,
+        sunset_offset=sunset_offset,
+    )
+
+    assert np.array_equal(result[:, 6], np.array(expected_is_night, dtype=float))
+
+
 # ── spin_angle tests ──────────────────────────────────────────────────────────
 
 
@@ -392,7 +424,8 @@ def test_compute_position_angle():
 def l1b_dataset_full():
     """Minimal L1B dataset with all variables required by HistogramL2.
 
-    Two epochs, four bins, 17 flags (all good).
+    Two epochs, four bins, 17 flags. Both epochs are daytime (is_night=1).
+    All other flags are 1 (good).
     """
     n_epochs, n_bins, n_angle_flags, n_time_flags = 2, 4, 4, 17
     fillval = GlowsConstants.HISTOGRAM_FILLVAL
@@ -402,6 +435,9 @@ def l1b_dataset_full():
     spin_angle = np.tile(np.linspace(0, 270, n_bins), (n_epochs, 1))
     histogram_flag_array = np.zeros((n_epochs, n_angle_flags, n_bins), dtype=np.uint8)
 
+    # All flags good (1). Index 6 is is_night: 1 = daytime (good).
+    flags = np.ones((n_epochs, n_time_flags), dtype=float)
+
     return xr.Dataset(
         {
             "histogram": (["epoch", "bins"], histogram),
@@ -413,10 +449,7 @@ def l1b_dataset_full():
                 histogram_flag_array,
             ),
             "number_of_bins_per_histogram": (["epoch"], [n_bins, n_bins]),
-            "flags": (
-                ["epoch", "flag_index"],
-                np.ones((n_epochs, n_time_flags)),
-            ),
+            "flags": (["epoch", "flag_index"], flags),
             "filter_temperature_average": (["epoch"], [20.0, 21.0]),
             "hv_voltage_average": (["epoch"], [1000.0, 1000.0]),
             "pulse_length_average": (["epoch"], [5.0, 5.0]),