diff --git a/CHANGELOG.md b/CHANGELOG.md
index e45d5e3bf..d27831c88 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -31,6 +31,7 @@ Code freeze date: YYYY-MM-DD
 - `Hazard.from_raster_xarray` now returns a sparse matrix instead of a sparse array [#1261](https://github.com/CLIMADA-project/climada_python/pull/1261).
 - `ImpactCalc.impact` now raises a clear `ValueError` when the supplied `Hazard` contains no events, instead of failing later inside `np.array_split` with an obscure message [#814](https://github.com/CLIMADA-project/climada_python/issues/814).
 - Fix TCTracks.from_FAST duplicate loading from year loop [#1269](github.com/CLIMADA-project/climada_python/pull/1269)
+- Fix a wrong computation of EAI in the `trajectory` module when the index of the `GeoDataFrame` of an Exposures object is not a 0 to n range. [#1297](https://github.com/CLIMADA-project/climada_python/pull/1297).
 - `TropCyclone.apply_climate_scenario_knu` now raises a `ValueError` if a frequency is negative [#1338](https://github.com/CLIMADA-project/climada_python/pull/1338).
 
 ### Deprecated
diff --git a/climada/trajectories/calc_risk_metrics.py b/climada/trajectories/calc_risk_metrics.py
index 39c0b8fca..56509c4fd 100644
--- a/climada/trajectories/calc_risk_metrics.py
+++ b/climada/trajectories/calc_risk_metrics.py
@@ -255,7 +255,12 @@ def calc_eai_gdf(self) -> pd.DataFrame:
 
         """
 
-        metric_df = pd.DataFrame(self.per_date_eai, index=self._date_idx)
+        metric_df = pd.DataFrame(
+            self.per_date_eai,
+            index=self._date_idx,
+            columns=self.snapshots[0].exposure.gdf.index,
+        )
+
         metric_df = metric_df.reset_index().melt(
             id_vars=DATE_COL_NAME, var_name=COORD_ID_COL_NAME, value_name=RISK_COL_NAME
         )
@@ -868,7 +873,11 @@ def calc_eai_gdf(self) -> pd.DataFrame:
         (notably for `value` and `group_id`).
 
         """
-        metric_df = pd.DataFrame(self.per_date_eai, index=self.date_idx)
+        metric_df = pd.DataFrame(
+            self.per_date_eai,
+            index=self._date_idx,
+            columns=self.snapshot_start.exposure.gdf.index,
+        )
         metric_df = metric_df.reset_index().melt(
             id_vars=DEFAULT_PERIOD_INDEX_NAME,
             var_name=COORD_ID_COL_NAME,
diff --git a/climada/trajectories/test/conftest.py b/climada/trajectories/test/conftest.py
index 57a1d3a27..5ff7e6a64 100644
--- a/climada/trajectories/test/conftest.py
+++ b/climada/trajectories/test/conftest.py
@@ -39,7 +39,8 @@
 EXP_DESC = "Test exposure dataset"
 EXPOSURE_REF_YEAR = 2020
 EXPOSURE_VALUE_UNIT = "USD"
-VALUES = np.array([0, 1000, 2000, 3000, 4000, 5000])
+EXPOSURE_INDICES = np.array([0, 1, 2, 7, 8, 5])
+EXPOSURE_VALUES = np.array([0, 1000, 2000, 3000, 4000, 5000])
 CATEGORIES = np.array([1, 1, 2, 1, 1, 3])
 
 # Exposure coordinates
@@ -60,7 +61,7 @@
 # Hazard events
 EVENT_IDS = np.array([1, 2, 3, 4, 5])
 EVENT_NAMES = ["ev1", "ev2", "ev3", "ev4", "ev5"]
-DATES = np.array([1, 2, 3, 4, 5])
+DATES = np.array([1, 2, 7, 4, 5])
 
 # Frequency are choosen so that they cumulate nicely
 # to correspond to 250, 100, 50, and 20y return periods (for impacts)
@@ -80,9 +81,9 @@
 IMPF_NAME = "IMPF_1"
 
 # Sanity checks
-for const in [VALUES, CATEGORIES, EXP_LONS, EXP_LATS]:
+for const in [EXPOSURE_VALUES, CATEGORIES, EXP_LONS, EXP_LATS]:
     assert len(const) == len(
-        VALUES
+        EXPOSURE_VALUES
     ), "VALUES, REGIONS, CATEGORIES, EXP_LONS, EXP_LATS should all have the same lengths."
 
 for const in [EVENT_IDS, EVENT_NAMES, DATES, FREQUENCY]:
@@ -93,7 +94,7 @@
 
 @pytest.fixture(scope="session")
 def exposure_values():
-    return VALUES.copy()
+    return EXPOSURE_VALUES.copy()
 
 
 @pytest.fixture(scope="session")
@@ -124,6 +125,7 @@ def _make_exposures(
                 "geometry": exposure_geometry,
             },
             crs=CRS_WGS84,
+            index=EXPOSURE_INDICES,
         )
         if group_id is not None:
             gdf["group_id"] = group_id