diff --git a/.github/workflows/legacy_test.yml b/.github/workflows/legacy_test.yml
new file mode 100644
index 00000000..7e9e7699
--- /dev/null
+++ b/.github/workflows/legacy_test.yml
@@ -0,0 +1,33 @@
+name: Legacy test
+
+on: 
+    push:
+        branches:
+            - main
+    pull_request:
+        branches:
+            - main
+
+jobs:
+    test:
+        runs-on: ubuntu-latest
+        
+        steps:
+        - uses: actions/checkout@v3
+        - name: Set up Python
+          uses: actions/setup-python@v4
+          with:
+            python-version: 3.8
+
+        - name: Install older dependencies
+          run: |
+            python -m pip install --upgrade pip
+            pip install -r requirements_min.txt
+            pip install pytest
+
+        - name: Install modelskill
+          run: |
+            pip install .[test]
+        - name: Test with pytest
+          run: |
+            pytest --ignore tests/notebooks/
diff --git a/modelskill/comparison/_comparison.py b/modelskill/comparison/_comparison.py
index 8b678e6b..1f291622 100644
--- a/modelskill/comparison/_comparison.py
+++ b/modelskill/comparison/_comparison.py
@@ -50,16 +50,14 @@
 
 
 class Scoreable(Protocol):
-    def score(self, metric: str | Callable, **kwargs: Any) -> Dict[str, float]:
-        ...
+    def score(self, metric: str | Callable, **kwargs: Any) -> Dict[str, float]: ...
 
     def skill(
         self,
         by: str | Iterable[str] | None = None,
         metrics: Iterable[str] | Iterable[Callable] | str | Callable | None = None,
         **kwargs: Any,
-    ) -> SkillTable:
-        ...
+    ) -> SkillTable: ...
 
     def gridded_skill(
         self,
@@ -69,8 +67,7 @@ def gridded_skill(
         metrics: Iterable[str] | Iterable[Callable] | str | Callable | None = None,
         n_min: int | None = None,
         **kwargs: Any,
-    ) -> SkillGrid:
-        ...
+    ) -> SkillGrid: ...
 
 
 def _parse_dataset(data: xr.Dataset) -> xr.Dataset:
@@ -922,7 +919,7 @@ def _to_long_dataframe(
 
         df = (
             data.to_dataframe()
-            .reset_index(names="time")
+            .reset_index()
             .melt(
                 value_vars=self.mod_names,
                 var_name="model",
diff --git a/modelskill/timeseries/_timeseries.py b/modelskill/timeseries/_timeseries.py
index 86262c4c..981478ea 100644
--- a/modelskill/timeseries/_timeseries.py
+++ b/modelskill/timeseries/_timeseries.py
@@ -3,7 +3,6 @@
 from dataclasses import dataclass
 from typing import ClassVar, Optional, TypeVar, Any
 import numpy as np
-from numpy.typing import NDArray
 import pandas as pd
 import xarray as xr
 
@@ -121,7 +120,9 @@ class TimeSeries:
     """Time series data"""
 
     data: xr.Dataset
-    plotter: ClassVar = MatplotlibTimeSeriesPlotter  # TODO is this the best option to choose a plotter? Can we use the settings module?
+    plotter: ClassVar = (
+        MatplotlibTimeSeriesPlotter  # TODO is this the best option to choose a plotter? Can we use the settings module?
+    )
 
     def __init__(self, data: xr.Dataset) -> None:
         self.data = data if self._is_input_validated(data) else _validate_dataset(data)
@@ -213,7 +214,7 @@ def y(self) -> Any:
     def y(self, value: Any) -> None:
         self.data["y"] = value
 
-    def _coordinate_values(self, coord: str) -> float | NDArray[Any]:
+    def _coordinate_values(self, coord: str) -> float | np.ndarray:
         vals = self.data[coord].values
         return np.atleast_1d(vals)[0] if vals.ndim == 0 else vals
 
@@ -222,7 +223,7 @@ def _is_modelresult(self) -> bool:
         return bool(self.data[self.name].attrs["kind"] == "model")
 
     @property
-    def values(self) -> NDArray[Any]:
+    def values(self) -> np.ndarray:
         """Values as numpy array"""
         return self.data[self.name].values
 
diff --git a/pyproject.toml b/pyproject.toml
index e8646dcf..f1c02cb6 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -9,8 +9,8 @@ exclude = ["notebooks", "tests"]
 name="modelskill"
 version="1.0.dev24"
 dependencies = [
-    "numpy",
-    "pandas",
+    "numpy  >= 1.20.0",
+    "pandas >= 1.4",
     "mikeio >= 1.2",
     "matplotlib",
     "xarray",
diff --git a/requirements_min.txt b/requirements_min.txt
new file mode 100644
index 00000000..13303fb1
--- /dev/null
+++ b/requirements_min.txt
@@ -0,0 +1,3 @@
+numpy==1.20.0
+pandas==1.4.0
+xarray==2022.3.0
\ No newline at end of file
diff --git a/tests/model/test_dfsu.py b/tests/model/test_dfsu.py
index 9740e5b5..996a9642 100644
--- a/tests/model/test_dfsu.py
+++ b/tests/model/test_dfsu.py
@@ -1,4 +1,3 @@
-import numpy as np
 import pytest
 
 import mikeio
@@ -208,15 +207,6 @@ def test_dfsu_extract_point(sw_dutch_coast, Hm0_EPL):
     mr_extr_2 = mr2.extract(Hm0_EPL.copy())
 
     assert list(mr_extr_1.data.data_vars) == list(mr_extr_2.data.data_vars)
-    assert np.all(mr_extr_1.data == mr_extr_2.data)
-
-    c1 = mr1.extract(Hm0_EPL.copy())
-    c2 = mr2.extract(Hm0_EPL.copy())
-    assert isinstance(c1, ms.PointModelResult)
-    assert isinstance(c2, ms.PointModelResult)
-    assert np.all(c1.data == c2.data)
-    # c1.observation.itemInfo == Hm0_EPL.itemInfo
-    # assert len(c1.observation.data.index.difference(Hm0_EPL.data.index)) == 0
 
 
 def test_dfsu_extract_point_aux(sw_dutch_coast, Hm0_EPL):
@@ -243,15 +233,6 @@ def test_dfsu_extract_track(sw_dutch_coast, Hm0_C2):
     ds2 = mr_track2.data
 
     assert list(ds1.data_vars) == list(ds2.data_vars)
-    assert np.all(ds1 == ds2)
-
-    c1 = mr1.extract(Hm0_C2.copy())
-    c2 = mr2.extract(Hm0_C2.copy())
-    assert isinstance(c1, ms.TrackModelResult)
-    assert isinstance(c2, ms.TrackModelResult)
-    assert np.all(c1.data == c2.data)
-    # c1.observation.itemInfo == Hm0_C2.itemInfo
-    # assert len(c1.observation.data.index.difference(Hm0_C2.data.index)) == 0
 
 
 def test_dfsu_extract_track_aux(sw_dutch_coast, Hm0_C2):
diff --git a/tests/test_aggregated_skill.py b/tests/test_aggregated_skill.py
index d2ff7be7..dfcafa9f 100644
--- a/tests/test_aggregated_skill.py
+++ b/tests/test_aggregated_skill.py
@@ -163,6 +163,7 @@ def test_skill_mm_swaplevel(cc2):
     assert np.all(sk2.index.levels[0] == sk.index.levels[1])
 
 
+@pytest.mark.skipif(pd.__version__ < "2.0.0", reason="requires newer pandas")
 def test_skill_mm_sort_index(cc2):
     sk = cc2.skill(metrics=["rmse", "bias"])
     assert list(sk.index.get_level_values(1)) == [
@@ -321,6 +322,7 @@ def test_skill_plot_grid(cc2):
     assert "only possible for MultiIndex" in str(wn[0].message)
 
 
+@pytest.mark.skipif(pd.__version__ < "1.5.0", reason="requires Pandas 1.5.0 or higher")
 def test_skill_style(cc2):
     sk = cc2.skill(metrics=["bias", "rmse", "lin_slope", "si"])
     sk.style()
@@ -331,6 +333,7 @@ def test_skill_style(cc2):
     sk.style(cmap="viridis_r", show_best=False)
 
 
+@pytest.mark.skipif(pd.__version__ < "1.5.0", reason="requires Pandas 1.5.0 or higher")
 def test_styled_skill_can_be_rendered(cc2):
     sk = cc2.skill()
     # _repr_html_ is called by Jupyter
diff --git a/tests/test_comparer.py b/tests/test_comparer.py
index ae8f15be..9b01f9a9 100644
--- a/tests/test_comparer.py
+++ b/tests/test_comparer.py
@@ -634,6 +634,7 @@ def test_skill_dt(pc):
     assert list(sk.data.index.levels[1]) == [1, 2, 3, 4, 5]  # Tuesday to Saturday
 
 
+@pytest.mark.skipif(pd.__version__ < "2.0.0", reason="requires newer pandas")
 def test_skill_freq(pc):
     assert pc.time.freq == "D"
 
diff --git a/tests/test_grid_skill.py b/tests/test_grid_skill.py
index 3747333a..5d41e416 100644
--- a/tests/test_grid_skill.py
+++ b/tests/test_grid_skill.py
@@ -80,6 +80,7 @@ def test_gridded_skill_sel_model(cc2) -> None:
         gs.sel(model="bad_model")
 
 
+@pytest.mark.skipif(pd.__version__ < "2.0.0", reason="requires newer pandas")
 def test_gridded_skill_is_subsettable(cc2) -> None:
     gs = cc2.gridded_skill(bins=3, metrics=["rmse", "bias"])
     gs.data.rmse.sel(x=2, y=53.5, method="nearest").values == pytest.approx(0.10411702)