Dependency refresh, sunset 3.11

.github/workflows/publish.yaml

@@ -13,9 +13,9 @@ jobs:
     runs-on: ubuntu-latest
 
     steps:
-      - uses: actions/checkout@v4
+      - uses: actions/checkout@v6
 
-      - uses: actions/setup-python@v5
+      - uses: actions/setup-python@v7
         with:
           python-version: "3.x"
 
@@ -25,7 +25,7 @@ jobs:
           python -m build
 
       - name: Upload distributions
-        uses: actions/upload-artifact@v4
+        uses: actions/upload-artifact@v7
         with:
           name: release-dists
           path: dist/
@@ -48,7 +48,7 @@ jobs:
 
     steps:
       - name: Retrieve release distributions
-        uses: actions/download-artifact@v4
+        uses: actions/download-artifact@v8
         with:
           name: release-dists
           path: dist/
@@ -74,7 +74,7 @@ jobs:
 
     steps:
       - name: Retrieve release distributions
-        uses: actions/download-artifact@v4
+        uses: actions/download-artifact@v8
         with:
           name: release-dists
           path: dist/

.github/workflows/run-pre-commit.yaml

@@ -14,20 +14,22 @@ jobs:
     runs-on: ubuntu-latest
     strategy:
       matrix:
-        python-version: ["3.10", "3.11", "3.12", "3.13"]
+        python-version: ["3.12", "3.13", "3.14"]
     steps:
-      - uses: actions/checkout@v4
+      - uses: actions/checkout@v6
 
-      - name: Set up Python env with uv
-        uses: astral-sh/setup-uv@v4
+      - name: "Set up Python"
+        uses: actions/setup-python@v7
         with:
-          version: "0.5.6"
-          enable-cache: true
+          python-version: ${{ matrix.python-version }}
 
-      - name: "Set up Python"
-        uses: actions/setup-python@v5
+      - name: Set up Python env with uv
+        uses: astral-sh/setup-uv@v9.0.0
         with:
+          version: "0.12.3"
           python-version: ${{ matrix.python-version }}
+          enable-cache: true
+          cache-suffix: ${{ matrix.python-version }}
 
       - name: Install the project
         run: uv sync --all-extras --all-groups

.github/workflows/run-tests.yaml

@@ -14,20 +14,22 @@ jobs:
     runs-on: ubuntu-latest
     strategy:
       matrix:
-        python-version: ["3.10", "3.11", "3.12", "3.13"]
+        python-version: ["3.12", "3.13", "3.14"]
     steps:
-      - uses: actions/checkout@v4
+      - uses: actions/checkout@v6
 
-      - name: Set up Python env with uv
-        uses: astral-sh/setup-uv@v4
+      - name: "Set up Python"
+        uses: actions/setup-python@v7
         with:
-          version: "0.5.6"
-          enable-cache: true
+          python-version: ${{ matrix.python-version }}
 
-      - name: "Set up Python"
-        uses: actions/setup-python@v5
+      - name: Set up Python env with uv
+        uses: astral-sh/setup-uv@v9.0.0
         with:
+          version: "0.12.3"
           python-version: ${{ matrix.python-version }}
+          enable-cache: true
+          cache-suffix: ${{ matrix.python-version }}
 
       - name: Install the project
         run: uv sync --all-extras --all-groups

pyproject.toml

@@ -16,53 +16,55 @@ classifiers = [
     "License :: OSI Approved :: MIT License",
     "Operating System :: OS Independent",
     "Programming Language :: Python :: 3",
-    "Programming Language :: Python :: 3.11",
     "Programming Language :: Python :: 3.12",
     "Programming Language :: Python :: 3.13",
+    "Programming Language :: Python :: 3.14",
     "Topic :: Scientific/Engineering :: Atmospheric Science",
 ]
-requires-python = ">=3.11,<3.14"
+requires-python = ">=3.12,<3.15"
 dependencies = [
     "dacite>=1.8.1",
     "gcsfs>=2024.12.0",
     "geopandas>=1.0.1",
     "joblib>=1.4.2",
     "numpy>=2.2.0",
-    "pandas>=2.2.3",
+    "pandas>=2.2.3,<3",
+    "pooch>=1.8.2",
     "pyarrow>=19.0.1",
     "pyyaml>=6.0.2",
     "regionmask>=0.13.0",
     "s3fs>=2024.12.0",
     "scores>=2.4.0",
+    "shapely>=2.0",
     "tqdm>=4.67.1",
     "xarray[io]>=2025.7.1",
     "zarr>=3.0.9,<4",
-    "kerchunk[dev]",
+    "kerchunk>=0.2.10",
+    "fastparquet>=2024.11.0",
     "flox>=0.10.0",
     "click>=8.1.8",
     "polars>=1.32.1",
-    "eccodes==2.42.0",
-    "frozenlist==1.5.0",
-    "pyogrio==0.10.0",
     "scikit-image>=0.25.2",
-    "cartopy==0.24.1",
-    "numba>=0.63.1",
+    "numba>=0.66.0",
     "sparse>=0.17.0",
-    "icechunk>=1.1.14",
+    "icechunk>=2.1.2",
     "dask[complete]>=2025.1.0",
     "distributed>=2025.1.0",
     "lazy-loader>=0.4",
 ]
 
 [project.optional-dependencies]
 data-prep = [
+    # No cp314 wheels are published; installing on 3.14 would build from
+    # source and require system GEOS and PROJ.
+    "cartopy>=0.24.1; python_version < '3.14'",
     "fsspec>=2024.12.0",
     "matplotlib>=3.10.0",
     "obstore>=0.8.2",
     "scipy>=1.13",
     "seaborn>=0.13.2",
     "ujson>=5.10.0",
-    "virtualizarr==2.1.2",
+    "virtualizarr>=2.7.3",
     "dask[complete]>=2025.1.0",
     "distributed>=2025.1.0",
 ]
@@ -117,7 +119,6 @@ ewb = "extremeweatherbench.evaluate_cli:cli_runner"
 
 [tool.uv.sources]
 extremeweatherbench = { workspace = true }
-kerchunk = { git = "https://github.com/fsspec/kerchunk" }
 
 [tool.ruff]
 line-length = 88

src/extremeweatherbench/utils.py

@@ -8,10 +8,11 @@
 import pathlib
 from typing import Any, Callable, Literal, Optional, Sequence, Union
 
-import cartopy.io.shapereader as shpreader
+import geopandas as gpd
 import numpy as np
 import numpy.typing as npt
 import pandas as pd  # type: ignore[import-untyped]
+import pooch
 import regionmask
 import shapely
 import sparse
@@ -24,6 +25,13 @@
 
 logger = logging.getLogger(__name__)
 
+# Natural Earth vector data has been hosted on S3 since 2021; see
+# https://github.com/nvkelso/natural-earth-vector/issues/445
+NATURAL_EARTH_URL = (
+    "https://naturalearth.s3.amazonaws.com/{resolution}_{category}/"
+    "ne_{resolution}_{name}.zip"
+)
+
 operators = {
     ">": operator.gt,
     ">=": operator.ge,
@@ -767,6 +775,35 @@ def reduce_dataarray(
         raise TypeError(f"method must be str or callable, got {type(method)}")
 
 
+def load_natural_earth_geometries(
+    name: str,
+    resolution: Literal["10m", "50m", "110m"] = "50m",
+    category: str = "physical",
+) -> list[shapely.geometry.base.BaseGeometry]:
+    """Download and read a Natural Earth vector layer.
+
+    The zipped shapefile is cached on disk by pooch, so repeat calls within and
+    across sessions do not re-download it.
+
+    Args:
+        name: Natural Earth layer name, e.g. 'land', 'lakes', or 'ocean'.
+        resolution: Natural Earth resolution ('10m', '50m', or '110m').
+            Defaults to '50m'.
+        category: Natural Earth category. Defaults to 'physical'.
+
+    Returns:
+        The layer's geometries in EPSG:4326.
+    """
+    path = pooch.retrieve(
+        url=NATURAL_EARTH_URL.format(
+            resolution=resolution, category=category, name=name
+        ),
+        known_hash=None,
+        path=pooch.os_cache("extremeweatherbench"),
+    )
+    return list(gpd.read_file(f"zip://{path}").geometry)
+
+
 def load_land_geometry(
     resolution: Literal["10m", "50m", "110m"] = "50m",
 ) -> shapely.geometry.Polygon:
@@ -780,18 +817,12 @@ def load_land_geometry(
         The land geometry as a shapely Polygon with lakes and
         ocean-connected water bodies (bays, estuaries, seas) excluded.
     """
-    land = shpreader.natural_earth(
-        category="physical", name="land", resolution=resolution
-    )
-    land_geoms = list(shpreader.Reader(land).geometries())
+    land_geoms = load_natural_earth_geometries("land", resolution=resolution)
     land_union = shapely.ops.unary_union(land_geoms)
 
     # Exclude lakes to avoid false landfall detections
     try:
-        lakes = shpreader.natural_earth(
-            category="physical", name="lakes", resolution=resolution
-        )
-        lake_geoms = list(shpreader.Reader(lakes).geometries())
+        lake_geoms = load_natural_earth_geometries("lakes", resolution=resolution)
         if lake_geoms:
             lakes_union = shapely.ops.unary_union(lake_geoms)
             land_union = land_union.difference(lakes_union)
@@ -820,10 +851,7 @@ def load_ocean_geometry(
     Returns:
         The ocean geometry as a unified shapely Polygon.
     """
-    ocean = shpreader.natural_earth(
-        category="physical", name="ocean", resolution=resolution
-    )
-    ocean_geoms = list(shpreader.Reader(ocean).geometries())
+    ocean_geoms = load_natural_earth_geometries("ocean", resolution=resolution)
     return shapely.ops.unary_union(ocean_geoms)
 
 

tests/test_inputs.py

@@ -9,7 +9,7 @@
 import sparse
 import xarray as xr
 
-from extremeweatherbench import inputs
+from extremeweatherbench import inputs, utils
 
 
 class TestInputBase:
@@ -1252,7 +1252,9 @@ def test_ghcn_custom_convert_to_dataset_no_duplicates(self, sample_ghcn_datafram
 
         # Should have no NaN values if no duplicates were dropped
         original_count = len(clean_data)
-        result_count = result.surface_air_temperature.count().item()
+        result_count = utils.maybe_densify_dataarray(
+            result.surface_air_temperature.count()
+        ).item()
         assert result_count == original_count
 
     def test_ghcn_custom_convert_to_dataset_single_duplicate(
@@ -1282,7 +1284,9 @@ def test_ghcn_custom_convert_to_dataset_single_duplicate(
 
         # Should have dropped one duplicate, so count should equal original
         original_count = len(clean_data)
-        result_count = result.surface_air_temperature.count().item()
+        result_count = utils.maybe_densify_dataarray(
+            result.surface_air_temperature.count()
+        ).item()
         assert result_count == original_count
 
     def test_ghcn_custom_convert_to_dataset_many_duplicates(
@@ -1319,7 +1323,9 @@ def test_ghcn_custom_convert_to_dataset_many_duplicates(
 
         # Should have dropped all duplicates, so count should equal original
         original_count = len(clean_data)
-        result_count = result.surface_air_temperature.count().item()
+        result_count = utils.maybe_densify_dataarray(
+            result.surface_air_temperature.count()
+        ).item()
         assert result_count == original_count
 
     def test_ghcn_custom_convert_to_dataset_exception_handling(self):

tests/test_metrics.py

@@ -2484,8 +2484,8 @@ def side_effect(track_data, **kw):
                 assert coord_name in result.coords, (
                     f"Missing metadata coord: {coord_name}"
                 )
-            assert abs(float(result.forecast_landfall_latitude) - 16.1) < 0.01
-            assert abs(float(result.target_landfall_latitude) - 16.0) < 0.01
+            assert abs(result.forecast_landfall_latitude.item() - 16.1) < 0.01
+            assert abs(result.target_landfall_latitude.item() - 16.0) < 0.01
 
     def test_first_approach_skips_init_with_late_track_start(self):
         """Verify approach='first' filters out init_times whose
@@ -2757,7 +2757,7 @@ def side_effect(track_data, **kw):
                 f"Displacement {result.values[0]:.1f} km should be small (FL vs FL)"
             )
             # Verify metadata uses the Florida target
-            assert abs(float(result.target_landfall_latitude) - 29.5) < 0.01
+            assert abs(result.target_landfall_latitude.item() - 29.5) < 0.01
 
     def test_multi_landfall_displacement_next_approach(self):
         """Verify approach='next' also uses first forecast landfall
@@ -4802,9 +4802,11 @@ def _to_backend(forecast, target, backend):
 
     @staticmethod
     def _resolve(result):
-        """Compute dask results; pass through numpy/sparse."""
+        """Compute dask results and densify sparse ones; pass through numpy."""
         if hasattr(result, "compute"):
-            return result.compute()
+            result = result.compute()
+        if isinstance(result.data, sparse.COO):
+            result = result.copy(data=result.data.todense())
         return result
 
     @pytest.mark.parametrize("backend", ["numpy", "dask", "sparse"])