Update pyproj >=3.6.1

This newer versions of pyproj have a 1nm difference in the calculated elevation
output, which is not large enough to matter in transformations based on
parameters accurate to the mm level

Author: trey-stafford

pyproject.toml

@@ -29,7 +29,7 @@ dependencies = [
   "pandas >=2.2",
   "h5py >=3.11",
   "gps-timemachine >=1.1.4",
-  "pyproj ~=3.6.1",
+  "pyproj >=3.6.1",
   "shapely >=2.0.5",
   "pandera[mypy] ~=0.23.0",
   "pydantic >=2.8.2",

tests/unit/test_itrf_converter.py

@@ -40,7 +40,12 @@ def test_transform_itrf():
     # https://github.com/nsidc/NSIDC-Data-Tutorials/blob/a35203f18841456d258fac3a11dc04f44f839d9d/notebooks/iceflow/corrections.ipynb
     assert result.latitude.to_numpy()[0] == 69.99999974953995
     assert result.longitude.to_numpy()[0] == -50.0000001319163
-    assert result.elevation.to_numpy()[0] == 1.0052761882543564
+    # Rounding to 8 decimals in a field assumed to be measured in meters is 10nm accuracy!
+    # This rounding became necessary when pyproj updated to a newer than 3.6.1
+    # version, leading to a difference of ~1nm in the resulting calculation.
+    assert np.round(result.elevation.to_numpy()[0], 8) == np.round(
+        1.0052761882543564, 8
+    )
 
 
 def _itrf_cart_coords_to_wgs84(x, y, z):