compare means with GUNW cubes/rasters representing similar items

tests/test_packaging.py

@@ -1,6 +1,8 @@
 import json
 from pathlib import Path
 
+import numpy as np
+import rasterio
 from numpy.testing import assert_almost_equal
 
 from isce2_topsapp.packaging import (
@@ -59,11 +61,41 @@ def test_mean_of_geocoded_isce_outputs():
         merged_dir="https://gunw-development-testing.s3.us-west-2.amazonaws.com/sample_merged_data"
     )
     expected_out = {
-        "mean_filtered_coherence_without_water_mask": 0.3126283,
-        "mean_filtered_coherence_with_water_mask": 0.32342613,
-        "mean_incidence_angle": 28.281992,
-        "mean_azimuth_angle": -99.18769836425781,
+        "mean_filtered_coherence_without_water_mask": 0.4395995,
+        "mean_filtered_coherence_with_water_mask": 0.4649538,
+        "mean_unfiltered_coherence_without_water_mask": 0.33125195,
+        "mean_unfiltered_coherence_with_water_mask": 0.33531728,
+        "mean_incidence_angle": 38.845047,
+        "mean_azimuth_angle": -169.84756469726562,
     }
 
     for key in expected_out:
         assert_almost_equal(expected_out[key], out[key], decimal=5)
+
+    # Use output GUNW and compare means
+    gunw_s3_path = (
+        "/vsis3/gunw-development-testing/sample_merged_data"
+        "/S1-GUNW-A-R-164-tops-20220212_20220131-222829-00071W_00047N-PP-3d6c-v3_0_0.nc"
+    )
+
+    vars_data_0 = {var: f"//science/grids/data/{var}" for var in ["unfilteredCoherence", "coherence"]}
+    vars_data_1 = {var: f"//science/grids/imagingGeometry/{var}" for var in ["azimuthAngle", "incidenceAngle"]}
+    vars_data = {**vars_data_0, **vars_data_1}
+
+    gunw_mean_data = {}
+    for var, var_path in vars_data.items():
+        with rasterio.open(f"HDF5:{gunw_s3_path}:{var_path}") as ds:
+            X = ds.read()
+            # 0 is the nodata for all the variables we consider
+            gunw_mean_data[var] = np.mean(X[X != 0])
+
+    # Ensure the GUNW cubes and rasters (with lower resolution than geocoded ISCE2 data) are within a reasonable range
+    # of the geocoded rasters that ICE2 produces
+    assert np.abs(out['mean_filtered_coherence_without_water_mask'] - gunw_mean_data['coherence']) < .01
+    assert np.abs(out['mean_unfiltered_coherence_without_water_mask'] - gunw_mean_data['unfilteredCoherence']) < .01
+    assert np.abs(out['mean_azimuth_angle'] - gunw_mean_data['azimuthAngle']) < 1
+    assert np.abs(out['mean_incidence_angle'] - gunw_mean_data['incidenceAngle']) < 1
+
+    # Sanity checks - make sure that applying a water mask *increases* the coherence as water has low coherence
+    assert out['mean_filtered_coherence_without_water_mask'] < out['mean_filtered_coherence_with_water_mask']
+    assert out['mean_unfiltered_coherence_without_water_mask'] < out['mean_unfiltered_coherence_with_water_mask']