(CF Compliance) `valid_min` and `valid_max` Attribute Dtype Mismatch

Description: The valid_min and valid_max attributes have different data types than the datasets they describe. This is a related (but separate) issue from #297 .

CF Requirement (Section 2.5.1):

"The attributes valid_min, valid_max and valid_range ... are all of the same type as the variable to which they are attached."

— CF Conventions Version 1.7, Section 2.5.1: Missing Data

Source: http://cfconventions.org/Data/cf-conventions/cf-conventions-1.7/cf-conventions.html#_missing_data

Note: This issue primarily impacts RSLC and the five InSAR. (GSLC and GCOV resolved this issue via #213 ).

Impact:

• Inconsistent behavior across different CF-compliant software
• Tools may reject values that should be valid or accept invalid values

Fix Required: Ensure valid_min and valid_max use the same dtype as their parent dataset.

Common Patterns:

1. float32 datasets with float64 valid_min/max (most common)

   • Geolocation/radarGrid metadata cubes
   • Affects: RSLC, RIFG, RUNW, ROFF, GUNW, GOFF

2. uint32 datasets with int64 valid_min (no valid_max)

   • Mask datasets
   • Affects: RIFG, RUNW, ROFF, GUNW, GOFF

Occurrences by Product:

RSLC

File: NISAR_L1_PR_RSLC_001_149_D_080_2005_QPDH_A_20251003T130952_20251003T131016_X05016_N_P_J_001.h5

Dataset: /science/LSAR/RSLC/metadata/geolocationGrid/alongTrackUnitVectorX
  Dataset dtype: float32
  valid_min (actual): -1.0 (dtype: float64)
  valid_min (should be): np.float32(-1.0)
  valid_max (actual): 1.0 (dtype: float64)
  valid_max (should be): np.float32(1.0)

Other affected datasets (same pattern):
  - /science/LSAR/RSLC/metadata/geolocationGrid/alongTrackUnitVectorY
  - /science/LSAR/RSLC/metadata/geolocationGrid/elevationAngle (valid_min: 0.0, valid_max: 90.0)
  - /science/LSAR/RSLC/metadata/geolocationGrid/incidenceAngle (valid_min: 0.0, valid_max: 90.0)
  - /science/LSAR/RSLC/metadata/geolocationGrid/losUnitVectorX
  - /science/LSAR/RSLC/metadata/geolocationGrid/losUnitVectorY

RIFG

File: NISAR_L1_PR_RIFG_013_118_A_171_014_2000_SH_20260222T102514_20260222T102520_20260306T102515_20260306T102521_X05016_P_P_J_001.h5

Pattern: /science/LSAR/RIFG/metadata/geolocationGrid/*
  - alongTrackUnitVectorX, alongTrackUnitVectorY (float32 dataset, float64 valid_min/max)
  - elevationAngle, incidenceAngle (float32 dataset, float64 valid_min/max)
  - losUnitVectorX, losUnitVectorY (float32 dataset, float64 valid_min/max)

Additional valid_min only (no valid_max):
  - /science/LSAR/RIFG/grids/frequencyA/pixelOffsets/mask (uint32 dataset, int64 valid_min)
  - /science/LSAR/RIFG/grids/frequencyA/unwrappedPhase/mask (uint32 dataset, int64 valid_min)

RUNW

File: NISAR_L1_PR_RUNW_007_106_A_140_008_7700_SH_20251211T140849_20251211T140926_20251223T140850_20251223T140926_X05016_N_P_J_001.h5

Same pattern as RIFG:
Pattern: /science/LSAR/RUNW/metadata/geolocationGrid/* (6 datasets with both valid_min and valid_max)

Additional valid_min only:
  - /science/LSAR/RUNW/grids/frequencyA/pixelOffsets/mask (uint32 → int64)
  - /science/LSAR/RUNW/grids/frequencyA/unwrappedPhase/mask (uint32 → int64)

ROFF

File: NISAR_L1_PR_ROFF_007_106_A_140_008_7700_SH_20251211T140849_20251211T140926_20251223T140850_20251223T140926_X05016_N_P_J_001.h5

Pattern: /science/LSAR/ROFF/metadata/geolocationGrid/* (6 datasets with both valid_min and valid_max)
  - alongTrackUnitVectorX/Y, elevationAngle, incidenceAngle, losUnitVectorX/Y
  - All: float32 dataset with float64 valid_min/max

Additional valid_min only:
  - /science/LSAR/ROFF/grids/frequencyA/pixelOffsets/mask (uint32 → int64)

GUNW

File: NISAR_L2_PR_GUNW_001_062_A_019_006_2000_SH_20250927T114508_20250927T114542_20251126T114516_20251126T114545_X05016_P_F_J_001.h5

Pattern: /science/LSAR/GUNW/metadata/radarGrid/* (6 datasets with both valid_min and valid_max)
  - alongTrackUnitVectorX/Y, elevationAngle, incidenceAngle, losUnitVectorX/Y
  - All: float32 dataset with float64 valid_min/max

Additional valid_min only (uint32 datasets with int64 valid_min):
  - /science/LSAR/GUNW/grids/frequencyA/pixelOffsets/mask
  - /science/LSAR/GUNW/grids/frequencyA/unwrappedInterferogram/mask
  - /science/LSAR/GUNW/grids/frequencyA/wrappedInterferogram/mask
  (Similar for frequencyB)

GOFF

File: NISAR_L2_PR_GOFF_005_106_A_140_006_7700_SH_20251117T140848_20251117T140925_20251129T140849_20251129T140925_X05016_P_P_J_001.h5

Pattern: /science/LSAR/GOFF/metadata/radarGrid/* (6 datasets with both valid_min and valid_max)
  - alongTrackUnitVectorX/Y, elevationAngle, incidenceAngle, losUnitVectorX/Y
  - All: float32 dataset with float64 valid_min/max

Additional valid_min only:
  - /science/LSAR/GOFF/grids/frequencyA/pixelOffsets/mask (uint32 → int64)

GSLC and GCOV: CORRECT

GSLC and GCOV correctly uses matching dtypes for valid_min/max attributes.
Example from GSLC:

Dataset: /science/LSAR/GSLC/metadata/radarGrid/alongTrackUnitVectorX
  Dataset dtype: float32
  valid_min: -1.0 (dtype: float32) ✓ CORRECT
  valid_max: 1.0 (dtype: float32) ✓ CORRECT

cc: @hfattahi @bhawkins @xhuang-jpl @seongsujeong @gshiroma

The issue names sample HDF5 files and dataset paths but no source file or test; start by locating the RSLC, RIFG, RUNW, ROFF, GUNW, and GOFF product writers, then compare their attribute handling with GSLC/GCOV and #213. Check the affected geolocation/radarGrid and mask datasets. Done means valid_min and valid_max match each parent dataset's dtype across the listed products.