Spectral Preprocessing

1. Scope & Processing Contract

Spectral Preprocessing converts sensor-encoded optical imagery into a physically meaningful radiometric domain and, when appropriate, Surface Reflectance. The current implementation is sensor/product aware: the raster is inspected before a calibration path is selected.

2. Sensor & Product Detection

The application offers Auto plus a manual Sensor / Product Type selection. Auto detection combines project/layer context, raster metadata/tags, band metadata and source/product naming. Explicit semantic metadata such as “surface reflectance”, “TOA reflectance” or “radiance” takes priority over filename heuristics.

3. Radiometric Calibration — 6 Choices

3.1 Generic scale and offset

3.2 DN to TOA radiance

3.3 DN to TOA reflectance

3.4 Radiance to TOA reflectance

4. Atmospheric Correction — 6 Choices

4.1 Dark Object Subtraction

4.2 6S implementation

For each selected band, the implementation resolves a center wavelength, runs 6S, retrieves coefficients x_a, x_b, x_c, and applies:

Conditional parameters include atmosphere profile, aerosol profile, AOT550, target altitude, solar/view zenith and azimuth overrides, and acquisition month/day.

4.3 Empirical Line

The current implementation uses mean observed values inside the two ROIs and requires at least three valid pixels in each ROI for every selected band.

5. Level-2 / Surface Reflectance Safety Lock

Sentinel-2 L2A
Landsat L2SP / L2SR
MOD09 / MYD09
EMIT L2A RFL
ASTER SR-type product
        ↓
Decode scale/offset if needed
        ↓
Atmospheric = Preserve 🔒

Landsat L1
Sentinel-2 L1C
MODIS L1B
ASTER L1T
EMIT L1B Radiance
        ↓
Radiometric path
        ↓
TOA / Radiance
        ↓
Atmospheric method when compatible

6. Application Parameters

7. Output & Provenance

The output is a Float32 Cloud Optimized GeoTIFF using the source geospatial structure. The analysis metadata records both requested and effective methods so that automatic routing and safety locks remain auditable.

{
  "product": {
    "sensor_profile": "landsat_oli",
    "processing_level": "L2SP/L2SR",
    "reflectance_state": "surface_reflectance",
    "detection_confidence": "high",
    "preprocessing_locked": true
  },
  "stages": [
    {"stage":"radiometric_calibration","effective_method":"metadata_scale_offset"},
    {"stage":"atmospheric_correction","effective_method":"preserve_existing_surface_reflectance"}
  ],
  "final_radiometric_state": "surface_reflectance"
}

8. Recommended Workflows

Input L2 SR
↓
Auto detection
↓
Scale/decode
↓
Preserve SR
↓
Analysis

Input L1 DN
↓
DN → TOA Reflectance
↓
DOS or 6S
↓
Surface Reflectance

Input physical image
+
Dark/Bright ROIs
↓
Empirical Line
↓
Surface Reflectance

For aquatic workflows, atmospheric/radiometric preprocessing comes before Hydro Preprocessing. For cross-sensor time series, Surface Reflectance comes before Spectral Harmonization.

9. Quality Control

• Confirm the detected product level and radiometric state before processing.
• Never force atmospheric correction on an already corrected Level-2 SR product.
• Do not use Metadata Scale & Offset unless a non-trivial scale/offset exists.
• Radiance→TOA Reflectance requires valid solar irradiance and solar geometry.
• 6S requires wavelength, solar/view geometry, date and an operational Py6S/6S installation.
• For Empirical Line, use homogeneous reference targets with independently known reflectance and adequate valid pixels.
• Inspect negative or >1 reflectance values scientifically; do not automatically clip them unless the downstream method requires it.
• Keep provenance when using the output for hyperspectral matching, harmonization, water analysis or ML.

10. Scientific & Technical References

1. U.S. Geological Survey. Using the USGS Landsat Level-1 Data Product — TOA radiance and reflectance equations. USGS.
2. U.S. Geological Survey. Landsat Collection 2 Surface Reflectance — Level-2 SR and scaling. USGS.
3. NASA MODIS Land Team. MODIS Surface Reflectance (MOD09/MYD09). NASA.
4. NASA Earthdata. EMIT L2A Estimated Surface Reflectance and Uncertainty and Masks. NASA Earthdata.
5. Vermote, E. F. et al. 6S/6SV radiative-transfer heritage for atmospheric correction; MODIS Land validation resources summarize use of 6SV. NASA MODIS Land.
6. Smith, G. M. & Milton, E. J. (1999). The use of the empirical line method to calibrate remotely sensed data to reflectance. International Journal of Remote Sensing 20(13), 2653–2662. DOI: 10.1080/014311699211994.