Hi there,
i am experiencing bugs when calling the thermal noise removal node with snap.
I tried to create a comprehensive markdown report below. I am happy for any help.
The report is here:
SNAP_ThermalNoiseRemoval_Bug_Report.pdf (540.2 KB)
also as pdf available (markdown rendering does not seem to be supported on this page)
# SNAP ThermalNoiseRemoval Operator Bug Report
## Summary
The SNAP `ThermalNoiseRemoval` operator produces severely corrupted output for certain Sentinel-1 GRD products acquired between late March and mid-April 2026. The `removeThermalNoise` parameter is ignored — setting it to `false` has no effect. Data corruption only ceases when the `ThermalNoiseRemoval` node is entirely removed from the processing graph.
## Environment
| Component | Version |
|—|—|
| SNAP Desktop / Engine | 13.0.0 |
| Java | 21.0.6 |
| ESA IPF (all affected products) | 004.03 |
## Affected Product
`S1A_IW_GRDH_1SDV_20260322T040426_20260322T040451_063735_08035B_7DD0`
- **Footprint:** `POLYGON ((32.99 46.81, 33.43 48.31, 29.97 48.71, 29.63 47.22, 32.99 46.81))` (approx. lat 46.8°–48.7° N, lon 29.6°–33.4° E)
Further images intersecting this footprint:
- At least 3 consecutive acquisition dates on orbit D138 are affected (2026-03-22, 2026-04-03, 2026-04-15). Most neighbouring dates (2026-02-02, 2026-02-14, 2026-03-10 and 2026-04-27) on the same orbit are healthy.
- 2026-02-26 is also corrupt, but with different signature (just skewed values. not further investigated here, root cause is probably the same.)
## Reproduction
**Minimum corrupt graph** (Read → TNR → Calibration → TC → dB → Write):
```xml
1.0
Read${source}
ThermalNoiseRemovaltrue
Calibrationtrue
Terrain-CorrectionSigma0_VH,Sigma0_VVCopernicus 30m Global DEM10.0EPSG:32636
LinearToFromdB
Write${output}ENVI
```
**Minimum working graph** — remove the `ThermalNoiseRemoval` node (Read → Calibration → TC → dB → Write):
```xml
1.0
Read${source}
Calibrationtrue
Terrain-CorrectionSigma0_VH,Sigma0_VVCopernicus 30m Global DEM10.0EPSG:32636
LinearToFromdB
Write${output}ENVI
```
## Visual Comparison
**With ThermalNoiseRemoval node (corrupt):**
**Without ThermalNoiseRemoval node (healthy):**
## Additional Notes
- The `removeThermalNoise=false` parameter is ignored — the operator produces identical (corrupted) output regardless of this flag.
- The raw GRD measurement data (uint16 DN) is healthy and uniform across all subswaths, confirming the corruption is introduced by SNAP.
- The artifact is subswath-dependent (IW1/IW2 affected, IW3 not), suggesting the operator reads corrupted noise annotation vectors from the SAFE product.