Persistent horizontal/diagonal NoData gap along Sentinel-1 burst boundaries across water body in dolphin workflow

[oguzhannysr]

Hi @scottstanie ,
I am running a Sentinel-1 InSAR time-series stack analysis over the Izmir/Karaburun region (Turkey) using sweets and dolphin (v0.26.0). My Area of Interest (AOI) covers a coastal area heavily intersected by a major water body (Aegean Sea / Izmir Gulf).

I am experiencing a persistent, sharp, diagonal/horizontal NoData gap that perfectly aligns with the Sentinel-1 burst boundaries (between bursts 279457, 279458, and 279459 of Track 131, IW1).

Initially, I suspected a SNAPHU tiling/stitching artifact. To eliminate this, I explicitly disabled SNAPHU spatial tiling by setting ntiles: [1, 1] and n_parallel_tiles: 1. De-ramping was kept at default. However, even with single-tile unwrapping successfully executed, the exact same gap persists in the final velocity map (see attached image).

Configuration & Log Analysis
Looking at my dolphin.log and dolphin_config.yaml, the phase linking and single-tile unwrapping steps complete without any explicit errors. However, I noticed that run_burst_align is currently set to false.

Here is the snippet of the parameters used during the run:

YAML
unwrap_options:
run_unwrap: true
unwrap_method: snaphu
run_burst_align: false
burst_align_planar_ramp: false
snaphu_options:
ntiles: [1, 1]
n_parallel_tiles: 1
init_method: mcf
cost: smooth
single_tile_reoptimize: true
From the logs, stitching takes place successfully across 3 images per date, but it seems to permanently drop the overlap pixels right where the land transitions into the low-coherence water body:

Plaintext
Stitching interferograms by date.
(datetime.datetime(2022, 12, 31, 0, 0), datetime.datetime(2023, 1, 12, 0, 0)): Stitching 3 images.
Questions
Could this issue be resolved by enabling run_burst_align: true and burst_align_planar_ramp: true in the configuration to better align the phase ramps across the water boundary before stitching?

Alternatively, does this indicate a known Sentinel-1 burst synchronization/timing anomaly (similar to ESA Quality Disclaimer #273) on a specific date in the stack that forces the workflow to discard this overlapping valid region entirely for the whole stack?

How can I force dolphin or isce3 to preserve or interpolate this overlap region across coastlines during the burst stitching phase?

Any guidance or workarounds would be greatly appreciated.

dolphin_.log
izmir_menemensol_config.yaml
dolphin_config.yaml

[scottstanie]

Hi, I agree that looks like another bust quality issue for a day at two... if it doesn't include that May 1 2025 QA time, perhaps you could check the interferogram or correlation files themselves, maybe make a bunch of tiny plots to quickly scan? I think finding and removing the one or two bad dates will fix this. That will be better than interpolating. And the ramp align option is for when there is overlap (which is not the problem here)