How to avoid "blobs" in produced velocity map at the seemingly disconnected areas ?

[Shima-shoki]

Hi, thank you very much for the great project as always.
I have been able to run dolphin for a while, and most of the cases the results are quire satisfactly.

However, there are notable "blob" in the produced velocity maps, especially the disconnected areas from the nearby lands.
For example, the attached image shows the Chiba prefecture, Japan, which is known to be affected by ground subsidence which we can see in the middle as an ellipsoidal negative los velocity.

On the other hand, it is notable that there is an almost uniform "positive" los velocity area in the top-left, which would be unlikely to occur. I noticed that this phenomena looks happening if the area is disconnected from the nearby areas.

Are there any good options in dolphin to circumbent this issue?
I attached the sweet + dolphin options to run the processings below.

import subprocess
from pathlib import Path
repo_dir = Path("/content/sweets")

bl=35.6104293826326, 140.2
tr=35.79940065664542, 140.54242516810402

bbox = [bl[1], bl[0], tr[1], tr[0]]
start = "2023-01-01"
end = "2024-01-01"

region='Chiba'

work_dir = repo_dir / f"{region}_demo"
out_dir = repo_dir / f"{region}_data"
config_path = work_dir / "sweets_config.yaml"

work_dir.mkdir(exist_ok=True)
out_dir.mkdir(exist_ok=True)

cmd_config = [
    "pixi", "run", "sweets", "config",
    "--bbox", str(bbox[0]), str(bbox[1]), str(bbox[2]), str(bbox[3]),
    "--start", start,
    "--end", end,
    "--track", "46",
    "--swaths", "IW1","IW2",
    "--out-dir", str(out_dir),
    "--work-dir", str(work_dir),
    "--output", str(config_path),

    "--no-gpu-enabled",
    "--n-workers", "20",
    "--threads-per-worker", "10",

    "--dolphin.strides", "3", "6",
    "--dolphin.n-parallel-bursts", "20",
    "--dolphin.n-parallel-unwrap", "20",
]

subprocess.run(cmd_config, check=True, cwd=repo_dir)

!pixi run sweets run Chiba_demo/sweets_config.yaml

We had a similar issue in Niigata, Japan, where a clear uplift was estimated by dolphin, even though the area is known for long-term land subsidence.

Thanks in advance!

[scottstanie]

ah, put in a sweets PR for

dolphin/src/dolphin/workflows/config/_unwrap_options.py

Lines 305 to 308 in 5edd4e7

	run_interpolation: bool = Field(
	False,
	description="Whether to run interpolation step on wrapped interferogram.",
	)

. We should expose it there.

Interpolation usually works to smooth across small rivers like that.
see disp-s1's choices for this: https://github.com/opera-adt/disp-s1/blob/main/configs/algorithm_parameters_forward_20260205.yaml#L74-L76 , https://github.com/opera-adt/disp-s1/blob/main/configs/algorithm_parameters_forward_20260205.yaml#L91-L101

[Shima-shoki]

Thank you very much as always Dr. @scottstanie ! I'll try the suggested method to resolve the issue.