[Feature] Process N neareast neighbor pairs with On Demand Sentinel-1 Processing in Vertex

[alhandwerger]

The current On Demand Sentinel-1 Processing in Vertex workflow supports seasonal filtering, which is extremely helpful to exclude snow-covered acquisitions (or other seasonal challenges). However, after seasonal filtering, acquisitions are often too sparsely distributed in time to form a connected interferogram network using only temporal baseline constraints. This limits SBAS time-series inversions in alpine and high-latitude regions (or other seasonally challenged areas). A current work around is to add custom pairs that span the snow-free to snow-free period but this can be simplified with the following request:

Request
Add an optional nearest-neighbor interferogram pairing mode to the SBAS workflow (e.g., limit each acquisition to N nearest temporal neighbors after seasonal filtering). This would allow interferograms to span from snow-free to snow-free periods across winter gaps while still avoiding snow-covered scenes.

Ive attached an example of what the end result could look like.

[Alex-Lewandowski]

Hi @alhandwerger, We have been working on adding functionality to [asf_search](https://github.com/asfadmin/Discovery-asf_search/tree/feature/stacking/asf_search) that will support this. A Vertex update would build off that. We have a Pair class in there already. I am working on a Stack class PR this sprint. Then, an SBASNetwork class will follow that supports creating connected, seasonal SBAS stacks without the need for manually creating bridge pairs. This will also support Sentinel-1 multi-burst stacks.

These updates are based on some code I wrote last year. The current round of development will make some changes to generalize the Pair and Stack classes so that they are useful beyond SBAS, but there is code living in a dev branch that does what you are after, here here, and it includes an example notebook.

[asjohnston-asf]

@ccfleming417 any insight into whether time will be invested adding features to the SBAS tool this year?

[alhandwerger]

hi @Alex-Lewandowski , thank you for the info and your response. The link you provided did not work, can you please send again?

Thanks!

[Alex-Lewandowski]

@alhandwerger Sorry about that! Here you go: asfadmin/Discovery-asf_search#332

[asjohnston-asf]

For ARIA_S1_GUNW jobs in particular, the asf_enumeration python library provides support for building SBAS stacks. Here's a short example of submitting nearest-neighbor jobs for a specific ARIA frame ID:

import hyp3_sdk
from asf_enumeration import aria_s1_gunw

frame_id = 27236
acquisitions = aria_s1_gunw.get_acquisitions(frame=frame_id)
# add any desired filtering here, e.g. removing acquisitions from snow-covered months

nearest_neighbor_jobs = [
    {
        'frame_id': frame_id,
        'reference_date': pair[1].date.strftime('%Y-%m-%d'),
        'secondary_date': pair[0].date.strftime('%Y-%m-%d'),
    } for pair in zip(acquisitions, acquisitions[1:])
]

hyp3 = hyp3_sdk.HyP3()
for job in nearest_neighbor_jobs:
    hyp3.submit_aria_s1_gunw_job(**job)

The asf_search work that @Alex-Lewandowski pointed to is intended to be a more general-purpose solution supporting all Sentinel-1 InSAR workflows, ARIA or otherwise.