Skip to content

Latest commit

 

History

History
38 lines (23 loc) · 1.95 KB

README.md

File metadata and controls

38 lines (23 loc) · 1.95 KB
Raw

HPC_isce_stack

Author: Oliver Stephenson, 2021

Scripts for submitting tops stack stages as a series of chained SLURM jobs Each job uses SLURM arrays to manage the processing that can be done in parallel Avoids the issue with wasting resources, hopefully deals better with very large numbers of jobs

The resources_array.cfg file gives the resources allocated to each array element, not to the whole job step combined

Different resource files are for different sized jobs - resources_array_full_eff.cfg is the efficient ('eff') allocation of resources for the full T115a track (25 to 32N)

Optional before running processing codes

(ykliu 2022-03-08)

  1. Run s1_select_ion.py to filter your data acquisitions (SLC zip files). Acquisitions not satisfying the requirements would be put into a separate not_used folders and they are not used by stackSentinel.py. The filering is based on the:

    • IPF versions
    • Frames with different IPF versions
    • gaps between frames
    • starting ranges
    • latitude coverage

  2. Use the appropriate .cfg for resourse allocation. Copy the customed .cfg to overwrite resources_array.cfg

    • run_16_unwrap set to 3 hr for long tracks (more than 6~7 latitude degrees)
    • run_20_unwrap_ion set to 3 hr for long tracks (more than 6~7 latitude degrees)
    • run_23_filtIon set memory usage to 30G (more than 6~7 latitude degrees)

  3. stack_sentinel_cmd.sh: the bbox -b corresponds to the -b argument in stackSentinel.py. It will process all frames that intersect with this bbox.

  4. Remove extra pairs by using -x option in stackSentienl.py.

  5. stack_sentinel_cmd.sh: add --polarization hh for Antartica data (HH polarization)

  6. analysis_time.py: If re-submitting jobs, go ahead and erase the redundant header rows in the log files time_unix.txt and timing.txt.