New Open-Source Tool Brings Order to Satellite Radar Data Before the Download Even Starts

Satellite radar interferometry has quietly become one of the most powerful ways to watch the ground move beneath our feet. By comparing radar images of the same patch of Earth taken at different times, scientists can measure millimetre-scale shifts caused by earthquakes, volcanoes, sinking cities, landslides and thawing permafrost. Europe’s Sentinel-1 satellites feed a vast public archive of such data, and an automated processing system called LiCSAR turns it into ready-made interferograms and coherence maps that anyone can download through the COMET-LiCS portal. But a new open-source software package, described in the journal SoftwareX, argues that having access to the archive is not the same as having a usable dataset — and it sets out to close that gap with a tool called LiCSAR-WinFetch.

Developed by Qingsong Du and released under the GNU General Public License, LiCSAR-WinFetch version 1.3.0 is a Python workflow that handles the unglamorous but critical stage between the public archive and the time-series analysis packages researchers actually use. It discovers and resolves remote products, downloads a selected single-frame dataset, organises the results into the exact directory hierarchy expected by LiCSBAS — the widely adopted open-source package for quality control, displacement time-series estimation and velocity mapping — and validates the structure of what has arrived on disk. Crucially, it also does something most downloaders do not: it can plan a wide-area study entirely from inventory metadata before a single large raster file is transferred.

The motivation is practical. A single LiCSAR frame can contain hundreds of acquisition dates and thousands of interferogram pairs, and a browser-based retrieval session or a minimal script offers little protection against interrupted connections, partially written files, repeated transfers and uncertainty about which missing objects are actually required. For a time-series analysis to work, every interferogram pair needs both its unwrapped-phase raster and its coherence raster; if either is absent, the pair is structurally incomplete. An optional atmospheric correction product, by contrast, should be reported as unavailable without being misclassified as a failed required pair. LiCSAR-WinFetch keeps these product roles strictly separate, ending each file in an explicit state: downloaded, already complete, planned, unavailable or failed.

The tool also confronts a messier reality of public archives: they change. The COMET-LiCS portal reported a storage-system migration in October 2025 and advised users to update their download code. During such transitions, an expected file may appear as a direct object, a redirected target, or a small HTML or plain-text manifest containing the real object URL. A naive client that treats every entry as an ordinary directory can falsely classify available data as missing. LiCSAR-WinFetch’s resolver tests the expected URL and supported storage variants, resolves HTML anchors relative to the final response location, searches plain-text manifests for absolute targets, and de-duplicates candidates before any transfer begins. Its behaviour under these cases is pinned down by a fifteen-test deterministic suite, all of which passed for version 1.3.0.

Download integrity receives equally careful treatment. Transfers run through a configurable thread pool, with bytes first written to a temporary .part file; the final filename appears only after a size check against the remote object succeeds. If a partial file exists, the client sends an HTTP Range request to resume from the local byte count, and if a server ignores that request and returns a complete object, the temporary file is reopened cleanly rather than corrupted by incompatible appending. Every run produces a machine-readable CSV report of file-level outcomes, a text summary and a log — records that can be archived alongside a study so that a later investigator can distinguish a changed remote object from an interrupted local transfer or a deliberately excluded optional product.

The most scientifically interesting component, however, is the multi-frame temporal planner. Wide-area studies often span many adjacent radar frames, and those frames may cover different date ranges. If neighbouring frames sample different seasonal windows or event intervals, a nominally common analysis can silently compare incomparable things. The planner scans the remote inventories of an entire frame group and computes the maximum common calendar interval shared by all frames, the exact intersection of acquisition dates and interferogram-pair networks within a same-orbit group, and a retention statistic describing how much of the smaller pair network survives strict common-pair selection. It then evaluates four temporal-selection modes of increasing strictness — independent, common-period, common-epochs and exact common-pairs — and reports the graph-theoretic consequences of each.

A demonstration on 22 LiCSAR frames covering the Qilian Mountains — eleven ascending and eleven descending frames in twelve relative-orbit groups — shows why this planning layer matters. All 22 inventory scans completed without downloading any imagery, revealing a 22-frame maximum common calendar interval from November 2014 to September 2023. Across the ten two-frame groups, exact common-pair retention ranged from 71.4 percent to 97.1 percent, and nine of the resulting networks stayed connected. But the 106D group delivered a cautionary result: it retained 94.0 percent of the smaller within-period pair network — 1,970 of 2,096 pairs — yet the strict filtering fragmented the graph into four connected components and left three common acquisition dates isolated. Two of those dates were already isolated in the shorter frame’s independent network; the third, August 2023, became isolated only after pair intersection. A high retention percentage, in other words, can coexist with hidden fragmentation that would only surface after terabytes of data had been downloaded.

The package’s interoperability claims were tested against the maintained LiCSBAS step 01 downloader on a bounded benchmark: frame 124D_04854_171313 over four acquisition epochs in September and October 2016, yielding six interferogram pairs. Both tools produced all 20 expected output paths, and 19 of the 20 files were byte-identical by SHA-256 hashing, including all 12 interferogram products. The single difference was the representative multilook-intensity raster, traced to a differing source-epoch selection rule rather than an incomplete transfer. The LiCSAR-WinFetch output then passed through LiCSBAS’s step 02 preprocessing script under Ubuntu 22.04 on the Windows Subsystem for Linux without any manual renaming, processing all six interferograms and returning exit code zero. Cross-platform continuous integration on Windows, Ubuntu and macOS under Python 3.10 and 3.12 passed all six jobs, confirming that the tool, though born on Windows desktops, is genuinely platform-neutral.

The authors are careful about scope. LiCSAR-WinFetch does not generate interferograms, assess phase-unwrapping quality, estimate deformation time series, harmonise cross-frame references, mosaic results, model the atmosphere or perform line-of-sight decomposition. A structurally complete directory certifies only that the requested products are present and organised, not that the underlying geophysical measurements are sound. The validation is bounded to the tested archive representations, frame selections and LiCSBAS step 02, and says nothing about transfer speed or future archive compatibility. Yet within those limits, the tool changes the order of operations for wide-area Sentinel-1 studies: researchers can now check calendar overlap, network connectivity and pair retention before committing storage and processing resources, and archive the machine-readable evidence of every decision. For studies of tectonic deformation, volcanic unrest, subsidence and infrastructure monitoring, that makes the path from remote inventory to analysis-ready input explicit, inspectable and reproducible — a quiet but meaningful advance in how satellite radar science gets done.

Subject of Research: Open-source software for reproducible acquisition, temporal planning and LiCSBAS-ready organisation of Sentinel-1 InSAR products from the COMET-LiCS portal

Article Title: LiCSAR-WinFetch: Reproducible LiCSAR acquisition, multi-frame temporal planning, and LiCSBAS-ready organisation of LiCSAR products

Article References: Du, Q. (2026). LiCSAR-WinFetch: Reproducible LiCSAR acquisition, multi-frame temporal planning, and LiCSBAS-ready organisation of LiCSAR products. SoftwareX, 36, Article 103119. https://doi.org/10.1016/j.softx.2026.103119

Image Credits: AI Generated

DOI: 10.1016/j.softx.2026.103119

Keywords: InSAR, LiCSAR, LiCSBAS, Sentinel-1, LiCSAR-WinFetch, open-source software, radar interferometry, ground deformation, reproducibility, temporal planning, COMET-LiCS portal, FAIR data