The InSAR post-processing workflow

InSAR processing is usually described as if it ended when the software exports a velocity map. In engineering practice it ends when a report is accepted — and between those two points lies a chain of conversions, checks and judgement calls that the processing tutorials never cover. This page walks the full chain in eight stages, names the quality gate at each stage, and maps every stage to the tool that performs it. Run the stages in order for a first campaign; afterwards the tools stand alone whenever a specific question arises.

Stage 0 — Site pre-check (before anything else)

Tool: Site Suitability Pre-Checker. Processing a site that physics will not support — dense forest under C-band, open water, slopes steeper than the incidence angle — wastes a week and produces a confidently useless map. The pre-check scores expected coherence, layover/shadow exposure and atmospheric burden from four honest inputs, raises red flags with their physical mechanisms, and recommends sensor, track plan and algorithm. Quality gate: if the score is Poor with unfixable mechanisms, the deliverable should be a documented negative pre-check, not a forced map.

Stage 1 — Data acquisition & processing

Tool: your InSAR processor — MintPy, LiCSBAS, HyP3 or SNAP. This workbench deliberately does not compete with the processors; it starts where they stop. Acquire SLC pairs, run your SBAS or PS chain, and export velocities, per-point geometry (incidence angle, LOS azimuth) and time series. Quality gate: export both the velocity layer and the geometry files — the conversion stage needs the geometry, and retrofitting it later is painful.

Stage 2 — Coordinate preparation

Tool: Batch Coordinate Converter. Survey data arrives in UTM; satellite results live in longitude and latitude. Convert whole tables in either direction, keeping one zone and one datum per file, and verify the result on the plan-view scatter before merging anything. Quality gate: a single wrong zone silently relocates the site — check the scatter, every time.

Stage 3 — LOS to vertical conversion

Tool: LOS → Vertical Converter. The projection correction that separates defensible numbers from the most common InSAR misinterpretation. Convert every point with its own incidence geometry, note the along-slope scenario where slopes exist, and carry the propagated uncertainty forward. Quality gate: sign conventions verified on one known area; the Data Tray now holds vertical rates for everything downstream.

Stage 4 — Spatial hazard screening

Tool: Subsidence Gradient & Hazard Classifier. Differential settlement damages assets; gradients localize it. Classify points against the criteria package matching the asset type, with the neighbour radius chosen for the structural span. Quality gate: class labels are always reported with their package name and radius; flagged clusters, not isolated points, drive follow-up.

Stage 5 — Temporal characterization

Tool: Time-series Calculator. Velocity hides timing. Where the classifier says where, the time series says how it unfolded — steady, accelerating, or interrupted by a discontinuity that is usually an unwrapping event rather than geology. Quality gate: accelerating points are cross-checked against seasonal context; step flags are traced to their interferograms before entering the report as findings.

Stage 6 — Ground-truth validation

Tool: InSAR ↔ GNSS Validation. Independent observations anchor the whole exercise. Match ground points within a defensible radius, read bias and residual spread separately, flag anomalous pairs, and copy the standardized comparison paragraph into the report. Quality gate: every flagged pair is accounted for — explained, investigated, or carried as a caveat; nothing is silently deleted.

Stage 7 — Uncertainty statement

Tool: Error Budget Calculator. The budget predicts what the validation measured. Combine the five standard error sources, identify the dominant one, and write the quality-assurance paragraph with eyes open about which correction would actually improve the numbers. Quality gate: budget values are justified by the project's own processing parameters, not left at screening defaults in a final deliverable.

Stage 8 — Reporting

Tool: Report & Interpretation Note Generator. Assemble everything into a scenario-tailored screening note with injected statistics, standardized limitations and verification language. The generator writes structure and neutral phrasing; the engineer writes the findings and owns the conclusions. Quality gate: no placeholder reaches the client; regenerated after any re-analysis.

The failure modes reviewers actually look for

A screening report fails review far more often on procedural details than on the calculation itself. The patterns experienced reviewers check first: a velocity map with no documented processing chain (stack size, orbit, algorithm, atmospheric model) — if the chain is absent, the numbers are not reproducible and the report cannot be accepted as evidence; geometry applied inconsistently, for example converting some points with their incidence angle and others with a single average, which introduces a slope-dependent bias of 5–20 % in the vertical rates; validation performed on points that were excluded from the comparison because they "did not look right" — exclusions must be documented, or the validation is treated as cherry-picking; error budgets left at screening defaults in a final deliverable, signalling that the uncertainty statement was pasted rather than computed; and sign or reference errors — a rate reported as subsidence when the convention of the stack defines negative LOS as motion towards the satellite, or a displacement referenced to the wrong epoch. Each of these failure modes is caught by exactly one quality gate in the chain above, which is why the gates are not optional decoration: they are the checklist a competent reviewer mentally re-runs while reading your report. Run the chain honestly and your deliverable carries the procedural fingerprints that reviewers recognise as professional work.

The workflow in one paragraph

Pre-check the site; process with your usual tools; convert coordinates; correct the LOS geometry; classify gradients in space; characterize them in time; validate against the ground; state the uncertainty; write it up. Each stage leaves an export that the next stage consumes, the Data Tray carries the points between tools, and every number in the final report traces back to a reproducible step. That traceability — not any single calculation — is what makes an InSAR screening deliverable survive review. It is also what makes the chain teachable: a junior engineer who follows the stages in order learns where the judgements live, and a reviewer who reads the report can reconstruct exactly which judgements were made and where — the two properties that let a screening study be trusted, re-run and defended years after it was written, which is the practical meaning of professional traceability in monitoring work for infrastructure.

Start now

The fastest introduction is the demonstration dataset: open the LOS Converter, load the demo, and follow the chain through classification, validation and reporting. The sample dataset guide documents every number you should see along the way, and the InSAR basics page covers the underlying concepts if the geometry terms above went by too quickly. If a stage produces numbers you cannot reproduce by hand, treat that as a stop condition: an unreproducible intermediate value will not survive review.