Viewer data sources

Where every value the frame viewer shows comes from, and how it is derived. Values are listed by where they live: per frame, per granule (GSLC) or interferogram (GUNW), the granule flags, and the QA metrics.

Four sources feed the page:

Source	What it is	Access
TrackFrame database	NISAR_TrackFrame_L_*.gpkg, one row per NISAR frame (nisar-db download-frame-db)	public download
CMR	NASA's Common Metadata Repository, https://cmr.earthdata.nasa.gov/search/granules.umm_json, collections NISAR_L2_GSLC_PROVISIONAL_V1 and NISAR_L2_GUNW_PROVISIONAL_V1. UMM-JSON leaves some granules out of its results while counting them in CMR-Hits (seen October 2026), so every search is checked against the atom-JSON granules.json listing and the missing records are fetched from search/concepts/<id> (details)	public
QA_STATS.h5	the small (60-160 kB) statistics file the NISAR QA software writes next to every product: https://nisar.asf.earthdatacloud.nasa.gov/NISAR/<collection>/<granule>/<granule>_QA_STATS.h5	Earthdata login
Repository files	CalVal sites, rollout regions, blackout dates, the flag and QA caches under catalog/	in the repo

The product HDF5 itself (several GB) is read only to learn the RFI mitigation flag of a release not seen before (one product per release); see RFI mitigation.

Frames

Value	Source	Derivation
frame_idx	TrackFrame database	row index of the frame in the database (the number OPERA uses)
track, frame, pass direction, outline	TrackFrame database	track, frame, passDirection, geometry; coordinates rounded to 1e-4 degrees
isSNWG, isDNC	TrackFrame database	as stored
hasLand (Land frames only)	TrackFrame database	hasLand, which is fractionLand > 0: any land at all
isCalVal (CalVal frames only)	scripts/disp_s1_calval_sites.geojson	frames covering at least 10 % of a DISP-S1 validation site (plus the Mexico City basin)
rollout options	rollout regions GeoJSON in the repo	regions the frame intersects
blackout months, windows	catalog/opera-nisar-disp-blackout-dates.json	per-frame snow windows from scripts/snow-analysis, summarised by blackout_summary
reference (reset) dates	latest release's opera-nisar-disp-reference-dates-*.json	as published
gslc_count, n_unique, n_duplicate, n_modes, n_full, n_partial	CMR GSLC search	counted from the frame's granules; a unique acquisition is one (date, mode, coverage); the _sel variants recount under the chips and date range
consistent mode / coverage	CMR GSLC search, or the published consistent-GSLC catalog	common_mode_coverage: among science modes 4005 / 2005, full frame over partial (unless partials dominate), then most acquisitions, then mode priority; the published catalog wins when given
gunw_count, gunw_pairs, GUNW modes, polarizations, dt range	CMR GUNW search	counted from the frame's interferograms
GUNW network	CMR GUNW search	connected components of the reference-secondary graph, computed in the page
flag colourings (all / some / none)	granule flags below	share of the frame's granules (under the chips) whose flag is set; not collected when none has been read
QA colourings	QA metrics below	median, worst decile or share past a threshold over the frame's granules under the chips

Granules and interferograms

Read from the granule name (GSLC and GUNW ids are the CMR GranuleUR), split on _:

Field	GSLC position	GUNW position	Example
cycle	4	4 (reference), 8 (secondary)	031
track	5	5	155
direction	6	6	D
frame	7	7	084
mode	8	9	4005, 4000
polarization	9	10	DHDH, SH
start time	11	11 (reference), 13 (secondary)	20260928T231125
composite release (CRID)	13	15	P05023
coverage (F full / P partial)	15	17	F
product version	last	last	001

Granule flags

Six flags per granule, keyed j f o r m d in catalog/granule_flags.json.gz (booleans as 0 / 1). Collected by scripts/collect_granule_qa.py --flags-output and by the local search & rebuild.

Key	Flag	Source	GSLC	GUNW
j	joint observation	QA_STATS.h5; also CMR JOINT_OBSERVATION	science/LSAR/identification/isJointObservation	referenceIsJointObservation OR secondaryIsJointObservation (same group)
f	full frame	QA_STATS.h5; also CMR FULL_FRAME and the name's coverage field	science/LSAR/identification/isFullFrame	same
o	orbit type	CMR ORBIT_TYPE	the granule's orbit (MOE, NOE, FOE, POE)	the reference acquisition's orbit (see below)
r	RFI mitigation applied	inferred per product type and CRID	0 for every GSLC read so far	1 for every GUNW read so far
m	mixed mode	QA_STATS.h5	science/LSAR/identification/isMixedMode	same
d	dithered	QA_STATS.h5	science/LSAR/identification/isDithered	same

Notes:

• Free with the search. j, f and o are in every CMR record, and r follows from the release, so a local rebuild shows those four for every granule without downloading anything. m and d need the QA_STATS.h5 download and read not read until then.
• RFI mitigation (r) is recorded only inside the product (.../GSLC/metadata/processingInformation/parameters/rfiMitigationApplied; for a GUNW the reference/ and secondary/ copies, both required). It is set by the processing configuration, and every (product type, CRID) group read so far agrees on one value, so a granule inherits its group's value. A group not seen before is learned from one product read; a group whose cached granules disagree is not inherited.
• GUNW orbit type (o). CMR records only the reference acquisition's orbit. GUNW pairs read from the product before (.../GUNW/metadata/orbit/{reference,secondary}/orbitType) carry both as REF/SEC when they differ (6 of ~10,000 checked); the viewer labels them orbit ref/sec and the others ref. orbit. Reading the secondary back is in TODO.md.
• Checked against the product reads of the earlier collector: 14,800 re-read globe granules and stratified samples over all 16 (type, mode, polarization, CRID) groups matched on every flag; CMR j / f matched on 1,175 granules and o on 2,032.

QA metrics

Keyed in catalog/granule_qa.json.gz, all from QA_STATS.h5. A granule withdrawn from the archive (the file returns 404) is cached as empty.

GUNW, from the frequency-A co-pol (HH, else VV) group science/LSAR/QA/data/frequencyA/unwrappedInterferogram/<pol>/:

Key	Metric	Dataset(s) under the group	Derivation
cm / ca	coherence median / mean	coherenceMagnitude/histogramBins, histogramDensity	from the histogram (the QA software's own mean_value reads near zero for some granules, so it is not used)
v	valid unwrapped pixels (%)	connectedComponents/percentPixelsWithNonZeroCC, unwrappedPhase/percentNan	100 * nonZeroCC / (100 - percentNan): share of the data area, not of the raster
l	largest connected component (%)	connectedComponents/percentPixelsInLargestCC, unwrappedPhase/percentNan	same scaling
n	connected components	connectedComponents/numValidConnectedComponents	as stored
im / imd	ionosphere phase screen mean / median (rad)	ionospherePhaseScreen/histogram*	from the histogram
is	ionosphere phase screen spread (rad)	ionospherePhaseScreen/histogram*	standard deviation from the histogram
iu	ionosphere uncertainty (rad)	ionospherePhaseScreenUncertainty/histogram*	mean from the histogram

GSLC:

Key	Metric	Dataset(s)	Derivation
rl	RFI likelihood	science/LSAR/RFI/data/frequency*/<pol>/rfiLikelihood	the largest over frequencies and polarizations; documented as 0-1 but up to ~1e31 in the archive, so stored as reported and coloured on a log scale

Both:

Key	Value	Dataset
qv	QA software version	science/LSAR/QA/processing/QASoftwareVersion

Histogram statistics weight each bin centre by density * bin width; the median interpolates inside the bin where the cumulative weight crosses one half (hist_stats in scripts/collect_granule_qa.py).

Browse and QA images

Image	Source	Access
browse image	https://nisar.asf.earthdatacloud.nasa.gov/BROWSE/<collection>/<granule>/<granule>_LATLON.png	public
QA report layers	pages of <granule>_QA_REPORT.pdf, cut into layers by the local helper	Earthdata login, through scripts/qa_browse_server.py