Method and sources

Three reasons, each checked against something it could have failed, and each with its own limits

Graded claim by claim on the satellite page, which also says what would settle what is open. The full audit, with what is not established and the corrections adopted after an independent review, is METHOD_AUDIT.md; every run’s parameters, code and data are linked from its details panel on the satellite page.

Every figure says whether it is simulated, measured, claimed or planned

Every material property is sourced, derived from a source, or marked assumed

P and S wave speeds in m/s; ′ derived from a source by a stated ratio; * assumed. Densities are assumed throughout until an open source is found.

Two communities, one vocabulary

Seismologists and radar engineers rarely read each other. The terms this site uses from each, in plain words; they are underlined where they appear.

ambient-noise interferometry seismic

Cross-correlating the background hum recorded at two sensors turns it into the wave that would travel from one to the other, so noise can stand in for a source.

azimuth radar

The direction along the satellite’s track, as opposed to range, the direction of the beam.

coherence radar

How similar two radar looks of the same ground are, from 0 to 1. Phase can only be compared where coherence is high.

crosshole seismic

A survey with sources down one borehole and receivers down another, so rays pass through the ground between them rather than skimming the surface.

Doppler frequency radar

The frequency shift of an echo from the satellite’s motion. In a focused image it maps to the time and direction from which a target was seen.

Doppler tomography radar

The claimed method: reading ground vibration from the differences between sub-apertures of one image and focusing it into depth. Published 2022, retracted 2026.

eikonal equation seismic

The high-frequency limit of the wave equation: it gives the earliest arrival time everywhere, and the rays are its steepest descent.

first break seismic

The first arrival of energy on a seismic record. Its timing, picked on every record, is the data of travel-time tomography.

Fisher information seismic · radar

How sharply the probability of the data changes with a parameter. Its inverse bounds the variance of any unbiased estimate (the Cramér–Rao bound), and for a small change it sets, to first order, how well any test can tell the two cases apart; for a finite change it bounds that with an explicit remainder.

full-waveform inversion seismic

Fitting whole seismic records, not just arrival times, by adjusting a model of the ground; each step is a gradient computed by propagating the misfit backwards in time.

geophone seismic

A ground-motion sensor recording velocity; three-component geophones record east, north and up.

imprint seismic · radar

The difference a buried chamber makes to how the surface moves as a wave passes: what any passive method would have to detect.

line of sight radar

The direction from the satellite to the ground; radar measures motion only along it.

look radar

An image formed from part of the Doppler band. In a dwell the band is time, so a look sees the ground as it was during a few seconds, and averages any motion over them.

microseism seismic

The ground’s constant hum from ocean swell, strongest between 0.1 and 0.3 Hz, felt everywhere on Earth as surface waves kilometres long.

P wave seismic

The compressional wave: the fastest, so the first to arrive. About 3,300 m/s in Giza limestone.

PML seismic

Perfectly matched layer: an absorbing boundary that lets simulated waves leave the grid without reflecting.

points per wavelength seismic

How many grid cells span the shortest wave simulated; too few and the waves travel at the wrong speed.

Rayleigh wave seismic

A wave that rolls along the ground surface, slightly slower than S, carrying most of a hammer blow’s energy. It samples to about a third of its wavelength in depth.

reverse-time migration seismic

Imaging by sending recorded waves backwards through a model of the ground and marking where they meet the waves sent forward from the source. It is the first step of full-waveform inversion.

S wave seismic

The shear wave: slower than P (about 1,800 m/s in the same rock) and unable to cross air or water.

SLC radar

Single-look complex image: a focused radar image that keeps amplitude and phase for every pixel.

speckle radar

The grainy texture of radar images of natural ground: many small scatterers in each cell interfering. Each Doppler bin holds an independent speckle pattern.

Spotlight Dwell radar

An ICEYE mode that keeps the beam on one scene for tens of seconds, here 24.5 s, far longer than a normal image needs.

sub-aperture radar

A part of the synthetic aperture processed on its own: a look from a slightly different direction and time.

synthetic aperture radar

The stretch of orbit over which echoes are combined into one image; a long aperture gives fine resolution along the track.

tomogram seismic · radar

A three-dimensional image of a property (here, wave speed) reconstructed from measurements taken around it.

travel-time tomography seismic

Finding the speed of the ground from how long first arrivals take between many sources and receivers, by tracing their rays and adjusting the speeds until the times fit.

virtual source seismic

A sensor made to act as a source: the correlation of its noise record with another sensor’s is the wave a real source there would have sent. Noise sources spread over the ground give it as minus the lag derivative of the correlation.

X-band radar

Radar near 9.6 GHz (3.1 cm wavelength). It enters dry limestone by centimetres, not metres.

Three commands rebuild everything

cd sim && uv sync && uv run pytest                     # the simulation and its checks (~40 s)
uv run python experiments/p2_07_whole_chain.py         # any run, e.g. the whole chain (~5 min)
uv run python -m katabasis.export                      # everything the site reads
cd ../web && npm ci && npm run build                   # the site, into docs/

The lab’s Python package keeps its first name, katabasis. Real-image steps need the ICEYE products, which are not in the repository; their acquisition records are, so everything else reruns anywhere.

One command runs a method over any area of a product and adds the result to the lab

cd sim
uv run --with scikit-image python -m katabasis.lab gated --name khafre --site giza --centre 0,0 --twin 101
uv run python -m katabasis.export                      # the run appears in the lab, under the satellite

gated runs the stricter gated reconstruction, the publicReplication and Verification of the Biondi Protocol (v1.8 standalone package, commit ac6f2500), unchanged but for its input paths, along east–west lines --step metres apart across a square of the site (--half-ew, --half-ns), each line placed through the product’s RPC over the site’s surface (--lines ns lays them north–south). Its fit scores are hung below the surface into a volume, unsmoothed and empty where nothing passed; --twin SEED runs the same raster on a motionless copy of the image, the control every picture needs, and --producttakes another pass, or a synthetic product. In the lab, every place a method was run shows at once, and the pass and the lines become choices: things a real feature should survive. Each run writes sim/results/lab_<name>/: its record (date, commit, the method’s source hash, every parameter) and its volumes, which the export adds to the lab with the record attached. More methods will join it as they are built.

Every picture on this site comes from one of these 55 runs, and each can be made again

The record each run wrote: when, from which commit, how long it took, and the command that repeats it (from sim/). A commit marked + had uncommitted changes to the lab’s code when the run saved (8 of 55); rerunning at that commit may differ in small ways. Runs that read a real ICEYE product need it on disk, and the 18 runs of the gated reconstruction need its public package, whose hash each of those records carries.

One repository, three parts