BCMG AIGeospatial · Movement Index
← bcmg.ai
BCMG AI · Nexus Geospatial · Public Briefing

The BCMG AI Movement Index
Palos Verdes Landslide Market

A 0–100 land-movement score for individual parcels · a decade of satellite radar, fused with the City’s licensed GPS monument surveys · Palos Verdes Peninsula, California
01 · The Information Gap

A famous landslide, a premium market — and almost no movement data for buyers

The south flank of the Palos Verdes Peninsula carries one of North America’s most active landslide complexes. Portuguese Bend sits within an ancient slide mass and has been moving continuously since 1956, when road construction reactivated it. After the exceptional back-to-back storm seasons of 2023, the slide accelerated dramatically — documented rates reached feet per month at the 2024 peak, fast enough to sever utilities and reshape roads, before the City’s emergency dewatering wells cut those rates sharply.

Yet the surrounding market is one of coastal Los Angeles’s most expensive, with multi-million-dollar transactions closing steadily on and around the peninsula. A buyer browsing listings encounters a striking silence: homes in and near the active zone commonly carry no land-movement language at all. The single most consequential fact about a property here — is the ground under it moving, and how fast? — is effectively absent from the marketplace.

BCMG AI built the Movement Index to close that gap: a 0–100 score for an individual parcel, derived from a decade of satellite radar measurements and fused with the City of Rancho Palos Verdes’s licensed GPS monument surveys — the ground-truth record that keeps working even where satellites cannot.

1956
moving continuously since
ft / month
documented peak rates, 2024
340+
radar passes analyzed, one decade
0–100
per-parcel movement index
02 · The Instrument

Radar interferometry: millimeter surveying from orbit

Before satellite radar, ground movement here was tracked two ways. The City commissions licensed GPS monument surveys — highly precise, but a sparse set of fixed points, measured periodically. And individual owners commission geotechnical borings — a single location on a single date, at a cost of thousands of dollars. Both are essential; neither tells a buyer what the whole hillside has been doing, everywhere, for years.

InSAR (interferometric synthetic-aperture radar) does. The Sentinel-1 radar satellite passes over the peninsula roughly every 12 days and records not just an image but the phase of the returning radar wave — where in its cycle the wave was when it bounced back. Comparing the phase between passes reveals ground motion at the millimeter scale, across the entire peninsula at once. BCMG AI’s pipeline analyzes more than 340 such passes spanning a decade — a dense movement history for every radar cell of ground, not just where a monument happens to stand.

Fig. 1 · What the radar actually measures
opposite-direction pass radar satellite ~12-day revisit look angle viewing ray · phase measured true motion component along the ray = what the radar sees
The radar measures motion along its viewing ray — one component of the true 3-D movement. Combining ascending and descending passes (opposite viewing directions) recovers vertical and east–west motion, the components that matter on a coastal slope.
Fig. 2 · Coherence: when the signal holds — and when it breaks
STABLE GROUND FAST-MOVING GROUND pass 1pass 2 pass 3pass 4 crests align → coherent, measurable pass 1pass 2 pass 3pass 4 echoes scramble → coherence lost the same ground, revisited pass after pass — phase either repeats, or it doesn’t
Ground moving faster than the radar’s measuring ceiling scrambles the signal. At Portuguese Bend’s core this is exactly what happens — so BCMG fuses in the City’s licensed GPS monument surveys. Absence of a radar score is treated as a warning, never as safety.
Fig. 3 · One radar measurement = this exact ground
Parcel-scale aerial view with one radar measurement cell outlined by a dashed square
Parcel scale, from the BCMG quality-control pipeline: the dashed square is a single radar measurement cell laid over the actual ground it averages — a sample parcel and its neighbors. The score a buyer sees is built from the cells that touch that specific lot, not from a peninsula-wide average.

Fig. 4 · The peninsula, scored — and deliberately veiled
Peninsula-wide overview with a plaid privacy pattern and the beyond-ceiling zone highlighted in red
Public overview from the Movement Index pipeline. The red patch at the coast is the Portuguese Bend–Abalone Cove complex — ground moving beyond the satellite’s measuring ceiling, where the record comes from City GPS monuments instead. The plaid coloration elsewhere is a decorative privacy pattern, not measurement data: parcel-level scores are delivered per address, to clients — not broadcast over every neighbor’s roof. How a data product handles discretion is itself part of the product.
03 · The Next Instrument

A longer wavelength for faster ground — ALOS-2 L-bandExperimental

The workhorse satellite, Sentinel-1, operates in C-band: a 5.6 cm radar wave. JAXA’s ALOS-2 operates in L-band: 24 cm — more than four times longer. The longer wave penetrates vegetation to reach the soil itself, and because the motion ceiling scales with wavelength, it tolerates roughly 4× faster movement per pass before the signal scrambles. BCMG has demonstrated L-band holding coherent lock across the Portuguese Bend complex itself — terrain where C-band cannot keep up (Fig. 6). ALOS-2 is currently integrated into BCMG’s Nexus platform as an experimental instrument: a supplement under validation, not yet part of the scored product.

Fig. 5 · Two radar wavelengths, to scale
Sentinel-1 · C-band (the workhorse) λ = 5.6 cm · motion ceiling ≈ 1.4 cm per pass ALOS-2 · L-band (sees through brush, tolerates speed) λ = 24 cm · motion ceiling ≈ 6 cm per pass
Drawn to true relative scale. The per-pass motion tolerance is about a quarter wavelength — the longer L-band wave keeps its lock on ground moving roughly 4× faster than C-band can follow.
Fig. 6 · L-band holding lock over Portuguese Bend
ALOS-2 L-band coherence map of the peninsula showing sustained coherence across the Portuguese Bend box
BCMG quality-control output from an ALOS-2 L-band pair (28 days apart). Green-to-yellow = coherent signal. Inside the outlined Portuguese Bend box, L-band coherence holds at the same level as the surrounding stable peninsula — the longer wave keeps measuring where C-band goes blind. Under validation as a supplement to the scored product.
04 · The Method

Why a computational pathologist reads this ground

BCMG AI’s principal, Dr. Cone, is a physician — a computational pathologist. Reading a gigapixel pathology slide and reading a decade of satellite imagery are, computationally, the same discipline: multi-resolution image pyramids (peninsula → hillside → parcel, exactly as slide → region → cell), and texture and pattern recognition at scale, where the finding is a subtle textural shift an untrained eye passes over. The same modern computer vision runs on aerial imagery — structure and damage detection with human-in-the-loop review, every machine flag checked by an expert eye.

Above all, pathology brings controls. No pathologist trusts a stain without a control slide, and no BCMG analysis ships without the geospatial equivalent:

Negative control
GPS-instrumented quiet ground — independently surveyed, known stable — must score quiet through the identical pipeline.
Positive control
Documented moving ground — independently surveyed, known active — must light up through the identical pipeline.
Identical pipeline
Controls and client parcels run through the same code, same parameters, same satellite record. No special handling.

When both controls behave, the result in between can be trusted the way a pathologist trusts a properly stained slide. Biospatial to geospatial: same eyes, different tissue.