From Fault Fabric to Early Warning: High-Resolution Geomorphology and Enhanced DInSAR Over the SP-0B Saltworks (Dead Sea, Jordan) - Mini Review

The Dead Sea has dropped ~40m since 1980, removing ~0.48-0.49MPa of hydrostatic buttress from the basin walls. This rapid base-level fall steepens hydraulic gradients, depressurizes adjacent aquifers, and accelerates dissolution of evaporites in the Lisan Peninsula. The emergent wave-cut platform around Cape Costigan was converted to industrial salt-evaporation ponds by the Arab Potash Company (APC). SP-0B, bordered on the west by earthen Dike 19, catastrophically failed on 22 March 2000, releasing ~56Mm³ of brine and revealing solution voids beneath the platform. Although Dike 19 has since been rebuilt and densely instrumented, the substrate remains karst-prone and structurally complex, necessitating a quantitative, operations-ready early-warning approach.

This work asks two questions: (i) Can we map, at basin scale, the structural corridors (faults, diapiric joints, sinkhole belts) that localize karst beneath SP-0B; and (ii) Can spaceborne deformation be converted quickly into an early-warning stream that informs day-today dike management?

Two complementary streams are fused (1) A six-decade morpho-tectonic reconstruction uses aerial photographs (1953, 1992, 1999), CORONA (1971-1974), Landsat/ASTER/ Sentinel-2 (1973-2025), Space-Shuttle photography and sub-meter VHR satellites (2000- 2012) to delineate transform splays, diapiric joints, teepee ridges, sag ponds and shoreline retreat. (2) ERS-1/2 (1992-1999) and Sentinel-1 (2015-2025) Differential InSAR (DInSAR) time series are processed within SBAS. A “sibling-coherence” filter clusters only neighbors with matching multi-date amplitude statistics, preserving sharp deformation gradients and increasing the density of reliable pixels by ~20-30% relative to boxcar averaging, yielding 6- to 12-day LOS-displacement cubes at centimetric precision. All layers are integrated in UTM- 36N and tied to the continuous Dead Sea gauge.

The Lisan Peninsula overlies an actively rising salt diapir cross-cut by Dead Sea Transform (DST) splays. Two stacked fluids govern near-surface hydrology: (i) A dense, hypersaline brine wedge (~1.24g cm⁻³) extending inland; (ii) a 2-6m perched lens of fresh to brackish water (~1.00-1.02g cm⁻³) above it. The DST’s N-S sinistral faults and a NE-SW normal-fault array create a high-permeability lattice that funnels recharge toward the retreating shoreline, where freshwater-brine mixing drives evaporite dissolution and sinkhole development. As lake level fell from ~-395m in the 1950s to ~-440m by 2025, the operative halokarst column thickened and the wave-planed platform widened, exposing structurally guided lineaments, teepee structures and sag ponds.

Annualized shorelines and lineaments reconstructed from CORONA and Landsat show a west-widening platform sloping 1-2° lakeward. Lineaments align with the DST strike-slip corridor (N-S) and the western normal-fault array (NE-SW), confirming tectonic control on groundwater flow. A longitudinal discharge corridor developed along the N-S strike-slip zone, marked by aligned teepees and a cluster of sag ponds; this corridor later overlapped the western margin of SP-0B.

Time-series InSAR reveals a bowl-shaped subsidence field in northern SP-0B reaching -40 to -48 cm yr⁻¹, superposed with localized uplift patches up to +5cm yr⁻¹. The March 22, 2000 dike failure occurred within the maximum-subsidence zone, linking regional drawdown, dissolution along structural corridors, and engineered failure. While geotechnical analyses emphasized weak, sensitive laminated clays and load-induced pore-pressure rise, the present synthesis shows these materials concentrate within a structurally inherited accommodation space (pull-apart/minibasin/ salt-induced synform) that predisposed the site to settlement and collapse.

Two high-quality interferograms bracket key pre-refilling intervals: (a) 23 Feb-7 Mar 2025 (descending, 12 days, coherence ~0.98; height of ambiguity ~1,899m) and (b) 10 Jan-16 Apr 2025 (ascending, 96 days). Both show widespread subsidence in the southeastern sector of SP-0B and persistent localized uplift to the northwest, consistent with ERS patterns and the interplay of dissolution, groundwater flow, anhydrite-to-gypsum hydration, and halokinetic doming. DInSAR motions are relative to a presumed stable reference; absolute magnitudes require GNSS/leveling.

Conventional boxcar coherence over-smooths heterogeneous terrains and overestimates coherence in mixed pixels. The siblingcoherence approach clusters pixels by similarity in amplitude statistics, then computes coherence and phase within this adaptive ensemble, followed by second-kind statistical filtering. Implemented as a SNAP plug-in within an SBAS workflow, it retains narrow, high-gradient rupture/sinkhole zones and increases reliable-pixel density by ~20-30%. Over a tectonically quiescent Inner-Mongolia test site, sibling selection reduced phase bias/ variance versus the boxcar+Goldstein chain.

The 40m drawdown since 1980 translates to ~490 kPa of lost hydrostatic buttress, steepening hydraulic gradients that drive fresh/brackish water through fractured, soluble units. A simple Darcy estimate gives q ~ 8×10⁻⁶ m s⁻¹--several-fold higher than pre-1980 flux--supplying chemical work for halite dissolution and gypsum hydration plus mechanical work that fractures roofs and collapses voids. ERS and Sentinel-1 time series thus capture a “leaky-capacitor” behavior: stress accumulates during dissolution until roof failure releases it as accelerated subsidence.

A static hazard layer (faults, diapiric joints, sinkhole belts, discharge corridors) is fused with Sentinel-1 displacement cubes to partition SP-0B into watch cells. Trial rules escalate responses when |dLOS| > 6mm in 12 days or when acceleration turns positive: field inspections intensify, sensor sampling densifies, and brine inflow is temporarily throttled; three simultaneous cell breaches halt impoundment. InSAR streams are cross-validated with extensometers, piezometers and GNSS to anchor relative motions to absolute frames.

Limitations and outlook. Current displacements are referenced to assumed-stable zones; Sentinel-1’s ~14 m GSD misses metrescale ruptures; thresholds are rule-based rather than fully coupled hydro-mechanical responses. Planned upgrades include tying the InSAR grid to a permanent GNSS backbone, extending coherence selection with single-date deep-learning classifiers, and coupling displacement cubes to 3-D transient hydro-mechanical models driven by the continuing Dead-Sea drawdown.

Contribution and transferability. The framework shifts SP- 0B management from reactive repair to proactive mitigation by (i) establishing a structural baseline that identifies dissolution corridors; (ii) delivering near-real-time ground-motion cubes with higher fidelity in heterogeneous karst terrain; and (iii) operationalizing thresholds that link geodetic precursors to concrete dam-management actions. Because Sentinel-1 data are free and global, and processing uses open tools (ESA SNAP + SNAPHU), the workflow offers a reproducible template for other evaporite basins undergoing rapid base-level fall.