Abstract
Stone columns mitigate earthquake-induced liquefaction through three mechanisms: densification, shear-stress redistribution, and radial drainage. The first two are established at installation and are permanent; the third is not. Fines migrate into the column under cyclic hydraulic gradients, permeability falls, drainage resistance rises, and the pore pressure the design was intended to suppress re-accumulates. An improved site therefore migrates from a three-mechanism to a two-mechanism system during its service life, a transition current design practice does not recognize. This paper synthesizes that degradation quantitatively: a structured database of test configurations was extracted from a corpus of 188 records and pooled using a study-clustered random-effects procedure. The pooled bearing capacity ratio is 1.52 and the pooled settlement ratio 0.75, but every contributing configuration was tested statically, so the quantities the literature reports densely are not the quantities seismic design depends upon. That mismatch between what the literature measures densely and what seismic design actually requires is the principal empirical finding of this synthesis. Shaking table evidence shows ordinary columns clogging so rapidly that liquefaction onset is delayed by about one second relative to untreated ground at low base acceleration, after which the columns punch 26 to 43 percent of their length into the deposit; encased and sand-filtered columns settle with the ground by 9 percent and prevent liquefaction at four times that acceleration. Analytical models admitting clogging report post-earthquake settlement increases up to 229 percent and second-event pore-pressure ratios more than 36 percent higher. Against this, every fully coupled numerical framework in current use holds column permeability constant by assumption. The paper formalizes a clogging-induced hydro-mechanical deterioration framework with governing equations, shows the classical filter rule and the seismic drainage requirement to be mutually incompatible, and specifies the falsifiable experiment that would close the gap. Two limitations bound these pooled values and are stated here rather than buried: the bearing capacity ratio rests on three independent studies and the settlement ratio on one, so neither should be read as a stable population estimate, and both derive from static tests and therefore cannot be transferred to seismic drainage durability.