An analytical method has been developed to predict the location, depth and size of caverns created at the interface between salt and overlying formations. A governing hyperbolic equation is used in a statistical analysis of the ground survey data to determine the cavern location, maximum subsidence, maximum surface slope and surface curvature under the sub-critical and critical conditions. The regression produces a set of subsidence components and a representative profile of the surface subsidence under sub-critical and critical conditions. Finite difference analyses using FLAC correlate the subsidence components with the cavern size and depth under a variety of overburden strengths and deformation moduli. Empirical equations are proposed which correlate the subsidence components with the cavern configurations and overburden properties, allowing prediction of cavern configurations from the subsidence components.