Imaging Dyke-Induced Deformation in the Lab

The presence of a dike in Earth's crust can be inferred from a combination of seismological and geodetic techniques. Often, geodetic measurements (e.g. GPS and InSAR) of surface deformation near a dike are inverted for dike parameters using the equations for rectangular tensile dislocations in an elastic, homogeneous half-space, as presented in Okada (1985). Here, we create artificial dikes in the lab and simulate surface geodetic observations. Because the dike parameters are known, or can be directly observed, we are able to test the validity of standard geodetic models for dike injections. A fluid injection into gelatin is analogous to dike ascent driven by magma buoyancy in the crust; gelatin is an elastic medium (Menand and Tait, 2002) and away from the tip of a magma dike, strains are small, suggesting that deformation of the host rock may also be largely elastic (Delaney and Pollard, 1981). Gelatin has been used in several investigations into the propagation of fluid-filled cracks in the Earth's crust, including the shape and velocity of fluid-filled fractures (Takada, 1990; Dahm, 2000), propagation in layered media (Rivalta et al., 2005) and laccolith growth (Johnson and Pollard, 1973), among others. This study focuses on the deformation at the surface caused by this propagation. We seeded the upper surface of a homogeneous gelatin mass with markers and injected some fluid at the bottom of the gelatin container. We recorded the ascent of the resulting buoyancy-driven fluid-filled fracture from above with two camcorders and from the side with an additional camcorder. Surfaces markers were observed by the camcorders and tracked from one frame to the next. 3D positions were determined using photogrammetry after matching the markers. The resultant time series of surface deformation at each marker are analogous to continuous GPS observations from real dikes. The horizontal accuracy obtainable with HD camcorders was about 0.1 mm. We inverted the surface deformation using the standard Okada model, and a more realistic model based on fluid-filled fractures theory, and compared the inversion results to the direct measurements of dike parameters. We find that the Okada model underestimated the size of the dike, but overestimated the dike thickness. The parameters recovered assuming a propagating fluid-filled fracture shape were in much better agreement with the real observations.