The Geodynamics of Relamination, Delamination, and Foundering of the Lower Crust
- Stoner, Ryan Kevin
- Advisor(s): Hacker, Bradley R
Abstract
\textit{What is the fate of the lower crust during continental collision?} Chapter 2 incorporates geologic constraints of lower crustal dynamics. Specifically, crustal xenoliths from the Pamir are used to constrain the geodynamic style of the loss of a lithospheric root. Delamination—the peeling back of a dense layer—is favored based on xenolith pressure-temperature-time paths. Moreover, delamination is caused by feedbacks between the lateral escape of the lower crust and cooling at the margins of the orogen. This contrasts with the assumption that delamination and channel flow are decoupled mechanisms. Instead, laterally flowing buoyant lower crust can founder after cooling. \textit{How do sediment diapirs differentiate?} Recent models of sediment diapirs in subduction zones assume diapirs are perfectly round with no internal transport of solid-state material. Chapter 3 tests these assumptions by applying porous and solid-state flow to diapiric differentiation. Not only can diapirs split into a buoyant portion and a dense residue but mushy material flows upward along the diapir margins, driving convection \textit{within} the diapir. Such churning could promote reactions between different lithologies in diapirs, meaning that hybridized material need not only be created in a subduction channel but inside the diapir itself. More broadly, however, geodynamic models of diapirism, delamination, and tectonic processes in general often require many more input parameters than can be tested individually, limiting the robustness of results. \textit{How then can the sensitivity of many input parameters in geodynamic models be tested?} This work streamlines the calculation of linearized sensitivities. Specifically, the technique of \textit{reverse automatic differentiation} is applied to models with many (154) inputs to obtain sensitivities of \textit{all} inputs for a reasonable computational cost ($<$6x forward pass). This technique is applied to models of sediment diapirs to lay a framework to test far more parameters than would be possible by the individual variation of inputs.