In order to demonstrate the interactions between plate boundary forces and preexisting zones of weakness, and how those control the localization of continental deformation inside the lithosphere, the thermomechanical evolution of an initially thinned and thermally perturbed continental lithosphere in compression was analyzed within the framework of the Pyrenees. While a concensus exists as to the shallow structures, the extent and geometry of deformation at depth have been the subject of great controversy. The thermomechanical implications of recent geological hypotheses on the initial state, prior to the compression, recently derived from balanced cross-section reconstructions are here quantified and compared to the recent ECORS deep seismic reflection profile across the Pyrenees. A thermomechanical model based on finite elastoplastic-viscoplastic deformations has therefore been formulated using phenomenological constitutive equations derived from experimental studies. The thermally coupled problem is solved by an appropriate vertical two-dimensional plane strain finite element approximation. In those models most of the shortening is accommodated inside a local zone of intense plastic shearing that creates a sharp wedge-shaped root progressively sinking deep into the mantle. The main thickening takes place inside the lower crust, where it can reach a factor 2. This indentation and crustal imbrication process is controlled by a bucking instability occuring inside the uppermost part of the mantle.

The evolution of the topography can be related to these phenomena. Lateral conductive heat transfer overtakes the advective transfer in the vicinity of the inflow of colder crustal material inside a warmer mantle. The deformation pattern can be explained by a three-layer mechanical lithosphere, with a soft ductile internal layer of potential decoupling between two more competent layers characterized by a significan elastic stored energy. The overall response of this composite plate is not much influenced by smaller scale variations of crustal rheology but is very much dependent on the contrast in rates of viscous dissipation between the lower crust and the upper mantle and on the possibility of singling out a layer of significant elastic stored energy by the balance between rate of loading and rate of dissipation inside the uppermost mantle. Predicted deformation pattern is found to agree with the seismic reflection and gravity data. Even though extrapolation to other rates of shortening must be verified by appropriate numerical modelling, these physical mechanisms may be regarded as a paradigm for other crustal imbrication phenomena found in the Himalayas, Tibet, or the Alps. ¿American Geophysical Union 1991