As tragically demonstrated by the 2011 Tohoku Japan earthquake, one of the most challenging issues in improving seismic hazard assessment consists in better forecasting the size/magnitude of future great (M>8) earthquakes. This requires exploring many fundamental but unresolved questions in earth sciences: What controls the lateral variation of large earthquakes occurrence along major seismic faults? What governs the transition from stick-slip behaviour to steady sliding? How do earthquake rupture zones recover and reload? How do large and small earthquakes fundamentally differ, if they do? To answer these questions active oceanic subduction zones are obvious targets because most of the great earthquakes occur there. However the presence of water prevents from measuring surface deformation within the frontal most ~100 km of major seismic structures. Much effort has been done to deploy ocean bottom instrumentation (including seismic or geodetic measurements), but the sparse coverage of these datasets still prevents detailed studies of seismic zones and of their state of stress. Here, we will rather favour the study of an emerged area: the Himalayan belt, which – as a former oceanic subduction zone – exhibits a first-order along-strike continuity of major faults and tectono-stratigraphic units that can be mapped over an E-W distance of ~2500 km. Some recent major historical earthquakes have been documented. However both the maximum earthquake size that struck the Himalayan front in the past and the probability of occurrence of a magnitude 9 megaquake in the next decades are still debated. Over the last decades most studies along the Himalayas have focused only on Central Nepal. Therefore lateral variations of the state of stress on frontal faults and the size of great Himalayan earthquakes are still poorly constrained. Based on previous studies and our own work in Nepal and Bhutan we have decided to address the question of lateral variations in seismic coupling along the Himalayan arc by extensive and detailed description of local loading (present-day convergence and seismicity rate, late Quaternary shortening rate, past seismic events), and crustal structural geometry (major faults, Moho depth, Indian plate flexure) from western Nepal to Bhutan. The proposed approach is clearly multi-disciplinary and aims at integrating deformation of the Himalayan arc over various spatial and temporal scales. Our methodology encompasses a large panel of up-to-date and innovative complementary techniques in gravity, seismology, geodesy, morpho-tectonics, paleo-seismology and thermo-mechanical numerical modelling. Gathering a high level of expertise from national (Montpellier, Paris, Nancy, Chambéry-Grenoble) and international (Switzerland, USA, India, Nepal and Bhutan) co-operation, this project will also contribute to the development of innovative methods in the InSAR processing chain, analysis of GOCE gravity data, determining accurate hypocentral location and numerical modelling. Ultimately, this project will contribute to provide the first real 3D image of the state of stress along a continental thrust-fault system, which is a crucial step in improving seismic hazard assessment in the areas producing most of the largest earthquakes.