Earthquakes pose one of the greatest natural threats to vast populations. In the last century, earthquakes have caused 2.3 million deaths (1 million in the last 30 years alone) and US$820 billion of financial losses. Earthquakes are generated by movement along lines of geological weakness called "active faults" which, in some places, can be observed on the Earth's surface. Unlike other natural hazards, advances in scientific understanding have not yet led to a reduction in fatalities from earthquakes. Predicting the timing, location and magnitude of individual earthquakes is likely impossible, but estimating the spatial distribution of earthquake hazard is manageable, and of great importance to the global population and the insurance economy. However, there are difficulties in calculating the earthquake hazard because we are currently reliant on present-day measurements of the rates of movement of faults and historical records of past damaging earthquakes. We cannot simply observe earthquakes for longer, therefore we must develop 'geologically richer' numerical simulations to build synthetic earthquake records and seismic hazard models to improve our understanding of the fundamental processes that control earthquakes. This project will develop a new, geologically-rich, fully integrated physics-based approach to modelling all aspects of the earthquake cycle. The earthquake cycle is the cyclical nature of earthquakes occurring, with tectonic stress building up and then releasing in a series of earthquakes over time. The physical processes that control the earthquake cycle operate on different time-scales, from seconds during the earthquake to millennia between earthquakes recurring on the same fault. The shape and spacing of faults also affect how earthquakes are generated, but it is not always easy to see the true shape of faults at the Earth's surface. There are three stages of the earthquake cycle that are currently modelled separately. These are; 1. the dynamic process of fault slip occurring over seconds to minutes during the earthquake, 2. the resulting deformation and stress transfer onto surrounding faults and 3. the evolution and accumulation of tectonic stress between earthquakes. Each of these three stages can be modelled individually and are used to speculate on different aspects of the earthquake cycle. However, because they are presently not integrated, the effects of each one on the others are poorly understood. Several active and inactive systems of extensional faults will be studied. The seismically active central and southern Italian Apennines will be studied because there is a wealth of data available; the faults are well-exposed at the surface and there is a 700 years record of damaging earthquakes and therefore high seismic hazard. The inactive fault systems that will be studied are offshore Norway, Australia and New Zealand. These inactive systems are important to study because we can use seismic reflection (like echo-location of the ground under the sea bed) to image the faults, to see their 3D shape and study how that has evolved with time. The slip rate on these faults can be quantified by studying the age and offset across these faults. It's important to study a range of different systems to synthesise the different data sets available in these regions. In summary, earthquake hazard forecasting is currently lagging behind forecasting of other natural hazards. By combining three different physics-based modelling approaches and testing the resulting model on two data-rich natural fault systems, this project will generate a truly physical and geological model of a fault system - this has not been attempted before. These models will output synthetic earthquake catalogues that can be compared to historical records (hindcasting), used to speculate on the future locations of earthquakes (forecasting) and used to inform and understand uncertainty in seismic hazard mode