Valhall is a large chalk field in the Norwegian sector of the North Sea. The reservoirs consist of highly overpressured (0.84 psi/ft) chalk. The high overpressure and early oil migration has resulted in very well preserved porosity, exceeding 50% in parts of the field. This highly porous chalk is extremely weak, which results in liquefaction of the chalk at certain conditions during production. The weak nature of the chalk does also result in significant reservoir compaction, exceeding 10 meters in some locations in the reservoir. The compaction is providing an excellent source of reservoir energy, accounting for 50-60% in certain areas of the field. The compaction is transferred to the seafloor in the form of subsidence and is currently exceeding 6 meters below the central platform complex.The seafloor subsidence has resulted in a diminishing air gap for the platforms at Valhall, resulting in demobilisation of personnel during very rough winter storms. Water injection started in the field in 2004 with a gradual step up in rate. An offset chalk field in the area have experienced an increasing subsidence rate due to water weakening of the chalk. The micro-scale physical and chemical processes involved in water weakening of chalk have not been quantitatively determined, but a strategy to handle this had to be developed. This paper documents the development of a prediction model for the subsidence at Valhall between 2002 and 2008. The paper covers the development of a suitable and effective constitutive model to account for strain rate dependent reservoir compaction during depletion, re-pressurisation and waterflooding. The constitutive deformation model is driven by pore pressure and watersaturation calculated in a reservoir flow model and imported into the geomechanics model.It will be described how the constitutive model for the reservoir was populated. The overburden, underburden and sideburden were also modelled and required properties. The initialisation requires an initial stress state and it will be presented how this was defined. We also included faults in the model and check for re-activation and amount of slip on them. The Valhall model can also be constrained by a relative large amount of field and surveillance data. It will be presented how this data was used in the history matching process. The data used for the history matching process includes vertical and horizontal platform movements from GPS, time lapse seafloor bathymetry maps, radioactive markers in the reservoir and overburden as well as 4D seismic. It will be presented how the model matched the observed data from the field, showing a gradual decrease in subsidence rate from around 25 cm/year with the risk to increase due to water weakening, to the current reduced rate at 11 cm/year. This response is different to the one reported from the offset field. We will also present other high impact business applications of the model prediction, besides seafloor subsidence forecasting.