Approaches to the modelling of long-term, large-scale (1–100 year and drainage basin to coastal zone) behaviour are outlined for the catchments of the Humber and Tamar estuaries, UK, and San Francisco Bay, USA. The estuarine environment is particularly difficult to model. Once fluvial materials enter an estuary, sharp chemical gradients and complex physical and biochemical dynamics can result in large changes to the amount, chemical nature and timing of the flux of nutrients and other materials that are transported to the coastal zone. Examples are given both of catchment models for the three systems and of an estuarine model that has been used to estimate contemporary fluxes of some nitrogen species from the Humber catchment to the North Sea, in an attempt to forecast the possible effects associated with climate change. Other applications of long-term models are discussed, including predictions of oxygen isotope ratio distributions in San Francisco Bay, which are of value to palaeo studies, and the dependence of salt intrusion on rising sea level and drought conditions there. Some of the more difficult modelling problems involve the determination of estuarine sediment transport and its consequences. A problem which has received little attention up to now, but which is becoming a very active area of research, is the influence of biota on sediment transport. Long-term modelling of fine sediment transport and its application to contaminant (including radionuclide) transport and dispersion to the coastal zone are discussed for the Tamar Estuary. Other long-term estuarine problems that are discussed include the modelling of seasonal heat fluxes from the Tamar to its coastal zone, and the dispersion of tracers within the coastal zone using three-dimensional simulations of heat, salinity and flow. The article is mainly concerned with simulations that require only reasonably short amounts of time to run on a PC and with models that are not intended to give solutions of great temporal or spatial resolution. Generally, they are intended to provide insights into phenomena and systems. To achieve this, a common feature of the long-term models has been their parameterisation of some intratidal processes, such as sediment resuspension and deposition, or tidal mixing. Provided the parameterisations are physically based and lead to good comparisons with available data, when these exist, then the approach would appear to be a valid ‘use’ rather than an invalid ‘abuse’ of models. [Copyright &y& Elsevier]