Geoscience modellers develop numerical models where constraints are placed on how the model domain and sampling relates to location in the modelled world and the real world. The shape of the model domain is a significant factor for numerical algorithms and computational solvers. This leads to a number of interesting definitions of coordinate reference systems. I will summarise some requirements the modelling community have for specifying and working with coordinate reference systems. Post processing and presentation of analyses are important factors; archiving for future use is a crucial consideration. I will present examples of horizontal and vertical coordinate system definitions in common use in the meteorology and oceanography domains and the challenges they may bring. The conclusion will be a discussion how specifications and tools for defining, interpreting and transforming coordinate reference systems, such as Well Known Text (WKT), European Petroleum Survey Group (EPSG) and PROJ.4 are able or unable to meet the requirements of a geosciences modeller.