Meso-scale processes in the turbulent ocean act on length scales larger than the micro-scale (O(10km)) and smaller than basin-scales or those of large ocean fronts (O(100-500km)). They can be compared somehow with low-pressure systems or large storm systems in the atmosphere. Meso-scale eddies account for a huge amount of the energy transport from larger to smaller scales (the so-called energy cascade). Furthermore, such eddies have a major impact on the transport of water masses across ocean fronts and thus on the mixing.
Ocean models need to have a minimum horizontal resolution (nodes per length scale) in order to simulate meso-scale processes properly. Often, the internal Rossby radius of deformation is used to assess the capability of a model to resolve mesoscale eddies (see also e.g. Hallberg (2013).
A numerical ocean model is a computer program which calculates various physical, chemical and biological parameters by solving mathmatical equations on a pre-defined descrete grid using numerical procedures. In some cases, it is not possible to describe a relation bewteen two or more parameters with purely physical-mathmatical equations or its computation would be too costly. Then an empirical relation between predictable parameters is used instead (a so-called parametrisation).
The program code itself is often written in FORTRAN language and can esaily exceed 100.000 lines of code. Due to the complexity of modern model code it is essential to ensure that the development is well documented and traceable (e.g. using versioning tools like git or svn) and the compilation of the program executable is reproducible.
Such ocean models are then utilised to run numerous simulations. High-resolution simulations in particular require a huge amount of computing resources and therefore only supercomputers providing plenty of CPUs and Memory (High-Performance Computers - HPC) are capable of carrying them out within a reasonable period of time. Furthermore, in climate (ocean) science it is inevitable to analyse the model output of the complete ocean and over the full simulated time span. And thus the storage capabilities of the computing facility is also of special importance.
A general introduction into ocean modelling can be found....
A simulation (or experiment) is the execution of a certain model code using specific preadjustments and input data. The simulated time span is often some period from the past or lays in the future. The most realistic simulations are typically labeled as "reference" experiments, while "sensitivity" experiments are used to investigate the effects of only one modified parameter on the ocean state in the model.
So-called ocean-only simulations require additional atmospheric forcing data sets as input for the surface boundary conditions of the ocean. An experiment that aims to simulate the ocean state of the last decades with such a forcing is called a "hindcast" experiment. A "forecast" experiment (or "projection") in contrast makes only sense when run in coupled mode with an atmospheric model. If the simulated time span lays further in the past (e.g. millions of years) using some reconstructed, artificial atmospheric forcing or a coupled ocean-atmosphere model, the experiment is called a paleo-simulation.