A short history of GETM

The idea for GETM was born in May 1997 in Arcachon, France during a workshop of the PhaSE project which was sponsored by the European Community in the framework of the MAST-III programme. It was planned to set up an idealised numerical model for the Eastern Scheldt, The Netherlands for simulating the effect of vertical mixing of nutrients on filter feeder growth rates. A discussion between the first author of this report, Peter Herman (NIOO, Yerseke, The Netherlands) and Walter Eifler (JRC Ispra, Italy) had the result that the associated processes were inherently three-dimensional (in space), and thus, only a three-dimensional model could give satisfying answers. Now the question arose, which numerical model to use. An old wadden sea model by Burchard (1995) including a two-equation turbulence model was written in $ z$-coordinates with fixed geopotential layers (which could be added or removed for rising and sinking sea surface elevation, respectively) had proven to be too noisy for the applications in mind. Furthermore, the step-like bottom approximation typical for such models did not seem to be sufficient. Other Public Domain models did not allow for drying and flooding of inter-tidal flats, such as the Princeton Ocean Model (POM). There was thus the need for a new model. Most of the ingredients were however already there. The first author of this report had already written a $ k$- $ \varepsilon $ turbulence model, see Burchard and Baumert (1995), the forerunner of GOTM. A two-dimensional code for general vertical coordinates had been written as well, see Burchard and Petersen (1997). And the first author of this report had already learned a lot about mode splitting models from Jean-Marie Beckers (University of Liege, Belgium). Back from Arcachon in Ispra, Italy at the Joint Research Centre of the European Community, the model was basically written during six weeks, after which an idealised tidal simulation for the Sylt-Rømø Bight in the wadden sea area between Germany and Denmark could be successfully simulated, see Burchard (1998). By that time this model had the little attractive name MUDFLAT which at least well accounted for the models ability to dry and flood inter-tidal flats. At the end of the PhaSE project in 1999, the idealised simulation of mussel growth in the Eastern Scheldt could be finished (not yet published, pers. comm. Francois Lamy and Peter Herman).

In May 1998 the second author of this report joined the development of MUDFLAT. He first fully rewrote the model from a one-file FORTRAN77 code to a modular FORTRAN90/95 code, made the interface to GOTM (such that the original $ k$- $ \varepsilon $ model was not used any more), integrated the netCDF-library into the model, and prepared the parallelisation of the model. And a new name was created, GETM, General Estuarine Transport Model. As already in GOTM, the word "General" does not imply that the model is general, but indicates the motivation to make it more and more general.

At that time, GETM has actually been applied for simulating currents inside the Mururoa atoll in the Pacific Ocean, see Mathieu et al. (2002).

During the year 2001, GETM was then extended by the authors of this report to be a fully baroclinic model with transport of active and passive tracers, calculation of density, internal pressure gradient and stratification, surface heat and momentum fluxes and so forth. During a stay of the first author at the Université Catholique de Louvain, Institut d'Astronomie et de Géophysique George Lemaître, Belgium (we are grateful to Eric Deleersnijder for this invitation and many discussions) the high-order advection schemes have been written. During another invitation to Belgium, this time to the GHER at the Université de Liège, the first author had the opportunity to discuss numerical details of GETM with Jean-Marie Beckers, who originally motivated us to use the mode splitting technique.

The typical challenging application in mind of the authors was always a simulation of the tidal Elbe, where baroclinicity and drying and flooding of inter-tidal flats play an important role. Furthermore, the tidal Elbe is long, narrow and bended, such that the use of Cartesian coordinates would require an indexing of the horizontal fields, see e.g. Duwe (1988). Thus, the use of curvi-linear coordinates which follow the course of the river has already been considered for a long time. However, the extensions just listed above, give the model also the ability to simulate shelf sea processes in fully baroclinic mode, such that the name General Estuarine Transport Model is already a bit too restrictive.

kklingbe 2017-10-02