Ocean Dynamics
https://doi.org/10.1007/s 10236-020-01353-9
(N)
Improvements in turbulence model realizability for enhanced
stability of ocean forecast and its importance
for downstream components
Thorger Brüning '®©
Check for
updates
Received: 27 February 2019 / Accepted: 5 February 2020
© The Author(s) 2020
Abstract
The numerical stability of ocean circulation models is of high significance in operational forecasting. A substantial
improvement in numerical stability of the 3D-ocean model HBM could be achieved by the implementation of new
realizability criteria in the turbulence closure scheme. Realizability criteria which were already well documented for
closure functions without double diffusion were therefore extended to those using double diffusion. A purely technical
validation method called e-test which is suitable for the detection of numerical stability problems is presented, and the
effect of the development in turbulence model is demonstrated under severe weather conditions during extreme storm
events. Evaluation of statistics of longer simulations indicate that instabilities appeared only locally and temporary;
nevertheless, a significant impact on drift products relying on the current forecasts could be demonstrated, which under-
lines the importance of realizability in turbulence closure schemes in comprehensive operational model systems including
ocean circulation and downstream drift components.
Keywords turbulence closure - numerical stability / realizability - operational forecasting systems
1 Introduction
A comprehensive operational oceanography service today
should consist of both observation data (remote sensing
data, in situ measurements) and model data (forecasts and
reanalysis) of both physical and biogeochemical parame-
ters (She et al. 2016). Such a service exists both at
European level (CMEMS — Le Traon et al. (2017)) and
mostly also at national level with a greater focus on the
immediate coastal waters (e.g. in Germany, Brüning et al.
(2014); in the Netherlands, De Kleermaeker et al. (2012);
in Portugal, Mateus et al. (2012)). In particular, at national
level, there is often an additional downstream drift
This article is part of the Topical Collection on the /9th Joint Numerical
Sea Modelling Group Conference, Florence, Italy, 17-19 October 2018
Responsible Editor: Martin Verlaan
X Thorger Brüning
thorger.bruening@bsh.de
Bundesamt für Seeschifffahrt und Hydrographie (BSH),
Bernhard-Nocht-Straße 78, 20359 Hamburg, Germany
Published online: 24 February 2020
prediction component of great importance in crisis situa-
tions such as oil spills (Broström et al. 2011; Maßmann
st al. 2014) or search and rescue operations (Breivik
et al. 2013). Obviously the current data of the operational
circulation models are the most important input for the drift
orediction. Due to the lack of direct current measurements,
operational current forecasts are regularly validated indi-
rectly by comparing drift model results with observations
from surface drifters (Callies et al. 2017). Unfortunately,
such drifter experiments usually take place only under nor-
mal conditions, while drift predictions often have their
highest relevance in extreme situations, whereas the qual-
ity of current forecasts during normal conditions cannot
easıly be transferred to extreme situations. Accidentally,
instabilities resulting in unrealistic currents in the
CMEMS-Baltic MFC NRT-forecast product (Le Traon
et al. 2017) during storm events limited in time and space
have been noticed, which were not detected by the very
sxtensive, usual calibration and validation procedures of
the Baltic MFC including the validation of extreme sea
level events (CMEMS (n.d.); Golbeck et al. (2015)).
More detailed analyses of these events showed that be-
neath current profiles especially diffusivity profiles were
unrealistic, so that the turbulence model was considered
ZA Springer
