newsletter6
Dear GOTM user,
the last GOTM update is already about 13 months ago, and thus we are
happy to inform you that
we have now released GOTM 2.4 with several new physical and technical
features.
These new features are:
-
Generic two-equation model according to Umlauf and Burchard
[2001].
-
Various TVD schemes for vertical advection.
-
Arbitrary spatial grid to be read from file.
Furthermore, we prepared four new test cases, all documented by manuscripts
submitted
to peer-reviewed international journals:
-
Lago Maggiore 1995
-
Northern North Sea 1998
-
Liverpool Bay 1999
-
Test case for generic model
This long period in which no update has been made does not imply that GOTM
has been loosing
some of its dynamics. The contrary is true. The new physical features
and the new test cases are
rather complex and thus took us quite some time to develop and test.
We have actively participated in two
projects funded by the European Community in which turbulence modelling
plays a key role (PROVESS
and CARTUM).
The user group is constantly growing, by today we count 112 members. Each
day, we
have about 20 logins from all over the world, adding up to more than
8000 logins since the first GOTM
released in June 1999. GOTM users send several mails per week to the
GOTM developers
with questions and suggestions concerning GOTM. We are always happy
to receive and answer these mails
since such we learn a lot about what the users do with GOTM and where
the problems are.
We publish this new version just before the final workshop of the CARTUM
project, which is held in Brussels,
Belgium, from December 3 - 5, 2001. GOTM profited a lot from the CARTUM
project and we feel
that it was in a certain sense also the other way around. This new
GOTM version and the GOTM scenarios
will be copied to a CD-ROM, which will be attached to
the CARTUM book, the major deliverable of the CARTUM project. In the
part "Modelling and boundary layers",
four sections will be dedicated to GOTM, entitled "Physical concepts",
"Computational concepts", "Selected results"
and "Coupling with 3D models". We will keep you informed about the
status of this book.
GOTM depends very much on the co-operation and feedback from its users.
So, please write us what your experience with GOTM is, tell us about any
problems, send us your publications based on GOTM, provide links to GOTM
from your home pages and recommend GOTM to interested colleagues.
We hope that you enjoy our new GOTM version and the new scenarios.
Kind regards,
The GOTM Team.
New features
Generic two-equation model
The basic idea of the generic two-equation model is to allow for a wide
range of length scale related transport equations of the form knem
with the turbulent kinetic energy k and the dissipation rate e.
The structure of the dynamical equation for this new quantity is adopted
from the e-equation. Via the relation L~k3/2e-1,
the macro length scale needed can be calculated. For different choices
of n and m, classical two-equations model can be retained:
n=0 and m=1 gives the k-e
model, n=-1 and m=1 gives the k-w
model with w=k-1e,
and n=5/2 and m=-1 gives the k-kL model by
Mellor and Yamada [1982]. The advantage
of the generic two-equation model is its greater flexibility since for
each pair (n,m) a new turbulence model is designed.
This method offers a rational way of determining the optimal choice
for the second variable in two-equation models.
The physical
relevance of such a model is guaranteed by choosing the empirical parameters
appropriately. For this, we have implemented two methods:
1. The empirical parameters are tuned such that the same log-law behaviour
than for the k-e model is retained. An extensive
report by Hans Burchard including this
generic two-equation model can be downloaded as postscript
or pdf
file.
2. The empirical parameters are individually tuned to various observations
of boundary layer flow. In unstratified boundary layer flow, five basic
observations are used for fixing five empirical parameters. When also considering
shear-free turbulent flow generated by an oscillating grid, the relation
between n and m is fixed as well. A submitted manuscript
by Lars Umlauf and Hans
Burchard can be downloaded as postscript
or pdf
file.
TVD advection schemes
The Flux Corrected Transport (FCT) scheme which has been used in GOTM so
far has been replaced by a number of higher order schemes including some
TVD (Total variation Diminishing) limiters. The available schemes are:
-
First-order upstream (monotone but diffusive)
-
Third-order polynomial (accurate but not monotone)
-
TVD with Superbee limiter (monotone, but anti-diffusive)
-
TVD with MUSCL limiter (monotone, but slightly diffusive)
-
TVD with ULTIMATE QUICKEST (monotone and very little diffusive, the best
we have)
These schemes will be used in the sediment module for the sinking of particle
concentrations and in all tracer concentration routines when a vertical
displacement velocity is given.
The performance of these schemes can be easily tested within the vertical
_advection scenario in which a thermocline is moved up and down by a prescribed
vertical velocity. The maximum Courant number is chosen to be larger than
one, which is no problem for our schemes since the vertical-advection subroutine
first calculates the Courant number and then iterates the advection procedure
with a sufficient number of cycles. It should be noted that these advection
schemes are extracted from the three-dimensional circulation model GETM
(General Estuarine Transport Model) about which a comprehensive report
by Hans Burchard and Karsten
Bolding is in preparation.
Vertical grid read from file
The turbulence module of GOTM is now integrated in a number of three-dimensional
models, among them are the z-coordinate models Modular Ocean Model
(MOM, contact Encho Demirov)
and Hamburg Ocean Primitive Equation Model (HOPE, contact Johann
Jungclaus and Raimon Hernandez-Roura).
In order to study the impact of various vertical grids on the model performance,
we have implemented the possibility to read any non-equidistant grid from
a file and to use it inside GOTM. This helped us for example to find out
that a vertical resolution of 20 m is too coarse for a world ocean model
with a k-e model used for vertical turbulent
exchange. In co-operation with the MPI in Hamburg, we found that at least
10 m are needed for vertical resolution.
For the scenarios liverpool_bay and ows_papa, grid files are provided. For
liverpool_bay, set grid_method in gotmmean.inp to 1 for
s-coordinates
and for ows_papa set grid_method in gotmmean.inp to 2 for
z-coordinates.
New scenarios
The following new scenarios have been prepared for GOTM now:
Lago Maggiore 1995
In December 1995, turbulence observations have been carried out under strongly
convective condition in the Lago Maggiore near the shore of Ispra at a
water depth of 42 m. These observations have been conducted by a group
from the Joint Research Centre of the European
Communities led by Adolf Stips.
The simulation results with GOTM are reported in a manuscript by Adolf
Stips, Hans Burchard, Karsten Bolding and Walter Eifler which has been
submitted to Ocean Dynamics.
Northern North Sea 1998
In the framework of the European Communities PROVESS
(Processes of Vertical Exchange in Shelf Seas, MAS3-CT97-0025) project,
intensive observations have been carried out in the Northern North Sea
during two months from September - November 1998. Various turbulence observations
have been made as well. During 24 hours, the turbulent dissipaption rate
has been observed independently in parallel from two ship with two different
types of shear probes. Many of these observations have been prepared for
forcing and validating a GOTM simulation. Depending on the users intention,
GOTM can be run for the whole two months or only during the period in which
dissipation rate observations are available. Two manuscripts based on these
simulations have been submitted to international journals, one by Karsten
Bolding, Hans Burchard, Thomas Pohlmann, and Adolf Stips on annual and
seasonal simulations ([ps],
[pdf])
and one by Hans Burchard, Karsten Bolding, Tom Rippeth, Adolf Stips, John
Simpson, and Jürgen Sündermann on short term turbulence dynamics
([ps],
[pdf]).
Liverpool Bay 1999
In summer 1999, detailed mean flow and turbulence observations have been
carried out in Liverpool Bay in the eastern Irish Sea. This area is strongly
influences by tides and the river run-off from several rivers in England.
Thus, a persistent horizontal density gradient is present, leading in conjunction
with the tides to the SIPS (strain-induced periodic stratification) phenomenon,
described in detail by Simpson et al. 1990 (Estuaries 26, 1579-1590).
The measurements have been carried out by Tom Rippeth, John Simpson and
Neil Fisher and are reported in detail by Rippeth et al. 2001 (Journal
Phys. Oceanogr. 31, 2458-2471). During the CARTUM
workshop in Marseille, France in March 2000, John Simpson had offered these
data to the numerical modelling community. The GOTM Team took the chance
to use these data for testing, whether such turbulence model as the k-e
model are able to quantify turbulent mixing in such complex coastal flow
with strain-induced convective mixing. The simulations proved that we could
reproduce the dissipation rate data without any tuning of GOTM (we did
not receive the dissipation rate data before finishing the simulations).
We are now confident that estuarine dynamics (which are e.g. responsible
for estuarine turbidity zones) can be simulated with sufficient accuracy
with this modelling level.
Test case for generic model
This test case has been set up to provide a test bed for the generic two-equation
model with the methods presented by Umlauf and Burchard [2001]. Here a
constant surface stress is applied to a homogeneous water column. A positive
flux of turbulent kinetic energy, a simple way of modelling
the effect of breaking surface waves,
is applied as well, such that a near-surface
layer of enhanced turbulence is built up. It can be clearly seen how the
length scale in this enhance turbulence layer has a slope of about 0.2
whereas below a shear layer is built up with a slope of 0.4, which is the
van Karman number, a value typical for the logarithmic layer.