source: NEMO/trunk/doc/latex/NEMO/subfiles/chap_misc.tex @ 11543

Last change on this file since 11543 was 11543, checked in by nicolasmartin, 23 months ago

Implementation of convention for labelling references + files renaming
Now each reference is supposed to have the information of the chapter in its name
to identify quickly which file contains the reference (\label{$prefix:$chap_...)

Rename the appendices from 'annex_' to 'apdx_' to conform with the prefix used in labels (apdx:...)
Suppress the letter numbering

File size: 22.1 KB
4% ================================================================
5% Chapter --- Miscellaneous Topics
6% ================================================================
7\chapter{Miscellaneous Topics}
14% ================================================================
15% Representation of Unresolved Straits
16% ================================================================
17\section{Representation of unresolved straits}
20In climate modeling, it often occurs that a crucial connections between water masses is broken as
21the grid mesh is too coarse to resolve narrow straits.
22For example, coarse grid spacing typically closes off the Mediterranean from the Atlantic at
23the Strait of Gibraltar.
24In this case, it is important for climate models to include the effects of salty water entering the Atlantic from
25the Mediterranean.
26Likewise, it is important for the Mediterranean to replenish its supply of water from the Atlantic to
27balance the net evaporation occurring over the Mediterranean region.
28This problem occurs even in eddy permitting simulations.
29For example, in ORCA 1/4\deg several straits of the Indonesian archipelago (Ombai, Lombok...)
30are much narrow than even a single ocean grid-point.
32We describe briefly here the two methods that can be used in \NEMO\ to handle such
33improperly resolved straits. The methods consist of opening the strait while ensuring
34that the mass exchanges through the strait are not too large by either artificially
35reducing the cross-sectional area of the strait grid-cells or, locally increasing the
36lateral friction.
38% -------------------------------------------------------------------------------------------------------------
39%       Hand made geometry changes
40% -------------------------------------------------------------------------------------------------------------
41\subsection{Hand made geometry changes}
44The first method involves reducing the scale factor in the cross-strait direction to a
45value in better agreement with the true mean width of the strait
46(\autoref{fig:MISC_strait_hand}).  This technique is sometime called "partially open face"
47or "partially closed cells".  The key issue here is only to reduce the faces of $T$-cell
48(\ie\ change the value of the horizontal scale factors at $u$- or $v$-point) but not the
49volume of the $T$-cell.  Indeed, reducing the volume of strait $T$-cell can easily produce
50a numerical instability at that grid point which would require a reduction of the model
51time step.  Thus to instigate a local change in the width of a Strait requires two steps:
55\item Add \texttt{e1e2u} and \texttt{e1e2v} arrays to the \np{cn\_domcfg} file. These 2D
56arrays should contain the products of the unaltered values of: $\texttt{e1u}*\texttt{e2u}$
57and $\texttt{e1u}*\texttt{e2v}$ respectively. That is the original surface areas of $u$-
58and $v$- cells respectively.  These areas are usually defined by the corresponding product
59within the \NEMO\ code but the presence of \texttt{e1e2u} and \texttt{e1e2v} in the
60\np{cn\_domcfg} file will suppress this calculation and use the supplied fields instead.
61If the model domain is provided by user-supplied code in \mdl{usrdef\_hgr}, then this
62routine should also return \texttt{e1e2u} and \texttt{e1e2v} and set the integer return
63argument \texttt{ie1e2u\_v} to a non-zero value. Values other than 0 for this argument
64will suppress the calculation of the areas.
66\item Change values of \texttt{e2u} or \texttt{e1v} (either in the \np{cn\_domcfg} file or
67via code in  \mdl{usrdef\_hgr}), whereever a Strait reduction is required. The choice of
68whether to alter \texttt{e2u} or \texttt{e1v} depends. respectively,  on whether the
69Strait in question is North-South orientated (\eg\ Gibraltar) or East-West orientated (\eg
75The second method is to increase the viscous boundary layer thickness by a local increase
76of the fmask value at the coast. This method can also be effective in wider passages.  The
77concept is illustarted in the second part of  \autoref{fig:MISC_strait_hand} and changes
78to specific locations can be coded in \mdl{usrdef\_fmask}. The \forcode{usr_def_fmask}
79routine is always called after \texttt{fmask} has been defined according to the choice of
80lateral boundary condition as discussed in \autoref{sec:LBC_coast}. The default version of
81\mdl{usrdef\_fmask} contains settings specific to ORCA2 and ORCA1 configurations. These are
82meant as examples only; it is up to the user to verify settings and provide alternatives
83for their own configurations. The default \forcode{usr_def_fmask} makes no changes to
84\texttt{fmask} for any other configuration.
88  \begin{center}
89    \includegraphics[width=\textwidth]{Fig_Gibraltar}
90    \includegraphics[width=\textwidth]{Fig_Gibraltar2}
91    \caption{
92      \protect\label{fig:MISC_strait_hand}
93      Example of the Gibraltar strait defined in a $1^{\circ} \times 1^{\circ}$ mesh.
94      \textit{Top}: using partially open cells.
95      The meridional scale factor at $v$-point is reduced on both sides of the strait to account for
96      the real width of the strait (about 20 km).
97      Note that the scale factors of the strait $T$-point remains unchanged.
98      \textit{Bottom}: using viscous boundary layers.
99      The four fmask parameters along the strait coastlines are set to a value larger than 4,
100      \ie\ "strong" no-slip case (see \autoref{fig:LBC_shlat}) creating a large viscous boundary layer that
101      allows a reduced transport through the strait.
102    }
103  \end{center}
109  \begin{center}
110    \includegraphics[width=\textwidth]{Fig_closea_mask_example}
111    \caption{
112      \protect\label{fig:MISC_closea_mask_example}
113      Example of mask fields for the closea module. \textit{Left}: a
114      closea\_mask field; \textit{Right}: a closea\_mask\_rnf
115      field. In this example, if ln\_closea is set to .true., the mean
116      freshwater flux over each of the American Great Lakes will be
117      set to zero, and the total residual for all the lakes, if
118      negative, will be put into the St Laurence Seaway in the area
119      shown.
120    }
121  \end{center}
125% ================================================================
126% Closed seas
127% ================================================================
128\section[Closed seas (\textit{closea.F90})]
129{Closed seas (\protect\mdl{closea})}
132Some configurations include inland seas and lakes as ocean
133points. This is particularly the case for configurations that are
134coupled to an atmosphere model where one might want to include inland
135seas and lakes as ocean model points in order to provide a better
136bottom boundary condition for the atmosphere. However there is no
137route for freshwater to run off from the lakes to the ocean and this
138can lead to large drifts in the sea surface height over the lakes. The
139closea module provides options to either fill in closed seas and lakes
140at run time, or to set the net surface freshwater flux for each lake
141to zero and put the residual flux into the ocean.
143Prior to \NEMO\ 4 the locations of inland seas and lakes was set via
144hardcoded indices for various ORCA configurations. From \NEMO\ 4 onwards
145the inland seas and lakes are defined using mask fields in the
146domain configuration file. The options are as follows.
149\item{{\bfseries No ``closea\_mask'' field is included in domain configuration
150  file.} In this case the closea module does nothing.}
152\item{{\bfseries A field called closea\_mask is included in the domain
153configuration file and ln\_closea=.false. in namelist namcfg.} In this
154case the inland seas defined by the closea\_mask field are filled in
155(turned to land points) at run time. That is every point in
156closea\_mask that is nonzero is set to be a land point.}
158\item{{\bfseries A field called closea\_mask is included in the domain
159configuration file and ln\_closea=.true. in namelist namcfg.} Each
160inland sea or group of inland seas is set to a positive integer value
161in the closea\_mask field (see \autoref{fig:MISC_closea_mask_example}
162for an example). The net surface flux over each inland sea or group of
163inland seas is set to zero each timestep and the residual flux is
164distributed over the global ocean (ie. all ocean points where
165closea\_mask is zero).}
167\item{{\bfseries Fields called closea\_mask and closea\_mask\_rnf are
168included in the domain configuration file and ln\_closea=.true. in
169namelist namcfg.} This option works as for option 3, except that if
170the net surface flux over an inland sea is negative (net
171precipitation) it is put into the ocean at specified runoff points. A
172net positive surface flux (net evaporation) is still spread over the
173global ocean. The mapping from inland seas to runoff points is defined
174by the closea\_mask\_rnf field. Each mapping is defined by a positive
175integer value for the inland sea(s) and the corresponding runoff
176points. An example is given in
177\autoref{fig:MISC_closea_mask_example}. If no mapping is provided for a
178particular inland sea then the residual is spread over the global
181\item{{\bfseries Fields called closea\_mask and closea\_mask\_emp are
182included in the domain configuration file and ln\_closea=.true. in
183namelist namcfg.} This option works the same as option 4 except that
184the nonzero net surface flux is sent to the ocean at the specified
185runoff points regardless of whether it is positive or negative. The
186mapping from inland seas to runoff points in this case is defined by
187the closea\_mask\_emp field.}
190There is a python routine to create the closea\_mask fields and append
191them to the domain configuration file in the utils/tools/DOMAINcfg directory.
193% ================================================================
194% Sub-Domain Functionality
195% ================================================================
196\section{Sub-domain functionality}
199\subsection{Simple subsetting of input files via NetCDF attributes}
201The extended grids for use with the under-shelf ice cavities will result in redundant rows
202around Antarctica if the ice cavities are not active.  A simple mechanism for subsetting
203input files associated with the extended domains has been implemented to avoid the need to
204maintain different sets of input fields for use with or without active ice cavities.  This
205subsetting operates for the j-direction only and works by optionally looking for and using
206a global file attribute (named: \np{open\_ocean\_jstart}) to determine the starting j-row
207for input.  The use of this option is best explained with an example:
210\noindent Consider an ORCA1
211configuration using the extended grid domain configuration file: \ifile{eORCA1\}
212This file define a horizontal domain of 362x332.  The first row with
213open ocean wet points in the non-isf bathymetry for this set is row 42 (\fortran indexing)
214then the formally correct setting for \np{open\_ocean\_jstart} is 41.  Using this value as
215the first row to be read will result in a 362x292 domain which is the same size as the
216original ORCA1 domain.  Thus the extended domain configuration file can be used with all
217the original input files for ORCA1 if the ice cavities are not active (\np{ln\_isfcav =
218.false.}).  Full instructions for achieving this are:
221\item  Add the new attribute to any input files requiring a j-row offset, i.e:
223ncatted  -a open_ocean_jstart,global,a,d,41
226\item Add the logical switch \np{ln\_use\_jattr} to \nam{cfg} in the configuration
227namelist (if it is not already there) and set \np{.true.}
230\noindent Note that with this option, the j-size of the global domain is (extended
231j-size minus \np{open\_ocean\_jstart} + 1 ) and this must match the \texttt{jpjglo} value
232for the configuration. This means an alternative version of \ifile{eORCA1\} must
233be created for when \np{ln\_use\_jattr} is active. The \texttt{ncap2} tool provides a
234convenient way of achieving this:
237ncap2 -s 'jpjglo=292'
240The domain configuration file is unique in this respect since it also contains the value of \jp{jpjglo}
241that is read and used by the model.
242Any other global, 2D and 3D, netcdf, input field can be prepared for use in a reduced domain by adding the
243\texttt{open\_ocean\_jstart} attribute to the file's global attributes.
244In particular this is true for any field that is read by \NEMO\ using the following optional argument to
245the appropriate call to \np{iom\_get}.
251Currently, only the domain configuration variables make use of this optional argument so
252this facility is of little practical use except for tests where no other external input
253files are needed or you wish to use an extended domain configuration with inputs from
254earlier, non-extended configurations. Alternatively, it should be possible to exclude
255empty rows for extended domain, forced ocean runs using interpolation on the fly, by
256adding the optional argument to \texttt{iom\_get} calls for the weights and initial
257conditions. Experimenting with this remains an exercise for the user.
259% ================================================================
260% Accuracy and Reproducibility
261% ================================================================
262\section[Accuracy and reproducibility (\textit{lib\_fortran.F90})]
263{Accuracy and reproducibility (\protect\mdl{lib\_fortran})}
266\subsection[Issues with intrinsinc SIGN function (\texttt{\textbf{key\_nosignedzero}})]
267{Issues with intrinsinc SIGN function (\protect\key{nosignedzero})}
270The SIGN(A, B) is the \fortran intrinsic function delivers the magnitude of A with the sign of B.
271For example, SIGN(-3.0,2.0) has the value 3.0.
272The problematic case is when the second argument is zero, because, on platforms that support IEEE arithmetic,
273zero is actually a signed number.
274There is a positive zero and a negative zero.
276In \fninety, the processor was required always to deliver a positive result for SIGN(A, B) if B was zero.
277Nevertheless, in \fninety, the processor is allowed to do the correct thing and deliver ABS(A) when
278B is a positive zero and -ABS(A) when B is a negative zero.
279This change in the specification becomes apparent only when B is of type real, and is zero,
280and the processor is capable of distinguishing between positive and negative zero,
281and B is negative real zero.
282Then SIGN delivers a negative result where, under \fninety rules, it used to return a positive result.
283This change may be especially sensitive for the ice model,
284so we overwrite the intrinsinc function with our own function simply performing :   \\
285\verb?   IF( B >= 0.e0 ) THEN   ;   SIGN(A,B) = ABS(A)  ?    \\
286\verb?   ELSE                   ;   SIGN(A,B) =-ABS(A)     ?  \\
287\verb?   ENDIF    ? \\
288This feature can be found in \mdl{lib\_fortran} module and is effective when \key{nosignedzero} is defined.
289We use a CPP key as the overwritting of a intrinsic function can present performance issues with
290some computers/compilers.
293\subsection{MPP reproducibility}
296The numerical reproducibility of simulations on distributed memory parallel computers is a critical issue.
297In particular, within \NEMO\ global summation of distributed arrays is most susceptible to rounding errors,
298and their propagation and accumulation cause uncertainty in final simulation reproducibility on
299different numbers of processors.
300To avoid so, based on \citet{he.ding_JS01} review of different technics,
301we use a so called self-compensated summation method.
302The idea is to estimate the roundoff error, store it in a buffer, and then add it back in the next addition.
304Suppose we need to calculate $b = a_1 + a_2 + a_3$.
305The following algorithm will allow to split the sum in two
306($sum_1 = a_{1} + a_{2}$ and $b = sum_2 = sum_1 + a_3$) with exactly the same rounding errors as
307the sum performed all at once.
309   sum_1 \ \  &= a_1 + a_2 \\
310   error_1     &= a_2 + ( a_1 - sum_1 ) \\
311   sum_2 \ \  &= sum_1 + a_3 + error_1 \\
312   error_2     &= a_3 + error_1 + ( sum_1 - sum_2 ) \\
313   b \qquad \ &= sum_2 \\
315An example of this feature can be found in \mdl{lib\_fortran} module.
316It is systematicallt used in glob\_sum function (summation over the entire basin excluding duplicated rows and
317columns due to cyclic or north fold boundary condition as well as overlap MPP areas).
318The self-compensated summation method should be used in all summation in i- and/or j-direction.
319See \mdl{closea} module for an example.
320Note also that this implementation may be sensitive to the optimization level.
322\subsection{MPP scalability}
325The default method of communicating values across the north-fold in distributed memory applications (\key{mpp\_mpi})
326uses a \textsc{MPI\_ALLGATHER} function to exchange values from each processing region in
327the northern row with every other processing region in the northern row.
328This enables a global width array containing the top 4 rows to be collated on every northern row processor and then
329folded with a simple algorithm.
330Although conceptually simple, this "All to All" communication will hamper performance scalability for
331large numbers of northern row processors.
332From version 3.4 onwards an alternative method is available which only performs direct "Peer to Peer" communications
333between each processor and its immediate "neighbours" across the fold line.
334This is achieved by using the default \textsc{MPI\_ALLGATHER} method during initialisation to
335help identify the "active" neighbours.
336Stored lists of these neighbours are then used in all subsequent north-fold exchanges to
337restrict exchanges to those between associated regions.
338The collated global width array for each region is thus only partially filled but is guaranteed to
339be set at all the locations actually required by each individual for the fold operation.
340This alternative method should give identical results to the default \textsc{ALLGATHER} method and
341is recommended for large values of \np{jpni}.
342The new method is activated by setting \np{ln\_nnogather} to be true (\nam{mpp}).
343The reproducibility of results using the two methods should be confirmed for each new,
344non-reference configuration.
346% ================================================================
347% Model optimisation, Control Print and Benchmark
348% ================================================================
349\section{Model optimisation, control print and benchmark}
356Options are defined through the  \nam{ctl} namelist variables.
358\subsection{Vector optimisation}
360\key{vectopt\_loop} enables the internal loops to collapse.
361This is very a very efficient way to increase the length of vector calculations and thus
362to speed up the model on vector computers.
364% Add here also one word on NPROMA technique that has been found useless, since compiler have made significant progress during the last decade.
366% Add also one word on NEC specific optimisation (Novercheck option for example)
368\subsection{Control print}
370The \np{ln\_ctl} switch was originally used as a debugging option in two modes:
373\item{\np{ln\_ctl}: compute and print the trends averaged over the interior domain in all TRA, DYN, LDF and
374ZDF modules.
375This option is very helpful when diagnosing the origin of an undesired change in model results. }
377\item{also \np{ln\_ctl} but using the nictl and njctl namelist parameters to check the source of differences between
378mono and multi processor runs.}
381However, in recent versions it has also been used to force all processors to assume the
382reporting role. Thus when \np{ln\_ctl} is true all processors produce their own versions
383of files such as: ocean.output, layout.dat, etc.  All such files, beyond the the normal
384reporting processor (narea == 1), are named with a \_XXXX extension to their name, where
385XXXX is a 4-digit area number (with leading zeros, if required). Other reporting files
386such as run.stat (and its netCDF counterpart: and tracer.stat contain global
387information and are only ever produced by the reporting master (narea == 1). For version
3884.0 a start has been made to return \np{ln\_ctl} to its original function by introducing
389a new control structure which allows finer control over which files are produced. This
390feature is still evolving but it does already allow the user to: select individually the
391production of run.stat and tracer.stat files and to toggle the production of other files
392on processors other than the reporting master. These other reporters can be a simple
393subset of processors as defined by a minimum, maximum and incremental processor number.
395Note, that production of the run.stat and tracer.stat files require global communications.
396For run.stat, these are global min and max operations to find metrics such as the gloabl
397maximum velocity. For tracer.stat these are global sums of tracer fields. To improve model
398performance these operations are disabled by default and, where necessary, any use of the
399global values have been replaced with local calculations. For example, checks on the CFL
400criterion are now done on the local domain and only reported if a breach is detected.
402Experienced users may wish to still monitor this information as a check on model progress.
403If so, the best compromise will be to activate the files with:
406     sn_cfctl%l_config = .TRUE.
407       sn_cfctl%l_runstat = .TRUE.
408       sn_cfctl%l_trcstat = .TRUE.
411and to use the new time increment setting to ensure the values are collected and reported
412at a suitably long interval. For example:
415       sn_cfctl%ptimincr  = 25
418will carry out the global communications and write the information every 25 timesteps. This
419increment also applies to the time.step file which is otherwise updated every timestep.
421% ================================================================
Note: See TracBrowser for help on using the repository browser.