source: trunk/LMDZ.MARS/doc/io.tex @ 3537

\chapter{Input/Output}
\label{sc:io}

\section{NetCDF format}

%%%%%%%%%%%%%%%%%%%%%%%%%%
GCM input/output data are written in {\bf NetCDF} format
(Network Common Data Form). NetCDF is an interface used to store and access
geophysical data, and a library that provides an implementation of this
interface. The NetCDF library also defines a machine-independent format for
representing scientific data. 
Together, the interface, library and format support the creation, access and
sharing of scientific data. NetCDF was developed at the Unidata Program Center
in Boulder, Colorado. The freely available source can be obtained from
the Unidata website:
\begin{verbatim}
http://www.unidata.ucar.edu/software/netcdf
\end{verbatim}


%%%%%%%%%%%%%%%%%%%%%%%%%%

A data set in NetCDF format is a single file, as it is self-descriptive.

\subsection{NetCDF text representation: ncdump}

This utility is included in the NetCDF library.
It generates the CDL representation (text format of the file content) to the standard output
from the NetCDF file specified as input. 

\paragraph{Main options for the ncdump command}

\begin{center}
{\it ncdump diagfi.nc}
\end{center}

\noindent
dump contents of NetCDF file {\tt diagfi.nc} to standard output
(i.e. the screen).

\begin{center}
{\it ncdump -c diagfi.nc}
\end{center}

\noindent
Displays the {\bf coordinate} variable values (variables which are also
dimensions), as well as the declarations, variables and attribute values.
The values of the non-coordinate variable data are not displayed at
the output.

\begin{center}
{\it ncdump -h diagfi.nc}
\end{center}

\noindent
Shows only the informative header of the file, which is the declaration
of the dimensions, variables and attributes, but not the values of these
variables. The output is identical to that in option {\bf -c} except for
the fact that the coordinated variable values are not included.

\begin{center}
{\it ncdump -v var1,...,varn diagfi.nc}
\end{center}

\noindent
The output includes the specific variable values,
as well as all the dimensions, variables and attributes.
More that one variable can be specified in the list following this option.
The list must be a simple argument for the command, and must not contain any
spaces. If no variable is specified, the command displays all the values of
the variables in the file by default.  


\subsection{Graphic visualization of NetCDF files using GrAds}

GrAdS (The Grid Analysis and Display System) is a graphic software developed
by Brian Doty at the "Center for Ocean-Land-Atmosphere (COLA)".

One of its functions is to enable data stored in NetCDF format to be
visualized directly. In figure~\ref{fg:grads} for example, we can see the
GrADS visualization of the temperature data at a given moment.
%
\begin{figure}
\centering
\includegraphics[width=0.5\textwidth,angle=270]{Fig/grads.pdf}
\caption{Example of temperature data at a given time using
GrADS visualization\label{fg:grads}}
\end{figure}
%
However, unlike NetCDF, GrADS only recognizes files where all the variables are stored on the same horizontal grid.
These variables can be in 1, 2, 3 or 4 dimensions (X,Y,Z and t).\\

GrADS can also be obtained from:
\begin{verbatim}
http://grads.iges.org/grads/
\end{verbatim}

\subsection{Graphic visualization of NetCDF files using Ferret}

Ferret may also be used to visualize the contents of NetCDF files. Download intruction and documentation are available from the official website:
\begin{verbatim}
https://ferret.pmel.noaa.gov/Ferret/
\end{verbatim} 

\section{Input and parameter files}

%{\bf \it Examples of initialization files can be found in directory
%\begin{verbatim}$PATH1/LMDZ.MARS/deftank \end{verbatim}}
\label{loc:entrees}

The (3D version of the) GCM requires
the input of two initialization files (in NetCDF format):\\
-{\bf start.nc}
contains the initial states of the dynamical variables.\\
-{\bf startfi.nc}
contains the initial states of the physical variables.\\
Note that collections of initial states can be retreived at:\\
\verb+http://www.lmd.jussieu.fr/~lmdz/planets/mars/starts+ \\
Extracting {\tt start.nc} and {\tt startfi.nc} from these archived
requires using program {\tt newstart}, as described in
section~\ref{sc:newstart}.\\

\noindent
To run, the GCM also requires the  four following
parameter files (ascii text files):\\
-{\bf run.def} the parameters of the dynamical part of the program,
and the temporal integration of the model.\\
-{\bf callphys.def} the parameters for calling the physical part.\\
-{\bf traceur.def} the names of the tracer to use.\\
-{\bf z2sig.def}
 the vertical distribution of the atmospheric layers.\\
Examples of these parameter files can be found in the
\verb+LMDZ.MARS/deftank+ directory.

\subsection{run.def}
\label{vb:run.def}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% run.def: les param sont lus dans dyn3d/defrun.F
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

A typical {\tt run.def} file is given as an example below.
The choice of variables to be set is simple (e.g.
 {\tt nday} number of modeled days to run),
while the others do not need to be changed for normal use.\\
The format of the {\tt run.def} file is quite straightforward
(and flexible): values given to parameters must be given as:
\begin{verbatim}
  parameter = value
\end{verbatim}
Any blank line or line beginning with symbol {\bf \#} is
a comment, and instruction lines may be written in any order.
Moreover, not specifying a parameter/value set (e.g. deleting it
or commenting it out) means you want the GCM to use a default built-in value.
Additionally, one may use a specific keyword {\bf INCLUDEDEF} to specify
another (text) file in which to also read values of parameters; e.g.:
\begin{verbatim}
INCLUDEDEF=callphys.def
\end{verbatim}


\noindent Here are some details about some of the parameters which may be
set in {\tt run.def}:
\begin{itemize}
\item {\bf day\_step}, the number of dynamical steps per day to use for
the time integration. This needs to be large enough for the model
to remain stable (this is related to the CFL stability criterion
which essentially depends on the horizontal resolution of the model).
On Mars, in theory, the GCM can run with
{\tt day\_step}=480 using the 64$\times$48 grid, but model stability
improves when this number is higher: {\tt day\_step}=960 is recommended
 when using the 64$\times$48 grid. According to the CFL criterion,
{\tt day\_step} should vary in proportion with the resolution: for example
{\tt day\_step}=480 using the 32$\times$24 horizontal resolution.
Note that {\tt day\_step} must also be divisible by {\tt iperiod}.

\item {\bf tetagdiv, tetagrot, tetatemp} control the dissipation intensity.
It is better to limit the dissipation intensity
(tetagdiv, tetagrot, tetatemp should not be too low).
However the model diverges if tetagdiv, tetagrot, tetatemp are too high,
especially if there is a lot of dust in the atmosphere. \\
Example used with nitergdiv=1 and  nitergrot=niterh=2 : \\
- using the 32$\times$24 grid tetagdiv=6000~s ; tetagrot=tetatemp=30000~s \\
- using the 64$\times$48 grid: tetagdiv=2500~s ; tetagrot=tetatemp=5000~s

\item {\bf idissip} is the time step used for the dissipation:
dissipation is computed and added every {\tt idissip} dynamical
time step. If {\tt idissip} is
too short, the model waste time in these calculations. But if idissip is too
long, the  dissipation will not be parametrized correctly and the  model will
be more likely to diverge. 
A check must be made, so that:
{\tt idissip}~$<$~{\tt tetagdiv}$\times${\tt daystep}/88775
(same rule for {\tt tetagrot} and {\tt tetatemp}).
This is tested automatically during the run.

\item {\bf iphysiq} is the time step used for the physics:
physical tendencies are computed every {\tt iphysiq} dynamical time step.
In practice, we
usually set the physical time step to be of the order of half an hour. 
We thus generally set {\tt iphysiq}= {\tt day\_step}/48

\end{itemize}

\noindent
{\it Example of run.def file: }
{\footnotesize
\input{input/run.tex}
}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%


\subsection{callphys.def}
\label{sc:callphys.def}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% callphys.def: les param sont lus dans phymars/inifis.F
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
The {\tt callphys.def} file (along the same format
as the {\tt run.def} file) contains parameter/value sets
for the physics.\\

 
\noindent
{\it Example of callphys.def file: }
{\footnotesize
\input{input/callphys.tex}
}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

\subsection{traceur.def}
\label{sc:traceur.def}
Tracers in input ({\tt start.nc} and {\tt startfi.nc}) and output
files ({\tt restart.nc} and {\tt restartfi.nc}) are stored using
individual tracer names (e.g. {\tt co2} for CO2 gas, {\tt h2o\_vap}
for water vapour, {\tt h2o\_ice} for water ice, ...).\\
The first line of the {\tt traceur.def} file (an ASCII file) must
contain the number of tracers to load and use (this number should
be the same as given to the {\tt -t} option of the {\tt makegcm}
script when the GCM was compiled), followed by the tracer names
(one per line). Note that if the corresponding tracers are not
found in input files {\tt start.nc} and {\tt startfi.nc}, then the
tracer is initialized to zero.\\


\noindent {\it Example of a traceur.def file}:
with CO2, dust distribution moments, (water ice) cloud condensation nuclei moments, water vapour and water ice tracers
{\footnotesize
\begin{verbatim}
7
co2
dust_number
dust_mass
ccn_number
ccn_mass
h2o_ice
h2o_vap
\end{verbatim}
}

\subsection{z2sig.def}
The {\tt Z2sig.def} file contains the pseudo-altitudes
(in km) at which the user wants to set the vertical levels.\\
Note that levels should be unevenly spread, with a higher resolution
near the surface in order to capture the rapid variations of variables
there. It is recommended to use the altitude levels as set in the
{\tt z2sig.def} file provided in the {\tt deftank} directory.\\


\noindent
{\it Example of  z2sig.def file
(this version for 49 layers between  0 and 300~km):}
{\footnotesize
\input{input/z2sig.tex}
}

\subsection{Initialization files: start and startfi}

%
\begin{figure}[h]
\centering
\framebox[0.8\textwidth][c]{\includegraphics[width=0.7\textwidth]{Fig/netcdf.pdf}}
\caption{Organization of NetCDF files \label{fg:netcdf}}
\end{figure}
%
Files {\tt start.nc} and {\tt startfi.nc}, like all the NetCDF files of
the GCM,
are constructed on the same model (see NetCDF file composition,
figure~\ref{fg:netcdf}). They contain:\\
- a header with a ``control'' variable followed by a series of variables
defining the (physical and dynamical) grids \\
- a series of non temporal variables that give information about surface 
conditions on the planet.\\
- a ``time'' variable giving the values of the different instants at which 
the temporal variables are stored 
(a single time value (t=0) for start,
as it describes the dynamical initial states,
and no time values for startfi, as it describes only a physical state).\\

To dump (in text format) the contents of a {\tt start.nc} file using the
{\tt ncdump} command:\\

\noindent
{\it ncdump -h start.nc}\\
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% List START
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\input{input/dyn_list.tex}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

\noindent
List of contents of a {\tt startfi.nc} file:\\

\noindent
{\it ncdump -h startfi.nc}\\
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% List startfi
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\input{input/fi_list.tex}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%



%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%    Description des start et startfi
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

\paragraph{Physical and dynamical headers}

There are two types of headers: one for the physical headers,
and one for the dynamical headers.
The headers always begin with a ``control' variable
(described below), that is allocated differently in the physical and
dynamical parts.
The other variables in the header concern the (physical and dynamical) grids.
They are the following:\\

\noindent
the horizontal coordinates\\
- {\bf rlonu}, {\bf rlatu}, {\bf rlonv}, {\bf rlatv} for the dynamical part,\\
- {\bf lati}, {\bf long} for the physical part,\\

\noindent
the coefficients for passing from the physical grid to the dynamical grid\\
- {\bf cu},{\bf cv} only in the dynamical header\\

\noindent
and finally, the grid box areas\\
- {\bf aire} for the dynamical part,\\
- {\bf area} for the physical part.\\

\paragraph{Surface conditions}

The surface conditions are mostly given in the physical NetCDF files by
variables:\\
- {\bf phisfi} for the initial state of surface geopotential,\\
- {\bf albedodat} for the bare ground albedo,\\
- {\bf inertiedat} for the surface thermal inertia,\\
- {\bf zmea}, {\bf zstd}, {\bf zsig}, {\bf zgam} and {\bf zthe} for
  the subgrid scale topography.\\

\noindent
For the dynamics:\\
- {\bf physinit} for the initial state of surface geopotential\\

\noindent
Remark: variables {\bf phisfi} and {\bf physinit} contain the same information
(surface geopotential), but {\bf phisfi} gives the geopotential values on the
physical grid, while {\bf physinit} give the values on the dynamical grid.\\

\paragraph{Physical and dynamical state variables}
To save disk space, the initialization files store the variables used by
the model, rather than the ``natural'' variables.\\

\noindent
For the dynamics:
\begin{description}
\item - {\bf ucov} and {\bf vcov} the covariant winds\\
These variables are linked to the ``natural'' winds by\\
\verb+ucov = cu * u+ and \verb+vcov = cv * v+
\item - {\bf teta} the potential temperature,\\
  or more precisely, the potential enthalpy linked to temperature {\bf T} by
  $\theta = T\dep{\frac{P}{Pref}}^{-K}$
\item - the tracers,
\item - {\bf ps} surface pressure.
\item - {\bf masse} the atmosphere mass in each grid box.
\end{description}

\noindent
``Vectorial'' variables {\bf ucov} and {\bf vcov} are stored on
``staggered'' grids u and v respectively (in the dynamics)
(see section \ref{fg:grid}).\\
Scalar variables {\bf h}, {\bf q} (tracers), {\bf ps}, {\bf masse} are stored
on the ``scalar'' grid of the dynamical part.\\

\noindent
For the physics:
\begin{description}
\item - {\bf co2ice} surface dry ice,
\item - {\bf tsurf} surface temperature,
\item - {\bf tsoil} temperatures at different layers under the surface,
\item - {\bf emis} surface emissivity,
\item - {\bf q2} wind variance,\\
or more precisely, the square root of the turbulent kinetic energy.
\item - the surface ``tracer'' budget
 (kg.m$^{-2}$),\\
\end{description}

\noindent
All these variables are stored on the ``physical'' grid
(see section \ref{fg:grid}).\\
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

\paragraph{The ``control'' array}

\indent
Both physical and dynamical headers of the GCM NetCDF files start with
a {\bf controle} variable. This variable is an array of 100 reals (the vector
called {\tt tab\_cntrl} in the program), which contains the program control
parameters. 
Parameters differ between the physical and dynamical sections, and examples
of both are listed below. The contents of table {\tt tab\_cntrl} can also
be checked with the command {\tt ncdump -ff -v controle}.\\

\noindent
{\bf The "control" array in the header of a dynamical NetCDF file:
start}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% tab_cntrl (dynamique) dans dyn3d/inimomo.F
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\input{input/dyn_cntl.tex}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

\noindent
{\bf The "controle" array in the header of a physical NetCDF file:
startfi.nc}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% tab_cntrl (physique) dans phymars/iniwritefi.F
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\input{input/fi_cntl.tex}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
 
\newpage
\section{Output files}

\subsection{NetCDF restart files - restart.nc and restartfi.nc}
These files are of the exact same format as {\tt start.nc} and
{\tt startfi.nc}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Description des fichiers de Sortie
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

\subsection{ NetCDF file - diagfi.nc}
NetCDF file {\tt diagfi.nc} stores the instantaneous physical variables
throughout the simulation at regular intervals
(set by the value of parameter {\tt ecritphy} in 
parameter file {\tt run.def}; note that {\tt ecritphy} should be a
multiple of {\tt iphysiq} as well as a divisor of {\tt day\_step}).

\noindent
{\bf Any variable from any sub-routine of the physics can be stored
by calling subroutine} {\tt writediagfi}.
Moreover, one may add a {\tt diagfi.def} file containing only the names
of variables to output (one per line) in the directory where the GCM is
run, in order to have only thoses listed outputed in the {\tt diagfi.nc}
file.\\

\noindent
Illustrative example of the contents of a {\tt diagfi.nc}
file (using ncdump):\\
\noindent
{\it ncdump -h diagfi.nc}\\
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% List DIAGFI
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% temporaire!!!
\input{input/diag_list.tex}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

\noindent
The structure of the file is thus as follows: 
\begin{description}
\item- the dimensions 
\item- variable ``time'' containing the time of the timestep stored in the
 file (in Martian days since the beginning of the run)
\item- variable ``control'' containing many parameters, as described above.
\item- from `` rhonu'' to 'phisinit'': a list of data describing the
 geometrical coordinates of the data file, plus the surface topography
\item- finally, all the 2D or 3D data stored in the run. 
\end{description}


\subsection{Stats files}

As an option ({\tt stats} must be set to {\tt .true.} in {\tt callphys.def}),
the model can accumulate any
variable from any subroutine of the physics by calling
subroutine \verb+ wstat+ 
\\ \\ 
\noindent
This save is performed at regular intervals 12 times a day.
An average of the daily evolutions over the whole run is calculated
(for example, for a 10 day run, the averages of the variable values at
0hTU, 2hTU, 4hTU,...24hTU are calculated), along with RMS standard
deviations of the variables. This ouput is given in 
file {\tt stats.nc}.\\


\noindent
Illustrative example of the contents of a {\tt stats.nc} file (using ncdump):\\
\noindent
{\it ncdump -h stats.nc}\\
\input{input/stats_list.tex}

\noindent
The structure of the file is simillar to the {\tt diagfi.nc} file,
except that, as stated before, the average of variables are given for
12 times of the day and that RMS standard deviation are also provided.