source:NEMO/trunk/doc/latex/NEMO/subfiles/chap_STO.tex@10442

Last change on this file since 10442 was 10442, checked in by nicolasmartin, 22 months ago

Front page edition, cleaning in custom LaTeX commands and add index for single subfile compilation

• Use \thanks storing cmd to refer to the ST members list for 2018 in an footnote on the cover page
• NEMO and Fortran in small capitals
• Removing of unused or underused custom cmds, move local cmds to their respective .tex file
• Addition of new ones (\zstar, \ztilde, \sstar, \stilde, \ie, \eg, \fortran, \fninety)
• Fonts for indexed items: italic font for files (modules and .nc files), preformat for code (CPP keys, routines names and namelists content)
File size: 8.2 KB
Line
1\documentclass[../main/NEMO_manual]{subfiles}
2
3\begin{document}
4% ================================================================
5% Chapter stochastic parametrization of EOS (STO)
6% ================================================================
7\chapter{Stochastic Parametrization of EOS (STO)}
8\label{chap:STO}
9
10Authors: P.-A. Bouttier
11
12\minitoc
13
14\newpage
15
16The stochastic parametrization module aims to explicitly simulate uncertainties in the model.
17More particularly, \cite{Brankart_OM2013} has shown that,
18because of the nonlinearity of the seawater equation of state, unresolved scales represent a major source of
19uncertainties in the computation of the large scale horizontal density gradient (from T/S large scale fields),
20and that the impact of these uncertainties can be simulated by
21random processes representing unresolved T/S fluctuations.
22
23The stochastic formulation of the equation of state can be written as:
24\begin{equation}
25  \label{eq:eos_sto}
26  \rho = \frac{1}{2} \sum_{i=1}^m\{ \rho[T+\Delta T_i,S+\Delta S_i,p_o(z)] + \rho[T-\Delta T_i,S-\Delta S_i,p_o(z)] \}
27\end{equation}
28where $p_o(z)$ is the reference pressure depending on the depth and,
29$\Delta T_i$ and $\Delta S_i$ are a set of T/S perturbations defined as
30the scalar product of the respective local T/S gradients with random walks $\mathbf{\xi}$:
31\begin{equation}
32  \label{eq:sto_pert}
33  \Delta T_i = \mathbf{\xi}_i \cdot \nabla T \qquad \hbox{and} \qquad \Delta S_i = \mathbf{\xi}_i \cdot \nabla S
34\end{equation}
35$\mathbf{\xi}_i$ are produced by a first-order autoregressive processes (AR-1) with
36a parametrized decorrelation time scale, and horizontal and vertical standard deviations $\sigma_s$.
37$\mathbf{\xi}$ are uncorrelated over the horizontal and fully correlated along the vertical.
38
39
40\section{Stochastic processes}
41\label{sec:STO_the_details}
42
43The starting point of our implementation of stochastic parameterizations in NEMO is to observe that
44many existing parameterizations are based on autoregressive processes,
45which are used as a basic source of randomness to transform a deterministic model into a probabilistic model.
46A generic approach is thus to add one single new module in NEMO,
47generating processes with appropriate statistics to simulate each kind of uncertainty in the model
48(see \cite{Brankart_al_GMD2015} for more details).
49
50In practice, at every model grid point,
51independent Gaussian autoregressive processes~$\xi^{(i)},\,i=1,\ldots,m$ are first generated using
52the same basic equation:
53
54\begin{equation}
55  \label{eq:autoreg}
56  \xi^{(i)}_{k+1} = a^{(i)} \xi^{(i)}_k + b^{(i)} w^{(i)} + c^{(i)}
57\end{equation}
58
59\noindent
60where $k$ is the index of the model timestep and
61$a^{(i)}$, $b^{(i)}$, $c^{(i)}$ are parameters defining the mean ($\mu^{(i)}$) standard deviation ($\sigma^{(i)}$) and
62correlation timescale ($\tau^{(i)}$) of each process:
63
64\begin{itemize}
65\item
66  for order~1 processes, $w^{(i)}$ is a Gaussian white noise, with zero mean and standard deviation equal to~1,
67  and the parameters $a^{(i)}$, $b^{(i)}$, $c^{(i)}$ are given by:
68
69  $70 % \label{eq:ord1} 71 \left\{ 72 \begin{array}{l} 73 a^{(i)} = \varphi \\ 74 b^{(i)} = \sigma^{(i)} \sqrt{ 1 - \varphi^2 } \qquad\qquad\mbox{with}\qquad\qquad \varphi = \exp \left( - 1 / \tau^{(i)} \right) \\ 75 c^{(i)} = \mu^{(i)} \left( 1 - \varphi \right) \\ 76 \end{array} 77 \right. 78$
79
80\item
81  for order~$n>1$ processes, $w^{(i)}$ is an order~$n-1$ autoregressive process, with zero mean,
82  standard deviation equal to~$\sigma^{(i)}$;
83  correlation timescale equal to~$\tau^{(i)}$;
84  and the parameters $a^{(i)}$, $b^{(i)}$, $c^{(i)}$ are given by:
85
86  \begin{equation}
87    \label{eq:ord2}
88    \left\{
89      \begin{array}{l}
90        a^{(i)} = \varphi \\
91        b^{(i)} = \frac{n-1}{2(4n-3)} \sqrt{ 1 - \varphi^2 }
93        \varphi = \exp \left( - 1 / \tau^{(i)} \right) \\
94        c^{(i)} = \mu^{(i)} \left( 1 - \varphi \right) \\
95      \end{array}
96    \right.
97  \end{equation}
98
99\end{itemize}
100
101\noindent
102In this way, higher order processes can be easily generated recursively using the same piece of code implementing
103(\autoref{eq:autoreg}), and using succesively processes from order $0$ to~$n-1$ as~$w^{(i)}$.
104The parameters in (\autoref{eq:ord2}) are computed so that this recursive application of
105(\autoref{eq:autoreg}) leads to processes with the required standard deviation and correlation timescale,
106with the additional condition that the $n-1$ first derivatives of the autocorrelation function are equal to
107zero at~$t=0$, so that the resulting processes become smoother and smoother as $n$ is increased.
108
109Overall, this method provides quite a simple and generic way of generating a wide class of stochastic processes.
110However, this also means that new model parameters are needed to specify each of these stochastic processes.
111As in any parameterization of lacking physics, a very important issues then to tune these new parameters using
112either first principles, model simulations, or real-world observations.
113
114\section{Implementation details}
115\label{sec:STO_thech_details}
116
117%---------------------------------------namsbc--------------------------------------------------
118
119\nlst{namsto}
120%--------------------------------------------------------------------------------------------------------------
121
122The computer code implementing stochastic parametrisations can be found in the STO directory.
123It involves three modules :
124\begin{description}
125\item[\mdl{stopar}:]
126  define the Stochastic parameters and their time evolution.
127\item[\mdl{storng}:]
128  a random number generator based on (and includes) the 64-bit KISS (Keep It Simple Stupid) random number generator
129  distributed by George Marsaglia
130  (see \href{https://groups.google.com/forum/#!searchin/comp.lang.fortran/64-bit$20KISS$20RNGs}{here})
131\item[\mdl{stopts}:]
132  stochastic parametrisation associated with the non-linearity of the equation of seawater,
133  implementing \autoref{eq:sto_pert} and specific piece of code in
134  the equation of state implementing \autoref{eq:eos_sto}.
135\end{description}
136
137The \mdl{stopar} module has 3 public routines to be called by the model (in our case, NEMO):
138
139The first routine (\rou{sto\_par}) is a direct implementation of (\autoref{eq:autoreg}),
140applied at each model grid point (in 2D or 3D), and called at each model time step ($k$) to
141update every autoregressive process ($i=1,\ldots,m$).
142This routine also includes a filtering operator, applied to $w^{(i)}$,
143to introduce a spatial correlation between the stochastic processes.
144
145The second routine (\rou{sto\_par\_init}) is an initialization routine mainly dedicated to
146the computation of parameters $a^{(i)}, b^{(i)}, c^{(i)}$ for each autoregressive process,
147as a function of the statistical properties required by the model user
148(mean, standard deviation, time correlation, order of the process,\ldots).
149
150Parameters for the processes can be specified through the following \ngn{namsto} namelist parameters:
151\begin{description}
152\item[\np{nn\_sto\_eos}:]   number of independent random walks
153\item[\np{rn\_eos\_stdxy}:] random walk horz. standard deviation (in grid points)
154\item[\np{rn\_eos\_stdz}:]  random walk vert. standard deviation (in grid points)
155\item[\np{rn\_eos\_tcor}:]  random walk time correlation (in timesteps)
156\item[\np{nn\_eos\_ord}:]   order of autoregressive processes
157\item[\np{nn\_eos\_flt}:]   passes of Laplacian filter
158\item[\np{rn\_eos\_lim}:]   limitation factor (default = 3.0)
159\end{description}
160This routine also includes the initialization (seeding) of the random number generator.
161
162The third routine (\rou{sto\_rst\_write}) writes a restart file
163(which suffix name is given by \np{cn\_storst\_out} namelist parameter) containing the current value of
164all autoregressive processes to allow restarting a simulation from where it has been interrupted.
165This file also contains the current state of the random number generator.
166When \np{ln\_rststo} is set to \forcode{.true.}),
167the restart file (which suffix name is given by \np{cn\_storst\_in} namelist parameter) is read by
168the initialization routine (\rou{sto\_par\_init}).
169The simulation will continue exactly as if it was not interrupted only
170when \np{ln\_rstseed} is set to \forcode{.true.},
171\ie when the state of the random number generator is read in the restart file.
172
173\biblio
174
175\pindex
176
177\end{document}
Note: See TracBrowser for help on using the repository browser.