/[MITgcm]/MITgcm_contrib/articles/ceaice/ceaice.tex
ViewVC logotype

Annotation of /MITgcm_contrib/articles/ceaice/ceaice.tex

Parent Directory Parent Directory | Revision Log Revision Log | View Revision Graph Revision Graph


Revision 1.14 - (hide annotations) (download) (as text)
Tue Feb 26 00:13:20 2008 UTC (17 years, 4 months ago) by dimitri
Branch: MAIN
Changes since 1.13: +37 -4 lines
File MIME type: application/x-tex
Added a placeholder introduction (including first paragraph of PH's
CLIVAR newsletter) more in line with this morning's discussion.

1 dimitri 1.14 % $Header: /u/gcmpack/MITgcm_contrib/articles/ceaice/ceaice.tex,v 1.13 2008/02/25 23:45:46 dimitri Exp $
2 dimitri 1.10 % $Name: $
3 dimitri 1.1 \documentclass[12pt]{article}
4 mlosch 1.4
5 mlosch 1.5 \usepackage[]{graphicx}
6     \usepackage{subfigure}
7 dimitri 1.1
8     \usepackage[round,comma]{natbib}
9 dimitri 1.2 \bibliographystyle{bib/agu04}
10 dimitri 1.1
11     \usepackage{amsmath,amssymb}
12     \newcommand\bmmax{10} \newcommand\hmmax{10}
13     \usepackage{bm}
14    
15     % math abbreviations
16     \newcommand{\vek}[1]{\ensuremath{\mathbf{#1}}}
17     \newcommand{\mat}[1]{\ensuremath{\mathbf{#1}}}
18     \newcommand{\vtau}{\bm{{\tau}}}
19    
20     \newcommand{\degree}{\ensuremath{^\circ}}
21     \newcommand{\degC}{\,\ensuremath{\degree}C}
22     \newcommand{\degE}{\ensuremath{\degree}\,E}
23     \newcommand{\degS}{\ensuremath{\degree}\,S}
24     \newcommand{\degN}{\ensuremath{\degree}\,N}
25     \newcommand{\degW}{\ensuremath{\degree}\,W}
26    
27     % cross reference scheme
28     \newcommand{\reffig}[1]{Figure~\ref{fig:#1}}
29     \newcommand{\reftab}[1]{Table~\ref{tab:#1}}
30     \newcommand{\refapp}[1]{Appendix~\ref{app:#1}}
31     \newcommand{\refsec}[1]{Section~\ref{sec:#1}}
32     \newcommand{\refeq}[1]{\,(\ref{eq:#1})}
33     \newcommand{\refeqs}[2]{\,(\ref{eq:#1})--(\ref{eq:#2})}
34    
35     \newlength{\stdfigwidth}\setlength{\stdfigwidth}{20pc}
36     %\newlength{\stdfigwidth}\setlength{\stdfigwidth}{\columnwidth}
37     \newlength{\mediumfigwidth}\setlength{\mediumfigwidth}{39pc}
38     %\newlength{\widefigwidth}\setlength{\widefigwidth}{39pc}
39     \newlength{\widefigwidth}\setlength{\widefigwidth}{\textwidth}
40 mlosch 1.4 \newcommand{\fpath}{figs}
41    
42     % commenting scheme
43     \newcommand{\ml}[1]{\textsf{\slshape #1}}
44 dimitri 1.1
45     \title{A Dynamic-Thermodynamic Sea ice Model for Ocean Climate
46     Estimation on an Arakawa C-Grid}
47    
48     \author{Martin Losch, Dimitris Menemenlis, Patrick Heimbach, \\
49     Jean-Michel Campin, and Chris Hill}
50     \begin{document}
51    
52     \maketitle
53    
54     \begin{abstract}
55 dimitri 1.10 As part of ongoing efforts to obtain a best possible synthesis of most
56 dimitri 1.12 available, global-scale, ocean and sea ice data, a dynamic and thermodynamic
57     sea-ice model has been coupled to the Massachusetts Institute of Technology
58     general circulation model (MITgcm). Ice mechanics follow a viscous plastic
59     rheology and the ice momentum equations are solved numerically using either
60     line successive relaxation (LSR) or elastic-viscous-plastic (EVP) dynamic
61     models. Ice thermodynamics are represented using either a zero-heat-capacity
62     formulation or a two-layer formulation that conserves enthalpy. The model
63     includes prognostic variables for snow and for sea-ice salinity. The above
64     sea ice model components were borrowed from current-generation climate models
65     but they were reformulated on an Arakawa C-grid in order to match the MITgcm
66     oceanic grid and they were modified in many ways to permit efficient and
67     accurate automatic differentiation. This paper describes the MITgcm sea ice
68     model; it presents example Arctic and Antarctic results from a realistic,
69     eddy-permitting, global ocean and sea-ice configuration; it compares B-grid
70     and C-grid dynamic solvers in a regional Arctic configuration; and it presents
71     example results from coupled ocean and sea-ice adjoint-model integrations.
72 dimitri 1.10
73 dimitri 1.1 \end{abstract}
74    
75     \section{Introduction}
76     \label{sec:intro}
77    
78 dimitri 1.14 The availability of an adjoint model as a powerful research
79     tool complementary to an ocean model was a major design
80     requirement early on in the development of the MIT general
81     circulation model (MITgcm) [Marshall et al. 1997a,
82     Marotzke et al. 1999, Adcroft et al. 2002]. It was recognized
83     that the adjoint permitted very efficient computation
84     of gradients of various scalar-valued model diagnostics,
85     norms or, generally, objective functions with respect
86     to external or independent parameters. Such gradients
87     arise in at least two major contexts. If the objective function
88     is the sum of squared model vs. obervation differences
89     weighted by e.g. the inverse error covariances, the gradient
90     of the objective function can be used to optimize this measure
91     of model vs. data misfit in a least-squares sense. One
92     is then solving a problem of statistical state estimation.
93     If the objective function is a key oceanographic quantity
94     such as meridional heat or volume transport, ocean heat
95     content or mean surface temperature index, the gradient
96     provides a complete set of sensitivities of this quantity
97     with respect to all independent variables simultaneously.
98    
99     References to existing sea-ice adjoint models, explaining that they are either
100     for simplified configurations, for ice-only studies, or for short-duration
101     studies to motivate the present work.
102    
103 dimitri 1.1 Traditionally, probably for historical reasons and the ease of
104     treating the Coriolis term, most standard sea-ice models are
105     discretized on Arakawa-B-grids \citep[e.g.,][]{hibler79, harder99,
106 dimitri 1.14 kreyscher00, zhang98, hunke97}. From the perspective of coupling a
107 dimitri 1.1 sea ice-model to a C-grid ocean model, the exchange of fluxes of heat
108     and fresh-water pose no difficulty for a B-grid sea-ice model
109     \citep[e.g.,][]{timmermann02a}. However, surface stress is defined at
110     velocities points and thus needs to be interpolated between a B-grid
111     sea-ice model and a C-grid ocean model. While the smoothing implicitly
112     associated with this interpolation may mask grid scale noise, it may
113     in two-way coupling lead to a computational mode as will be shown. By
114 dimitri 1.13 choosing a C-grid for the sea-ice model, we circumvent this difficulty
115 dimitri 1.1 altogether and render the stress coupling as consistent as the
116     buoyancy coupling.
117    
118     A further advantage of the C-grid formulation is apparent in narrow
119     straits. In the limit of only one grid cell between coasts there is no
120     flux allowed for a B-grid (with no-slip lateral boundary counditions),
121     whereas the C-grid formulation allows a flux of sea-ice through this
122 dimitri 1.14 passage for all types of lateral boundary conditions. We
123 dimitri 1.1 demonstrate this effect in the Candian archipelago.
124    
125 dimitri 1.14 Talk about problems that make the sea-ice-ocean code very sensitive and
126     changes in the code that reduce these sensitivities.
127    
128     This paper describes the MITgcm sea ice
129     model; it presents example Arctic and Antarctic results from a realistic,
130     eddy-permitting, global ocean and sea-ice configuration; it compares B-grid
131     and C-grid dynamic solvers in a regional Arctic configuration; and it presents
132     example results from coupled ocean and sea-ice adjoint-model integrations.
133    
134 dimitri 1.13 \section{Model}
135     \label{sec:model}
136    
137 dimitri 1.1 \subsection{Dynamics}
138     \label{sec:dynamics}
139    
140 dimitri 1.13 The momentum equation of the sea-ice model is
141 dimitri 1.1 \begin{equation}
142     \label{eq:momseaice}
143     m \frac{D\vek{u}}{Dt} = -mf\vek{k}\times\vek{u} + \vtau_{air} +
144 dimitri 1.13 \vtau_{ocean} - mg \nabla{\phi(0)} + \vek{F},
145 dimitri 1.1 \end{equation}
146 dimitri 1.13 where $m=m_{i}+m_{s}$ is the ice and snow mass per unit area;
147     $\vek{u}=u\vek{i}+v\vek{j}$ is the ice velocity vector;
148     $\vek{i}$, $\vek{j}$, and $\vek{k}$ are unit vectors in the $x$, $y$, and $z$
149     directions, respectively;
150     $f$ is the Coriolis parameter;
151     $\vtau_{air}$ and $\vtau_{ocean}$ are the wind-ice and ocean-ice stresses,
152     respectively;
153     $g$ is the gravity accelation;
154     $\nabla\phi(0)$ is the gradient (or tilt) of the sea surface height;
155     $\phi(0)$ is the sea surface height potential in response to ocean dynamics
156     and to atmospheric pressure loading;
157     and $\vek{F}=\nabla\cdot\sigma$ is the divergence of the internal ice stress
158     tensor $\sigma_{ij}$.
159     When using the rescaled vertical coordinate system, z$^\ast$, of
160     \citet{cam08}, $\phi(0)$ also includes a term due to snow and ice loading, $mg$.
161     Advection of sea-ice momentum is neglected. The wind and ice-ocean stress
162     terms are given by
163 dimitri 1.1 \begin{align*}
164 dimitri 1.13 \vtau_{air} = & \rho_{air} C_{air} |\vek{U}_{air} -\vek{u}|
165     R_{air} (\vek{U}_{air} -\vek{u}), \\
166     \vtau_{ocean} = & \rho_{ocean}C_{ocean} |\vek{U}_{ocean}-\vek{u}|
167 dimitri 1.1 R_{ocean}(\vek{U}_{ocean}-\vek{u}), \\
168     \end{align*}
169     where $\vek{U}_{air/ocean}$ are the surface winds of the atmosphere
170 dimitri 1.13 and surface currents of the ocean, respectively; $C_{air/ocean}$ are
171     air and ocean drag coefficients; $\rho_{air/ocean}$ are reference
172     densities; and $R_{air/ocean}$ are rotation matrices that act on the
173     wind/current vectors.
174 dimitri 1.1
175     For an isotropic system this stress tensor can be related to the ice
176     strain rate and strength by a nonlinear viscous-plastic (VP)
177     constitutive law \citep{hibler79, zhang98}:
178     \begin{equation}
179     \label{eq:vpequation}
180     \sigma_{ij}=2\eta(\dot{\epsilon}_{ij},P)\dot{\epsilon}_{ij}
181     + \left[\zeta(\dot{\epsilon}_{ij},P) -
182     \eta(\dot{\epsilon}_{ij},P)\right]\dot{\epsilon}_{kk}\delta_{ij}
183     - \frac{P}{2}\delta_{ij}.
184     \end{equation}
185     The ice strain rate is given by
186     \begin{equation*}
187     \dot{\epsilon}_{ij} = \frac{1}{2}\left(
188     \frac{\partial{u_{i}}}{\partial{x_{j}}} +
189     \frac{\partial{u_{j}}}{\partial{x_{i}}}\right).
190     \end{equation*}
191 mlosch 1.5 The maximum ice pressure $P_{\max}$, a measure of ice strength, depends on
192     both thickness $h$ and compactness (concentration) $c$:
193 mlosch 1.4 \begin{equation}
194 mlosch 1.5 P_{\max} = P^{*}c\,h\,e^{[C^{*}\cdot(1-c)]},
195 mlosch 1.9 \label{eq:icestrength}
196 mlosch 1.4 \end{equation}
197     with the constants $P^{*}$ and $C^{*}$. The nonlinear bulk and shear
198     viscosities $\eta$ and $\zeta$ are functions of ice strain rate
199     invariants and ice strength such that the principal components of the
200     stress lie on an elliptical yield curve with the ratio of major to
201     minor axis $e$ equal to $2$; they are given by:
202 dimitri 1.1 \begin{align*}
203 mlosch 1.5 \zeta =& \min\left(\frac{P_{\max}}{2\max(\Delta,\Delta_{\min})},
204     \zeta_{\max}\right) \\
205     \eta =& \frac{\zeta}{e^2} \\
206 dimitri 1.1 \intertext{with the abbreviation}
207     \Delta = & \left[
208     \left(\dot{\epsilon}_{11}^2+\dot{\epsilon}_{22}^2\right)
209     (1+e^{-2}) + 4e^{-2}\dot{\epsilon}_{12}^2 +
210     2\dot{\epsilon}_{11}\dot{\epsilon}_{22} (1-e^{-2})
211     \right]^{-\frac{1}{2}}
212     \end{align*}
213 mlosch 1.5 The bulk viscosities are bounded above by imposing both a minimum
214     $\Delta_{\min}=10^{-11}\text{\,s}^{-1}$ (for numerical reasons) and a
215     maximum $\zeta_{\max} = P_{\max}/\Delta^*$, where
216     $\Delta^*=(5\times10^{12}/2\times10^4)\text{\,s}^{-1}$. For stress
217 dimitri 1.13 tensor computation the replacement pressure $P = 2\,\Delta\zeta$
218 mlosch 1.5 \citep{hibler95} is used so that the stress state always lies on the
219     elliptic yield curve by definition.
220    
221 mlosch 1.6 In the so-called truncated ellipse method the shear viscosity $\eta$
222     is capped to suppress any tensile stress \citep{hibler97, geiger98}:
223     \begin{equation}
224     \label{eq:etatem}
225     \eta = \min(\frac{\zeta}{e^2}
226     \frac{\frac{P}{2}-\zeta(\dot{\epsilon}_{11}+\dot{\epsilon}_{22})}
227     {\sqrt{(\dot{\epsilon}_{11}+\dot{\epsilon}_{22})^2
228     +4\dot{\epsilon}_{12}^2}}
229     \end{equation}
230    
231 dimitri 1.1 In the current implementation, the VP-model is integrated with the
232     semi-implicit line successive over relaxation (LSOR)-solver of
233     \citet{zhang98}, which allows for long time steps that, in our case,
234     is limited by the explicit treatment of the Coriolis term. The
235     explicit treatment of the Coriolis term does not represent a severe
236     limitation because it restricts the time step to approximately the
237     same length as in the ocean model where the Coriolis term is also
238     treated explicitly.
239    
240     \citet{hunke97}'s introduced an elastic contribution to the strain
241 dimitri 1.13 rate elastic-viscous-plastic in order to regularize
242 dimitri 1.1 Eq.\refeq{vpequation} in such a way that the resulting
243 dimitri 1.13 elastic-viscous-plastic (EVP) and VP models are identical at steady
244 dimitri 1.1 state,
245     \begin{equation}
246     \label{eq:evpequation}
247     \frac{1}{E}\frac{\partial\sigma_{ij}}{\partial{t}} +
248     \frac{1}{2\eta}\sigma_{ij}
249     + \frac{\eta - \zeta}{4\zeta\eta}\sigma_{kk}\delta_{ij}
250     + \frac{P}{4\zeta}\delta_{ij}
251     = \dot{\epsilon}_{ij}.
252     \end{equation}
253     %In the EVP model, equations for the components of the stress tensor
254     %$\sigma_{ij}$ are solved explicitly. Both model formulations will be
255     %used and compared the present sea-ice model study.
256     The EVP-model uses an explicit time stepping scheme with a short
257     timestep. According to the recommendation of \citet{hunke97}, the
258     EVP-model is stepped forward in time 120 times within the physical
259     ocean model time step (although this parameter is under debate), to
260     allow for elastic waves to disappear. Because the scheme does not
261     require a matrix inversion it is fast in spite of the small timestep
262     \citep{hunke97}. For completeness, we repeat the equations for the
263     components of the stress tensor $\sigma_{1} =
264     \sigma_{11}+\sigma_{22}$, $\sigma_{2}= \sigma_{11}-\sigma_{22}$, and
265     $\sigma_{12}$. Introducing the divergence $D_D =
266     \dot{\epsilon}_{11}+\dot{\epsilon}_{22}$, and the horizontal tension
267     and shearing strain rates, $D_T =
268     \dot{\epsilon}_{11}-\dot{\epsilon}_{22}$ and $D_S =
269 dimitri 1.13 2\dot{\epsilon}_{12}$, respectively, and using the above abbreviations,
270 dimitri 1.1 the equations can be written as:
271     \begin{align}
272     \label{eq:evpstresstensor1}
273     \frac{\partial\sigma_{1}}{\partial{t}} + \frac{\sigma_{1}}{2T} +
274     \frac{P}{2T} &= \frac{P}{2T\Delta} D_D \\
275     \label{eq:evpstresstensor2}
276     \frac{\partial\sigma_{2}}{\partial{t}} + \frac{\sigma_{2} e^{2}}{2T}
277     &= \frac{P}{2T\Delta} D_T \\
278     \label{eq:evpstresstensor12}
279     \frac{\partial\sigma_{12}}{\partial{t}} + \frac{\sigma_{12} e^{2}}{2T}
280     &= \frac{P}{4T\Delta} D_S
281     \end{align}
282     Here, the elastic parameter $E$ is redefined in terms of a damping timescale
283     $T$ for elastic waves \[E=\frac{\zeta}{T}.\]
284     $T=E_{0}\Delta{t}$ with the tunable parameter $E_0<1$ and
285     the external (long) timestep $\Delta{t}$. \citet{hunke97} recommend
286     $E_{0} = \frac{1}{3}$.
287    
288     For details of the spatial discretization, the reader is referred to
289     \citet{zhang98, zhang03}. Our discretization differs only (but
290     importantly) in the underlying grid, namely the Arakawa C-grid, but is
291     otherwise straightforward. The EVP model in particular is discretized
292     naturally on the C-grid with $\sigma_{1}$ and $\sigma_{2}$ on the
293     center points and $\sigma_{12}$ on the corner (or vorticity) points of
294     the grid. With this choice all derivatives are discretized as central
295     differences and averaging is only involved in computing $\Delta$ and
296     $P$ at vorticity points.
297    
298     For a general curvilinear grid, one needs in principle to take metric
299 dimitri 1.13 terms into account that arise in the transformation of a curvilinear grid
300     on the sphere. For now, however, we can neglect these metric terms
301 dimitri 1.1 because they are very small on the cubed sphere grids used in this
302     paper; in particular, only near the edges of the cubed sphere grid, we
303     expect them to be non-zero, but these edges are at approximately
304     35\degS\ or 35\degN\ which are never covered by sea-ice in our
305     simulations. Everywhere else the coordinate system is locally nearly
306     cartesian. However, for last-glacial-maximum or snowball-earth-like
307     simulations the question of metric terms needs to be reconsidered.
308     Either, one includes these terms as in \citet{zhang03}, or one finds a
309 dimitri 1.13 vector-invariant formulation for the sea-ice internal stress term that
310 dimitri 1.1 does not require any metric terms, as it is done in the ocean dynamics
311     of the MITgcm \citep{adcroft04:_cubed_sphere}.
312    
313     Moving sea ice exerts a stress on the ocean which is the opposite of
314     the stress $\vtau_{ocean}$ in Eq.\refeq{momseaice}. This stess is
315     applied directly to the surface layer of the ocean model. An
316     alternative ocean stress formulation is given by \citet{hibler87}.
317     Rather than applying $\vtau_{ocean}$ directly, the stress is derived
318     from integrating over the ice thickness to the bottom of the oceanic
319     surface layer. In the resulting equation for the \emph{combined}
320     ocean-ice momentum, the interfacial stress cancels and the total
321     stress appears as the sum of windstress and divergence of internal ice
322     stresses: $\delta(z) (\vtau_{air} + \vek{F})/\rho_0$, \citep[see also
323     Eq.\,2 of][]{hibler87}. The disadvantage of this formulation is that
324     now the velocity in the surface layer of the ocean that is used to
325     advect tracers, is really an average over the ocean surface
326     velocity and the ice velocity leading to an inconsistency as the ice
327     temperature and salinity are different from the oceanic variables.
328    
329     Sea ice distributions are characterized by sharp gradients and edges.
330     For this reason, we employ a positive 3rd-order advection scheme
331     \citep{hundsdorfer94} for the thermodynamic variables discussed in the
332     next section.
333    
334     \subparagraph{boundary conditions: no-slip, free-slip, half-slip}
335    
336     \begin{itemize}
337     \item transition from existing B-Grid to C-Grid
338     \item boundary conditions: no-slip, free-slip, half-slip
339     \item fancy (multi dimensional) advection schemes
340     \item VP vs.\ EVP \citep{hunke97}
341     \item ocean stress formulation \citep{hibler87}
342     \end{itemize}
343    
344     \subsection{Thermodynamics}
345     \label{sec:thermodynamics}
346    
347     In the original formulation the sea ice model \citep{menemenlis05}
348     uses simple thermodynamics following the appendix of
349     \citet{semtner76}. This formulation does not allow storage of heat
350     (heat capacity of ice is zero, and this type of model is often refered
351     to as a ``zero-layer'' model). Upward heat flux is parameterized
352     assuming a linear temperature profile and together with a constant ice
353     conductivity. It is expressed as $(K/h)(T_{w}-T_{0})$, where $K$ is
354     the ice conductivity, $h$ the ice thickness, and $T_{w}-T_{0}$ the
355     difference between water and ice surface temperatures. The surface
356     heat budget is computed in a similar way to that of
357     \citet{parkinson79} and \citet{manabe79}.
358    
359     There is considerable doubt about the reliability of such a simple
360     thermodynamic model---\citet{semtner84} found significant errors in
361     phase (one month lead) and amplitude ($\approx$50\%\,overestimate) in
362     such models---, so that today many sea ice models employ more complex
363     thermodynamics. A popular thermodynamics model of \citet{winton00} is
364     based on the 3-layer model of \citet{semtner76} and treats brine
365     content by means of enthalphy conservation. This model requires in
366     addition to ice-thickness and compactness (fractional area) additional
367     state variables to be advected by ice velocities, namely enthalphy of
368     the two ice layers and the thickness of the overlying snow layer.
369    
370 mlosch 1.9
371 dimitri 1.1 \subsection{C-grid}
372     \begin{itemize}
373     \item no-slip vs. free-slip for both lsr and evp;
374     "diagnostics" to look at and use for comparison
375     \begin{itemize}
376     \item ice transport through Fram Strait/Denmark Strait/Davis
377     Strait/Bering strait (these are general)
378     \item ice transport through narrow passages, e.g.\ Nares-Strait
379     \end{itemize}
380     \item compare different advection schemes (if lsr turns out to be more
381     effective, then with lsr otherwise I prefer evp), eg.
382     \begin{itemize}
383     \item default 2nd-order with diff1=0.002
384     \item 1st-order upwind with diff1=0.
385     \item DST3FL (SEAICEadvScheme=33 with diff1=0., should work, works for me)
386     \item 2nd-order wit flux limiter (SEAICEadvScheme=77 with diff1=0.)
387     \end{itemize}
388     That should be enough. Here, total ice mass and location of ice edge
389     is interesting. However, this comparison can be done in an idealized
390     domain, may not require full Arctic Domain?
391     \item
392     Do a little study on the parameters of LSR and EVP
393     \begin{enumerate}
394     \item convergence of LSR, how many iterations do you need to get a
395     true elliptic yield curve
396     \item vary deltaTevp and the relaxation parameter for EVP and see when
397     the EVP solution breaks down (relative to the forcing time scale).
398     For this, it is essential that the evp solver gives use "stripeless"
399     solutions, that is your dtevp = 1sec solutions/or 10sec solutions
400     with SEAICE\_evpDampC = 615.
401     \end{enumerate}
402     \end{itemize}
403    
404     \section{Forward sensitivity experiments}
405     \label{sec:forward}
406    
407     A second series of forward sensitivity experiments have been carried out on an
408     Arctic Ocean domain with open boundaries. Once again the objective is to
409     compare the old B-grid LSR dynamic solver with the new C-grid LSR and EVP
410     solvers. One additional experiment is carried out to illustrate the
411     differences between the two main options for sea ice thermodynamics in the MITgcm.
412    
413     \subsection{Arctic Domain with Open Boundaries}
414     \label{sec:arctic}
415    
416 dimitri 1.12 The Arctic domain of integration is illustrated in Fig.~\ref{fig:arctic1}. It
417 mlosch 1.6 is carved out from, and obtains open boundary conditions from, the
418     global cubed-sphere configuration of the Estimating the Circulation
419     and Climate of the Ocean, Phase II (ECCO2) project
420     \citet{menemenlis05}. The domain size is 420 by 384 grid boxes
421     horizontally with mean horizontal grid spacing of 18 km.
422 dimitri 1.1
423 dimitri 1.12 \begin{figure}
424     %\centerline{{\includegraphics*[width=0.44\linewidth]{\fpath/arctic1.eps}}}
425     \caption{Bathymetry of Arctic Domain.\label{fig:arctic1}}
426     \end{figure}
427    
428 dimitri 1.1 There are 50 vertical levels ranging in thickness from 10 m near the surface
429     to approximately 450 m at a maximum model depth of 6150 m. Bathymetry is from
430     the National Geophysical Data Center (NGDC) 2-minute gridded global relief
431     data (ETOPO2) and the model employs the partial-cell formulation of
432 mlosch 1.6 \citet{adcroft97:_shaved_cells}, which permits accurate representation of the bathymetry. The
433 dimitri 1.1 model is integrated in a volume-conserving configuration using a finite volume
434     discretization with C-grid staggering of the prognostic variables. In the
435 mlosch 1.6 ocean, the non-linear equation of state of \citet{jackett95}. The ocean model is
436 dimitri 1.1 coupled to a sea-ice model described hereinabove.
437    
438 mlosch 1.6 This particular ECCO2 simulation is initialized from rest using the
439     January temperature and salinity distribution from the World Ocean
440     Atlas 2001 (WOA01) [Conkright et al., 2002] and it is integrated for
441     32 years prior to the 1996--2001 period discussed in the study. Surface
442     boundary conditions are from the National Centers for Environmental
443     Prediction and the National Center for Atmospheric Research
444     (NCEP/NCAR) atmospheric reanalysis [Kistler et al., 2001]. Six-hourly
445     surface winds, temperature, humidity, downward short- and long-wave
446     radiations, and precipitation are converted to heat, freshwater, and
447     wind stress fluxes using the \citet{large81, large82} bulk formulae.
448     Shortwave radiation decays exponentially as per Paulson and Simpson
449     [1977]. Additionally the time-mean river run-off from Large and Nurser
450     [2001] is applied and there is a relaxation to the monthly-mean
451     climatological sea surface salinity values from WOA01 with a
452     relaxation time scale of 3 months. Vertical mixing follows
453     \citet{large94} with background vertical diffusivity of
454     $1.5\times10^{-5}\text{\,m$^{2}$\,s$^{-1}$}$ and viscosity of
455     $10^{-3}\text{\,m$^{2}$\,s$^{-1}$}$. A third order, direct-space-time
456     advection scheme with flux limiter is employed \citep{hundsdorfer94}
457     and there is no explicit horizontal diffusivity. Horizontal viscosity
458     follows \citet{lei96} but
459     modified to sense the divergent flow as per Fox-Kemper and Menemenlis
460     [in press]. Shortwave radiation decays exponentially as per Paulson
461     and Simpson [1977]. Additionally, the time-mean runoff of Large and
462     Nurser [2001] is applied near the coastline and, where there is open
463     water, there is a relaxation to monthly-mean WOA01 sea surface
464     salinity with a time constant of 45 days.
465 dimitri 1.1
466     Open water, dry
467     ice, wet ice, dry snow, and wet snow albedo are, respectively, 0.15, 0.85,
468     0.76, 0.94, and 0.8.
469    
470     \begin{itemize}
471     \item Configuration
472     \item OBCS from cube
473     \item forcing
474     \item 1/2 and full resolution
475     \item with a few JFM figs from C-grid LSR no slip
476     ice transport through Canadian Archipelago
477     thickness distribution
478     ice velocity and transport
479     \end{itemize}
480    
481     \begin{itemize}
482     \item Arctic configuration
483     \item ice transport through straits and near boundaries
484     \item focus on narrow straits in the Canadian Archipelago
485     \end{itemize}
486    
487     \begin{itemize}
488     \item B-grid LSR no-slip
489     \item C-grid LSR no-slip
490     \item C-grid LSR slip
491     \item C-grid EVP no-slip
492     \item C-grid EVP slip
493 mlosch 1.6 \item C-grid LSR + TEM (truncated ellipse method, no tensile stress, new flag)
494 dimitri 1.1 \item C-grid LSR no-slip + Winton
495     \item speed-performance-accuracy (small)
496     ice transport through Canadian Archipelago differences
497     thickness distribution differences
498     ice velocity and transport differences
499     \end{itemize}
500    
501     We anticipate small differences between the different models due to:
502     \begin{itemize}
503     \item advection schemes: along the ice-edge and regions with large
504     gradients
505 mlosch 1.6 \item C-grid: less transport through narrow straits for no slip
506     conditons, more for free slip
507 dimitri 1.1 \item VP vs.\ EVP: speed performance, accuracy?
508     \item ocean stress: different water mass properties beneath the ice
509     \end{itemize}
510    
511     \section{Adjoint sensiivities of the MITsim}
512    
513     \subsection{The adjoint of MITsim}
514    
515     The ability to generate tangent linear and adjoint model components
516     of the MITsim has been a main design task.
517     For the ocean the adjoint capability has proven to be an
518     invaluable tool for sensitivity analysis as well as state estimation.
519     In short, the adjoint enables very efficient computation of the gradient
520     of scalar-valued model diagnostics (called cost function or objective function)
521     with respect to many model "variables".
522     These variables can be two- or three-dimensional fields of initial
523     conditions, model parameters such as mixing coefficients, or
524     time-varying surface or lateral (open) boundary conditions.
525     When combined, these variables span a potentially high-dimensional
526     (e.g. O(10$^8$)) so-called control space. Performing parameter perturbations
527     to assess model sensitivities quickly becomes prohibitive at these scales.
528     Alternatively, (time-varying) sensitivities of the objective function
529     to any element of the control space can be computed very efficiently in
530     one single adjoint
531     model integration, provided an efficient adjoint model is available.
532    
533     [REFERENCES]
534    
535    
536     The adjoint operator (ADM) is the transpose of the tangent linear operator (TLM)
537     of the full (in general nonlinear) forward model, i.e. the MITsim.
538     The TLM maps perturbations of elements of the control space
539     (e.g. initial ice thickness distribution)
540     via the model Jacobian
541     to a perturbation in the objective function
542     (e.g. sea-ice export at the end of the integration interval).
543     \textit{Tangent} linearity ensures that the derivatives are evaluated
544     with respect to the underlying model trajectory at each point in time.
545     This is crucial for nonlinear trajectories and the presence of different
546     regimes (e.g. effect of the seaice growth term at or away from the
547     freezing point of the ocean surface).
548     Ensuring tangent linearity can be easily achieved by integrating
549     the full model in sync with the TLM to provide the underlying model state.
550     Ensuring \textit{tangent} adjoints is equally crucial, but much more
551     difficult to achieve because of the reverse nature of the integration:
552     the adjoint accumulates sensitivities backward in time,
553     starting from a unit perturbation of the objective function.
554     The adjoint model requires the model state in reverse order.
555     This presents one of the major complications in deriving an
556     exact, i.e. \textit{tangent} adjoint model.
557    
558     Following closely the development and maintenance of TLM and ADM
559     components of the MITgcm we have relied heavily on the
560     autmomatic differentiation (AD) tool
561     "Transformation of Algorithms in Fortran" (TAF)
562     developed by Fastopt (Giering and Kaminski, 1998)
563     to derive TLM and ADM code of the MITsim.
564     Briefly, the nonlinear parent model is fed to the AD tool which produces
565     derivative code for the specified control space and objective function.
566     Following this approach has (apart from its evident success)
567     several advantages:
568     (1) the adjoint model is the exact adjoint operator of the parent model,
569     (2) the adjoint model can be kept up to date with respect to ongoing
570     development of the parent model, and adjustments to the parent model
571     to extend the automatically generated adjoint are incremental changes
572     only, rather than extensive re-developments,
573     (3) the parallel structure of the parent model is preserved
574     by the adjoint model, ensuring efficient use in high performance
575     computing environments.
576    
577     Some initial code adjustments are required to support dependency analysis
578     of the flow reversal and certain language limitations which may lead
579     to irreducible flow graphs (e.g. GOTO statements).
580     The problem of providing the required model state in reverse order
581     at the time of evaluating nonlinear or conditional
582     derivatives is solved via balancing
583     storing vs. recomputation of the model state in a multi-level
584     checkpointing loop.
585     Again, an initial code adjustment is required to support TAFs
586     checkpointing capability.
587 mlosch 1.6 The code adjustments are sufficiently simple so as not to cause
588 dimitri 1.1 major limitations to the full nonlinear parent model.
589     Once in place, an adjoint model of a new model configuration
590     may be derived in about 10 minutes.
591    
592     [HIGHLIGHT COUPLED NATURE OF THE ADJOINT!]
593    
594     \subsection{Special considerations}
595    
596     * growth term(?)
597    
598     * small active denominators
599    
600     * dynamic solver (implicit function theorem)
601    
602     * approximate adjoints
603    
604    
605     \subsection{An example: sensitivities of sea-ice export through Fram Strait}
606    
607     We demonstrate the power of the adjoint method
608     in the context of investigating sea-ice export sensitivities through Fram Strait
609     (for details of this study see Heimbach et al., 2007).
610 mlosch 1.6 %\citep[for details of this study see][]{heimbach07}. %Heimbach et al., 2007).
611 dimitri 1.1 The domain chosen is a coarsened version of the Arctic face of the
612     high-resolution cubed-sphere configuration of the ECCO2 project
613 mlosch 1.6 \citep[see][]{menemenlis05}. It covers the entire Arctic,
614 dimitri 1.1 extends into the North Pacific such as to cover the entire
615     ice-covered regions, and comprises parts of the North Atlantic
616     down to XXN to enable analysis of remote influences of the
617     North Atlantic current to sea-ice variability and export.
618     The horizontal resolution varies between XX and YY km
619     with 50 unevenly spaced vertical levels.
620     The adjoint models run efficiently on 80 processors
621     (benchmarks have been performed both on an SGI Altix as well as an
622     IBM SP5 at NASA/ARC).
623    
624 mlosch 1.6 Following a 1-year spinup, the model has been integrated for four
625     years between 1992 and 1995. It is forced using realistic 6-hourly
626     NCEP/NCAR atmospheric state variables. Over the open ocean these are
627     converted into air-sea fluxes via the bulk formulae of
628     \citet{large04}. Derivation of air-sea fluxes in the presence of
629     sea-ice is handled by the ice model as described in \refsec{model}.
630 dimitri 1.1 The objective function chosen is sea-ice export through Fram Strait
631 mlosch 1.6 computed for December 1995. The adjoint model computes sensitivities
632     to sea-ice export back in time from 1995 to 1992 along this
633     trajectory. In principle all adjoint model variable (i.e., Lagrange
634     multipliers) of the coupled ocean/sea-ice model are available to
635     analyze the transient sensitivity behaviour of the ocean and sea-ice
636     state. Over the open ocean, the adjoint of the bulk formula scheme
637     computes sensitivities to the time-varying atmospheric state. Over
638     ice-covered parts, the sea-ice adjoint converts surface ocean
639     sensitivities to atmospheric sensitivities.
640    
641     \reffig{4yradjheff}(a--d) depict sensitivities of sea-ice export
642     through Fram Strait in December 1995 to changes in sea-ice thickness
643     12, 24, 36, 48 months back in time. Corresponding sensitivities to
644     ocean surface temperature are depicted in
645     \reffig{4yradjthetalev1}(a--d). The main characteristics is
646     consistency with expected advection of sea-ice over the relevant time
647     scales considered. The general positive pattern means that an
648     increase in sea-ice thickness at location $(x,y)$ and time $t$ will
649     increase sea-ice export through Fram Strait at time $T_e$. Largest
650     distances from Fram Strait indicate fastest sea-ice advection over the
651     time span considered. The ice thickness sensitivities are in close
652     correspondence to ocean surface sentivitites, but of opposite sign.
653     An increase in temperature will incur ice melting, decrease in ice
654     thickness, and therefore decrease in sea-ice export at time $T_e$.
655 dimitri 1.1
656     The picture is fundamentally different and much more complex
657     for sensitivities to ocean temperatures away from the surface.
658 mlosch 1.6 \reffig{4yradjthetalev10??}(a--d) depicts ice export sensitivities to
659 dimitri 1.1 temperatures at roughly 400 m depth.
660     Primary features are the effect of the heat transport of the North
661     Atlantic current which feeds into the West Spitsbergen current,
662     the circulation around Svalbard, and ...
663    
664     \begin{figure}[t!]
665     \centerline{
666     \subfigure[{\footnotesize -12 months}]
667 mlosch 1.4 {\includegraphics*[width=0.44\linewidth]{\fpath/run_4yr_ADJheff_arc_lev1_tim072_cmax2.0E+02.eps}}
668 dimitri 1.1 %\includegraphics*[width=.3\textwidth]{H_c.bin_res_100_lev1.pdf}
669     %
670     \subfigure[{\footnotesize -24 months}]
671 mlosch 1.4 {\includegraphics*[width=0.44\linewidth]{\fpath/run_4yr_ADJheff_arc_lev1_tim145_cmax2.0E+02.eps}}
672 dimitri 1.1 }
673    
674     \centerline{
675     \subfigure[{\footnotesize
676     -36 months}]
677 mlosch 1.4 {\includegraphics*[width=0.44\linewidth]{\fpath/run_4yr_ADJheff_arc_lev1_tim218_cmax2.0E+02.eps}}
678 dimitri 1.1 %
679     \subfigure[{\footnotesize
680     -48 months}]
681 mlosch 1.4 {\includegraphics*[width=0.44\linewidth]{\fpath/run_4yr_ADJheff_arc_lev1_tim292_cmax2.0E+02.eps}}
682 dimitri 1.1 }
683     \caption{Sensitivity of sea-ice export through Fram Strait in December 2005 to
684     sea-ice thickness at various prior times.
685     \label{fig:4yradjheff}}
686     \end{figure}
687    
688    
689     \begin{figure}[t!]
690     \centerline{
691     \subfigure[{\footnotesize -12 months}]
692 mlosch 1.4 {\includegraphics*[width=0.44\linewidth]{\fpath/run_4yr_ADJtheta_arc_lev1_tim072_cmax5.0E+01.eps}}
693 dimitri 1.1 %\includegraphics*[width=.3\textwidth]{H_c.bin_res_100_lev1.pdf}
694     %
695     \subfigure[{\footnotesize -24 months}]
696 mlosch 1.4 {\includegraphics*[width=0.44\linewidth]{\fpath/run_4yr_ADJtheta_arc_lev1_tim145_cmax5.0E+01.eps}}
697 dimitri 1.1 }
698    
699     \centerline{
700     \subfigure[{\footnotesize
701     -36 months}]
702 mlosch 1.4 {\includegraphics*[width=0.44\linewidth]{\fpath/run_4yr_ADJtheta_arc_lev1_tim218_cmax5.0E+01.eps}}
703 dimitri 1.1 %
704     \subfigure[{\footnotesize
705     -48 months}]
706 mlosch 1.4 {\includegraphics*[width=0.44\linewidth]{\fpath/run_4yr_ADJtheta_arc_lev1_tim292_cmax5.0E+01.eps}}
707 dimitri 1.1 }
708 mlosch 1.6 \caption{Same as \reffig{4yradjheff} but for sea surface temperature
709 dimitri 1.1 \label{fig:4yradjthetalev1}}
710     \end{figure}
711    
712    
713    
714     \section{Discussion and conclusion}
715     \label{sec:concl}
716    
717     The story of the paper could be:
718     B-grid ice model + C-grid ocean model does not work properly for us,
719     therefore C-grid ice model with advantages:
720     \begin{enumerate}
721     \item stress coupling simpler (no interpolation required)
722     \item different boundary conditions
723     \item advection schemes carry over trivially from main code
724     \end{enumerate}
725     LSR/EVP solutions are similar with appropriate bcs, evp parameters as
726     a function of forcing time scale (when does VP solution break
727     down). Same for LSR solver, provided that it works (o:
728     Which scheme is more efficient for the resolution/time stepping
729     parameters that we use here. What about other resolutions?
730    
731     \paragraph{Acknowledgements}
732     We thank Jinlun Zhang for providing the original B-grid code and many
733 mlosch 1.6 helpful discussions. ML thanks Elizabeth Hunke for multiple explanations.
734 dimitri 1.1
735 dimitri 1.7 \bibliography{bib/journal_abrvs,bib/seaice,bib/genocean,bib/maths,bib/mitgcmuv,bib/fram}
736 dimitri 1.1
737     \end{document}
738    
739     %%% Local Variables:
740     %%% mode: latex
741     %%% TeX-master: t
742     %%% End:
743 mlosch 1.4
744    
745     A Dynamic-Thermodynamic Sea ice Model for Ocean Climate
746     Estimation on an Arakawa C-Grid
747    
748     Introduction
749    
750     Ice Model:
751     Dynamics formulation.
752     B-C, LSR, EVP, no-slip, slip
753     parallellization
754     Thermodynamics formulation.
755     0-layer Hibler salinity + snow
756     3-layer Winton
757    
758     Idealized tests
759     Funnel Experiments
760     Downstream Island tests
761     B-grid LSR no-slip
762     C-grid LSR no-slip
763     C-grid LSR slip
764     C-grid EVP no-slip
765     C-grid EVP slip
766    
767     Arctic Setup
768     Configuration
769     OBCS from cube
770     forcing
771     1/2 and full resolution
772     with a few JFM figs from C-grid LSR no slip
773     ice transport through Canadian Archipelago
774     thickness distribution
775     ice velocity and transport
776    
777     Arctic forward sensitivity experiments
778     B-grid LSR no-slip
779     C-grid LSR no-slip
780     C-grid LSR slip
781     C-grid EVP no-slip
782     C-grid EVP slip
783     C-grid LSR no-slip + Winton
784     speed-performance-accuracy (small)
785     ice transport through Canadian Archipelago differences
786     thickness distribution differences
787     ice velocity and transport differences
788    
789     Adjoint sensitivity experiment on 1/2-res setup
790     Sensitivity of sea ice volume flow through Fram Strait
791     *** Sensitivity of sea ice volume flow through Canadian Archipelago
792    
793     Summary and conluding remarks

  ViewVC Help
Powered by ViewVC 1.1.22