/[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 1.4 - (hide annotations) (download) (as text)
Thu Jan 10 15:47:32 2008 UTC (17 years, 6 months ago) by mlosch
Branch: MAIN
Changes since 1.3: +165 -19 lines
File MIME type: application/x-tex
saving todays works:
- write something about the funnel experiments and start describing
  downstream island experiments relating to Hunke (2001)
- fix a few paths and usepackage statements

1 dimitri 1.1 \documentclass[12pt]{article}
2 mlosch 1.4
3     \usepackage{graphicx,subfigure}
4 dimitri 1.1
5     \usepackage[round,comma]{natbib}
6 dimitri 1.2 \bibliographystyle{bib/agu04}
7 dimitri 1.1
8     \usepackage{amsmath,amssymb}
9     \newcommand\bmmax{10} \newcommand\hmmax{10}
10     \usepackage{bm}
11    
12     % math abbreviations
13     \newcommand{\vek}[1]{\ensuremath{\mathbf{#1}}}
14     \newcommand{\mat}[1]{\ensuremath{\mathbf{#1}}}
15     \newcommand{\vtau}{\bm{{\tau}}}
16    
17     \newcommand{\degree}{\ensuremath{^\circ}}
18     \newcommand{\degC}{\,\ensuremath{\degree}C}
19     \newcommand{\degE}{\ensuremath{\degree}\,E}
20     \newcommand{\degS}{\ensuremath{\degree}\,S}
21     \newcommand{\degN}{\ensuremath{\degree}\,N}
22     \newcommand{\degW}{\ensuremath{\degree}\,W}
23    
24     % cross reference scheme
25     \newcommand{\reffig}[1]{Figure~\ref{fig:#1}}
26     \newcommand{\reftab}[1]{Table~\ref{tab:#1}}
27     \newcommand{\refapp}[1]{Appendix~\ref{app:#1}}
28     \newcommand{\refsec}[1]{Section~\ref{sec:#1}}
29     \newcommand{\refeq}[1]{\,(\ref{eq:#1})}
30     \newcommand{\refeqs}[2]{\,(\ref{eq:#1})--(\ref{eq:#2})}
31    
32     \newlength{\stdfigwidth}\setlength{\stdfigwidth}{20pc}
33     %\newlength{\stdfigwidth}\setlength{\stdfigwidth}{\columnwidth}
34     \newlength{\mediumfigwidth}\setlength{\mediumfigwidth}{39pc}
35     %\newlength{\widefigwidth}\setlength{\widefigwidth}{39pc}
36     \newlength{\widefigwidth}\setlength{\widefigwidth}{\textwidth}
37 mlosch 1.4 \newcommand{\fpath}{figs}
38    
39     % commenting scheme
40     \newcommand{\ml}[1]{\textsf{\slshape #1}}
41 dimitri 1.1
42     \title{A Dynamic-Thermodynamic Sea ice Model for Ocean Climate
43     Estimation on an Arakawa C-Grid}
44    
45     \author{Martin Losch, Dimitris Menemenlis, Patrick Heimbach, \\
46     Jean-Michel Campin, and Chris Hill}
47     \begin{document}
48    
49     \maketitle
50    
51     \begin{abstract}
52     Some blabla
53     \end{abstract}
54    
55     \section{Introduction}
56     \label{sec:intro}
57    
58     more blabla
59    
60     \section{Model}
61     \label{sec:model}
62    
63     Traditionally, probably for historical reasons and the ease of
64     treating the Coriolis term, most standard sea-ice models are
65     discretized on Arakawa-B-grids \citep[e.g.,][]{hibler79, harder99,
66     kreyscher00, zhang98, hunke97}. From the perspective of coupling a
67     sea ice-model to a C-grid ocean model, the exchange of fluxes of heat
68     and fresh-water pose no difficulty for a B-grid sea-ice model
69     \citep[e.g.,][]{timmermann02a}. However, surface stress is defined at
70     velocities points and thus needs to be interpolated between a B-grid
71     sea-ice model and a C-grid ocean model. While the smoothing implicitly
72     associated with this interpolation may mask grid scale noise, it may
73     in two-way coupling lead to a computational mode as will be shown. By
74     choosing a C-grid for the sea-ice model, we circumvene this difficulty
75     altogether and render the stress coupling as consistent as the
76     buoyancy coupling.
77    
78     A further advantage of the C-grid formulation is apparent in narrow
79     straits. In the limit of only one grid cell between coasts there is no
80     flux allowed for a B-grid (with no-slip lateral boundary counditions),
81     whereas the C-grid formulation allows a flux of sea-ice through this
82     passage for all types of lateral boundary conditions. We (will)
83     demonstrate this effect in the Candian archipelago.
84    
85     \subsection{Dynamics}
86     \label{sec:dynamics}
87    
88     The momentum equations of the sea-ice model are standard with
89     \begin{equation}
90     \label{eq:momseaice}
91     m \frac{D\vek{u}}{Dt} = -mf\vek{k}\times\vek{u} + \vtau_{air} +
92     \vtau_{ocean} - m \nabla{\phi(0)} + \vek{F},
93     \end{equation}
94     where $\vek{u} = u\vek{i}+v\vek{j}$ is the ice velocity vectory, $m$
95     the ice mass per unit area, $f$ the Coriolis parameter, $g$ is the
96     gravity accelation, $\nabla\phi$ is the gradient (tilt) of the sea
97     surface height potential beneath the ice. $\phi$ is the sum of
98     atmpheric pressure $p_{a}$ and loading due to ice and snow
99     $(m_{i}+m_{s})g$. $\vtau_{air}$ and $\vtau_{ocean}$ are the wind and
100     ice-ocean stresses, respectively. $\vek{F}$ is the interaction force
101     and $\vek{i}$, $\vek{j}$, and $\vek{k}$ are the unit vectors in the
102     $x$, $y$, and $z$ directions. Advection of sea-ice momentum is
103     neglected. The wind and ice-ocean stress terms are given by
104     \begin{align*}
105     \vtau_{air} =& \rho_{air} |\vek{U}_{air}|R_{air}(\vek{U}_{air}) \\
106     \vtau_{ocean} =& \rho_{ocean} |\vek{U}_{ocean}-\vek{u}|
107     R_{ocean}(\vek{U}_{ocean}-\vek{u}), \\
108     \end{align*}
109     where $\vek{U}_{air/ocean}$ are the surface winds of the atmosphere
110     and surface currents of the ocean, respectively. $C_{air/ocean}$ are
111     air and ocean drag coefficients, $\rho_{air/ocean}$ reference
112     densities, and $R_{air/ocean}$ rotation matrices that act on the
113     wind/current vectors. $\vek{F} = \nabla\cdot\sigma$ is the divergence
114     of the interal stress tensor $\sigma_{ij}$.
115    
116     For an isotropic system this stress tensor can be related to the ice
117     strain rate and strength by a nonlinear viscous-plastic (VP)
118     constitutive law \citep{hibler79, zhang98}:
119     \begin{equation}
120     \label{eq:vpequation}
121     \sigma_{ij}=2\eta(\dot{\epsilon}_{ij},P)\dot{\epsilon}_{ij}
122     + \left[\zeta(\dot{\epsilon}_{ij},P) -
123     \eta(\dot{\epsilon}_{ij},P)\right]\dot{\epsilon}_{kk}\delta_{ij}
124     - \frac{P}{2}\delta_{ij}.
125     \end{equation}
126     The ice strain rate is given by
127     \begin{equation*}
128     \dot{\epsilon}_{ij} = \frac{1}{2}\left(
129     \frac{\partial{u_{i}}}{\partial{x_{j}}} +
130     \frac{\partial{u_{j}}}{\partial{x_{i}}}\right).
131     \end{equation*}
132     The pressure $P$, a measure of ice strength, depends on both thickness
133 mlosch 1.4 $h$ and compactness (concentration) $c$:
134     \begin{equation}
135     P = P^{*}c\,h\,e^{[C^{*}\cdot(1-c)]},
136     \label{icestrength}
137     \end{equation}
138     with the constants $P^{*}$ and $C^{*}$. The nonlinear bulk and shear
139     viscosities $\eta$ and $\zeta$ are functions of ice strain rate
140     invariants and ice strength such that the principal components of the
141     stress lie on an elliptical yield curve with the ratio of major to
142     minor axis $e$ equal to $2$; they are given by:
143 dimitri 1.1 \begin{align*}
144     \zeta =& \frac{P}{2\Delta} \\
145     \eta =& \frac{P}{2\Delta{e}^2} \\
146     \intertext{with the abbreviation}
147     \Delta = & \left[
148     \left(\dot{\epsilon}_{11}^2+\dot{\epsilon}_{22}^2\right)
149     (1+e^{-2}) + 4e^{-2}\dot{\epsilon}_{12}^2 +
150     2\dot{\epsilon}_{11}\dot{\epsilon}_{22} (1-e^{-2})
151     \right]^{-\frac{1}{2}}
152     \end{align*}
153     In the current implementation, the VP-model is integrated with the
154     semi-implicit line successive over relaxation (LSOR)-solver of
155     \citet{zhang98}, which allows for long time steps that, in our case,
156     is limited by the explicit treatment of the Coriolis term. The
157     explicit treatment of the Coriolis term does not represent a severe
158     limitation because it restricts the time step to approximately the
159     same length as in the ocean model where the Coriolis term is also
160     treated explicitly.
161    
162     \citet{hunke97}'s introduced an elastic contribution to the strain
163     rate elatic-viscous-plastic in order to regularize
164     Eq.\refeq{vpequation} in such a way that the resulting
165     elatic-viscous-plastic (EVP) and VP models are identical at steady
166     state,
167     \begin{equation}
168     \label{eq:evpequation}
169     \frac{1}{E}\frac{\partial\sigma_{ij}}{\partial{t}} +
170     \frac{1}{2\eta}\sigma_{ij}
171     + \frac{\eta - \zeta}{4\zeta\eta}\sigma_{kk}\delta_{ij}
172     + \frac{P}{4\zeta}\delta_{ij}
173     = \dot{\epsilon}_{ij}.
174     \end{equation}
175     %In the EVP model, equations for the components of the stress tensor
176     %$\sigma_{ij}$ are solved explicitly. Both model formulations will be
177     %used and compared the present sea-ice model study.
178     The EVP-model uses an explicit time stepping scheme with a short
179     timestep. According to the recommendation of \citet{hunke97}, the
180     EVP-model is stepped forward in time 120 times within the physical
181     ocean model time step (although this parameter is under debate), to
182     allow for elastic waves to disappear. Because the scheme does not
183     require a matrix inversion it is fast in spite of the small timestep
184     \citep{hunke97}. For completeness, we repeat the equations for the
185     components of the stress tensor $\sigma_{1} =
186     \sigma_{11}+\sigma_{22}$, $\sigma_{2}= \sigma_{11}-\sigma_{22}$, and
187     $\sigma_{12}$. Introducing the divergence $D_D =
188     \dot{\epsilon}_{11}+\dot{\epsilon}_{22}$, and the horizontal tension
189     and shearing strain rates, $D_T =
190     \dot{\epsilon}_{11}-\dot{\epsilon}_{22}$ and $D_S =
191     2\dot{\epsilon}_{12}$, respectively and using the above abbreviations,
192     the equations can be written as:
193     \begin{align}
194     \label{eq:evpstresstensor1}
195     \frac{\partial\sigma_{1}}{\partial{t}} + \frac{\sigma_{1}}{2T} +
196     \frac{P}{2T} &= \frac{P}{2T\Delta} D_D \\
197     \label{eq:evpstresstensor2}
198     \frac{\partial\sigma_{2}}{\partial{t}} + \frac{\sigma_{2} e^{2}}{2T}
199     &= \frac{P}{2T\Delta} D_T \\
200     \label{eq:evpstresstensor12}
201     \frac{\partial\sigma_{12}}{\partial{t}} + \frac{\sigma_{12} e^{2}}{2T}
202     &= \frac{P}{4T\Delta} D_S
203     \end{align}
204     Here, the elastic parameter $E$ is redefined in terms of a damping timescale
205     $T$ for elastic waves \[E=\frac{\zeta}{T}.\]
206     $T=E_{0}\Delta{t}$ with the tunable parameter $E_0<1$ and
207     the external (long) timestep $\Delta{t}$. \citet{hunke97} recommend
208     $E_{0} = \frac{1}{3}$.
209    
210     For details of the spatial discretization, the reader is referred to
211     \citet{zhang98, zhang03}. Our discretization differs only (but
212     importantly) in the underlying grid, namely the Arakawa C-grid, but is
213     otherwise straightforward. The EVP model in particular is discretized
214     naturally on the C-grid with $\sigma_{1}$ and $\sigma_{2}$ on the
215     center points and $\sigma_{12}$ on the corner (or vorticity) points of
216     the grid. With this choice all derivatives are discretized as central
217     differences and averaging is only involved in computing $\Delta$ and
218     $P$ at vorticity points.
219    
220     For a general curvilinear grid, one needs in principle to take metric
221     terms into account that arise in the transformation a curvilinear grid
222     on the sphere. However, for now we can neglect these metric terms
223     because they are very small on the cubed sphere grids used in this
224     paper; in particular, only near the edges of the cubed sphere grid, we
225     expect them to be non-zero, but these edges are at approximately
226     35\degS\ or 35\degN\ which are never covered by sea-ice in our
227     simulations. Everywhere else the coordinate system is locally nearly
228     cartesian. However, for last-glacial-maximum or snowball-earth-like
229     simulations the question of metric terms needs to be reconsidered.
230     Either, one includes these terms as in \citet{zhang03}, or one finds a
231     vector-invariant formulation fo the sea-ice internal stress term that
232     does not require any metric terms, as it is done in the ocean dynamics
233     of the MITgcm \citep{adcroft04:_cubed_sphere}.
234    
235     Moving sea ice exerts a stress on the ocean which is the opposite of
236     the stress $\vtau_{ocean}$ in Eq.\refeq{momseaice}. This stess is
237     applied directly to the surface layer of the ocean model. An
238     alternative ocean stress formulation is given by \citet{hibler87}.
239     Rather than applying $\vtau_{ocean}$ directly, the stress is derived
240     from integrating over the ice thickness to the bottom of the oceanic
241     surface layer. In the resulting equation for the \emph{combined}
242     ocean-ice momentum, the interfacial stress cancels and the total
243     stress appears as the sum of windstress and divergence of internal ice
244     stresses: $\delta(z) (\vtau_{air} + \vek{F})/\rho_0$, \citep[see also
245     Eq.\,2 of][]{hibler87}. The disadvantage of this formulation is that
246     now the velocity in the surface layer of the ocean that is used to
247     advect tracers, is really an average over the ocean surface
248     velocity and the ice velocity leading to an inconsistency as the ice
249     temperature and salinity are different from the oceanic variables.
250    
251     Sea ice distributions are characterized by sharp gradients and edges.
252     For this reason, we employ a positive 3rd-order advection scheme
253     \citep{hundsdorfer94} for the thermodynamic variables discussed in the
254     next section.
255    
256     \subparagraph{boundary conditions: no-slip, free-slip, half-slip}
257    
258     \begin{itemize}
259     \item transition from existing B-Grid to C-Grid
260     \item boundary conditions: no-slip, free-slip, half-slip
261     \item fancy (multi dimensional) advection schemes
262     \item VP vs.\ EVP \citep{hunke97}
263     \item ocean stress formulation \citep{hibler87}
264     \end{itemize}
265    
266     \subsection{Thermodynamics}
267     \label{sec:thermodynamics}
268    
269     In the original formulation the sea ice model \citep{menemenlis05}
270     uses simple thermodynamics following the appendix of
271     \citet{semtner76}. This formulation does not allow storage of heat
272     (heat capacity of ice is zero, and this type of model is often refered
273     to as a ``zero-layer'' model). Upward heat flux is parameterized
274     assuming a linear temperature profile and together with a constant ice
275     conductivity. It is expressed as $(K/h)(T_{w}-T_{0})$, where $K$ is
276     the ice conductivity, $h$ the ice thickness, and $T_{w}-T_{0}$ the
277     difference between water and ice surface temperatures. The surface
278     heat budget is computed in a similar way to that of
279     \citet{parkinson79} and \citet{manabe79}.
280    
281     There is considerable doubt about the reliability of such a simple
282     thermodynamic model---\citet{semtner84} found significant errors in
283     phase (one month lead) and amplitude ($\approx$50\%\,overestimate) in
284     such models---, so that today many sea ice models employ more complex
285     thermodynamics. A popular thermodynamics model of \citet{winton00} is
286     based on the 3-layer model of \citet{semtner76} and treats brine
287     content by means of enthalphy conservation. This model requires in
288     addition to ice-thickness and compactness (fractional area) additional
289     state variables to be advected by ice velocities, namely enthalphy of
290     the two ice layers and the thickness of the overlying snow layer.
291    
292     \section{Funnel Experiments}
293     \label{sec:funnel}
294    
295 mlosch 1.4 For a first/detailed comparison between the different variants of the
296     MIT sea ice model an idealized geometry of a periodic channel,
297     1000\,km long and 500\,m wide on a non-rotating plane, with converging
298     walls forming a symmetric funnel and a narrow strait of 40\,km width
299     is used. The horizontal resolution is 5\,km throughout the domain
300     making the narrow strait 8 grid points wide. The ice model is
301     initialized with a complete ice cover of 50\,cm uniform thickness. The
302     ice model is driven by a constant along channel eastward ocean current
303     of 25\,cm/s that does not see the walls in the domain. All other
304     ice-ocean-atmosphere interactions are turned off, in particular there
305     is no feedback of ice dynamics on the ocean current. All thermodynamic
306     processes are turned off so that ice thickness variations are only
307     caused by convergent or divergent ice flow. Ice volume (effective
308     thickness) and concentration are advected with a third-order scheme
309     with a flux limiter \citep{hundsdorfer94} to avoid undershoots. This
310     scheme is unconditionally stable and does not require additional
311     diffusion. The time step used here is 1\,h.
312    
313     \reffig{funnelf0} compares the dynamic fields ice concentration $c$,
314     effective thickness $h_{eff} = h\cdot{c}$, and velocities $(u,v)$ for
315     five different cases at steady state (after 10\,years of integration):
316     \begin{description}
317     \item[B-LSRns:] LSR solver with no-slip boundary conditions on a B-grid,
318     \item[C-LSRns:] LSR solver with no-slip boundary conditions on a C-grid,
319     \item[C-LSRfs:] LSR solver with free-slip boundary conditions on a C-grid,
320     \item[C-EVPns:] EVP solver with no-slip boundary conditions on a C-grid,
321     \item[C-EVPfs:] EVP solver with free-slip boundary conditions on a C-grid,
322     \end{description}
323     \ml{[We have not implemented the EVP solver on a B-grid.]}
324     \begin{figure*}[htbp]
325     \includegraphics[width=\widefigwidth]{\fpath/all_086280}
326     \caption{Ice concentration, effective thickness [m], and ice
327     velocities [m/s]
328     for 5 different numerical solutions.}
329     \label{fig:funnelf0}
330     \end{figure*}
331     At a first glance, the solutions look similar. This is encouraging as
332     the details of discretization and numerics should not affect the
333     solutions to first order. In all cases the ice-ocean stress pushes the
334     ice cover eastwards, where it converges in the funnel. In the narrow
335     channel the ice moves quickly (nearly free drift) and leaves the
336     channel as narrow band.
337    
338     A close look reveals interesting differences between the B- and C-grid
339     results. The zonal velocity in the narrow channel is nearly the free
340     drift velocity ( = ocean velocity) of 25\,cm/s for the C-grid
341     solutions, regardless of the boundary conditions, while it is just
342     above 20\,cm/s for the B-grid solution. The ice accelerates to
343     25\,cm/s after it exits the channel. Concentrating on the solutions
344     B-LSRns and C-LSRns, the ice volume (effective thickness) along the
345     boundaries in the narrow channel is larger in the B-grid case although
346     the ice concentration is reduces in the C-grid case. The combined
347     effect leads to a larger actual ice thickness at smaller
348     concentrations in the C-grid case. However, since the effective
349     thickness determines the ice strength $P$ in Eq\refeq{icestrength},
350     the ice strength and thus the bulk and shear viscosities are larger in
351     the B-grid case leading to more horizontal friction. This circumstance
352     might explain why the no-slip boundary conditions in the B-grid case
353     appear to be more effective in reducing the flow within the narrow
354     channel, than in the C-grid case. Further, the viscosities are also
355     sensitive to details of the velocity gradients. Via $\Delta$, these
356     gradients enter the viscosities in the denominator so that large
357     gradients tend to reduce the viscosities. This again favors more flow
358     along the boundaries in the C-grid case: larger velocities
359     (\reffig{funnelf0}) on grid points that are closer to the boundary by
360     a factor $\frac{1}{2}$ than in the B-grid case because of the stagger
361     nature of the C-grid lead numerically to larger tangential gradients
362     across the boundary; these in turn make the viscosities smaller for
363     less tangential friction and allow more tangential flow along the
364     boundaries.
365    
366     The above argument can also be invoked to explain the small
367     differences between the free-slip and no-slip solutions on the C-grid.
368     Because of the non-linearities in the ice viscosities, in particular
369     along the boundaries, the no-slip boundary conditions has only a small
370     impact on the solution.
371    
372     The difference between LSR and EVP solutions is largest in the
373     effective thickness and meridional velocity fields. The velocity field
374     appears to be a little noisy. This noise has been address by
375     \citet{hunke01}. It can be dealt with by reducing EVP's internal time
376     step (increasing the number of iterations) or by regularizing the bulk
377     and shear viscosities. We revisit the latter option by reproducing the
378     results of \citet{hunke01} for the C-grid no-slip cases.
379     \begin{figure*}[htbp]
380     \includegraphics[width=\widefigwidth]{\fpath/hun12days}
381     \caption{Hunke's test case.}
382     \label{fig:hunke01}
383     \end{figure*}
384    
385 dimitri 1.1 \begin{itemize}
386     \item B-grid LSR no-slip
387     \item C-grid LSR no-slip
388     \item C-grid LSR slip
389     \item C-grid EVP no-slip
390     \item C-grid EVP slip
391     \end{itemize}
392    
393     \subsection{B-grid vs.\ C-grid}
394     Comparison between:
395     \begin{itemize}
396     \item B-grid, lsr, no-slip
397     \item C-grid, lsr, no-slip
398     \item C-grid, evp, no-slip
399     \end{itemize}
400     all without ice-ocean stress, because ice-ocean stress does not work
401     for B-grid.
402    
403     \subsection{C-grid}
404     \begin{itemize}
405     \item no-slip vs. free-slip for both lsr and evp;
406     "diagnostics" to look at and use for comparison
407     \begin{itemize}
408     \item ice transport through Fram Strait/Denmark Strait/Davis
409     Strait/Bering strait (these are general)
410     \item ice transport through narrow passages, e.g.\ Nares-Strait
411     \end{itemize}
412     \item compare different advection schemes (if lsr turns out to be more
413     effective, then with lsr otherwise I prefer evp), eg.
414     \begin{itemize}
415     \item default 2nd-order with diff1=0.002
416     \item 1st-order upwind with diff1=0.
417     \item DST3FL (SEAICEadvScheme=33 with diff1=0., should work, works for me)
418     \item 2nd-order wit flux limiter (SEAICEadvScheme=77 with diff1=0.)
419     \end{itemize}
420     That should be enough. Here, total ice mass and location of ice edge
421     is interesting. However, this comparison can be done in an idealized
422     domain, may not require full Arctic Domain?
423     \item
424     Do a little study on the parameters of LSR and EVP
425     \begin{enumerate}
426     \item convergence of LSR, how many iterations do you need to get a
427     true elliptic yield curve
428     \item vary deltaTevp and the relaxation parameter for EVP and see when
429     the EVP solution breaks down (relative to the forcing time scale).
430     For this, it is essential that the evp solver gives use "stripeless"
431     solutions, that is your dtevp = 1sec solutions/or 10sec solutions
432     with SEAICE\_evpDampC = 615.
433     \end{enumerate}
434     \end{itemize}
435    
436     \section{Forward sensitivity experiments}
437     \label{sec:forward}
438    
439     A second series of forward sensitivity experiments have been carried out on an
440     Arctic Ocean domain with open boundaries. Once again the objective is to
441     compare the old B-grid LSR dynamic solver with the new C-grid LSR and EVP
442     solvers. One additional experiment is carried out to illustrate the
443     differences between the two main options for sea ice thermodynamics in the MITgcm.
444    
445     \subsection{Arctic Domain with Open Boundaries}
446     \label{sec:arctic}
447    
448     The Arctic domain of integration is illustrated in Fig.~\ref{???}. It is
449     carved out from, and obtains open boundary conditions from, the global
450     cubed-sphere configuration of the Estimating the Circulation and Climate of
451     the Ocean, Phase II (ECCO2) project \cite{men05a}. The domain size is 420 by
452     384 grid boxes horizontally with mean horizontal grid spacing of 18 km.
453    
454     There are 50 vertical levels ranging in thickness from 10 m near the surface
455     to approximately 450 m at a maximum model depth of 6150 m. Bathymetry is from
456     the National Geophysical Data Center (NGDC) 2-minute gridded global relief
457     data (ETOPO2) and the model employs the partial-cell formulation of
458     \cite{adc97}, which permits accurate representation of the bathymetry. The
459     model is integrated in a volume-conserving configuration using a finite volume
460     discretization with C-grid staggering of the prognostic variables. In the
461     ocean, the non-linear equation of state of \cite{jac95}. The ocean model is
462     coupled to a sea-ice model described hereinabove.
463    
464     This particular ECCO2 simulation is initialized from rest using the January
465     temperature and salinity distribution from the World Ocean Atlas 2001 (WOA01)
466     [Conkright et al., 2002] and it is integrated for 32 years prior to the
467     1996-2001 period discussed in the study. Surface boundary conditions are from
468     the National Centers for Environmental Prediction and the National Center for
469     Atmospheric Research (NCEP/NCAR) atmospheric reanalysis [Kistler et al.,
470     2001]. Six-hourly surface winds, temperature, humidity, downward short- and
471     long-wave radiations, and precipitation are converted to heat, freshwater, and
472     wind stress fluxes using the Large and Pond [1981, 1982] bulk
473     formulae. Shortwave radiation decays exponentially as per Paulson and Simpson
474     [1977]. Additionally the time-mean river run-off from Large and Nurser [2001]
475     is applied and there is a relaxation to the monthly-mean climatological sea
476     surface salinity values from WOA01 with a relaxation time scale of 3
477     months. Vertical mixing follows Large et al. [1994] with background vertical
478     diffusivity of 1.5 × 10-5 m2 s-1 and viscosity of 10-3 m2 s-1. A third order,
479     direct-space-time advection scheme with flux limiter is employed and there is
480     no explicit horizontal diffusivity. Horizontal viscosity follows Leith [1996]
481     but modified to sense the divergent flow as per Fox-Kemper and Menemenlis [in
482     press]. Shortwave radiation decays exponentially as per Paulson and Simpson
483     [1977]. Additionally, the time-mean runoff of Large and Nurser [2001] is
484     applied near the coastline and, where there is open water, there is a
485     relaxation to monthly-mean WOA01 sea surface salinity with a time constant of
486     45 days.
487    
488     Open water, dry
489     ice, wet ice, dry snow, and wet snow albedo are, respectively, 0.15, 0.85,
490     0.76, 0.94, and 0.8.
491    
492     \begin{itemize}
493     \item Configuration
494     \item OBCS from cube
495     \item forcing
496     \item 1/2 and full resolution
497     \item with a few JFM figs from C-grid LSR no slip
498     ice transport through Canadian Archipelago
499     thickness distribution
500     ice velocity and transport
501     \end{itemize}
502    
503     \begin{itemize}
504     \item Arctic configuration
505     \item ice transport through straits and near boundaries
506     \item focus on narrow straits in the Canadian Archipelago
507     \end{itemize}
508    
509     \begin{itemize}
510     \item B-grid LSR no-slip
511     \item C-grid LSR no-slip
512     \item C-grid LSR slip
513     \item C-grid EVP no-slip
514     \item C-grid EVP slip
515     \item C-grid LSR no-slip + Winton
516     \item speed-performance-accuracy (small)
517     ice transport through Canadian Archipelago differences
518     thickness distribution differences
519     ice velocity and transport differences
520     \end{itemize}
521    
522     We anticipate small differences between the different models due to:
523     \begin{itemize}
524     \item advection schemes: along the ice-edge and regions with large
525     gradients
526     \item C-grid: more transport through narrow straits for no slip
527     conditons, less for free slip
528     \item VP vs.\ EVP: speed performance, accuracy?
529     \item ocean stress: different water mass properties beneath the ice
530     \end{itemize}
531    
532     \section{Adjoint sensiivities of the MITsim}
533    
534     \subsection{The adjoint of MITsim}
535    
536     The ability to generate tangent linear and adjoint model components
537     of the MITsim has been a main design task.
538     For the ocean the adjoint capability has proven to be an
539     invaluable tool for sensitivity analysis as well as state estimation.
540     In short, the adjoint enables very efficient computation of the gradient
541     of scalar-valued model diagnostics (called cost function or objective function)
542     with respect to many model "variables".
543     These variables can be two- or three-dimensional fields of initial
544     conditions, model parameters such as mixing coefficients, or
545     time-varying surface or lateral (open) boundary conditions.
546     When combined, these variables span a potentially high-dimensional
547     (e.g. O(10$^8$)) so-called control space. Performing parameter perturbations
548     to assess model sensitivities quickly becomes prohibitive at these scales.
549     Alternatively, (time-varying) sensitivities of the objective function
550     to any element of the control space can be computed very efficiently in
551     one single adjoint
552     model integration, provided an efficient adjoint model is available.
553    
554     [REFERENCES]
555    
556    
557     The adjoint operator (ADM) is the transpose of the tangent linear operator (TLM)
558     of the full (in general nonlinear) forward model, i.e. the MITsim.
559     The TLM maps perturbations of elements of the control space
560     (e.g. initial ice thickness distribution)
561     via the model Jacobian
562     to a perturbation in the objective function
563     (e.g. sea-ice export at the end of the integration interval).
564     \textit{Tangent} linearity ensures that the derivatives are evaluated
565     with respect to the underlying model trajectory at each point in time.
566     This is crucial for nonlinear trajectories and the presence of different
567     regimes (e.g. effect of the seaice growth term at or away from the
568     freezing point of the ocean surface).
569     Ensuring tangent linearity can be easily achieved by integrating
570     the full model in sync with the TLM to provide the underlying model state.
571     Ensuring \textit{tangent} adjoints is equally crucial, but much more
572     difficult to achieve because of the reverse nature of the integration:
573     the adjoint accumulates sensitivities backward in time,
574     starting from a unit perturbation of the objective function.
575     The adjoint model requires the model state in reverse order.
576     This presents one of the major complications in deriving an
577     exact, i.e. \textit{tangent} adjoint model.
578    
579     Following closely the development and maintenance of TLM and ADM
580     components of the MITgcm we have relied heavily on the
581     autmomatic differentiation (AD) tool
582     "Transformation of Algorithms in Fortran" (TAF)
583     developed by Fastopt (Giering and Kaminski, 1998)
584     to derive TLM and ADM code of the MITsim.
585     Briefly, the nonlinear parent model is fed to the AD tool which produces
586     derivative code for the specified control space and objective function.
587     Following this approach has (apart from its evident success)
588     several advantages:
589     (1) the adjoint model is the exact adjoint operator of the parent model,
590     (2) the adjoint model can be kept up to date with respect to ongoing
591     development of the parent model, and adjustments to the parent model
592     to extend the automatically generated adjoint are incremental changes
593     only, rather than extensive re-developments,
594     (3) the parallel structure of the parent model is preserved
595     by the adjoint model, ensuring efficient use in high performance
596     computing environments.
597    
598     Some initial code adjustments are required to support dependency analysis
599     of the flow reversal and certain language limitations which may lead
600     to irreducible flow graphs (e.g. GOTO statements).
601     The problem of providing the required model state in reverse order
602     at the time of evaluating nonlinear or conditional
603     derivatives is solved via balancing
604     storing vs. recomputation of the model state in a multi-level
605     checkpointing loop.
606     Again, an initial code adjustment is required to support TAFs
607     checkpointing capability.
608     The code adjustments are sufficiently simply so as not to cause
609     major limitations to the full nonlinear parent model.
610     Once in place, an adjoint model of a new model configuration
611     may be derived in about 10 minutes.
612    
613     [HIGHLIGHT COUPLED NATURE OF THE ADJOINT!]
614    
615     \subsection{Special considerations}
616    
617     * growth term(?)
618    
619     * small active denominators
620    
621     * dynamic solver (implicit function theorem)
622    
623     * approximate adjoints
624    
625    
626     \subsection{An example: sensitivities of sea-ice export through Fram Strait}
627    
628     We demonstrate the power of the adjoint method
629     in the context of investigating sea-ice export sensitivities through Fram Strait
630     (for details of this study see Heimbach et al., 2007).
631     The domain chosen is a coarsened version of the Arctic face of the
632     high-resolution cubed-sphere configuration of the ECCO2 project
633     (see Menemenlis et al. 2005). It covers the entire Arctic,
634     extends into the North Pacific such as to cover the entire
635     ice-covered regions, and comprises parts of the North Atlantic
636     down to XXN to enable analysis of remote influences of the
637     North Atlantic current to sea-ice variability and export.
638     The horizontal resolution varies between XX and YY km
639     with 50 unevenly spaced vertical levels.
640     The adjoint models run efficiently on 80 processors
641     (benchmarks have been performed both on an SGI Altix as well as an
642     IBM SP5 at NASA/ARC).
643    
644     Following a 1-year spinup, the model has been integrated for four years
645     between 1992 and 1995.
646     It is forced using realistic 6-hourly NCEP/NCAR atmospheric state variables.
647     Over the open ocean these are converted into
648     air-sea fluxes via the bulk formulae of Large and Yeager (2004).
649     Derivation of air-sea fluxes in the presence of sea-ice is handled
650     by the ice model as described in Section XXX.
651     The objective function chosen is sea-ice export through Fram Strait
652     computed for December 1995
653     The adjoint model computes sensitivities to sea-ice export back in time
654     from 1995 to 1992 along this trajectory.
655     In principle all adjoint model variable (i.e. Lagrange multipliers)
656     of the coupled ocean/sea-ice model
657     are available to analyze the transient sensitivity behaviour
658     of the ocean and sea-ice state.
659     Over the open ocean, the adjoint of the bulk formula scheme
660     computes sensitivities to the time-varying atmospheric state.
661     Over ice-covered parts, the sea-ice adjoint converts
662     surface ocean sensitivities to atmospheric sensitivities.
663    
664     Fig. XXX(a--d) depict sensitivities of sea-ice export through Fram Strait
665     in December 1995 to changes in sea-ice thickness
666     12, 24, 36, 48 months back in time.
667     Corresponding sensitivities to ocean surface temperature are
668     depicted in Fig. XXX(a--d).
669     The main characteristics is consistency with expected advection
670     of sea-ice over the relevant time scales considered.
671     The general positive pattern means that an increase in
672     sea-ice thickness at location $(x,y)$ and time $t$ will increase
673     sea-ice export through Fram Strait at time $T_e$.
674     Largest distances from Fram Strait indicate fastest sea-ice advection
675     over the time span considered.
676     The ice thickness sensitivities are in close correspondence to
677     ocean surface sentivitites, but of opposite sign.
678     An increase in temperature will incur ice melting, decrease in ice thickness,
679     and therefore decrease in sea-ice export at time $T_e$.
680    
681     The picture is fundamentally different and much more complex
682     for sensitivities to ocean temperatures away from the surface.
683     Fig. XXX (a--d) depicts ice export sensitivities to
684     temperatures at roughly 400 m depth.
685     Primary features are the effect of the heat transport of the North
686     Atlantic current which feeds into the West Spitsbergen current,
687     the circulation around Svalbard, and ...
688    
689     \begin{figure}[t!]
690     \centerline{
691     \subfigure[{\footnotesize -12 months}]
692 mlosch 1.4 {\includegraphics*[width=0.44\linewidth]{\fpath/run_4yr_ADJheff_arc_lev1_tim072_cmax2.0E+02.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_ADJheff_arc_lev1_tim145_cmax2.0E+02.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_ADJheff_arc_lev1_tim218_cmax2.0E+02.eps}}
703 dimitri 1.1 %
704     \subfigure[{\footnotesize
705     -48 months}]
706 mlosch 1.4 {\includegraphics*[width=0.44\linewidth]{\fpath/run_4yr_ADJheff_arc_lev1_tim292_cmax2.0E+02.eps}}
707 dimitri 1.1 }
708     \caption{Sensitivity of sea-ice export through Fram Strait in December 2005 to
709     sea-ice thickness at various prior times.
710     \label{fig:4yradjheff}}
711     \end{figure}
712    
713    
714     \begin{figure}[t!]
715     \centerline{
716     \subfigure[{\footnotesize -12 months}]
717 mlosch 1.4 {\includegraphics*[width=0.44\linewidth]{\fpath/run_4yr_ADJtheta_arc_lev1_tim072_cmax5.0E+01.eps}}
718 dimitri 1.1 %\includegraphics*[width=.3\textwidth]{H_c.bin_res_100_lev1.pdf}
719     %
720     \subfigure[{\footnotesize -24 months}]
721 mlosch 1.4 {\includegraphics*[width=0.44\linewidth]{\fpath/run_4yr_ADJtheta_arc_lev1_tim145_cmax5.0E+01.eps}}
722 dimitri 1.1 }
723    
724     \centerline{
725     \subfigure[{\footnotesize
726     -36 months}]
727 mlosch 1.4 {\includegraphics*[width=0.44\linewidth]{\fpath/run_4yr_ADJtheta_arc_lev1_tim218_cmax5.0E+01.eps}}
728 dimitri 1.1 %
729     \subfigure[{\footnotesize
730     -48 months}]
731 mlosch 1.4 {\includegraphics*[width=0.44\linewidth]{\fpath/run_4yr_ADJtheta_arc_lev1_tim292_cmax5.0E+01.eps}}
732 dimitri 1.1 }
733     \caption{Same as Fig. XXX but for sea surface temperature
734     \label{fig:4yradjthetalev1}}
735     \end{figure}
736    
737    
738    
739     \section{Discussion and conclusion}
740     \label{sec:concl}
741    
742     The story of the paper could be:
743     B-grid ice model + C-grid ocean model does not work properly for us,
744     therefore C-grid ice model with advantages:
745     \begin{enumerate}
746     \item stress coupling simpler (no interpolation required)
747     \item different boundary conditions
748     \item advection schemes carry over trivially from main code
749     \end{enumerate}
750     LSR/EVP solutions are similar with appropriate bcs, evp parameters as
751     a function of forcing time scale (when does VP solution break
752     down). Same for LSR solver, provided that it works (o:
753     Which scheme is more efficient for the resolution/time stepping
754     parameters that we use here. What about other resolutions?
755    
756     \paragraph{Acknowledgements}
757     We thank Jinlun Zhang for providing the original B-grid code and many
758     helpful discussions.
759    
760     \bibliography{bib/journal_abrvs,bib/seaice,bib/genocean,bib/maths,bib/mitgcmuv,bib/fram}
761    
762     \end{document}
763    
764     %%% Local Variables:
765     %%% mode: latex
766     %%% TeX-master: t
767     %%% End:
768 mlosch 1.4
769    
770     A Dynamic-Thermodynamic Sea ice Model for Ocean Climate
771     Estimation on an Arakawa C-Grid
772    
773     Introduction
774    
775     Ice Model:
776     Dynamics formulation.
777     B-C, LSR, EVP, no-slip, slip
778     parallellization
779     Thermodynamics formulation.
780     0-layer Hibler salinity + snow
781     3-layer Winton
782    
783     Idealized tests
784     Funnel Experiments
785     Downstream Island tests
786     B-grid LSR no-slip
787     C-grid LSR no-slip
788     C-grid LSR slip
789     C-grid EVP no-slip
790     C-grid EVP slip
791    
792     Arctic Setup
793     Configuration
794     OBCS from cube
795     forcing
796     1/2 and full resolution
797     with a few JFM figs from C-grid LSR no slip
798     ice transport through Canadian Archipelago
799     thickness distribution
800     ice velocity and transport
801    
802     Arctic forward sensitivity experiments
803     B-grid LSR no-slip
804     C-grid LSR no-slip
805     C-grid LSR slip
806     C-grid EVP no-slip
807     C-grid EVP slip
808     C-grid LSR no-slip + Winton
809     speed-performance-accuracy (small)
810     ice transport through Canadian Archipelago differences
811     thickness distribution differences
812     ice velocity and transport differences
813    
814     Adjoint sensitivity experiment on 1/2-res setup
815     Sensitivity of sea ice volume flow through Fram Strait
816     *** Sensitivity of sea ice volume flow through Canadian Archipelago
817    
818     Summary and conluding remarks

  ViewVC Help
Powered by ViewVC 1.1.22