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

Contents of /MITgcm_contrib/articles/ceaice/ceaice.tex

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


Revision 1.9 - (show annotations) (download) (as text)
Mon Jan 21 08:06:00 2008 UTC (17 years, 6 months ago) by mlosch
Branch: MAIN
Changes since 1.8: +14 -1 lines
File MIME type: application/x-tex
fix a label, add a paragraph (really a comment), add $Header: $,
$Name: $ for better tracking

1 % $Header: $
2 % $Name: $
3 \documentclass[12pt]{article}
4
5 \usepackage[]{graphicx}
6 \usepackage{subfigure}
7
8 \usepackage[round,comma]{natbib}
9 \bibliographystyle{bib/agu04}
10
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 \newcommand{\fpath}{figs}
41
42 % commenting scheme
43 \newcommand{\ml}[1]{\textsf{\slshape #1}}
44
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 Some blabla
56 \end{abstract}
57
58 \section{Introduction}
59 \label{sec:intro}
60
61 more blabla
62
63 \section{Model}
64 \label{sec:model}
65
66 Traditionally, probably for historical reasons and the ease of
67 treating the Coriolis term, most standard sea-ice models are
68 discretized on Arakawa-B-grids \citep[e.g.,][]{hibler79, harder99,
69 kreyscher00, zhang98, hunke97}. From the perspective of coupling a
70 sea ice-model to a C-grid ocean model, the exchange of fluxes of heat
71 and fresh-water pose no difficulty for a B-grid sea-ice model
72 \citep[e.g.,][]{timmermann02a}. However, surface stress is defined at
73 velocities points and thus needs to be interpolated between a B-grid
74 sea-ice model and a C-grid ocean model. While the smoothing implicitly
75 associated with this interpolation may mask grid scale noise, it may
76 in two-way coupling lead to a computational mode as will be shown. By
77 choosing a C-grid for the sea-ice model, we circumvene this difficulty
78 altogether and render the stress coupling as consistent as the
79 buoyancy coupling.
80
81 A further advantage of the C-grid formulation is apparent in narrow
82 straits. In the limit of only one grid cell between coasts there is no
83 flux allowed for a B-grid (with no-slip lateral boundary counditions),
84 whereas the C-grid formulation allows a flux of sea-ice through this
85 passage for all types of lateral boundary conditions. We (will)
86 demonstrate this effect in the Candian archipelago.
87
88 \subsection{Dynamics}
89 \label{sec:dynamics}
90
91 The momentum equations of the sea-ice model are standard with
92 \begin{equation}
93 \label{eq:momseaice}
94 m \frac{D\vek{u}}{Dt} = -mf\vek{k}\times\vek{u} + \vtau_{air} +
95 \vtau_{ocean} - m \nabla{\phi(0)} + \vek{F},
96 \end{equation}
97 where $\vek{u} = u\vek{i}+v\vek{j}$ is the ice velocity vectory, $m$
98 the ice mass per unit area, $f$ the Coriolis parameter, $g$ is the
99 gravity accelation, $\nabla\phi$ is the gradient (tilt) of the sea
100 surface height potential beneath the ice. $\phi$ is the sum of
101 atmpheric pressure $p_{a}$ and loading due to ice and snow
102 $(m_{i}+m_{s})g$. $\vtau_{air}$ and $\vtau_{ocean}$ are the wind and
103 ice-ocean stresses, respectively. $\vek{F}$ is the interaction force
104 and $\vek{i}$, $\vek{j}$, and $\vek{k}$ are the unit vectors in the
105 $x$, $y$, and $z$ directions. Advection of sea-ice momentum is
106 neglected. The wind and ice-ocean stress terms are given by
107 \begin{align*}
108 \vtau_{air} =& \rho_{air} |\vek{U}_{air}|R_{air}(\vek{U}_{air}) \\
109 \vtau_{ocean} =& \rho_{ocean} |\vek{U}_{ocean}-\vek{u}|
110 R_{ocean}(\vek{U}_{ocean}-\vek{u}), \\
111 \end{align*}
112 where $\vek{U}_{air/ocean}$ are the surface winds of the atmosphere
113 and surface currents of the ocean, respectively. $C_{air/ocean}$ are
114 air and ocean drag coefficients, $\rho_{air/ocean}$ reference
115 densities, and $R_{air/ocean}$ rotation matrices that act on the
116 wind/current vectors. $\vek{F} = \nabla\cdot\sigma$ is the divergence
117 of the interal stress tensor $\sigma_{ij}$.
118
119 For an isotropic system this stress tensor can be related to the ice
120 strain rate and strength by a nonlinear viscous-plastic (VP)
121 constitutive law \citep{hibler79, zhang98}:
122 \begin{equation}
123 \label{eq:vpequation}
124 \sigma_{ij}=2\eta(\dot{\epsilon}_{ij},P)\dot{\epsilon}_{ij}
125 + \left[\zeta(\dot{\epsilon}_{ij},P) -
126 \eta(\dot{\epsilon}_{ij},P)\right]\dot{\epsilon}_{kk}\delta_{ij}
127 - \frac{P}{2}\delta_{ij}.
128 \end{equation}
129 The ice strain rate is given by
130 \begin{equation*}
131 \dot{\epsilon}_{ij} = \frac{1}{2}\left(
132 \frac{\partial{u_{i}}}{\partial{x_{j}}} +
133 \frac{\partial{u_{j}}}{\partial{x_{i}}}\right).
134 \end{equation*}
135 The maximum ice pressure $P_{\max}$, a measure of ice strength, depends on
136 both thickness $h$ and compactness (concentration) $c$:
137 \begin{equation}
138 P_{\max} = P^{*}c\,h\,e^{[C^{*}\cdot(1-c)]},
139 \label{eq:icestrength}
140 \end{equation}
141 with the constants $P^{*}$ and $C^{*}$. The nonlinear bulk and shear
142 viscosities $\eta$ and $\zeta$ are functions of ice strain rate
143 invariants and ice strength such that the principal components of the
144 stress lie on an elliptical yield curve with the ratio of major to
145 minor axis $e$ equal to $2$; they are given by:
146 \begin{align*}
147 \zeta =& \min\left(\frac{P_{\max}}{2\max(\Delta,\Delta_{\min})},
148 \zeta_{\max}\right) \\
149 \eta =& \frac{\zeta}{e^2} \\
150 \intertext{with the abbreviation}
151 \Delta = & \left[
152 \left(\dot{\epsilon}_{11}^2+\dot{\epsilon}_{22}^2\right)
153 (1+e^{-2}) + 4e^{-2}\dot{\epsilon}_{12}^2 +
154 2\dot{\epsilon}_{11}\dot{\epsilon}_{22} (1-e^{-2})
155 \right]^{-\frac{1}{2}}
156 \end{align*}
157 The bulk viscosities are bounded above by imposing both a minimum
158 $\Delta_{\min}=10^{-11}\text{\,s}^{-1}$ (for numerical reasons) and a
159 maximum $\zeta_{\max} = P_{\max}/\Delta^*$, where
160 $\Delta^*=(5\times10^{12}/2\times10^4)\text{\,s}^{-1}$. For stress
161 tensor compuation the replacement pressure $P = 2\,\Delta\zeta$
162 \citep{hibler95} is used so that the stress state always lies on the
163 elliptic yield curve by definition.
164
165 In the so-called truncated ellipse method the shear viscosity $\eta$
166 is capped to suppress any tensile stress \citep{hibler97, geiger98}:
167 \begin{equation}
168 \label{eq:etatem}
169 \eta = \min(\frac{\zeta}{e^2}
170 \frac{\frac{P}{2}-\zeta(\dot{\epsilon}_{11}+\dot{\epsilon}_{22})}
171 {\sqrt{(\dot{\epsilon}_{11}+\dot{\epsilon}_{22})^2
172 +4\dot{\epsilon}_{12}^2}}
173 \end{equation}
174
175 In the current implementation, the VP-model is integrated with the
176 semi-implicit line successive over relaxation (LSOR)-solver of
177 \citet{zhang98}, which allows for long time steps that, in our case,
178 is limited by the explicit treatment of the Coriolis term. The
179 explicit treatment of the Coriolis term does not represent a severe
180 limitation because it restricts the time step to approximately the
181 same length as in the ocean model where the Coriolis term is also
182 treated explicitly.
183
184 \citet{hunke97}'s introduced an elastic contribution to the strain
185 rate elatic-viscous-plastic in order to regularize
186 Eq.\refeq{vpequation} in such a way that the resulting
187 elatic-viscous-plastic (EVP) and VP models are identical at steady
188 state,
189 \begin{equation}
190 \label{eq:evpequation}
191 \frac{1}{E}\frac{\partial\sigma_{ij}}{\partial{t}} +
192 \frac{1}{2\eta}\sigma_{ij}
193 + \frac{\eta - \zeta}{4\zeta\eta}\sigma_{kk}\delta_{ij}
194 + \frac{P}{4\zeta}\delta_{ij}
195 = \dot{\epsilon}_{ij}.
196 \end{equation}
197 %In the EVP model, equations for the components of the stress tensor
198 %$\sigma_{ij}$ are solved explicitly. Both model formulations will be
199 %used and compared the present sea-ice model study.
200 The EVP-model uses an explicit time stepping scheme with a short
201 timestep. According to the recommendation of \citet{hunke97}, the
202 EVP-model is stepped forward in time 120 times within the physical
203 ocean model time step (although this parameter is under debate), to
204 allow for elastic waves to disappear. Because the scheme does not
205 require a matrix inversion it is fast in spite of the small timestep
206 \citep{hunke97}. For completeness, we repeat the equations for the
207 components of the stress tensor $\sigma_{1} =
208 \sigma_{11}+\sigma_{22}$, $\sigma_{2}= \sigma_{11}-\sigma_{22}$, and
209 $\sigma_{12}$. Introducing the divergence $D_D =
210 \dot{\epsilon}_{11}+\dot{\epsilon}_{22}$, and the horizontal tension
211 and shearing strain rates, $D_T =
212 \dot{\epsilon}_{11}-\dot{\epsilon}_{22}$ and $D_S =
213 2\dot{\epsilon}_{12}$, respectively and using the above abbreviations,
214 the equations can be written as:
215 \begin{align}
216 \label{eq:evpstresstensor1}
217 \frac{\partial\sigma_{1}}{\partial{t}} + \frac{\sigma_{1}}{2T} +
218 \frac{P}{2T} &= \frac{P}{2T\Delta} D_D \\
219 \label{eq:evpstresstensor2}
220 \frac{\partial\sigma_{2}}{\partial{t}} + \frac{\sigma_{2} e^{2}}{2T}
221 &= \frac{P}{2T\Delta} D_T \\
222 \label{eq:evpstresstensor12}
223 \frac{\partial\sigma_{12}}{\partial{t}} + \frac{\sigma_{12} e^{2}}{2T}
224 &= \frac{P}{4T\Delta} D_S
225 \end{align}
226 Here, the elastic parameter $E$ is redefined in terms of a damping timescale
227 $T$ for elastic waves \[E=\frac{\zeta}{T}.\]
228 $T=E_{0}\Delta{t}$ with the tunable parameter $E_0<1$ and
229 the external (long) timestep $\Delta{t}$. \citet{hunke97} recommend
230 $E_{0} = \frac{1}{3}$.
231
232 For details of the spatial discretization, the reader is referred to
233 \citet{zhang98, zhang03}. Our discretization differs only (but
234 importantly) in the underlying grid, namely the Arakawa C-grid, but is
235 otherwise straightforward. The EVP model in particular is discretized
236 naturally on the C-grid with $\sigma_{1}$ and $\sigma_{2}$ on the
237 center points and $\sigma_{12}$ on the corner (or vorticity) points of
238 the grid. With this choice all derivatives are discretized as central
239 differences and averaging is only involved in computing $\Delta$ and
240 $P$ at vorticity points.
241
242 For a general curvilinear grid, one needs in principle to take metric
243 terms into account that arise in the transformation a curvilinear grid
244 on the sphere. However, for now we can neglect these metric terms
245 because they are very small on the cubed sphere grids used in this
246 paper; in particular, only near the edges of the cubed sphere grid, we
247 expect them to be non-zero, but these edges are at approximately
248 35\degS\ or 35\degN\ which are never covered by sea-ice in our
249 simulations. Everywhere else the coordinate system is locally nearly
250 cartesian. However, for last-glacial-maximum or snowball-earth-like
251 simulations the question of metric terms needs to be reconsidered.
252 Either, one includes these terms as in \citet{zhang03}, or one finds a
253 vector-invariant formulation fo the sea-ice internal stress term that
254 does not require any metric terms, as it is done in the ocean dynamics
255 of the MITgcm \citep{adcroft04:_cubed_sphere}.
256
257 Moving sea ice exerts a stress on the ocean which is the opposite of
258 the stress $\vtau_{ocean}$ in Eq.\refeq{momseaice}. This stess is
259 applied directly to the surface layer of the ocean model. An
260 alternative ocean stress formulation is given by \citet{hibler87}.
261 Rather than applying $\vtau_{ocean}$ directly, the stress is derived
262 from integrating over the ice thickness to the bottom of the oceanic
263 surface layer. In the resulting equation for the \emph{combined}
264 ocean-ice momentum, the interfacial stress cancels and the total
265 stress appears as the sum of windstress and divergence of internal ice
266 stresses: $\delta(z) (\vtau_{air} + \vek{F})/\rho_0$, \citep[see also
267 Eq.\,2 of][]{hibler87}. The disadvantage of this formulation is that
268 now the velocity in the surface layer of the ocean that is used to
269 advect tracers, is really an average over the ocean surface
270 velocity and the ice velocity leading to an inconsistency as the ice
271 temperature and salinity are different from the oceanic variables.
272
273 Sea ice distributions are characterized by sharp gradients and edges.
274 For this reason, we employ a positive 3rd-order advection scheme
275 \citep{hundsdorfer94} for the thermodynamic variables discussed in the
276 next section.
277
278 \subparagraph{boundary conditions: no-slip, free-slip, half-slip}
279
280 \begin{itemize}
281 \item transition from existing B-Grid to C-Grid
282 \item boundary conditions: no-slip, free-slip, half-slip
283 \item fancy (multi dimensional) advection schemes
284 \item VP vs.\ EVP \citep{hunke97}
285 \item ocean stress formulation \citep{hibler87}
286 \end{itemize}
287
288 \subsection{Thermodynamics}
289 \label{sec:thermodynamics}
290
291 In the original formulation the sea ice model \citep{menemenlis05}
292 uses simple thermodynamics following the appendix of
293 \citet{semtner76}. This formulation does not allow storage of heat
294 (heat capacity of ice is zero, and this type of model is often refered
295 to as a ``zero-layer'' model). Upward heat flux is parameterized
296 assuming a linear temperature profile and together with a constant ice
297 conductivity. It is expressed as $(K/h)(T_{w}-T_{0})$, where $K$ is
298 the ice conductivity, $h$ the ice thickness, and $T_{w}-T_{0}$ the
299 difference between water and ice surface temperatures. The surface
300 heat budget is computed in a similar way to that of
301 \citet{parkinson79} and \citet{manabe79}.
302
303 There is considerable doubt about the reliability of such a simple
304 thermodynamic model---\citet{semtner84} found significant errors in
305 phase (one month lead) and amplitude ($\approx$50\%\,overestimate) in
306 such models---, so that today many sea ice models employ more complex
307 thermodynamics. A popular thermodynamics model of \citet{winton00} is
308 based on the 3-layer model of \citet{semtner76} and treats brine
309 content by means of enthalphy conservation. This model requires in
310 addition to ice-thickness and compactness (fractional area) additional
311 state variables to be advected by ice velocities, namely enthalphy of
312 the two ice layers and the thickness of the overlying snow layer.
313
314 \section{Funnel Experiments}
315 \label{sec:funnel}
316
317 For a first/detailed comparison between the different variants of the
318 MIT sea ice model an idealized geometry of a periodic channel,
319 1000\,km long and 500\,m wide on a non-rotating plane, with converging
320 walls forming a symmetric funnel and a narrow strait of 40\,km width
321 is used. The horizontal resolution is 5\,km throughout the domain
322 making the narrow strait 8 grid points wide. The ice model is
323 initialized with a complete ice cover of 50\,cm uniform thickness. The
324 ice model is driven by a constant along channel eastward ocean current
325 of 25\,cm/s that does not see the walls in the domain. All other
326 ice-ocean-atmosphere interactions are turned off, in particular there
327 is no feedback of ice dynamics on the ocean current. All thermodynamic
328 processes are turned off so that ice thickness variations are only
329 caused by convergent or divergent ice flow. Ice volume (effective
330 thickness) and concentration are advected with a third-order scheme
331 with a flux limiter \citep{hundsdorfer94} to avoid undershoots. This
332 scheme is unconditionally stable and does not require additional
333 diffusion. The time step used here is 1\,h.
334
335 \reffig{funnelf0} compares the dynamic fields ice concentration $c$,
336 effective thickness $h_{eff} = h\cdot{c}$, and velocities $(u,v)$ for
337 five different cases at steady state (after 10\,years of integration):
338 \begin{description}
339 \item[B-LSRns:] LSR solver with no-slip boundary conditions on a B-grid,
340 \item[C-LSRns:] LSR solver with no-slip boundary conditions on a C-grid,
341 \item[C-LSRfs:] LSR solver with free-slip boundary conditions on a C-grid,
342 \item[C-EVPns:] EVP solver with no-slip boundary conditions on a C-grid,
343 \item[C-EVPfs:] EVP solver with free-slip boundary conditions on a C-grid,
344 \end{description}
345 \ml{[We have not implemented the EVP solver on a B-grid.]}
346 \begin{figure*}[htbp]
347 \includegraphics[width=\widefigwidth]{\fpath/all_086280}
348 \caption{Ice concentration, effective thickness [m], and ice
349 velocities [m/s]
350 for 5 different numerical solutions.}
351 \label{fig:funnelf0}
352 \end{figure*}
353 At a first glance, the solutions look similar. This is encouraging as
354 the details of discretization and numerics should not affect the
355 solutions to first order. In all cases the ice-ocean stress pushes the
356 ice cover eastwards, where it converges in the funnel. In the narrow
357 channel the ice moves quickly (nearly free drift) and leaves the
358 channel as narrow band.
359
360 A close look reveals interesting differences between the B- and C-grid
361 results. The zonal velocity in the narrow channel is nearly the free
362 drift velocity ( = ocean velocity) of 25\,cm/s for the C-grid
363 solutions, regardless of the boundary conditions, while it is just
364 above 20\,cm/s for the B-grid solution. The ice accelerates to
365 25\,cm/s after it exits the channel. Concentrating on the solutions
366 B-LSRns and C-LSRns, the ice volume (effective thickness) along the
367 boundaries in the narrow channel is larger in the B-grid case although
368 the ice concentration is reduces in the C-grid case. The combined
369 effect leads to a larger actual ice thickness at smaller
370 concentrations in the C-grid case. However, since the effective
371 thickness determines the ice strength $P$ in Eq\refeq{icestrength},
372 the ice strength and thus the bulk and shear viscosities are larger in
373 the B-grid case leading to more horizontal friction. This circumstance
374 might explain why the no-slip boundary conditions in the B-grid case
375 appear to be more effective in reducing the flow within the narrow
376 channel, than in the C-grid case. Further, the viscosities are also
377 sensitive to details of the velocity gradients. Via $\Delta$, these
378 gradients enter the viscosities in the denominator so that large
379 gradients tend to reduce the viscosities. This again favors more flow
380 along the boundaries in the C-grid case: larger velocities
381 (\reffig{funnelf0}) on grid points that are closer to the boundary by
382 a factor $\frac{1}{2}$ than in the B-grid case because of the stagger
383 nature of the C-grid lead numerically to larger tangential gradients
384 across the boundary; these in turn make the viscosities smaller for
385 less tangential friction and allow more tangential flow along the
386 boundaries.
387
388 The above argument can also be invoked to explain the small
389 differences between the free-slip and no-slip solutions on the C-grid.
390 Because of the non-linearities in the ice viscosities, in particular
391 along the boundaries, the no-slip boundary conditions have only a small
392 impact on the solution.
393
394 The difference between LSR and EVP solutions is largest in the
395 effective thickness and meridional velocity fields. The EVP velocity
396 fields appears to be a little noisy. This noise has been address by
397 \citet{hunke01}. It can be dealt with by reducing EVP's internal time
398 step (increasing the number of iterations along with the computational
399 cost) or by regularizing the bulk and shear viscosities. We revisit
400 the latter option by reproducing some of the results of
401 \citet{hunke01}, namely the experiment described in her section~4, for
402 our C-grid no-slip cases: in a square domain with a few islands the
403 ice model is initialized with constant ice thickness and linearly
404 increasing ice concentration to the east. The model dynamics are
405 forced with a constant anticyclonic ocean gyre and by variable
406 atmospheric wind whose mean direction is diagnonal to the north-east
407 corner of the domain; ice volume and concentration are held constant
408 (no thermodynamics and no advection by ice velocity).
409 \reffig{hunke01} shows the ice velocity field, its divergence, and the
410 bulk viscosity $\zeta$ for the cases C-LRSns and C-EVPns, and for a
411 C-EVPns case, where \citet{hunke01}'s regularization has been
412 implemented; compare to Fig.\,4 in \citet{hunke01}. The regularization
413 contraint limits ice strength and viscosities as a function of damping
414 time scale, resolution and EVP-time step, effectively allowing the
415 elastic waves to damp out more quickly \citep{hunke01}.
416 \begin{figure*}[htbp]
417 \includegraphics[width=\widefigwidth]{\fpath/hun12days}
418 \caption{Ice flow, divergence and bulk viscosities of three
419 experiments with \citet{hunke01}'s test case: C-LSRns (top),
420 C-EVPns (middle), and C-EVPns with damping described in
421 \citet{hunke01} (bottom).}
422 \label{fig:hunke01}
423 \end{figure*}
424
425 In the far right (``east'') side of the domain the ice concentration
426 is close to one and the ice should be nearly rigid. The applied wind
427 tends to push ice toward the upper right corner. Because the highly
428 compact ice is confined by the boundary, it resists any further
429 compression and exhibits little motion in the rigid region on the
430 right hand side. The C-LSRns solution (top row) allows high
431 viscosities in the rigid region suppressing nearly all flow.
432 \citet{hunke01}'s regularization for the C-EVPns solution (bottom row)
433 clearly suppresses the noise present in $\nabla\cdot\vek{u}$ and
434 $\log_{10}\zeta$ in the
435 unregularized case (middle row), at the cost of reduced viscosities.
436 These reduced viscosities lead to small but finite ice velocities
437 which in turn can have a strong effect on solutions in the limit of
438 nearly rigid regimes (arching and blocking, not shown).
439
440 \ml{[Say something about performance? This is tricky, as the
441 perfomance depends strongly on the configuration. A run with slowly
442 changing forcing is favorable for LSR, because then only very few
443 iterations are required for convergences while EVP uses its fixed
444 number of internal timesteps. If the forcing in changing fast, LSR
445 needs far more iterations while EVP still uses the fixed number of
446 internal timesteps. I have produces runs where for slow forcing LSR
447 is much faster than EVP and for fast forcing, LSR is much slower
448 than EVP. EVP is certainly more efficient in terms of vectorization
449 and MFLOPS on our SX8, but is that a criterion?]}
450
451 \subsection{C-grid}
452 \begin{itemize}
453 \item no-slip vs. free-slip for both lsr and evp;
454 "diagnostics" to look at and use for comparison
455 \begin{itemize}
456 \item ice transport through Fram Strait/Denmark Strait/Davis
457 Strait/Bering strait (these are general)
458 \item ice transport through narrow passages, e.g.\ Nares-Strait
459 \end{itemize}
460 \item compare different advection schemes (if lsr turns out to be more
461 effective, then with lsr otherwise I prefer evp), eg.
462 \begin{itemize}
463 \item default 2nd-order with diff1=0.002
464 \item 1st-order upwind with diff1=0.
465 \item DST3FL (SEAICEadvScheme=33 with diff1=0., should work, works for me)
466 \item 2nd-order wit flux limiter (SEAICEadvScheme=77 with diff1=0.)
467 \end{itemize}
468 That should be enough. Here, total ice mass and location of ice edge
469 is interesting. However, this comparison can be done in an idealized
470 domain, may not require full Arctic Domain?
471 \item
472 Do a little study on the parameters of LSR and EVP
473 \begin{enumerate}
474 \item convergence of LSR, how many iterations do you need to get a
475 true elliptic yield curve
476 \item vary deltaTevp and the relaxation parameter for EVP and see when
477 the EVP solution breaks down (relative to the forcing time scale).
478 For this, it is essential that the evp solver gives use "stripeless"
479 solutions, that is your dtevp = 1sec solutions/or 10sec solutions
480 with SEAICE\_evpDampC = 615.
481 \end{enumerate}
482 \end{itemize}
483
484 \section{Forward sensitivity experiments}
485 \label{sec:forward}
486
487 A second series of forward sensitivity experiments have been carried out on an
488 Arctic Ocean domain with open boundaries. Once again the objective is to
489 compare the old B-grid LSR dynamic solver with the new C-grid LSR and EVP
490 solvers. One additional experiment is carried out to illustrate the
491 differences between the two main options for sea ice thermodynamics in the MITgcm.
492
493 \subsection{Arctic Domain with Open Boundaries}
494 \label{sec:arctic}
495
496 The Arctic domain of integration is illustrated in Fig.~\ref{???}. It
497 is carved out from, and obtains open boundary conditions from, the
498 global cubed-sphere configuration of the Estimating the Circulation
499 and Climate of the Ocean, Phase II (ECCO2) project
500 \citet{menemenlis05}. The domain size is 420 by 384 grid boxes
501 horizontally with mean horizontal grid spacing of 18 km.
502
503 There are 50 vertical levels ranging in thickness from 10 m near the surface
504 to approximately 450 m at a maximum model depth of 6150 m. Bathymetry is from
505 the National Geophysical Data Center (NGDC) 2-minute gridded global relief
506 data (ETOPO2) and the model employs the partial-cell formulation of
507 \citet{adcroft97:_shaved_cells}, which permits accurate representation of the bathymetry. The
508 model is integrated in a volume-conserving configuration using a finite volume
509 discretization with C-grid staggering of the prognostic variables. In the
510 ocean, the non-linear equation of state of \citet{jackett95}. The ocean model is
511 coupled to a sea-ice model described hereinabove.
512
513 This particular ECCO2 simulation is initialized from rest using the
514 January temperature and salinity distribution from the World Ocean
515 Atlas 2001 (WOA01) [Conkright et al., 2002] and it is integrated for
516 32 years prior to the 1996--2001 period discussed in the study. Surface
517 boundary conditions are from the National Centers for Environmental
518 Prediction and the National Center for Atmospheric Research
519 (NCEP/NCAR) atmospheric reanalysis [Kistler et al., 2001]. Six-hourly
520 surface winds, temperature, humidity, downward short- and long-wave
521 radiations, and precipitation are converted to heat, freshwater, and
522 wind stress fluxes using the \citet{large81, large82} bulk formulae.
523 Shortwave radiation decays exponentially as per Paulson and Simpson
524 [1977]. Additionally the time-mean river run-off from Large and Nurser
525 [2001] is applied and there is a relaxation to the monthly-mean
526 climatological sea surface salinity values from WOA01 with a
527 relaxation time scale of 3 months. Vertical mixing follows
528 \citet{large94} with background vertical diffusivity of
529 $1.5\times10^{-5}\text{\,m$^{2}$\,s$^{-1}$}$ and viscosity of
530 $10^{-3}\text{\,m$^{2}$\,s$^{-1}$}$. A third order, direct-space-time
531 advection scheme with flux limiter is employed \citep{hundsdorfer94}
532 and there is no explicit horizontal diffusivity. Horizontal viscosity
533 follows \citet{lei96} but
534 modified to sense the divergent flow as per Fox-Kemper and Menemenlis
535 [in press]. Shortwave radiation decays exponentially as per Paulson
536 and Simpson [1977]. Additionally, the time-mean runoff of Large and
537 Nurser [2001] is applied near the coastline and, where there is open
538 water, there is a relaxation to monthly-mean WOA01 sea surface
539 salinity with a time constant of 45 days.
540
541 Open water, dry
542 ice, wet ice, dry snow, and wet snow albedo are, respectively, 0.15, 0.85,
543 0.76, 0.94, and 0.8.
544
545 \begin{itemize}
546 \item Configuration
547 \item OBCS from cube
548 \item forcing
549 \item 1/2 and full resolution
550 \item with a few JFM figs from C-grid LSR no slip
551 ice transport through Canadian Archipelago
552 thickness distribution
553 ice velocity and transport
554 \end{itemize}
555
556 \begin{itemize}
557 \item Arctic configuration
558 \item ice transport through straits and near boundaries
559 \item focus on narrow straits in the Canadian Archipelago
560 \end{itemize}
561
562 \begin{itemize}
563 \item B-grid LSR no-slip
564 \item C-grid LSR no-slip
565 \item C-grid LSR slip
566 \item C-grid EVP no-slip
567 \item C-grid EVP slip
568 \item C-grid LSR + TEM (truncated ellipse method, no tensile stress, new flag)
569 \item C-grid LSR no-slip + Winton
570 \item speed-performance-accuracy (small)
571 ice transport through Canadian Archipelago differences
572 thickness distribution differences
573 ice velocity and transport differences
574 \end{itemize}
575
576 We anticipate small differences between the different models due to:
577 \begin{itemize}
578 \item advection schemes: along the ice-edge and regions with large
579 gradients
580 \item C-grid: less transport through narrow straits for no slip
581 conditons, more for free slip
582 \item VP vs.\ EVP: speed performance, accuracy?
583 \item ocean stress: different water mass properties beneath the ice
584 \end{itemize}
585
586 \section{Adjoint sensiivities of the MITsim}
587
588 \subsection{The adjoint of MITsim}
589
590 The ability to generate tangent linear and adjoint model components
591 of the MITsim has been a main design task.
592 For the ocean the adjoint capability has proven to be an
593 invaluable tool for sensitivity analysis as well as state estimation.
594 In short, the adjoint enables very efficient computation of the gradient
595 of scalar-valued model diagnostics (called cost function or objective function)
596 with respect to many model "variables".
597 These variables can be two- or three-dimensional fields of initial
598 conditions, model parameters such as mixing coefficients, or
599 time-varying surface or lateral (open) boundary conditions.
600 When combined, these variables span a potentially high-dimensional
601 (e.g. O(10$^8$)) so-called control space. Performing parameter perturbations
602 to assess model sensitivities quickly becomes prohibitive at these scales.
603 Alternatively, (time-varying) sensitivities of the objective function
604 to any element of the control space can be computed very efficiently in
605 one single adjoint
606 model integration, provided an efficient adjoint model is available.
607
608 [REFERENCES]
609
610
611 The adjoint operator (ADM) is the transpose of the tangent linear operator (TLM)
612 of the full (in general nonlinear) forward model, i.e. the MITsim.
613 The TLM maps perturbations of elements of the control space
614 (e.g. initial ice thickness distribution)
615 via the model Jacobian
616 to a perturbation in the objective function
617 (e.g. sea-ice export at the end of the integration interval).
618 \textit{Tangent} linearity ensures that the derivatives are evaluated
619 with respect to the underlying model trajectory at each point in time.
620 This is crucial for nonlinear trajectories and the presence of different
621 regimes (e.g. effect of the seaice growth term at or away from the
622 freezing point of the ocean surface).
623 Ensuring tangent linearity can be easily achieved by integrating
624 the full model in sync with the TLM to provide the underlying model state.
625 Ensuring \textit{tangent} adjoints is equally crucial, but much more
626 difficult to achieve because of the reverse nature of the integration:
627 the adjoint accumulates sensitivities backward in time,
628 starting from a unit perturbation of the objective function.
629 The adjoint model requires the model state in reverse order.
630 This presents one of the major complications in deriving an
631 exact, i.e. \textit{tangent} adjoint model.
632
633 Following closely the development and maintenance of TLM and ADM
634 components of the MITgcm we have relied heavily on the
635 autmomatic differentiation (AD) tool
636 "Transformation of Algorithms in Fortran" (TAF)
637 developed by Fastopt (Giering and Kaminski, 1998)
638 to derive TLM and ADM code of the MITsim.
639 Briefly, the nonlinear parent model is fed to the AD tool which produces
640 derivative code for the specified control space and objective function.
641 Following this approach has (apart from its evident success)
642 several advantages:
643 (1) the adjoint model is the exact adjoint operator of the parent model,
644 (2) the adjoint model can be kept up to date with respect to ongoing
645 development of the parent model, and adjustments to the parent model
646 to extend the automatically generated adjoint are incremental changes
647 only, rather than extensive re-developments,
648 (3) the parallel structure of the parent model is preserved
649 by the adjoint model, ensuring efficient use in high performance
650 computing environments.
651
652 Some initial code adjustments are required to support dependency analysis
653 of the flow reversal and certain language limitations which may lead
654 to irreducible flow graphs (e.g. GOTO statements).
655 The problem of providing the required model state in reverse order
656 at the time of evaluating nonlinear or conditional
657 derivatives is solved via balancing
658 storing vs. recomputation of the model state in a multi-level
659 checkpointing loop.
660 Again, an initial code adjustment is required to support TAFs
661 checkpointing capability.
662 The code adjustments are sufficiently simple so as not to cause
663 major limitations to the full nonlinear parent model.
664 Once in place, an adjoint model of a new model configuration
665 may be derived in about 10 minutes.
666
667 [HIGHLIGHT COUPLED NATURE OF THE ADJOINT!]
668
669 \subsection{Special considerations}
670
671 * growth term(?)
672
673 * small active denominators
674
675 * dynamic solver (implicit function theorem)
676
677 * approximate adjoints
678
679
680 \subsection{An example: sensitivities of sea-ice export through Fram Strait}
681
682 We demonstrate the power of the adjoint method
683 in the context of investigating sea-ice export sensitivities through Fram Strait
684 (for details of this study see Heimbach et al., 2007).
685 %\citep[for details of this study see][]{heimbach07}. %Heimbach et al., 2007).
686 The domain chosen is a coarsened version of the Arctic face of the
687 high-resolution cubed-sphere configuration of the ECCO2 project
688 \citep[see][]{menemenlis05}. It covers the entire Arctic,
689 extends into the North Pacific such as to cover the entire
690 ice-covered regions, and comprises parts of the North Atlantic
691 down to XXN to enable analysis of remote influences of the
692 North Atlantic current to sea-ice variability and export.
693 The horizontal resolution varies between XX and YY km
694 with 50 unevenly spaced vertical levels.
695 The adjoint models run efficiently on 80 processors
696 (benchmarks have been performed both on an SGI Altix as well as an
697 IBM SP5 at NASA/ARC).
698
699 Following a 1-year spinup, the model has been integrated for four
700 years between 1992 and 1995. It is forced using realistic 6-hourly
701 NCEP/NCAR atmospheric state variables. Over the open ocean these are
702 converted into air-sea fluxes via the bulk formulae of
703 \citet{large04}. Derivation of air-sea fluxes in the presence of
704 sea-ice is handled by the ice model as described in \refsec{model}.
705 The objective function chosen is sea-ice export through Fram Strait
706 computed for December 1995. The adjoint model computes sensitivities
707 to sea-ice export back in time from 1995 to 1992 along this
708 trajectory. In principle all adjoint model variable (i.e., Lagrange
709 multipliers) of the coupled ocean/sea-ice model are available to
710 analyze the transient sensitivity behaviour of the ocean and sea-ice
711 state. Over the open ocean, the adjoint of the bulk formula scheme
712 computes sensitivities to the time-varying atmospheric state. Over
713 ice-covered parts, the sea-ice adjoint converts surface ocean
714 sensitivities to atmospheric sensitivities.
715
716 \reffig{4yradjheff}(a--d) depict sensitivities of sea-ice export
717 through Fram Strait in December 1995 to changes in sea-ice thickness
718 12, 24, 36, 48 months back in time. Corresponding sensitivities to
719 ocean surface temperature are depicted in
720 \reffig{4yradjthetalev1}(a--d). The main characteristics is
721 consistency with expected advection of sea-ice over the relevant time
722 scales considered. The general positive pattern means that an
723 increase in sea-ice thickness at location $(x,y)$ and time $t$ will
724 increase sea-ice export through Fram Strait at time $T_e$. Largest
725 distances from Fram Strait indicate fastest sea-ice advection over the
726 time span considered. The ice thickness sensitivities are in close
727 correspondence to ocean surface sentivitites, but of opposite sign.
728 An increase in temperature will incur ice melting, decrease in ice
729 thickness, and therefore decrease in sea-ice export at time $T_e$.
730
731 The picture is fundamentally different and much more complex
732 for sensitivities to ocean temperatures away from the surface.
733 \reffig{4yradjthetalev10??}(a--d) depicts ice export sensitivities to
734 temperatures at roughly 400 m depth.
735 Primary features are the effect of the heat transport of the North
736 Atlantic current which feeds into the West Spitsbergen current,
737 the circulation around Svalbard, and ...
738
739 \begin{figure}[t!]
740 \centerline{
741 \subfigure[{\footnotesize -12 months}]
742 {\includegraphics*[width=0.44\linewidth]{\fpath/run_4yr_ADJheff_arc_lev1_tim072_cmax2.0E+02.eps}}
743 %\includegraphics*[width=.3\textwidth]{H_c.bin_res_100_lev1.pdf}
744 %
745 \subfigure[{\footnotesize -24 months}]
746 {\includegraphics*[width=0.44\linewidth]{\fpath/run_4yr_ADJheff_arc_lev1_tim145_cmax2.0E+02.eps}}
747 }
748
749 \centerline{
750 \subfigure[{\footnotesize
751 -36 months}]
752 {\includegraphics*[width=0.44\linewidth]{\fpath/run_4yr_ADJheff_arc_lev1_tim218_cmax2.0E+02.eps}}
753 %
754 \subfigure[{\footnotesize
755 -48 months}]
756 {\includegraphics*[width=0.44\linewidth]{\fpath/run_4yr_ADJheff_arc_lev1_tim292_cmax2.0E+02.eps}}
757 }
758 \caption{Sensitivity of sea-ice export through Fram Strait in December 2005 to
759 sea-ice thickness at various prior times.
760 \label{fig:4yradjheff}}
761 \end{figure}
762
763
764 \begin{figure}[t!]
765 \centerline{
766 \subfigure[{\footnotesize -12 months}]
767 {\includegraphics*[width=0.44\linewidth]{\fpath/run_4yr_ADJtheta_arc_lev1_tim072_cmax5.0E+01.eps}}
768 %\includegraphics*[width=.3\textwidth]{H_c.bin_res_100_lev1.pdf}
769 %
770 \subfigure[{\footnotesize -24 months}]
771 {\includegraphics*[width=0.44\linewidth]{\fpath/run_4yr_ADJtheta_arc_lev1_tim145_cmax5.0E+01.eps}}
772 }
773
774 \centerline{
775 \subfigure[{\footnotesize
776 -36 months}]
777 {\includegraphics*[width=0.44\linewidth]{\fpath/run_4yr_ADJtheta_arc_lev1_tim218_cmax5.0E+01.eps}}
778 %
779 \subfigure[{\footnotesize
780 -48 months}]
781 {\includegraphics*[width=0.44\linewidth]{\fpath/run_4yr_ADJtheta_arc_lev1_tim292_cmax5.0E+01.eps}}
782 }
783 \caption{Same as \reffig{4yradjheff} but for sea surface temperature
784 \label{fig:4yradjthetalev1}}
785 \end{figure}
786
787
788
789 \section{Discussion and conclusion}
790 \label{sec:concl}
791
792 The story of the paper could be:
793 B-grid ice model + C-grid ocean model does not work properly for us,
794 therefore C-grid ice model with advantages:
795 \begin{enumerate}
796 \item stress coupling simpler (no interpolation required)
797 \item different boundary conditions
798 \item advection schemes carry over trivially from main code
799 \end{enumerate}
800 LSR/EVP solutions are similar with appropriate bcs, evp parameters as
801 a function of forcing time scale (when does VP solution break
802 down). Same for LSR solver, provided that it works (o:
803 Which scheme is more efficient for the resolution/time stepping
804 parameters that we use here. What about other resolutions?
805
806 \paragraph{Acknowledgements}
807 We thank Jinlun Zhang for providing the original B-grid code and many
808 helpful discussions. ML thanks Elizabeth Hunke for multiple explanations.
809
810 \bibliography{bib/journal_abrvs,bib/seaice,bib/genocean,bib/maths,bib/mitgcmuv,bib/fram}
811
812 \end{document}
813
814 %%% Local Variables:
815 %%% mode: latex
816 %%% TeX-master: t
817 %%% End:
818
819
820 A Dynamic-Thermodynamic Sea ice Model for Ocean Climate
821 Estimation on an Arakawa C-Grid
822
823 Introduction
824
825 Ice Model:
826 Dynamics formulation.
827 B-C, LSR, EVP, no-slip, slip
828 parallellization
829 Thermodynamics formulation.
830 0-layer Hibler salinity + snow
831 3-layer Winton
832
833 Idealized tests
834 Funnel Experiments
835 Downstream Island tests
836 B-grid LSR no-slip
837 C-grid LSR no-slip
838 C-grid LSR slip
839 C-grid EVP no-slip
840 C-grid EVP slip
841
842 Arctic Setup
843 Configuration
844 OBCS from cube
845 forcing
846 1/2 and full resolution
847 with a few JFM figs from C-grid LSR no slip
848 ice transport through Canadian Archipelago
849 thickness distribution
850 ice velocity and transport
851
852 Arctic forward sensitivity experiments
853 B-grid LSR no-slip
854 C-grid LSR no-slip
855 C-grid LSR slip
856 C-grid EVP no-slip
857 C-grid EVP slip
858 C-grid LSR no-slip + Winton
859 speed-performance-accuracy (small)
860 ice transport through Canadian Archipelago differences
861 thickness distribution differences
862 ice velocity and transport differences
863
864 Adjoint sensitivity experiment on 1/2-res setup
865 Sensitivity of sea ice volume flow through Fram Strait
866 *** Sensitivity of sea ice volume flow through Canadian Archipelago
867
868 Summary and conluding remarks

  ViewVC Help
Powered by ViewVC 1.1.22