/[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 1.12 - (show annotations) (download) (as text)
Mon Feb 25 22:06:17 2008 UTC (17 years, 4 months ago) by dimitri
Branch: MAIN
Changes since 1.11: +23 -12 lines
File MIME type: application/x-tex
abstract modified as per morning's discussion

1 % $Header: /u/gcmpack/MITgcm_contrib/articles/ceaice/ceaice.tex,v 1.11 2008/02/25 19:30:56 mlosch Exp $
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
56 As part of ongoing efforts to obtain a best possible synthesis of most
57 available, global-scale, ocean and sea ice data, a dynamic and thermodynamic
58 sea-ice model has been coupled to the Massachusetts Institute of Technology
59 general circulation model (MITgcm). Ice mechanics follow a viscous plastic
60 rheology and the ice momentum equations are solved numerically using either
61 line successive relaxation (LSR) or elastic-viscous-plastic (EVP) dynamic
62 models. Ice thermodynamics are represented using either a zero-heat-capacity
63 formulation or a two-layer formulation that conserves enthalpy. The model
64 includes prognostic variables for snow and for sea-ice salinity. The above
65 sea ice model components were borrowed from current-generation climate models
66 but they were reformulated on an Arakawa C-grid in order to match the MITgcm
67 oceanic grid and they were modified in many ways to permit efficient and
68 accurate automatic differentiation. This paper describes the MITgcm sea ice
69 model; it presents example Arctic and Antarctic results from a realistic,
70 eddy-permitting, global ocean and sea-ice configuration; it compares B-grid
71 and C-grid dynamic solvers in a regional Arctic configuration; and it presents
72 example results from coupled ocean and sea-ice adjoint-model integrations.
73
74 \end{abstract}
75
76 \section{Introduction}
77 \label{sec:intro}
78
79 \section{Model}
80 \label{sec:model}
81
82 Traditionally, probably for historical reasons and the ease of
83 treating the Coriolis term, most standard sea-ice models are
84 discretized on Arakawa-B-grids \citep[e.g.,][]{hibler79, harder99,
85 kreyscher00, zhang98, hunke97}. From the perspective of coupling a
86 sea ice-model to a C-grid ocean model, the exchange of fluxes of heat
87 and fresh-water pose no difficulty for a B-grid sea-ice model
88 \citep[e.g.,][]{timmermann02a}. However, surface stress is defined at
89 velocities points and thus needs to be interpolated between a B-grid
90 sea-ice model and a C-grid ocean model. While the smoothing implicitly
91 associated with this interpolation may mask grid scale noise, it may
92 in two-way coupling lead to a computational mode as will be shown. By
93 choosing a C-grid for the sea-ice model, we circumvene this difficulty
94 altogether and render the stress coupling as consistent as the
95 buoyancy coupling.
96
97 A further advantage of the C-grid formulation is apparent in narrow
98 straits. In the limit of only one grid cell between coasts there is no
99 flux allowed for a B-grid (with no-slip lateral boundary counditions),
100 whereas the C-grid formulation allows a flux of sea-ice through this
101 passage for all types of lateral boundary conditions. We (will)
102 demonstrate this effect in the Candian archipelago.
103
104 \subsection{Dynamics}
105 \label{sec:dynamics}
106
107 The momentum equations of the sea-ice model are standard with
108 \begin{equation}
109 \label{eq:momseaice}
110 m \frac{D\vek{u}}{Dt} = -mf\vek{k}\times\vek{u} + \vtau_{air} +
111 \vtau_{ocean} - m \nabla{\phi(0)} + \vek{F},
112 \end{equation}
113 where $\vek{u} = u\vek{i}+v\vek{j}$ is the ice velocity vectory, $m$
114 the ice mass per unit area, $f$ the Coriolis parameter, $g$ is the
115 gravity accelation, $\nabla\phi$ is the gradient (tilt) of the sea
116 surface height potential beneath the ice. $\phi$ is the sum of
117 atmpheric pressure $p_{a}$ and loading due to ice and snow
118 $(m_{i}+m_{s})g$. $\vtau_{air}$ and $\vtau_{ocean}$ are the wind and
119 ice-ocean stresses, respectively. $\vek{F}$ is the interaction force
120 and $\vek{i}$, $\vek{j}$, and $\vek{k}$ are the unit vectors in the
121 $x$, $y$, and $z$ directions. Advection of sea-ice momentum is
122 neglected. The wind and ice-ocean stress terms are given by
123 \begin{align*}
124 \vtau_{air} =& \rho_{air} |\vek{U}_{air}|R_{air}(\vek{U}_{air}) \\
125 \vtau_{ocean} =& \rho_{ocean} |\vek{U}_{ocean}-\vek{u}|
126 R_{ocean}(\vek{U}_{ocean}-\vek{u}), \\
127 \end{align*}
128 where $\vek{U}_{air/ocean}$ are the surface winds of the atmosphere
129 and surface currents of the ocean, respectively. $C_{air/ocean}$ are
130 air and ocean drag coefficients, $\rho_{air/ocean}$ reference
131 densities, and $R_{air/ocean}$ rotation matrices that act on the
132 wind/current vectors. $\vek{F} = \nabla\cdot\sigma$ is the divergence
133 of the interal stress tensor $\sigma_{ij}$.
134
135 For an isotropic system this stress tensor can be related to the ice
136 strain rate and strength by a nonlinear viscous-plastic (VP)
137 constitutive law \citep{hibler79, zhang98}:
138 \begin{equation}
139 \label{eq:vpequation}
140 \sigma_{ij}=2\eta(\dot{\epsilon}_{ij},P)\dot{\epsilon}_{ij}
141 + \left[\zeta(\dot{\epsilon}_{ij},P) -
142 \eta(\dot{\epsilon}_{ij},P)\right]\dot{\epsilon}_{kk}\delta_{ij}
143 - \frac{P}{2}\delta_{ij}.
144 \end{equation}
145 The ice strain rate is given by
146 \begin{equation*}
147 \dot{\epsilon}_{ij} = \frac{1}{2}\left(
148 \frac{\partial{u_{i}}}{\partial{x_{j}}} +
149 \frac{\partial{u_{j}}}{\partial{x_{i}}}\right).
150 \end{equation*}
151 The maximum ice pressure $P_{\max}$, a measure of ice strength, depends on
152 both thickness $h$ and compactness (concentration) $c$:
153 \begin{equation}
154 P_{\max} = P^{*}c\,h\,e^{[C^{*}\cdot(1-c)]},
155 \label{eq:icestrength}
156 \end{equation}
157 with the constants $P^{*}$ and $C^{*}$. The nonlinear bulk and shear
158 viscosities $\eta$ and $\zeta$ are functions of ice strain rate
159 invariants and ice strength such that the principal components of the
160 stress lie on an elliptical yield curve with the ratio of major to
161 minor axis $e$ equal to $2$; they are given by:
162 \begin{align*}
163 \zeta =& \min\left(\frac{P_{\max}}{2\max(\Delta,\Delta_{\min})},
164 \zeta_{\max}\right) \\
165 \eta =& \frac{\zeta}{e^2} \\
166 \intertext{with the abbreviation}
167 \Delta = & \left[
168 \left(\dot{\epsilon}_{11}^2+\dot{\epsilon}_{22}^2\right)
169 (1+e^{-2}) + 4e^{-2}\dot{\epsilon}_{12}^2 +
170 2\dot{\epsilon}_{11}\dot{\epsilon}_{22} (1-e^{-2})
171 \right]^{-\frac{1}{2}}
172 \end{align*}
173 The bulk viscosities are bounded above by imposing both a minimum
174 $\Delta_{\min}=10^{-11}\text{\,s}^{-1}$ (for numerical reasons) and a
175 maximum $\zeta_{\max} = P_{\max}/\Delta^*$, where
176 $\Delta^*=(5\times10^{12}/2\times10^4)\text{\,s}^{-1}$. For stress
177 tensor compuation the replacement pressure $P = 2\,\Delta\zeta$
178 \citep{hibler95} is used so that the stress state always lies on the
179 elliptic yield curve by definition.
180
181 In the so-called truncated ellipse method the shear viscosity $\eta$
182 is capped to suppress any tensile stress \citep{hibler97, geiger98}:
183 \begin{equation}
184 \label{eq:etatem}
185 \eta = \min(\frac{\zeta}{e^2}
186 \frac{\frac{P}{2}-\zeta(\dot{\epsilon}_{11}+\dot{\epsilon}_{22})}
187 {\sqrt{(\dot{\epsilon}_{11}+\dot{\epsilon}_{22})^2
188 +4\dot{\epsilon}_{12}^2}}
189 \end{equation}
190
191 In the current implementation, the VP-model is integrated with the
192 semi-implicit line successive over relaxation (LSOR)-solver of
193 \citet{zhang98}, which allows for long time steps that, in our case,
194 is limited by the explicit treatment of the Coriolis term. The
195 explicit treatment of the Coriolis term does not represent a severe
196 limitation because it restricts the time step to approximately the
197 same length as in the ocean model where the Coriolis term is also
198 treated explicitly.
199
200 \citet{hunke97}'s introduced an elastic contribution to the strain
201 rate elatic-viscous-plastic in order to regularize
202 Eq.\refeq{vpequation} in such a way that the resulting
203 elatic-viscous-plastic (EVP) and VP models are identical at steady
204 state,
205 \begin{equation}
206 \label{eq:evpequation}
207 \frac{1}{E}\frac{\partial\sigma_{ij}}{\partial{t}} +
208 \frac{1}{2\eta}\sigma_{ij}
209 + \frac{\eta - \zeta}{4\zeta\eta}\sigma_{kk}\delta_{ij}
210 + \frac{P}{4\zeta}\delta_{ij}
211 = \dot{\epsilon}_{ij}.
212 \end{equation}
213 %In the EVP model, equations for the components of the stress tensor
214 %$\sigma_{ij}$ are solved explicitly. Both model formulations will be
215 %used and compared the present sea-ice model study.
216 The EVP-model uses an explicit time stepping scheme with a short
217 timestep. According to the recommendation of \citet{hunke97}, the
218 EVP-model is stepped forward in time 120 times within the physical
219 ocean model time step (although this parameter is under debate), to
220 allow for elastic waves to disappear. Because the scheme does not
221 require a matrix inversion it is fast in spite of the small timestep
222 \citep{hunke97}. For completeness, we repeat the equations for the
223 components of the stress tensor $\sigma_{1} =
224 \sigma_{11}+\sigma_{22}$, $\sigma_{2}= \sigma_{11}-\sigma_{22}$, and
225 $\sigma_{12}$. Introducing the divergence $D_D =
226 \dot{\epsilon}_{11}+\dot{\epsilon}_{22}$, and the horizontal tension
227 and shearing strain rates, $D_T =
228 \dot{\epsilon}_{11}-\dot{\epsilon}_{22}$ and $D_S =
229 2\dot{\epsilon}_{12}$, respectively and using the above abbreviations,
230 the equations can be written as:
231 \begin{align}
232 \label{eq:evpstresstensor1}
233 \frac{\partial\sigma_{1}}{\partial{t}} + \frac{\sigma_{1}}{2T} +
234 \frac{P}{2T} &= \frac{P}{2T\Delta} D_D \\
235 \label{eq:evpstresstensor2}
236 \frac{\partial\sigma_{2}}{\partial{t}} + \frac{\sigma_{2} e^{2}}{2T}
237 &= \frac{P}{2T\Delta} D_T \\
238 \label{eq:evpstresstensor12}
239 \frac{\partial\sigma_{12}}{\partial{t}} + \frac{\sigma_{12} e^{2}}{2T}
240 &= \frac{P}{4T\Delta} D_S
241 \end{align}
242 Here, the elastic parameter $E$ is redefined in terms of a damping timescale
243 $T$ for elastic waves \[E=\frac{\zeta}{T}.\]
244 $T=E_{0}\Delta{t}$ with the tunable parameter $E_0<1$ and
245 the external (long) timestep $\Delta{t}$. \citet{hunke97} recommend
246 $E_{0} = \frac{1}{3}$.
247
248 For details of the spatial discretization, the reader is referred to
249 \citet{zhang98, zhang03}. Our discretization differs only (but
250 importantly) in the underlying grid, namely the Arakawa C-grid, but is
251 otherwise straightforward. The EVP model in particular is discretized
252 naturally on the C-grid with $\sigma_{1}$ and $\sigma_{2}$ on the
253 center points and $\sigma_{12}$ on the corner (or vorticity) points of
254 the grid. With this choice all derivatives are discretized as central
255 differences and averaging is only involved in computing $\Delta$ and
256 $P$ at vorticity points.
257
258 For a general curvilinear grid, one needs in principle to take metric
259 terms into account that arise in the transformation a curvilinear grid
260 on the sphere. However, for now we can neglect these metric terms
261 because they are very small on the cubed sphere grids used in this
262 paper; in particular, only near the edges of the cubed sphere grid, we
263 expect them to be non-zero, but these edges are at approximately
264 35\degS\ or 35\degN\ which are never covered by sea-ice in our
265 simulations. Everywhere else the coordinate system is locally nearly
266 cartesian. However, for last-glacial-maximum or snowball-earth-like
267 simulations the question of metric terms needs to be reconsidered.
268 Either, one includes these terms as in \citet{zhang03}, or one finds a
269 vector-invariant formulation fo the sea-ice internal stress term that
270 does not require any metric terms, as it is done in the ocean dynamics
271 of the MITgcm \citep{adcroft04:_cubed_sphere}.
272
273 Moving sea ice exerts a stress on the ocean which is the opposite of
274 the stress $\vtau_{ocean}$ in Eq.\refeq{momseaice}. This stess is
275 applied directly to the surface layer of the ocean model. An
276 alternative ocean stress formulation is given by \citet{hibler87}.
277 Rather than applying $\vtau_{ocean}$ directly, the stress is derived
278 from integrating over the ice thickness to the bottom of the oceanic
279 surface layer. In the resulting equation for the \emph{combined}
280 ocean-ice momentum, the interfacial stress cancels and the total
281 stress appears as the sum of windstress and divergence of internal ice
282 stresses: $\delta(z) (\vtau_{air} + \vek{F})/\rho_0$, \citep[see also
283 Eq.\,2 of][]{hibler87}. The disadvantage of this formulation is that
284 now the velocity in the surface layer of the ocean that is used to
285 advect tracers, is really an average over the ocean surface
286 velocity and the ice velocity leading to an inconsistency as the ice
287 temperature and salinity are different from the oceanic variables.
288
289 Sea ice distributions are characterized by sharp gradients and edges.
290 For this reason, we employ a positive 3rd-order advection scheme
291 \citep{hundsdorfer94} for the thermodynamic variables discussed in the
292 next section.
293
294 \subparagraph{boundary conditions: no-slip, free-slip, half-slip}
295
296 \begin{itemize}
297 \item transition from existing B-Grid to C-Grid
298 \item boundary conditions: no-slip, free-slip, half-slip
299 \item fancy (multi dimensional) advection schemes
300 \item VP vs.\ EVP \citep{hunke97}
301 \item ocean stress formulation \citep{hibler87}
302 \end{itemize}
303
304 \subsection{Thermodynamics}
305 \label{sec:thermodynamics}
306
307 In the original formulation the sea ice model \citep{menemenlis05}
308 uses simple thermodynamics following the appendix of
309 \citet{semtner76}. This formulation does not allow storage of heat
310 (heat capacity of ice is zero, and this type of model is often refered
311 to as a ``zero-layer'' model). Upward heat flux is parameterized
312 assuming a linear temperature profile and together with a constant ice
313 conductivity. It is expressed as $(K/h)(T_{w}-T_{0})$, where $K$ is
314 the ice conductivity, $h$ the ice thickness, and $T_{w}-T_{0}$ the
315 difference between water and ice surface temperatures. The surface
316 heat budget is computed in a similar way to that of
317 \citet{parkinson79} and \citet{manabe79}.
318
319 There is considerable doubt about the reliability of such a simple
320 thermodynamic model---\citet{semtner84} found significant errors in
321 phase (one month lead) and amplitude ($\approx$50\%\,overestimate) in
322 such models---, so that today many sea ice models employ more complex
323 thermodynamics. A popular thermodynamics model of \citet{winton00} is
324 based on the 3-layer model of \citet{semtner76} and treats brine
325 content by means of enthalphy conservation. This model requires in
326 addition to ice-thickness and compactness (fractional area) additional
327 state variables to be advected by ice velocities, namely enthalphy of
328 the two ice layers and the thickness of the overlying snow layer.
329
330
331 \subsection{C-grid}
332 \begin{itemize}
333 \item no-slip vs. free-slip for both lsr and evp;
334 "diagnostics" to look at and use for comparison
335 \begin{itemize}
336 \item ice transport through Fram Strait/Denmark Strait/Davis
337 Strait/Bering strait (these are general)
338 \item ice transport through narrow passages, e.g.\ Nares-Strait
339 \end{itemize}
340 \item compare different advection schemes (if lsr turns out to be more
341 effective, then with lsr otherwise I prefer evp), eg.
342 \begin{itemize}
343 \item default 2nd-order with diff1=0.002
344 \item 1st-order upwind with diff1=0.
345 \item DST3FL (SEAICEadvScheme=33 with diff1=0., should work, works for me)
346 \item 2nd-order wit flux limiter (SEAICEadvScheme=77 with diff1=0.)
347 \end{itemize}
348 That should be enough. Here, total ice mass and location of ice edge
349 is interesting. However, this comparison can be done in an idealized
350 domain, may not require full Arctic Domain?
351 \item
352 Do a little study on the parameters of LSR and EVP
353 \begin{enumerate}
354 \item convergence of LSR, how many iterations do you need to get a
355 true elliptic yield curve
356 \item vary deltaTevp and the relaxation parameter for EVP and see when
357 the EVP solution breaks down (relative to the forcing time scale).
358 For this, it is essential that the evp solver gives use "stripeless"
359 solutions, that is your dtevp = 1sec solutions/or 10sec solutions
360 with SEAICE\_evpDampC = 615.
361 \end{enumerate}
362 \end{itemize}
363
364 \section{Forward sensitivity experiments}
365 \label{sec:forward}
366
367 A second series of forward sensitivity experiments have been carried out on an
368 Arctic Ocean domain with open boundaries. Once again the objective is to
369 compare the old B-grid LSR dynamic solver with the new C-grid LSR and EVP
370 solvers. One additional experiment is carried out to illustrate the
371 differences between the two main options for sea ice thermodynamics in the MITgcm.
372
373 \subsection{Arctic Domain with Open Boundaries}
374 \label{sec:arctic}
375
376 The Arctic domain of integration is illustrated in Fig.~\ref{fig:arctic1}. It
377 is carved out from, and obtains open boundary conditions from, the
378 global cubed-sphere configuration of the Estimating the Circulation
379 and Climate of the Ocean, Phase II (ECCO2) project
380 \citet{menemenlis05}. The domain size is 420 by 384 grid boxes
381 horizontally with mean horizontal grid spacing of 18 km.
382
383 \begin{figure}
384 %\centerline{{\includegraphics*[width=0.44\linewidth]{\fpath/arctic1.eps}}}
385 \caption{Bathymetry of Arctic Domain.\label{fig:arctic1}}
386 \end{figure}
387
388 There are 50 vertical levels ranging in thickness from 10 m near the surface
389 to approximately 450 m at a maximum model depth of 6150 m. Bathymetry is from
390 the National Geophysical Data Center (NGDC) 2-minute gridded global relief
391 data (ETOPO2) and the model employs the partial-cell formulation of
392 \citet{adcroft97:_shaved_cells}, which permits accurate representation of the bathymetry. The
393 model is integrated in a volume-conserving configuration using a finite volume
394 discretization with C-grid staggering of the prognostic variables. In the
395 ocean, the non-linear equation of state of \citet{jackett95}. The ocean model is
396 coupled to a sea-ice model described hereinabove.
397
398 This particular ECCO2 simulation is initialized from rest using the
399 January temperature and salinity distribution from the World Ocean
400 Atlas 2001 (WOA01) [Conkright et al., 2002] and it is integrated for
401 32 years prior to the 1996--2001 period discussed in the study. Surface
402 boundary conditions are from the National Centers for Environmental
403 Prediction and the National Center for Atmospheric Research
404 (NCEP/NCAR) atmospheric reanalysis [Kistler et al., 2001]. Six-hourly
405 surface winds, temperature, humidity, downward short- and long-wave
406 radiations, and precipitation are converted to heat, freshwater, and
407 wind stress fluxes using the \citet{large81, large82} bulk formulae.
408 Shortwave radiation decays exponentially as per Paulson and Simpson
409 [1977]. Additionally the time-mean river run-off from Large and Nurser
410 [2001] is applied and there is a relaxation to the monthly-mean
411 climatological sea surface salinity values from WOA01 with a
412 relaxation time scale of 3 months. Vertical mixing follows
413 \citet{large94} with background vertical diffusivity of
414 $1.5\times10^{-5}\text{\,m$^{2}$\,s$^{-1}$}$ and viscosity of
415 $10^{-3}\text{\,m$^{2}$\,s$^{-1}$}$. A third order, direct-space-time
416 advection scheme with flux limiter is employed \citep{hundsdorfer94}
417 and there is no explicit horizontal diffusivity. Horizontal viscosity
418 follows \citet{lei96} but
419 modified to sense the divergent flow as per Fox-Kemper and Menemenlis
420 [in press]. Shortwave radiation decays exponentially as per Paulson
421 and Simpson [1977]. Additionally, the time-mean runoff of Large and
422 Nurser [2001] is applied near the coastline and, where there is open
423 water, there is a relaxation to monthly-mean WOA01 sea surface
424 salinity with a time constant of 45 days.
425
426 Open water, dry
427 ice, wet ice, dry snow, and wet snow albedo are, respectively, 0.15, 0.85,
428 0.76, 0.94, and 0.8.
429
430 \begin{itemize}
431 \item Configuration
432 \item OBCS from cube
433 \item forcing
434 \item 1/2 and full resolution
435 \item with a few JFM figs from C-grid LSR no slip
436 ice transport through Canadian Archipelago
437 thickness distribution
438 ice velocity and transport
439 \end{itemize}
440
441 \begin{itemize}
442 \item Arctic configuration
443 \item ice transport through straits and near boundaries
444 \item focus on narrow straits in the Canadian Archipelago
445 \end{itemize}
446
447 \begin{itemize}
448 \item B-grid LSR no-slip
449 \item C-grid LSR no-slip
450 \item C-grid LSR slip
451 \item C-grid EVP no-slip
452 \item C-grid EVP slip
453 \item C-grid LSR + TEM (truncated ellipse method, no tensile stress, new flag)
454 \item C-grid LSR no-slip + Winton
455 \item speed-performance-accuracy (small)
456 ice transport through Canadian Archipelago differences
457 thickness distribution differences
458 ice velocity and transport differences
459 \end{itemize}
460
461 We anticipate small differences between the different models due to:
462 \begin{itemize}
463 \item advection schemes: along the ice-edge and regions with large
464 gradients
465 \item C-grid: less transport through narrow straits for no slip
466 conditons, more for free slip
467 \item VP vs.\ EVP: speed performance, accuracy?
468 \item ocean stress: different water mass properties beneath the ice
469 \end{itemize}
470
471 \section{Adjoint sensiivities of the MITsim}
472
473 \subsection{The adjoint of MITsim}
474
475 The ability to generate tangent linear and adjoint model components
476 of the MITsim has been a main design task.
477 For the ocean the adjoint capability has proven to be an
478 invaluable tool for sensitivity analysis as well as state estimation.
479 In short, the adjoint enables very efficient computation of the gradient
480 of scalar-valued model diagnostics (called cost function or objective function)
481 with respect to many model "variables".
482 These variables can be two- or three-dimensional fields of initial
483 conditions, model parameters such as mixing coefficients, or
484 time-varying surface or lateral (open) boundary conditions.
485 When combined, these variables span a potentially high-dimensional
486 (e.g. O(10$^8$)) so-called control space. Performing parameter perturbations
487 to assess model sensitivities quickly becomes prohibitive at these scales.
488 Alternatively, (time-varying) sensitivities of the objective function
489 to any element of the control space can be computed very efficiently in
490 one single adjoint
491 model integration, provided an efficient adjoint model is available.
492
493 [REFERENCES]
494
495
496 The adjoint operator (ADM) is the transpose of the tangent linear operator (TLM)
497 of the full (in general nonlinear) forward model, i.e. the MITsim.
498 The TLM maps perturbations of elements of the control space
499 (e.g. initial ice thickness distribution)
500 via the model Jacobian
501 to a perturbation in the objective function
502 (e.g. sea-ice export at the end of the integration interval).
503 \textit{Tangent} linearity ensures that the derivatives are evaluated
504 with respect to the underlying model trajectory at each point in time.
505 This is crucial for nonlinear trajectories and the presence of different
506 regimes (e.g. effect of the seaice growth term at or away from the
507 freezing point of the ocean surface).
508 Ensuring tangent linearity can be easily achieved by integrating
509 the full model in sync with the TLM to provide the underlying model state.
510 Ensuring \textit{tangent} adjoints is equally crucial, but much more
511 difficult to achieve because of the reverse nature of the integration:
512 the adjoint accumulates sensitivities backward in time,
513 starting from a unit perturbation of the objective function.
514 The adjoint model requires the model state in reverse order.
515 This presents one of the major complications in deriving an
516 exact, i.e. \textit{tangent} adjoint model.
517
518 Following closely the development and maintenance of TLM and ADM
519 components of the MITgcm we have relied heavily on the
520 autmomatic differentiation (AD) tool
521 "Transformation of Algorithms in Fortran" (TAF)
522 developed by Fastopt (Giering and Kaminski, 1998)
523 to derive TLM and ADM code of the MITsim.
524 Briefly, the nonlinear parent model is fed to the AD tool which produces
525 derivative code for the specified control space and objective function.
526 Following this approach has (apart from its evident success)
527 several advantages:
528 (1) the adjoint model is the exact adjoint operator of the parent model,
529 (2) the adjoint model can be kept up to date with respect to ongoing
530 development of the parent model, and adjustments to the parent model
531 to extend the automatically generated adjoint are incremental changes
532 only, rather than extensive re-developments,
533 (3) the parallel structure of the parent model is preserved
534 by the adjoint model, ensuring efficient use in high performance
535 computing environments.
536
537 Some initial code adjustments are required to support dependency analysis
538 of the flow reversal and certain language limitations which may lead
539 to irreducible flow graphs (e.g. GOTO statements).
540 The problem of providing the required model state in reverse order
541 at the time of evaluating nonlinear or conditional
542 derivatives is solved via balancing
543 storing vs. recomputation of the model state in a multi-level
544 checkpointing loop.
545 Again, an initial code adjustment is required to support TAFs
546 checkpointing capability.
547 The code adjustments are sufficiently simple so as not to cause
548 major limitations to the full nonlinear parent model.
549 Once in place, an adjoint model of a new model configuration
550 may be derived in about 10 minutes.
551
552 [HIGHLIGHT COUPLED NATURE OF THE ADJOINT!]
553
554 \subsection{Special considerations}
555
556 * growth term(?)
557
558 * small active denominators
559
560 * dynamic solver (implicit function theorem)
561
562 * approximate adjoints
563
564
565 \subsection{An example: sensitivities of sea-ice export through Fram Strait}
566
567 We demonstrate the power of the adjoint method
568 in the context of investigating sea-ice export sensitivities through Fram Strait
569 (for details of this study see Heimbach et al., 2007).
570 %\citep[for details of this study see][]{heimbach07}. %Heimbach et al., 2007).
571 The domain chosen is a coarsened version of the Arctic face of the
572 high-resolution cubed-sphere configuration of the ECCO2 project
573 \citep[see][]{menemenlis05}. It covers the entire Arctic,
574 extends into the North Pacific such as to cover the entire
575 ice-covered regions, and comprises parts of the North Atlantic
576 down to XXN to enable analysis of remote influences of the
577 North Atlantic current to sea-ice variability and export.
578 The horizontal resolution varies between XX and YY km
579 with 50 unevenly spaced vertical levels.
580 The adjoint models run efficiently on 80 processors
581 (benchmarks have been performed both on an SGI Altix as well as an
582 IBM SP5 at NASA/ARC).
583
584 Following a 1-year spinup, the model has been integrated for four
585 years between 1992 and 1995. It is forced using realistic 6-hourly
586 NCEP/NCAR atmospheric state variables. Over the open ocean these are
587 converted into air-sea fluxes via the bulk formulae of
588 \citet{large04}. Derivation of air-sea fluxes in the presence of
589 sea-ice is handled by the ice model as described in \refsec{model}.
590 The objective function chosen is sea-ice export through Fram Strait
591 computed for December 1995. The adjoint model computes sensitivities
592 to sea-ice export back in time from 1995 to 1992 along this
593 trajectory. In principle all adjoint model variable (i.e., Lagrange
594 multipliers) of the coupled ocean/sea-ice model are available to
595 analyze the transient sensitivity behaviour of the ocean and sea-ice
596 state. Over the open ocean, the adjoint of the bulk formula scheme
597 computes sensitivities to the time-varying atmospheric state. Over
598 ice-covered parts, the sea-ice adjoint converts surface ocean
599 sensitivities to atmospheric sensitivities.
600
601 \reffig{4yradjheff}(a--d) depict sensitivities of sea-ice export
602 through Fram Strait in December 1995 to changes in sea-ice thickness
603 12, 24, 36, 48 months back in time. Corresponding sensitivities to
604 ocean surface temperature are depicted in
605 \reffig{4yradjthetalev1}(a--d). The main characteristics is
606 consistency with expected advection of sea-ice over the relevant time
607 scales considered. The general positive pattern means that an
608 increase in sea-ice thickness at location $(x,y)$ and time $t$ will
609 increase sea-ice export through Fram Strait at time $T_e$. Largest
610 distances from Fram Strait indicate fastest sea-ice advection over the
611 time span considered. The ice thickness sensitivities are in close
612 correspondence to ocean surface sentivitites, but of opposite sign.
613 An increase in temperature will incur ice melting, decrease in ice
614 thickness, and therefore decrease in sea-ice export at time $T_e$.
615
616 The picture is fundamentally different and much more complex
617 for sensitivities to ocean temperatures away from the surface.
618 \reffig{4yradjthetalev10??}(a--d) depicts ice export sensitivities to
619 temperatures at roughly 400 m depth.
620 Primary features are the effect of the heat transport of the North
621 Atlantic current which feeds into the West Spitsbergen current,
622 the circulation around Svalbard, and ...
623
624 \begin{figure}[t!]
625 \centerline{
626 \subfigure[{\footnotesize -12 months}]
627 {\includegraphics*[width=0.44\linewidth]{\fpath/run_4yr_ADJheff_arc_lev1_tim072_cmax2.0E+02.eps}}
628 %\includegraphics*[width=.3\textwidth]{H_c.bin_res_100_lev1.pdf}
629 %
630 \subfigure[{\footnotesize -24 months}]
631 {\includegraphics*[width=0.44\linewidth]{\fpath/run_4yr_ADJheff_arc_lev1_tim145_cmax2.0E+02.eps}}
632 }
633
634 \centerline{
635 \subfigure[{\footnotesize
636 -36 months}]
637 {\includegraphics*[width=0.44\linewidth]{\fpath/run_4yr_ADJheff_arc_lev1_tim218_cmax2.0E+02.eps}}
638 %
639 \subfigure[{\footnotesize
640 -48 months}]
641 {\includegraphics*[width=0.44\linewidth]{\fpath/run_4yr_ADJheff_arc_lev1_tim292_cmax2.0E+02.eps}}
642 }
643 \caption{Sensitivity of sea-ice export through Fram Strait in December 2005 to
644 sea-ice thickness at various prior times.
645 \label{fig:4yradjheff}}
646 \end{figure}
647
648
649 \begin{figure}[t!]
650 \centerline{
651 \subfigure[{\footnotesize -12 months}]
652 {\includegraphics*[width=0.44\linewidth]{\fpath/run_4yr_ADJtheta_arc_lev1_tim072_cmax5.0E+01.eps}}
653 %\includegraphics*[width=.3\textwidth]{H_c.bin_res_100_lev1.pdf}
654 %
655 \subfigure[{\footnotesize -24 months}]
656 {\includegraphics*[width=0.44\linewidth]{\fpath/run_4yr_ADJtheta_arc_lev1_tim145_cmax5.0E+01.eps}}
657 }
658
659 \centerline{
660 \subfigure[{\footnotesize
661 -36 months}]
662 {\includegraphics*[width=0.44\linewidth]{\fpath/run_4yr_ADJtheta_arc_lev1_tim218_cmax5.0E+01.eps}}
663 %
664 \subfigure[{\footnotesize
665 -48 months}]
666 {\includegraphics*[width=0.44\linewidth]{\fpath/run_4yr_ADJtheta_arc_lev1_tim292_cmax5.0E+01.eps}}
667 }
668 \caption{Same as \reffig{4yradjheff} but for sea surface temperature
669 \label{fig:4yradjthetalev1}}
670 \end{figure}
671
672
673
674 \section{Discussion and conclusion}
675 \label{sec:concl}
676
677 The story of the paper could be:
678 B-grid ice model + C-grid ocean model does not work properly for us,
679 therefore C-grid ice model with advantages:
680 \begin{enumerate}
681 \item stress coupling simpler (no interpolation required)
682 \item different boundary conditions
683 \item advection schemes carry over trivially from main code
684 \end{enumerate}
685 LSR/EVP solutions are similar with appropriate bcs, evp parameters as
686 a function of forcing time scale (when does VP solution break
687 down). Same for LSR solver, provided that it works (o:
688 Which scheme is more efficient for the resolution/time stepping
689 parameters that we use here. What about other resolutions?
690
691 \paragraph{Acknowledgements}
692 We thank Jinlun Zhang for providing the original B-grid code and many
693 helpful discussions. ML thanks Elizabeth Hunke for multiple explanations.
694
695 \bibliography{bib/journal_abrvs,bib/seaice,bib/genocean,bib/maths,bib/mitgcmuv,bib/fram}
696
697 \end{document}
698
699 %%% Local Variables:
700 %%% mode: latex
701 %%% TeX-master: t
702 %%% End:
703
704
705 A Dynamic-Thermodynamic Sea ice Model for Ocean Climate
706 Estimation on an Arakawa C-Grid
707
708 Introduction
709
710 Ice Model:
711 Dynamics formulation.
712 B-C, LSR, EVP, no-slip, slip
713 parallellization
714 Thermodynamics formulation.
715 0-layer Hibler salinity + snow
716 3-layer Winton
717
718 Idealized tests
719 Funnel Experiments
720 Downstream Island tests
721 B-grid LSR no-slip
722 C-grid LSR no-slip
723 C-grid LSR slip
724 C-grid EVP no-slip
725 C-grid EVP slip
726
727 Arctic Setup
728 Configuration
729 OBCS from cube
730 forcing
731 1/2 and full resolution
732 with a few JFM figs from C-grid LSR no slip
733 ice transport through Canadian Archipelago
734 thickness distribution
735 ice velocity and transport
736
737 Arctic forward sensitivity experiments
738 B-grid LSR no-slip
739 C-grid LSR no-slip
740 C-grid LSR slip
741 C-grid EVP no-slip
742 C-grid EVP slip
743 C-grid LSR no-slip + Winton
744 speed-performance-accuracy (small)
745 ice transport through Canadian Archipelago differences
746 thickness distribution differences
747 ice velocity and transport differences
748
749 Adjoint sensitivity experiment on 1/2-res setup
750 Sensitivity of sea ice volume flow through Fram Strait
751 *** Sensitivity of sea ice volume flow through Canadian Archipelago
752
753 Summary and conluding remarks

  ViewVC Help
Powered by ViewVC 1.1.22