1: \documentclass[12pt,preprint]{aastex}
2: %\documentclass{emulateapj}
3:
4: \newcommand{\ms}{\mbox{m s$^{-1}$}}
5: \newcommand{\degrees}{\mbox{$^\mathrm{o}$}}
6:
7: \begin{document}
8:
9: \title{On Signatures of Atmospheric Features in Thermal Phase Curves
10: of Hot Jupiters}
11:
12: \author{Emily Rauscher, Kristen Menou,}
13:
14: \affil{Department of Astronomy, Columbia University,\\ 550 W. 120th
15: Street, New York, NY 10027, USA}
16:
17: \author{James Y-K.\ Cho,}
18:
19: \affil{Astronomy Unit, School of
20: Mathematical Sciences, Queen Mary, University of London, \\ Mile End
21: Road, London E1 4NS, UK}
22:
23:
24: \author{Sara Seager}
25:
26: \affil{Dept. of Earth, Atmospheric, and Planetary sciences, and
27: Dept. of Physics, Massachusetts Institute of Technology, 54-1626, 77
28: Massachusetts Ave., Cambridge, MA, 02139, USA}
29:
30: \and
31:
32: \author{Bradley M.\ S.\ Hansen,}
33:
34: \affil{Department of Physics and Astronomy and Institute for
35: Geophysics and Planetary Physics, University of California,
36: 475 Portola Plaza, Box 951547, Los Angeles, CA 90095, USA}
37:
38:
39: \begin{abstract}
40: Turbulence is ubiquitous in Solar System planetary atmospheres. In
41: hot Jupiter atmospheres, the combination of moderately slow rotation
42: and thick pressure scale height may result in dynamical weather
43: structures with {unusually} large, planetary-size scales. Using
44: equivalent-barotropic, turbulent circulation models, we illustrate
45: how such structures can generate a variety of features in the
46: thermal phase curves of hot Jupiters, including phase shifts and
47: deviations from periodicity. {Such features may have been spotted in
48: the recent infrared phase curve of HD~189733b.} Despite inherent
49: difficulties with the interpretation of disk-integrated quantities,
50: phase curves promise to offer unique constraints on the nature of
51: the circulation regime present on hot Jupiters.
52: \end{abstract}
53:
54: \keywords{Infrared: General, Infrared: Stars, Stars: Planetary Systems, Stars: Atmospheres, Turbulence}
55:
56:
57: \section{Introduction}
58:
59: Just over two years ago the \emph{Spitzer Space Telescope} was used
60: for the first direct detections of light from extrasolar planets
61: \citep{Charbonneau2005,Deming2005}, heralding the birth of a new field
62: of observational astronomy: the characterization of exoplanet
63: atmospheres. So far these direct measurements are only possible for
64: the shortest period ``hot Jupiter'' exoplanets, as these planets are
65: heated enough to radiate $\sim 10^{-3}$ of their parent stars'
66: infrared emission, making {the planet's component of the} signal just
67: separable from the stellar component. In the last couple {of} years,
68: many new results have emerged in this field, including several more
69: infrared direct detections of hot Jupiters (and a Neptune!) through
70: secondary eclipse measurements
71: \citep{Deming2006,Deming2007,Harrington2007}, measurements of infrared
72: emission spectra \citep{Grillmair2007,Richardson2007}, and orbital
73: phase variation from the day-night temperature contrasts of eclipsing
74: and non-eclipsing planets
75: \citep{Harrington2006,Cowan2007,Knutson2007}. Further exciting
76: results are expected from the last round of cryogenic \emph{Spitzer}
77: observations, the ``warm \emph{Spitzer}'' phase \citep{Deming2007b},
78: and continuing on with the vastly increased photometric sensitivity of
79: the \emph{James Webb Space Telescope} ({\it JWST}). The launch of
80: \emph{JWST} will enable high precision measurements and the
81: acquisition of detailed information about the atmospheres of these
82: planets, including the ability to use the known geometry of eclipsing
83: systems to make resolved maps of their day sides
84: \citep{Williams2006,Rauscher2007b}.
85:
86: As the number of directly observed systems continues to grow, it {is
87: becoming} increasingly apparent that not all hot Jupiters are alike.
88: Differences in planetary radii, brightness temperatures or inferred
89: efficiencies of heat redistribution from permanent day sides to cold
90: night sides have led to theoretical speculations on possible ways to
91: subclassify hot Jupiters \citep{Burrows2007,Fortney2007a,Hansen2007b}.
92: Perhaps we should not be surprised to discover diversity in these
93: systems \citep[see, e.g.,][]{Redfield2007} since even our own Solar
94: System planets exhibit a wide range of atmospheric conditions. For
95: example, it is currently not well understood what sets the wind speeds
96: at cloud-deck level on the Solar System gas giants. Since atmospheric
97: circulation can be largely driven by insolation, one might expect those
98: planets closer to the Sun to experience stronger winds, but the fact
99: that Neptune receives much less solar illumination than Jupiter and
100: yet has faster winds illustrates how misleading simple arguments may
101: be when applied to these complex, nonlinear flows (\citealt{Cho2006};
102: see review by \citealt{Showman2007}).
103:
104: Even a thorough understanding of the atmospheric dynamics of the
105: planets in the Solar System would not necessarily allow us to
106: extrapolate to hot Jupiters, {which are characterized by a
107: very} different atmospheric regime. These presumably tidally locked,
108: short-period planets experience intense stellar irradiation on their
109: constant day sides, {much in excess of the internally generated
110: heat flux}, while their night sides remain in perpetual shadow. In
111: addition, based on their relatively slow rotation periods, dynamical
112: arguments suggest that large scale atmospheric structures can form on
113: these planets \citep{Cho2003,Menou2003,Showman2002}, in contrast to
114: the many thin jets and smaller structures observed on Jupiter, for
115: example. Given this unusual combination of asymmetric heating and
116: slow rotation, there have been several recent attempts to model the
117: atmospheric flow on hot Jupiters \citep[see review by Showman et
118: al. 2007]{Showman2002,Cho2003,Cho2006,C&S05,Cooper2006,Fortney2006,Langton2007,Dobbs-Dixon2007}.
119:
120: The combination of strong external irradiation (which promotes
121: vertical static stability), moderately slow rotation and large
122: pressure scale heights in hot Jupiter atmospheres results in Rossby
123: deformation radii comparable to or larger than the planetary radii
124: themselves \citep{Showman2002,Cho2003,Menou2003,Showman2007}.
125: {The Rossby deformation radius acts as a limiting scale for
126: dynamics in a rotating stratified flow. It sets the typical horizontal
127: size of large vortices in a turbulent atmosphere and determines the
128: scale of the most unstable baroclinic modes. The large deformation
129: radii in hot Jupiter atmospheres may thus limit the importance of
130: baroclinic instabilities and favor a largely barotropic\footnote{In
131: this context, barotropic refers to a flow in which baroclinic
132: instabilities, which result in a form of horizontal convection, play a
133: minimal role (see \citet{Showman2007} for details).} atmospheric
134: flow.}
135:
136: These considerations led \citet{Cho2003,Cho2006} to apply the
137: two-dimensional, equivalent-barotropic formulation of the primitive
138: equations of meteorology \citep{Salby1989} to hot Jupiter atmospheres.
139: {This vertically integrated formulation emphasizes the
140: quasi-two dimensional, barotropic nature of the flow. While only
141: adiabatic simulations have been reported so far, a great advantage of
142: the equivalent-barotropic model is that small horizontal scales are
143: resolved in a fully turbulent flow \citep[see, e.g.,][for turbulent
144: energy spectra]{Cho1996}. An additional advantage of this approach is
145: that the set of physical and numerical model parameters is limited and
146: it includes the unknown magnitude of global average wind speed which
147: is treated as an input parameter \citep{Cho2003,Cho2006}.} This
148: allows us to {perform an analysis in which we explore a large range of
149: possible equivalent-barotropic flows for hot Jupiter atmospheres and
150: study how these diverse flows may translate into a diversity of
151: thermal phase curves for hot Jupiters.}
152:
153:
154: We begin by describing the models used and our method for computing
155: phase curves in $\S$~2, move on to discuss predicted model phase
156: curves in $\S$~3, consider implications for current and future
157: observations in $\S$~4, and finally conclude in $\S$~5.
158:
159: \section{Method}
160:
161: We use the adiabatic, equivalent-barotropic model of
162: \citet{Cho2003,Cho2006} for the planet HD~209458b throughout our
163: analysis. Our main results for thermal phase curves do not strongly
164: depend on the specific planetary parameters adopted---as long
165: as one is concerned with hot Jupiters of the same, broad dynamical
166: class as HD~209458b \citep[see, e.g.,][]{Menou2003}. The reader is
167: referred to \citet{Cho2006} for a detailed description of the
168: circulation model. Briefly, the stellar radiative forcing imposes a
169: steady day-night temperature difference on the flow. In the adiabatic
170: equivalent-barotropic model, this forcing is represented by a forced
171: bending of the model's bottom boundary surface, with a specified bend
172: amplitude $\eta$. {This corresponds to a maximum day-night
173: temperature variation of amplitude $\Delta T=\pm \eta \bar{T}$, where
174: $\bar{T}$ is the average equivalent temperature of the modeled layer.
175: The pre-existing background flow, with a prescribed average global
176: wind speed, $\bar{U}$, produces temperature variations, which are
177: superimposed on those due to the steady day-night difference. The
178: combined temperature range covered by each model, as well as the range
179: due to radiative forcing only (case $\bar{U}=0$), are given in
180: Table~1.} The model runs have been performed at T63 spectral
181: resolution (equivalent to a fully-resolved 192 $\times$ 96
182: longitude-latitude grid), which has been found to be sufficient to
183: capture the formation of dominant atmospheric structures
184: \citep{Rauscher2007}. Following the initialization procedure
185: described in \citet{Cho2003,Cho2006}, each run {has been} allowed to
186: relax for 40 {planetary} orbits, after which we output a snapshot of
187: the temperature structure on the planet, 100 times per orbit, for the
188: next 60 orbits.
189:
190: We create the phase curves presented here by rotating each temperature
191: map to the appropriate viewing orientation for its orbital phase angle
192: and assumed inclination, and orthographically projecting it onto a two
193: dimensional disk. Unless otherwise specified, a $90$\degrees
194: ~inclination (edge-on orientation) is assumed by default. We sample
195: the disk with a grid of $[r=20,\theta=40]$ resolution. We solve for
196: three values of emergent flux from each disk element area: a
197: bolometric value $\propto T^4$, and \emph{Spitzer} band fluxes
198: calculated either from a blackbody spectrum based on the local
199: temperature $T$ or a spectral model interpolated from a grid of
200: cloudless 1-D radiative transfer calculations \citep{Seager2005}. As
201: we shall see, our main conclusions depend only weakly on the specific
202: emission model adopted. {After obtaining the flux}, we integrate over
203: the disk, weighting the emergent flux from each grid point by its
204: relative area on the disk. Our procedure is similar to that adopted
205: in \citet{Rauscher2007}.
206:
207: We performed resolution tests of our phase curve analysis, both in the
208: number of temperature maps sampled per orbit and the number of grid
209: elements used to sample each map. For computational speed, we use
210: $[r=20,\theta=40]$ as the lowest possible resolution that is immune to
211: noise from under-sampling. The phase curves produced using this
212: resolution are accurate to within a few percent of those produced
213: at higher resolutions. Similarly, we find that by sampling the
214: temperature map of the planet 50 times per orbit, we are able to
215: adequately track the orbital flux variation without losing details in
216: the structure of the phase curve. We have also found that 10 orbits
217: are sufficient to sample the variable patterns seen in the phase
218: curves, making it unnecessary to analyze the full 60 orbits of our
219: model outputs.
220:
221:
222: \section{Phase Curves from Equivalent-Barotropic Models}
223:
224: The temperature structures predicted by {the equivalent-barotropic
225: model} consist of two superimposed components: a global day-night
226: ``fixed'' temperature gradient {(the forced bending---see \S~2)} whose
227: strength depends on the imposed radiative forcing (via the parameter
228: $\eta$), and a variable pattern of weather features created by
229: atmospheric winds. The most prominent weather features are cold
230: circumpolar vortices whose strength {(vorticity)} increases with
231: higher values of $\bar{U}$. The vortices rotate around the poles with
232: a period comparable, but not equal, to the spin {(=orbital)} period of
233: the planet. While these vortices are the weather features that most
234: strongly affect the emitted thermal flux in our models, other large
235: scale, possibly transient, features generally exist in {an}
236: atmosphere. For example, in the models presented here we observe a
237: broader warm area that develops on the side of the planet opposite
238: from the circumpolar vortices, and on occasion cold transient features
239: that develop on the night side. {These features constitute the
240: weather on these model planets.} Figure~\ref{fig:images} shows
241: temperature snapshots for the $\eta=0.10$, $\bar{U}$ = 800 \ms~model,
242: exemplifying the type of weather features that develop. Our
243: discussion will primarily focus on the circumpolar vortices, as they
244: are the {dynamical} features that most strongly affect the orbital
245: phase curves, but it is worth noting that they do not completely
246: determine the observable signatures in the model phase curves.
247:
248: The nature of the global temperature structure is determined by the
249: relative strengths of the two superimposed components mentioned above.
250: The models with high $\eta$ and low $\bar{U}$ values have strong
251: day-night temperature contrasts with weak variable weather patterns.
252: Their thermal emission phase curves are thus primarily a simple
253: day-night (peak-trough) pattern with only minor deviations from weak
254: weather features. The low $\eta$, high $\bar{U}$ models, on the other
255: hand, retain some day-night temperature difference, but they can have
256: temperature structures strongly influenced by the variable weather
257: features. The corresponding phase curves are highly variable from one
258: orbit to the next, {since they are strongly dependent on the
259: location of cold polar vortices relative to the day-night gradient.}
260: Here we survey the $(\eta, \bar{U})$ parameter space to determine the
261: range of possible characteristics of model phase curves.
262:
263: \subsection{The Influence of Wind Strength} \label{sec:ubar}
264:
265: First we consider models with the radiative forcing parameter $\eta$
266: fixed ($\eta=0.05$). Figure~\ref{fig:u_values} shows the bolometric
267: phase curves resulting from runs with average global wind strength
268: $\bar{U}$ ranging from 100 to 800 \ms. One immediately apparent
269: result is that the amount of predicted flux variation increases with
270: $\bar{U}$. This is to be expected, since rotating cold polar vortices
271: and other weather features have an increasingly dominant role in the
272: temperature structures of models with stronger wind strengths. For
273: example, when the vortices are on the night side, there is a stronger
274: day-night temperature difference than would result from radiative
275: forcing alone, and this allows for more orbital flux variation than in
276: models where the vortex-induced temperatures are weak compared to
277: the imposed day-night temperature difference.
278:
279:
280: We do not consider global wind strengths in excess of 800 \ms \,in
281: this study. It is unclear whether much larger wind speeds
282: \citep[$\sim$ several km/s; e.g.,][]{C&S05,Dobbs-Dixon2007} can be
283: realized in a turbulent, rotationally-balanced flow on hot
284: Jupiters. We have found that large wind speeds of a few km/s are not
285: possible in our specific setup since they either lead to atmospheric
286: holes in the initially balanced state or to a rapid blow up of the
287: flow field \citep{Cho2006}.
288:
289:
290:
291: It is worth noting that in these adiabatic models the strength of the
292: imposed day-night temperature is decoupled from the assumed global
293: mean wind strength: it depends solely on $\eta$. {Winds of
294: greater strength} may cause a decreased day-night temperature
295: difference by, for example, more efficiently redistributing the heat
296: between the two sides. This means that the trend of increasing phase
297: curve amplitude with $\bar{U}$ shown here may not be self-consistent.
298: Coupled, diabatic circulation models would address this issue more
299: satisfactorily.
300:
301: The next feature that emerges in {cases} with higher $\bar{U}$ is the
302: presence of {oscillations} in the phase curves. Whenever the cold
303: vortices are visible on the planetary disk, they create a dip in the
304: disk-integrated thermal flux. These dips can weaken the day side flux
305: peak or strengthen the night side flux trough, when the vortices are
306: predominantly at the sub- or anti-stellar point, respectively. The
307: vortices and other weather features also produce small {perturbations} seen in
308: the phase curves (most apparent for the $\bar{U}=800$ \ms~model), when
309: apparent at some intermediate phase between peak and trough.
310:
311: The smallest of the identifiable {perturbations} is about 30\degrees~wide in
312: orbital phase. It is not trivial to relate the phase size of this
313: {perturbation} back to a physical scale in the planet's {atmosphere}, since
314: there exists some degeneracy in interpretation between the size and
315: brightness of any atmospheric feature. {F}or instance, a small and
316: extremely bright feature that {is present during the entire passage
317: across the visible disk could result} in a phase curve {perturbation} of
318: almost 180\degrees. If the feature were dimmer, perhaps it would only
319: significantly contribute to the observed flux when it was near the
320: ``sub-observer'' point, and the size of the resulting {perturbation} would
321: therefore be limited in phase. However, a diffuse feature of only
322: slightly elevated brightness could produce a similar {perturbation}. Thus the
323: size and brightness of features can counterbalance each other, making
324: it difficult to place constraints on either property. This
325: exemplifies the ambiguity present in any interpretation of orbital
326: phase curve data---each point in the orbit is a measure of the
327: disk-integrated emission so that, in principle, it is possible to
328: construct a variety of temperature structures that reproduce a given
329: orbital variation.
330:
331: In our model {runs}, when the circumpolar vortices are close to being
332: aligned with the substellar point, but are some degrees of longitude
333: displaced, the peak of the emitted phase curve is somewhat offset from
334: the phase at which the substellar point faces the observer. The
335: resulting apparent phase shift could be interpreted as a hot spot
336: advected away from the substellar point, while in this example it is
337: actually the result of the disk-integrated combination of emission
338: from the hot substellar region and the asymmetry from the cold
339: pattern. This point is worth emphasizing, as it shows that there can
340: be degeneracy between models in the interpretation of even a single
341: orbit's phase curve. Examining Figure~\ref{fig:u_values} we find
342: phase offsets of up to $\pm$40\degrees~(with extreme offsets of orbit
343: 8 at 85\degrees~and orbit 3 at 100\degrees) for the $\bar{U}=800$
344: \ms~model {run}. The $\bar{U}=400$ and 200 \ms~model {runs} typically
345: produce phase shifts of $\pm$30\degrees~and $\pm$10\degrees,
346: respectively.
347:
348:
349: \subsection{The Influence of Imposed Radiative Forcing}
350:
351: Next we consider the effect of increasing the amount of radiative
352: forcing imposed on the model atmospheres, controlled by the parameter
353: $\eta$. Figure~\ref{fig:eta_values} shows phase curves for the full
354: range of $\bar{U}$ values for {simulations} with $\eta = 0.10$ and
355: 0.20. As expected, the increased radiative forcing causes the regular
356: day-night variation to dominate over the features produced by the
357: variable weather structure. The day-night variation is completely
358: dominant for the weakest wind models; for $\eta=0.20$, a wind speed of
359: 800 \ms~is required to produce deviations of more than a few percent
360: away from the peak-trough pattern. Therefore, increasing the value of
361: $\eta$ results in phase curves with larger amplitudes and a more
362: regular peak-trough pattern. The cold vortices are too weak, compared
363: to the day-night gradient, to produce the same small features seen in
364: the phase curves of the $\eta=0.05$ models, and they now only work to
365: change the amplitude of the variation from one orbit to the next, or
366: to produce small shifts in the phase of the peak flux. The amount of
367: phase offset is greatly reduced from the $\eta=0.05$ model. For
368: $\eta=0.10$, $\bar{U}=800$ \ms~leads to $\pm$20\degrees ~offsets,
369: $\bar{U}=400$ \ms~leads to $\pm$10\degrees ~offsets and there are no
370: shifts for $\bar{U}=200$ \ms~or less. For $\eta=0.20$, the only model
371: whose curve shows a shift is $\bar{U}=800$ \ms, with an offset of
372: $\pm$10\degrees.
373:
374:
375: \subsection{The Role of Inclination}
376:
377: In addition to sampling the $(\eta,\bar{U})$ parameter space itself,
378: in this work we also consider the role of orbital inclination on model
379: phase curves. A challenge in interpreting data from non-eclipsing
380: systems is that the unknown inclination is partially degenerate with
381: the amount of heat redistributed to the night side
382: \citep[e.g.,][]{Harrington2006}. In Figure~\ref{fig:incl_values} we
383: show the effect of varying the assumed inclination in our calculations
384: of the phase curves for model {runs} with $\eta=0.05$ and 0.10, and
385: $\bar{U}=800$ \ms. We choose to plot the highest wind {runs} (and
386: likewise omit the $\eta=0.20$ {run}) {since} the role of inclination
387: on phase curves from atmospheres dominated by the day-night
388: temperature contrast is simple: as the inclination is reduced (i.e.,
389: as the orbit is viewed from an increasingly polar orientation), there
390: is a trivial decrease in the amplitude of orbital phase variation.
391: (At the $i=0$\degrees~extreme there is no variation since equal parts
392: of the day and night sides are constantly in view.) We see this
393: reduction in amplitude in our model curves as well, especially for the
394: $\eta=0.10$ model, where the thermal emission is dominated by the
395: day-night temperature contrast.
396:
397: Another effect of decreasing the inclination is the emergence of more
398: small features in the model phase curves. Since the cold vortices
399: revolve around the poles of the planet, as the orbital inclination is
400: decreased, they are located closer to the center of the observed
401: planetary disk, and therefore are given more weighting in the
402: disk-integrated emitted flux. For a visual explanation of this
403: effect, see Figure~\ref{fig:images}, where we have plotted temperature
404: snapshots from one orbital rotation of the $\eta=0.10$, $\bar{U}=800$
405: \ms~model {run}, shown at $i =$ 90, 60, and 30\degrees. Here we show
406: the $\eta=0.10$ model {run} because it has strong enough radiative
407: forcing that the day-night gradient is clearly seen, while
408: $\bar{U}=800$ \ms~creates strong enough vortices that they are
409: likewise easily visible in the global temperature maps. The top,
410: middle, and bottom rows of Figure~\ref{fig:images} correspond to the
411: top, middle, and bottom curves of Figure~\ref{fig:incl_values}b,
412: respectively. We see that at $i=30$\degrees, the cold vortices are
413: more centrally located on the apparent disk and the day-night
414: temperature difference plays a much less significant role in producing
415: variations of the disk-integrated flux. The decrease in day-night
416: variation, combined with the more prominent role of cold vortices,
417: results in the increased presence of small features in the model phase
418: curve.
419:
420:
421: \subsection{Wavelength-Dependent Phase Curves}
422:
423: Throughout our discussion so far we have been analyzing bolometric
424: phase curves---\emph{i.e.}, light curves calculated under the simple
425: assumption that the emission from any area on the planetary disk is
426: $\propto T^4_{local}$. {Here} we also compute the light curves that
427: would be measured in each \emph{Spitzer} instrumental band to
428: determine the wavelength dependence of these model curves. We
429: calculate two values for the flux measured by each \emph{Spitzer}
430: band: one uses a blackbody for the emitted spectrum from each element
431: area on the planetary disk, and the other one interpolates from a grid
432: of model atmosphere spectra from \citet{Seager2005}. This is identical
433: to the procedure adopted in \citet{Rauscher2007,Rauscher2007b}.
434: Figure~\ref{fig:bbody} shows the wavelength-dependent phase curves for
435: the $\eta=0.05$, $\bar{U}=800$ \ms~{simulation}, assuming local
436: blackbody emission. The amplitude of variation decreases with
437: increasing wavelength. {We have verified that, in the range of
438: temperatures covered by our model {runs} (see Table~1), this is simply
439: a consequence of the Planck function varying more with temperature at
440: shorter wavelengths than it does at longer wavelengths.}
441:
442: This same general behavior is seen in Figure~\ref{fig:smodel}, in
443: which we plot model phase curves for the same \emph{Spitzer} bands,
444: but using the detailed spectral models for local emission. Just as
445: with a blackbody, the modeled spectra are more temperature-sensitive
446: at shorter wavelengths. The detailed wavelength-temperature
447: dependence of these spectral models differs from that of a blackbody,
448: however, largely because of the presence of absorption bands,
449: principally those due to water in the atmosphere. Although the
450: presence of water {vapor spectral signatures} in hot Jupiter
451: atmospheres is currently debated
452: \citep{Barman2007,Burrows2007,Fortney2007b,Grillmair2007,Richardson2007,Tinetti2007},
453: the main point of Figures~\ref{fig:bbody} and~\ref{fig:smodel} remains
454: the same: for an atmosphere that spans the range of temperatures
455: expected on hot Jupiters, phase curves at shorter wavelengths may be
456: generally more sensitive to temperature differentials.
457:
458: One needs to interpret this trend cautiously, however. It is only one
459: of several effects that contribute to the detailed emission properties
460: from such an atmosphere. For example, our modeling strategy based on
461: a vertically-integrated formulation of the atmospheric flow does not
462: capture well the possibility that shorter wavelengths may probe deeper
463: into a planet's atmosphere \citep[e.g.,][]{Seager2005}, where the
464: circulation regime and the horizontal temperature field could in
465: principle differ from higher up. {While the model phase curves shown
466: in Figs.~\ref{fig:bbody} and~\ref{fig:smodel} emphasize the role of
467: horizontal temperature variations, additional wavelength-dependent
468: effects are expected from variations of temperature with atmospheric
469: depth. {The vertical profile therefore has an effect on the
470: wavelength dependence of phase curves which is unspecified in our
471: models and the behavior shown here can thus only be associated with
472: effect of horizontal temperature variations.} Three-dimensional
473: radiation-hydrodynamics models will be required to address this issue
474: reliably.}
475:
476:
477: \section{Implications for current and future observations of hot Jupiters}
478:
479: The equivalent-barotropic circulation models of \citet{Cho2003,
480: Cho2006} indicate the possible existence of large scale weather
481: structures {on hot Jupiters. These structures could produce
482: observable signatures in the thermal phase curves of hot Jupiters, in
483: the absence of obscuring elements such as clouds or haze \citep[see,
484: e.g.][]{Pont07}.} We have surveyed the parameter space of these
485: models and find a diverse set of resulting phase curves. This
486: diversity may be relevant in the interpretation of observations from
487: various systems, especially as we are beginning to find that the label
488: ``hot Jupiter'' may in fact encompass a range of planetary types
489: \citep[e.g.,][]{Burrows2007,Fortney2007a,Hansen2007b}.
490:
491: Even within the limited set of observed infrared phase curves, there
492: is already evidence for significant diversity. The first published
493: exoplanet phase curve was a set of five \emph{Spitzer} 24$\micron$
494: observations of $\upsilon$ Andromeda b during one of its 4.617-day
495: orbits \citep{Harrington2006}. These authors have found strong
496: variation, indicative of little to no heat redistribution, and a small
497: phase offset consistent with zero. \citet{Cowan2007} used the IRAC
498: instrument on \emph{Spitzer} to observe HD 209458, HD 179949, and 51
499: Peg. Their 8$\micron$ data placed an upper limit on the efficiency of
500: heat redistribution on HD 179949b, while both HD 209458b and 51 Peg b
501: were given lower limits. There was no obvious phase shift evident in
502: the data, and these authors chose to fit their data with zero
503: phase-offset models. \citet{Hansen2007a} report 24$\micron$ phase
504: curves, with observations at five epochs for $\upsilon$
505: And\footnote{These are the same data as reported in
506: \citet{Harrington2006}.}, $\tau$ Bootes, and 51 Peg, and at three
507: epochs for the fainter systems HD 179949 and HD 75289. Like for 51
508: Peg, and in contrast to $\upsilon$ And, there is some evidence for
509: heat redistribution on $\tau$ Boo. Particularly interesting is the
510: detection of an approximately -80\degrees~shift in the phase curve of
511: $\tau$ Boo. This is a phase lag, meaning that the minimum of the
512: curve occurs after transit. {Recent MOST photometry of the
513: system indicates that this phase variation may be due to the presence
514: of an active region on the star \citep{Walker2008}. In such a case, it
515: would be more difficult to disentangle the variation in emission due
516: to any longitudinal temperature differences on the planet from
517: additional variations due to the presence of magnetic structures.
518: Nevertheless, the possibility of both positive and negative phase
519: offsets, as found in the phase curves presented here, and as opposed
520: to systematically positive ones according to other dynamical models
521: \citep[e.g.,][]{Showman2002,C&S05,Fortney2006,Langton2007} may provide
522: important additional constraints on the regime of circulation present
523: in these atmospheres.}
524:
525: The most detailed observational results published to date are those of
526: \citet{Knutson2007}. While the interpretation of other phase
527: observations has been limited by the necessity to fit model curves to
528: a few noisy points spread throughout an orbital period,
529: \citet{Knutson2007} observed HD 189733b continuously for $\sim$33
530: hours, just over half an orbital period. These authors report an
531: amplitude of variation indicating heat redistribution, although at an
532: incomplete level. The flux maximum occurs 2.3 hours before secondary
533: eclipse, while the minimum is 6.7 hours after transit (=primary
534: eclipse), which corresponds to two separate phase shifts: a
535: +16\degrees~phase shift for the flux maximum, and a
536: $-$47\degrees~shift for the minimum. This means that the
537: data cannot be explained with a simple model that consists solely of a
538: day-night temperature gradient, nor a model whose only atmospheric
539: feature is a hot spot shifted away from the substellar point.
540:
541: The continuous nature of the \citet{Knutson2007} data allows for the
542: resolution of small features in the phase curve, which was not
543: possible with other published measurements. By postulating a steady
544: emission pattern over the observation time (in the frame rotating with
545: the synchronized planet), these authors reconstruct a longitudinal map
546: of the brightness across the planetary disk. The disk-integrated flux
547: reaches its minimum and maximum values in the same longitudinal
548: hemisphere and a noticeable {small scale} feature
549: $\sim$20\degrees~wide in phase is seen directly after transit.
550:
551: Interestingly, we find that similar {small scale perturbations}
552: can arise in the thermal phase curves of our equivalent-barotropic
553: models. In particular, the $\eta=0.05$, $\bar{U}=800$ \ms~model
554: (Figs.~\ref{fig:u_values} and~\ref{fig:bbody}) shows such features and
555: produces an overall amplitude of phase curve variations that is
556: qualitatively consistent with the \citet{Knutson2007} data.
557: {The min-to-max variation in the data is $\sim$50\%, a value
558: consistent with the approximate 20--80\% range of single-orbit flux
559: variation seen in this model.} Since {the features of this
560: model phase curve} rely on the presence of large-scale moving
561: atmospheric features, a consequence is that the shape of the phase
562: curve would be expected to vary from one orbit to the next according
563: to our models. This is in fact one of the main qualitative
564: predictions of high-resolution, equivalent-barotropic models for hot
565: Jupiter atmospheres: with strong weather patterns, the temporal
566: variability of the atmospheric temperature field should result in
567: possibly detectable variations for repeated observations of a single
568: system \citep[see also][]{Menou2003,Rauscher2007}.
569:
570: {This is particularly relevant in light of the recent
571: 24$\micron$ observation of HD 189733b \citep{Knutson2008}, in which
572: another continuous measurement of the system over a similar half-orbit
573: was performed. Although the circulation regime (e.g., global average
574: wind speeds) could be different in the atmospheric layers probed at 8
575: and 24$\micron$ \citep[e.g.,][]{Seager2005}, the phase curve in
576: \citet{Knutson2008} is similar in amplitude and shape to the
577: 8$\micron$ observation. Thus the global temperature structure seems
578: to have little vertical variation between the two corresponding
579: ``photospheres.'' Perhaps more importantly, as it pertains to our
580: models, there is little difference in the observed emission pattern
581: between these two orbits. This disfavors any atmospheric model for
582: this planet that predicts significant orbit-to-orbit variations in
583: temperature structure, and specifically the $\eta=0.05$, $\bar{U}=800$
584: \ms~model previously deemed qualitatively consistent with the
585: 8$\micron$ observations. While the small feature seen after transit
586: in the 8$\micron$ data is still indicative of some complexity in the
587: global temperature structure, and the 24$\micron$ data are not of
588: sufficient precision to observe a comparable feature
589: \citep{Knutson2008}, a comparison between these two phase curves
590: points toward a mostly steady global circulation pattern on HD
591: 189733b.}
592:
593: The interpretation of any orbital phase curve data is subtle.
594: Measurements that consist of several discrete observations throughout
595: one orbit may constrain the level of heat redistribution around the
596: planet, but there are degeneracies even with the simplest models when
597: trying to fit the data \citep{Hansen2007a}. This type of observation
598: may not easily constrain the level of atmospheric complexity inherent
599: in nonlinear circulation models---even for models as simple as the
600: ones used here. The interpretation of observational data
601: {sets} like {those} of \citet{Knutson2007,Knutson2008}
602: becomes ambiguous if variable weather systems are present in hot
603: Jupiter atmospheres, since they can typically vary on orbital
604: timescales. The assumption of a steady atmospheric field (in the
605: planet's rotating frame) is necessary for the construction of a
606: longitudinal temperature (or brightness) map, so that the resulting
607: inferred properties could become inaccurate if moving features exist.
608:
609: Repeated observations over multiple orbits help to constrain the
610: steadiness of the global temperature field, as opposed to one which
611: changes from orbit to orbit, like the high $\bar{U}$ models presented
612: here. The permanence of the day side field could also be constrained
613: with repeated secondary eclipse measurements \citep{Rauscher2007},
614: which are less observationally expensive than phase curve measurements
615: but lack information on the global day-night pattern. In the future,
616: these two observational techniques should usefully complement each
617: other when applied to eclipsing systems.
618:
619: Finally, while we have focused our study on thermal phase curves, it
620: is worth noting that the upcoming \emph{Kepler} mission could
621: potentially also constrain the presence of large-scale weather
622: features in (very) hot Jupiter atmospheres. Indeed, signatures
623: similar to those in thermal phase curves may emerge in reflected light
624: phase curves, if dynamical weather structures such as circumpolar
625: vortices act as distinct thermodynamic regions of the atmospheric
626: flow, with specific cloud and albedo properties. These issues would
627: be best addressed in the future with improved atmospheric circulation
628: models, coupled with radiation transfer calculations.
629:
630: \section{Conclusion}
631:
632: The interpretation of hot Jupiter phase curve data is subtle,
633: especially if variable weather features are present in their
634: atmospheres. We have explored the parameter space of the adiabatic,
635: equivalent-barotropic models of \citet{Cho2003,Cho2006} to investigate
636: the possible effects that large-scale features would have on hot
637: Jupiter thermal phase curves. Despite their relative simplicity, we
638: find that these models can produce phase curves with shapes that
639: change from one orbit to the next, have apparent phase shifts, and
640: contain {small transient perturbations}. The detection of such
641: observational signatures would probe the regime of circulation present
642: in these atmospheres. Even the absence of such features would
643: constrain the atmospheric dynamics regime.
644:
645: {We thank the anonymous referee for comments that helped us
646: clarify the manuscript.} This work was supported by NASA contract
647: NNG06GF55G.
648:
649: \begin{thebibliography}
650:
651: \bibitem[Barman(2007)]{Barman2007} Barman, T.\ 2007, \apjl, 661, L191
652:
653:
654: \bibitem[Burrows et al.(2007)]{Burrows2007} Burrows, A., Budaj, J., \&
655: Hubeny, I.\ 2007, ArXiv e-prints, arXiv:0709.4080
656:
657: \bibitem[Charbonneau et al.(2005)]{Charbonneau2005} Charbonneau, D., et al.\ 2005, \apj, 626, 523
658:
659: \bibitem[Cho et al.(2003)]{Cho2003} Cho, J.~Y.-K., Menou, K., Hansen, B.~M.~S., \& Seager, S.\ 2003, \apjl, 587, L117
660:
661: \bibitem[Cho et al.(2008)]{Cho2006} Cho, J.~Y.-K., Menou, K., Hansen,
662: B., \& Seager, S.\ 2008, ApJ accepted, astro-ph/0607338
663:
664: \bibitem[Cho \& Polvani(1996)]{Cho1996} Cho, J.~Y.-K., \& Polvani, L.~M.\ 1996, Physics of Fluids, 8, 1531
665:
666: \bibitem[Cooper \& Showman(2005)]{C&S05} Cooper, C.~S., \& Showman, A.~P.\ 2005, \apjl, 629, L45
667:
668: \bibitem[Cooper \& Showman(2006)]{Cooper2006} Cooper, C.~S., \& Showman, A.~P.\ 2006, \apj, 649, 1048
669:
670: \bibitem[Cowan et al.(2007)]{Cowan2007} Cowan, N.~B., Agol, E., \& Charbonneau, D.\ 2007, \mnras, 552
671:
672: \bibitem[Deming et al.(2007a)]{Deming2007b} Deming, D., Agol, E.,
673: Charbonneau, D., Cowan, N., Knutson, H., \& Marengo, M.\ 2007, ArXiv
674: e-prints, arXiv:0710.4145
675:
676: \bibitem[Deming et al.(2007b)]{Deming2007} Deming, D., Harrington, J., Laughlin, G., Seager, S., Navarro, S.~B., Bowman, W.~C., \& Horning, K.\ 2007, \apjl, 667, L199
677:
678: \bibitem[Deming et al.(2006)]{Deming2006} Deming, D., Harrington, J., Seager, S., \& Richardson, L.~J.\ 2006, \apj, 644, 560
679:
680: \bibitem[Deming et al.(2005)]{Deming2005} Deming, D., Seager, S., Richardson, L.~J., \& Harrington, J.\ 2005, \nat, 434, 740
681:
682: \bibitem[Dobbs-Dixon \& Lin(2007)]{Dobbs-Dixon2007} Dobbs-Dixon, I.,
683: \& Lin, D.~N.~C.\ 2007, ArXiv e-prints, arXiv:0704.3269
684:
685: \bibitem[Fortney et al.(2006)]{Fortney2006} Fortney, J.~J., Cooper,
686: C.~S., Showman, A.~P., Marley, M.~S., \& Freedman, R.~S.\ 2006, \apj,
687: 652, 746
688:
689: \bibitem[Fortney et al.(2007)]{Fortney2007a} Fortney, J.~J., Lodders,
690: K., Marley, M.~S., \& Freedman, R.~S.\ 2007, ArXiv e-prints,
691: arXiv:0710.2558
692:
693: \bibitem[Fortney \& Marley(2007)]{Fortney2007b} Fortney, J.~J., \& Marley, M.~S.\ 2007, \apjl, 666, L45
694:
695: \bibitem[Grillmair et al.(2007)]{Grillmair2007} Grillmair, C.~J., Charbonneau, D., Burrows, A., Armus, L., Stauffer, J., Meadows, V., Van Cleve, J., \& Levine, D.\ 2007, \apjl, 658, L115
696:
697: \bibitem[Hansen et al.(2008)]{Hansen2007a} Hansen, B. et al. 2008, to
698: be submitted to ApJ
699:
700: \bibitem[Hansen \& Barman(2007)]{Hansen2007b} Hansen, B.~M.~S., \&
701: Barman, T.\ 2007, ArXiv e-prints, arXiv:0706.3052
702:
703: \bibitem[Harrington et al.(2006)]{Harrington2006} Harrington, J. et al. 2006, Science, 314, 623
704:
705: \bibitem[Harrington et al.(2007)]{Harrington2007} Harrington, J., Luszcz, S., Seager, S., Deming, D., \& Richardson, L.~J.\ 2007, \nat, 447, 691
706:
707: \bibitem[Knutson et al.(2007)]{Knutson2007} Knutson, H.~A., et al.\ 2007, \nat, 447, 183
708:
709: \bibitem[Knutson et al.(2008)]{Knutson2008} Knutson, H.~A., et al.\
710: 2008, ArXiv e-prints, 802, arXiv:0802.1705
711:
712: \bibitem[Langton \& Laughlin(2007)]{Langton2007} Langton, J., \& Laughlin, G.\ 2007, \apjl, 657, L113
713:
714: \bibitem[Menou et al.(2003)]{Menou2003} Menou, K., Cho, J.~Y.-K.,
715: Seager, S. \& Hansen, B.~M.~S.\ 2003, \apjl, 587, L113
716:
717: \bibitem[Pont et al.(2007)]{Pont07} Pont, F., Knutson, H., Gilliland,
718: R. L., Moutou, C. \& Charbonneau, D.\ 2007, MNRAS, in press, arXiv
719: 0712.1374
720:
721:
722: \bibitem[Rauscher et al.(2007a)]{Rauscher2007} Rauscher, E., Menou, K., Cho, J.~Y.-K., Seager, S., \& Hansen, B.~M.~S.\ 2007a, \apjl, 662, L115
723:
724: \bibitem[Rauscher et al.(2007b)]{Rauscher2007b} Rauscher, E., Menou, K., Seager, S., Deming, D., Cho, J.~Y.-K., \& Hansen, B.~M.~S.\ 2007b, \apj, 664, 1199
725:
726: \bibitem[Redfield et al.(2007)]{Redfield2007} Redfield, S., Endl, M., Cochran, W.~D., \& Koesterke, L.\ 2007, ArXiv e-prints, 712, arXiv:0712.0761
727:
728: \bibitem[Richardson et al.(2007)]{Richardson2007} Richardson, L.~J., Deming, D., Horning, K., Seager, S., \& Harrington, J.\ 2007, \nat, 445, 892
729:
730: \bibitem[Salby(1989)]{Salby1989} Salby, M. L. 1989, Tellus, 41A, 48
731:
732: \bibitem[Seager et al.(2005)]{Seager2005} Seager, S., Richardson,
733: L.~J., Hansen, B.~M.~S., Menou, K., Cho, J.~Y.-K., \& Deming, D.\
734: 2005, \apj, 632, 1122
735:
736: \bibitem[Showman \& Guillot(2002)]{Showman2002} Showman, A.~P., \& Guillot, T.\ 2002, \aap, 385, 166
737:
738: \bibitem[Showman et al.(2007)]{Showman2007} Showman, A.~P., Menou, K.,
739: \& Cho, J. Y-K.\ 2007, ArXiv e-prints, arXiv:0710.2930
740:
741: \bibitem[Tinetti et al.(2007)]{Tinetti2007} Tinetti, G., et al.\ 2007, \nat, 448, 169
742:
743: \bibitem[Walker et al.(2008)]{Walker2008} Walker, G.~A.~H., et
744: al.\ 2008, ArXiv e-prints, 802, arXiv:0802.2732
745:
746: \bibitem[Williams et al.(2006)]{Williams2006} Williams, P.~K.~G., Charbonneau, D., Cooper, C.~S., Showman, A.~P., \& Fortney, J.~J.\ 2006, \apj, 649, 1020
747:
748: \end{thebibliography}
749:
750: \clearpage
751:
752: \begin{deluxetable}{cccc}
753: \tablewidth{0pt}
754: \tablecaption{Equivalent-barotropic models under consideration}
755: \tablehead{
756: \colhead{$\bar{U}$} & \multicolumn{3}{c}{Min-Max Layer Temperature (K)} \\
757: \colhead{(\ms)} & \colhead{$\eta=0.05$} & \colhead{$\eta=0.10$} & \colhead{$\eta=0.20$}
758: }
759: \startdata
760: 0\tablenotemark{a} & 1140-1260 & 1080-1320 & 960-1440 \\
761: 100 & 1148-1274 & 1085-1335 & 960-1452 \\
762: 200 & 1113-1284 & 1081-1338 & 950-1458 \\
763: 400 & 998-1313 & 948-1351 & 938-1465 \\
764: 800 & 665-1373 & 713-1415 & 722-1504 \\
765: \enddata
766:
767: \tablenotetext{a}{{Temperatures in the $\bar{U}=0$ models isolate the
768: contribution from imposed day-night radiative forcing. The bottom
769: surface of the model layer is bent by an amount corresponding to a
770: day-night temperature range of $(1\pm \eta)\times 1200$K
771: \citep[see][for details]{Cho2006}. Trivial $\bar{U}=0$ models are
772: shown for reference only.}}
773: \end{deluxetable}
774:
775: \clearpage
776:
777: \begin{figure}[ht]
778: \begin{center}
779: \includegraphics[width=\textwidth]{f1.eps}
780: \caption{Bolometric orbital phase curves for equivalent-barotropic
781: model {simulations} with $\eta=0.05$ and $\bar{U} =$ 800
782: (\emph{dash-dotted}), 400 (\emph{dotted}), 200 (\emph{dashed}),
783: and 100 (\emph{solid}) \ms. Each curve is normalized to its
784: absolute minimum, emphasizing the relative variation for each
785: model. The phase curves become increasingly complex in models
786: with stronger winds because the temperature field is increasingly
787: dominated by variable weather features. Observable signatures of
788: atmospheric features include small {perturbations} in the phase curve,
789: offsets in the phase of min./max. flux, and variations in the peak
790: to trough amplitude, from one orbit to the
791: next.}\label{fig:u_values}
792: \end{center}
793: \end{figure}
794:
795: \begin{figure}[ht]
796: \begin{center}
797: \includegraphics[width=\textwidth]{f2.eps}
798: \caption{Same as Figure~\ref{fig:u_values} (from top to bottom:
799: $\bar{U}=800, 400, 200, 100$ \ms), but for {simulations} with
800: $\eta=0.10$ (\emph{a}) and $\eta=0.20$ (\emph{b}). As the amount
801: of radiative forcing is increased, the imposed day-night
802: temperature contrast becomes the dominant component of the
803: planet's temperature field, rendering the phase curves more
804: regular in their variation. Nevertheless, large-scale weather
805: features continue to produce deviations away from a strictly
806: regular peak-trough pattern, even for the strong radiative forcing
807: case.} \label{fig:eta_values}
808: \end{center}
809: \end{figure}
810:
811: \begin{figure}[ht]
812: \begin{center}
813: \includegraphics[width=\textwidth]{f3.eps}
814: \caption{\small Orbital phase curves for $\bar{U}=800$
815: \ms~equivalent-barotropic models with $\eta=0.05$ (\emph{a}) and
816: $\eta=0.10$ (\emph{b}). Three cases, with orbital inclinations $i
817: =$ 90\degrees~(edge-on; \emph{dash-dot}),
818: 60\degrees~(\emph{dotted}), and 30\degrees~(\emph{solid}), are
819: shown in each panel. As the system's inclination is tilted away
820: from edge-on {orientation}, the amount of phase variation due to
821: the imposed day-night temperature contrast is decreased, while the
822: signatures of variable weather features become increasingly more
823: apparent in the phase curves. The features that contribute most
824: strongly to the wiggly shapes of these model phase curves are the
825: cold circumpolar vortices, since the change in inclination brings
826: them more into view.} \label{fig:incl_values}
827: \end{center}
828: \end{figure}
829:
830: \begin{figure}[ht]
831: \begin{center}
832: \includegraphics[width=\textwidth]{f4.eps}
833: \caption{Snapshots of the apparent temperature field in the model
834: with $\eta=0.10$, $\bar{U}=800$ \ms,~as it rotates through one
835: orbit. Temperatures range from 700 to 1400K. From left to right,
836: the central longitudes are: 0\degrees~(substellar; phase $-0.5$),
837: 90\degrees, 180\degrees~(antistellar; phase $0$), 270\degrees, and
838: 360\degrees. The maps in the top row are shown at an inclination
839: of 90\degrees~(edge-on), the middle ones at 60\degrees~and the
840: bottom ones at 30\degrees. As the inclination of the system is
841: titled away from edge-on {orientation}, the cold
842: circumpolar vortices assume a more central location on the
843: apparent planetary disk, so that they affect more strongly the
844: disk-integrated thermal emission. At the same time, more equal
845: areas of the hot day side and cool night side are visible, which
846: decreases the amplitude of the regular peak-trough pattern in the
847: phase curves. \emph{[See the electronic edition of the Journal for
848: a color version of this figure.]}}\label{fig:images}
849: \end{center}
850: \end{figure}
851:
852: \begin{figure}[ht]
853: \begin{center}
854: \includegraphics[width=\textwidth]{f5.eps}
855: \caption{Orbital phase curves as they would be observed in various
856: \emph{Spitzer} instrumental bands, for the $\eta=0.05$,
857: $\bar{U}=800$ \ms~equivalent-barotropic model, assuming simple
858: blackbody emission. From top to bottom the curves are: bolometric
859: (\emph{black}), 3.6\micron~(\emph{purple}),
860: 4.5\micron~(\emph{blue}), 6\micron~(\emph{green}),
861: 8\micron~(\emph{yellow}), 16\micron~(\emph{orange}), and
862: 24\micron~(\emph{red}). The shorter wavelengths are more variable
863: for the range of temperatures covered by this model. \emph{[See
864: the electronic edition of the Journal for a color version of
865: this figure.]}}
866: \label{fig:bbody}
867: \end{center}
868: \end{figure}
869:
870: \begin{figure}[ht]
871: \begin{center}
872: \includegraphics[width=\textwidth]{f6.eps}
873: \caption{Orbital phase curves as they would be observed in various
874: \emph{Spitzer} instrumental bands, for the $\eta=0.05$,
875: $\bar{U}=800$ \ms~equivalent-barotropic model, using detailed
876: atmospheric spectral models for emission. From top to bottom the
877: curves are: 3.6\micron~(\emph{purple}), bolometric (\emph{black}),
878: followed by overlapping 4.5\micron~(\emph{blue}),
879: 6\micron~(\emph{green}), and 8\micron~(\emph{yellow}), and finally
880: 16\micron~(\emph{orange}) and 24\micron~(\emph{red}). The results
881: for any given band are strongly affected by the absorption
882: properties of constituents included in the atmospheric spectral
883: model (especially water), but shorter wavelengths are generally
884: more variable for the range of temperatures covered by our
885: circulation model. \emph{[See the electronic edition of the
886: Journal for a color version of this figure.]}}
887: \label{fig:smodel}
888: \end{center}
889: \end{figure}
890:
891:
892:
893: \end{document}