Dry and Moist Atmospheric Circulation with Uniform Sea-Surface Temperature
DDry and Moist Atmospheric Circulation with UniformSea-Surface Temperature
D.L. Suhas ∗ , Jai Sukhatme , and Nili Harnik Centre for Atmospheric and Oceanic Sciences, Indian Institute of Science, Bangalore,560012, India Divecha Centre for Climate Change, Indian Institute of Science, Bangalore, 560012,India Department of Geosciences, Tel Aviv University, Tel Aviv 69978, Israel
Abstract
The steady and transient response of ”dynamically” dry and moist atmospheres to uni-form sea-surface temperature (SST) is studied. Specifically, the latent heat ( L v ) of watervapor is varied, so that for small L v , water substance is essentially a passive tracer froma dynamical point of view. Despite the lack of SST gradients, a general circulation withHadley and Ferrel cells is observed for relatively stronger moist coupling. Organized precip-itation patterns via equatorial waves appear to play a significant role in tropical ascent, andalong with the equatorial deformation radius, the Hadley cell width increases with couplingstrength. An abrupt switch to a much shallower tropical cell is noted when the system be-comes completely passive. In all cases, the Hadley cell is thermally indirect and is influencedby eddy fluxes which are strong in the upper and lower troposphere. Moist static energyis transported equatorward in the tropics and a larger amount is directed poleward in themidlatitudes. As a whole, there is an almost invariant poleward transport of moist staticenergy for relatively strong coupling of water substance. Transient extratropical activityis seen in the form of intense warm-core vortices for strong coupling, and these systemsbecome weaker and smaller as L v decreases. The drift of these moist vortices results inthe observed poleward energy transport in the midlatitudes. In the tropics, intraseasonalvariability is dominant and systematically shifts to longer time periods with stronger cou-pling. In fact, large-scale, low-frequency Kelvin waves and MJO-like modes disappear aswater vapor becomes passive in nature. Finally, extreme rainfall events associated withcyclonic storms vanish as water vapor becomes dynamically inactive, however, moderateprecipitation events increase leading to higher total precipitation for weaker coupling of wa-ter substance. Tropospheric heating due to a saturation of the outgoing longwave radiationresults in an increase in the stability of the atmosphere for strong coupling, and provides aplausible physical mechanism for interpreting the behavior of precipitation. Key Words : Water vapor, Hadley cell, Time-dependent flow, Uniform surface temperature
The explicit response of the atmosphere to uniform lower boundary forcing has proven to bea useful idealization in the climate modeling hierarchy (Maher et al., 2019). Simulations of ∗ Corresponding author: D.L. Suhas, [email protected] a r X i v : . [ phy s i c s . a o - ph ] J a n his kind have yielded insight into possible routes to an organized general circulation (Sumi,1992; Kirtman and Schneider, 2000; Barsugli et al., 2005; Horinouchi, 2012) and the formationof tropical storm-like vortices and their sensitivity to environmental parameters on a f -plane(Held and Zhao, 2008; Khairoutdinov and Emanuel, 2013; Zhou et al., 2014; Ram´ırez-Reyes andYang, 2020) as well as on the globe (Shi and Bretherton, 2014; Merlis et al., 2016; Chavas andReed, 2019). A recent review of these efforts can be found in Merlis and Held (2019). Moisttropical transients, especially the Madden-Julian Oscillation (MJO) have also been studied inuniform sea-surface temperaure (SST) experiments (Grabowski, 2003; Grabowski and Moncrieff,2004). Further, constant SST experiments have allowed for an examination of convective self-aggregation in a variety of cloud resolving and parameterized general circulation models (see,for example, Wing et al., 2017). Another motivation comes from paleoclimate estimates thatsuggest a much weaker meridional gradient in SST during the early Pliocene (Brierley et al.,2009), and it is of interest to understand how such boundary conditions can affect the atmo-spheric circulation and transient phenomena, especially the frequency and strength of tropicaland extratropical cyclones (Fedorov et al., 2010, 2019).Taking a steady state view, aquaplanet experiments with constant SST showed the developmentof an organized tropical atmospheric circulation due to the interaction of moist convection withrotation (Sumi, 1992; Kirtman and Schneider, 2000). This took the form of a well definedequatorial convergence zone, along with tropical easterlies and subtropical westerlies (Kirtmanand Schneider, 2000). Further explorations along these lines showed that the nature of thetropical convergence zone was dependent on the magnitude of the uniform SST. In fact, single,double, symmetric and asymmetric convergence zones were seen to form under differing SSTstrengths (Barsugli et al., 2005). Interestingly, on varying the threshold relative humidity fortriggering the deep moist convection, a uniform SST yielded tropical meridionally overturningcells with vastly differing strengths. In particular, low and high thresholds produced a weakand strong moist Hadley cells, respectively (Horinouchi, 2012).Focusing on transient phenomena, uniform SST runs on a f –plane aimed at understandingrotating radiative-convective equilibrium, were seen to result in the spontaneous formation oftropical storm-like vortices (Held and Zhao, 2008). These results were expanded to largerdomains (Zhou et al., 2014), and seen to carry over to a spherical geometry where multiplevortices were seen to form, persist and drift poleward (Shi and Bretherton, 2014). Moreover, thenumber of systems produced was seen to be dependent on the magnitude of SST (Khairoutdinovand Emanuel, 2013; Merlis et al., 2016). The genesis of these storms, preferential latitudes andscales of the systems have also been examined in experiments with varying rates of planetaryrotation (Chavas and Reed, 2019). In the aforementioned simulations with varying thresholdsfor triggering deep convection, persistent cyclones along with organized tropical waves wereobserved in runs with high relative humidity threshold (Horinouchi, 2012). At long times,after the establishment of zonal mean states, sporadic well organized eastward propagatingconvective activity was observed in the tropics (Sumi, 1992). In fact, detailed simulations withcloud resolving convective parameterization have noted moisture-convection feedback and theemergence of modes resembling the MJO in flat SST experiments (Grabowski, 2003; Grabowskiand Moncrieff, 2004).From a self-aggregation perspective, uniform SST simulations have demonstrated the spon-taneous appearance of organized moist convection and this was attributed to the feedbacksinvolving cloud-radiative interactions (Bretherton et al., 2005). The sensitivity of aggregationto the resolution of models, the domain size and magnitude of SST have been examined andlinked to the differing distribution of clouds in these scenarios (Muller and Held, 2012). Fur-ther, the initiation of aggregation and its maintenance were seen to be primarily driven by theshortwave and longwave radiative processes (Wing and Emanuel, 2014). The self-aggregation2f moisture is believed to play a vital role in the genesis of tropical cyclones and its intensifica-tion (Shi and Bretherton, 2014; Muller and Romps, 2018), as well as in the emergence of MJO(Arnold and Randall, 2015).Here we adopt the viewpoint of treating water vapor as a dynamically active scalar field (Sobel,2002). In particular, we note that the coupling of water vapor, or any other condensablesubstance, is controlled by its latent heat ( L v ). When L v →
0, the condensable substance isadvected with the flow but will not be dynamically coupled to the equations of motion . Aninteresting question to ask in this uniform SST scenario is, what happens when water vapor isno longer (or only weakly) dynamically active? For example, does the interaction of convectionand rotation still result in an organized meridional flow? Do we observe a tropical Hadleycell? Is there a systematic transport of energy across latitudes even without a SST gradient,and if so, why? And how does the partition of latent and dry static energy change in thesesimulations? What is the nature of the transient tropical wave activity? Specifically, giventhat the coupling with water vapor is thought to be essential to its existence (Grabowski, 2003;Raymond and Fuchs, 2009; Sobel and Maloney, 2013; Adames and Kim, 2016), do we observean intraseasonal mode like the MJO when water vapor behaves more like a passive tracer?Further, if there is a change in tropical intraseasonal activity, is it abrupt? How does the natureof the midlatitude synoptic vorticity field change as one progresses from a dynamically dry tomoist atmosphere. Do we always see the formation of tropical storm-like vortices? How aboutthe nature of precipitation and cloud fraction in a world with small L v ?In all, we believe that these simulations, much like the moist and dry scenarios studied byFrierson et al. (2006, 2007), will help in a more robust view of planetary atmospheric circulationregimes in the presence of a condensable substance, especially ones with a different latent heatcompared to water vapor. Further, even on present-day Earth, our hope is that exploring theeffect of different coupling strengths will help in the understanding of moist geophysical systems— much like the dry and moist simulations of tropical cyclones that are yielding insight intothe fundamental nature of these cyclonic systems (Mroweic et al., 2011; Cronin and Chavas,2019; Wang and Lin, 2020). The modeling framework to address these questions is describedin Section 2. The results pertaining to the mean flow and transient activity are presentedin Sections 3 and 4, respectively. Section 5 collects and discusses the main findings of ourinvestigation. The aquaplanet experiments are carried out using the Community Atmosphere Model, Version5.3 (CAM 5.3), the atmospheric component of the Community Earth System Model (CESM),Version 1.2.2. The model uses a finite volume dynamical core with a resolution of 0.9 ◦ latitude × ◦ longitude, with 30 vertical levels. Based on the recommendations of Medeiros et al. (2016),aerosol effects are minimized by removing the aerosol emissions and by specifying constantdroplet and ice number concentrations in the microphysics. Zhang-McFarlane convection schemeis used for deep convection, while the University of Washington (UWSC) scheme is employedfor shallow convection. Following Shi and Bretherton (2014), we have used the aquaplanetconfiguration with a prescribed globally uniform SST of 27 ◦ C. Uniform solar insolation is usedwith no diurnal cycle (Barsugli et al., 2005; Kirtman and Schneider, 2000) and the solar constantwas set to 342 W / m . All other fields (like Ozone) were set to their annual horizontally averaged The so called advection-condensation framework sets L v = 0, and has proven useful in probing the distributionof water vapor in the troposphere and its radiative impacts (Pierrehumbert et al., 2007; O’Gorman and Schneider,2006; Sukhatme and Young, 2011) L v . Specifically, latent heat is varied everywhere except in the Clausius-Clapeyronequation as detailed in Frierson et al. (2006). The degree of moist coupling can also be controlledby varying the amount of moisture in the atmosphere (by varying the saturation vapor pressureat a reference temperature) and keeping L v fixed (Frierson et al., 2006), but we prefer varying L v due to the ease in modifying this parameter in the model and the fact that this directlycontrols the amount of heat released on condensation. Further, changing L v can potentially beuseful in understanding other planetary atmospheres, such as Mars (Read and Lewis, 2004) orTitan (Mitchell and Lora, 2016), that have condensable substances with very different latentheats than water vapor. Specifically, as L v →
0, water substance is transported by the flow andis active in the radiation budget, but does not lead to latent heating in the dynamical equationsof motion. Note that for L v = 0 run, we also explicitly switch off the evaporation in the model.When L v >
0, water substance becomes dynamically active and its strength, or coupling withthe flow increases with L v . Perhaps due to excessive amount of heat released, the model failsat latent heats around twice the latent heat of water. So we restrict our experiments to L v between 1.5X to 0X, where X = 2 . × J/kg. The model is run for 3 years, with the firstyear of data being discarded as a spin-up time.
Even with no solar and surface meridional thermal gradients imposed, a systematic Hadley celllike circulation emerges in all the cases. This is seen in the mean meridional mass streamfunc-tion presented in Figure 1 as a pair of tropical cells straddling the equator. In addition, weakeroppositely directed Ferrel-like circulation cells are seen in the midlatitudes. For L v = 1X, themeridional streamfunction is of the same order of magnitude, but narrower than that of theannual mean circulation of present-day Earth (Dima and Wallace, 2003; Walker and Schneider,2005). Usually, the annual mean value is less than the seasonal extremes, but in these simula-tions, there are no seasons and this may explain the comparable magnitude even without a SSTgradient. In all these cases, i.e., from L v = 1.5X to 0X, the vorticity based local Rossby numberat the extremities of the Hadley cell is between (0 . , .
6) — thus, idealized theories from limitingcases of Rossby number → L v . As water vapor becomes passive, the tropical cells become narrower (Figure 1c),while the Ferrel cells seem more clearly formed. In fact, there is a systematic decrease in theHadley cell width from about 25 ◦ to 15 ◦ as L v varies from 1.5X to 0X. Remarkably, for L v ≤ P r ( h P a ) (a) Lv = 1.5X
45 30 15 0 15 30 4510008006004002000 (b)
Lv = 1X
45 30 15 0 15 30 4510008006004002000 (c)
Lv = 0.5X
45 30 15 0 15 30 45
Latitude P r ( h P a ) (d) Lv = 0.25X
45 30 15 0 15 30 45
Latitude (e)
Lv = 0.1X
45 30 15 0 15 30 45
Latitude (f)
Lv = 0X -6.8e-06 0.0e+00 6.8e-06 ms -9.7e-06 0.0e+00 9.7e-06 ms -2.4e-05 0.0e+00 2.4e-05 ms -1.5e-05 0.0e+00 1.5e-05 ms -2.3e-05 0.0e+00 2.3e-05 ms -3.0e-06 0.0e+00 3.0e-06 ms Figure 1: Mean meridional mass stream function (black contours; solid clockwise and dashedcounterclockwise) and angular momentum (grey contours) averaged over the last two years of theruns. The divergence of eddy momentum flux is shown in color. The stream function contoursare logarithmic in nature, with its magnitude doubling between successive contours. The lowestcontour has a value of 2 . × kg s − and zero contour is not shown. Angular momentumcontours are in the intervals of 0 . a , with decreasing values away from the equator.5 P r ( h P a ) (a) - - - - - . - . - . - . . . . . Lv = 1.5X
90 60 30 0 30 60 9010008006004002000 (b) - - - - - . - . - . - . . . . . Lv = 1X
90 60 30 0 30 60 9010008006004002000 (c) - - - - - . - . - . - . . . . . . . . . Lv = 0.5X
90 60 30 0 30 60 90
Latitude P r ( h P a ) (d) - - - - . - . - . - . - . - . - . - . . . . . . . Lv = 0.25X
90 60 30 0 30 60 90
Latitude (e) - - - - . - . - . - . - . - . - . - . - . . . . . Lv = 0.1X
90 60 30 0 30 60 90
Latitude (f) - . - . - . - . . . Lv = 0X
16 4 1 0 1 4 16 ms
16 4 1 0 1 4 16 ms
16 4 1 0 1 4 16 ms
16 4 1 0 1 4 16 ms
16 4 1 0 1 4 16 ms
16 4 1 0 1 4 16 ms Figure 2: Zonal mean zonal wind (in color) along with zonal temperature anomalies (blackcontours) averaged over the last two years of the run. Temperature anomalies are computed byremoving the mean at each pressure level. The contour levels are logarithmic in nature with itsmagnitude doubling between successive contours. Tropopause is shown as a thick red line.this circulation (Walker and Schneider, 2006; Schneider and Bordoni, 2008). The divergence ofthe resolved eddy angular momentum flux, ∇· [cos φ ( u (cid:48) v (cid:48) , u (cid:48) ω (cid:48) )], is shown by colored shading inFigure 1. Unlike eddy fluxes in the present-day atmosphere (Ait-Chaalal and Schneider, 2015),the vertical component of momentum flux convergence is also significant, especially as watervapor becomes passive in nature. Also, these fluxes are dominant in both the upper troposphereand near the surface of the planet. In fact, these flat SST cases constitute an intermediate regimebetween present-day Earth where the eddy momentum fluxes are concentrated in the upper tro-posphere, and reverse surface temperature gradient scenarios on Earth as well as high obliquityplanets where eddy fluxes are confined near the surface (Ait-Chaalal and Schneider, 2015; Kanget al., 2019). Interestingly, the resolved eddy fluxes seem to account well for the crossing ofangular momentum surfaces by the overturning circulation in the lower troposphere, but notso well in the upper troposphere, where their structure is more complex than the overturningcirculation. This implies the existence of another momentum sink, possibly unresolved gravitywaves which arise near the tropopause, or a numerical sink of momentum (Toniazzo et al.,2020). Further, as L v becomes smaller, the eddy momentum flux is progressively restricted tothe tropical regions.The zonal mean zonal winds as a function of height and latitude are shown in Figure 2. Forall the cases, easterlies form through most of the tropical troposphere and a small region ofsuperrotation is seen directly above the equator near the tropopause. Indeed, low pole to equatorSST gradients are known to favour tropical superrotation (Lutsko, 2018). The magnitude ofsuperrotation becomes weaker, moves into the stratosphere and progressively disappears as L v →
0. For L v = 1.5X, 1X and 0.5X cases, westerlies form in the subtropics, though mostly in6 ms
1e 6603003060 L a t i t u d e (a) ms
1e 6603003060 (b) ms
1e 6603003060 (c)
20 10 0 10 20 ms ms ms Figure 3: Zonal momentum flux convergence (blue) and zonal mean zonal wind (orange) for L v = 1X. The plots are averaged over (a) 50-150 hPa, (b) 200-400 hPa and (c) 500-800 hParespectively, and is computed using the last two years of the run.the lower and middle levels. Whereas for lower latent heats, westerlies are pushed equatorwardand upwards between 200-400 hPa. The momentum flux convergence that is responsible fordriving the zonal mean zonal flow is shown in Figure 3. From a flux decomposition (Lee, 1999),we note that transient eddies play a dominant role in the momentum flux budget. This isexpected because there is no orography, land-sea contrast or imposed SST structure, hencestationary eddies are absent. Further, the mean meridional circulation also does not have aseasonal cycle and its contribution to the momentum flux convergence is small (Lee, 1999;Dima et al., 2005). As is illustrated for L v = 1X in Figure 3, the eddy flux convergence drivesthe upper equatorial superrotation (Figure 3a), the equatorial easterlies below 200 hPa (Figure3b) and the westerlies in the midlatitudes from 30-60 ◦ below 500 hPa (Figure 3c). Even in othercases, the eddy convergence seems to be primarily responsible for driving the zonal mean flow.In the midlatitudes, we also see a transition from lower level westerly jets in the Ferrel-like cellregions of the strong moisture coupled runs, to narrow upper-tropospheric and stratosphericwesterly jets for the weaker moisture coupled runs. We note that the change from a surface toan upper tropospheric westerly jet goes along with a change in the direction of the overturningcells. In fact, the sharp westerly jets between 200-400 hPa at around 15 ◦ -20 ◦ latitude in the L v = 0.1X runs are maintained by a vertical advection of zonal momentum by the strong ascentat around 600 hPa (Fig. 1e). The existence of easterlies at all latitudes for the low moisturecoupled runs implies a net gain of atmospheric angular momentum from the surface. This netgain of momentum is actually found in all runs (the globally averaged surface zonal winds areon average negative for all runs), and is consistent with the model not conserving momentum(Toniazzo et al., 2020).The zonal mean temperature anomalies, i.e., the deviation with respect to the mean on a givenpressure level, is also shown in Figure 2. Due to the absence of any prescribed horizontal thermalgradients, the temperature anomalies are much weaker than the typical Earth-like case. Whilethe anomalies are less than 0.5 ◦ C in the lower troposphere, the largest anomalies are seen in theupper atmosphere with magnitudes reaching up to 4 ◦ C. For L v ≥
00 220 240 260 280 300 K P r ( h P a ) (a) Lv = 1.5XLv = 1XLv = 0.25XLv = 0.1XLv = 0X 280 300 320 340 360 380 K (b) Lv = 1.5XLv = 1XLv = 0.25XLv = 0.1XLv = 0X 280 300 320 340 360 380 K (c) Lv = 1.5XLv = 1XLv = 0.25XLv = 0.1XLv = 0X
Figure 4: Vertical profile of (a) temperature, (b) potential temperature and (c) equivalentpotential temperature in the tropics (30 ◦ N/S). The profile is averaged over the last two yearsof the run.Interestingly, for weaker coupling of water vapor, the situation changes and the equatorial regionis marginally warmer than the poles. Note that this change in the equator to pole temperaturegradient is accompanied by a reversal in the sense of tropical circulation. Thus, in all cases,the tropical cell is an indirect circulation. In fact, the criterion for a moist direct circulation(Emanuel, 1995) is almost never satisfied by these simulations, and as noted above, the coherentcirculation cells in the troposphere are strongly influenced by the eddy fluxes. With a lack ofimposed temperature gradients at the surface, the sense of temperature anomalies in the uppertroposphere in Figure 2 is consistent with the overturning circulation and its reversal seen inFigure 1. Further, the direction of zonal mean flow in the upper midlatitudes is also in accordwith the meridional temperature gradient in the troposphere.Vertical profiles of temperature, potential temperature and equivalent potential temperatureare shown in Figure 4. The temperature lapse rate increases with decreasing moist coupling,with the L v = 0X approaching the dry lapse rate. Potential temperature increases with heightand in general, lower latent heat cases have lower potential temperature. Equivalent potentialtemperature is mostly constant till up to 200 hPa, especially for low coupling cases of L v =0.1X and 0.25X. For the more active cases, L v = 1.5X and 1X, it decreases till about 700hPa and then rises gradually up to a height of 200 hPa. As expected, with decreasing moistcoupling the equivalent potential temperature and potential temperature profile converges. Thedecrease of equivalent potential temperature with height (but not the potential temperature)indicates that the conditions are unstable for moist ascent but stable for dry ascent in the lowertroposphere. Further, higher up in the troposphere, we note an increase in stability with L v .This is reminiscent of the present-day tropics (Holton and Hakim, 2012), but note that, herethe vertical profiles shown in Figure 4 are similar in the tropical and midlatitudinal regions.Thus, the conditional instability observed for strong coupling is present in the tropics as wellas at higher latitudes. There is a sharp change in potential temperature at upper levels (Figure4b) roughly coinciding with the level of tropopause (shown by red line in Figure 2). But, forthe dry case, which has a constant potential temperature up to 600 hPa and rising rapidlyabove it, the identified tropopause level is significantly higher. This is due to our usage of theWMO criterion of 2 K/km lapse rate to identify the level of tropopause. As seen from Figure4a, for the dry case, even though the temperature lapse rate decreases above 600 hPa, its not8 .6 0.3 0.0 0.3 0.69060300306090 L a t i t u d e (a) Dry
Lv = 1.5XLv = 1XLv = 0.5X (b)
Latent
Lv = 1.5XLv = 1XLv = 0.5X (c)
Total
Lv = 1.5XLv = 1XLv = 0.5X PW L a t i t u d e (d) Lv = 0.25XLv = 0.1XLv = 0X PW (e) Lv = 0.25XLv = 0.1XLv = 0X PW (f) Lv = 0.25XLv = 0.1XLv = 0X
Figure 5: Vertically integrated dry, latent heat and moist static energy fluxes averaged over thelast two years of the run.sufficiently low to identify as the tropopause by the WMO criterion, and hence the discrepancy.But, from a dynamical point of view, the tropical overturning circulation and eddy fluxes forthe dry case are confined below 600 hPa.The vertically integrated dry and moist static energy fluxes are shown in Figure 5. Moiststatic energy (MSE), m = C p T + gz + L v q , comprises of the dry static energy C p T + gz and latent heat energy L v q . Vertically integrated moist static energy flux is then defined as2 πacosφ (cid:82) p s vmdP/g , where φ is the latitude, a is the radius of earth, g is the acceleration dueto gravity, p s is the surface pressure and overbar denotes a time and zonal mean (Frierson et al.,2007). Tropical latent heat flux is equatorward (poleward) for higher (lower) latent heats, i.e., itmimics the sense of mean circulation at the surface. The dry static energy flux opposes the latentterm in the tropics; thus, MSE transport is determined by the difference of these two opposingcomponents. For the more active cases, i.e., L v > L v ≤ L v ≤ L v values, regardless of the direction of the zonal mean MSEgradients (which are similar to the temperature anomalies, Figure 2). Thus, while for the weaklycoupled runs the midlatitude MSE fluxes are down gradient, for the strongly coupled runs, the9 .00 0.25 0.50 0.75 1.00 1.25 1.50 Latent Heat (x 2.5e6) T o t a l P o l e w a r d T r a n s p o r t ( P W ) (a) DryLatentTotal 0.00 0.25 0.50 0.75 1.00 1.25 1.50
Latent Heat (x 2.5e6) W / m (b) SWLWLatentSensibleNet
Figure 6: (a) Global mean poleward transport of energy as a function of latent heat. (b)Meridional gradient of various flux components (positive sign implying a net flux into theatmosphere). Here, gradient is taken as the difference between the averaged values in thetropics (30 ◦ N/S) and extra-tropics (30 ◦ - 90 ◦ latitudes). The plots are computed using the lasttwo years of the run.
90 60 30 0 30 60 906065707580 W / m (a) Net SW (TOA - Sfc)
Lv = 1.5XLv = 1XLv = 0.25XLv = 0.1X 90 60 30 0 30 60 90280260240220200180 (b)
OLR
Lv = 1.5XLv = 1XLv = 0.25XLv = 0.1X 90 60 30 0 30 60 9030507090110 (c)
LW (Sfc)
Lv = 1.5XLv = 1XLv = 0.25XLv = 0.1X90 60 30 0 30 60 90
Latitude W / m (d) Latent
Lv = 1.5XLv = 1XLv = 0.25XLv = 0.1X 90 60 30 0 30 60 90
Latitude (e)
Sensible
Lv = 1.5XLv = 1XLv = 0.25XLv = 0.1X 90 60 30 0 30 60 90
Latitude (f)
Net
Lv = 1.5XLv = 1XLv = 0.25XLv = 0.1X
Figure 7: Zonal mean time mean fluxes for four different simulations; specifically, L v = 1.5X,1X, 0.25X and 0.1X. The various components are, (a) net atmospheric absorption of shortwaveradiation (top - surface fluxes), (b) longwave radiation at the top (OLR), (c) longwave radiationat the surface, (d) latent heat, (e) sensible heat and (f) net flux into the atmosphere. The fluxesare averaged over the last two years of the run and positive sign implies a net flux into theatmosphere. 10idlatitude MSE fluxes are up gradient, suggesting the process driving the flux is not a simpleeddy diffusivity or baroclinic instability. Remarkably the magnitude of total MSE transport(Figure 6) remains roughly the same for strong coupling (till L v = 0.25X) and falls drasticallyas water substance becomes almost passive. The nearly constant transport is due to an increaseof dry static energy transport that compensates the decreasing latent heat component with L v .This is reminiscent of a similar compensation found by Frierson et al. (2007), when moisturecontent is varied but with a meridional surface temperature gradient. Similar to the meanmeridional circulation, the meridional energy transport is also an order smaller than the present-day Earth like conditions. Further, a decomposition of moist static energy flux into meanand eddy components suggests that the major contribution is from transient eddies in themidlatitudes, while the mean flow dominates the transport in the tropics. In all, as withthe appearance of a coherent circulation, despite the lack of an imposed SST gradient, theatmospheric flow systematically transports energy towards the poles and this is almost invariantfor cases with relatively strong moisture coupling.A net poleward MSE flux implies a net input of energy into the atmosphere at low latitudes,and a net output of energy at high latitudes (Trenberth and Stepaniak, 2003). We examinehow the net Energy Input into the Atmosphere (EIA), and it’s latitudinal gradient changeswith the strength of moisture coupling. Figure 7 shows the latitudinal profile of the time andzonal mean latent, sensible and net longwave radiation at the surface, as well as the outgoinglongwave radiation (OLR) at the top of the atmosphere and the net atmospheric absorptionof shortwave radiation for a few L v values. We see that as L v increases, the energy exchangeat the surface, which acts to heat the atmosphere, changes from being dominated by longwaveradiation to being dominated by latent heat (LH) fluxes. Consistently, the longwave radiationseems to saturate at low L v values (values similar for L v = 0.1X and L v = 0.25X, c.f. Figure7c), while LH starts to saturate at large L v (only a small increase between L v = 1X and L v =1.5X, Figure 7d). We also note an opposing dependence of LH and longwave radiation on theatmospheric temperature – warmer atmospheric temperatures increase LH due to the largermoisture holding capacity of the air, but they decrease the net longwave cooling of the surfacebecause of the larger downward flux from the atmosphere (while SST is fixed). As a result, themean latitudinal gradient of these fields (Figure 6b) changes following the temperature field –as L v increases, the mean latitudinal gradient of net surface longwave radiation increases, whilethat of LH decreases (directly via the increase in L v and indirectly through the temperatureeffect) . The sensible heat (SH) fluxes have a relatively small contribution to the overall surfaceenergy budget, however, their contribution to the mean latitudinal gradient is significant (Figure7e), with values being strongest in the tropics for all runs, but the tropics–high latitudes gradientis largest for large L v . This is probably due to two factors affecting the sensible heat fluxes–the strength of surface winds and the atmospheric temperatures. For large L v , with warmerpoles, the colder temperatures and stronger surface winds make the equatorial SH fluxes largerthan at the poles, but for the small L v runs, the warmer tropics act to reduce the equator–polegradient. The other atmospheric heating term is the net shortwave radiation absorption. Thisquantity depends mostly on the clouds, which change in a complex way (especially for small L v ), but its overall contribution peaks at high latitudes for large L v , while for small L v thelatitudinal gradient reverses, and changes strongly especially in the tropics for L v = 0.1X. Theabove heating terms are balanced by the OLR, which increases with L v , and saturates at around L v = 1X. This increase is consistent with the decrease in temperature lapse rate as L v increases,since the longwave emitted to space originates in the mid-upper troposphere (Pierrehumbert,2010). At smallest L v , OLR has a strong peak in the tropics, and a flat extratropical profile, but Note that for clarity of the presentation, Figure 6b shows the mean gradient of the total longwave radiationi.e., net upward flux at surface minus OLR. While these fluxes have an opposing dependence on atmospherictemperature, the variation or their mean latitudinal gradient with L v is similar for values L v ≥ L v is increased to emissions peaking at high latitudes for large L v . Summing allthese contributions to get the net EIA (Figure 7f), we see that the atmosphere gains energy inthe subtropics and emits energy in the extratropics and near the equator. This is consistent withthe subtropical peak of MSE flux for small L v and the stronger midlatitude peaks of MSE fluxfor large L v (Figure 5), and is of course quite different from present-day Earth with differentialsolar heating. Despite the somewhat noisy latitudinal profile of EIA, we see a clear tendency forthe latitudinal gradient to increase with L v , so that for large L v , the high latitudes cool moststrongly and the subtropics heat most strongly. This change in high minus low latitude EIA(Figure 6b) is consistent with the increase in mean poleward MSE flux as L v increases (Fig 6a).It is again interesting to point out that for strong moisture coupling, the MSE fluxes are up-gradient, i.e., from the colder tropics to the warmer poles. Consistently, we will see later onthat the poleward MSE flux is driven by the dynamics of vortices on a sphere, resulting inthe high latitudes being warmer than the tropics. At small L v values, the anomalies are morewave-like, and the EIA is dominated by radiative processes (OLR and SW absorption) to yielda net positive heating in the tropics. The wave-like anomalies, in this case, act diffusively andflux MSE poleward, reducing (but not reversing) the temperature gradient. In the previous section, we looked at the emergent steady state zonal mean circulation and itsmaintenance by diabatic processes and by eddy MSE and angular momentum fluxes. Given thatthere is no differential heating imposed in these runs, the underlying processes driving these eddyfluxes are not apriori clear. In this section we examine the characteristics of the transient eddiesand vortices in more detail. The globally integrated kinetic energy (KE) spectra for variousscenarios is shown in Figure 8. The KE dominates over the available potential energy (Shi andBretherton, 2014) and further, the rotational part of the spectrum has much more energy thanthe divergent component, so Figure 8 is essentially a plot of the rotational kinetic energy. Signsof power-law scaling emerge with stronger moist coupling, and the exponent of the spectrumtends to be around − − k E ( k ) Lv = 1.5XLv = 1XLv = 0.25XLv = 0X
Length (km)
Figure 8: Vertically integrated kinetic energy spectra averaged over the last two years of the run.The spectra is integrated from the surface till 200 hPa and is normalized for easy comparison.Black dashed lines have slopes of − −
4. Deformation radius is denoted by dashed verticallines.the stronger coupling runs. It is important to note that these wavenumber–frequency diagramsrepresent scales larger than those accounted for in the spectra of Figure 8. The intraseasonalmodes, especially the eastward moving MJO and Kelvin waves disappear with decreasing latentheat, i.e., as water substance becomes dynamically passive. The lack of a MJO-like mode forweak coupling suggests that this mode of low frequency activity — at least in this modelingframework — requires interactive moisture (Sobel and Maloney, 2013; Adames and Kim, 2016).The loss of Kelvin waves is evident by the L v = 0.25X case itself. In fact, together with evidenceon the disappearance of Kelvin modes with changes in parameters like the relative humiditythreshold for triggering convection (Horinouchi, 2012), convective time scales (Frierson, 2007)and background saturation fields (Suhas and Sukhatme, 2020), this points to the sensitivityof these waves in the tropical atmosphere. For the limiting case of L v = 0X (Figure 9g,h), itappears as though there isn’t much variability in the tropics at these large length scales.In all simulations, intraseasonal variability (around 20–90 days) is dominant in the tropics.Specifically, tropical zonal winds in lower atmosphere are averaged to produce a daily timeseries whose spectral properties are shown in Figure 10. A sharp distinct peak is observedwith stronger coupling, and the dominant period of activity shifts to longer time scales withincreasing latent heats (Figure 10a), i.e., as L v = 0.75X → L v = 1.25X and 1.5X there is a second windowof activity that is at shorter time scales, approximately 7 to 15 days. However, with weakenedcoupling, especially for the cases with L v = 0.25X and below, rather than distinct peaks, thewinds show more of a plateau that is spread around the period of a month (Figure 10b). Takentogether, a broad picture that emerges in the tropics is that of progressively longer periodintraseasonal variability that aligns with familiar tropical dispersion curves and co-exists withsmaller scale (predominantly rotational) turbulence for stronger moist coupling.With regard to transients at higher latitudes, a snapshot of the lower level relative vorticity(850 hPa) for Day 1000 is shown in Figure 11. Much like prior observations in f –plane (Heldand Zhao, 2008) and global (Shi and Bretherton, 2014; Chavas and Reed, 2019) uniform SSTsimulations, intense cyclonic structures can be seen in the extra-tropics for relatively strong13 L v = . X (a) Symmetric
15 10 5 0 5 10 150.10.20.30.40.5 F r e q u e n c y ( d a y ) (b) Anti-Symmetric
15 10 5 0 5 10 150.10.20.30.40.5 L v = X (c)
15 10 5 0 5 10 150.10.20.30.40.5 F r e q u e n c y ( d a y ) (d)
15 10 5 0 5 10 150.10.20.30.40.5 L v = . X (e)
15 10 5 0 5 10 150.10.20.30.40.5 F r e q u e n c y ( d a y ) (f)
15 10 5 0 5 10 15 kx L v = X (g)
15 10 5 0 5 10 15 kx F r e q u e n c y ( d a y ) (h) Figure 9: Frequency–wavenumber power spectra of zonal wind at 850 hPa. Symmetric andanti-symmetric components are shown in the first and second column, respectively and thebackground spectra is removed. The power spectra is calculated over the last two years of therun and is averaged over 15 ◦ N to 15 ◦ S latitudes. Superimposed are the dispersion curves (inblack) with an equivalent depth of 12, 25 and 50m.14 Frequency (day ) E ( f ) * f (a) Lv = 1.5XLv = 1.25XLv = 1XLv = 0.75X 10 Frequency (day ) (b) Lv = 0.5XLv = 0.25XLv = 0.1XLv = 0X360 180 90 30 7
Period (day)
360 180 90 30 7
Period (day)
Figure 10: Variance preserving spectra of zonal wind at 850 hPa averaged over the latitudes15 ◦ N to 15 ◦ S. The spectra is normalized and is computed using the last two years of the run.coupling. Animations of the vorticity field suggest that these storm-like vortices are born inthe subtropics, sometimes merge to form larger structures and progress westward and towardsthe poles, due to β –drift (Shi and Bretherton, 2014). In fact, these storms are responsible forthe poleward transport of both dry and latent energy as noted in Figure 5. The lifetime ofthese systems is approximately of the order of a month. The vertical profile of a typical stormfor different L v values is shown in Figure 12a, b. Here, the vorticity anomaly has a maximumin the lower troposphere and the system has a warm core with largest temperature anomaliesin the upper troposphere. Thus, even though these storms are observed at higher latitudes,they are broadly similar to present-day tropical cyclones (Wang and Jiang, 2019) and sometropical lows (Kushwaha et al., 2020). Indeed, for L v = 1X, the largest temperature anomalyis at approximately 300 hPa and has a value of about 6 K, both aspects being comparableto present-day category one tropical cyclones (Wang and Jiang, 2019). The fact that suchcyclones are supported at higher latitudes in these simulations is likely tied to the verticalthermal structure of the atmosphere, especially its similarity with tropical regions (Figure 4).The strength of these vortices, the magnitude of temperature anomalies and their spatial scaledecreases with lower latent heats and in the dry simulation (Figure 12 and Figure 11c and d).A similar pattern emerged in simulations of cyclones with increasing surface dryness (Croninand Chavas, 2019). This change in morphology is starkly captured in the probability densityfunction (PDF) of the 850 hPa relative vorticity field as shown in Figure 12c. Quite clearly, forstrong coupling the PDF is fat-tailed with a high propensity of extremes associated with intensecyclones, whereas for progressively smaller coupling the PDF tends to have exponential tails(seen as straight lines in the semi-log plot of Figure 12c). In addition to these deep troposphericvortices, the midlatitudes also show shallow near tropopause waves. Examples of these systemscan been seen in a snapshot of vorticity anomalies for a few L v cases in Figure 13. Specifically,in this snapshot, for L v = 1X, we observe near tropopause anomalies, while for L v = 0.25X wecapture both the near tropopause structures as well as the deeper poleward drifting vortices.Indeed, these disturbances are identified as waves by noting that they are associated with asystematic meridional heat transport (albeit much smaller than the MSE transport; Figure 5)of the form v (cid:48) T (cid:48) . In fact, the near tropopause waves are tied to the relatively large temperatureanomalies noted in Figure 2. 15 a) Lv = 1.5X (b)
Lv = 1X (c)
Lv = 0.25X (d)
Lv = 0X -4e-04 -2e-05 0e+00 2e-05 4e-04 s -4e-04 -2e-05 0e+00 2e-05 4e-04 s -4e-04 -2e-05 0e+00 2e-05 4e-04 s -4e-04 -2e-05 0e+00 2e-05 4e-04 s Figure 11: Snapshot of relative vorticity at 850 hPa on Day 1000 for four different simulations;specifically, L v = 1.5X, 1X, 0.25X and 0X. The contour levels are logarithmic in nature withits magnitude doubling between successive contours.16 .00 0.25 0.50 0.75 1.00 s
1e 410008006004002000 P r ( h P a ) (a) Lv = 1.5XLv = 1XLv = 0.5X K P r ( h P a ) (b) Lv = 1.5XLv = 1XLv = 0.5X s
1e 410 P D F (c) Lv = 1.5XLv = 1XLv = 0.25XLv = 0X
Figure 12: Vertical profile of (a) relative vorticity and (b) temperature anomaly inside thecyclonic structure. The profiles are computed by averaging over multiple cyclonic structures atDay 1000. Temperature anomaly is defined as the deviation from the global average at eachvertical level. Panel (c) shows the probability density function of relative vorticity field at 850hPa, and is computed over the last two years of the run.
30 40 50 60
Latitude P r ( h P a ) (a)
30 40 50 60
Latitude (b) -1e-04 0e+00 1e-04 s -1e-04 0e+00 1e-04 s Figure 13: Vertical section of vorticity anomalies for (a) L v = 1 X and (b) L v = 0 . X cases atDay 1000. Vorticity anomalies are the deviation from the vertical mean and is shown for 30 ◦ N- 60 ◦ N latitudes and 180 ◦ longitude. 17 a) Lv = 1.5X (b)
Lv = 1X (c)
Lv = 0.25X (d)
Lv = 0.1X mm/day mm/day mm/day mm/day Figure 14: Snapshots of the daily precipitation at Day 1000 for four different simulations;specifically, L v = 1.5X, 1X, 0.25X and 0.1X. Note that the colorbar is logarithmic in nature. Snapshots of the daily precipitation fields are shown in Figure 14. For stronger coupling (sayfor example, L v = 1.5X) intense rainfall events (dark colors in Figure 14a) have a very differentcharacter in the tropics and midlatitudes. The former have a wave-like nature while the latterare clearly associated with vortices. As latent heat decreases, rainfall is more evenly distributed,with relatively higher variability near the tropics and subtropics. Further, with weaker coupling,the precipitation has a peak at subtropics (also seen in Figure 15a), corresponding to theascending motion in the subtropical region (Figure 1). However, with stronger coupling theprecipitation has an equatorial peak (at least locally), consistent with the ascending motion inthe tropics (Figure 1). Thus, the sense of the Hadley cell seen in Figure 1, especially its change indirection of overturning is in line with the zonal mean precipitation. In fact, the near equatorialregion gets progressively devoid of strong rainfall events with decreasing L v which is consistentwith the loss of Kelvin and MJO-like modes (Figure 9). Rossby waves survive for smaller L v ,and are likely responsible for the between gyre or meridionally aligned events in Figure 14c, d(Wheeler et al., 2000; Suhas and Sukhatme, 2020). Given that there are no meridional gradients18 mm/day L a t i t u d e (a) Lv = 1.5XLv = 1XLv = 0.25XLv = 0.1X mm/day P D F (b) Lv = 1.5XLv = 1XLv = 0.25XLv = 0.1X
Figure 15: (a) Zonal mean precipitation averaged over the last two years of the run. (b)Probability density function of the daily mean precipitation computed over the the last twoyears of the run. Four different simulations with L v = 1.5X, 1X, 0.25X and 0.1X are shown.at the surface, the organization of moist convection in the tropics by equatorial waves likely giverise to precipitation patterns, which in turn, via latent heating couple to result in large-scaleascent and the overturning flow (Horinouchi, 2012). The connection of tropical wave activityto the Hadley cell is also supported by the fact that the width of these tropical cells, while notan exact match, follows the same pattern as the equatorial deformation scale in Figure 8.Interestingly, precipitation is actually higher for lower latent heat values at all latitudes as isseen in Figure 15a. The distribution of these precipitation events (Figure 15b) shows thatwhile the larger latent heat cases have a higher chance of extreme rainfall (up to 400 mm/day),the lower latent heat runs have a higher probability of comparatively moderate rainfall events(about 50-150 mm/day). A similar picture emerges in Figure 16a, with the global mean precipi-tation increasing with decreasing latent heat. Note that precipitable water increases with moistcoupling (Figure 16a), perhaps suggesting that with decrease in precipitation, more precipitablewater is stored in the atmosphere. Consistent with the behaviour of precipitation, cloud frac-tion, shown in Figure 16b, also increases with decreasing L v . This increase is mainly due tochanges in the medium and high cloud cover. We also see that outgoing longwave radiation(OLR) saturates at high L v (Figure 16b), possibly as a result of the increase in water vapor (Kolland Cronin, 2018). Overall, as OLR saturates and the SST is fixed, this results in an increasein tropospheric temperature with height (Figure 4b, c). This implies an increase in stability forlarger L v in the middle and upper troposphere. In turn, this possibly affects middle and highclouds and accounts for the observed decrease in precipitation and cloud fraction (Bony et al.,2016). Interestingly, recent explorations of moist and dry warm core cyclones have also shownan increase in cloud fraction for relatively drier conditions (Cronin and Chavas, 2019).19 .00 0.25 0.50 0.75 1.00 1.25 1.50 Latent Heat (x 2.5e6) mm / d a y (a) ConvectiveLarge ScaleTotal 0.00 0.25 0.50 0.75 1.00 1.25 1.50
Latent Heat (x 2.5e6) C l o u d F r a c t i o n (b) LowMediumHighTotal2030405060 k g / m Precipitable Water 190210230250270 W / m OLR
Figure 16: Globally averaged (a) precipitation & precipitable water and, (b) cloud fraction &OLR as a function of latent heat. The fields are averaged over the last two years of the run.
We have explored the various aspects of dynamically dry and moist atmospheric circulationsusing a 3–D aquaplanet model with uniform lower boundary conditions. The degree or strengthof moist coupling was controlled by systematically varying the latent heat of water vapor ( L v ),and with L v →
0, water substance is essentially a passive tracer from a dynamical point ofview. This allows one to not only contrast the moist and dry dynamics, but also look at theemergent dynamics in the intermediate moist coupling regimes. In all these experiments, eventhough the SST is fixed, the atmosphere is in energy balance.Despite the absence of meridional thermal gradients, we observe a general circulation withHadley and Ferrel cells. These cells are of comparable magnitude to those on present-day Earth.But, the nature of the Hadley cell is quite different from an Earth-like scenario where the surfacetemperature gradient and midlatitude baroclinic waves play a prominent role in determiningits terminus (Levine and Schneider, 2015); in fact, as described by Horinouchi (2012), ascentin these cells is possibly dictated by the organization of latent heating via equatorial waveswhich are prominent with strong coupling. Interestingly, for weaker moist coupling ( L v ≤ L v =1.5X to 0.25X, and then falls drastically as water substance becomes almost passive. Much likethe observations of Frierson et al. (2007), the nearly constant transport is due to an increaseof dry static energy transport that compensates the decreasing latent heat component with L v .The kinetic energy spectra for the relatively strong coupling runs have some similarities to theEarth’s atmosphere; specifically, the power spectrum is characterized by an approximate slopeof − − L v goes from 0.75X to1.5X, the dominant peak of activity moves from about a month to 90 days. Further, intrasea-sonal modes, especially the eastward moving MJO-like structure and Kelvin waves disappearas water substance becomes dynamically passive. The lack of a MJO-like mode for the pas-sive case suggests the vital and possibly an essential role played by interactive moisture in itsdynamics. In accord with the spatial energy spectrum, the passive case does not have muchenergy at large scales. The midlatitudes are characterised by multiple tropical storm-like warmcore vortices, drifting poleward and westward over the time period of a month. The lack ofbaroclinic instability due to the absence of imposed meridional thermal gradients, and the con-ditionally unstable nature of the temperature profiles possibly favour the existence of warm corevortices similar to the Earth-like tropical cyclones even at higher latitudes. These vortices areassociated with extreme rainfall events, and are also the reason for a poleward MSE transportin the midlatitudes despite the absence of any imposed temperature surface gradient. As withthe “dry” simulations of tropical cyclones Cronin and Chavas (2019), when water substancebecomes passive, the storm-like vortices become less intense, smaller in size and their temper-ature anomalies decrease. The change in the vorticity field is succinctly captured by the shiftfrom a PDF with fat-tails (large L v ) to one with a purely exponential form (small L v ). Thelargest temperature anomalies are observed in the upper troposphere, and in addition to thedeep storm-like vortices, we note the presence of near tropopause, shallow waves that contribute(albeit a very small amount) to the poleward dry static energy transport. The change fromwave-like disturbances to warm-core vortices seems to have very strong implications for thezonal mean energy budget and mean temperature structure. While wave disturbances act tomix background gradients diffusively, the moist and warm vortices which move poleward dueto a beta drift carry MSE poleward regardless of the direction of the background MSE gradi-ent. As a result, the vortex world has warmer poles and an up gradient MSE flux, while thewave-world has colder poles and down gradient MSE fluxes.21napshots of intense precipitation suggest a change in morphology between tropical and mid-latitude events, the former are wave-like while the latter are in the form of vortices. Withdecreasing L v , the region near the equator is progressively devoid of intense rainfall, this goeshand in hand with the loss of Kelvin and MJO-like modes. Interestingly, global mean precipi-tation increases with weaker moist coupling with an opposing trend shown by the precipitablewater. In essence, extreme rainfall events decrease, but there is more rainfall with weaker cou-pling of water substance. The decrease in rainfall and mid to upper level cloud fraction withincreasing L v appears to be tied to the mechanism proposed by Bony et al. (2016), involvingincreased static stability of the troposphere at middle and upper levels. More fundamentally,OLR saturates as L v increases, hence with fixed SSTs, the system responds with an increasein tropospheric temperature. This appears to be the cause for an increased static stability andresults in lower amounts of total precipitation with strong coupling of water vapor.We believe that the results from these simulations, much like the moist and dry scenariosstudied by Frierson et al. (2006, 2007), will help in developing a more robust view of planetaryatmospheric circulation regimes in the presence of a condensable substance. In particular, theflexibility of our approach could also be of use in probing the dynamics of planetary atmosphereswith condensable substances that have a significantly different latent heat than water vapor.From a general circulation point of view, these simulations show a very different way of obtainingtropical and extra-tropical regimes. Specifically, we have a Hadley cell with a well definedterminus but rather than demarcating the change to a baroclinic wave dominated regime (Levineand Schneider, 2015), here the transition is to storm-like warm core vortices. Indeed, the Hadleycell itself appears to be coupled to the equatorial waves and their modulation of latent heating,rather than a surface temperature gradient. In fact, the width of the Hadley cell is seen todecrease with L v , and this seems to be directly connected to the equatorial deformation scale.Further, similar to emerging details on the possibility of warm core cyclones in dry simulations(Mroweic et al., 2011; Cronin and Chavas, 2019; Wang and Lin, 2020), and their implicationsfor real-world cyclonic systems, the effects of varying moist coupling on transient activity couldprove to be of use in a more basic understanding of the moist modes of tropical intraseasonalvariability. Acknowledgements
We would like to thank the Supercomputing Education and Research Centre (SERC) at IIScfor computer facilitites where most of these simulations were carried out.
References
Adames, A. and Kim, D. (2016). The mjo as a dispersive, convectively coupled moisture wave:Theory and observations.
Journal of the Atmospheric Sciences , 73(3):913–941.Ait-Chaalal, F. and Schneider, T. (2015). Why eddy momentum fluxes are concentrated in theupper troposphere.
Journal of the Atmospheric Sciences , 72(4):1585–1604.Arnold, N. and Randall, D. (2015). Global-scale convective aggregation: Implications for themadden-julian oscillation.
Journal of Advances in Modeling Earth Systems , 7(4):1499–1518.Barsugli, J., Shin, S., and Sardeshmukh, P. (2005). Tropical climate regimes and global climatesensitivity in a simple setting.
Journal of the atmospheric sciences , 62(4):1226–1240.22artello, P. (1995). Geostrophic adjustment and inverse cascades in rotating stratified turbu-lence.
Journal of the atmospheric sciences , 52(24):4410–4428.Bony, S. et al. (2016). Thermodynamic control of anvil cloud amount.
Proceedings of theNational Academy of Sciences , 113:8927–8932.Bretherton, C., Blossey, P., and Khairoutdinov, M. (2005). An energy-balance analysis ofdeep convective self-aggregation above uniform sst.
Journal of the atmospheric sciences ,62(12):4273–4292.Brierley, C. et al. (2009). Greatly expanded tropical warm pool and weakened hadley circulationin the early pliocene.
Science , 323(5922):1714–1718.B¨uhler, O., Callies, J., and Ferrari, R. (2014). Wave-vortex decomposition of one-dimensionalship-track data.
Journal of Fluid Mechanics , 756:1007–1026.Chavas, D. and Reed, K. (2019). Dynamical aquaplanet experiments with uniform thermalforcing: System dynamics and implications for tropical cyclone genesis and size.
Journal ofthe Atmospheric Sciences , 76(8):2257–2274.Cronin, T. and Chavas, D. (2019). Dry and semidry tropical cyclones.
Journal of the Atmo-spheric Sciences , pages 1–20.Das, S. et al. (2016). Low-frequency intraseasonal variability in a zonally symmetric aquaplanetmodel.
Meteorology and Atmospheric Physics , 128(6):697–713.Dima, I. and Wallace, J. (2003). On the seasonality of the hadley cell.
Journal of the atmosphericsciences , 60(12):1522–1527.Dima, I., Wallace, J., and Kraucunas, I. (2005). Tropical zonal momentum balance in the ncepreanalyses.
Journal of the atmospheric sciences , 62(7):2499–2513.Emanuel, K. (1995). On thermally direct circulations in moist atmospheres.
Journal of theatmospheric sciences , 52(9):1529–1534.Fedorov, A., Brierley, C., and Emanuel, K. (2010). Tropical cyclones and permanent el ni˜no inthe early pliocene epoch.
Nature , 463(7284):1066–1070.Fedorov, A. et al. (2019). Tropical cyclogenesis in warm climates simulated by a cloud-systemresolving model.
Climate Dynamics , 52(1-2):107–127.Frierson, D. (2007). Convectively coupled kelvin waves in an idealized moist general circulationmodel.
Journal of the Atmospheric Sciences , 64(6):2076–2090.Frierson, D., Held, I., and Zurita-Gotor, P. (2006). A gray-radiation aquaplanet moist gcm. parti: Static stability and eddy scale.
Journal of the atmospheric sciences , 63(10):2548–2566.Frierson, D., Held, I., and Zurita-Gotor, P. (2007). A gray-radiation aquaplanet moist gcm. partii: Energy transports in altered climates.
Journal of the atmospheric sciences , 64(5):1680–1693.Grabowski, W. (2003). Mjo-like coherent structures: Sensitivity simulations using the cloud-resolving convection parameterization (crcp).
Journal of the Atmospheric Sciences , 60(6):847–864.Grabowski, W. and Moncrieff, M. (2004). Moisture–convection feedback in the tropics.
Quar-terly Journal of the Royal Meteorological Society: A journal of the atmospheric sciences,applied meteorology and physical oceanography , 130(604):3081–3104.23eld, I. and Zhao, M. (2008). Horizontally homogeneous rotating radiative–convective equilibriaat gcm resolution.
Journal of the Atmospheric Sciences , 65(6):2003–2013.Holton, J. and Hakim, G. (2012).
An Introduction to Dynamic Meteorology . Academic Press,New York, 5 edition.Horinouchi, T. (2012). Moist hadley circulation: Possible role of wave–convection coupling inaquaplanet experiments.
Journal of the atmospheric sciences , 69(3):891–907.Kang, W., Cai, M., and Tziperman, E. (2019). Tropical and extratropical general circulationwith a meridional reversed temperature gradient as expected in a high obliquity planet.
Icarus ,330:142–154.Khairoutdinov, M. and Emanuel, K. (2013). Rotating radiative-convective equilibrium simu-lated by a cloud-resolving model.
Journal of Advances in Modeling Earth Systems , 5(4):816–825.Kirtman, B. and Schneider, E. (2000). A spontaneously generated tropical atmospheric generalcirculation.
Journal of the atmospheric sciences , 57(13):2080–2093.Kitamura, Y. and Matsuda, Y. (2006). The kh- 3 and kh- 5/3 energy spectra in stratifiedturbulence.
Geophysical research letters , 33(5).Koll, D. and Cronin, T. (2018). Earth’s outgoing longwave radiation linear due to h o greenhouseeffect. Proceedings of the National Academy of Sciences , 115:10293–10298.Kushwaha, P., Sukhatme, J., and Nanjundiah, R. (2020). A global survey of mid-troposphericcyclones.
Monthy Weather Review , in revision.Lee, S. (1999). Why are the climatological zonal winds easterly in the equatorialupper tropo-sphere?
Journal of the atmospheric sciences , 56(10):1353–1363.Levine, X. and Schneider, T. (2015). Baroclinic eddies and the extent of the hadley circulation:An idealized gcm study.
Journal of the Atmospheric Sciences , 72:2744–2761.Lindborg, E. (2015). A helmholtz decomposition of structure functions and spectra calculatedfrom aircraft data.
Journal of Fluid Mechanics , 762:R4.Lutsko, N. (2018). The response of an idealized atmosphere to localized tropical heating:Superrotation and the breakdown of linear theory.
Journal of the Atmospheric Sciences ,75(1):3–20.Maher, P. et al. (2019). Model hierarchies for understanding atmospheric circulation.
Reviewsof Geophysics , 57(2):250–280.Medeiros, B., Williamson, D., and Olson, J. (2016). Reference aquaplanet climate in the commu-nity atmosphere model, version 5.
Journal of Advances in Modeling Earth Systems , 8(1):406–424.Merlis, T. et al. (2016). Surface temperature dependence of tropical cyclone-permitting sim-ulations in a spherical model with uniform thermal forcing.
Geophysical Research Letters ,43(6):2859–2865.Merlis, T. and Held, I. (2019). Aquaplanet simulations of tropical cyclones.
Current ClimateChange Reports , 5(3):185–195.Mitchell, J. and Lora, J. (2016). The climate of titan.
Annual Review of Earth and PlanetarySciences , 44:353–380. 24roweic, A., Garner, S., and Paulius, O. (2011). Axisymmetric hurricane in a dry atmosphere:Theoretical framework and numerical experiments.
Journal of the Atmospheric Sciences ,68:1607–1619.Muller, C. and Held, I. (2012). Detailed investigation of the self-aggregation of convection incloud-resolving simulations.
Journal of the Atmospheric Sciences , 69(8):2551–2565.Muller, C. and Romps, D. (2018). Acceleration of tropical cyclogenesis by self-aggregationfeedbacks.
Proceedings of the National Academy of Sciences , 115(12):2930–2935.Nastrom, G. and Gage, K. (1985). A climatology of atmospheric wavenumber spectra ofwind and temperature observed by commercial aircraft.
Journal of the Atmosheric Sciences ,42:950–960.O’Gorman, P. and Schneider, T. (2006). Stochastic models for the kinematics of moisturetransport and condensation in homogeneous turbulent flows.
Journal of the atmosphericsciences , 63(11):2992–3005.Pierrehumbert, R. (2002). The hydrologic cycle in deep-time climate problems.
Nature , 419:191–198.Pierrehumbert, R. (2010).
Principles of Planetary Climate . Cambridge University Press, Cam-bridge, UK, 1 edition.Pierrehumbert, R., Brogniez, H., and Roca, R. (2007).
On the relative humidity of the Earth’satmosphere . Princeton University Press, Princeton, NJ.Ram´ırez-Reyes, A. and Yang, D. (2020). Spontaneous cyclogenesis without radiative andsurface-flux feedbacks. arXiv preprint arXiv:2004.08662 .Raymond, D. and Fuchs, ˇZ. (2009). Moisture modes and the madden–julian oscillation.
Journalof Climate , 22(11):3031–3046.Read, P. and Lewis, S. (2004).
The Martian Climate Revisited . Springer Verlag, Berlin, 1edition.Rocha, C., Chereskin, T., and Gille, S. (2016). Mesoscale to submesoscale wavenumber spectrain drake passage.
Journal of Physical Oceanography , 46:601–620.Schneider, T. and Bordoni, S. (2008). Eddy-mediated regime transitions in the seasonal cycleof a hadley circulation and implications for monsoon dynamics.
Journal of the AtmosphericSciences , 65(3):915–934.Shi, X. and Bretherton, C. (2014). Large-scale character of an atmosphere in rotating radiative-convective equilibrium.
Journal of Advances in Modeling Earth Systems , 6(3):616–629.Shi, X. et al. (2018). Wishe-moisture mode in an aquaplanet simulation.
Journal of advancesin modeling earth systems , 10(10):2393–2407.Sobel, A. (2002). Water vapor as an active scalar in tropical atmospheric dynamics.
Chaos: AnInterdisciplinary Journal of Nonlinear Science , 12(2):451–459.Sobel, A. and Maloney, E. (2013). Moisture modes and the eastward propagation of the mjo.
Journal of the Atmospheric Sciences , 70(1):187–192.Suhas, D. and Sukhatme, J. (2020). Moist shallow water response to tropical forcing: Initialvalue problems.
Quarterly Journal of the Royal Meteorological Society .25ukhatme, J. et al. (2020). Near-surface ocean kinetic energy spectra and small-scale intermit-tency from ship-based adcp data in the bay of bengal.
Journal of Physical Oceanography ,50:2037–2052.Sukhatme, J. and Smith, L. (2008). Vortical and wave modes in 3d rotating stratified flows:random large-scale forcing.
Geophysical and Astrophysical Fluid Dynamics , 102(5):437–455.Sukhatme, J. and Young, W. (2011). The advection–condensation model and water-vapourprobability density functions.
Quarterly Journal of the Royal Meteorological Society ,137(659):1561–1572.Sumi, A. (1992). Pattern formation of convective activity over the aqua-planet with globallyuniform sea surface temperature (sst).
Journal of the Meteorological Society of Japan. Ser.II , 70(5):855–876.Toniazzo, T. et al. (2020). Enforcing conservation of axial angular momentum in the atmosphericgeneral circulation model cam6.
Geosci. Model Dev. , 13:685–705.Trenberth, K. and Stepaniak, D. (2003). Covariability of components of poleward atmosphericenergy transports on seasonal and interannual timescales.
Journal of Climate , 16:3691–3705.Vallgren, A., Deusebio, E., and Lindborg, E. (2011). Possible explanation of the atmospherickinetic and potential energy spectra.
Physical Review Letters , 107(26):268501.Walker, C. and Schneider, T. (2005). Response of idealized hadley circulations to seasonallyvarying heating.
Geophysical research letters , 32(6).Walker, C. and Schneider, T. (2006). Eddy influences on hadley circulations: Simulations withan idealized gcm.
Journal of the atmospheric sciences , 63(12):3333–3350.Wang, D. and Lin, Y. (2020). Size and structure of dry and moist reversible tropical cyclones.
Journal of the Atmospheric Sciences , 77:2091–2114.Wang, X. and Jiang, H. (2019). A 13-year global climatology of tropical cyclone warm-corestructures from airs data.
Monthly Weather Review , 147(3):773–790.Wheeler, M. and Kiladis, G. (1999). Convectively coupled equatorial waves: Analysis of cloudsand temperature in the wavenumber–frequency domain.
Journal of the Atmospheric Sciences ,56(3):374–399.Wheeler, M., Kiladis, G., and Webster, P. (2000). Large-scale dynamical fields associated withconvectively coupled equatorial waves.
Journal of the Atmospheric Sciences , 57(5):613–640.Wing, A. and Emanuel, K. (2014). Physical mechanisms controlling self-aggregation of con-vection in idealized numerical modeling simulations.
Journal of Advances in Modeling EarthSystems , 6(1):59–74.Wing, A. et al. (2017). Convective self-aggregation in numerical simulations: A review. In
Shallow Clouds, Water Vapor, Circulation, and Climate Sensitivity , pages 1–25. Springer.Zhou, W., Held, I., and Garner, S. (2014). Parameter study of tropical cyclones in rotatingradiative–convective equilibrium with column physics and resolution of a 25-km gcm.