betaFeedback

WorksPhysics and Astronomy

Condensation-related expansion work of water vapor as a major pathway for kinetic energy generation in the lower atmosphere

Abstract

Earth's atmosphere operates a steam cycle: water evaporates from the surface, expands as vapor, condenses in colder air, and returns as precipitation. Yet its power and relationship to atmospheric mechanical power remain poorly quantified. Applying the Clausius–Clapeyron relation between saturated vapor expansion work and latent heat converted at Carnot efficiency to vertically distributed condensation, we derive the steam-engine power as a sum of the gravitational and condensation-related contributions, Wv=Wg+WcW_v=W_g+W_c; it is not the net power of a complete moist-air cycle. Using twelve years of GPM/DPR spectral latent-heating profiles and MERRA-2 fields, we estimate Wv≃4.1W_v\simeq4.1 W m−2^{-2}, with Wc≃2.8W_c\simeq2.8 W m−2^{-2}. Separately, we diagnose total atmospheric mechanical power W=WP+WK≃4.3W=W_P+W_K\simeq4.3 W m−2^{-2}, comprising gravitational power of precipitation WP≃1.1W_P\simeq1.1 W m−2^{-2} and kinetic energy generation WK≃3.2W_K\simeq3.2 W m−2^{-2}, including 2.42.4 W m−2^{-2} below 500 hPa. Both correspondences Wv≃WW_v\simeq W and Wc≃WKW_c\simeq W_K lie within 20–30% uncertainty estimates. To explain these correspondences, we propose that WcW_c generates the potential energy associated with surface pressure gradients. Cross-isobaric flow in the frictional lower atmosphere converts this energy into kinetic energy. In a horizontally isothermal atmosphere, this positive pressure work would be nearly cancelled by compression aloft. An idealized circulation illustrates how differential heating can offset compression, allowing nearly isobaric upper return flow. In the proposed mechanism, condensation-driven circulation unlocks this thermal compensation thus providing a major pathway for lower-atmospheric kinetic energy generation. We discuss how changes in the coupling between condensation and differential heating could test this proposed causal relationship.

Introduction

Understanding the causes of changes in atmospheric circulation and their climatic consequences remains a major challenge in climate science ([42, 41]). An important part of this problem is identifying the physical mechanisms of atmospheric kinetic energy generation, particularly those associated with water vapor phase transitions. The global mean latent heat flux is approximately 80 W m−2^{-2}, nearly four times the sensible heat flux ([7]). While these energy fluxes are relatively well quantified, establishing a robust quantitative relationship between water vapor phase transitions and atmospheric kinetic energy generation remains a challenge.

Currently, the magnitude of atmospheric mechanical power remains uncertain. Among more recent estimates, Paulus and Dias (2012) stated that the total power associated with atmospheric motions is not known precisely but is likely on the order of 5 W m−2^{-2}. Paulus (2015) cited the results of Laliberté et al. [17] as supporting this magnitude. In contrast, Bohren and Albrecht [4] reported a broader range of 2–5 W m−2^{-2}, whereas Kleidon [16] argued, on theoretical grounds, that the rate should not exceed 2 W m−2^{-2}. Finally, observational estimates of Lorenz available potential energy generation yielded only about 1.2 W m−2^{-2} [40]. Part of this discrepancy concerns the distinction between kinetic energy generation by atmospheric winds, WKW_{\mathrm{K}}, and total atmospheric power, W=WK+WPW = W_{\mathrm{K}} + W_{\mathrm{P}}, which also includes the gravitational power required to lift precipitating water, WPW_{\mathrm{P}}, of approximately 1 W m−2^{-2} (Pauluis and Dias, 2012). Yet a robust observational assessment of both contributions using independent diagnostics remains lacking, thereby preventing a reliable validation of theoretical approaches against observations.

Theoretical predictions of atmospheric power generation are also subject to considerable uncertainty. In moist heat-engine formulations, mechanical work is commonly evaluated using entropy budgets that depend on irreversible processes associated with water vapor diffusion and cloud microphysics [44] (Pauluis and Held 2002). These processes are difficult to constrain independently, complicating the observational validation of theoretical power estimates [15]. This highlights the need for transparent, quantitative predictions that can be tested against independent observations.

An important finding of moist atmospheric heat-engine theory is that moisture-related irreversible processes can substantially reduce thermodynamic efficiency below the Carnot limit (Pauluis and Held, 2002; Pauluis, 2011). However, a lower efficiency does not necessarily imply lower mechanical power. The maximum Carnot efficiency corresponds to the reversible limit of an infinitely slow cycle, while the power generated by an actual circulation depends on both its efficiency and the rate of energy transfer. Furthermore, latent and sensible heat fluxes are not independent of atmospheric circulation. They depend on winds and atmospheric temperature and humidity distributions, and could be substantially different in the absence of the water cycle. Thus, understanding the role of water vapor phase transitions in atmospheric power generation requires us to consider not only how moisture affects the efficiency of a given thermodynamic cycle, but also what circulations become possible when the water cycle is present.

A different perspective on the role of moisture in atmospheric power generation dates back to Heinrich Hertz, who in 1885 proposed that the atmosphere operates as a steam engine, with evaporation at the Earth’s surface and condensation in clouds corresponding to the boiler and condenser, respectively [30]. Hertz estimated the mean temperature difference between evaporation and condensation to be approximately 15 K. At a temperature of about 300 K, this corresponds to a Carnot efficiency of approximately 5%. Applying this efficiency to the present-day global latent heat flux of about 80 W m−2^{-2} yields a steam-engine power of approximately 4 W m−2^{-2}, remarkably close to the estimates of total atmospheric mechanical power discussed above. In view of the robust finding that the water cycle reduces thermodynamic efficiency relative to the Carnot limit, an important question is what underlying physical relationship, if any, could account for this numerical correspondence. In this paper, we assess this numerical correspondence using modern observations and investigate its possible physical basis.

A possible physical mechanism linking the atmospheric water cycle to kinetic energy generation has been proposed within condensation-induced atmospheric dynamics (CIAD), developed since 2007 (for a review, see Makarieva et al., [26]). Condensation removes water vapor from the gas phase, introducing a mass sink into the gaseous continuity equation, while precipitation removes condensed water from the atmospheric column. Together with evaporation and atmospheric transport, these processes can maintain horizontal differences in atmospheric column mass and surface pressure, thereby contributing to kinetic energy generation in the lower atmosphere.

However, the relationship between condensation rate and power generation, while yielding meaningful predictions across diverse meteorological contexts, has remained largely heuristic. Its connection to the thermodynamic expansion work of the atmospheric steam cycle has not been explicitly established. This raises a broader question: does the water cycle determine atmospheric circulation, or does it primarily modify the efficiency of circulation driven by differential heating?

To address these questions, § 2 derives atmospheric steam-engine power from the Clausius–Clapeyron law, separates its gravitational and condensation-related components, and defines the corresponding contributions to atmospheric mechanical power. § 3 describes the observational data and diagnostic methods. § 4 presents the global power estimates and their uncertainties, and compares steam-engine and atmospheric mechanical power. In § 5, we examine the relationship to condensation-induced atmospheric dynamics, investigate how differential heating can prevent the cancellation of lower-level pressure work by compression aloft, and discuss thermodynamic efficiency, causality, and implications for predicting atmospheric circulation. Appendices § 6 and § 7 provide supporting derivations and computational details, while Appendix § 8 develops a simplified mass-related Lorenz-type available-energy framework to examine the relationship between surface pressure gradients and differential heating.

Water-vapor expansion and atmospheric mechanical power

Clausius–Clapeyron work and water-vapor expansion power

The Clausius–Clapeyron relation for the saturated water-vapor pressure pv∗p_v^* is usually written as

dpv∗dT=Lpv∗RT2,(1)\frac{\mathrm{d}p_v^*}{\mathrm{d}T} = \frac{\mathcal{L}p_v^*}{RT^2}, \tag*{(1)}

or, equivalently,

RTpv∗ dpv∗=LdTT,(2)\frac{RT}{p_v^*}\,\mathrm{d}p_v^* = \mathcal{L}\frac{\mathrm{d}T}{T}, \tag*{(2)}

where RR is the universal gas constant and L\mathcal{L} is the molar latent heat of vaporization. The corresponding mass-specific latent heat is L=L/MvL = \mathcal{L}/M_v, where MvM_v is the molar mass of water. In this form, the equation relates the expansion work of saturated water vapor to latent heat converted at Carnot efficiency.

The Clausius–Clapeyron equation is commonly derived from the equality of Gibbs free energies, or chemical potentials, of liquid water and vapor ([1, 4]). An alternative classical derivation considers an infinitesimal reversible Carnot cycle ([5, 18, 48]), whose net work is ΔV Δpv∗\Delta V\,\Delta p_v^*. Neglecting the liquid-water volume and assuming ideal-vapor behavior, ΔV≃RT/pv∗\Delta V \simeq RT/p_v^*. Equating this work to L ΔT/T\mathcal{L}\,\Delta T/T gives Eq. (2).

Equation (2) provides the starting point for evaluating the expansion work of saturated water vapor in the atmosphere. Since condensation occurs over a range of temperatures and along many trajectories, we weight the work by the amount of water condensing at each temperature. Let C≥0C \ge0 and E≥0E \ge0 denote the local rates of net condensation and net evaporation of water vapor, respectively (kg m−3 s−1\mathrm{kg\ m^{-3}\ s^{-1}}). These rates represent opposite signs of the local net phase change: at any given point and instant, at most one of them is nonzero1.

Consider the atmospheric volume Ω\Omega above a reference surface area SS. For any quantity XX associated with a condensation event, the condensation-weighted mean is defined as

⟨X⟩c≡∫ΩCX‾ dΩ∫ΩC‾ dΩ.(3)\langle X\rangle_c \equiv\frac{\int_{\Omega}\overline{CX}\,\mathrm{d}\Omega}{\int_{\Omega}\overline{C}\,\mathrm{d}\Omega}. \tag*{(3)}
Pc≡1S∫ΩC‾ dΩ.(4)P_c \equiv\frac{1}{S}\int_{\Omega}\overline{C}\,\mathrm{d}\Omega. \tag*{(4)}

Here an overbar denotes time averaging over the period considered. The condensation rate per unit surface area PcP_c need not equal the surface precipitation PP.

The water vapor condensing at temperature TT has previously evaporated elsewhere and, before condensing, reached saturation at temperature T1T_1. We refer to T1T_1 as the initial saturation temperature associated with this condensation contribution. Under saturated conditions, vapor pressure is uniquely determined by temperature. Consequently, the expansion work between the initial saturated state and the condensation state calculated from the Clausius–Clapeyron relation depends only on T1T_1 and TT, irrespective of the particular atmospheric trajectory connecting these states.

For each condensation contribution, integrating Eq. (2) from TT to T1T_1 gives the expansion work per mole of water vapor as the saturation pressure decreases from its initial value at T1T_1 to that at TT. Multiplying this work by the local molar condensation rate C/MvC/M_v, averaging over time, and integrating over the atmospheric volume, we obtain the water-vapor expansion power per unit surface area:

Wv=1SMv∫ΩC‾∫TT1L(T′)dT′T′ dΩ=PcMv⟨∫TT1L(T′)dT′T′⟩c≃LPc⟨ln⁡T1T⟩c,(5)\begin{aligned} W_v &= \frac{1}{SM_v}\int_{\Omega}\overline{C}\int_T^{T_1}\mathcal{L}(T')\frac{\mathrm{d}T'}{T'}\,\mathrm{d}\Omega\\ &= \frac{P_c}{M_v}\left\langle\int_T^{T_1}\mathcal{L}(T')\frac{\mathrm{d}T'}{T'}\right\rangle_c \\ &\simeq LP_c\left\langle\ln\frac{T_1}{T}\right\rangle_c, \tag*{(5)} \end{aligned}

where the final approximation neglects the temperature dependence of L\mathcal{L}.

Equation (5) relates water-vapor expansion power to the condensation rate and the temperatures of the initial saturated and condensation states. The distribution of condensation temperatures can be estimated from GPM/DPR latent-heating profiles combined with atmospheric temperature data, while the initial saturation temperatures can be estimated from lifting-condensation levels derived from MERRA-2. These estimates provide the basis for calculating WvW_v from observations, as described in the Methods section.

Separation of gravitational and condensation-related expansion work

We now separate water-vapor expansion work into gravitational and condensation-related components. Writing γ≡pv∗/pd\gamma\equiv p_v^*/p_d, where pdp_d is dry-air partial pressure, gives the identity

dln⁡pv∗=dln⁡pd+dln⁡γ.(6)\mathrm{d}\ln p_v^*=\mathrm{d}\ln p_d+\mathrm{d}\ln\gamma. \tag*{(6)}

Thus the decrease of saturation vapor pressure can be separated into changes of dry-air pressure and of the vapor-to-dry-air pressure ratio.

would be zero in steady state. An appropriate spatial scale should therefore be chosen to capture the separation of condensation and evaporation occurring at different environmental temperatures. Hydrostatic equilibrium relates the vertical variation of air pressure to height:

RT∂ln⁡p∂z=−Mg,(7)RT\frac{\partial\ln p}{\partial z}=-Mg, \tag*{(7)}

where MM is the mean molar mass of gaseous air, taken approximately constant, and gg is gravitational acceleration. To accuracy O(γ)O(\gamma), ∂zln⁡pd≃∂zln⁡p\partial_z\ln p_d\simeq\partial_z\ln p.

Let zz and z1z_1 denote the heights of the condensation and initial saturated states, respectively. Evaluating the vertical dry-air pressure difference in the local hydrostatic column for each condensation event gives its gravitational contribution to the expansion power:

Wg≡−RPcMv⟨∫z1zT(z′)∂ln⁡pd∂z′ dz′⟩c≃MMvgPc(zc−⟨z1⟩c).(8)\begin{aligned} W_g\equiv-\frac{RP_c}{M_v}\left\langle\int_{z_1}^{z}T(z')\frac{\partial\ln p_d}{\partial z'}\,\mathrm{d}z'\right\rangle_c \\ &\simeq\frac{M}{M_v}gP_c(z_c-\langle z_1\rangle_c). \tag*{(8)} \end{aligned}

where zc≡⟨z⟩cz_c\equiv\langle z\rangle_c is the mean height of condensation. While WvW_v depends on condensation temperatures, WgW_g depends on condensation heights, which we also estimate using GPM/DPR and MERRA-2 data.

The remaining condensation-related power is

Wc≡Wv−Wg≃LPc⟨ln⁡T1T⟩c−MMvgPc(zc−⟨z1⟩c).(9)\begin{aligned} W_c\equiv W_v-W_g \\ &\simeq LP_c\left\langle\ln\frac{T_1}{T}\right\rangle_c-\frac{M}{M_v}gP_c(z_c-\langle z_1\rangle_c). \tag*{(9)} \end{aligned}

Thus, the water-vapor expansion power decomposes as

Wv=Wg+Wc.(10)W_v=W_g+W_c. \tag*{(10)}

To clarify the physical meaning of WcW_c, we use Eqs. (2) and (6) to write the expansion work per mole associated with each condensation contribution as

∫TT1L(T′)dT′T′=−R∫pd(z1)pdT dln⁡pd′−R∫γ1γT dln⁡γ′,(11)\int_T^{T_1}\mathcal{L}(T')\frac{\mathrm{d}T'}{T'}=-R\int_{p_d(z_1)}^{p_d}T\,\mathrm{d}\ln p'_d-R\int_{\gamma_1}^{\gamma}T\,\mathrm{d}\ln\gamma', \tag*{(11)}

where the integrals on the right are taken between the initial saturated state and the condensation state. The vertical hydrostatic contribution of the first integral, after condensation weighting and multiplication by Pc/MvP_c/M_v, gives WgW_g in Eq. (8). For the idealized calculation below, we neglect horizontal and temporal changes in dry-air pressure. The remaining condensation-related power is then

Wc≃−RPcMv⟨∫γ1γT(γ′) dln⁡γ′⟩c.(12)W_c\simeq-\frac{RP_c}{M_v}\left\langle\int_{\gamma_1}^{\gamma}T(\gamma')\,\mathrm{d}\ln\gamma'\right\rangle_c. \tag*{(12)}

Consider an idealized saturated trajectory with a fixed amount of dry air, where γ\gamma decreases from γ1\gamma_1 to γ2\gamma_2 solely through condensation. Since γ\gamma equals the molar ratio of water vapor to dry air, the amount condensed is proportional to −dγ-\mathrm{d}\gamma. Approximating the temperature in the preceding integral by Tc≡⟨T⟩cT_c\equiv\langle T\rangle_c, we obtain

Wc≃RPcTcMv(γ1−γ2)∫γ2γ1ln⁡γ1γ dγ=RPcTcMv(1+κ),(13)\begin{aligned} W_c\simeq\frac{RP_cT_c}{M_v(\gamma_1-\gamma_2)}\int_{\gamma_2}^{\gamma_1}\ln\frac{\gamma_1}{\gamma}\,\mathrm{d}\gamma\\ &=\frac{RP_cT_c}{M_v}(1+\kappa), \tag*{(13)} \end{aligned}
κ≡γ2/γ11−γ2/γ1ln⁡γ2γ1≤0.\kappa\equiv\frac{\gamma_2/\gamma_1}{1-\gamma_2/\gamma_1}\ln\frac{\gamma_2}{\gamma_1}\leq0.

The ratio γ2/γ1\gamma_2/\gamma_1 represents the fraction of water vapor remaining uncondensed2. For γ2/γ1=0.3\gamma_2/\gamma_1 = 0.3, κ≃−0.52\kappa\simeq-0.52, while κ→0\kappa\to0 for complete condensation. In the opposite limit, γ2/γ1→1\gamma_2/\gamma_1 \to1, condensation vanishes, κ→−1\kappa\to-1, and Wc→0W_c \to0. With no condensation, the condensation-weighted mean quantities become undefined. Water vapor then behaves as a noncondensable gas, and separating its expansion work from that of the other atmospheric gases has no physical meaning.

For the atmosphere as a whole, an effective value of κ\kappa can be estimated from the condensation-related power WcW_c, the condensation rate per unit surface area PcP_c, and the mean condensation temperature TcT_c:

κeff=MvWcRPcTc−1=Mv(Wv−Wg)RPcTc−1.(14)\kappa_{\mathrm{eff}} = \frac{M_v W_c}{R P_c T_c} - 1 = \frac{M_v(W_v-W_g)}{R P_c T_c} - 1. \tag*{(14)}

Mechanical atmospheric power: W=WP+WKW=W_P+W_K

Atmospheric pressure work per unit surface area is

W=1S∫Ωp∇⋅v‾ dΩ=−1S∫Ωv⋅∇p‾ dΩ,(15)W = \frac{1}{S}\int_{\Omega} \overline{p\nabla\cdot\mathbf{v}}\,\mathrm{d}\Omega= -\frac{1}{S}\int_{\Omega} \overline{\mathbf{v}\cdot\nabla p}\,\mathrm{d}\Omega, \tag*{(15)}

where v=u+wez\mathbf{v}=\mathbf{u}+w\mathbf{e}_z is gaseous-air velocity, SS is the reference surface area, and the boundary flux of pvp\mathbf{v} is taken to vanish. Separating the vertical and horizontal contributions to pressure work and using hydrostatic equilibrium, we obtain

W=WP+WK,(16)W=W_P+W_K, \tag*{(16)}

where

WP=gPc(zc−ze)(17)W_P=gP_c(z_c-z_e) \tag*{(17)}

is the gravitational contribution to pressure work, and

WK=−1S∫Ωu⋅∇hp‾ dΩ(18)W_K=-\frac{1}{S}\int_{\Omega}\overline{\mathbf{u}\cdot\nabla_h p}\,\mathrm{d}\Omega \tag*{(18)}

is kinetic energy generation by horizontal pressure gradients. Here PcP_c, defined in Eq. (4), is the condensation rate per unit surface area, zcz_c and zez_e are the corresponding condensation- and evaporation-weighted heights (Appendix § 6).

In pressure coordinates, the evaluation of total atmospheric pressure work WW leads to the so-called αω\alpha\omega formulation, where α≡1/ρ\alpha\equiv1/\rho is specific volume and ω≡Dp/Dt\omega\equiv Dp/Dt is pressure vertical velocity. This formulation involves cancellation between large positive and negative contributions, so even small numerical errors in ω\omega or mass continuity may strongly affect the global residual ([46, 50]). Here we diagnose WKW_K directly from horizontal pressure gradients and calculate WPW_P from the height-weighted local net condensation and evaporation rates CC and EE.

The gravitational contribution WgW_g to water-vapor expansion is distinct from WPW_P. From Eqs. (8) and (17),

WgWP≃MMvzc−⟨z1⟩czc−ze.(19)\frac{W_g}{W_P}\simeq\frac{M}{M_v}\frac{z_c-\langle z_1\rangle_c}{z_c-z_e}. \tag*{(19)}

This ratio reduces to M/Mv≃1.6M/M_v\simeq1.6 when ⟨z1⟩c=ze\langle z_1\rangle_c=z_e, i.e. when water vapor enters saturation at the same mean height where it evaporated, which is not generally the case.

Data and methods

We estimate WvW_{\mathrm{v}}, WgW_{\mathrm{g}}, and WPW_{\mathrm{P}} from the GPM Dual-frequency Precipitation Radar (DPR) spectral latent-heating product (SLH V07; [10]), together with contemporaneous three-hourly MERRA-2 pressure-level data ([8]), and obtain Wc=Wv−WgW_{\mathrm{c}} = W_{\mathrm{v}} - W_{\mathrm{g}}. The analysis covers the entire period of availability of SLH V07, from 8 March 2014 to 28 February 2026, comprising twelve March–February periods.

The GPM analysis covers 65°S–65°N. To account for unequal latitude sampling, we first average within 5° latitude bands and then weight the results by the area of each band relative to the entire Earth’s surface area. Thus, the resulting fluxes and powers are normalized by the total Earth’s area, while contributions poleward of 65° are not included.

The GPM/DPR radar scans swaths approximately 250–280 km wide, with 49 footprints per cross-track scan. Each footprint is about 5 km across (approximately 25 km2^2), and SLH provides temperature tendencies T˙\dot{T} in 80 layers of 250 m thickness. For each footprint, we take MERRA-2 temperature, pressure, humidity, and density from the nearest grid point, interpolating between consecutive three-hourly records and to the SLH height levels. We estimate the local rates of net condensation CC and atmospheric evaporation EaE_{\mathrm{a}} as

C≃ρcpLmax⁡(T˙,0),Ea≃ρcpLmax⁡(−T˙,0),(20)C \simeq\frac{\rho c_{\mathrm{p}}}{L}\max(\dot{T},0), \qquad E_{\mathrm{a}} \simeq\frac{\rho c_{\mathrm{p}}}{L}\max(-\dot{T},0), \tag*{(20)}

where T˙\dot{T} is in K s−1^{-1}, ρ\rho is MERRA-2 air density, cp=1004c_{\mathrm{p}} = 1004 J kg−1^{-1} K−1^{-1}, and L=2.5×106L = 2.5 \times10^{6} J kg−1^{-1}. We approximate the retrieved heating by vapor–liquid phase change, without separately removing contributions from freezing and melting.

Surface evaporation is not retrieved by SLH and is inferred from steady-state mass balance. With Pe,aP_{\mathrm{e,a}} and Pe,sP_{\mathrm{e,s}} denoting atmospheric and surface evaporation rates per unit area, we define the signed latent-heating contributions

Qc=LPc,Qe,a=−LPe,a,Qe,s=−LPe,s=−(Qc+Qe,a),(21)\begin{aligned} Q_{\mathrm{c}} &= LP_{\mathrm{c}},\\ Q_{\mathrm{e,a}} &= -LP_{\mathrm{e,a}},\\ Q_{\mathrm{e,s}} &= -LP_{\mathrm{e,s}} = -(Q_{\mathrm{c}} + Q_{\mathrm{e,a}}), \tag*{(21)} \end{aligned}

using Pc=Pe,a+Pe,sP_{\mathrm{c}} = P_{\mathrm{e,a}} + P_{\mathrm{e,s}}. To account for unequal latitude sampling, we average within 5° latitude bands and then weight by band area over 65°S–65°N. The twelve-period Earth-area-normalized estimates are Qc≃74Q_{\mathrm{c}} \simeq74, Qe,a≃−16Q_{\mathrm{e,a}} \simeq-16, and Qe,s≃−58Q_{\mathrm{e,s}} \simeq-58 W m−2^{-2} (Figure 1).

Figure 1. Vertical arrangement of condensation, evaporation, and estimates of initial saturation. The dashed lines indicate the lifting condensation levels of surface air and air at atmospheric evaporation levels, used in the two-LCL estimate. The dotted line indicates zc,minz_{\mathrm{c,min}}, the positive-heating-weighted mean of the lowest levels of positive SLH heating in individual footprints, used in the one-onset estimate.

The heights used to calculate WPW_{\mathrm{P}} are expressed above mean sea level (ASL), as in Figure 1. The corresponding mean heights are zc≃4.44z_{\mathrm{c}} \simeq4.44 km for condensation, ze,a≃2.35z_{\mathrm{e,a}} \simeq2.35 km for atmospheric evaporation, and ze,s≃0.13z_{\mathrm{e,s}} \simeq0.13 km for inferred surface evaporation, estimated from the mean local surface elevation. Thus,

ze=Pe,aze,a+Pe,sze,sPc≃0.61 km.(22)z_{\mathrm{e}} = \frac{P_{\mathrm{e,a}}z_{\mathrm{e,a}} + P_{\mathrm{e,s}}z_{\mathrm{e,s}}}{P_{\mathrm{c}}} \simeq0.61\ \mathrm{km}. \tag*{(22)}

Equation (17) then gives WP=gPc(zc−ze)≃1.11W_{\mathrm{P}} = gP_{\mathrm{c}}(z_{\mathrm{c}} - z_{\mathrm{e}}) \simeq1.11 W m−2^{-2}. Setting ze,s=0z_{\mathrm{e,s}} = 0 instead increases WPW_{\mathrm{P}} to 1.14 W m−2^{-2}, a difference of approximately 3%.

We evaluate the initial saturated state of condensing vapor in two ways (Figure 1). In the two-LCL estimate, vapor supplied by surface evaporation, with fraction −Qe,s/Qc-Q_{\mathrm{e,s}}/Q_{\mathrm{c}}, is assigned the lifting condensation level (LCL) calculated from local surface-air temperature, pressure, and humidity at each condensation footprint. Vapor supplied by atmospheric evaporation, with fraction −Qe,a/Qc-Q_{\mathrm{e,a}}/Q_{\mathrm{c}}, is assigned the LCL calculated from the local air conditions at the atmospheric evaporation levels. The LCL is the height where an adiabatically lifted parcel first becomes saturated. Denoting the geometric-mean LCL temperatures weighted by positive and negative SLH heating magnitudes by T1,sT_{1,\mathrm{s}} and T1,aT_{1,\mathrm{a}}, respectively, we estimate

⟨ln⁡T1Tref⟩c≃−Qe,sQcln⁡T1,sTref−Qe,aQcln⁡T1,aTref−⟨ln⁡TTref⟩c,(23)\left\langle\ln\frac{T_1}{T_{\mathrm{ref}}} \right\rangle_{\mathrm{c}} \simeq-\frac{Q_{\mathrm{e,s}}}{Q_{\mathrm{c}}}\ln\frac{T_{1,\mathrm{s}}}{T_{\mathrm{ref}}}-\frac{Q_{\mathrm{e,a}}}{Q_{\mathrm{c}}}\ln\frac{T_{1,\mathrm{a}}}{T_{\mathrm{ref}}}-\left\langle\ln\frac{T}{T_{\mathrm{ref}}}\right\rangle_{\mathrm{c}}, \tag*{(23)}

where TT is MERRA-2 temperature at each positive SLH level, the last term is weighted by positive SLH heating, and the arbitrary TrefT_{\mathrm{ref}} cancels. Equations (5) and (8) then give WvW_{\mathrm{v}} and WgW_{\mathrm{g}}, with the latter using

⟨z1⟩c≃−Qe,sQcz1,s−Qe,aQcz1,a.(24)\langle z_1\rangle_{\mathrm{c}} \simeq-\frac{Q_{\mathrm{e,s}}}{Q_{\mathrm{c}}}z_{1,\mathrm{s}}-\frac{Q_{\mathrm{e,a}}}{Q_{\mathrm{c}}}z_{1,\mathrm{a}}. \tag*{(24)}

Here z1,sz_{1,\mathrm{s}} and z1,az_{1,\mathrm{a}} are the corresponding heating-weighted mean LCL heights.

For z1,s≃0.7 kmz_{1,\mathrm{s}}\simeq0.7\ \mathrm{km} and z1,a≃2.9 kmz_{1,\mathrm{a}}\simeq2.9\ \mathrm{km}, we find ⟨z1⟩c≃1.2 km\langle z_1\rangle_{\mathrm{c}}\simeq1.2\ \mathrm{km} and Wg≃1.5 W m−2W_{\mathrm{g}}\simeq1.5\ \mathrm{W\ m^{-2}}.

In the one-onset estimate, we instead estimate the mean initial saturation state of vapor supplied by both surface and atmospheric evaporation using the lowest level of positive SLH heating. In each footprint containing positive SLH heating, we identify the lowest level at which T˙>0\dot{T}>0. The positive-heating-weighted mean of these footprintwise levels is zc,minz_{\mathrm{c,min}}; Tc,minT_{\mathrm{c,min}} is the corresponding positive-heating-weighted geometric mean of MERRA-2 temperatures. We estimate

⟨ln⁡T1Tref⟩c≃ln⁡Tc,minTref,⟨z1⟩c≃zc,min.(25)\left\langle\ln\frac{T_1}{T_{\mathrm{ref}}}\right\rangle_{\mathrm{c}}\simeq\ln\frac{T_{\mathrm{c,min}}}{T_{\mathrm{ref}}},\qquad\langle z_1\rangle_{\mathrm{c}}\simeq z_{\mathrm{c,min}}. \tag*{(25)}

The twelve-period means are zc,min≃1.9 kmz_{\mathrm{c,min}}\simeq1.9\ \mathrm{km} and Tc,min≃284.6 KT_{\mathrm{c,min}}\simeq284.6\ \mathrm{K}.

We estimate WKW_{\mathrm{K}} directly from the instantaneous three-hourly MERRA-2 pressure-level wind and geopotential height fields. The local kinetic-energy generation per unit mass is

cK=−u⋅∇pΦc_K = -\mathbf{u} \cdot\nabla_p \Phi, where Φ=gZ\Phi= gZ and ZZ is geopotential height (the MERRA-2 HH field). We evaluate horizontal geopotential gradients on pressure surfaces and integrate cKc_K over the atmospheric mass column. For the unresolved surface layer, we interpolate cKc_K linearly in pressure from zero at the surface to its value at the lowest valid pressure level; alternative extrapolations are detailed in Appendix § 7.

OpenAI ChatGPT was used as an interactive assistant for manuscript editing, checking mathematical derivations, developing and verifying numerical analysis code, preparing figures, and reference verification.

Results

Kinetic energy generation WKW_K from MERRA-2

The interannual variation in WKW_K is relatively small (Figure 2a). Over the twelve March–February periods, the mean resolved power is WK=2.89±0.03W_K = 2.89 \pm0.03 W m−2^{-2} (mean ±\pm interannual standard deviation), of which 2.08±0.032.08 \pm0.03 W m−2^{-2} is generated below 500 hPa and 0.81±0.020.81 \pm0.02 W m−2^{-2} above (Figure 2b).

Figure 2. Kinetic energy generation diagnosed from MERRA-2. (a) Global WKW_K for resolved pressure levels and three treatments of the unresolved surface layer. (b) Resolved WKW_K below and above 500 hPa. (c) Latitude–pressure distribution of local kinetic energy generation cKc_K (mW kg−1^{-1}), with corresponding mean geopotential heights. (d) Vertical profile of resolved WKW_K density (W m−2^{-2} hPa−1^{-1}). Panels (c,d) show the study-period means for 8 March 2014–28 February 2026, displayed over 50–1000 hPa.

Our resolved estimate exceeds the 2.46 W m−2^{-2} reported by Huang and McElroy [13] for the earlier MERRA reanalysis over 1979–2010. The latter study used data at a lower horizontal resolution (1.25∘×1.25∘1.25^\circ\times1.25^\circ, compared with 0.5∘×0.625∘0.5^\circ\times0.625^\circ for MERRA-2). More recently, Oh et al. [31] reported a value of 2.87 W m−2^{-2} from the Lorenz conversion terms for MERRA-2 over 1980–2023, effectively identical to our estimate. Their corresponding values for ERA5 and JRA-3Q are 2.45 and 2.46 W m−2^{-2}, respectively, approximately 0.4 W m−2^{-2} (15%) lower than for MERRA-2.

Our separate comparison for 2025 indicates that much of this discrepancy arises above 500 hPa, where MERRA-2 generates more kinetic energy than ERA5 and JRA-3Q (Makarieva and Nefiodov, [28], data not shown here). At these heights, positive and negative local cKc_K partially cancel (Figure 2c), increasing the uncertainty of the net estimate. Below 500 hPa, kinetic energy generation is predominantly positive (Figure 2c), with much less cancellation. Furthermore, the reanalysis is more strongly constrained by observations at these levels. The lower-atmospheric contribution to WKW_K is therefore more robustly estimated.

The vertical profile of WKW_K density has a pronounced maximum in the lower atmosphere (Figure 2d). Thus, restricting the calculation to resolved pressure levels omits the contribution of the surface layer and underestimates the actual WKW_K. We considered three treatments of this unresolved layer: assuming cK=0c_K = 0 at the surface yields WK=3.17±0.04W_K = 3.17 \pm0.04 W m−2^{-2}; holding cKc_K constant at its lowest resolved value gives 3.45±0.043.45 \pm0.04 W m−2^{-2}; and linear extrapolation from the two lowest resolved levels gives 3.53±0.043.53 \pm0.04 W m−2^{-2}. In subsequent analyses, we adopt the most conservative of these estimates, WK=3.17W_K = 3.17 W m−2^{-2}, corresponding to cK=0c_K = 0 at the surface. Overall, these results support a global kinetic energy generation rate of approximately 3 W m−2^{-2}.

Gravitational power of precipitation WPW_P from GPM-SLH

Our GPM-SLH analysis covers 65∘65^\circS–65∘65^\circN, which contains approximately 97% of global precipitation, according to our independent analysis of the GPCP V3.3 monthly product (Huffman et al., [14]). We normalize WPW_P by the area of the entire Earth, the tropics (30∘30^\circS–30∘30^\circN), and the extratropics (30∘30^\circ–90∘90^\circ in both hemispheres), respectively. Contributions poleward of 65∘65^\circ are omitted. Their contribution to global WPW_P is expected to be small, since WP/PW_P/P generally decreases toward the poles (Figure 3a).

Figure 3. Gravitational power of precipitation WPW_P estimated from GPM/DPR spectral latent-heating profiles. (a) Latitude profiles of WPW_P and surface precipitation PP from GPM, compared with PP from GPCP V3.3, for March 2014–December 2025. All three quantities are averaged over 5∘5^\circ latitude bands. (b) WPW_P for twelve March–February periods (2014/15–2025/26), normalized by the area of the entire Earth (Total), the tropics (30∘30^\circS–30∘30^\circN), and the extratropics (30∘30^\circ–90∘90^\circ in both hemispheres), respectively. GPM observations cover 65∘65^\circS–65∘65^\circN; contributions poleward of 65∘65^\circ are not included, although the corresponding areas are included in the normalization.

The mean gravitational power of precipitation, normalized by the entire Earth’s surface area, is WP=1.11±0.02W_P = 1.11 \pm0.02 W m−2^{-2} (mean ±\pm interannual standard deviation). The corresponding regional means are 1.60 ±\pm 0.03 W m−2^{-2} in the tropics and 0.62 ±\pm 0.02 W m−2^{-2} in the extratropics (Figure 3b). The tropical estimate is close to the corrected value of 1.5 W m−2^{-2} reported by Paulius and Dias [36, 37] using observations from 1999–2007.

For March 2014–December 2025, the mean GPM-derived surface precipitation within 65∘65^\circS–65∘65^\circN is 2.36 mm day−1^{-1}, compared with 3.31 mm day−1^{-1} from GPCP V3.3 (Figure 3a), a difference of nearly 30%. Our GPM precipitation estimates come from the Ku-band radar-based V07C SLH product, rather than the combined radar–radiometer retrieval. This distinction is important when considering changes between product versions. Behrangi et al. [2] found that oceanic precipitation estimated from GPM KuPR decreased by about 8% from V06 to V07, whereas the combined radar–radiometer estimate increased by about 9%. Meanwhile, GPCP V3.3 reports 8.5% higher global precipitation than V3.2 over 1992–2023, mainly due to increases over the oceans (Behrangi et al., [3]). Part of this increase reflects the transition to the V07 combined radar–radiometer retrieval.

However, the approximately 30% discrepancy in surface precipitation does not necessarily imply an equally large underestimate of WPW_P. As shown by Makarieva et al. [23], an underestimate of surface precipitation confined to the lowest kilometre contributes relatively little to gravitational power: a 25% underestimate of precipitation in this layer corresponds to only about a 5% underestimate of WPW_P. Thus, although the GPM–GPCP discrepancy introduces uncertainty into our estimate of WPW_P, its magnitude cannot be inferred directly from the difference in surface precipitation.

Another source of uncertainty in our WPW_P estimate is the treatment of SLH profiles without separately accounting for melting and freezing. Our analysis of the microphysical heating profiles in Fig. 2-1 of Tao et al. [47] indicates that excluding these processes reduces WPW_P by about 2% in convective clouds but increases it by about 12% in stratiform clouds (data not shown). The latter effect may contribute to an underestimate of WPW_P, particularly at higher latitudes. Overall, an uncertainty in WPW_P of order 20% appears plausible.

Taking WP≃1.1W_P \simeq1.1 W m−2^{-2} as a representative global estimate, together with WK≃3.2W_K \simeq3.2 W m−2^{-2}, we obtain a total atmospheric power under the hydrostatic approximation of approximately W=WK+WP≃4.3W = W_K + W_P \simeq4.3 W m−2^{-2}, with a likely uncertainty of order 20%.

Steam-engine power Wv=Wg+WcW_v = W_g + W_c from GPM-SLH and MERRA-2

Besides precipitation uncertainties discussed in § 4.2, a major source of uncertainty in WvW_v and its components is the estimation of z1z_1 and T1T_1, the height and temperature at which vapor that subsequently condenses first enters saturation. The two-LCL method assumes undiluted ascent to saturation from the corresponding evaporation levels (Figure 1). This assumption is consistent with the model-derived microphysical profiles of Tao et al. [47], which show gross condensation extending down to approximately 0.30.3 km above the surface in both convective and stratiform regions, even though evaporation dominates over much of the lower stratiform atmosphere. Accounting for atmospheric evaporation, the two-LCL method yields a condensation-weighted mean initial saturation height ⟨z1⟩c≃1.18\langle z_1\rangle_c \simeq1.18 km.

Over the twelve March–February periods, the two-LCL method gives Wv=4.43±0.10W_v = 4.43 \pm0.10 W m−2^{-2}, comprising Wg=1.52±0.03W_g = 1.52 \pm0.03 W m−2^{-2} and Wc=2.91±0.07W_c = 2.91 \pm0.07 W m−2^{-2} (Figure 4; mean ±\pm interannual standard deviation).

Figure 4. Steam-engine expansion power WvW_v and its gravitational (WgW_g) and condensation-related (WcW_c) components estimated from GPM/DPR spectral latent heating and MERRA-2 using the two-LCL (solid lines) and one-onset (dashed lines) methods. The values are normalized by the area of the entire Earth for twelve March–February periods (2014/15–2025/26). The GPM analysis covers 65∘S65^\circ\mathrm{S}–65∘N65^\circ\mathrm{N}; contributions poleward of 65∘65^\circ are omitted, while the polar areas are included in the normalization.

An alternative, more conservative estimate is based on the lowest level of positive retrieved net latent heating in each footprint, whose positive-heating-weighted mean is zc,min⁡z_{c,\min}. The corresponding heights and MERRA-2 temperatures are used to estimate the mean effective initial saturation state for vapor originating from both surface and atmospheric evaporation. Because net evaporation can dominate the lower cloud layers and mask gross condensation there, we expect the lowest level of positive net latent heating to lie above the onset of gross condensation.

The one-onset method gives zc,min⁡≃1.9z_{c,\min} \simeq1.9 km and an effective initial saturation temperature Tc,min⁡≃284.6T_{c,\min} \simeq284.6 K, compared with the effective geometric-mean temperature T1≃287.1T_1 \simeq287.1 K for the two-LCL method.

The resulting powers are Wv=3.80±0.08W_v = 3.80 \pm0.08 W m−2^{-2}, Wg=1.18±0.02W_g = 1.18 \pm0.02 W m−2^{-2} and Wc=2.62±0.06W_c = 2.62 \pm0.06 W m−2^{-2} (Figure 4). These estimates are lower than the corresponding two-LCL values by approximately 14%, 22% and 10%, respectively.

The arithmetic means of the two estimates are Wv≃4.11 W m−2W_v \simeq4.11\ \mathrm{W\ m^{-2}}, Wg≃1.35 W m−2W_g \simeq1.35\ \mathrm{W\ m^{-2}} and Wc≃2.76 W m−2W_c \simeq2.76\ \mathrm{W\ m^{-2}}.

Using the Earth-area-normalized condensation heating Qc≃74 W m−2Q_c \simeq74\ \mathrm{W\ m^{-2}}, or Pc=Qc/L≃2.96×10−5 kg m−2 s−1P_c = Q_c/L \simeq2.96 \times10^{-5}\ \mathrm{kg\ m^{-2}\ s^{-1}}, and Tc≃271 KT_c \simeq271\ \mathrm{K}, we obtain from Eq. (14) κeff≃−0.21\kappa_{\mathrm{eff}} \simeq-0.21 for the two-LCL method and κeff≃−0.29\kappa_{\mathrm{eff}} \simeq-0.29 for the one-onset method. These correspond to effective incompleteness-of-condensation ratios γ2/γ1≃0.08\gamma_2/\gamma_1 \simeq0.08 and 0.120.12, respectively, or effective degrees of condensation 1−γ2/γ1≃0.921-\gamma_2/\gamma_1 \simeq0.92 and 0.880.88.

Discussion

Numerical results

Under the hydrostatic approximation, we estimated atmospheric mechanical power as W=WK+WPW = W_K + W_P, diagnosing WKW_K from MERRA-2 and WPW_P separately using GPM-SLH and MERRA-2 atmospheric fields.

The gravitational power of precipitation WPW_P was already identified and estimated by Gorshkov [9], who listed it separately from wind power among terrestrial energy flows. The distinction between mechanical work required to lift precipitation and kinetic energy generation by atmospheric motions was subsequently discussed by Pauluis et al. [33] and Pauluis and Dias (2012), the latter providing the first satellite-based estimate of precipitation-induced dissipation using TRMM observations. However, to our knowledge, the general decomposition W=WK+WPW = W_K + W_P from the atmospheric continuity equation has not been derived in the peer-reviewed literature. The conventional equivalence between total pressure work WW calculated by integrating −(1/ρ)Dp/Dt-(1/\rho) Dp/Dt over atmospheric mass and horizontal pressure-gradient work WKW_K calculated by integrating −u⋅∇hp-\mathbf{u} \cdot\nabla_h p over atmospheric volume, which neglects the contribution WPW_P arising from gaseous-mass sources and sinks, continues to be assumed in recent studies ([49, 31]).3

Our conservative surface-layer treatment gives WK=3.2 W m−2W_K = 3.2\ \mathrm{W\ m^{-2}} (2.9 W m−22.9\ \mathrm{W\ m^{-2}} in the resolved layers; § 4.1). Differences of about 15% among reanalyses, largely concentrated in the upper atmosphere ([31]), together with the uncertainty of the unresolved surface layer, suggest a representative WK≃3 W m−2W_K \simeq3\ \mathrm{W\ m^{-2}} with an uncertainty of order 20%.

We obtain WP=1.1 W m−2W_P = 1.1\ \mathrm{W\ m^{-2}} (§ 4.2), with a plausible uncertainty of order 20% arising primarily from satellite retrievals of precipitation and latent heating, possible underestimation of near-surface precipitation, and the treatment of melting and freezing. Together, the reference estimates give W=WK+WP≃4.3 W m−2W = W_K + W_P \simeq4.3\ \mathrm{W\ m^{-2}}, also with an uncertainty of order 20%. The smaller interannual standard deviations reported in the Results do not represent these observational and methodological uncertainties.

Kinetic energy generation is concentrated in the lower atmosphere (Figure 2b-Figure 2d). We denote the contributions below and above 500 hPa by WK,lW_{K,l} and WK,uW_{K,u}, respectively. Below 500 hPa, WK,l=2.1 W m−2W_{K,l} = 2.1\ \mathrm{W\ m^{-2}} in the resolved layers, increasing to 2.4 W m−22.4\ \mathrm{W\ m^{-2}} with the conservative surface-layer correction. The upper atmosphere contributes WK,u=0.8 W m−2W_{K,u} = 0.8\ \mathrm{W\ m^{-2}}, or approximately one quarter of the corrected total.

The relatively small positive net contribution aloft results from partial cancellation between positive and negative kinetic energy generation, including negative work where air moves against temperature-related pressure gradients, as in the Ferrel cells (Figure 2c). In contrast, lower-atmospheric kinetic energy generation is predominantly positive, with a pronounced near-surface maximum (Figure 2c,Figure 2d).

Our separately diagnosed positive saturated water-vapor expansion power is Wv=4.4 W m−2W_v = 4.4\ \mathrm{W}\,\mathrm{m}^{-2} with the two-LCL method and 3.8 W m−23.8\ \mathrm{W}\,\mathrm{m}^{-2} with the one-onset method (§ 4.3). The arithmetic mean is Wv≃4.1 W m−2W_v \simeq4.1\ \mathrm{W}\,\mathrm{m}^{-2}, with a representative uncertainty of order 30%.

Water-vapor expansion power consists of two physically distinct contributions. The gravitational component WgW_g arises because condensation removes gaseous mass during ascent, so that more gas molecules ascend and expand than descend and contract. This component is quantified by the hydrostatic decrease in total-air pressure during ascent (Eq. (8)).

The gravitational power of precipitation WPW_P likewise arises from this imbalance, but differs from WgW_g in two respects. First, WgW_g considers expansion only between z1z_1 and zcz_c, with z1>zez_1 > z_e, whereas WPW_P considers the full interval from evaporation at zez_e to condensation at zcz_c (Figure 1). Second, WgW_g is proportional to the molecular mass of air through the hydrostatic pressure decrease. Thus, if z1z_1 and zez_e coincided, we would obtain Wg≃(M/Mv)WP=1.6WPW_g \simeq(M/M_v)W_P = 1.6W_P, rather than the present Wg≃1.2WPW_g \simeq1.2W_P, based on the mean of the two estimates of WgW_g. The excess ascending water-vapor molecules are lighter than the surrounding air, contributing to buoyancy and hence to greater positive work.

The condensation-related component WcW_c arises from the additional decrease in saturated water-vapor partial pressure relative to total-air pressure during condensation (Eqs. (6) and (12)). For the two-LCL and one-onset methods, Wc=2.9W_c = 2.9 and 2.6 W m−22.6\ \mathrm{W}\,\mathrm{m}^{-2}, respectively, with a mean Wc≃2.8 W m−2W_c \simeq2.8\ \mathrm{W}\,\mathrm{m}^{-2}. Both estimates are comparable in magnitude to, and exceed, the estimated lower-atmospheric kinetic energy generation, WK,l≃2.4 W m−2W_{K,l} \simeq2.4\ \mathrm{W}\,\mathrm{m}^{-2}.

Figure 5 shows that WW and WvW_v differ by approximately 4%, while WKW_K and WcW_c differ by approximately 13%. Importantly, Wc≃2.8 W m−2W_c \simeq2.8\ \mathrm{W}\,\mathrm{m}^{-2} is also close to the lower-atmospheric WK,l≃2.4 W m−2W_{K,l} \simeq2.4\ \mathrm{W}\,\mathrm{m}^{-2}, whose physical relationship to WcW_c we consider below. These differences are smaller than the estimated uncertainties of 20–30%.

Figure 5. (a) Comparison of steam-engine expansion power WvW_v and its condensation-related (WcW_c) and gravitational (WgW_g) components with atmospheric mechanical power W=WK+WPW = W_K + W_P, kinetic energy generation WKW_K and its lower-atmospheric contribution WK,lW_{K,l} (below 500 hPa, including the surface-layer correction), and gravitational power of precipitation WPW_P. Blue solid lines show arithmetic means of the two-LCL and one-onset estimates; orange broken lines show separately diagnosed atmospheric powers. Shaded areas indicate ±10%\pm10\% of each series for visual comparison. (b) Corresponding period-mean power budgets, with WKW_K divided into lower- and upper-atmospheric contributions. The labels ±20%\pm20\% for WW and ±30%\pm30\% for WvW_v indicate representative uncertainties. For comparison, the green bar shows the global Lorenz available potential energy (APE) generation GL≃1.2 W m−2G_L \simeq1.2\ \mathrm{W\ m^{-2}} estimated by Romanski and Rossow [40] from predominantly observation-based diabatic heating and temperature fields for 1997–2000.

Our steam-engine analysis concerns the positive expansion work of saturated water vapor, which is energetically linked to latent heat release during condensation through the Clausius–Clapeyron relation. In a related but different approach, Pauluis and Held (2002) evaluated the global integral of pv∇⋅vp_v\nabla\cdot\boldsymbol{v}, likewise distinguishing the hydrostatic decrease of total-air pressure from changes in water-vapor abundance due to phase transitions. However, because their integral also included evaporation, the positive condensation-related contribution proportional to RTcRT_c was more than offset by the negative evaporation-related contribution proportional to RTeRT_e, with Te>TcT_e > T_c (Te/Tc≃1.08T_e/T_c \simeq1.08 in our data; Figure 1). Thus, the condensation and evaporation contributions largely cancelled in their global integral, which remained dominated by (M/Mv)WP(M/M_v)W_P. This cancellation does not imply that WcW_c vanishes: their global integral does not separately evaluate the condensation-related expansion work defined here, and its potential contribution to atmospheric power generation was not investigated in that framework.

Relation to condensation-induced atmospheric dynamics

In contrast, the condensation-related contribution to atmospheric power has been central to condensation-induced atmospheric dynamics (CIAD) from the outset, although its relationship to the expansion work of the water-vapor steam engine had not previously been established.

CIAD originated from the idea that the vertical distribution of water-vapor partial pressure maintained by condensation differs from that associated with the hydrostatic decrease of total-air pressure, and that this difference can generate atmospheric power proportional to the condensation rate (for a review, see Makarieva et al., [26]).

The proposed local power density is

σCIAD=−w(∂pv∂z−pvp∂p∂z)=−wp(1+γ)2∂γ∂z≃−wp∂γ∂z.(26)\sigma_{\mathrm{CIAD}}=-w\left(\frac{\partial p_v}{\partial z}-\frac{p_v}{p}\frac{\partial p}{\partial z}\right) =-\frac{wp}{(1+\gamma)^2}\frac{\partial\gamma}{\partial z}\simeq-wp\frac{\partial\gamma}{\partial z}. \tag*{(26)}

Using water-vapor and dry-air continuity, neglecting vapor diffusion, and assuming that vertical advection dominates the material change of γ\gamma, we obtain σCIAD≃RTC/Mv\sigma_{\mathrm{CIAD}}\simeq RTC/M_v to leading order in the vapor fraction, where CC is the local condensation rate in kg m−3 s−1\mathrm{kg\ m^{-3}\ s^{-1}}. Integration over the condensation region Ωc\Omega_c gives

WCIAD≡1S∫Ωcσ‾CIAD dΩ≃RPcTcMv=Wc1+κeff,(27)W_{\mathrm{CIAD}}\equiv\frac{1}{S}\int_{\Omega_c}\overline{\sigma}_{\mathrm{CIAD}}\,\mathrm{d}\Omega \simeq\frac{RP_cT_c}{M_v}=\frac{W_c}{1+\kappa_{\mathrm{eff}}}, \tag*{(27)}

where the final equality follows from Eq. (14). Thus WCIADW_{\mathrm{CIAD}} corresponds to WcW_c in the complete-condensation limit κeff→0\kappa_{\mathrm{eff}}\to0. For the two-LCL and one-onset estimates, κeff≃−0.21\kappa_{\mathrm{eff}}\simeq-0.21 and −0.29-0.29, respectively, so that WcW_c is reduced relative to the integrated CIAD expression by factors of approximately 0.79 and 0.71.

If σCIAD\sigma_{\mathrm{CIAD}} is identified, in order of magnitude, with local kinetic energy generation by horizontal pressure gradients, σCIAD∼−u⋅∇hp\sigma_{\mathrm{CIAD}}\sim-\boldsymbol{u}\cdot\nabla_h p, dimensional analysis gives a characteristic horizontal pressure difference. Taking u/w∼L/Hu/w\sim L/H, ∣∇hp∣∼Δp/L\lvert\nabla_h p\rvert\sim\Delta p/L, and ∣∂γ/∂z∣∼Δγ/H\lvert\partial\gamma/\partial z\rvert\sim\Delta\gamma/H, we obtain Δp∼pΔγ\Delta p\sim p\Delta\gamma.

For nearly complete condensation, Δγ∼γ1\Delta\gamma\sim\gamma_1, giving Δp∼pγ1\Delta p\sim p\gamma_1, of the order of the initial saturated water-vapor partial pressure, typically about 10 hPa in the lower atmosphere. This is comparable to pressure differences in large-scale circulations (L∼103 kmL\sim10^3\ \mathrm{km}) [21], including about 12 hPa across the Hadley circulation [25]. Holton (2004, p. 4) also noted pressure differences of order 10 hPa in systems with very different spatial scales, including tornadoes, squall lines, and hurricanes.

Within CIAD, this relative constancy is interpreted as a consequence of the constraint that condensation imposes on kinetic energy generation. For tropical cyclones, equating the condensation-domain integral of σCIAD\sigma_{\mathrm{CIAD}} to boundary-layer kinetic energy generation and applying cyclostrophic balance yields an upper limit on maximum wind velocity [27].

Why does atmospheric power approximately coincide with steam-engine power?

The comparison posed in the Introduction can be written as

Wv−W=(Wg−WP)+(Wc−WK).(28)W_v-W=(W_g-W_P)+(W_c-W_K). \tag*{(28)}

With our global estimates, Wc≃2.8 W m−2W_c\simeq2.8\ \mathrm{W\ m^{-2}} is close to both the total WK≃3.2 W m−2W_K\simeq3.2\ \mathrm{W\ m^{-2}} and the lower-atmospheric WK,l≃2.4 W m−2W_{K,l}\simeq2.4\ \mathrm{W\ m^{-2}}. The mechanism we propose primarily concerns the latter correspondence. We seek to explain how condensation-related power generates kinetic energy in the lower atmosphere. We leave the smaller, consistently positive difference Wg−WP≃0.2 W m−2W_g-W_P\simeq0.2\ \mathrm{W\ m^{-2}} and its possible connection to the observed upper-atmospheric kinetic energy generation WK,u≃0.8 W m−2W_{K,u}\simeq0.8\ \mathrm{W\ m^{-2}} for future research.

Condensation thus offers the entire air mixture an opportunity to expand toward the lower surface pressure it maintains. However, while condensed water does not require compression, the expanded air does. In a horizontally isothermal atmosphere, upper-level compression would nearly cancel the positive pressure work of the lower-level expansion. This compensation is illustrated using a simplified mass-related Lorenz-type available-energy formulation in Appendix § 8.

Here expansion and compression refer to horizontal pressure work. Under hydrostatic equilibrium, vertical pressure-gradient forces are balanced by gravity, with the gravitational contributions to steam-engine and atmospheric power accounted for separately by WgW_g and WPW_P, respectively.

The crucial difference between the lower and upper branches (Figure 6) is that the frictional surface flow requires appreciable pressure work, whereas the return flow aloft can proceed with negligible pressure work. We propose that the condensation-driven surface flow unlocks differential heating that warms the upper-level return flow, modifying the pressure field and preventing compression4. The return flow thus becomes nearly isobaric, with WK,u≃0W_{K,u} \simeq0.

Figure 6. Idealized circulation with lower-level flow from the cooler high-pressure region HH to the warmer low-pressure region LL and a return flow aloft. Dotted lines indicate isobars; ziz_i is the isobaric height. EE and PP denote net evaporation and precipitation. For WK,u≃0W_{K,u} \simeq0, thermal compensation implies WT≃WK,lW_T \simeq W_{K,l}.

Here, “unlocks” means that we propose that the same horizontal temperature differences would be unable to drive a comparable circulation without the condensation-related power input at the surface.

Schematically, in the idealized case WK,u≃0W_{K,u} \simeq0, the negative surface-pressure contribution to upper-level work, approximately −WK,l-W_{K,l} in the isothermal limit, is offset by the positive temperature-related contribution WTW_T. Thus, as illustrated in Figure 6, WK≃WK,l≃WTW_K \simeq W_{K,l} \simeq W_T: the same net work can be described through its lower-level pressure-work origin or through the differential heating that prevents its cancellation aloft (Appendix § 8, Eq. (64)). In the idealized case considered here, when WK,l≃WcW_{K,l} \simeq W_c, the condensation-related steam-engine power and temperature-related power coincide: WT≃WcW_T \simeq W_c.

Notably, WTW_T depends not on the horizontal temperature difference ΔT\Delta T alone, but on its product with the mean height of the upper flow. Thus, for a given surface pressure gradient, the return flow can be accommodated by a range of horizontal temperature gradients, depending on its height (Eq. (62)). This flexibility supports the proposed “unlocking” mechanism.

The foregoing argument considered an idealized case with WK,u≃0W_{K,u} \simeq0 and WK,l≃WcW_{K,l} \simeq W_c. In our global estimates, by contrast, WK,u≃0.29WcW_{K,u} \simeq0.29W_c, while WK,l≃0.86WcW_{K,l} \simeq0.86W_c.

The difference between WK,lW_{K,l} and WcW_c may reflect the less-than-unity efficiency of the mechanical transformation of steam-engine expansion work into the potential energy associated with surface pressure gradients. One factor affecting this efficiency may be the spatial and temporal overlap between evaporation and condensation. Where these processes overlap in space and time, their opposing contributions to the gaseous-mass budget partially cancel, reducing the column-integrated net mass source or sink. As discussed in Appendix § 8, mass-related available-energy generation depends on the covariance between this column-integrated net mass source and surface pressure, rather than on the condensation rate alone. Such overlap may therefore reduce the power available for kinetic energy generation in the lower atmosphere. This possibility requires further study.

Efficiency and mechanical power

Previous studies have obtained substantially reduced, but quantitatively different, estimates of the mechanical work associated with atmospheric moisture. Pauluis and Held (2002), evaluating the global integral of pv∇⋅vp_v\nabla\cdot\mathbf{v}, estimated water-vapor work to be approximately 20–30% of the work of a perfect heat engine transporting the same latent heat. Using a different thermodynamic formulation, Pauluis (2011) found that the efficiency of a fully unsaturated steam cycle under Earth-like conditions is approximately one-sixth to one-fifth of Carnot efficiency, increasing to Carnot efficiency for a fully saturated cycle. These estimates concern different quantities and assumptions, and their numerical values cannot be directly applied to the relationship between WcW_c and WK,lW_{K,l} considered here.

We agree with the general conclusion that the net mechanical work associated with moist atmospheric cycles can be considerably smaller than the Carnot benchmark. In our framework, two distinct reductions can be identified. First, the positive water-vapor expansion power corresponds to a fraction Wv/Qc≃0.056W_v/Q_c \simeq0.056 of the latent heat released during condensation, where Qc≃74 W m−2Q_c \simeq74\ \mathrm{W\,m^{-2}}. For the two-LCL and one-onset methods, this fraction is approximately 0.060 and 0.051, respectively. According to Eq. (5),

WvQc≃⟨ln⁡T1T⟩c(29)\frac{W_v}{Q_c} \simeq\left\langle\ln\frac{T_1}{T} \right\rangle_c \tag*{(29)}

which approaches the corresponding Carnot-like efficiency in the small temperature-difference limit. Here T1T_1 is the initial saturation temperature, rather than necessarily the evaporation temperature. The commonly cited Carnot efficiency of the atmospheric heat engine is approximately 0.1, based on large-scale atmospheric temperature contrasts [39, 4] (Pauluis and Dias 2012). This comparison reflects different temperature contrasts and should not be interpreted as a measure of irreversible efficiency loss.

Second, in the horizontally isothermal case, a further and potentially much larger reduction occurs because positive lower-level pressure work is almost entirely offset by negative pressure work in the return flow aloft. Consequently, WK≪WcW_K \ll W_c, and the net mechanical efficiency relative to the latent-heat input is much smaller than even the Carnot efficiency associated with the initial-saturation–condensation temperature difference. In the non-isothermal case, however, under the condition of a small upper-level pressure-work residual, WK≃Wc≃WTW_K \simeq W_c \simeq W_T; the net power can also be expressed in terms of differential heating, including its sensible-heat contribution, whose conversion efficiency can in principle approach the Carnot limit. The condensation-related and temperature-related expressions characterize the same mechanical power rather than additive contributions. Thus, a potentially efficient contribution from differential heating can coexist with a low net efficiency of the work attributable to latent heat alone.

The dominant role of condensation that we propose lies not in increasing thermodynamic efficiency, but in making a circulation of substantial mechanical power possible in the first place. Thus, a moist atmosphere may have both lower thermodynamic efficiency and higher mechanical power than a dry atmosphere.

Hertz suggested that in a dry atmosphere, temperature differences “would by themselves give rise merely to movements of minor significance” ([30]). In our interpretation, without the water cycle, differential heating, which itself depends on air motion and manifests itself in temperature-related pressure gradients aloft, would be unable to generate comparable surface pressure gradients and cross-isobaric flow in the lower atmosphere.

Disentangling causality

The observed correspondence between WcW_c and WK,lW_{K,l}, together with the approximate equality WK≃WTW_K \simeq W_T, does not by itself establish causality. Differential heating might generate WKW_K independently of the steam engine, rather than the steam engine unlocking the differential heating needed for the return flow. In the former case, the proximity of WK,lW_{K,l} and WcW_c could be attributed to chance. We can examine these alternatives in circulations with contrasting upper-level pressure work.

Hurricanes offer such a contrasting test. They develop over a largely isothermal surface, and their upper outflow can generate negative pressure work, unlike the small positive global-mean upper-atmospheric contribution (Figure 2b,Figure 2c). Such motion against a significant pressure gradient is possible because hurricanes convert a substantial fraction of the pressure potential energy into kinetic energy in the low-level inflow. This fraction can reach about 40%, compared with about 2% in large-scale, friction-dominated circulations such as Hadley cells.5

In the modeled hurricane of Smith et al. [45], the positive and negative horizontal pressure-work contributions at about day 3 were approximately +55 and −45 TW, leaving 10 TW net within a cylinder of radius 300 km and height 20 km (their Fig. 4; values estimated from volume-mean terms). Thus, negative work in the upper outflow consumes much of the kinetic energy generated by positive pressure work in the low-level inflow.

We can compare the maximum local surface dissipation in hurricanes with condensation-related expansion power estimated from the maximum eyewall precipitation rate, using the former as a proxy for local kinetic energy generation. According to Emanuel [6], an average Atlantic hurricane with a maximum wind speed Vmax⁡=50 m s−1V_{\max} = 50\ \mathrm{m\ s^{-1}} dissipates kinetic energy at the surface at a local rate D=ρCDV3D = \rho C_D V^3, where ρ=1 kg m−3\rho= 1\ \mathrm{kg\ m^{-3}} is air density and CD=2×10−3C_D = 2 \times10^{-3} is the drag coefficient. The corresponding maximum local dissipation is Dmax⁡≃250 W m−2D_{\max} \simeq250\ \mathrm{W\ m^{-2}}.

Independently, Yang et al. [51] obtained a maximum azimuthally averaged eyewall precipitation rate of 9.1 mm h−1^{-1} for Category 3 hurricanes, or about 220 mm day−1^{-1}. For an illustrative condensation temperature Tc=270 KT_c = 270\ \mathrm{K}, assuming complete condensation (κ=0\kappa= 0), we obtain

WcPc≃RTcMv≃1.25×105 J kg−1,Wc≃320 W m−2.(30)\frac{W_c}{P_c} \simeq\frac{RT_c}{M_v} \simeq1.25 \times10^5\ \mathrm{J\ kg^{-1}}, \qquad W_c \simeq320\ \mathrm{W\ m^{-2}}. \tag*{(30)}

The two estimates are within 30% of each other, although they represent characteristic local maxima rather than collocated measurements. Nevertheless, similar power scales in circulations with contrasting upper-level energetics provide an additional consistency check for the proposed condensation-driven lower-atmospheric pathway. Even when, as in hurricanes, horizontal temperature gradients associated with latent heat release are insufficient to produce positive kinetic energy generation aloft, the lower atmosphere still generates power close to WcW_c. More detailed considerations of hurricane energetics within the CIAD framework are given by Makarieva and Gorshkov [22]; Makarieva and Nefiodov [27].

Implications of neglecting the second equality in WK≃WT≃WcW_K \simeq W_T \simeq W_c

The relatively small net kinetic energy generation aloft (Figure 2d) creates a peculiar situation: kinetic energy is generated predominantly in the lower atmosphere by surface pressure gradients, yet the same power can be expressed in terms of horizontal temperature gradients, WK≃WTW_K \simeq W_T. Our results establish an additional correspondence: WK≃WcW_K \simeq W_c, where WcW_c is the condensation-related power of the steam engine.

In view of this second correspondence, distinguishing cause from effect is crucial for predicting atmospheric responses to different forcings. In a statistically steady atmosphere, kinetic energy generation must balance frictional dissipation. Turbulent by nature, this dissipation is unknown a priori. We therefore cannot predict how an Earth-like planet would circulate its air from first principles alone.

Current parameterizations of turbulent friction are based on observed relationships between wind velocities, atmospheric state variables, and surface properties. For the contemporary atmosphere, models employing these parameterizations reproduce the observed circulation, in which Wc≃WTW_c \simeq W_T under the existing coupling between differential heating and condensation. Reproducing this circulation, however, does not establish which of these processes is the primary driver. Extrapolating model predictions to scenarios in which this coupling changes requires correctly identifying the primary determinant of circulation and ensuring that its role is properly represented in the model. To give a practical example, a model that attributes circulation primarily to differential heating may predict that Amazon deforestation enhances atmospheric moisture import by increasing the land–ocean temperature contrast. In a global storm-resolving simulation of complete Amazon deforestation, Yoon and Hohenegger [52] found that circulation changes enhanced moisture convergence, largely compensating for reduced evapotranspiration, whereas most CMIP6 models in a global-deforestation experiment predicted reduced Amazon precipitation ([20]).

An incomplete representation of WcW_c, combined with an emphasis on WTW_T, may also contribute to differences between modeled and observed atmospheric circulations under present-day conditions. For example, in a storm-resolving configuration of ICON, Paccini and Stevens [32] found excessive afternoon precipitation and an underestimated and delayed secondary early-morning rainfall maximum. We suggest that examining the relative roles of WcW_c and WTW_T could provide an additional perspective on such discrepancies, particularly the excessive concentration of rainfall during hours of strong daytime heating.

These distinctions are particularly important for predicting the effects of land-cover change. If condensation determines whether circulation takes place, reducing native forest cover and hence the moisture supply could weaken circulation despite increasing temperature contrasts. Thus, different assumptions about the physical origin of the power dissipated by turbulent friction may lead to contrasting predictions under the same forcing. More generally, the spatial reorganization of evaporation and precipitation could have different predicted effects on atmospheric circulation depending on whether the mechanical role of condensation is recognized.

The evidence presented here for the correspondence between lower-atmospheric kinetic energy generation and condensation-related expansion work of water vapor suggests that this mechanism deserves explicit consideration in understanding and predicting atmospheric responses to environmental change.

Gravitational power of precipitation

Hydrostatic equilibrium, ∂zp=−ρg\partial_z p=-\rho g, with ρ\rho the gaseous-air density, gives the time-mean vertical contribution to pressure work:

WP≡−1S∫Ωw∂p∂z‾ dΩ=gS∫Ωρw‾ dΩ.(31)W_P \equiv-\frac{1}{S}\int_{\Omega}\overline{w\frac{\partial p}{\partial z}}\,\mathrm{d}\Omega= \frac{g}{S}\int_{\Omega}\overline{\rho w}\,\mathrm{d}\Omega. \tag*{(31)}

The nonnegative local net condensation and evaporation rates CC and EE are defined in § 2.1. Surface evaporation is included in EE as a source in a thin near-surface atmospheric layer. The instantaneous gaseous-air continuity equation is

∂ρ∂t+∇⋅(ρv)=E−C.(32)\frac{\partial\rho}{\partial t}+\nabla\cdot(\rho\mathbf{v})=E-C. \tag*{(32)}

Multiplying by height zz and using v⋅∇z=w\mathbf{v}\cdot\nabla z=w gives

∂(zρ)∂t+∇⋅(zρv)=ρw−z(C−E).(33)\frac{\partial(z\rho)}{\partial t}+\nabla\cdot(z\rho\mathbf{v})=\rho w-z(C-E). \tag*{(33)}

For a statistically steady atmosphere, the time-mean storage term ∂(zρ)/∂t‾\overline{\partial(z\rho)/\partial t} vanishes. Assuming also that the gaseous-air mass flux through the boundary of Ω\Omega vanishes, time averaging and spatial integration yield

1S∫Ωρw‾ dΩ=1S∫Ωz(C−E)‾ dΩ.(34)\frac{1}{S}\int_{\Omega}\overline{\rho w}\,\mathrm{d}\Omega=\frac{1}{S}\int_{\Omega}z\overline{(C-E)}\,\mathrm{d}\Omega. \tag*{(34)}

The time-mean gaseous-mass balance in a statistically steady state also gives

Pc=1S∫ΩE‾ dΩ,(35)P_c=\frac{1}{S}\int_{\Omega}\overline{E}\,\mathrm{d}\Omega, \tag*{(35)}

where PcP_c is defined in Eq. (4).

The corresponding mean condensation and evaporation heights are

zc=1SPc∫ΩzC‾ dΩ,ze=1SPc∫ΩzE‾ dΩ.(36)\begin{aligned} z_c&=\frac{1}{SP_c}\int_{\Omega}z\overline{C}\,\mathrm{d}\Omega,\\ z_e&=\frac{1}{SP_c}\int_{\Omega}z\overline{E}\,\mathrm{d}\Omega. \tag*{(36)} \end{aligned}

Substituting Eq. (34) into Eq. (31) gives WP=gPc(zc−ze)W_P=gP_c(z_c-z_e), as in Eq. (17).

Diagnosing WKW_K from MERRA-2

We diagnose kinetic energy generation by horizontal pressure gradients from three-hourly MERRA-2 pressure-level data. The calculation uses the horizontal wind components uu and vv, the height ZZ of each isobaric surface, surface pressure psp_s, and the pressure-level coordinate pp. The geopotential of an isobaric surface is

Φ(p,φ,λ,t)=gZ(p,φ,λ,t).(37)\Phi(p,\varphi,\lambda,t)=gZ(p,\varphi,\lambda,t). \tag*{(37)}

We use g=9.80665 m s−2g = 9.80665\ \mathrm{m\ s^{-2}}. In pressure coordinates, the horizontal pressure-gradient force per unit mass is −∇pΦ-\nabla_p\Phi. Thus the local rate of kinetic energy generation per unit mass is

cK(p,φ,λ,t)=−u⋅∇pΦ=−(u∂Φ∂x+v∂Φ∂y),(38)c_K(p,\varphi,\lambda,t)=-\mathbf{u}\cdot\nabla_p\Phi=-\left(u\frac{\partial\Phi}{\partial x}+v\frac{\partial\Phi}{\partial y}\right), \tag*{(38)}

with units W kg−1\mathrm{W\ kg^{-1}}. The local rate in Eq. (38) is the mass-specific counterpart of the pressure-work density −u⋅∇hp-\mathbf{u}\cdot\nabla_h p integrated in Eq. (18).

The horizontal derivatives in Eq. (38) were evaluated by centered finite differences on each pressure surface, retaining a grid point only when its centre and four immediate horizontal neighbours have valid uu, vv and ZZ and lie above the local surface. For longitude, periodic boundary conditions were used:

∂Φ∂x≃1REcos⁡φΦ(p,φ,λ+Δλ)−Φ(p,φ,λ−Δλ)2Δλ.(39)\frac{\partial\Phi}{\partial x}\simeq\frac{1}{R_E\cos\varphi}\frac{\Phi(p,\varphi,\lambda+\Delta\lambda)-\Phi(p,\varphi,\lambda-\Delta\lambda)}{2\Delta\lambda}. \tag*{(39)}

For latitude,

∂Φ∂y≃Φ(p,φ+Δφ,λ)−Φ(p,φ−Δφ,λ)2REΔφ.(40)\frac{\partial\Phi}{\partial y}\simeq\frac{\Phi(p,\varphi+\Delta\varphi,\lambda)-\Phi(p,\varphi-\Delta\varphi,\lambda)}{2R_E\Delta\varphi}. \tag*{(40)}

Here RE=6371 kmR_E=6371\ \mathrm{km} is Earth’s radius, and angular increments are in radians. The two polar rows are excluded from the cKc_K calculation, avoiding the singular zonal metric factor.

Values were treated as invalid if non-numeric, missing-valued or outside the permissive source-code screening bounds (−500-500 to 500 m s−1500\ \mathrm{m\ s^{-1}} for wind components, −1000-1000 to 100000 m100000\ \mathrm{m} for ZZ, and 1000010000 to 110000 Pa110000\ \mathrm{Pa} for surface pressure). Values greater than 101410^{14} in absolute magnitude, including the MERRA-2 fill flag, were also rejected. The full five-point stencil and the constraint p≤psp\leq p_s were applied at every valid pressure surface. Invalid stencils were excluded from the resolved layer integral; in the global area mean they contribute zero to a fixed-area denominator, without being regarded as observed zero generation. Separate validity fractions accompany the local zonal means.

At each horizontal grid point, the resolved pressure-column integral was computed by trapezoidal integration over adjacent valid pressure levels:

K0(φ,λ,t)=∑k=1Np−112(cK,k+cK,k+1)Δpkg,(41)K_0(\varphi,\lambda,t)=\sum_{k=1}^{N_p-1}\frac{1}{2}\left(c_{K,k}+c_{K,k+1}\right)\frac{\Delta p_k}{g}, \tag*{(41)}

where Δpk=∣pk+1−pk∣\Delta p_k=\lvert p_{k+1}-p_k\rvert. Since Δp/g\Delta p/g is the atmospheric mass per unit area between two pressure surfaces, K0K_0 has units W m−2\mathrm{W\ m^{-2}}. The pressure-layer contributions are retained separately for the study period and its constituent March–February intervals; their split at 500 hPa uses each layer’s pressure midpoint.

Because the lowest valid pressure level does not always coincide with the local surface pressure, we computed three surface-layer corrections. Let pbp_b be the lowest valid pressure level in a column, cbc_b the corresponding value of cKc_K, and Δps=ps−pb\Delta p_s=p_s-p_b. If Δps>0\Delta p_s>0, the correction used for the main estimate assumes that u⋅∇p=0\mathbf{u}\cdot\nabla_p=0 at the surface and that the mean local kinetic energy generation between pbp_b and psp_s is one half of its value at pbp_b:

Kcorr=K0+12cbΔpsg.(42)K_{\mathrm{corr}}=K_0+\frac{1}{2}c_b\frac{\Delta p_s}{g}. \tag*{(42)}

This is the value reported in the main text. For the constant-value surface-layer treatment shown in Figure 2a, we hold cK=cbc_K=c_b throughout the unresolved layer, giving

Kconst=K0+cbΔpsg.K_{\mathrm{const}}=K_0+c_b\frac{\Delta p_s}{g}.

As a sensitivity estimate, we also extrapolated cKc_K to the surface from the lowest valid level and its immediately adjacent valid upper level (the archived adjacent mode). If pup_u is that upper pressure level, with value cuc_u, then

cs=cb+(cb−cu)ps−pbpb−pu,(43)c_s = c_b + (c_b-c_u)\frac{p_s-p_b}{p_b-p_u}, \tag*{(43)}

and

Kextr=K0+12(cb+cs)Δpsg.(44)K_{\mathrm{extr}} = K_0 + \frac{1}{2}(c_b+c_s)\frac{\Delta p_s}{g}. \tag*{(44)}

If the immediately adjacent upper level was invalid, the extrapolated correction was replaced by Eq. (42), without interpolation across the missing level.

Finally, global means were obtained by area-weighting the column values:

K‾(t)=∑i,jKij(t)Aij∑i,jAij,(45)\overline{K}(t) = \frac{\sum_{i,j} K_{ij}(t)A_{ij}}{\sum_{i,j} A_{ij}}, \tag*{(45)}

where the grid-cell area was computed from latitude-band boundaries,

Aij=RE2Δλ[sin⁡(φi+1/2)−sin⁡(φi−1/2)].(46)A_{ij} = R_E^2\Delta\lambda\left[\sin\left(\varphi_{i+1/2}\right)-\sin\left(\varphi_{i-1/2}\right)\right]. \tag*{(46)}

The four estimates compared in Figure 2a are the resolved pressure-level integral K‾0\overline{K}_0, the half-bottom-value surface correction K‾corr\overline{K}_{\mathrm{corr}}, the constant-value surface correction K‾const\overline{K}_{\mathrm{const}}, and the extrapolated surface correction K‾extr\overline{K}_{\mathrm{extr}}. The value discussed in the paper and shown in Figure 5 is

WK=K‾corr.W_K = \overline{K}_{\mathrm{corr}}.

The differences between the adopted estimate and the constant-value and extrapolated estimates characterize the sensitivity of WKW_K to the unresolved surface layer (Figure 2a).

A Lorenz-type formulation for mass-related pressure work

Here we develop a simplified mass-related available-energy framework to examine the relationship between surface-pressure-gradient work and the temperature-related contribution to kinetic energy generation (§ 5.3).

Consider a horizontally closed, hydrostatic atmosphere with the column gaseous mass m=ps/gm=p_s/g, where psp_s is surface air pressure. Writing p^s≡S−1∫Sps dS\widehat{p}_s \equiv S^{-1}\int_S p_s\,dS, the mass-related available-energy functional is

Am=RT0MgS∫Spsln⁡psp^s dS.(47)A_m = \frac{RT_0}{MgS}\int_S p_s\ln\frac{p_s}{\widehat{p}_s}\,dS. \tag*{(47)}

This functional is nonnegative, vanishing only for horizontally uniform surface pressure.6

The local source of gaseous mass is E−CE-C, using the net evaporation and condensation rates defined in Section 2.1. Define the column horizontal gaseous-mass flux and the column net source as

J=∫0∞ρu dz,q=∫0∞(E−C) dz.(48)\mathbf{J} = \int_0^\infty\rho\mathbf{u}\,dz,\qquad q = \int_0^\infty(E-C)\,dz. \tag*{(48)}

Integrating the continuity equation (32) over height gives

1g∂ps∂t+∇h⋅J=q.(49)\frac{1}{g}\frac{\partial p_s}{\partial t}+\nabla_h\cdot\mathbf{J}=q. \tag*{(49)}

Differentiating Eq. (47), using Eq. (49), and integrating by parts over the closed horizontal domain yields

dAmdt=Gm−Wm,(50)\frac{\mathrm{d}A_m}{\mathrm{d}t}=G_m-W_m, \tag*{(50)}

where

Gm=RT0MS∫Sqln⁡psp^s dS(51)G_m=\frac{RT_0}{MS}\int_S q\ln\frac{p_s}{\hat{p}_s}\,\mathrm{d}S \tag*{(51)}

is the mass-related available-energy generation, and

Wm=−RT0MS∫SJ⋅∇hln⁡ps dS(52)W_m=-\frac{RT_0}{MS}\int_S \mathbf{J}\cdot\nabla_h\ln p_s\,\mathrm{d}S \tag*{(52)}

is the area-mean surface pressure-gradient work. In an isothermal atmosphere with constant molar mass MM and T=T0T=T_0, we have ∇hln⁡p(z)=∇hln⁡ps\nabla_h\ln p(z)=\nabla_h\ln p_s. In this case, WmW_m equals the kinetic energy generation WKW_{\mathrm{K}} defined by Eq. (18).

In steady state, Gm=WmG_m=W_m. Equation (51) shows that GmG_m depends on the spatial covariance between the column gaseous-mass source qq and surface pressure psp_s. For example, in the tropics, zonal-mean evaporation varies relatively smoothly with latitude, whereas precipitation exhibits pronounced maxima [43 Fig. 1a]. More generally, evaporation is relatively distributed in space and time, while intense condensation is concentrated in precipitation systems, often associated with lower surface pressure. Such a distribution of gaseous-mass sources and sinks can produce a positive covariance between qq and psp_s.

For an order-of-magnitude estimate, we assume a characteristic pressure difference Δps=10\Delta p_s=10 hPa between gaseous-mass source and sink regions. Taking the mean tropical precipitation rate P≃3.6P\simeq3.6 mm day−1^{-1} from GPCP V3.3 (Figure 3a), p^s=1000\hat{p}_s=1000 hPa, and T0=300T_0=300 K, we obtain

Gm≃PRT0MΔpsp^s≃0.04 W m−2.(53)G_m\simeq P\frac{RT_0}{M}\frac{\Delta p_s}{\hat{p}_s}\simeq0.04\ \mathrm{W\,m^{-2}}. \tag*{(53)}

The associated equivalent velocities for large-scale net mass transport are of order 1 cm s−1^{-1} or less (e.g., Lorenz, [19]). Such estimates have contributed to the conventional neglect of the dynamical importance of the condensation-induced mass sink.

The small magnitude of GmG_m does not imply that pressure work at individual heights is small. Consider the air-mass flow against the surface-pressure gradient, which generates positive kinetic energy near the surface. Mass continuity requires a compensating return flow aloft (Figure 6). We therefore separate the column gaseous-mass flux into lower and upper contributions, J=Jl+Ju\mathbf{J}=\mathbf{J}_l+\mathbf{J}_u, where

Jl=∫0z∗ρu dz,Ju=∫z∗∞ρu dz.(54)\mathbf{J}_l=\int_0^{z_*}\rho\mathbf{u}\,\mathrm{d}z,\qquad\mathbf{J}_u=\int_{z_*}^{\infty}\rho\mathbf{u}\,\mathrm{d}z. \tag*{(54)}

Here z∗z_* denotes the height separating the opposing mass flows, with Jl\mathbf{J}_l directed against the surface-pressure gradient and Ju\mathbf{J}_u in the opposite direction. This idealized separation need not coincide with the 500-hPa boundary used for the observational diagnostics in § 4.1.

Their contributions to the surface-pressure-gradient work are

Wm,l+Wm,u=Wm=Gm,(55)W_{m,l}+W_{m,u}=W_m=G_m, \tag*{(55)}

where

Wm,l(u)=−RT0MS∫SJl(u)⋅∇hln⁡ps dS.(56)W_{m,l(u)}=-\frac{RT_0}{MS}\int_S\mathbf{J}_{l(u)}\cdot\nabla_h\ln p_s\,\mathrm{d}S. \tag*{(56)}

Since the upper atmosphere is generally drier than the lower atmosphere, the return flow carries less water vapor. Denoting the difference in water-vapor mass mixing ratio between the lower and upper flows by Δr=rl−ru\Delta r = r_l-r_u, and their characteristic air-mass transport rate per unit area by FF, moisture conservation gives P≃FΔrP \simeq F\Delta r. For a representative surface-pressure difference Δps\Delta p_s, the positive lower-atmospheric work is therefore

Wm,l≃FRT0MΔpsp^s≃GmΔr.(57)W_{m,l} \simeq F\frac{RT_0}{M}\frac{\Delta p_s}{\hat{p}_s} \simeq\frac{G_m}{\Delta r}. \tag*{(57)}

For Δr≃0.01\Delta r \simeq0.01–0.020.02, the positive lower-atmospheric pressure work can exceed GmG_m by a factor of 50–100. With Gm=0.04 W m−2G_m = 0.04\ \mathrm{W\ m^{-2}}, this yields Wm,l≃2W_{m,l} \simeq2–4 W m−24\ \mathrm{W\ m^{-2}}, with an approximately compensating negative contribution Wm,uW_{m,u} from the upper atmosphere.

We now impose, as a dynamical constraint on our idealized circulation, that the net upper-atmospheric kinetic energy generation be negligible, WK,u≃0W_{K,u} \simeq0.

We consider small horizontal temperature and surface pressure differences about an isothermal reference at temperature T0T_0. Assuming that the horizontal temperature difference is approximately independent of height, hydrostatic balance gives, to first order,

RT0M∇hln⁡p(z)≃RT0M∇hln⁡ps+gzT0∇hT.(58)\frac{RT_0}{M}\nabla_h \ln p(z) \simeq\frac{RT_0}{M}\nabla_h \ln p_s + \frac{gz}{T_0}\nabla_h T. \tag*{(58)}

Thus, kinetic energy generation separates into

WK≃Wm+WT,(59)W_K \simeq W_m + W_T, \tag*{(59)}

where WmW_m, given by Eq. (52), is the surface-pressure-related contribution, while

WT=−gT0S∫SJ(z)⋅∇hT dS,J(z)≡∫0∞zρu dz(60)W_T = -\frac{g}{T_0\mathcal{S}}\int_{\mathcal{S}} \mathbf{J}^{(z)}\cdot\nabla_h T\,dS,\qquad\mathbf{J}^{(z)}\equiv\int_0^\infty z\rho\mathbf{u}\,dz \tag*{(60)}

is the temperature-related contribution. While WmW_m depends on the column-integrated mass flux J\mathbf{J}, WTW_T depends on its height-weighted counterpart J(z)\mathbf{J}^{(z)}. Thus, the opposing lower and upper mass flows can nearly cancel in the surface-pressure-related work WmW_m, while their different heights allow a substantial temperature-related contribution WTW_T.

Consider the upper return flow Ju\mathbf{J}_u defined above, with characteristic mass-flux-weighted height zˉu\bar{z}_u. Let ΔT>0\Delta T > 0 denote the temperature difference between the warmer low-pressure and cooler high-pressure air columns. The corresponding upper-atmospheric kinetic energy generation is

WK,u≃Wm,u+WT,u≃F(−RT0MΔpsp^s+gzˉuT0ΔT).(61)W_{K,u} \simeq W_{m,u}+W_{T,u} \simeq F\left(-\frac{RT_0}{M}\frac{\Delta p_s}{\hat{p}_s}+\frac{g\bar{z}_u}{T_0}\Delta T\right). \tag*{(61)}

The condition WK,u≃0W_{K,u} \simeq0 therefore requires

ΔTT0≃HzˉuΔpsp^s,H≡RT0Mg.(62)\frac{\Delta T}{T_0}\simeq\frac{H}{\bar{z}_u}\frac{\Delta p_s}{\hat{p}_s},\qquad H\equiv\frac{RT_0}{Mg}. \tag*{(62)}

where HH is the atmospheric scale height. This means that the temperature difference required to offset the negative contribution of a given surface-pressure gradient aloft is not unique. The higher the return flow, the smaller the required temperature difference. For example, with T0=300 KT_0 = 300\ \mathrm{K} and Δps=10 hPa\Delta p_s = 10\ \mathrm{hPa} at p^s=1000 hPa\hat{p}_s = 1000\ \mathrm{hPa}, an upper return flow at the tropical tropopause height zˉu=16 km\bar{z}_u = 16\ \mathrm{km} requires a temperature difference of only ΔT≃1.6 K\Delta T \simeq1.6\ \mathrm{K}, whereas for zˉu=H≃8.8 km\bar{z}_u = H \simeq8.8\ \mathrm{km}, ΔT≃3 K\Delta T \simeq3\ \mathrm{K} is required.

When the positive lower-atmospheric work is concentrated near the surface, its temperature-related correction is comparatively small. Consequently, WT≃WT,u≃−Wm,uW_T \simeq W_{T,u} \simeq-W_{m,u} and

WK≃Wm,l≃WT.(63)W_K \simeq W_{m,l} \simeq W_T. \tag*{(63)}

Combining Eqs. (57) and (63), and identifying the lower-atmospheric kinetic energy generation with the condensation-related expansion power WcW_{\mathrm{c}}, as proposed in § 5.3, we obtain for the idealized circulation

Wc≃WK,l≃WK≃WT≃GmΔτ.(64)W_{\mathrm{c}} \simeq W_{\mathrm{K,l}} \simeq W_{\mathrm{K}} \simeq W_{\mathrm{T}} \simeq\frac{G_{\mathrm{m}}}{\Delta\tau}. \tag*{(64)}

This relation illustrates how the small mass-related available-energy generation GmG_{\mathrm{m}} can correspond to a much larger kinetic energy generation in the lower atmosphere, while the same net work can be expressed through the temperature-related contribution WTW_{\mathrm{T}}. The physical interpretation of this correspondence is discussed in § 5.3.

Data and code availability

GPM/DPR spectral latent-heating (SLH) Level-2 V07C orbital data were obtained from NASA’s Precipitation Processing System (PPS) archive (https://arthurhouhttps.pps.eosdis.nasa.gov/gpmallversions/V07/). MERRA-2 three-hourly pressure-level data (M2I3NPASM, version 5.12.4) were obtained from NASA’s Goddard Earth Sciences Data and Information Services Center (GES DISC) (https://goldsmr5.gesdisc.eosdis.nasa.gov/data/MERRA2/M2I3NPASM.5.12.4/). The processing code, definitions of archived variables, computational procedures, intermediate results, and reproducibility checks are available in the accompanying Zenodo archive https://doi.org/10.5281/zenodo.24269780.

Funding statement

This work received institutional support from the Biotic Pump Greening Institute and the Institute for Advanced Study, Technical University of Munich. No specific grant funding was received.

Conflict of interest statement

The authors declare no conflict of interest.

Acknowledgments

The authors have no additional acknowledgments to report.

References

  1. [1]Andrews DG. 2010. An introduction to atmospheric physics. 2nd edn. Cambridge University Press: New York, ISBN 978-0-521-69318-9, doi:10.1017/CBO9780511800788.DOI
  2. [2]Behrangi A, Song Y, Huffman GJ, Adler RF. 2024. Comparative analysis of the latest global oceanic precipitation estimates from GPM V07 and GPCP V3.2 products. J. Hydrometeorol. 25(2): 293–309, doi:10.1175/JHM-D-23-0082.1.DOI
  3. [3]Behrangi A, Huffman GJ, Adler RF, Song Y, Kumah KK, Zandi O, Bolvin DT, Nelkin EJ, Gu G. 2026. GPCP Version 3.3: A refined global view of precipitation from satellite and gauge observations. J. Hydrometeorol. 27(10): 1745–1760, doi:10.1175/JHM-D-26-0048.1.DOI
  4. [4]Bohren CF, Albrecht BA. 2023. Atmospheric thermodynamics. 2nd edn. Oxford University Press: New York, ISBN 978-0-19-887270-2, doi:10.1093/oso/9780198872702.001.0001.DOI
  5. [5]Brunt D. 1934. Physical and dynamical meteorology. Cambridge University Press: Cambridge.
  6. [6]Emanuel KA. 1999. The power of a hurricane: An example of reckless driving on the information superhighway. Weather 54(4): 107–108, doi:10.1002/j.1477-8696.1999.tb06435.x.
  7. [7]Forster P, Storelvmo T, Armour K, Collins W, Dufresne JL, Frame D, Lunt DJ, Mauritsen T, Palmer MD, Watanabe M, Wild M, Zhang H. 2021. The Earth’s energy budget, climate feedbacks, and climate sensitivity. In: Climate Change 2021: The Physical Science Basis, Chapter 7, Cambridge University Press: 923–1054, doi:10.1017/9781009157896.009.
  8. [8]Gelaro R, McCarty W, Suárez MJ, Todling R, Molod A, Takacs L, Randles CA, Darmenov A, Bosilovich MG, Reichle R, Wargan K, Coy L, Cullather R, Draper C, Akella S, Buchard V, Conaty A, da Silva AM, Gu W, Kim GK, Koster R, Lucchesi R, Merkovа D, Nielsen JE, Partyka G, Pawson S, Putman W, Rienecker M, Schubert SD, Sienkiewicz M, Zhao B. 2017. The modern-era retrospective analysis for research and applications, version 2 (MERRA-2). J. Clim. 30(14): 5419–5454, doi:10.1175/JCLI-D-16-0758.1.
  9. [9]Gorshkov VG. 1995. Physical and biological bases of life stability: Man, biota, environment. 1st edn. Springer: Berlin, Heidelberg, ISBN 978-3-540-57049-3, doi:10.1007/978-3-642-85001-1.
  10. [10]Hamada A, Yokoyama C, Tsuji H, Ikuta Y, Shige S, Yamaji M, Kubota T, Takayabu YN. 2025. Spectral latent heating retrieval for the midlatitudes using GPM DPR. Part II: Development and consistency check of the retrieval algorithm. J. Appl. Meteorol. Climatol. 64(1): 45–61, doi:10.1175/JAMC-D-23-0218.1.
  11. [11]Holland GJ. 1980. An analytic model of the wind and pressure profiles in hurricanes. Mon. Wea. Rev. 108(8): 1212–1218, doi:10.1175/1520-0493(1980)108<1212:AAMOTW>2.0.CO;2.
  12. [13]Huang J, McElroy MB. 2015. A 32-year perspective on the origin of wind energy in a warming climate. Renew. Energy 77: 482–492, doi:10.1016/j.renene.2014.12.045.
  13. [14]Huffman GJ, Behrangi A, Bolvin DT, Nelkin EJ. 2025. GPCP Version 3.3 Satellite-Gauge (SG) Combined Precipitation Data Set (GPCPMON 3.3). NASA Goddard Earth Sciences Data and Information Services Center (GES DISC): Greenbelt, MD, doi:10.5067/MEASURES/GPCP/DATA306. Accessed 17 May 2026.
  14. [15]Jansen MF, Chou HH, Seidel S. 2026. A theory for the global wind energy dissipation across climates. AGU Advances 7(5): e2026AV002593, doi:10.1029/2026AV002593.
  15. [16]Kleidon A. 2021. Physical limits of wind energy within the atmosphere and its use as renewable energy: From the theoretical basis to practical implications. Meteorol. Z. 30(3): 203–225, doi:10.1127/metz/2021/1062.
  16. [17]Laliberté F, Zika J, Mudryk L, Kushner PJ, Kjellsson J, Döös K. 2015. Constrained work output of the moist atmospheric heat engine in a warming climate. Science 347(6221): 540–543, doi:10.1126/science.1257103.
  17. [18]Landau LD, Akhiezer AI, Lifshitz EM. 1967. General physics: Mechanics and molecular physics. 1st edn. Pergamon Press: Oxford, ISBN 978-0-08-009106-8, doi:10.1016/C2009-0-10924-8.DOI
  18. [19]Lorenz EN. 1967. The nature and theory of the general circulation of the atmosphere. World Meteorological Organization: Geneva, WMO-No. 218, TP 115, 161 pp.
  19. [20]Luo X, Ge J, Guo W, Fan L, Chen C, Liu Y, Yang L. 2022. The biophysical impacts of deforestation on precipitation: Results from the CMIP6 model intercomparison. J. Clim. 35(11): 3293–3311, doi:10.1175/JCLI-D-21-0689.1.DOI
  20. [21]Makarieva AM, Gorshkov VG. 2007. Biotic pump of atmospheric moisture as driver of the hydrological cycle on land. Hydrol. Earth Syst. Sci. 11(2): 1013–1033, doi:10.5194/hess-11-1013-2007.DOI
  21. [22]Makarieva AM, Gorshkov VG. 2011. Radial profiles of velocity and pressure for condensation-induced hurricanes. Phys. Lett. A 375(7): 1053–1058, doi:10.1016/j.physleta.2011.01.005.DOI
  22. [23]Makarieva AM, Gorshkov VG, Nefiodov AV, Sheil D, Nobre AD, Bunyard P, Li BL. 2013. The key physical parameters governing frictional dissipation in a precipitating atmosphere. J. Atmos. Sci. 70(9): 2916–2929, doi:10.1175/JAS-D-12-0231.1.DOI
  23. [24]Makarieva AM, Gorshkov VG, Nefiodov AV, Sheil D, Nobre AD, Li B-L. 2017a. Quantifying the global atmospheric power budget. arXiv preprint 1603.03706v4 [physics.ao-ph], https://arxiv.org/abs/1603.03706.
  24. [25]Makarieva AM, Gorshkov VG, Nefiodov AV, Sheil D, Nobre AD, Shearman PL, Li BL. 2017b. Kinetic energy generation in heat engines and heat pumps: The relationship between surface pressure, temperature and circulation cell size. Tellus A 69(1): 1272752, doi:10.1080/16000870.2016.1272752.DOI
  25. [26]Makarieva AM, Gorshkov VG, Nobre AD, Nefiodov AV, Sheil D, Nobre P, Li BL. 2019. Comments on “Is condensation-induced atmospheric dynamics a new theory of the origin of the winds?”. J. Atmos. Sci. 76(7): 2181–2185, doi:10.1175/JAS-D-18-0358.1.DOI
  26. [27]Makarieva AM, Nefiodov AV. 2025. Alternative expressions for the maximum potential intensity of tropical cyclones. Phys. Fluids 37(3): 036620, doi:10.1063/5.0253001.DOI
  27. [28]Makarieva AM, Nefiodov AV. 2026. The alpha–omega problem in the assessments of atmospheric power. arXiv 2610.38522, doi:10.48550/arXiv.2610.38522.DOI
  28. [29]Margules M. 1910. The mechanical equivalent of any given distribution of atmospheric pressure, and the maintenance of a given difference in pressure. In: The Mechanics of the Earth’s Atmosphere: A Collection of Translations, Smithsonian Miscellaneous Collections, vol. 51, no. 4, Public C ed., Smithsonian Institution: Washington, D.C., pp. 501–532. English translation of the 1901 German original.
  29. [30]Mulligan JF, Hertz HG. 1997. An unpublished lecture by Heinrich Hertz: “On the energy balance of the Earth”. Am. J. Phys. 65(1): 36–45, doi:10.1119/1.18565.DOI
  30. [31]Oh JS, Kim YH, Min SK, Park HS, Park JY, Park DSR, Kim MK. 2026. Systematic comparison of the Lorenz energy cycle in four reanalysis datasets. Clim. Dyn. 64(5): 220, doi:10.1007/s00382-026-08153-9.DOI
  31. [32]Paccini L, Stevens B. 2023. Assessing precipitation over the Amazon Basin as simulated by a storm-resolving model. J. Geophys. Res. Atmos. 128(4): e2022JD037436, doi:10.1029/2022JD037436.DOI
  32. [33]Pauluis O, Balaji V, Held IM. 2000. Frictional dissipation in a precipitating atmosphere. J. Atmos. Sci. 57(7): 989–994, doi:10.1175/1520-0469(2000)057<0989:FDIAPA>2.0.CO;2.doi.org/10.1175/1520-0469(2000)057<0989:fdiapa>2.0.co;2
  33. [36]Paulius O, Dias J. 2012. Satellite estimates of precipitation-induced dissipation in the atmosphere. Science 335(6071): 953–956, doi:10.1126/science.1215869.
  34. [37]Paulius O, Dias J. 2013. Corrections and clarifications. Science 339(6117): 271, doi:10.1126/science.339.6117.271-b.DOI
  35. [39]Peixoto JP, Oort AH. 1992. Physics of climate. American Institute of Physics: New York, ISBN 978-0-88318-712-8.
  36. [40]Romanski J, Rossow WB. 2013. Contributions of individual atmospheric diabatic heating processes to the generation of available potential energy. J. Clim. 26(12): 4244–4263, doi:10.1175/JCLI-D-12-00457.1.DOI
  37. [41]Shaw TA, Arblaster JM, Birner T, Butler AH, Domeisen DIV, Garfinkel CI, Garmy H, Grise KM, Karpechko AY. 2024. Emerging climate change signals in atmospheric circulation. AGU Advances 5(6): e2024AV001297, doi:10.1029/2024AV001297.DOI
  38. [42]Shepherd TG. 2014. Atmospheric circulation as a source of uncertainty in climate change projections. Nat. Geosci. 7(10): 703–708, doi:10.1038/ngeo2253.DOI
  39. [43]Siler N, Bailey A, Roe GH, Buizert C, Markle B, Noone D. 2021. The large-scale, long-term coupling of temperature, hydrology, and water isotopes. J. Clim. 34(16): 6725–6742, doi:10.1175/JCLI-D-20-0563.1.DOI
  40. [44]Singh MS, O’Neill ME. 2022. The climate system and the second law of thermodynamics. Rev. Mod. Phys. 94(1): 015001, doi:10.1103/RevModPhys.94.015001.DOI
  41. [45]Smith RK, Montgomery MT, Kilroy G. 2018. The generation of kinetic energy in tropical cyclones revisited. Q. J. R. Meteorol. Soc. 144(717): 2481–2490, doi:10.1002/qj.3332.DOI
  42. [46]Steinheimer M, Hantel M, Bechtold P. 2008. Convection in Lorenz’s global energy cycle with the ECMWF model. Tellus A 60(5): 1001–1022, doi:10.1111/j.1600-0870.2008.00348.x.DOI
  43. [47]Tao W-K, Takayabu YN, Lang S, Shige S, Olson WS, Hou A, Skofronick-Jackson G, Jiang X, Zhang C, Lau W, Krishnamurti T, Waliser D, Grecu M, Ciesielski PE, Johnson RH, Houze R, Kakar R, Nakamura K, Braun S, Hagos S, Oki R, Bhardwaj A. 2016. TRMM latent heating retrieval: Applications and comparisons with field campaigns and large-scale analyses. Meteorol. Monogr. 56(1): 2.1–2.34, doi:10.1175/AMSMONOGRAPHS-D-15-0013.1.DOI
  44. [48]Wallace JM, Hobbs PV. 2006. Atmospheric science: An introductory survey, 2nd edn, International Geophysics, vol. 92. Academic Press: Amsterdam, ISBN 978-0-12-732951-2, doi:10.1016/C2009-0-00034-8.DOI
  45. [49]Wang S, Yang J. 2023. Lorenz energy cycle: Another way to understand the atmospheric circulation on tidally locked terrestrial planets. Planet. Sci. J. 4(5): 84, doi:10.3847/PSJ/accf11.DOI
  46. [50]Wang CC, Lee WL, Hsu HH, Kuo WC, Lin YS. 2024. The global atmospheric energy cycle in TaiESM1: present and future. J. Geophys. Res. Atmos. 129(20): e2024JD041108, doi:10.1029/2024JD041108.DOI
  47. [51]Yang S, Lao V, Bankert R, Whitcomb TR, Cossuth J. 2021. Improved climatology of tropical cyclone precipitation from satellite passive microwave measurements. J. Clim. 34(11): 4521–4537. doi:10.1175/JCLI-D-20-0196.1.DOI
  48. [52]Yoon A, Hohenegger C. 2025. Muted Amazon rainfall response to deforestation in a global storm-resolving model. Geophys. Res. Lett. 52(4): e2024GL110503, doi:10.1029/2024GL110503.DOI

Paper details

vlnc:W_wnd57zz7
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents