Small-scale Convection Research Articles

Modeling the interplay between sea ice formation and the oceanic mixed layer: Limitations of simple brine rejection parameterizations

The subtle interplay between sea ice formation and ocean vertical mixing is hardly represented in current large-scale models designed for climate studies. Convective mixing caused by the brine release when ice forms is likely to prevail in leads and thin ice areas, while it occurs in models at the much larger horizontal grid cell scale. Subgrid-scale parameterizations have hence been developed to mimic the effects of small-scale convection using a vertical distribution of the salt rejected by sea ice within the mixed layer, instead of releasing it in the top ocean layer. Such a brine rejection parameterization is included in the global ocean–sea ice model NEMO-LIM3. Impacts on the simulated mixed layers and ocean temperature and salinity profiles, along with feedbacks on the sea ice cover, are then investigated in both hemispheres. The changes are overall relatively weak, except for mixed layer depths, which are in general excessively reduced compared to observation-based estimates. While potential model biases prevent a definitive attribution of this vertical mixing underestimation to the brine rejection parameterization, it is unlikely that the latter can be applied in all conditions. In that case, salt rejections do not play any role in mixed layer deepening, which is unrealistic. Applying the parameterization only for low ice–ocean relative velocities improves model results, but introduces additional parameters that are not well constrained by observations.

Passive scalar convective-diffusive subrange for low Prandtl numbers in isotropic turbulence.

In this Rapid Communication, we study the behavior of a strongly diffusive passive scalar field T submitted to a freely decaying, homogeneous and isotropic turbulence with eddy-damped quasinormal Markovian simulations. We present a new subrange located between the k(-17/3) inertial-diffusive subrange and the Kolmogorov wave number k(η). This subrange is generated by small-scale convection linked to k(η) that balances diffusion effects. Thus, we build a typical length scale k(CD)(-1) based on convection and diffusion and give an expression for the shape of the passive scalar spectrum in this subrange E(T)∼√[Pr]k(-11/3) using physical arguments. This result unifies two different theories coming from Batchelor [G. K. Batchelor, J. Fluid. Mech. 5, 113 (1959)] and Chasnov [J. Chasnov et al., Phys. Fluids A 1, 1698 (1989)] and explains results previously obtained experimentally.

Melting Driven Convection at the Ice-seawater Interface

We performed direct numerical simulations to investigate the interaction of a vertical ice interface with seawater, and particularly the small-scale convection generated when a wall of Antarctic ice melts in the salty and warmer seawater. The three coupled interface equations are used, along with the Boussinesq and non-hydrostatic governing equations of motion and equation of state for seawater, to solve for interface temperature, salinity and melt rate. Fluxes of both heat and salt to the interface play significant roles in rate of ice melting. The main focus is on the dissolving of ice (at ambient water temperatures between −1° C and 6° C and salinity around 35 ‰) as characterizes many sites around Antarctica. Under these conditions the diffusion of salt to the ice-waterinterface depresses the freezing point and further enhances heat diffusion to the ice.

HIGH-RESOLUTION CALCULATION OF THE SOLAR GLOBAL CONVECTION WITH THE REDUCED SPEED OF SOUND TECHNIQUE. II. NEAR SURFACE SHEAR LAYER WITH THE ROTATION

We present a high-resolution, highly stratified numerical simulation of rotating thermal convection in a spherical shell. Our aim is to study in detail the processes that can maintain a near surface shear layer (NSSL) as inferred from helioseismology. Using the reduced speed of sound technique we can extend our global convection simulation to $0.99\,R_{\odot}$ and include near the top of our domain small-scale convection with short time scales that is only weakly influenced by rotation. We find the formation of a NSSL preferentially in high latitudes in the depth range $r=0.95-0.975R_\odot$. The maintenance mechanisms are summarized as follows. Convection under weak influence of rotation leads to Reynolds stresses that transport angular momentum radially inward in all latitudes. This leads to the formation of a strong poleward directed meridional flow and a NSSL, which is balanced in the meridional plane by forces resulting from the $\langle v'_r v'_\theta\rangle$ correlation of turbulent velocities. The origin of the required correlations depends to some degree on latitude. In high latitudes a positive correlation $\langle v'_rv'_\theta\rangle$ is induced in the NSSL by the poleward meridional flow whose amplitude increases with the radius, while a negative correlation is generated by the Coriolis force in bulk of the convection zone. In low latitudes a positive correlation $\langle v'_rv'_\theta\rangle$ results from rotationally aligned convection cells ("banana cells"). The force caused by these Reynolds stresses is in balance with the Coriolis force in the NSSL.

Role of random thermal perturbations in the magmatic segmentation of mid-oceanic ridges: Insights from numerical simulations

Using a random thermal perturbation (RTP) model this study investigates the process of magmatic segmentation along mid-oceanic ridge (MOR) axes as a function of the upwelling dynamics, controlled by coupled solidification–melting processes. The RTP model suggests that the variation in along-axis velocity (VL) fields constitutes the underlying mechanism of segmentation in natural MORs, showing temperature variations within a steady-state range, irrespective of large initial thermal perturbations imposed at the model base. The VL patterns are initially transient, characterized by multi-order segments, but attain a stable configuration with dominantly large segments (average size ~100km) within a time scale of 2.3Ma. Buoyant-melt driven thermal convection explains this transient segmentation. Small scale convection cells are found to be progressively consumed by larger cells, resulting in a stable convection structure over a similar time scale. Slow- and fast-spreading ridges (SSR and FSR) undergo upwelling with contrasting melt flow patterns. SSRs involve melt feeding into the ridge axis by horizontal flows from segment centers, trailing into large-scale conduits at an early stage. With time, vertical upwelling occurs throughout the segment. In the case of FSRs, both melt supply avenues prevail throughout their development. We also evaluate the variation of the across-axis flow velocity (VT) to investigate the mode of geometric evolution of MORs. Time series VT maps suggest that a ridge structure develops through localization of discrete axes (VT=0) with offsets varying up to 15km, which coalesce with one another to form a single axis. The matured ridge, however, retains higher-order offsets (up to 9km).

Mantle dynamics in the Mediterranean

The Mediterranean offers a unique opportunity to study the driving forces of tectonic deformation within a complex mobile belt. Lithospheric dynamics are affected by slab rollback and collision of two large, slowly moving plates, forcing fragments of continental and oceanic lithosphere to interact. This paper reviews the rich and growing set of constraints from geological reconstructions, geodetic data, and crustal and upper mantle heterogeneity imaged by structural seismology. We proceed to discuss a conceptual and quantitative framework for the causes of surface deformation. Exploring existing and newly developed tectonic and numerical geodynamic models, we illustrate the role of mantle convection on surface geology. A coherent picture emerges which can be outlined by two, almost symmetric, upper mantle convection cells. The downwellings are found in the center of the Mediterranean and are associated with the descent of the Tyrrhenian and the Hellenic slabs. During plate convergence, these slabs migrated backward with respect to the Eurasian upper plate, inducing a return flow of the asthenosphere from the backarc regions towards the subduction zones. This flow can be found at large distance from the subduction zones, and is at present expressed in two upwellings beneath Anatolia and eastern Iberia. This convection system provides an explanation for the general pattern of seismic anisotropy in the Mediterranean, first-order Anatolia and Adria microplate kinematics, and may contribute to the high elevation of scarcely deformed areas such as Anatolia and Eastern Iberia. More generally, the Mediterranean is an illustration of how upper mantle, small-scale convection leads to intraplate deformation and complex plate boundary reconfiguration at the westernmost terminus of the Tethyan collision.

Effect of temperature-dependent viscosity on mantle convection

Two-dimensional numerical calculations in cylindrical shell geometry have been carried out to investigate the effect of the temperature-dependent viscosity on the pattern and the characteristic parameters of the thermal convection occurring in the Earth’s mantle. Systematic model runs established that the viscosity decreasing with the temperature is reduced around the hot core-mantle boundary (CMB) which facilitates ‘the heat transport’ from ‘the core to the mantle’. On the other hand, the viscosity increases near the cold surface which hinders the heat outcome and results in higher mantle temperature and lower surface velocity. A power law relation was revealed between the strength of the temperature-dependence and the observed parameters, such as the velocity, surface mobility, heat flow, average temperature and viscosity. Two additional ‘mantle-like’ models were built up with extra strong temperature-dependent viscosity to imitate the flow in the Earth’s mantle. In model 1, in which the viscosity decreases seven orders of magnitude with the temperature increase, a highly viscous stagnant lid evolves along the cold surface which does not participate in the convection. The existence of the stagnant surface lid reduces the surface heat flow and generates a low viscosity asthenosphere beneath the lid with vigorous small-scale convection. In model 2, in which the viscosity decreases only six orders of magnitude with the temperature and the pressure-dependent viscosity is stronger, does not form a surface stagnant lid, highly viscous ‘slabs’ submerge to the CMB and effectively influence the hot upwelling plumes. Based on our numerical results it is necessary to implicate the yield stress into the simulations in order to obtain a highly viscous, ‘rigid’ surface lid, the lithosphere which can be broken up and subduct down to the mantle.

Mixing at mid-ocean ridges controlled by small-scale convection and plate motion

The geochemical variability of lavas erupted at mid-ocean ridges is lowest where plate spreading rates are high, implying that large-scale plate motions mix the mantle—yet lavas erupted at slow-spreading ridges are also quite homogeneous. Numerical simulations suggest that small-scale convection in the mantle mixes and homogenizes the lavas erupted at slow-spreading ridges. Oceanic lavas are thought to be derived from different sources within the Earth’s mantle, each with a distinct composition1,2,3,4. Large-scale plate motions provide the primary mechanism for mixing these sources, yet the geochemical signature of lavas erupted at different mid-ocean ridges can still vary significantly5,6. Geochemical variability is low where plate spreading rates are high, consistent with plate-scale mixing5,6. However, slow-spreading centres, such as the Southwest Indian Ridge in the Indian Ocean, are also geochemically homogeneous, which is inconsistent with plate-scale mixing6,7. Here we use numerical simulations of mantle flow to study mantle mixing at mid-ocean ridges, under conditions with variable plate length and spreading rate. Our simulations reveal that small-scale convection in the mantle contributes significantly to mantle mixing at slow spreading rates; faster plate velocities and smaller plates inhibit small-scale convection. We conclude that whereas fast-spreading ridge lavas are well mixed by plate-scale flow, slow-spreading ridge lavas are mixed by small-scale convection.

Effects of explicit atmospheric convection at high CO2.

The effect of clouds on climate remains the largest uncertainty in climate change predictions, due to the inability of global climate models (GCMs) to resolve essential small-scale cloud and convection processes. We compare preindustrial and quadrupled CO2 simulations between a conventional GCM in which convection is parameterized and a "superparameterized" model in which convection is explicitly simulated with a cloud-permitting model in each grid cell. We find that the global responses of the two models to increased CO2 are broadly similar: both simulate ice-free Arctic summers, wintertime Arctic convection, and enhanced Madden-Julian oscillation (MJO) activity. Superparameterization produces significant differences at both CO2 levels, including greater Arctic cloud cover, further reduced sea ice area at high CO2, and a stronger increase with CO2 of the MJO.

열대 지역 Gross Moist Stability 특징 분석 및 미래 변화

This study investigates the characteristics of the Gross Moist Stability (GMS) over the tropics. The GMS summarizes the relationship between large-scale entropy forcing due to radiation and surface fluxes and the response of smaller-scale convection. The GMS is able to explain both to where moist entropy is advected by the atmospheric circulation and how deep the moisture flux convergence is in the tropical region. In the deep convective region, positive GMS appears over the warm pool region due to the strong column-integrated moisture convergence and the ensuing export of moist entropy to the environment. The vertical advection of moist entropy dominates over the horizontal advection in this region. Meanwhile, over the eastern tropical ITCZ region, which is characterized by shallow convective area, import of moist entropy by horizontal winds is dominant compared to the vertical moist entropy advection. Future changes in the GMS are also examined using the 22 CMIP5 model simulations. A decrease in the GMS appears widely across the tropics, but its increase occurs over the western-central equatorial Pacific. It is evident that the increased GMS region corresponds to an increased region of precipitation, implying that strengthened convection in the future due to increased entropy forcing exports the enhanced moist energy to stabilize the environment.

Importance of initial buoyancy field on evolution of mantle thermal structure: Implications of surface boundary conditions

Although there has been significant progress in the seismic imaging of mantle heterogeneity, the outstanding issue that remains to be resolved is the unknown distribution of mantle temperature anomalies in the distant geological past that give rise to the present-day anomalies inferred by global tomography models. To address this question, we present 3-D convection models in compressible and self-gravitating mantle initialised by different hypothetical temperature patterns. A notable feature of our forward convection modelling is the use of self-consistent coupling of the motion of surface tectonic plates to the underlying mantle flow, without imposing prescribed surface velocities (i.e., plate-like boundary condition). As an approximation for the surface mechanical conditions before plate tectonics began to operate we employ the no-slip (rigid) boundary condition. A rigid boundary condition demonstrates that the initial thermally-dominated structure is preserved, and its geographical location is fixed during the evolution of mantle flow. Considering the impact of different assumed surface boundary conditions (rigid and plate-like) on the evolution of thermal heterogeneity in the mantle we suggest that the intrinsic buoyancy of seven superplumes is most-likely resolved in the tomographic images of present-day mantle thermal structure. Our convection simulations with a plate-like boundary condition reveal that the evolution of an initial cold anomaly beneath the Java-Indonesian trench system yields a long-term, stable pattern of thermal heterogeneity in the lowermost mantle that resembles the present-day Large Low Shear Velocity Provinces (LLSVPs), especially below the Pacific. The evolution of subduction zones may be, however, influenced by the mantle-wide flow driven by deeply-rooted and long-lived superplumes since Archean times. These convection models also detect the intrinsic buoyancy of the Perm Anomaly that has been identified as a unique slow feature distinct from the two principal LLSVPs. We find there is no need for dense chemical ‘piles’ in the lower mantle to generate a stable distribution of temperature anomalies that are correlated to the LLSVPs and the Perm Anomaly. Our tomography-based convection simulations also demonstrate that intraplate volcanism in the south-east Pacific may be interpreted in terms of shallow small-scale convection triggered by a superplume beneath the East Pacific Rise.

HIGH-RESOLUTION CALCULATIONS OF THE SOLAR GLOBAL CONVECTION WITH THE REDUCED SPEED OF SOUND TECHNIQUE. I. THE STRUCTURE OF THE CONVECTION AND THE MAGNETIC FIELD WITHOUT THE ROTATION

We carry out non-rotating high-resolution calculations of the solar global convection, which resolve convective scales of less than 10 Mm. To cope with the low Mach number conditions in the lower convection zone, we use the reduced speed of sound technique (RSST), which is simple to implement and requires only local communication in the parallel computation. In addition, the RSST allows us to expand the computational domain upward to about $0.99 R_{\odot}$ as it can also handle compressible flows. Using this approach, we study the solar convection zone on the global scale, including small-scale near-surface convection. In particular, we investigate the influence of the top boundary condition on the convective structure throughout the convection zone as well as on small-scale dynamo action. Our main conclusions are: 1. The small-scale downflows generated in the near-surface layer penetrate into deeper layers to some extent and excite small-scale turbulence in the region $>0.9R_\odot$, where $R_\odot$ is the solar radius. 2. In the deeper convection zone ($<0.9R_\odot$), the convection is not influenced by the location of the upper boundary. 3. Using an LES approach, we can achieve small-scale dynamo action and maintain a field of about $0.15-0.25 B_\mathrm{eq}$ throughout the convection zone, where $B_\mathrm{eq}$ is the equipartition magnetic field to the kinetic energy. 4. The overall dynamo efficiency varies significantly in the convection zone as a consequence of the downward directed Poynting flux and the depth variation of the intrinsic convective scales.

Sublithospheric small-scale convection—A mechanism for collision zone magmatism

We studied the effect of increased water content on the dynamics of the lithosphere-asthenosphere boundary in a postsubduction setting. Results from numerical mantle convection models show that the resultant decrease in mantle viscosity and the peridotite solidus produce small-scale convection at the lithosphere-asthenosphere boundary and magmatism that follows the spatially and temporally scattered style and volumes typical for collision magmatism, such as the late Cenozoic volcanism of the Turkish-Iranian Plateau. An inherent feature in small-scale convection is its chaotic nature that can lead to temporally isolated volcanic centers tens of millions of years after initial continental collision, without evident tectonic cause. We also conclude that water input into the upper mantle during and after subduction under the circum-Mediterranean area and the Tibetan Plateau can account for the observed magmatism in these areas. Only fractions (200–600 ppm) of the water input need to be retained after subduction to induce small-scale convection and magmatism on the scale of those observed from the Turkish-Iranian Plateau.

Double layer anisotropy beneath the New Madrid seismic zone and adjacent areas: insights from teleseismic shear wave splitting

A total of 93 well-defined PKS, 54 SKKS, and 126 SKS shear-wave splitting parameters are determined at 25 broadband seismic stations in an approximately 1000 by 1000 km&lt;sup&gt;2&lt;/sup&gt; area centered at the New Madrid seismic zone (NMSZ) in order to test the existence of two anisotropic layers and to map the direction and strength of mantle fabrics. The individual splitting parameters suggest a significant and systematic spatial and azimuthal variation in the splitting parameters. The azimuthal variations at most stations can be explained as the results of present SW ward asthenospheric flow and NNE trending lithospheric fabrics formed during past orogenic events. In the NMSZ, rift-parallel fast directions (potentially related to a long-rift flow) and rift-orthogonal fast directions from small-scale mantle convection are not observed. In addition, reduction in splitting times as a result of vertical asthenospheric flow is not observed.

Upper mantle structure of the Cascades from full-wave ambient noise tomography: Evidence for 3D mantle upwelling in the back-arc

Melt generation and volcanism at subduction zones may result from several possible processes: hydration of the mantle wedge by fluid released from the slab, subduction-induced mantle upwelling beneath the back-arc, and heating of downgoing sediments/oceanic crust atop the slab. Each process predicts a distinctly different spatial pattern of melt generation and can thus be distinguished with high-resolution seismic imaging. Here we construct an upper mantle model of the Pacific Northwest using a full-wave ambient noise tomographic method. Normalized vertical components of continuous seismic records at station pairs are cross-correlated to extract empirical Green's functions at periods of 7–200 s. We simulate wave propagation within the 3D Earth structure using a finite-difference method and calculate sensitivity kernels of Rayleigh waves to perturbations of Vp and Vs based on the Strain Green's Tensor database. Phase delays are extracted by cross-correlating the observed and synthetic waveforms at multiple frequency bands.Our tomographic result reveals three separate low shear-wave velocity anomalies along the back-arc in the upper mantle ∼200 km east of the Cascade volcanic arc, with the central one being the largest in size and lowest in velocity. These back-arc low-velocity anomalies are spatially correlated with the three arc-volcano clusters. The geometry of the low-velocity volumes relative to the slab and arc is consistent with the pattern of subduction-induced decompressional melting in the back-arc. Their along-strike variation suggests that the large-scale plate-motion-induced flow in the back-arc mantle wedge is modulated by small-scale convection, resulting in a highly 3D process that defines the segmentation of volcanism along the Cascade arc.

A systematic 2‐D investigation into the mantle wedge's transient flow regime and thermal structure: Complexities arising from a hydrated rheology and thermal buoyancy

Arc volcanism at subduction zones is likely regulated by the mantle wedge's flow regime and thermal structure and, hence, numerous studies have attempted to quantify the principal controls on mantle wedge conditions. In this paper, we build on these previous studies by undertaking a systematic 2‐D numerical investigation into how a hydrated rheology and thermal buoyancy influence the wedge's flow regime and associated thermal structure. We quantify the role of a range of plausible: (i) water contents (0–5000 H/106Si); (ii) subduction velocities (2–10 cm/yr); and (iii) upper‐plate ages (50–120 Myr), finding that small‐scale convection (SSC), resulting from Rayleigh‐Taylor instabilities, or drips, off the base of the overriding lithosphere, is a typical occurrence. The morphology of SSC varies with viscosity and subduction parameters, with drips at their most prominent when subduction velocities and wedge viscosities are low. Our results confirm that high subduction velocities and wedge viscosities promote a dominantly corner‐flow regime, and strong upper‐plate erosion below the arc region. By contrast, we find that back‐arc upper‐plate erosion by SSC is largely controlled by wedge viscosity, occurring when: (i) viscosities are &lt; 5·1018 Pa s; and (ii) the length of the upper plate, available for destabilization, exceeds the characteristic wavelength of instabilities. Thus, if hydrous weakening of wedge rheology extends at least 100–150 km from the trench, our 2‐D models predict an unstable flow regime, resulting in temperature fluctuations of 50–100 K, which are sufficient to influence melting and the stability of hydrous minerals.

A Numerical Verification of Self-Similar Multiplicative Theory for Small-Scale Atmospheric Turbulent Convection

自我类似的趋于增加的理论(SSM 理论) ，解释温度结构的可伸缩的行为的目的工作。在现在的纸，作者报告源于大气的狂暴的传送对流的数字模拟以便验证这个理论。模拟基于从简化大气的传送对流方程被推出的一个壳模型。数字结果与从第一份订单放大法律到第八份订单的理论预言同意了很好。他们也比由 Kolmogorovs 理论交上 1941 的证明这个理论的预言好， lognormal，和模型理论。

3D spherical models of Martian mantle convection constrained by melting history

While most of Tharsis rise was in place by end of the Noachian period, at least one volcano on Tharsis swell (Arsia Mons) has been active within the last 2 Ma. This places an important constraint on mantle convection and on the thermal evolution of Mars. The existence of recent volcanism on Mars implies that adiabatic decompression melting and, hence, upwelling convective flow in the mantle remains important on Mars at present. The thermal history on Mars can be constrained by the history of melt production, specifically generating sufficient melt in the first billion years of the planets history to produce Tharsis rise as well as present day melt to explain recent volcanism. In this work, mantle convection simulations were performed using finite element code CitcomS in a 3D sphere starting from a uniformly hot mantle and integrating forward in time for the age of the solar system. We implement constant and decaying radioactive heat sources; and vary the partitioning of heat sources between the crust and mantle, and consider decreasing core–mantle boundary temperature and latent heat of melting. The constant heat source calculations produce sufficient melt to create Tharsis early in Martian history and continue to produce significant melt to the present. Calculations with decaying radioactive heat sources generate excessive melt in the past, except when all the radiogenic elements are in the crust, and none produce melt after 2 Gyr. Producing a degree-1 or degree-2 structure may not be pivotal to explain the Tharsis rise: we present multi-plume models where not every plume produces melt. The Rayleigh number controls the timing of the first peak of volcanism while late-stage volcanism is controlled more by internal mantle heating. Decreasing the Rayleigh number increases the lithosphere thickness (i.e., depth), and increasing lithosphere thickness increases the mean mantle temperature. Increasing pressure reduces melt production while increasing temperature increases melt production; hence predicting melt production from convection parameters is not straightforward. Generating enough melt in the mantle to create Tharsis early on and also to explain recent volcanism may require other mechanisms such as small-scale convection or lowering the thermal conductivity of the crust.

Small-volume melts of lithospheric mantle during continental collision: Late Cenozoic lavas of Mahabad, NW Iran

Basanites and alkali basalts from Mahabad in the West Azerbaijan province of Iran are part of a widespread series of Late Miocene–Quaternary mantle-derived magmas erupted within the Turkish–Iranian orogenic plateau, itself part of the active Arabia–Eurasia collision zone. New elemental and Sr–Nd isotopic results are combined with geophysical and geological constraints to suggest that these lavas formed predominantly by small degrees of partial melting of the thick (≫100km) Eurasian lithospheric mantle within the garnet facies. Samples are highly enriched in large ion lithophile elements (LILE) and the light rare earth elements (LREE), up to 600 times chondritic values. They mostly possess negative primitive mantle-normalised Rb, K, Nb–Ta, Zr–Hf and Ti anomalies, with an overall signature that indicates a mantle source metasomatised by fluids or melts derived from crust during continental collision or the Tethyan oceanic subduction that preceded it. Sr–Nd isotopic values are similar to other Quaternary centres in NW Iran; 87Sr/86Sr is slightly depleted with respect to Bulk Silicate Earth, at ∼0.7045, and 143Nd/144Nd is slightly enriched, at ∼0.5127. Crustal contamination does not appear to be an important process in the chemistry of these samples. Possible triggers for melting may include: breakdown of hydrous phases during lithospheric thickening; hydration of the mantle lithosphere by underthrusting of the Arabian passive margin; small-scale sub-lithospheric convection due to a significant thickness gradient in the Zagros lithosphere. Such processes may account for small-volume syn-collisional mantle-derived magmatism elsewhere in regions of thick lithosphere where recent slab break-off or lithospheric delamination cannot be proven.

Horizontal convection dynamics: insights from transient adjustment

AbstractThe dynamics of horizontal convection are revealed by examining transient adjustment toward thermal equilibrium. We restrict attention to high Rayleigh numbers (of $O(1{0}^{12} )$) and a Prandtl number ${\sim }5$ that characterize many practical applications, and consider responses to small changes in the thermal boundary conditions, using laboratory experiments, three-dimensional direct numerical simulations (DNS) and simple theoretical models. The experiments and the mechanical energy budget from the DNS demonstrate that unsteady forcing can produce flow dramatically more active than horizontal convection under steady forcing. The physical mechanisms at work are indicated by the time scales of approach to the new equilibrium, and we show that these can range over two orders of magnitude depending on the imposed change in boundary conditions. Changes that lead to a net destabilizing buoyancy flux give rapid adjustments: for applied heat flux conditions the whole of the circulation is controlled by conduction through the stable portion of the boundary layer, whereas for applied temperature difference the circulation is controlled by small-scale convection within the unstable part of the boundary layer. The experiments, DNS and models are in close agreement and show that the time scale under applied temperatures is as small as 0.01 vertical diffusion time scales, a factor of four smaller than for imposed flux. Both cases give adjustments too rapid for diffusion in the interior to play a significant role, at least through 99 % of the adjustment, and we conclude that diffusion through the full depth is not significant in setting the equilibrium state. Boundary changes leading to a net stabilizing buoyancy flux give a very different response, causing the convection to quickly form a shallow circulation cell, followed eventually by a return to full-depth overturning through a combination of penetrative convection and conduction. The time scale again varies by orders of magnitude, depending on the boundary conditions and the location of the imposed boundary perturbation.