Hostname: page-component-7b7567c8c-2n9c9 Total loading time: 0 Render date: 2026-11-16T03:56:55.951Z Has data issue: false hasContentIssue false

Turbulent convection in subglacial lakes

Published online by Cambridge University Press:  11 March 2021

Louis-Alexandre Couston*
Affiliation:
British Antarctic Survey, Cambridge CB3 0ET, UKUniv Lyon, ENS de Lyon, Univ Claude Bernard, CNRS, Laboratoire de Physique, F-69342 Lyon, France
*
†Email address for correspondence: louis.couston@ens-lyon.fr

Abstract

Subglacial lakes are isolated, low-temperature and high-pressure water environments hidden under ice sheets. Here, we use two-dimensional direct numerical simulations in order to investigate the characteristic temperature fluctuations and velocities in freshwater subglacial lakes as functions of the ice overburden pressure, $p_i$, the water depth, $h$, and the geothermal flux, $F$. Geothermal heating is the unique forcing mechanism as we consider a flat ice–water interface. Subglacial lakes are fully convective when $p_i$ is larger than the critical pressure $p_*\approx 2848$ dbar, but self-organize into a lower convective bulk and an upper stably stratified layer when $p_i < p_*$, because of the existence at low pressure of a density maximum at temperature $T_d$ greater than the freezing temperature $T_f$. For both high and low $p_i$, we demonstrate that the Nusselt number, $Nu$, and Reynolds number, $Re$, satisfy classical scaling laws provided that an effective Rayleigh number $Ra_{eff}$ is considered. We show that the convective and stably stratified layers at low pressure are dynamically decoupled at leading order because plume penetration is weak and induces limited entrainment of the stable fluid. From the empirical equation for $Nu$ with $Ra_{eff}$, we derive two sets of closed-form expressions for several variables of interest, including the unknown bottom temperature, in terms of the problem parameters $p_i$, $h$ and $F$. The two predictions correspond to two limiting regimes obtained when the effective thermal expansion coefficient is either approximately constant or linearly proportional to the temperature difference driving the convection.

Information

Type
JFM Papers
Copyright
© The Author(s), 2021. Published by Cambridge University Press
Figure 0

Figure 1.(a) Problem schematic. The green shading highlights the region of the water column that stays stably stratified when $p_i \leq p_*$. (b) Thermal expansion coefficient $\beta (T,p)$. The solid black (respectively red) line shows $T_f$ (respectively $T_d$) while the dashed line shows the $p_*$ isobar. The arrows highlight the ice overburden pressures considered in the simulations. (c) Density variations with depth at $t=0$ for each of the four simulation cases $\mathcal {S}_i^1 (i=0,1,2,3)$ of the first experiment corresponding to the different ice pressures $p_i$ shown by arrows in (b).

Figure 1

Table 1.Physical parameters and polynomial approximations for $T_f$, $T_d$, $\rho _1$ and $C$ with temperatures in $^{\circ }$C, pressures in dbar, densities in $\textrm {kg}\ \textrm {m}^{-3}$ and $C$ in $\textrm {kg}\ \textrm {m}^{-3}\ ^{\circ }\textrm {C}^{-2}$.

Figure 2

Figure 2.Graphical illustration of our two sets of numerical experiments in physical space with (a) $h=0.5$ m fixed in the first experiment and (b) $F$ (in W m$^{-2}$) fixed in the second (cf. details in table 2). Thinner lines correspond to a larger ice overburden pressure. (c) Corresponding coverage in dimensionless space $(\bar {S},\overline {Ra}_F)$. The red shading highlights the region of fully convective lakes.

Figure 3

Table 2.Dimensional parameters for the two sets of numerical experiments, which consider four distinct ice overburden pressures $p_i$ (in dbar) each and either a broad range of geothermal fluxes (7th column) or a broad range of water depths (last column). Ice thickness $H$ and water depths $h$, $h_c$ and $h_t$ are in metres and fluxes $F$, $F_c$ and $F_t$ are in W m$^{-2}$; s.n. means simulation name. Note that figure 2 provides a graphical illustration of the dimensional and dimensionless parameter spaces explored.

Figure 4

Figure 3.(a) Snapshots of dimensionless temperature $\tilde {T}$ for simulation $\mathcal {S}_1^1$ with $F$ increasing from top to bottom. Here, $\tilde {T}_d$ is the dimensionless temperature of maximum density whereas $\langle \tilde {T}_b\rangle _h$ is the dimensionless horizontally averaged (but time-dependent) bottom temperature. (b) Same as (a) but for $\mathcal {S}_3^1$. (c) Dimensionless time and horizontally averaged temperature profiles $\langle \tilde {T} \rangle$ with depth at different stages (i.e. different $F$) for $\mathcal {S}_1^1$. (d) Same as (c) but for $\mathcal {S}_3^1$. The line colours go from dark to light as $F$ increases from small to large values (lines shifting from right to left as shown by the black arrows). The thin black lines show the conductive profiles at $\tilde {t}=0$. Panels (e) and (f) show the time evolution of $\langle \tilde {T}_b\rangle _h$ for $\mathcal {S}_1^1$ and $\mathcal {S}_3^1$, respectively. The vertical dashed lines highlight the times $\tilde {t}=i$ ($i=1,2,3\ldots$) when the control parameter ($F$ or $h$) starts increasing (smoothly) and the simulation stage changes, with a new statistical steady state reached before $\tilde {t}=i+0.5$.

Figure 5

Figure 4.(a,b) Dimensionless effective temperature difference $\tilde {\varDelta }_{eff}$ (i.e. driving the convection) as a function of (a) the normalized geothermal flux $(F-F_t)/(F_c-F_t)$ for the simulations of the first experiment and (b) the normalized water depth $(h-h_t)/(h_c-h_t)$ for the simulations of the second experiment (cf. legends and table 2). (c,d) Same as (a,b) but for the dimensionless effective thermal expansion coefficient $\tilde {\beta }_{eff}$. (e–h) Show the same variables as (a–d) but in dimensional form, i.e. with $\varDelta _{eff}=\varDelta \tilde {\varDelta }_{eff}$ and ${\beta }_{eff}=\bar {\beta }_b\tilde {\beta }_{eff}$. The solid lines show scaling laws as discussed in § 3.3 and listed in table 4. Note that the symbol size and line thickness are inversely proportional to $p_i$, i.e. large (respectively small) symbols and thick (respectively thin) lines highlight results for small (respectively large) $p_i$.

Figure 6

Figure 5.(a) Effective Rayleigh number $Ra_{eff}$ as a function of $\overline {Ra}_F$ for all simulations, including the Oberbeck–Boussinesq (OB) simulations $\mathcal {S}_{OB}$ whose results are shown by black crosses. Note that, as in figure 4, large (respectively small) symbols highlight simulation results with small (respectively large) $p_i$. The thick black and thin green solid lines show the predictive expressions for $Ra_{eff}$ as a function of $\overline {Ra}_F$ for simulations $\mathcal {S}_{OB}$ (OB regime) and $\mathcal {S}_{2}^1$ (linear expansion coefficient, or LEC, regime), respectively (cf. table 4); (b) $Ra_{eff}$ as a function of $1+\bar {S}$. The vertical dotted line highlights $\bar {S}=0$; (c) $Ra_{eff}$ for the first experiment only as a function of the normalized heat flux $(F-F_t)/(F_c-F_t)$; (d) $Ra_{eff}$ for the second experiment only as a function of the normalized water depth $(h-h_t)/(h_c-h_t)$. The solid lines in (c) and (d) show the theoretical scalings for $Ra_{eff}$ with $F$ and $h$ (cf. table 4) in the OB regime (thin red lines) and in the LEC regime (all other lines) with $\alpha$ and $\gamma$ listed in table 3.

Figure 7

Figure 6.(a) Nusselt number $Nu$ as a function of $Ra_{eff}$ for all simulations (we use the same symbol colours and size chart as in figures 4 and 5). (b) Reynolds number $Re$ as a function of $Ra_{eff}$. The black solid lines in (a,b) show the power law fit (3.9) and (3.10) for $Nu$ and $Re$ for the results of simulation $\mathcal {S}_{OB}$ (cf. table 3). (c) Compensated Nusselt number as a function of $\overline {Ra}_F$. (d) Compensated Reynolds number as a function of $\overline {Ra}_F$. The black (respectively green) solid lines in (c,d) show the power law fits provided in table 4 for the results of simulation $\mathcal {S}_{OB}$ (respectively $\mathcal {S}_2^1$), i.e. in the OB regime (respectively LEC regime). The light-coloured markers in (a,b,c) show the results for simulations $\mathcal {S}_{i}^{j}$ ($i=0,1$; $j=1,2$) with $Nu$ multiplied by 2/3 and $Ra_{eff}$ multiplied by 3/2 (see details in § 3.4).

Figure 8

Table 3.Best-fit coefficients for the power laws $Nu=a Ra_{eff}^{\alpha }$ and $Re=cRa_{eff}^{\gamma }$ for $Ra_{eff}>10^5$ for all simulations, including the OB simulation $\mathcal {S}_{OB}$. Here, s.n. means simulation name while Rel. err. denotes the maximum relative error between the simulation results and the predictive best-fit power law.

Figure 9

Table 4.Key expressions in terms of the control parameters $\overline {Ra}_F$ and $\bar {S}$ (cf. (2.7a–c)) and asymptotic scaling laws in terms of the water depth $h$ and heat flux $F$ for the dimensionless and dimensional variables of interest to the study of turbulent convection in subglacial lakes. The starting point is the algebraic equation for $\tilde {\varDelta }_{eff}$ (3.13). The difference between the limiting OB regime and LEC regime arises from the independence or variability of the effective thermal expansion coefficient $\tilde {\beta }_{eff}$ with the effective temperature difference $\tilde {\varDelta }_{eff}$. The asymptotic scalings with $F$ and $h$ are valid in the limit $(Fh/k)/|T_d-T_f| \ll 1$ ($\bar {S}\approx -1$) for the OB regime and $(Fh/k)/|T_d-T_f| \gg 1$ ($\bar {S}\approx 0$) for the LEC regime.

Figure 10

Figure 7.(a) Effective stiffness $S_{eff}$ (cf. (3.20)) as a function of $Ra_{eff}$ for simulations $\mathcal {S}_i^j$ ($i=0,1$; $j=1,2$), i.e. with a top stably stratified layer. (b) Ratio between the bulk temperature $\tilde {T}_{bulk}$ (cf. (3.21)) in excess of $\tilde {T}_d(\tilde {T}_d>0)$ and the mean bottom temperature $\langle \tilde {T}_b \rangle$ in excess of $\tilde {T}_d(\tilde {T}_d>0)$ as a function of $Ra_{eff}$. The results are shown for all simulations of the first and second experiments and the dashed line highlights the 1/4 ordinate. (c) Effective stiffness $S_{eff}$ as a function of the expected dimensionless convective layer depth $\tilde {h}_{eff}$ (cf. (3.5)) for all simulations with a top stably stratified layer.

Figure 11

Figure 8.(a–d) Convective plume-driven $q_{cp}$ (dashed lines), entrainment-driven $q_{ce}$ (solid lines) and conductive $q_d$ (thin dotted lines) heat fluxes as functions of water depth $z/h$ ($y$ axis) for simulations $\mathcal {S}_0^1$, $\mathcal {S}_1^1$, $\mathcal {S}_0^2$ and $\mathcal {S}_1^2$, respectively. (e–h) Top of the bottom conductive layer $\tilde {z}=L_{db}$ (big bottom circles), plume-driven convective layer $\tilde {z}=L_{db}+L_{cp}$ (lower horizontal bars), entrainment-driven convective layer $\tilde {z}=L_{db}+L_{cp}+L_{ce}$ (upper horizontal bars) and stably stratified layer $\tilde {z}=L_{db}+L_{cp}+L_{ce}+L_{dt}$ (top squares) as functions of $Ra_{eff}$ for simulations $\mathcal {S}_0^1$, $\mathcal {S}_1^1$, $\mathcal {S}_0^2$ and $\mathcal {S}_1^2$, respectively. The crosses show the height $\tilde {z}_d$ of the mean $\tilde {T}_d$ isotherm. The lines’ colour and symbol colour become lighter as the heat flux or water depth increases.

Figure 12

Figure 9.(a) Thickness of the entrained layer $L_{ce}$ (cf. (3.24a–d)) as a function of $Ra_{eff}$. (b) Entrainment parameter $\mathcal {E}$ as a function of $Ra_{eff}$. (c) Dimensionless height $\tilde {z}_d$ of the $\tilde {T}_d$ isotherm as a function of the expected convective layer thickness $\tilde {h}_{eff}$. The solid line shows $\tilde {z}_d=\tilde {h}_{eff}$.

Figure 13

Figure 10.(a) Colour map of $1+\bar {S}$, with $\bar {S}$ the base-state stiffness parameter, in $(p_i,h)$ space relevant to Antarctic subglacial lakes with heat flux $F=68$ mW m$^{-2}$. The hashed region in the top left corner highlights subglacial lakes that are fully stable because $F=68$ mW m$^{-2}$ is smaller than the threshold heat flux $F_t$. The horizontal dashed line highlights the $p_*$ isobar and corresponds also to the isocontour $1+\bar {S}=1$. The dotted blue lines are isocontours of $1+\bar {S}$ and the coloured circles (see legend) highlight the location of some of the well-known subglacial lakes in parameter space (Couston & Siegert 2021). (b) Same as (a) but for the dimensionless effective thermal driving $\tilde {\varDelta }_{eff}$. The dotted blue lines still show the isocontours of $1+\bar {S}$ whereas the green solid lines show isocontours of $\tilde {\varDelta }_{eff}$.

Figure 14

Figure 11.(a) Time history of the dimensionless effective temperature difference $\langle \tilde {T}_b\rangle _h-\tilde {T}_d(p_i)$ for simulation $\mathcal {S}_{14}^2$ (thick orange line), which is simulation $\mathcal {S}_{1}^2$ discussed in the main text with $h = 4$ m and constant $\tilde {T}_d$, and simulation $\mathcal {S}_{14+}^2$ (thin black line), which includes hydrostatic effects in $\tilde {T}_d$. (b) Same as (a) but for the Reynolds number. (c) Same as (a,b) but for the dynamic pressure r.m.s. $p'_{rms}$ normalized by $10^{-9}$ dbar. (d) Mean vertical profiles of the dimensionless temperature $\tilde {T}$ (solid lines) and of the dimensionless temperature of maximum density $\tilde {T}_d$ (dashed lines) for simulations $\mathcal {S}_{14}^2$ (thick orange lines) and $\mathcal {S}_{14+}^2$ (thin black lines).