The April 2010 North African heatwave: when the water vapor greenhouse effect drives nighttime temperatures

Abstract

North Africa experienced a severe heatwave in April 2010 with daily maximum temperatures ($T_{max}$) frequently exceeding $40\,^{\circ }\mathrm{C}$ and daily minimum temperatures ($T_{min}$) over $27\,^{\circ }\mathrm{C}$ for more than five consecutive days in extended Saharan and Sahelian areas. Observations show that areas and periods affected by the heatwave correspond to strong positive anomalies of surface incoming longwave fluxes ($LW_{in}$) and negative anomalies of incoming shortwave fluxes ($SW_{in}$). The latter are explained by clouds in the Sahara, and by both clouds and dust loadings in the Sahel. However, the strong positive anomalies of $LW_{in}$ are hardly related to cloud or aerosol radiative effects. An analysis based on climate-model simulations (CNRM-AM) complemented by a specially-designed conceptual soil-atmospheric surface layer model (SARAWI) shows that this positive anomaly of $LW_{in}$ is mainly due to a water vapor greenhouse effect. SARAWI, which represents the two processes driving temperatures, namely turbulence and longwave radiative transfer between the soil and the atmospheric surface layer, points to the crucial impact of synoptic low-level advection of water vapor on $T_{min}$. By increasing the atmospheric infrared emissivity, the advected water vapor dramatically increases the nocturnal radiative warming of the soil surface, then in turn reducing the nocturnal cooling of the atmospheric surface layer, which remains warm throughout the night. Over Western Sahel, this advection is related to an early northward incursion of the monsoon flow. Over Sahara, the anomalously high precipitable water is due to a tropical plume event. Both observations and simulations support this major influence of the low-level water vapor radiative effect on $T_{min}$ during this spring heatwave.

Appendix: Configuration of the SARAWI simulations and tuned coefficients

In the SARAWI simulations used in the present study, physiographic and physical parameters are statistically tuned using the monthly-average values resolved by CNRM-AM. We also differentiate nighttime and daytime conditions when the considered parameter physically depends on static stability. This leads to:

$\epsilon _a$: Coefficients $a_i$ of Eq. 9 are estimated using longwave fluxes simulated by CNRM-AM which are regressed with the atmospheric specific humidity and air temperature. We can consider $a_i$ coefficients obtained from a regression that include all points in North Africa, or alternatively use $a_i$ coefficients which vary depending on the climate zone (Sahara, Sahel, Guinea). Both approaches lead to similar results (with a 5.3 W/m$^2$ or $1.3\%$ uncertainty on $LW_{in}$ and a $0.28\,^{\circ }\mathrm{C}$ uncertainty on $T_{2m}$). A regional fitting over North Africa gives $a_1=0.667$, $a_2=1.17\times 10^{-2}$ with hus in g/kg and $a_3=4.55\times 10^{-4}$ with $T_a$ in $\,^{\circ }\mathrm{C}$.

$\epsilon _s$: As for $\epsilon _a$, we use CNRM-AM longwave fluxes to estimate $\epsilon _s$ (which uses the ECOCLIMAP database, Champeaux et al. 2005; Faroux et al. 2013). It is almost constant and equal to $0.9946\pm 0.0065$ over all North Africa in CNRM-AM. We take this mean-value as a constant for all continental grid points (this leads to a 0.11 W/m$^2$ or 0.03% uncertainty on $LW_{in}$ and a $0.08\,^{\circ }\mathrm{C}$ uncertainty on $T_{2m}$ as compared to the local value for each grid point).

$C_s$: In order to correctly fit $C_s$, we use Eq. (2) with the resolved fluxes and temperatures given by CNRM-AM, which takes its soil physiographic characteristics from the ECOCLIMAP database (Champeaux et al. 2005; Faroux et al. 2013). We average the different terms for each grid point separately over daytime and nighttime, from which we estimate two physiographic 2D fields $C_s^{night}(lon,lat)$ and $C_s^{day}(lon,lat)$.

$h_{rad}$: We compute $h_{rad}=c_{rad}.\delta z$ at each grid point by determining the value of $h_{rad}$ that minimizes the root mean square error between the CNRM-AM values of $\frac{\partial T_a}{\partial t}_{rlw}$ and the estimated values of that tendency according to Eq. (6). Results show that the value of $c_{rad}$ is very homogeneous over all continental North Africa, so we choose to keep one constant value in SARAWI equal to the average over the continental area: $h_{rad}=4.74.\delta z$. Physically, $h_{rad}$ corresponds to a characteristic penetration depth of the upwelling longwave flux emitted by the surface, or alternatively to the depth of the layer radiatively warmed (or cooled) by the surface.

$h_{turb}$: It is fixed equal to the height between the first and the second layers of the CNRM-AM simulation (35 m).

$c_{t2m}$: The parameterization available in CNRM-AM (Mahfouf et al. 1995) is used here to prescribe $c_{t2m}$, in order to facilitate comparison with the diagnosed $T_{2m}$ in CNRM-AM simulation.

$C_h$, $K_s$, $K_h$: In coupled soil-atmospheric models, the drag coefficient $C_h$ usually depends on the static stability (see Noilhan and Mahfouf 1996 for the ISBA model used in CNRM-AM). Similarly, the turbulent diffusivity in the low layers also strongly varies with the static stability (e.g. Yasuda 1988; Largeron et al. 2010).

Here, we choose to use constant daytime values: $C_h=4.10^{-3}$, $K_s=1.6\times 10^{-4}$, $K_h=0.94$$\mathrm{m}^2\,{\cdot}\, \mathrm{s}^{-1}$; and constant nighttime values about 8 times lower: $C_h=5\times 10^{-4}$, $K_s=2\times 10^{-5}$, $K_h=0.08$$\mathrm{m}^2\cdot\mathrm{s}^{-1}$. Values are tuned to recover the heat fluxes given by the CNRM-AM simulation.