Volume 119, Issue 10 p. 5845-5863
Research Article
Free Access

Aerosols in the convective boundary layer: Shortwave radiation effects on the coupled land-atmosphere system

Eduardo Barbaro

Eduardo Barbaro

Meteorology and Air Quality Section, Wageningen University, Wageningen, Netherlands

Search for more papers by this author
Jordi Vilà-Guerau de Arellano

Jordi Vilà-Guerau de Arellano

Meteorology and Air Quality Section, Wageningen University, Wageningen, Netherlands

Search for more papers by this author
Huug G. Ouwersloot

Huug G. Ouwersloot

Meteorology and Air Quality Section, Wageningen University, Wageningen, Netherlands

Max Planck Institute for Chemistry, Mainz, Germany

Search for more papers by this author
Joel S. Schröter

Joel S. Schröter

Meteorology and Air Quality Section, Wageningen University, Wageningen, Netherlands

Search for more papers by this author
David P. Donovan

David P. Donovan

Royal Netherlands Meteorological Institute, De Bilt, Netherlands

Search for more papers by this author
Maarten C. Krol

Maarten C. Krol

Meteorology and Air Quality Section, Wageningen University, Wageningen, Netherlands

Institute for Marine and Atmospheric Research Utrecht, Utrecht, Netherlands

Search for more papers by this author
First published: 06 May 2014
Citations: 43
Correspondence to:

Correspondence to: E. Barbaro,

[email protected]

Abstract

By combining observations and numerical simulations, we investigated the responses of the surface energy budget and the convective boundary layer (CBL) dynamics to the presence of aerosols. A detailed data set containing (thermo)dynamic observations at CESAR (Cabauw Experimental Site for Atmospheric Research) and aerosol information from the European Integrated Project on Aerosol, Cloud, Climate, and Air Quality Interactions was employed to design numerical experiments reproducing two typical clear-sky days, each characterized by contrasting thermodynamic initial profiles: (i) residual layer above a strong surface inversion and (ii) well-mixed CBL connected to the free troposphere by a capping inversion, without the residual layer in between. A large-eddy simulation (LES) model and a mixed-layer (MXL) model, coupled to a broadband radiative transfer code and a land surface model, were used to study the impacts of aerosols on shortwave radiation. Both the LES model and the MXL model results reproduced satisfactorily the observations for both days. A sensitivity analysis on a wide range of aerosol properties was conducted. Our results showed that higher loads of aerosols decreased irradiance imposing an energy restriction at the surface, delaying the morning onset of the CBL and advancing its afternoon collapse. Moderately to strongly absorbing aerosols increased the heating rate contributing positively to increase the afternoon CBL height and potential temperature and to decrease Bowen ratio. In contrast, scattering aerosols were associated with smaller heating rates and cooler and shallower CBLs. Our findings advocate the need for accounting for the aerosol influence in analyzing surface and CBL dynamics.

Key Points

  • Over grassland, aerosols reduce the sensible heat more than the latent heat flux
  • Scattering aerosols cool and absorbing aerosols warm the boundary layer
  • Absorbing aerosols increase the afternoon convective boundary layer height

1 Introduction

Tropospheric aerosols influence the Earth's climate by absorbing and scattering shortwave (SW) radiation [Boucher and Anderson, 1995; Stier et al., 2007; Paasonen et al., 2013]. Depending on the aerosol characteristics [Liu and Ou, 1990; Wang et al., 2009], chemical composition [Yu and Zhang, 2011], and vertical distribution [Knapp et al., 2002; Raut and Chazette, 2008; Sakaeda et al., 2011], their effects on the surface radiation/energy budgets range from negligible to very significant [see also Kaufman et al., 2002; Matthias and Bosenberg, 2002; Forster et al., 2007; Arneth et al., 2010; Costabile et al., 2013]. More specifically, aerosols modify the convective boundary layer (CBL) dynamics changing its depth, vertical structure, and entrainment zone characteristics [Ackerman, 1977; Venkatram and Viskanta, 1977; Zhang et al., 2010; Barbaro et al., 2013].

In this paper we study observations of typical midlatitude CBLs formed over grassland characterized by different mechanisms governing their growth: (i) breakup of a strong surface inversion and (ii) well-mixed CBL growth due to entrainment. We emphasize the importance of the vertical distribution of aerosol extinction (optical depth ∼0.2 and single scattering albedo ∼0.9) on the CBL growth, the engulfing of the aerosol layer during the morning transition, and the subsequent influence on the (i) surface energy budget (SEB) and (ii) thermodynamic state diurnal variation.

Very few data sets combine aerosol measurements, radiation observations, surface, and detailed CBL vertical structure data. Therefore, these processes have been normally investigated separately or by means of simplified CBL numerical models [Venkatram and Viskanta, 1977; Tunved et al., 2010], regional models (see Baklanov et al. [2013] for a complete review), mesoscale models [Zhang et al., 2010; Wong et al., 2012], and climate models combined with satellite observations and aerosol parameterizations [Haywood et al., 1999; Chen et al., 2013].

Previous studies have shown that very broad ranges of aerosol optical depth (expressed as τ) and single scattering albedo (expressed as ω) are found in the atmosphere. Kaufman [1993] summarizes several measurements of τ (at 550 nm) varying from nonpolluted cases (τ≈0) up to very polluted situations (τ≈1.5) for different areas around the globe. Zubler et al. [2011a] present diurnal means of τ for 13 AERONET stations (all at 550 nm) varying on average between 0.14 and 0.29. Hewitt and Jackson [2009] suggest τ ranging from 0.2 up to 0.8 for polluted continental conditions. For the northern India region, Tripathi et al. [2005] obtained averaged τ values (at 500 nm) of around 0.77 ± 0.29. Israelevich et al. [2012] found that heavy loads of Saharan dust are transported over the Mediterranean Sea with τ values up to 0.6 (at 550 nm). For the single scattering albedo, Jacobson [2001] estimated an annually global averaged at ω≈0.85 (at 550 nm) over land. Lyamani et al. [2010] found ω as low as 0.65 ± 0.07(at 670 nm) for Granada, an urban location in Spain. Tripathi et al. [2005] obtained ω≈0.76(at 500 nm). Takemura et al. [2002] simulated global distributions of ω (at 550 nm) varying between 0.8 and 1.0.

With respect to observations, by isolating the aerosol impacts on the SW radiation, Hatzianastassiou et al. [2005] and Wang et al. [2009] have shown that detailed aerosol temporal evolution and vertical profiles are needed to reproduce the observed direct/diffuse partitioning and global irradiance [see also Meywerk and Ramanathan 1999; Singh et al., 2010]. The available observational campaigns studying the aerosol impacts on the troposphere [e.g., de Wekker et al., 2004; Johnson et al., 2008] focused mainly on instrument comparisons and data validation without discussing in detail how the aerosols influence the boundary-layer vertical structure and the CBL heat budget. At the same time, experiments aiming to understand (urban) boundary-layer turbulent processes [Angevine et al., 1998; Masson et al., 2008] do not further explore how the aerosol SW radiation absorption and scattering affect the SEB or the redistribution of heat throughout the CBL.

Regarding numerical modeling, Li et al. [1997] and Zhang et al. [2010] mentioned that the lack of observational data leads to major uncertainties in the aerosol representation. They argued that numerical studies need more observations to improve our understanding of the complex interaction of aerosols and SW radiation, the land-atmosphere system, and the CBL vertical structure. None of the studies mentioned above have merged detailed aerosol information to a three-dimensional high-resolution model able to reproduce detailed CBL dynamics. Here extending previous studies [e.g., Venkatram and Viskanta, 1977; Yu et al., 2002], we systematically investigate and quantify how the surface fluxes respond to the aerosol absorption and scattering of SW radiation. By doing so, we also consider the subsequent impact of the aerosols on the diurnal evolution of the CBL dynamics. Our approach integrates at diurnal scales radiation-surface-atmosphere using a large-eddy simulation (LES) model coupled to a land-atmosphere model and a radiative transfer code. We also design a parameter space to study typical τ and ω conditions aiming to further quantify the aerosol effects on the land-atmosphere system.

The paper is organized as follows: Section 2 describes the numerical models, the available observations, and the experimental design. The validation of the models is discussed in section 3. In section 4 we present a sensitivity analysis varying the amount and the aerosol characteristics. The main results are summarized in section 5.

2 Description of the Numerical Models

We use the Dutch Atmospheric LES (DALES version 3.2, see Heus et al. [2010] for a full description of the model) to explicitly simulate the most energetic dynamical structures of the CBL and parameterize the scales smaller than the adopted numerical grid size [Moeng, 1984]. The SW radiation is represented by means of the broadband Delta-Eddington (DE) approximation [Joseph and Wiscombe, 1976]. To connect the radiation budget to the surface energy budget, we use the land surface model described in van Heerwaarden et al. [2010, and references within].

Concerning the radiation calculations, in this work we effectively divide the atmosphere into an aerosol-free Rayleigh layer which sits above an aerosol layer subdivided according to the DALES vertical spacial resolution (see Table 1). In order to calculate the SW radiation fluxes within the aerosol layer, the two-stream DE approximation fulfills the purposes of this paper while providing rapid and precise calculations of the direct and diffuse components of the SW radiation [Shettle and Weinman, 1970; Joseph and Wiscombe, 1976; Liou, 2002]. To calculate the aerosol SW radiation absorption and scattering, the method requires the temporal evolution of the τ, ω, and the asymmetry parameter (g) [see Liou, 2002]. The Rayleigh scattering above the aerosol layer is calculated based on the Elterman [1968] standard profile. The atmospheric net transmissivity from the top of the atmosphere to up to the top of the aerosol layer is parameterized following Burridge and Gadd [1974 [see also Stull, 1988] and provides the total downward SW radiation used as an upper boundary condition for the aerosol layer DE radiative transfer calculations. Here we neglect the contribution of gaseous SW radiation heating and longwave (LW) radiation cooling to the CBL total heating rate (HR). As a result, in our simulations the HR is entirely due to the effect of the aerosols. According to Angevine et al. [1998] and Stull [1988] and further corroborated by radiative transfer simulations using the libRadtran code [Mayer and Kylling, 2005] the heating due to gaseous SW radiation absorption and the LW cooling nearly offset each other from 9.5 UTC to 15.5 UTC for midlatitudes (not shown). In contrast, for the morning and afternoon transitions the LW radiation cooling is larger than the SW radiation gaseous absorption. For the cases discussed here the differences remain smaller than −0.7 K d−1. Note that a full radiation treatment might yield smaller values of HR both early in the morning and late in the afternoon. However, we expect that this bias has a small influence on our findings for the convective period since the neglect of SW gaseous absorption and LW cooling is consistent between the numerical experiments independent of the amount of aerosols. Moreover, early in the morning the divergence of the turbulent heat flux is the dominant term in the potential temperature budget equation (see section 4). As a result, we expect that our approach produces a proper representation of the effect of the aerosol HR during the convective period. Here we focus only on the convective period; therefore, the impact of aerosols and the HR imbalance on the stable boundary layer has not been analyzed.

Table 1. LES Initial and Prescribed Values for CONTROL, AERO+, and CLEAR Experiments for Both CESAR2008 and CESAR2003 Data Setsa
Boundary Layer Properties CESAR2008 CESAR2003
Initial boundary layer height, zi (m) 150 114
Initial residual layer height (m) 1700 600
Initial aerosol layer height, ha (m) 1700 600
Geostrophic wind (Ug, Vg) (m s−1) (8,0) (6,3)
Large-scale vertical velocity (ws) (m s−1) 0.0 0.0
Spatial domain (x,y,z) (m) (12800,12800,3000) (12800,12800,3000)
Spatial resolution (dx,dy,dz) (m) (50,50,15) (50,50,12)
Integration total time (h) 12 12
Heat, Moisture and Extinction Vertical Profiles θ(K) q(g kg−1) βa (km−1)
CESAR2008
z< 250 m 286.0 + 0.032z 7.1 − 0.01z 0.11
250 <z< 800 m 294.0 4.5 0.11
800 <z< 1700 m 294 + 0.0006(z − 800) 4.5 − 0.0016(z − 800) 0.11
1700 <z< 1950 m 294.5 + 0.012(z − 1700) 3.1 − 0.0093(z − 1700) 0.00
z> 1950 m 298.1 + 0.0045(z − 1950) 0.3 0.0
CESAR2003
z< 114 m 284.0 4.3 0.30
114 <z< 138 m 284.0 + 0.167(z − 114.0) 4.3 − 0.033(z − 114.0) 0.30
138 <z< 600 m 288.0 + 0.0036(z − 138.0) 3.5 − 0.0012(z − 138.0) 0.30
z> 600 m 288.0 + 0.0036(z − 138.0) 3.5 − 0.0012(z − 138.0) 0.00
Land Surface Properties CESAR2008 CESAR2003
Volumetric water content (m3 m−3) 0.43 0.39
Saturated volumetric water content (m3 m−3) 0.60 0.60
Volumetric water content field capacity (m3 m−3) 0.491 0.491
Volumetric water content wilting point (m3 m−3) 0.314 0.314
Vegetation fraction (-) 0.9 0.9
Temperature top soil layer (K) 286.5 282.0
Temperature deeper soil layer (K) 289.5 285.0
Leaf area index (-) 2 2
Minimum resistance transpiration (-) 110 110
Minimum resistance soil evaporation (-) 50 50
Surface albedo (-) 0.25 0.25
Leaf area index (-) 2 2
Aerosol Properties (555 nm)
Single scattering albedo, ω (-), t < 13.62 h 0.925 + 0.055 cos(31.5(t + 20))
Single scattering albedo, ω (-), t > 13.62 h 0.975 − 0.06(1− exp(15.0 − 1.1t))
Asymmetry parameter, g (-) 0.645 + 0.025 cos(22.5(t + 27))
Aerosol optical depth, τ (-)
CONTROL 0.185 + 0.062 cos(24.75(t + 12))
AERO+ 0.555 + 0.186 cos(24.75(t + 12))
CLEAR 0.0
  • a The UTC time (t) is given in hours.

The vertical HR profile is calculated directly by taking the divergence of net radiation, which is obtained from the SW calculation at a single wavelength. We select the wavelength of 555 nm as representative for the aerosol properties, and the ground is assumed to be a Lambertian surface [Madronich, 1987] with an albedo of 0.25 [Beljaars and Bosveld, 1997; van Heerwaarden et al., 2010]. The downward longwave (LW) radiation is prescribed by the clear-sky empirical formula proposed by Brunt [1932], and the upward LW radiation assumes the surface emissivity taken as unity [Holtslag and de Bruin, 1988].

Concerning the land surface model, the surface resistance is calculated by means of the Jarvis-Stewart model [Jarvis, 1976], the energy budget and the soil temperature equations are based on Duynkerke [1991], and the soil moisture equations are based on Noilhan and Planton [1989]. We couple the radiation model and the surface model to our LES to account for the effects of aerosols on SW radiation and on the land-atmosphere system.

To further quantify the aerosol impact on the land-atmosphere system, we use a zeroth-order mixed-layer (MXL) model [Lilly, 1968; Garratt, 1992] to perform a series of systematic simulations exploring a wide range of ω and τ. To this end, the same radiation and land surface models used in our LES are implemented in the MXL model [Heus et al., 2010; van Heerwaarden et al., 2010]. The MXL model used here is an extension of that described by van Heerwaarden et al. [2010] and Barbaro et al. [2013], with the inclusion of the DE model for the SW radiation calculations. The simplified CBL dynamics in the MXL model assumes a well-mixed CBL with constant vertical profiles of the state variables. In the MXL model, the entrainment zone is represented by an infinitesimally thin inversion layer (zeroth-order approach [see Lilly, 1968; Betts, 1974]) and the entrainment ratio is defined as being constant and held equal to 0.2 [Stull, 1988].

2.1 Observational Data Set

The numerical experiments are based on observations taken at CESAR (Cabauw Experimental Site for Atmospheric Research). CESAR (http://www.cesar-observatory.nl) is located in a flat terrain covered with grass in the Netherlands (51.971°N, 4.927°E). Observations of radiation, surface fluxes, and thermodynamic variables along the 213 m mast (2, 10, 20, 40, 80, 140, and 200 m) are measured at a 10 Hz frequency [Beljaars and Bosveld, 1997]. From 2006 on, CESAR became part of the Baseline Surface Radiation Network and diffuse and direct shortwave radiation components were integrated to the radiation data set [Knap et al., 2010]. The prescribed CESAR land surface information is based on the Cabauw climatological values described in van Heerwaarden et al. [2010].

In May 2008, the intensive observational campaign called European Integrated Project on Aerosol, Cloud, Climate, and Air Quality Interactions (IMPACT/EUCAARI) took place in Cabauw (see Kulmala et al. [2009, 2011] for details). During the campaign, several additional meteorological observations were collected, including vertical profiles from radiosondes, detailed aerosol information, and remote sensing data. In view of this unique data set, 8 May 2008 (henceforth called CESAR2008) is chosen as a control case because of its cloudless conditions, constant weak to moderate easterly winds ranging from 4 to 6 m s−1, and the absence of large-scale heat advection. The chemistry and radiation of the CESAR2008 have already been described in detail (see Derksen et al. [2011], Mensah et al. [2012], Aan de Brugh et al. [2013], and van Beelen et al. [2013] for atmospheric chemistry and Wang et al. [2009] for clear-sky radiative closure). The CESAR2008 synoptic conditions and the pollution situation over Cabauw were summarized by Hamburger et al. [2011] and Grob et al. [2013]. Both the aerosol optical properties and the loadings represent typical continental aerosols transported from central Europe (Poland and Germany) over the Netherlands due to the easterly circulation associated with the persistent anticyclone located over Denmark, Northern Germany, and the Benelux states. The synoptics conditions observed around the CESAR2008 campaign—Scandinavian High, Ridge Central Europe, and high over Central Europe—are commonly observed in Europe during the month of May [James, 2006].

As we will show, CESAR2008 case is characterized by a residual layer (RL) above a surface inversion. The RL well-mixed vertical structure allows a rapid CBL growth after the breakup of the morning potential temperature inversion. To complement our analysis, we study another convective day with a well-mixed boundary layer at CESAR (25 September 2003—henceforth called CESAR2003). This day is characterized by negligible large-scale heat advection, constant moderate winds (4–7 m s−1), very few clouds, and the absence of a residual layer. For a complete description of the CBL (thermo)dynamics and the synoptics, see Casso-Torralba et al. [2008]. The synoptic situation observed for CESAR2003—anticyclone associated with southeasterly winds—is similar to CESAR2008. The CESAR2003 dynamics and land surface properties have been systematically studied. van Heerwaarden et al. [2010] described the feedback and forcings of the CBL dynamics and land surface on the time evolution of evapotranspiration. Pino et al. [2012] quantified the relation between CBL dynamics and CO2 budget and surface exchanges. Since the aerosol observations were only available during the IMPACT/EUCAARI campaign, τ, ω,and g values are assumed to be representative of a typical day in the Netherlands (urn:x-wiley:jgrd:media:jgrd51411:jgrd51411-math-0001g m−3 [see Schaap et al., 2009]), resulting in the same aerosol properties for both numerical experiments. To ensure that the aerosol layer is engulfed by the CBL, and in absence of any observational evidence, we assume an idealized initial aerosol layer height (ha) of 600 m. By using this value we obtain the closest match between the numerical experiments and the CESAR observations of (thermo)dynamic variables. Variations around ha have very little impact on the CBL (thermo)dynamics evolution, however.

2.2 Experimental Design

We present in Figure 1 the temporal evolutions of τ, ω, and gfor the CESAR2008 case. The aerosol observations described in Figure 1 are prescribed in our MXL and LES simulations by means of fitting functions (red dashes). The initial vertical profile of the aerosol extinction coefficient is prescribed, as shown in Figure 2c. The scalar spatial evolution is solved simultaneously with the other thermodynamic variables in the LES (three-dimensional) and MXL (bulk) models for every time step. By constraining the aerosol properties we aim for an accurate simulation of the SW radiation, with a special focus on the reproducibility of its direct and diffuse components.

Details are in the caption following the image
Aerosol (a) optical depth, (b) single scattering albedo, and (c) asymmetry parameter (all at 555 nm) observed diurnal evolution. The dashes in Figures 1a–1c represent the prescribed values in the LES and MXL models.
Details are in the caption following the image
Vertical profiles of (a) potential temperature, (b) specific humidity, and (c) aerosol extinction coefficient. The continuous black lines are radiosondes shown in Figures 2a and 2b and lidar measurements shown in Figure 2c. The blue dashed lines shown in Figures 2a and 2b are the CESAR mast observations. The red continuous lines represent the initial profiles at 06 UTC for the LES model (thick) and for the MXL model (thin).

In Figure 2 we present the mast data (06 UTC) and radiosonde vertical profile (10 UTC) for (Figure 2a) potential temperature and (Figure 2b) specific humidity, and (Figure 2c) the aerosol extinction coefficient (βa) obtained from a lidar profile. The combination of the radiosondes (representing the mixed layer) and the mast data (surface layer) is used to prescribe the initial conditions for our LES and MXL experiments (red vertical profiles). Note that we adapt the MXL model to also take into account the vertical extension of the layer above the developing CBL (residual layer). As long as the RL is present, air is entrained into the boundary layer from the RL with an entrainment ratio held constant and equal to 0.2. When the mixed-layer and the residual layer connect, the lower ground inversion disappears and the mixed-layer merges with the RL. At that moment, entrainment between the mixed-layer and the free troposphere becomes active, also with an entrainment ratio of 0.2.

The 06 UTC initial profiles of θ and q in the surface layer are based on mast observations (up to 200 m) in order to reproduce the thermodynamics of the morning potential temperature inversion jump. For potential temperature, at 200 m the radiosonde and the mast data coincide and the radiosonde data are taken as initial profile for the well-mixed layer and the free atmosphere. For specific humidity we note a discrepancy between radiosonde and mast data of about 1 g kg −1 at 200 m height. We follow the mast data closer to the surface and above 800 m; the radiosonde is considered to be representative for the initial condition in the upper atmosphere.

The lidar profile (06 UTC) is measured using a backscatter lidar operating at 355 nm. The lidar data were inverted using the Klett-Fernald method [Klett, 1985] assuming a value of the lidar backscatter-to-extinction ratio (S) of 50(±20)sr−1. The uncertainty in Scan result in significant uncertainty in the resulting aerosol extinction profile. However, here the lidar profile is simply scaled to the corresponding 555 nm τ value and the resulting extinction profile is then used as a qualitative indication of the aerosol extinction vertical profile. The lidar profiles, however, are not trustworthy below 500 m because the transmitted beam and the receiver telescope field of view do not fully overlap [Biavati et al., 2011]. We therefore assume a well-mixed aerosol extinction profile below 500 m down to the surface.

As observed in Figure 2, CESAR2008 is characterized by a residual layer extending from 200 m to around 1700 m—also containing the aerosol layer—see Figure 2c. This boundary layer prototype enables us to study the role of aerosols in delaying/advancing the breakup of the inversion layer and subsequent onset of the morning CBL. The main difference between the CESAR2008 and CESAR2003 cases is the absence of a residual layer in the latter. Therefore, the CBL growth during the morning transition in CESAR2008 is driven by the rapid incorporation of the well-mixed RL, while in CESAR2003 the CBL growth is more continuous and is driven by entrainment and surface heat fluxes. To study the impact of the aerosols on the CBL development, we design two extra LES numerical experiments (called CLEAR and AERO+) for each CESAR control simulation. The experiments “CLEAR” and “AERO+” simulate the same boundary layer and surface properties as in CESAR2003/CESAR2008. The difference lies only in the τ: for the “CLEAR” simulations τ is set to zero and for the “AERO+” simulations the τis tripled compared to the control cases. The AERO+ loadings are therefore characteristic of a moderately polluted urban area [Tripathi et al., 2005; Hewitt and Jackson, 2009], Saharan dust [Zubler et al., 2011b; Israelevich et al., 2012; Kinne et al., 2013], or biomass burning [Myhre et al., 2003].

The initial and boundary conditions of the LES experiments are described in Table 1. Note that we keep the design of the MXL model experiments as close as possible to the LES.

3 Model Validation

We start by validating the MXL and LES numerical experiments against the observations (CESAR2008 and CESAR2003). In Figure 3 we show the temporal evolution of the simulated and observed radiation/energy budget, boundary-layer height, potential temperature, and specific humidity for CESAR2008. We present in Figure 4 the results obtained for the CESAR2003 case. Note that we use the same radiation transfer model and the land surface parameterization in the MXL and LES models. Therefore, the validation of the radiation/energy budget of the MXL is omitted. The sensible (SH) and latent (LE) heat fluxes are directly measured by means of the eddy-correlation technique. The surface energy budget imbalance is proportionally distributed over SH and LE according to the Bowen ratio (urn:x-wiley:jgrd:media:jgrd51411:jgrd51411-math-0002) (see Beljaars and Bosveld [1997], Twine et al. [2000], and Foken [2008] for further details). The CBL height is retrieved from the wind profiler measurements by the modified maximum signal-to-noise ratio detection algorithm described by Bianco and Wilczak [2002].

Details are in the caption following the image
LES and MXL validation against the CESAR2008 data set. (a) Temporal evolution of the radiation budget. The dots represent the LES data, and the continuous lines represent observations. The dark gray line is the direct SW radiation, and the light gray line is the diffuse SW radiation. SWD and LWD (solid lines) are respectively the downward components of the observed SW irradiance (black) and LW radiation (red)—analog for the upward components (SWU and LWU—dashed lines). All the downward components include error bars, although these are too small to be depicted. (b) Global (left Y axis) and diffuse (right Y axis) SW radiation scatterplot—the thin black line is a 1:1 reference line. (c) Temporal evolution of surface energy budget. SH, LE, and G0 are respectively the sensible, latent, and soil heat fluxes; QNET=(SWD + LWD + SWU + LWU) is the net radiation. Temporal evolution of (d) boundary-layer height, (e) potential temperature, and (f) specific humidity. The dashes in Figures 3d–3f indicate the MXL model results—omitted for Figures 3a–3c; see text.
Details are in the caption following the image
Same as Figure 3 but for CESAR2003 case. Note in Figures 4a and 4c that we have extended the x axis to show the complete diurnal cycle of SW radiation and QNET.

We see in Figure 3a that the DE radiative transfer calculations constrained by aerosol observations (Figure 1) accurately reproduce the SW radiation components for this clear-sky day. The LW radiation components are also well captured. It is observed in Figure 3b that we are able to reproduce the global SW radiation (SWD), with a small mean bias error and root-mean-square error (MBE =2.1 W m−2 and RMSE =8.4W m−2, respectively). Also, the diffuse SW observations are well reproduced (MBE =7.0 W m−2 and RMSE =9.7W m−2). Our results are comparable to the more advanced radiation modeling study of Wang et al. [2009]— using a spectrally resolved doubling adding radiation transfer model—for the same day. We show in Figure 3c that also the modeled net radiation agrees with the observations. The prescribed land surface properties (e.g., soil moisture, temperature, and field capacity) are based on van Heerwaarden et al. [2010] and observations taken at CESAR. Combined with the radiation terms, this lead to an accurate reproduction of the surface fluxes (SH, LE, and the soil heat flux).

A proper determination of the surface fluxes is crucial for an accurate simulation of the CBL height evolution, Figure 3d. Both the LES and the MXL models are able to capture the breakup of the potential temperature inversion and the subsequent engulfing of the RL. The timing of this breakup differs only 10–20 min from the wind profiler observations. The RL from the previous day is then incorporated in the mixed-layer, and the boundary layer grows further to around 1600–1700 m. After 11 UTC the sensible heat flux (∼110 W m−2) diminishes, reaching negative values already around 15.5 UTC. Therefore, the boundary layer grows fairly little after 11 UTC (∼400 m in 6 h). The 7 K temperature increase at 10 m from 06 UTC till 10 UTC is well reproduced, Figure 3e, as well as the time at which 10 m and 200 m θ observations converge (09 UTC) indicating the formation of a well-mixed CBL. Finally, the specific humidity evolution, Figure 3f, compares well with the observations, capturing the entrainment of drier air from the layers above (mainly from 8.5 UTC to 10.5 UTC). Next, in Figure 4 we present the same validation for the CESAR2003 case.

Although we do not have information on the aerosol optical properties, the comparison with observations is also satisfactory for all the variables. Again, the radiative transfer calculations reproduce the observed SW radiation fluxes, except the downward SW at the middle of the day when scattered clouds were observed [Pino et al., 2012]. Our hypothesis is that these clouds are located in the upper troposphere, therefore not affecting the downward LW radiation measured at the surface, since most of the LW emission (90%) comes from the lowest 1000 m of the atmosphere [Dürr and Philipona, 2004]. We find in Figure 4b that the drop in the observed SWD does not affect the good fit between LES and the observations but considerably enlarges the MBE (8.1 W m−2) and RMSE (32.3 W m−2) if compared to the CESAR2008 values. Notwithstanding, both MBE and RMSE decrease by a factor of 3 by simply excluding the SWD observations during the cloudy period. In Figure 4c the modeled surface fluxes match the observations, showing, in Figure 4d, a continuously growing CBL. This indicates that the prescribed initial conditions are appropriate [see Casso-Torralba et al., 2008]. The potential temperature, Figure 4e, and the specific humidity, Figure 4f, evolutions also show that we are able—with the LES and the MXL models—to capture the physics that drive the CBL evolution. A similar good agreement between the LES and the MXL models is found by Pino et al. [2006], albeit without the coupling to a land surface model.

4 Aerosol Effects on the Surface Fluxes and Aerosol Heating Rate

In this section, we aim to further analyze the impact of aerosol scattering/absorption on the surface SW radiation and on the CBL vertical structure. To this end, we compare the CESAR2008 and CESAR2003 control experiments to the CLEAR (τ = 0) and AERO+ (tripled τCONTROL) simulations—see Table 1. Since our findings are similar for both cases under study we focus only on the results for CESAR2008.

We first quantify the aerosol impact on (i) the net SW radiation at the surface and above the CBL and (ii) the SEB. In Figure 5 we show the temporal evolutions of (a) the net SW radiation differences compared to the CONTROL experiment at 2500 m (always above the CBL) and at the surface, (b) the surface flux ratios relative to the CONTROL experiment, and (c) Bowen ratio.

Details are in the caption following the image
Temporal evolution of the (a) net SW radiation difference compared to CONTROL at 2500 m (dashed line) and at the surface (continuous lines), (b) SH and LE ratios referent to the CONTROL experiment, and (c) Bowen ratio. The thin black line in Figures 5a and 5b represents the reference case. Note that in Figure 5a we have extended the x axis to 16 UTC in order to show the maximum/minimum differences in the SW net radiation compared to CONTROL. The results are shown only for CESAR2008.

As shown in Figure 5a, aerosols directly reduce the net irradiance at the surface. The differences between the AERO+ and CONTROL experiments range from −43 W m−2 (8 UTC) down to −69 W m−2(15 UTC) less available energy at the surface. In turn, an increase of up to 46 W m−2 (at 16 UTC) is observed for the CLEAR experiment compared to CONTROL. These results are in agreement with the findings of Tripathi et al. [2005] who show diurnal averages ranging from −31 W m−2 to −98 W m−2 less available energy for moderately to highly polluted industrial cities in India. Haywood et al. [2001] obtained a clear-sky direct solar radiative effect as strong as −60 W m−2 for Saharan dust. Myhre et al. [2003] show values roughly between −50 W m−2 and −90 W m−2 for biomass burning aerosol in Southern Africa. At the top of the CBL we also observe less (more) available energy for AERO+ (CLEAR) experiments. The differences, however, are significantly smaller than at the surface because scattering aerosols increase the outgoing SW radiation at the top of the CBL due to backscattering. The temporal variation of ΔSWNET is explained by the diurnal cycle of the irradiance and by the temporal evolution of the aerosol properties (see Figure 1c).

The impact of the aerosols on the SEB is presented in Figure 5b. For AERO+ (the opposite response is found for CLEAR), the aerosol SW radiation attenuation throughout the aerosol layer causes a relatively small reduction of LE, ranging from −20% up to −10% during the day. The ratio urn:x-wiley:jgrd:media:jgrd51411:jgrd51411-math-0003 is also reduced but, in contrast, shows a more pronounced diurnal variation. The physical explanation is as follows: under the studied surface conditions, i.e., well-watered grassland, LE uses the available surface energy more efficiently compared to SH [see Gentine et al., 2011; Bateni and Entekhabi, 2012]. Since the reduction in LE is much less pronounced than in SH, the evaporation (not shown) for AERO+ compared with CLEAR decreases relatively little (5–15%). From 10.5 UTC on, the attenuation of SW radiation by aerosols leads to a stronger reduction of the irradiance, diminishing the available energy at the surface and therefore strongly decreasing SH. As a result, aerosols diminish β (Figure 5c) by about 50%more for AERO+ than for CLEAR in the afternoon.

We show in Figures 6a and 6b the 10 min averaged CBL heat budget components as a function of height for 09 UTC and 13.5 UTC. Similar to Angevine et al. [1998] and Barbaro et al. [2013] we calculate the two contributions of the heat budget: (i) potential temperature turbulent vertical flux divergence (-urn:x-wiley:jgrd:media:jgrd51411:jgrd51411-math-0004, TDθ) and (ii) aerosol heating rate (urn:x-wiley:jgrd:media:jgrd51411:jgrd51411-math-0005, HR). To complete the CBL vertical structure analysis we also show the potential temperature vertical profile at 09 UTC and 13.5 UTC (Figure 6c).

Details are in the caption following the image
Vertical profiles of (a) potential temperature turbulent vertical flux divergence (TDθ), (b) aerosol HR, and (c) potential temperature. To improve visualization, in Figure 6a the max/min of the turbulent term at 9 UTC is indicated in the figure. In all panels, the dashed lines (lighter tones) represent 9 UTC and the continuous lines (darker tones) represent 13.5 UTC. The results are only shown for CESAR2008.

As we observe in Figure 6a, the turbulent sensible heat flux divergence is the main contributor to the very rapid increase of the potential temperature (dilution of the 8 K inversion layer jump) during the morning period, i.e., TDθ>> HR. Values up to 45 K d−1 close to the surface are observed for this case at 9 UTC. During the afternoon, the HR (Figure 6b) becomes as relevant as TDθ contributing 15%and 49% to the CBL heat budget (at 500 m) for the CONTROL and AERO+ experiments, respectively. We show in Figure 6b that increasing τ(AERO+) leads to an approximately linear increase of the absorption of SW radiation [Sakaeda et al., 2011]. Note that despite the well-mixed aerosol vertical distribution, the HR induced by the absorbing aerosols is larger at the top of the CBL. The reason is the higher SW downward radiation flux at higher levels, especially for larger optical depths (AERO+). Therefore, the temporal evolution of the HR vertical profile is mainly driven by (i) the CBL dynamics, (ii) the diurnal cycle of irradiance, and (iii) changes in both the aerosol properties and aerosol depth [see also Penner et al., 1994]. Note that for the HR, the decrease in solar irradiance at larger zenith angles is compensated by the increase in path length of the light through the atmosphere.

In analyzing the potential temperature vertical profiles, we find for the AERO+ experiment that during the morning the CBL is colder and the residual layer is warmer (−0.6 K and +0.2 K, respectively) in comparison to the CONTROL experiment (dotted lines in Figure 6c). The explanation is that during the morning ha>zi (see Figure 7b). In this situation, aerosols are also present in the residual layer, reducing the amount of SW radiation at the surface (see Figure 5a) and heating the residual layer (see Figure 6b). Later, the aerosol layer becomes part of the CBL, slightly increasing the CBL potential temperature by about 0.25 K at 13.5 UTC, (continuous lines in Figure 6c).

Details are in the caption following the image
Temporal evolution of the boundary layer depth for (a) CESAR2008 and (b) CESAR2003 cases. The dotted lines represent the depth of the aerosol layer in the LES model.

4.1 Impact of the Aerosols on the CBL Depth Development

In this section we quantify the aerosol impact on the breakup of the morning inversion layer and on the CBL collapse in the afternoon. The overall aerosol effect on the temporal evolution of the CBL height (zi) for weakly absorbing aerosols (ω > 0.9) is summarized in Figure 7. During the morning, aerosols tend to delay the breakup of the ground inversion layer (∼40 min for AERO+). The final CBL height for the CESAR2008 case is less influenced by the aerosols because of the well-mixed residual layer (see Figure 2). For the CESAR2003 experiments we find that part of the aerosols are also located above the CBL till around noon. In this configuration aerosols reduce QNET and therefore the surface fluxes (see Figure 5) without heating the shallow CBL compared to the residual layer. This effect explains the significantly shallower CBL for AERO+ (and CONTROL) during the morning. In turn, during the afternoon aerosols act as a second source of energy, warming the CBL in comparison to the free troposphere. This effect counterbalances the decrease in the surface available energy. Warming the CBL leads to a weakening of Δθ and a subsequent faster CBL growth. As a result, the CBL for CONTROL and AERO+ reach a similar depth as in CLEAR experiment due to the different contributions in the heat budget.

Our findings indicate a series of connected effects of absorbing aerosols impacting the afternoon CBL depths: (i) the decrease of turbulent surface fluxes and entrainment (negative effect), (ii) weakening of Δθ by increasing the CBL potential temperature (positive effect), and (iii) development of a deeper and less strongly stable inversion layer in the lower part of the entrainment zone (negative effect) due to the relatively stronger heating at the top of the CBL. The latter process leads to an earlier stabilization of the potential temperature, deepening the entrainment zone, and thereby increasing the resistance for the penetration of the eddies [Ackerman, 1977; Barbaro et al., 2013]. Aiming to further quantify the impact of upper CBL heating on the turbulent field we calculate the CBL anisotropy at 500 m, quantified by urn:x-wiley:jgrd:media:jgrd51411:jgrd51411-math-0006, for 13.5 UTC. By doing so, we quantify the suppression of the upward movements in the CBL. We find 0.69, 0.66, and 0.60 for CLEAR, CONTROL, and AERO+, respectively, indicating a greater conversion of vertical into horizontal motions in the AERO+ case. This shows that the turbulent kinetic energy diminishes and the turbulent structures find more resistance to develop if abundant aerosols heat the CBL.

The net effect of the aerosols, in these particular cases, is to slightly warm the afternoon CBL (up to 0.25K if compared to CONTROL, Figure 6c) and to reduce the surface fluxes (down to 55%if compared to CONTROL, Figure 5b), which leads to a slightly shallower CBL (down to 80 m if compared to CONTROL, Figure 7). These results agree with Yu et al. [2002] and Wong et al. [2012] with respect to the net effect of scattering aerosols to slightly decrease the afternoon CBL depth. The reduced surface fluxes decrease the entrainment of dry air from the free troposphere. These effects combined result in negligible changes in the specific humidity (not shown).

4.2 Sensitivity Analysis to Aerosol Optical Properties

In the previous LES numerical experiments we studied only weakly absorbing aerosols, based on the CESAR2008 observations (ω > 0.9). To complete our research, we employ the MXL model to extend the analysis to a wider range of aerosol conditions (τ and ω) by performing 336 experiments varying systematically τ(0.0 to 1.0) and ω(0.7 to 1.0) for both CESAR2008 (Figure 8) and CESAR2003 (Figure 9) cases. We base the choice of the τ and ω ranges on previous studies that characterized aerosol properties for different parts of the globe [see Kaufman, 1993; Tripathi et al., 2005; Hewitt and Jackson, 2009; Israelevich et al.; 2012]. Note that these ranges encompass most situations, except for extreme events such as biomass burning or dust storms.

Details are in the caption following the image
(a) CBL height (shades), downward irradiance (red continuous contour in W m−2), and time delay (in hours) of the morning inversion layer breakup compared to the CLEAR case (white-dashed contour). (b) Potential temperature (shades) and aerosol HR (black continuous contour in K d−1) as a function of the aerosol optical depth and single scattering albedo. The black (blue) dots represent the CONTROL (AERO+) ω and τ conditions. All variables are shown for CESAR2008 case at 10 UTC, except the time delay of the morning inversion breakup, which is based on the morning onset time of CLEAR.
Details are in the caption following the image
Same as Figure 8 but for the CESAR2003 (13.5 UTC) except in Figure 9a where the black dashed contours are the time advance (in hours) of the CBL afternoon collapse compared to the CLEAR case.

4.2.1 MXL Model Validation Against LES Results

To show the validity of this approach, we first compare the results of the MXL model to the LES model for CESAR2008 (10 UTC) and CESAR2003 (13.5 UTC) in Tables 2 and 3, respectively. Note that we performed an additional LES experiment similar to AERO+ but with ω = 0.7 to further evaluate the response of the MXL model to strongly absorbing aerosols.

Table 2. MXL Model and LES Model Results for CESAR2008 at 10 UTCa
Variable/Experiment CONTROL CLEAR AERO+ AERO+ω = 0.7
θ (K) 295.2 (+0.20) 295.2 (+0.15) 294.9 (+0.20) 293.9 (+0.20)
HR (K d−1) 0.7 (+0.3) - 2.4 (+0.6) 10.5 (+0.8)
zi (m) 1420 (−266) 1600 (−148) 730 (−230) 240 (−54)
Δt (min) 12 (+6) 0 (+9) 40 (+24) 156 (+30)
  • a The brackets show the MXL model deviations from the LES results. Δt is the time delay of the CBL morning onset compared to the CLEAR case. The initial conditions of the AERO+ω = 0.7 experiment are equal to the ones for AERO+, except ω = 0.7. The LES results are the vertically averaged values from the surface until zi.
Table 3. MXL Model and LES Model Results for CESAR2003 at 13.5 UTC
Variable/Experiment CONTROL CLEAR AERO+ AERO+ω = 0.7
θ (K) 290.8 (−0.05) 290.7 (+0.05) 291.0 (−0.25) 291.8 (−0.15)
HR (K d−1) 1.3 (−0.10) - 3.7 (+0.25) 11.4 (+0.45)
zi (m) 1185 (+35) 1185 (+71) 1155 (+15) 1310 (+75)
Δt (min) −13 (−42) 0 (−6) −50 (+10) −97 (+19)
  • a The brackets show the MXL model deviations from the LES results. Δtis the time advance of the CBL afternoon collapse compared to the CLEAR case. The initial conditions of the AERO+ω = 0.7 experiment are equal to the ones for AERO+, except ω = 0.7. The LES results are the vertically averaged values from the surface until zi.

We observe that the MXL model properly captures the (thermo)dynamical responses to aerosol heating. Like the LES, in the MXL model the morning potential temperature response (Table 2) to aerosol forcing is small, unless strongly absorbing aerosols are abundant. In this situation—AERO+ω = 0.7— the morning CBL is cooled by 1 K in both models and the heating rates are similar (deviations <10%). The underestimation of the CBL height is due to the slight delay of the CBL morning onset observed in the MXL model results (see also Figure 3d). The delay of the CBL morning onset is explained by the slightly different initial conditions used for the LES and the MXL models (see Figure 2). For the afternoon CBL (CESAR2003, Table 3) a similar performance of the MXL model is found. The deepening of the afternoon CBL caused by strongly absorbing aerosols is slightly overestimated in the MXL model because the earlier stabilization of the potential temperature is not taken into account due to the well-mixed nature of the MXL model. Finally, the significant advance of the afternoon CBL collapse upon the presence of aerosols is well captured. In conclusion, the performance of the MXL suffices to map the aerosol effect on the CBL dynamics for a wider range of aerosol characteristics. The satisfactory performance of the MXL model indicates that the mixed-layer theory assumptions, i.e., infinitesimally thin inversion layer and the well-mixed CBL, do not dramatically change the CBL thermodynamics behavior, if compared to the LES results. Despite the good performance, we note, however, that the availability of an LES model remains important to validate the MXL model results, like presented in Tables 2 and 3.

4.2.2 Aerosol Impacts on the Morning CBL

As shown in Figure 8, we find shallower morning CBLs if τ increases. For example, similar to the CONTROL experiment (black dot), for ω = 0.92 and τ = 0.2, the CBL height at 10 UTC reaches about 1400 m decreasing to around 700 m for ω = 0.92 and τ = 0.6(AERO+ experiment, blue dot). We find that the sensitivity of the CBL growth (and irradiance) to τ increases if the aerosols become more absorbing, i.e., ωdecreases. In the case of scattering aerosols more energy reaches the surface because forward scattering prevails. If aerosols absorb instead of scatter SW radiation, some of this forward scattered radiation is absorbed in the CBL, further reducing the SW radiation reaching the surface. By increasing the amount of strongly absorbing aerosols (toward the lower right corner of Figure 8), the sensible heat flux is reduced significantly (see Figure 5b). Consequently, there is insufficient energy to overcome the temperature inversion and the residual layer remains present above the CBL (see also Figure 10b) the entire day (dashed contours in Figure 8a).

Details are in the caption following the image
Surface and (thermo)dynamics variables as a function of the aerosol optical depth for CESAR2008. The shades indicate the range of variation between strong absorption ω = 0.7 (continuous lines) and purely scattering aerosols ω = 1.0 (dashed lines). (a) Diurnally averaged QNET (black), LE (red), and SH (cyan). (b) Maximum CBL height (black dots and dotted line) and diurnal averages of total energy (blue, right axis) and CBL heat input (red, right axis). In Figure 10b the numbers next to the dots (ω = 0.7) indicate the time delay of the morning inversion layer breakup compared to the CLEAR case; for τ > 0.6 and ω = 0.7 the response of hMAX is schematized because the CBL does not overcome hRL (1700 m, thin black dotted line). Note the different scales for the y axes. (c) Maximum (red) surface temperature and diurnally and mixed-layer averaged potential temperature difference compared to CLEAR (black, right axis).

Our results also show that the morning CBL (10 UTC) is significantly warmer either for clear conditions (τ close to zero) or for purely scattering aerosols (ω close to unity). The physical explanation is the following: forward scattering aerosols (or clearer atmospheres) allow more SW radiation to reach the surface, leading to higher sensible heat fluxes, warming the morning CBL. As shown in Figures 6a and 6b, at 10 UTC the heat budget is mainly driven by the combination of the surface and entrainment turbulent fluxes, explaining the increase in the CBL potential temperature. If τ increases, more SW radiation is absorbed within the aerosol layer (increasing the HR for absorbing aerosols). However, under the conditions of CESAR2008 case, the surface inversion layer jump is not weakened because part of the aerosols reside above the CBL, i.e., ha>zi. The CBL remains therefore shallow, and θ is mainly driven by the turbulent surface fluxes, which are reduced due to aerosol extinction of SW radiation. These findings emphasize the importance of representing adequately the aerosol layer since its depth plays a crucial role in the surface energy partitioning and CBL growth.

4.2.3 Aerosol Impacts on the Afternoon CBL

Analogously, we explore in Figure 9 the results for CESAR2003. For the aerosol impacts on the afternoon CBL we focus on the CESAR2003 case because for CESAR2008 the surface inversion is not broken for certain aerosol conditions (see Figures 8a and 10). We select 13.5 UTC as a representative time for this case study because SH >0 and the aerosols are completely incorporated within the CBL for all the experiments.

The main difference with Figure 8 is the fact that during the afternoon the heating due to aerosols alters more significantly the heat budget (see Figure 6b). Our results agree with the findings of Yu et al. [2002] pointing out that strongly scattering aerosols lead to an afternoon CBL that is shallower and colder. The physical explanation is the following: scattering aerosols reduce irradiance—backscattering SW radiation—without incrementing the HR (solid contours in Figure 9b). The opposite is observed for absorbing aerosols because the reduction of the irradiance is overcompensated by the increase of the HR. For the CBL height, the HR compensates the irradiance reduction at ω ∼ 0.83. Interestingly, for potential temperature—at 13.5 UTC—the effects counterbalance at ω ∼ 0.91. This value is slightly higher than ω ∼ 0.86 found by Lyamani et al. [2010].

We observe a shortening of the convective period—defined as the part of the diurnal cycle when SH >0— if aerosols are present (see also SHAERO+ ratio in Figure 5b). Our results indicate that the earlier decay of turbulence advances the afternoon collapse of the CBL up to 2.5 h for high loads of strongly absorbing aerosols (dashed contours in Figure 9a). This effect has important implications for the establishment of the stable boundary layer [see Nair et al., 2011]. As shown in Figure 9a (lower right portion), the stable boundary layer is formed earlier (up to 2.5 h) if strongly absorbing aerosols are present.

4.2.4 Aerosol Impacts on the Diurnal Evolution of the CBL

In this section we apply the MXL model to quantify the impact of aerosol optical properties—τ and ω—on surface and atmospheric variables. We present maxima and diurnal averages (denoted by an overbar) for the CESAR2008 case. Our purpose here is to study how aerosols affect important state variables that are normally used to investigate atmospheric phenomena with larger temporal and spatial scales than the boundary layer scale. By doing so, we broad our understanding on how the land-atmosphere system responds to aerosol forcing. The impacts of purely scattering (ω = 1.0) and strongly absorbing (ω = 0.7) aerosols on the diurnal evolution of the CBL are shown in Figure 10. We use SH >0 as a criteria to determine the duration of the convective period. Note that we quantify the aerosol impact by normalizing the differences by the CLEAR case value (urn:x-wiley:jgrd:media:jgrd51411:jgrd51411-math-0007), where φis the variable under study.

In Figure 10a we show the diurnally averaged surface fluxes. Aerosols significantly diminish the amount of available energy at the surface. The reduction in urn:x-wiley:jgrd:media:jgrd51411:jgrd51411-math-0008 is more pronounced for strongly absorbing aerosols (44%) than for purely scattering aerosols (17%). Even for heavy loads of strongly scattering aerosols (τ = 1.0and ω = 1.0) the reductions in urn:x-wiley:jgrd:media:jgrd51411:jgrd51411-math-0009(15%) and urn:x-wiley:jgrd:media:jgrd51411:jgrd51411-math-0010(22%) remain relatively small compared to CLEAR. In contrast, for heavy loads of strongly absorbing aerosols (τ = 1.0 and ω = 0.7), urn:x-wiley:jgrd:media:jgrd51411:jgrd51411-math-0011 and urn:x-wiley:jgrd:media:jgrd51411:jgrd51411-math-0012 diminish by 43% and by 50%, respectively. Note that under the well-watered surface conditions, the percentual reduction is more significant for urn:x-wiley:jgrd:media:jgrd51411:jgrd51411-math-0013 than for urn:x-wiley:jgrd:media:jgrd51411:jgrd51411-math-0014 (see also Figures 5b and 5c). The reduction observed for urn:x-wiley:jgrd:media:jgrd51411:jgrd51411-math-0015 has direct implications for diurnal average evaporation (not shown), decreasing from 6.4 mm d−1 for the CLEAR case to 3.7 mm d−1 for heavy loads of strongly absorbing aerosols. For purely scattering aerosols a slight reduction of 1.0 mm d−1 is observed for τ = 1.0. These results confirm the findings of Biasutti and Giannini [2006], where aerosols intensified dryness in the Sahel region.

As shown in Figure 10b, we find for purely scattering aerosols that the reduction in urn:x-wiley:jgrd:media:jgrd51411:jgrd51411-math-0016 diminishes the maximum CBL height with 95 m for τ = 1.0. The total energy in the system integrated over the CBL and the residual layer also decreases with 17%. We define the CBL heat input (HI) similarly as Barbaro et al. [2013]:
urn:x-wiley:jgrd:media:jgrd51411:jgrd51411-math-0017(1)

For purely scattering aerosols urn:x-wiley:jgrd:media:jgrd51411:jgrd51411-math-0018 also decreases to 78% of the CLEAR value. Depending on τ, strongly absorbing aerosols have a different impact on the growth of the CBL. We find that the maximum CBL height increases by up to 65 m and that the urn:x-wiley:jgrd:media:jgrd51411:jgrd51411-math-0019 increases by up to 46% compared to CLEAR for τ≤0.4. Note the maxima in CBL height and CBL urn:x-wiley:jgrd:media:jgrd51411:jgrd51411-math-0020 for τ = 0.4. For τ > 0.4 the maximum CBL height diminishes because the reduction of urn:x-wiley:jgrd:media:jgrd51411:jgrd51411-math-0021 significantly delays the onset of the CBL growth. If τ > 0.6 the heat introduced by urn:x-wiley:jgrd:media:jgrd51411:jgrd51411-math-0022 is insufficient to totally overcome the residual layer (see also Figure 8a). Note that for τ > 0.8 the increase of urn:x-wiley:jgrd:media:jgrd51411:jgrd51411-math-0023 compensates the less acute decrease in both zi and urn:x-wiley:jgrd:media:jgrd51411:jgrd51411-math-0024, leading to a nearly constant CBL urn:x-wiley:jgrd:media:jgrd51411:jgrd51411-math-0025. The maximum CBL height equals to 278 m for τ = 1.0. For strongly absorbing aerosols, a significant part of the energy that does not reach the surface is absorbed by the aerosols located within the RL and the CBL. Therefore, we note an increase in the total energy up to 37% compared to CLEAR as τ increases. Note that despite of the shallower CBLs, the total energy in the system always increases because of the heating of the residual layer. The impact of aerosols on the total energy significantly modifies the effective albedo (αEFF) by altering the irradiance and the outgoing SW radiation (see Figure 5a). Compared to CLEAR, αEFF decreases by 40% for strongly absorbing aerosols and increases by 64%for purely scattering aerosols.

We show in Figure 10c the maximum surface temperature (TMAX), and diurnally and mixed-layer averaged potential temperature difference compared to the CLEAR case (urn:x-wiley:jgrd:media:jgrd51411:jgrd51411-math-0026). We observe that irrespective of their absorption properties, aerosols always reduce the daytime surface temperature. TMAX diminishes, by 0.8 K (ω = 0.7) and by 1.2 K (ω = 1.0), for the entire τ range. The differences in TMAX are small among the absorption regimes. TMAX is always slightly larger for purely scattering aerosols except for high loads of absorbing aerosols due to the less accentuated reduction in urn:x-wiley:jgrd:media:jgrd51411:jgrd51411-math-0027. The urn:x-wiley:jgrd:media:jgrd51411:jgrd51411-math-0028 evolution follows the CBL urn:x-wiley:jgrd:media:jgrd51411:jgrd51411-math-0029. For strongly absorbing aerosols and τ > 0.8 the CBL urn:x-wiley:jgrd:media:jgrd51411:jgrd51411-math-0030 remains nearly constant and is used to heat a shallower CBL, increasing urn:x-wiley:jgrd:media:jgrd51411:jgrd51411-math-0031.

5 Conclusions

In this study we quantified the effects of aerosol scattering and absorption of shortwave (SW) radiation on the surface energy budget and on the convective boundary layer (CBL) dynamics. To this end, we coupled an atmospheric large-eddy simulation (LES) model and a mixed-layer (MXL) model to (i) a land-surface model and (ii) a SW radiation transfer model. We successfully validated our LES and MXL model results using measurements of (thermo)dynamic variables and aerosol properties for two typical CBL prototypes: CESAR2008 and CESAR2003. During CESAR2008, the early morning potential temperature profile was characterized by a well-mixed residual layer above a strong near surface inversion, leading to a rapid onset of the CBL during the morning transition. CESAR2003, in contrast, was characterized by a continuous growth of the CBL. Given the good agreement between the LES and MXL model results, we explored the aerosol effect on the land-atmosphere system for a wide range of optical depths and single scattering albedos.

The LES and MXL model results showed that over the studied well-watered grassland, aerosols reduced the sensible heat flux more than the latent heat flux. As a result, relatively scattering aerosols decreased the Bowen ratio from around 0.47 at 08 UTC down to values as low as 0.12 at 14 UTC. Aerosols also delayed (up to 4.5 h) or even prevented the CBL morning onset and hasten (up to 2.5 h) its afternoon collapse. Not only the vertical distribution of the aerosols played an important role on the CBL evolution but also the initial temperature profile. In the CESAR2008 case, for instance, we found that strongly absorbing aerosols in the residual layer could maintain a persistent near-surface inversion for the entire day. When the aerosols were entrained in the CBL, we observed a strong dependence of the single scattering albedo on the afternoon CBL (thermo)dynamics: for CESAR2003 the strongly absorbing aerosols (ω = 0.7) deepened and warmed (+140m and +1.2K, respectively), while purely scattering aerosols shallowed and cooled (−280 m and −1.0 K, respectively) the afternoon CBL if compared to the CLEAR case. The diurnally averaged surface net radiation for CESAR2008 showed a strong dependence on the type of aerosol, decreasing by 68 W m−2 for heavy loads of purely scattering aerosols and by 180 W m−2 for the same load of strongly absorbing aerosols. Under the studied surface conditions, the diurnal average evaporation decreased by 16%and by 42%for purely scattering and strongly absorbing aerosols, respectively, if compared to CLEAR.

Due to the comprehensiveness of the observational data set and the LES results discussed here, our study can be used as a benchmark to evaluate the coupling and the performance of the parameterizations for SW radiation, land surface, and boundary layer schemes, implemented in mesoscale or global chemistry transport models. In particular, we showed the intrinsic nonlinear couplings within the land-atmosphere system. The impact of aerosols for different surfaces and heterogeneous conditions as well their effects on the LW cooling will be investigated in a future study.

Acknowledgments

We thank Ping Wang and Wouter Knap for providing us with the EUCAARI experiment data set. We are grateful to Henk Klein Baltink and Fred Bosveld for the CESAR data set (www.cesar-database.nl). E. Barbaro thanks Stephan de Roode for his very useful discussion of the radiation results. M. Krol is supported by the EU FP7 IP PEGASOS (FP7-ENV-2010/265148). The research was supported by the NWO grant for computing time (SH-060-13). Finally, we gratefully thank the three anonymous reviewers for their helpful comments that contributed to improving this manuscript.