Long-term Aerosol-mediated Changes in Cloud Radiative Forcing of Deep Clouds at the Top and Bottom of the Atmosphere over the Southern Great Plains

Aerosols can alter the macro-and micro-physical properties of deep convective clouds (DCCs) and their ra-diative forcing (CRF). This study presents what is arguably the first long-term estimate of the aerosol-mediated changes in CRF (AMCRF) for deep cloud systems derived from decade-long continuous ground-based and satellite observations , model simulations, and reanalysis data. Measurements were made at the US Department of Energy's Atmospheric Radiation Measurement Program's Southern Great Plains (SGP) site. Satellite retrievals are from the Geostationary Operational Environmental Satellite. Increases in aerosol loading were accompanied by the thickening of DCC cores and the expansion and thinning of anvils, due presumably to the aerosol invigoration effect (AIV) and the aerosol microphysi-cal effect. Meteorological variables dictating these cloud processes were investigated. Consistent with previous findings, the AIV is most significant when the atmosphere is moist and unstable with weak wind shear. Such aerosol-mediated systematic changes in DCC core thickness and anvil size alter CRF at the top of atmosphere (TOA) and at the surface. Using extensive observations, ∼ 300 DCC systems were identified over a 10 years period at the SGP site (2000–2011) and analyzed. Daily mean AMCRF at the TOA and at the surface are 29.3 W m −2 and 22.2 W m −2 , respectively. This net warming effect due to changes in DCC microphysics offsets the cooling resulting from the first aerosol indirect effect.


Introduction
Aerosols can alter the radiative energy of the earth's surfaceatmosphere system by directly attenuating solar radiation and/or by indirectly modifying cloud macro-physical (areal coverage, structure, altitude) and microphysical properties (droplet size, phase) in many different ways.Studies have shown that high aerosol loadings can produce a large number of tiny cloud droplets (Twomey, 1977), which suppress the warm rain-forming process through indirect effects (Albrecht, 1989).This allows more cloud particles to ascend above the freezing level and convert to ice hydrometeors.During this ice process, the release of more latent heat invigorates the vertical development of clouds and enhances precipitation and lighting (Rosenfeld et al., 2008a;Yuan et al., 2011).This effect is the so-called aerosol invigoration effect (AIV), which was first noted by Williams et al. (2002) who found that aerosols could fuel cloud vertical development and lift cloud-top heights.This was reinforced by an airborne experiment conducted in the Amazon (Andreae et al., 2004).The AIV is also associated with the expansion of cloud anvil extent (Koren et al., 2010a).Such aerosolmediated changes in cloud parameters alter cloud radiative forcing (CRF), which is a component of aerosol radiative forcing (ARF).To differentiate it from ARF exerted under clear-sky conditions, it is often referred to as aerosol indirect forcing or the aerosol indirect effect.It is called the aerosolmediated CRF (AMCRF) in this study.Among the aerosolinduced changes in radiative energy (IPCC, 2013), AMCRF Published by Copernicus Publications on behalf of the European Geosciences Union.
has the largest uncertainty because of the poor understanding of the mechanisms behind it, as reviewed recently by Rosenfeld et al. (2013).To date, only radiative forcing due to the first aerosol indirect effect has been considered in most climate models.Using observations to quantify the AIV and its impact on climate is very important for climate modeling and climate prediction.
The AIV has been demonstrated during several aircraft field campaigns, such as those studying tropical clouds in the Amazon (Andreae et al., 2004), hail storms in Argentina (Rosenfeld et al., 2006), winter storms in California (Rosenfeld et al., 2008a), and summer monsoon clouds in India (Freud and Rosenfeld, 2012).Satellite measurements have also been used to quantify the AIV over the Amazon basin (Lin et al., 2006), over the Atlantic Ocean (Koren et al., 2005;Storer et al., 2014), and over the entire global tropics (Niu and Li, 2012).The AIV occurs only under certain microphysical, dynamic and thermodynamic conditions (Khain and Pokrovsky, 2004;Khain et al., 2004Khain et al., , 2005Khain et al., , 2008;;Wang, 2005;Seifert and Beheng, 2006;van den Heever et al., 2006;Fan et al., 2007Fan et al., , 2009)).Rosenfeld et al. (2008b) postulated that the AIV is most significant for deep convective clouds (DCCs) with warm cloud bases and cold cloud tops.This was confirmed by an analysis of 10 years of observation data collected at the Atmospheric Radiation Measurement (ARM) Southern Great Plains (SGP) site (Li et al., 2011).The modeling study of Fan et al. (2009) further showed that weak wind shear and a moist environment favors the AIV.The expansion of anvil clouds associated with DCCs can induce positive radiative forcing during the invigoration process, while the thickening of the cores of DCCs may lead to cooling, rendering an uncertain net radiative effect (Koren et al., 2010a;Fan et al., 2012).The diurnal variation of DCCs is a key factor dictating the net effect because the shortwave cooling effect is most prominent at noon, and is nil at night.
Following Li et al. (2011) who revealed the impact of aerosols on cloud vertical development and precipitation frequency using ground-based measurements, this study aims to (1) identify factors governing the AIV and AIV-induced changes in CRF under various thermodynamic and dynamic environmental conditions, and (2) estimate the long-term AIV-induced changes in the CRF at the top of the atmosphere (TOA) and at the surface using a combination of geostationary satellite retrievals and ground-based observations made at the US Department of Energy's Atmospheric Radiation Measurement Program's Southern Great Plains (SGP) site.
A brief description of various data sets used in this study is presented in Sect.2, along with a description of the procedure used to calculate AMCRF.Section 3 shows the aerosol effect on DCC properties, such as cloud-top height, size of the cloud anvil and convective core, and radiative forcing arising from changes in cloud properties induced by the AIV.Section 4 summarizes the findings.

Surface observations
The SGP site was the first field observation site established by the US Department of Energy's Atmospheric Radiation Measurement Program (Stokes and Schwartz, 1994;Ackerman and Stokes, 2003) and has collected continuous measurements of extensive aerosol, cloud and meteorological variables for the past 25 years.This study uses data from 2000 to 2011 and is focused on convective clouds, which are more susceptible to aerosols (Tao et al., 2012).Condensation nuclei (CN) number concentration measured by a TSI model 3010 condensation particle counter is used as a proxy for aerosols.Cloud condensation nuclei (CCN) would be a better proxy, but there are not as many measurements available at the SGP site and many gaps in the times series are present due to the more delicate nature of the measurement technique.Relative to the more widely available aerosol optical depth (AOD), CN has a closer relationship with cloud droplets.
Cloud base and top boundary information are inferred from a suite of active sensors including the millimeter wavelength cloud radar (MMCR), the ceilometer, and the micropulse lidar (MPL) (Clothiaux et al., 2000).The cloudtop height might be blurred under heavy rain condition as the radar signal would be attenuated, but they are also valid for the majority of DCCs that do not rain.Combined with temperature profiles from the European Centre for Medium-Range Weather Forecast (ECMWF) diagnostic analyses, cloud base and top temperatures can be generated.The cloud-top temperature (CTT) helps identify the phase of a cloud and the cloud-base temperature (CBT) indicates the likelihood of interactions between clouds and aerosols that are located chiefly in the planetary boundary layer.As in Li et al. (2011), clouds with CBT greater than 15 • C are selected.This maximizes the likelihood that the CN measured on the ground will likely have an impact on cloud properties.ARM data, sampled every 10 s, are averaged over 30 min intervals in this study.This choice of averaging interval is discussed in Sect.2.3.
Other parameters from the ECMWF diagnostic analyses used in the study include vertical velocity, wind shear, and convective available potential energy (CAPE).AOD used in radiative transfer calculations under clear-sky conditions is retrieved from multi-filter rotating shadowband radiometer (MFRSR) measurements (Min and Harrison, 1996;Min et al., 2004Min et al., , 2008;;Wang and Min, 2008).Columnintegrated amounts of water vapor (PWV) and the cloud liquid water path (LWP) are retrieved from microwave radiometer (MWR) measurements of brightness temperatures at 23.8 GHz and 31.4GHz (Liljegren, 1999).Fractional sky cover retrievals from the total sky imager (TSI) are also used.Hemispheric radiative fluxes measured by pyranometers and pyrgeometers provide information about the radiation budget

Satellite retrievals
The Geostationary Operational Environmental Satellite (GOES) platform provides continuous pixel-level (4 km) cloud properties and TOA radiative fluxes.Broadband shortwave (SW) and longwave (LW) radiative fluxes at the TOA are estimated from narrowband GOES data using narrowband-to-broadband (NB-BB) conversion formulas (Minnis et al., 1995).Formula coefficients have historically been derived by regressing matched polar orbiting satellite BB fluxes or radiances with their NB counterparts from GOES.The visible infrared solar-infrared split window technique (VISST: Minnis et al., 2011;Huang et al., 2005)  for investigating the AIV and to estimate AMCRF at the surface.

Methodology
The study uses both ground-based and satellite measurements, which cover different spatial areas and have different temporal frequencies.Ground measurements are made continuously at the SGP site at a high frequency.Satellite observations are made over a large area around the SGP site, but not as frequently.Cloud fractions for different spatial domains centered on the SGP site were calculated from satellite retrievals and cloud frequencies for different averaging intervals were calculated from ground measurements.The optimal match is achieved when the satellite product is averaged over a 20 km × 20 km area around the SGP site and when surface observations are averaged over 30 min, centered on the satellite overpass time.Note that the spatial and temporal match between the two types of measurements depends on the cloud regime (Dong et al., 2002;Li et al., 2005;Yan et al., 2011).

The AIV
A variety of cloud types can appear over the SGP site, which experiences large seasonal variations in temperature and humidity.Analyses of long-term cloud products from ARM measurements reveal that stratus and cirrus clouds are the major cloud types seen at the SGP site (  , 2000;Kollias et al., 2007).Cumulonimbus clouds are not frequent there.Theoretical and observation-based studies have demonstrated that the AIV is significant for mixedphase warm-based clouds (Andreae et al., 2004;Koren et al., 2005;Rosenfeld et al., 2008b;Li et al., 2011).Invigorated clouds have a greater chance of developing into DCCs by lifting cloud-top heights and expanding cloud anvil fractions (Koren et al., 2010a).Taking advantage of pixel-level cloud optical depth (COD, or τ ) from GOES retrievals, DCCs are further classified as deep convective cores (pixels with τ > 10) and anvils (pixels with τ < 10), as defined by Koren et al. (2010a).GOES data has a resolution of 4 km which is too coarse to resolve clouds, especially warm cumulus clouds.However, this study is concerned with large DCC systems, which are identified using ground observations.Figure 1a shows the relationship between CTT and CN concentration for single-layered clouds.In the case of a precipitating cloud layer, CN concentrations measured 1.5 h before the observed rainfall were used.During the decade-long period, the total number of matched cloudy cases identified using ground-based and GOES data is 22, 820.Of the total, 861 are single-layer clouds with CBT greater than 15 • C. The number of warm shallow clouds (CTT > 0 • C) and mixed-phase clouds (CTT < −4 • C) are 516 and 299, respectively.CTT decreases from −46.0 • C to −51.4 • C as the aerosol loading increases from 1000 cm −3 to 5000 cm −3 for cases where CTT < −4 • C. The CTT for warm shallow clouds, however, show little change with increase in CN concentration.This finding is similar to that of Li et al. (2011).For mixed-phase warm-based clouds, an increase in aerosol loading reduces the size of cloud droplets.These droplets are more likely to be lifted by updrafts to become mixed-phase clouds due to buoyancy from the release of latent heat, which invigorates convection.Figure 1b shows the relationship between cloud fraction and CN concentration for three subsets of DCCs: DCCs, DCC cores, and anvils associated with the DCCs.While DCCs and DCC core cloud fractions decrease with increasing CN, anvil cloud fractions increase as the aerosol loading increases, consistent with statistical analyses and model simulations performed over other regions of the globe (Koren et al., 2010a;van den Heever et al., 2011).
Although aerosol particles play a key role in the formation of cloud droplets, cloud development is mainly driven by large-scale dynamic conditions.Koren et al. (2010b) used satellite data to study correlations between cloud properties and all variables from reanalysis data and found that vertical velocity and humidity are the primary meteorological controls on a cloud system.Through sensitivity studies using a cloud-resolving model, Fan et al. (2009) found that wind shear plays a dominant role in influencing the intensity of the AIV.Li et al. (2011) examined the correlation relationships between the CN concentration and some meteorological variables influencing cloud development and found that they were weak.This suggests that changes in cloud properties in- duced by atmospheric dynamics/thermodynamics can be effectively removed by using a very large number of samples.
To identify environmental factors that might affect the properties of clouds selected for this study, relationships between cloud properties and aerosol loading are analyzed according to vertical velocity (vv), CAPE, PWV and wind shear (ws).
Wind shear is calculated as the maximum of horizontal wind speed (u) minus the minimum of u within the lowest 7 km of the atmosphere (Fan et al., 2009).Figure 2 shows CTT as a function of CN concentration for the different environmental conditions selected here.Each meteorological parameter was divided into two groups representing diametrically possible values, e.g., rising motion and subsidence for the case of vv.Each group contains approximately equal numbers of samples and enough samples to reduce the noise generated by the analyses.No matter if the environment is unstable (vv > 0.0 ms −1 , CAPE > 500 Jm −2 ) or stable (vv < 0.0 ms −1 , CAPE < 500 Jm −2 ), moist (PWV > 5 cm) or dry (PWV < 5cm), CTT decreases as CN concentration increases (Fig. 2a-c).Under unstable and/or moist conditions, increasing the aerosol concentration in mixed-phase warmbased clouds generates more latent heat which, in turn, invigorates convection.The magnitude of the slope in the CTT-CN relationship is then larger than that under stable and/or dry conditions (Fig. 2b and c). Figure 2d shows that increasing the aerosol loading likely suppresses convection under strong wind shear conditions, but invigorates convection under weak wind shear conditions.Under weak wind shear conditions, the increase in condensational heating is more than the increase in evaporative cooling and/or entrainment as the aerosol load increases, leading to an increase in net latent heat release and stronger convection.With strong wind shear, the opposite holds, leading to the suppression of convection.
The above analyses together with those made in an earlier study by Li et al. (2011) for the same period and in the same region are consistent with the phenomenon of the AIV postulated by Rosenfeld et al. (2008).Causes for the AIV were recently revised significantly by Fan et al. (2013).While the thermodynamic role of aerosols in modulating latent heat release to fuel the development of a DCCs is still valid, it is the aerosol microphysical effect that seems to play a more important role, especially in the expansion of the anvil fraction.The extensive model simulation results obtained by Fan et al. (2013) agree well with observation-based findings, which may thus serve as another piece of evidence supporting aerosol and DCC interactions.

Determination of AMCRF
CRF is defined as the difference in net radiative fluxes under all-sky conditions minus those under clear-sky conditions:  Figure 4 shows SW, LW, and net CRF at the TOA (lefthand panels) and at the surface (right-hand panels) for DCCs as a function of SZA for different ranges of CN concentration and as a function of CN concentration.As expected, LW CRF is independent of SZA, and increases with increasing CN.In-stantaneous SW CRF is largely influenced by the intensity of solar radiation and ranges from 0 to ∼ −600 Wm −2 .As the CN range changes from 0-2000 cm −3 to 2000-4000 cm −3 , a noticeable and systematic increase in the magnitude of CRF is clearly seen, despite the larger uncertainty due to the smaller number of samples for the larger CN range.There are much fewer data samples in the largest CN range bin, leading to a larger uncertainty in the estimate of CRF.The magnitude of surface SW CRF is larger than that of TOA SW CRF, while the opposite is the case for LW CRF.
Instantaneous AMCRF is defined as In principal, CN 0 should be equal to zero, which is physically impossible.So an unperturbed CN value representative of relatively clean conditions is used instead.Diurnal mean AMCRF is defined as .µ = cos (SZA) = sin ϕ sin δ + cos ϕ cos δ cos 2πt 24 ( 4) where T is the half-day length which varies according to latitude and season, and ϕ and δ are latitude and declination, respectively.Equations ( 2)-( 4) are adopted from Li and Garand (1994).Note that the diurnal variation of SW CRF not only includes the effect of varying SZA, but also the diurnal changes in DCCs.The diurnally-averaged CRF in Eq. ( 2) is determined numerically as where n is the number of data points in Fig. 4a-c and eg.The discrete values of µ i and AMCRF(µ i ) at the TOA and at the surface have been substituted into these equations and the trapezoidal rule has been applied for each interval.Diurnal mean values for the individual components are given both in numbers (upper six panels) and plotted as functions of CN (lower two panels).
The daily mean TOA LW CRF increases monotonically from 113.1 ± 3.6 Wm −2 to 127.8 ± 8.3 Wm −2 as the mean CN concentration increases from 1000 cm −3 to 5000 cm −3 .This is accompanied by a decrease in mean CTT from −46.0 • C to −51.4 • C (Fig. 4d).The slight enhancement of LW warming at the TOA is due to the reduction in CTT, leading to a decrease in outgoing thermal radiative fluxes.At the surface level, the daily mean LW CRF changes from 66.7 ± 1.9 Wm −2 to 72.7 ± 4.3 Wm −2 (Fig. 4h).The much smaller increase is consistent with the finding that cloud bases are much less sensitive to CN, as shown by Li et al. (2011).
The daily mean TOA SW CRF remains constant (∼ −354 W m −2 ) as the mean CN concentration increases from 1000 cm −3 to 3000 cm −3 .It then sharply increases from −353.6 ± 37.1 Wm −2 to −265.3 ± 56.3 Wm −2 when the CN concentration increases from 3000 cm −3 to 5000 cm −3 .At the surface, changes in SW CRF are similar to those at the TOA, but with smaller magnitudes.The relationship between SW CRF and CN concentration is complicated because other factors are at play, such as cloud phase, τ , and cloud droplet size distribution.Aerosols could influence all cloud variables, but their influence is unclear.For a developing convective cloud system, the change in SW CRF is always negative because of the enhanced albedo effect.On the other hand, when mixed-phase clouds develop into mature DCCs, the cloud albedo effect easily saturates.
The relationship between net CRF and CN concentration follows the same pattern as that of the SW CRF at the top and bottom of the atmosphere because SW CRF is much larger in magnitude than LW CRF.The expansion of anvil sizes associated with the AIV produces more semi-transparent cirrus clouds, which could lead to a strong positive CRF.
R cloudy (SZA, CN) at the TOA is a mean value obtained by averaging all cloudy pixels in a 20 km × 20 km box centered on the SGP site.Changes in cloud coverage induced by the AIV are also taken into account.The competition between the cooling effect from the convective core and the warming effect from anvils in DCCs ultimately determines the sign of the net CRF induced by the AIV.Koren et al. (2010a) reached the same conclusion from model simulations of idealized tropical DCCs.In this regard, the observational findings presented here support their theoretical argument.
Figure 5 attempts to gain further insight into the two competing effects.DCCs were divided into thicker convective clouds (COD > 10), including deep but small convective cores, and thinner stratiform clouds (COD < 10), including thin but extensive anvils, following Koren et al. (2010a).For simplicity, they are referred to as convective and core clouds (CCCs) and stratiform anvil clouds (SACs).Ideally, a DCC should be classified as a deep convective core associated with heavy precipitation, thick stratiform raining clouds, and thin non-precipitating clouds, as was done by Feng et al. (2011) following an integrative analysis of ground-based scanning radar and geostationary satellite data.A small subset of such DCC cases was identified at the SGP site during the 10-year period under study.
Figure 5 shows COD, cloud-top height, and CRF (SW, LW, and net) as a function of CN concentration for CCCs and SACs.The COD associated with SACs remains constant as aerosol loading increases.The COD associated with CCC drops from 53.9 to 40.7 as CN concentrations increase from 1000 cm −3 to 5000 cm −3 .This finding supports a new theory about the role of aerosols in the development of DCC (Fan et al., 2013).The conventional wisdom concerning the AIV (Rosenfeld et al., 2008a) is that by serving as CCN, more aerosol particles lead to smaller cloud droplets that are difficult to grow into raindrops, and thus are more likely to be elevated to above the freezing level.The release of additional latent heat further fuels convection to render clouds deeper and larger.This effect is referred to as the aerosol thermodynamic effect.Through extensive modeling and comparisons with observations, Fan et al. (2013) found that the thermodynamic effect plays an essential role in the initial stage of DCC development only.As a DCC enters into its mature and decaying stages, the aerosol microphysical effect becomes more dominant.Gravitational settling of a larger number of smaller ice crystals does not occur, so these ice crystals expand horizontally due to the capping near the upper troposphere and tropopause.During this process, more CCCs are transformed into SACs.This prevents the thickening of the CCC as the SAC, especially its anvil, spreads.As CN concentration increases from 1000 cm −3 to 5000 cm −3 , the mean cloud thickness averaged over the whole domain of 400 km 2 increases from 10.3 km to 11.0 km.As a result of the somewhat artificial division of COD = 10, the mean COD of the CCC decreases.However, this does not mean that cloud-top height decreases with increasing CN.On the contrary, mean cloud-top heights for both CCCs and SACs increase with increasing CN, and the rate of increase for SACs is more pronounced than that for CCCs.As the cloud expands horizontally, the cloud base is elevated, which is more significant for the SAC than for the CCC part of the DCC.
The elevation of SAC cloud-top heights leads to a LW warming effect.Decreases in COD and increases in CCC top height can lead to both SW and LW warming effects.Moreover, Fig. 1b illustrates that the areal coverage of SACs increases and the size of a given CCC shrink as the aerosol load increases, which could amplify the warming effect.For SACs, as the CN concentration increases from 1000 cm −3 to 5000 cm −3 , SW cooling decreases from −206.9 Wm −2 to −135.9 Wm −2 , LW warming increases from 95.2 Wm −2 to 122.4 Wm −2 , and the net effect is warming (Fig. 5c).Variations in SW and LW CRF for CCCs are more complicated, but the trend is generally warming, which is similar to what is shown in Fig. 4.
To further understand changes in CRF with changes in CN concentration, the frequency of cloud occurrence as a function of the COD and height (Z), using GOES pixellevel (4 km) data, is examined (Fig. 6).In COD-Z space, clouds grow taller and spread out at the top with lower optical depths, intensifying the warming effect.Under relatively clean conditions (CN concentration: 0 to 2000 cm −3 ), frequencies of cloud occurrence greater than 10 % (reddish color) occur when COD > 45 and the cloud-top height is greater than 10 km.Shifting to moderately polluted conditions (CN concentration: 2000 to 4000 cm −3 ), the largest frequency of cloud occurrence is in the bin with cloud-top heights between 11.5 km and 13.0 km, and with COD > 45.0, which indicates a well-developed convective core.The frequencies for bins with COD < 15 increase slightly and the overall altitudes of cloud-top height increase.This suggests that CCCs contribute the most to the CRF under moderately polluted conditions, even as the anvils are growing.Under heavily polluted conditions (CN concentration: 4000 to 6000 cm −3 ), the largest frequency of cloud occurrence is in the bin with cloud-top heights greater than 13.0 km, and with COD < 15.0.Grossly speaking, high CN concentrations are associated with the more frequent occurrence of higher and thinner SACs.
Over the SGP site, the frequency of occurrence of these kinds of mixed-phase warm-based clouds in different CN concentration bins (0 to 2000 cm −3 , 2000 to 4000 cm −3 and greater than 4000 cm −3 ) is 36.4%, 38.1 %, and 25.4 %, respectively.Using clear-sky (CN concentration: 0 to 2000 cm −3 ) CRF as a baseline, TOA CRFs induced by the AIV are 8.69 Wm −2 and 102.37 Wm −2 under moderately and heavily polluted conditions, respectively.A rough estimate of daily mean TOA net CRF induced by the AIV is about 29.3 Wm −2 (calculated by weighting CRFs by their corresponding frequencies of occurrence, i.e., 38.1 % • 8.69 + 25.4 % • 102.37).The daily mean surface net CRF induced by the AIV is 22.2 Wm −2 .The real climate impact of the AIV at this site is unknown because of the overall low frequency of occurrence of DCCs, which comprise 1.5 % of all cloud types, according to long-term cloud statistics developed for the SGP site.Nevertheless, a warming effect offsetting the general cooling effect of the first aerosol indirect effect is observed.

Conclusions
Previous studies have shown that increasing aerosol loading increases the chance for warm-based mixed-phase clouds to develop into DCCs.This invigoration plays an important role in the hydrological cycle and on atmospheric general circulation through the release of latent heat which would modify atmospheric heating profiles and change the radiation This study is a systematic analysis of the AIV using a longterm, continuous ground-based data set of measurements from the ARM SGP site, together with GOES satellite-based retrievals, and a reanalysis data set.Results show that an increase in aerosol loading can invigorate convective clouds by increasing cloud-top heights, thicknesses, and the expansion of anvil cloud fractions when the atmosphere is unstable and moist.Weak wind shear promotes the lifting of cloud-top heights, but discourages anvil expansion as aerosol loading increases.
CRF of DCCs for different ranges of CN concentration are calculated by combining TOA radiative fluxes retrieved from GOES observations and simulated by a radiative transfer model.Cooling from the convective core and warming from the anvils are competing effects in DCCs.Whether the sign of the climate impact is negative or positive depends largely on the stage in the life cycle of the convective system.Over the SGP site, the long-term daily mean aerosolmediated TOA and surface CRF due to the AIV is positive in sign, and 29.3 Wm −2 and 22.2 Wm −2 in magnitude, respectively.This is based upon data from ∼ 300 DCCs identified over a 10-year period at the SGP site (about 1.5 % of all clouds scenes occurring during this time).
The AIV on clouds is assessed using ground-based measurements, while cloud radiative forcing related to the AIV is calculated using an independent data set consisting of satellite retrievals and simulations from a radiative transfer model.Results from different approaches can support each other and corroborate the same phenomenon.Results of this study are robust and the quantified estimations of radiative forcing are valuable to weather and climate modeling.The method used in this study can be easily applied to data from other surface observatories in order to determine how the climate forcing of the AIV behaves for different cloud regimes.

Figure 1 .
Figure 1.(a) Cloud-top temperature and (b) cloud fraction as a function of aerosol CN concentration.

Figure 2 .
Figure 2. Cloud-top temperature as a function of aerosol CN concentration for deep convective clouds under different (a) vertical velocity (vv), (b) CAPE, (c) precipitable water vapor amount (PWV), and (d) wind shear (ws) conditions.Slopes and correlation coefficients of the different regressions are given.

Figure 3 .
Figure 3. Santa Barbara DISORT Atmospheric Radiative Transfer (SBDART)-modeled flux as a function of satellite-retrieved flux for (a) SW fluxes at the TOA, (b) LW fluxes at the TOA, (c) SW fluxes at the surface, and (d) LW fluxes at the surface.Units are W m −2 .Standard deviation, bias, correlation coefficient, and number of data points (N) are given.

Figure 4 .
Figure 4. Deep convective cloud radiative forcing at the TOA (left-hand panels) and at the surface (right-hand panels) as a function of solar zenith angle for different ranges of CN concentration (a-c, e-g) and as a function of aerosol CN concentration (d, h).

Figure 5 .
Figure 5. Deep convective cloud (a) optical depth, (b) top height, and (c) radiative forcing (SW, LW, and net) as a function of aerosol CN concentration.Solid lines are for convective cores and dotted lines represent anvils.

Table 1 .
Data sets used in the study.

www.atmos-chem-phys.net/14/7113/2014/ Atmos. Chem. Phys., 14, 7113-7124, 2014 budget
of the climate system.The indirect effect of aerosols on clouds is the largest source of all uncertainties with regard to global climate forcing(IPCC, 2013).Small changes in cloud shape, structure, or lifetime can significantly change the local radiative balance.