1 Introduction

Aerosols are important to the energy budget of earth–atmosphere system and various atmospheric chemical–physical processes. Massive use of fossil fuel has made great contribution to the continuous growth of the anthropogenic aerosols since the pre-industrial era. The anthropogenic aerosols usually consist of four major categories: black carbon (BC), sulfate, organic carbon (OC), and nitrate. BC and OC aerosols are emitted from the combustion of carbon-based materials, e.g., coal, biomass, etc. (Bond et al., 2004, 2007). Sulfate aerosols are usually generated in the uses of petroleum fuel and volcano eruptions (Okada, 1983; Haywood and Boucher, 2000). Nitrate aerosols mainly come from the organic decomposition, fossil fuel combustion, and industrial activities (Ohara et al., 2007). Before 1850, most of aerosols were from natural sources. However, human activities have become a major contributor to BC, OC, and sulfate aerosols in recent decades. Therefore, we define BC, OC, and sulfate as anthropogenic aerosols in this study. Each kind of anthropogenic aerosols has unique impacts on atmospheric thermodynamics, hydrological circulation, and many complex climatic processes (Boucher et al., 2013; Myhre et al., 2013).

The external mixing (EM) scheme of aerosols is commonly used in most model simulations. In EM model, aerosol particles are treated as independent single-component spheres. However, lots of observations have confirmed that internally mixed aerosols are common all over the globe (Andreae et al., 1986; Jacobson, 2000). The description of a particle can be very different according to the selected internal mixing model: homogenous or inhomogeneous, spherical or nonspherical, single particle or cluster. However, one thing in common is that the internally mixed particles are formed by at least two different substances. The differences in particle structure between EM and PIM (partial internal mixing) schemes can bring about significant changes to particle optical properties and thus the direct radiative forcing (DRF) of aerosols. It has been found by many studies that the internal mixing of BC and soluble aerosol substances can largely increase the absorption of BC (Jacobson, 2001; Zhou et al., 2013; Zhang et al., 2015a), since the translucent soluble substances covering the BC core can act as a lens that focuses radiation on the absorptive core (Pósfai et al., 2003), especially when the relative humidity is below deliquescence point of the soluble substance (Lesins et al., 2002). The fact that the internal mixing of BC is beneficial to greater positive DRF and atmospheric warming has been proven in many experiments (Chung et al., 2002; Sato et al., 2003; Li et al., 2010).

Internal mixing can also alter the overall hygroscopic properties as well as the number concentration of aerosol particles, leading to changes in the aerosol–cloud interaction. Bauer et al. (2010) reported that the amount of pure soluble aerosols could be reduced after internally mixed with BC, and thus suppressed the generation of low clouds. However, coated BC aerosol can serve as CCNs (cloud condensation nuclei) under specific conditions, which can increase the brightness of low cloud. The quantitative relationship between coating and the timescale for hydrophobic-to-hydrophilic conversion is poorly constrained, and changes with many environmental variables (McMeeking et al., 2011). Therefore, quantitative evaluation of the impacts of internal mixing on aerosol–cloud interaction is still of large uncertainty.

Until now, the differences in climatic effects of anthropogenic aerosols have not been systematically evaluated between external and internal mixing treatments. Zhao et al. (2017) studied the impact of anthropogenic aerosols on aridity by using EM scheme. In this study, we used the aerosol–climate online coupled model of BCC_AGCM2.0_CUACE/Aero to simulate the effective radiative forcing (ERF) due to PIM anthropogenic aerosols and their effects on surface aridity change since the pre-industrial. In order to simulate aerosol–radiation and aerosol–cloud interactions, we improved the aerosol–cloud interaction module of BCC_AGCM2.0_CUACE/Aero, and included a set of detailed optical properties look-up tables of internally mixed aerosols (Zhang et al., 2015a). Therefore, the influences of internal mixing can be simulated under different volume mixing ratios and relative humidity (RH). Here, we further compare the differences in global aridity due to EM and PIM schemes using the aridity index. Section 2 presents the detailed calculation of aerosol optical properties and aerosol hydrophilism, and then describes the test setup and the aerosol–climate online coupled model, BCC_AGCM2.0_CUACE/Aero. In Section 3, we analyze the influences of the mixing treatments of anthropogenic aerosols on their ERF. Section 4 discusses the changes in global aridity and the related climate factors due to EM and PIM schemes. Finally, Section 5 presents the conclusions.

2 Methods and model

BC, sulfate, and OC are considered as “anthropogenic aerosols” in the BCC_AGCM2.0_CUACE/Aero. Most influences of the nature emissions of BC, OC, and sulfate on radiative forcing, and climate has been excluded by subtracting the results of pre-industrial scene from those of present-day scene. Two mixing schemes, EM and PIM, were evaluated in this study. In EM scheme, all of the aerosol particles were considered independently in both the radiative transfer and hydrophilic processes. In PIM scheme, 30% of the anthropogenic aerosols were internally mixed based on the available observational data (Martins et al., 1998; Abel et al., 2003; Deboudt et al., 2010) into BC–sulfate and BC–OC aerosols, while the rest of the aerosols were externally mixed. The concept of ERF (IPCC, 2013) was applied to describe the influences of aerosols on radiation energy budget.

2.1 Methods

Based on the concept of ERF, the calculations can be summarized into two major parts: aerosol–radiation (ERF_ari) and aerosol–cloud (ERF_aci) interactions. The main part in simulating the ERF_ari interaction is detailed calculations of the optical properties of aerosols. In this study, we use “core–shell” model to describe the internally mixed particles. The optical properties were calculated by using Mie theory for externally mixed aerosols, whereas by the coated sphere Mie theory (Bohren and Huffman, 1998) for internally mixed aerosols. Zhang et al. (2015a) have evaluated other internal mixing models (Maxwell–Garnett and Bruggeman equivalent media models) using a single bin radiative transferring model, and found that they had similar impacts on DRF as the core–shell model. The complex refractive indices of dry aerosol substances were adopted from the HITRAN (High Resolution Transmission) (2004) (Rothma et al., 2005) database. The complex refractive indices corresponding to different RHs were calculated according to Eq. (1) (Bohren and Huffman, 1998). The changes in RH do not influence the refractive indices of BC since BC is treated as insoluble aerosol in BCC_AGCM2.0_CUACE/Aero.

${{m = }}{{{m}}_{{\rm w}}}{{ + (}}{{{m}}_{{\rm d}}}{{ - }}{{{m}}_{{\rm w}}}{{)}}\frac{{{{{r}}_{{\rm d}}}^{{3}}}}{{{{{r}}_{{\rm m}}}^{{3}}}},$

(1)

where m is the complex refractive index of a soluble substance after hygroscopic growth; m_w and m_d are the complex refractive indices of water and a dry soluble substance, respectively; r_d and r_m are the equivalent radii of the soluble substance before and after hygroscopic growth, respectively. The hygroscopic growth rates of the soluble substance were obtained by using the к-Köhler theory (Petters and Kreidenweis, 2007). A set of look-up tables of internal mixing aerosol optical properties were created based on Zhang et al. (2015a). The optical properties between each bin of radius, mixing proportion, and humidity were obtained by using linear interpolation in real-time simulation. In the simulation, anthropogenic aerosols were binarily mixed into BC–sulfate and BC–OC aerosols. Both the volume mixing ratio of each substance and the proportion of each mixture were determined by the mass concentrations of anthropogenic aerosols within each grid box in real-time.

The к-Köhler theory was used to simulate the aerosol–cloud interaction. In the к-Köhler theory, the calculations of hygroscopic growth and cloud condensation nucleus activity depend only on a single hygroscopicity parameter, к. In the classic Köhler theory, the CCN activity is calculated by using several aerosol physicochemical properties, i.e., solute mass, molecular weight, bulk density, dissociable ions, and activity coefficient. However, calculation has been greatly simplified in the к-Köhler theory. The calculation of к is based on:

$\kappa { { = }}\frac{{{ {4}}{{ {A}}^{ {3}}}}}{{{ {27}}{{ {D}}_{ {\rm d}}}^{ {3}}{ {\ln}}^{ {2}}}{{ {S}}_{ {\rm c}}}},\quad$

(2)

$\!\!\!\!\!\!\!\!\!\!\!\!\!\!{ {A = }}\frac{{{ {4}}{\sigma _{{ {\rm s/a}}}}{{ {M}}_{ {\rm w}}}}}{{{ {RT}}{\rho _{ {\rm w}}}}},$

(3)

where к is the hygroscopicity parameter, D_d is the dry particle diameter, S_c is the critical supersaturation, σ_s/a is the surface tension of the solution/air interface, M_w is the molecular weight of water, R is the universal gas constant, T is the temperature, and ρ_w is the density of water. After obtaining к, the hygroscopic diameter growth factor (gf) can be calculated by:

$\frac{{\rm RH}}{{\exp \left( {\displaystyle\frac{A}{{{D_{\rm d}}{\rm gf}}}} \right)}} = \frac{{{\rm g}{{\rm f}^3} - 1}}{{{\rm g}{{\rm f}^3} - (1 - \kappa )}},$

(4)

where RH is the relative humidity as a fraction. The overall кvalue for the internally mixed particles can be calculated by:

$\kappa { { = }}\mathop \sum \nolimits_{ {i}} {\varepsilon _{ {i}}}{\kappa _{ {i}}},$

(5)

where ε_i and к_i are the volume fraction and the hygroscopicity parameter of substance i, respectively. The size distribution spectra of internally mixed anthropogenic aerosols were calculated based on the size distribution and the volume mixing ratio of BC in real-time simulation.

2.2 Model description

The general circulation model of BCC_AGCM2.0 (Wu et al., 2010) was developed by the Beijing Climate Center with a T42 horizontal resolution and 26 vertical layers. A new Monte Carlo Independent Column Approximation of McICA (Jing and Zhang, 2012) and a new radiation scheme of BCC_RAD (Zhang et al., 2014) were included in the model with updated aerosol (Wei and Zhang, 2011) and ice cloud (Zhang et al., 2015b) optical properties. BCC_AGCM2.0 was further coupled with the aerosol model of CUACE/Aero (Zhang et al., 2012), to build BCC_AGCM2.0_CUACE/Aero, including five typical aerosol species (BC, OC, sulfate, dust, and sea salt). The concentration of the aerosols was discretized into 12 bins of radii 0.005–20.480 μm. The natural emissions, such as sea salt and soil dust, were calculated online. In this study, the к-Köhler theory was applied into the aerosol–cloud interaction module of BCC_AGCM2.0_CUACE/Aero.

Wu et al. (2010) evaluated the capability of BCC_AGCM2.0 and CAM 3.0 by comparing to observations, and found that BCC_AGCM2.0 can produce major meteorological fields more accurate than CAM3.0, especially the tropical/subtropical wind field, latent heat flux, oceanic sensible heat flux, and precipitation. BCC_AGCM2.0_CUACE/Aero can simulate sulfate, sea salt, and dust generally consistent with those provided by AEROCOM (Aerosol Comparisons between Observations and Models)–MEDIAN, but the ranges and values of BC and OC were lower (Zhang et al., 2012). However, the simulations of BC and OC distributions, cloud properties, and the net radiative forcing at the top of atmosphere have been notably improved after introducing a new cloud microphysical scheme into the model (Wang et al., 2014). The simulated optical depth, single scattering albedo, and asymmetry parameter of total aerosols at 550 nm were generally consistent with AERONET observations (Zhang et al., 2012; Zhao et al., 2014).

2.3 Test setup

Two groups of tests were carried out. Group 1 represented the pre-industrial scene with the emission data of 1850. Tests in group 1 only considered EM due to the low atmospheric burden of anthropogenic aerosols. Group 2 represented the present-day scene with the emission data of 2000. Tests in group 2 considered both PIM and EM. The emissions data of BC, OC, sulfate, and dimethyl sulfide (DMS) were from AeroCom (Nightingale et al., 2000; Bond et al., 2004; van der Werf et al., 2004), and other emissions data were from the Emission Database for Global Atmospheric Research (EDGAR) version 3.2 (Olivier et al., 2001). In each scene, the model was ran for 20 yr (10-yr spin-up), with a prescribed monthly-mean sea surface temperature (SST) and sea ice cover (SI) to calculate the ERF due to anthropogenic aerosols. In addition, the model coupled with the slab ocean model (SOM) was ran for 80 yr (30-yr spin-up) to simulate the climatic features in each scene. A lower bound of cloud droplet number concentration (CDNC) was imposed everywhere as a proxy for natural background aerosols, which were not included in the model (Kirkevåg et al., 2008; Wang et al., 2016).

The results obtained in this study were compared with an earlier study of anthropogenic aerosols conducted by Zhang et al. (2016) using the same model with a similar experimental setup. However, the early study considered neither the lower bound of CDNC (the lack of the lower bound of CDNC would result in a more negative ERF) nor internal mixing, and thus obtained the global annual mean ERF of EM anthropogenic aerosols since 1850 to be –0.30 W m^–2 as ERF_ari and –2.19 W m^–2 as ERF_aci, respectively. The major impacts of EM anthropogenic aerosols on the climate and radiative forcing were similar in both studies. More result validation can be found in Zhang et al. (2016).

3 The effective radiative forcing of PIM and EM anthropogenic aerosols

The new concept of ERF has been proposed in IPCC AR5 to describe the change in the energy balance of earth–climate system due to certain factors. ERF allows the adjustment of all physical variables, except those regarding the ocean and sea ice. The inclusion of these adjustments makes ERF a better indicator of the eventual temperature response (Myhre et al., 2013). Compared with earlier concepts of radiative forcing, ERF performs better in predicting climate responses to forcing factors that have multiple feedbacks, especially for short-lifetime components such as aerosols. The change in the energy balance of the earth–atmosphere system due to anthropogenic aerosols will eventually lead to the change in climate field. Therefore, a brief discussion of ERF is needed before exploring the impacts of anthropogenic aerosols on global aridity.

The global annual mean ERFs of anthropogenic aerosols since the pre-industrial are –1.68 and –1.02 W m^–2 in EM and PIM schemes, respectively. ERF contains various complex feedbacks of the initial changes in radiation flux due to anthropogenic aerosols, and thus shows little corresponding with the geographical distribution of anthropogenic aerosols (compared to early concept of DRF). In both schemes, negative ERF dominates most of the globe and is strongest in the midlatitude Northern Hemisphere due to the high atmospheric burden of sulfate aerosols (Fig. 1). Sulfate aerosol is ideal CCN and is beneficial for generating more clouds, thus causing markedly negative ERF over such areas. The negative ERF is especially strong over Northeast Asia and the North Pacific Ocean where the anthropogenic aerosols emitted from Northeast China are mainly transported to here. Some areas show weak positive ERF, e.g., Sahara Desert, Arabian Peninsula to India, North Alaska, small patches over the oceans, etc., mainly due to the rapid adjustments of the direct radiative effects of BC. The peak of negative ERF appears over the Tibetan Plateau with the peak of positive ERF to the east of it. The negative peak is possibly due to the enhanced water vapor import from northern Indian Ocean (figure omitted, please refer to Zhou et al., 2018), which causes more low cloud as the terrain elevation increases. The positive peak is mostly due to the aerosol–radiation interaction of BC, which is amplified by the rapid change in terrain elevation and high surface albedo of Tibetan Plateau. Compared to EM scheme, the ERF of anthropogenic aerosols has changed obviously in PIM scheme. The negative ERF over the source regions of sulfate and OC aerosols are markedly reduced mainly due to the lower CCN number. The mixture of BC and soluble aerosols is lower in hygroscopicity parameter than the pure soluble aerosols, thus suppresses the generation of clouds, and reduces the cloud albedo. The positive ERF over the oceans and East Siberia are enhanced in PIM scheme as well, which is mostly due to the enhanced rapid adjustments of the direct radiative effects of BC, since the absorption of BC is much stronger in internal mixing than in external mixing. The peaks of both negative and positive ERFs appear over the same regions in two mixing schemes, but are much weaker in PIM scheme. Overall, the PIM scheme can cause positive change in ERF of anthropogenic aerosols over the globe, and reduces the cooling effect of anthropogenic aerosols compared to EM scheme.

Detailed discussion of the impacts of mixing methods on aerosol loading, optical depth as well as the related physical explanations can be found in Zhou et al. (2018).

4 The impacts of anthropogenic aerosols on global aridity 4.1 Calculation of surface aridity degree

In this study, we use the Aridity Index (AI) recommended by the United Nations Environment Programme (UNEP) to describe and classify the surface aridity degrees. Compared to the early used classification standards, AI comprehensively considered the influences of precipitation, energy budget, near surface humidity and wind speed, and temperature on surface aridity. Therefore, AI is more accurate in evaluating the impacts of certain factors on surface aridity in climate state (Kottek et al., 2006; Feng and Fu, 2013; Zhao et al., 2015). The calculation of AI is based on Eqs. (6) and (7):

${\rm E}{{\rm T}_0}{{ = }}\frac{{0{{.408}}\Delta \left( {{{{R}}_ {\rm{n}}}{{ - G}}} \right){{ + }}\gamma \frac{{{{900}}}}{{{T}}}{\mu _{{2}}}\left( {{{{e}}_ {\rm{s}}}{{ - }}{{{e}}_ {\rm{a}}}} \right)}}{{\Delta {{ + }}\gamma \left( {{{1 + 0}}{{.34}}{\mu _{{2}}}} \right)}},$

(6)

${{{\rm AI} = }}\frac{{{P}}}{{{\rm E}{{\rm T}_0}}}.\quad\quad\quad\quad\quad\quad\quad\quad\quad\quad\quad\;$

(7)

In Eq. (6), ET₀ (mm day^–1) is the potential evaporation capacity, representing the evaporation demand on the surface. R_n (MJ m^–2 day^–1) is the net radiation flux on the surface; G (MJ m^–2 day^–1) is the soil heat flux density; T (°C), μ₂ (m s^–1), e_s (kPa), and e_a (kPa) are the temperature, wind speed, saturated water vapor pressure, and actual water vapor pressure at 2 m above the surface; γ (kPa °C^–1) is a psychrometric constant; △ (kPa °C^–1) is the slope vapor-pressure curve. In Eq. (7), P (mm day^–1) is the precipitation rate, and AI represents the water supply−demand ratio, which is proportional to the regional moisture level. According to AI, surface aridity degree can be classified into 6 aridity levels: hyper-arid (AI < 0.05), arid (0.05 ≤ AI < 0.20), semi-arid (0.2 ≤ AI < 0.5), dry sub-humid (0.50 ≤ AI < 0.65), sub-humid (0.65 ≤ AI < 1), and humid (AI ≥ 1) ( Fig. 2). To reduce the influences of the systemic error of model on outcomes, we use observational data to correct the results corresponding to present-day (Dai, 2011), and reserve the differences in the relative climate factors to analyze the influences of different mixing schemes of anthropogenic aerosols on global aridity since the pre-industrial.

4.2 Influences of anthropogenic aerosols on AI-related climate factors

The land annual mean atmospheric temperature at 2 m above the surface (TSA) has decreased by 2.18 and 1.61 K since 1850 in EM and PIM schemes, respectively. The decrease is especially notable over the mid and high latitudes of the Northern Hemisphere due to the strong negative ERF produced by the highly concentrated sulfate aerosol (Fig. 3, upper panels). PIM scheme can largely reduce the negative ERF of anthropogenic aerosols, and suppresses the decrease in TSA compared to EM scheme, especially over the high latitudes of the Northern Hemisphere.

The influences of mixing schemes on RH are complex, since the change in RH is a comprehensive result of the variations of temperature, precipitation, evaporation, circulation, etc. The land annual mean RHs have increased by 1.03% and 0.82% since the pre-industrial in EM and PIM schemes, respectively. The increases in RH are mainly due to the lower surface temperature. In both mixing schemes, increases in RH are apparent over most of the globe, and are most notable over central Asia and the Tibetan Plateau (Fig. 3, middle panels). Weak decreases in RH are apparent over the narrow region south to the Sahara Desert and northern South America. Compared to EM scheme, PIM anthropogenic aerosols cause increases in RH mainly concentrated over the northeast part of China, whereas causing decreases in RH over the northeast part of the USA, the northwest part of South America, the Arabian Peninsula, and the northeast part of Africa, North India, and Australia. The decreases in RH due to PIM scheme are mostly due to the higher surface temperature, which raises the saturation vapor pressure.

Anthropogenic aerosols can also alter wind speed at 2 m above the surface (μ₂) through various feedbacks (Fig. 3, bottom panels). Throughout the industrial era, land annual mean μ₂ has slightly increased by 0.07 and 0.06 m s^–2 in EM and PIM schemes, respectively. How-ever, the geographic variations of μ₂ are large. Marked increase in μ₂ can be found over the midlatitudinal Eur-asia, and the increase reaches peak over the Tibetan Plateau. Strong increases in μ₂ are also apparent over the narrow region south to the Sahara Desert, central North America, Greenland, and eastern Siberia. Compared to the EM scheme, decreases in μ₂ mostly appear over the midlatitudal Northern Hemisphere and Greenland, while increases in μ₂ mostly appear over northern Canada, Africa, central Europe to central Siberia, Arabian Peninsula, and Iran Plateau in PIM scheme.

Anthropogenic aerosols cause general reduction in precipitation over land mainly due to two reasons: first, the reduction in surface temperature suppresses water evaporation; second, increased anthropogenic aerosol emission provides more CCNs, and thus reduces the cloud droplets radii. Anthropogenic aerosols can also alter the precipitation distribution by changing the large scale circulation, but the influences are obvious only over the oceans (this change cannot be seen in Fig. 4). The land annual mean precipitation rate (P) has decreased by 0.12 and 0.09 mm day^–1 since the pre-industrial in EM and PIM schemes, respectively. Decreases in precipitation are apparent over most land areas in the Northern Hemisphere, and are relatively large over the Tibetan Plateau, central and northern China, central India, and Southeast Asia (Fig. 4, upper panels). The strongest decrease was shown over the Panama region since the water vapor import from the Atlantic Ocean within 0−30°N was considerably reduced. The sole notable increase in precipitation is over northeastern Brazil Plateau, since the enhancement in sea–land breeze provides a better water vapor supply for terrain rainfall. Compared to EM scheme, PIM anthropogenic aerosols show less inhibition on precipitation over most land areas due to higher evaporation. Notable increases can be found over Canada, Europe, northeastern China, and central Africa.

Due to the increase in cloud amount and high scattering capacity of anthropogenic aerosols, the land annual mean net radiation fluxes on the surface (R_n) have decreased by 5.06 and 3.90 W m^–2 since the pre-industrial in EM and PIM schemes, respectively. The changes in R_n are geographically similar to the change in ERF, and show globally decreases in two mixing schemes (Fig. 4, middle panels). In both mixing schemes, major decreases in R_n are concentrated over the mid and high latitudes of the Northern Hemisphere due to sulfate emission, and relative strong decreases are apparent over South America and central Africa due to OC emission. Compared to EM scheme, R_n is largely increased in PIM scheme due to lower cloud amount/albedo and lower aerosol single scattering albedo. The increase is most obvious over western Siberia, central Asia, southern Canada, and northern South America, and the peak of increase appears over the Tibetan Plateau.

As a comprehensive result of decrease inR_n and TSA and simultaneous increase in RH, the potential evaporation capacity is inhibited over most land areas. The land annual mean potential evaporation capacity (ET₀) has reduced by 0.21 and 0.14 mm day^–1 since the pre-industrial in EM and PIM schemes, respectively (Fig. 4, bottom panels). In both mixing schemes, most land areas show decreases in ET₀, especially northwestern Africa, Iran Plateau, and southern China. Strong increases in ET₀ are concentrated over the narrow region south to the Sahara Desert, southern Arabian Peninsula, and southern Iran Plateau. Compared to EM scheme, PIM anthropogenic aerosols can cause increases in ET₀ over most land areas, with strong increases concentrated over the Arabian Peninsula, North Africa, and northern South America.

4.3 Influences of anthropogenic aerosols on AI and surface aridity classification

The distributions of AIs have markedly changed since the pre-industrial, and the differences in AI between mixing schemes are distinct (Fig. 5). Increases in AIs since the pre-industrial are notable over most land areas in both mixing schemes, especially over the mid and high latitudes of the Northern Hemisphere. The main reason for the global humidification is the decrease in ET₀. The land annual mean AIs have increased by 0.12 and 0.06 since the pre-industrial in EM and PIM schemes, respectively. In both mixing schemes, AI increases markedly over northeastern Canada, western Siberia, and the Tibetan Plateau. Weak decreases in AI are concentrated over northwestern China, Southeast Asia, central Africa, Mexico, and northern South America. Large decreases in AI can be found within the Arctic Circle due to high sensitivity to climate change of the North Pole. However, AI value is very high over this region, and the decreases do not indicate actual aridification.

Decreases in AI are distinctly larger over the mid and high latitudes of the Northern Hemisphere in PIM scheme than in EM scheme, mostly due to the higher R_n and TSA. PIM anthropogenic aerosols can reduce AI over northeastern South America, Sahara Desert, Arabian Peninsula, and Iran Plateau, causing further intensification on the current aridity of these regions. Compared to EM scheme, most areas of China, Southeast Asia, central Africa, and Mexico show AI increases in PIM scheme, which are mostly due to the increases in precipitation and RH, also the decreases in near surface wind speed.

The area of humid and sub-humid areas has markedly expanded due to anthropogenic aerosols since the pre-industrial, along with certain expansion of arid areas over some regions (Fig. 6). The influences of EM and PIM schemes on surface aridity are generally similar. Compared to pre-industrial, large expansions in humid and sub-humid areas are apparent over the north shore of Mediterranean, central and West Asia, central Africa, and central South America. Semi-arid area is the transition zone between arid and humid areas, and thus provides a good indication for the change in aridity. According to Huang et al. (2015) and Guan et al. (2016), the semi-arid and semi-wet regions (especially the area with high population density and population change) are also responsible for more than 53% of the anthropogenic dust aerosol emission. Many grid points have changed from arid and semi-arid to sub-humid in Iran Plateau, central Asia, and Australia, which indicates considerable shrink in arid area and possible reduction in local dust aerosol emission. In both mixing schemes, changes from sub-humid to semi-arid are apparent mostly over northern China and the Tibetan Plateau, indicating notable aridification. Similar aridification also appears over western USA and Mexico. In general, throughout the industrial era, anthropogenic aerosols have caused mostly humidification over large part of the globe, but also have caused notable aridification over large part of China.

Compared to EM scheme, PIM scheme can cause notable differences in the surface aridity degrees. Multiple grid points over northeastern China and Tibetan Plateau originally classified as “semi-arid” and “dry sub-humid” in EM scheme show humidification in PIM scheme. Humidification is also apparent over central Asia, northern Mexico, central Africa, and central South America, mainly due to decreased near surface wind speed and increased precipitation. However, PIM scheme can cause great aridification over Arabian Peninsula, East Africa, northern Sahara Desert, and Australia. Weak aridification is apparent over the northern shore of the Mediterranean and western South America. In general, compared to EM scheme, PIM scheme can markedly abbreviate the aridification over northern China, the Tibetan Plateau, and northern Mexico, but intensifies the aridity over Arabian Peninsula and East Africa while weakening the humidification over western Australia.

5 Conclusions

We used the aerosol–climate online coupled model, BCC_AGCM2.0_CUACE/Aero, to simulate the ERF due to PIM anthropogenic aerosols and their impacts on global aridity since the pre-industrial era and compared the results with those of EM ones in detail. The core–shell model and the к-Köhler theory were used to calculate the optical properties and the hygroscopic capacity of internally mixed aerosols, respectively.

The global annual mean ERFs of EM and PIM anthropogenic aerosols since the pre-industrial are –1.68 and –1.02 W m^–2, respectively. PIM scheme causes markedly positive changes in ERF on a global scale due to the higher absorption and lower hygroscopicity parameter. The positive change in ERF due to PIM scheme is the strongest over the mid and high latitudes of the Northern Hemisphere. The changes in the climatic factors related to aridity are also discussed. Since the pre-industrial, the near surface temperature and the net radiation flux have decreased, whereas the near surface wind speed and RH have increased over most land areas in both mixing schemes. These changes lead to a globally decrease in potential evaporation capacity. However, PIM scheme can markedly increase the near surface temperature and net radiation flux on the surface over most land areas compared to EM scheme, resulting in globally increase in potential evaporation capacity, especially over the arid and semi-arid areas (e.g. Sahara Desert, Arabian Peninsula, Iran Plateau, etc.). AI, as a factor representing the regional water supply–demand ratio, has increased over most of the land area since the pre-industrial in both mixing schemes, indicating a general surface humidification due to anthropogenic aerosols. According to surface aridity classification based on AI, large areas in central Asia, East Europe, central South America, and central Africa have changed to more humid climate states since the pre-industrial. However, notable aridification is apparent over northern China and the Tibetan Plateau. Compared to EM scheme, the aridification is notably alleviated over China but greatly intensified over Arabian Peninsula, Iran Plateau, and East Africa in PIM scheme.