Zhenxi Zhang , Wen Zhou

a College of Energy and Power Engineering, Inner Mongolia University of Technology, Hohhot, China

b School of Energy and Environment, City University of Hong Kong, Hong Kong, Tat Chee Avenue, Kowloon, Hong Kong, China

Keywords:Dust aerosol Tropical cyclone Eastern pacific Semi-direct effect Potential intensity theory

ABSTRACT The thermodynamic state of the tropical atmosphere plays an important role in the development of tropical cyclone (TC) intensity. This study reports new results that demonstrate a negative association between eastern Pacific TC intensity in offshore regions and dust aerosol optical depth (AOD) for the years 1980—2019. At the same time, a negative response of cloud water in the lower troposphere to dust AOD is reported by analyzing MERRA-2 reanalysis data and GCM simulations from CMIP6. This response can be explained by the dust semi-direct effect,in which dust aerosols absorb solar radiation, promoting the evaporation of clouds. In principle, this aerosoldriven vaporization modification could affect the enthalpy of the air surrounding a TC. Using potential intensity theory, the authors demonstrate that change in TC intensity related to dust AOD conditions is a consequence of the anomalous enthalpy of the air surrounding a TC caused by the dust semi-direct effect.

1. Introduction

A tropical cyclone (TC) is a nonfrontal low-pressure weather system that forms over warm tropical or subtropical oceans and has wellorganized convection and circulation ( Landsea and Gray, 1992 ). TC intensity is usually measured in terms of the maximum speed of tangential winds in the lower troposphere, within 50 km of the center( Ooyama, 1982 ). TC-induced increases in local sea level ( Anthes, 1982 ),vertical mixing of heat and nutrients in the ocean, and horizontal oceanic heat transport ( Emanuel, 2001 ; Jansen and Ferrari, 2009 ) are all related to TC intensity. TC intensity is controlled by several environmental influential meteorological factors, e.g., sea surface temperature(SST), vertical shear of horizontal wind, vorticity, disturbances in the upper troposphere, and humidity ( Bosart et al., 1999 ; Emanuel, 2013 ;Lin et al., 2015 ).

Atmospheric aerosols may impact TCs ( Zhang et al., 2007 ;Rosenfeld et al., 2012 ; Dunstone et al., 2013 ) because aerosols have profound impacts on atmospheric radiation and thermodynamics through direct radiative effects, indirect effects, and other cloud-mediated effects ( Andreae and Rosenfeld, 2008 ). Both modeling and observational analysis have indicated that aerosols induce a decrease in TC intensity( Wang et al., 2014 ; Zhao et al., 2018 ).

As absorbing aerosols, mineral dust absorbs solar radiation (known as radiative forcing), which induces a reduction in surface temperature ( Bangert et al., 2012 ) and enhances cloud evaporation (known as the semi-direct effect) ( Huang et al., 2006 ; Lau et al., 2009 ). On the other hand, mineral dust can act as effective cloud condensation nuclei( Koehler et al., 2009 ; Karydis et al., 2011 ) and ice nuclei ( Chen et al.,1998 ; Hoose and Mohler, 2012 ; Cziczo et al., 2013 ), known as the microphysical effect. Through long-range transport, dust aerosols can affect a distant area of the ocean ( Ginoux et al., 2001 ). Sun and Zhao (2020) investigated the Saharan dust effect on the North Atlantic environmental influential meteorological factors that affect TC development. In this study, we investigate the influence of dust aerosols on TC intensity over the tropical eastern Pacific.

2. Data and methods

The methods of this study includes two parts: (1) to study the longterm statistical relationship between dust aerosol optical depth (AOD)and TC intensity; and (2) to find a physical mechanism to account for the above relationship between dust AOD and TC intensity.

TC track data were obtained from NOAA’s Tropical Prediction Center (1980—2018) and the United States Central Pacific Hurricane Center(2019). The six-hour recorded maximum wind speed (ms), track (longitude (0°—360°), latitude (°)), and time (month, day of the month, and hour (GMT)) are used in this study.

Fig. 1. Comparison of TC intensity (units: m s?1 ) at 4° grid resolution between the eight-year strongest and weakest dust conditions. The images represent annual average values. Purple lines are the dust AOD contours of 0.01, 0.012, 0.016, 0.02, and 0.03, averaged for the corresponding eight years. The solid black outline shows the negative response region of TC intensity to dust.

MERRA is a satellite-era atmospheric reanalysis produced with the GEOS atmospheric model and data assimilation system ( Rienecker et al.,2008 ). Based on improvements in precipitation and water vapor climatology, MERRA focuses on historical analyses of the global hydrological cycle ( Rienecker et al., 2011 ). MERRA-2 is the new version of MERRA and covers the period from January 1980 to the present.Aerosol fields in MERRA-2, including AOD and other aerosol properties, are produced with the aerosol module of the GOCART model( Chin et al., 2002 ; Colarco et al., 2010 ) and assimilation of satelliteobserved AOD and ground-based measurements of AOD ( Randles et al.,2017 ), which are used to investigate aerosol—weather and aerosol—climate interactions ( Bellouin et al., 2013 ; Reale et al., 2014 ). The AOD observations in MERRA-2 include AVHRR (from January 1980 to August 2002), AERONET (station-dependent, from January 1999 to October 2014 ), MISR (from February 2000 to June 2014), and MODIS(Terra: March 2000 onward; Aqua: August 2002 onward) ( Randles et al.,2017 ). Bias correlation of these observations is implemented using a bias-correction approach that involves cloud screening and homogenization of the observing system by means of a neural net retrieval that translates cloud-cleared observed radiances into AERONET-calibrated AOD ( Randles et al., 2017 ).

This study uses the monthly MERRA-2 dust AOD, cloud liquid water mixing ratio, moist heating rate, and liquid water path (LWP; the vertical integration of liquid water in the air column) from 1980 to 2019, with a spatial resolution of 0.625°longitude and 0.5°latitude.

3. Analysis and results

3.1. Association patterns of TCs with dust

Dust AOD in the tropical eastern Pacific occurs mainly over offshore regions. Therefore, dust AOD is averaged over the eastern Pacific region of 5°—15°N and 85°—110°W, and eight years with the strongest and weakest averaged dust AOD are selected during 1980—2019 to construct the strongest and weakest dust conditions, respectively. Fig. 1 shows the regions with TC intensity statistics (by summing the total TC intensity in a 4°grid cell) in the strongest and weakest dust conditions. A comparison between the strong and weak dust conditions reveals the impact of dust AOD on TC activity. Corresponding to strong dust AOD, TC intensity in the tropical eastern Pacific offshore regions becomes weaker compared to weak dust AOD, and the impact area in July, August, and September is larger than that in June. In July and September, accompanied by the strong dust AOD, TC tracks along the shoreline and adjacent region decrease.

Although Fig. 1 presents the situation under extreme dust AOD conditions, the long-term statistical relationship between TC intensity and dust AOD is not clear. TC intensity in eastern Pacific offshore regions during the 40 years from 1980 to 2019 is divided into eight groups based on rankings of the mean dust AOD over the tropical eastern Pacific.Using the group mean data, we present a series of scatterplots of eastern Pacific TC intensity against summer dust AOD ( Fig. 2 ). These show an anticorrelation, which further demonstrates that there are fewer,weaker TCs in the years with stronger dust AOD, and more numerous and stronger TCs in the years with weaker dust AOD. Moreover, all the scatterplots in June, July, and August indicate that this association of TC intensity with dust AOD becomes evident for the strong group mean dust AOD.

Fig. 2. A series of scatterplots grouping the mean TC intensity (units: m s ? 1 ) exponent against dust AOD (on a logarithmic scale). The 40-year (from 1980 to 2019)TC intensity sum (on a logarithmic scale) over the region of 8°—32°N and 82°—106°W is divided into eight groups based on rankings of the mean dust AOD over the eastern Pacific (5°—15°N, 85°—110°W). Ranking of the dust AOD in each month (June, July, August, and September) is independent. Each group includes five years.The boldface lines in the panels indicate the linear fit to the data, and R is the linear correlation coefficient.

The strong correlation of TC intensity with dust AOD does not necessarily mean that dust AOD itself is the dominant mechanism responsible for the correlation. TC intensity depends on TC energy. According to the theory of potential intensity developed by Emanuel (1986 , 1995 ,1997 , 2012) , the energy of a TC comes from its Carnot heating engine,which acquires entropy from ocean—air heat and moisture fluxes and ultimately releases this entropy in the much lower temperatures of the tropopause. High SST can produce more evaporation from the surface,inducing larger ocean—air heat and moisture fluxes. Dust aerosols can cool the surface through direct radiative forcing ( Cavazos et al., 2009 ).However, dust AOD in the tropical eastern Pacific offshore regions has no obvious negative correlation with SST (figure not shown). By condensing moisture into cloud and precipitation, the Carnot heating engine in a TC converts latent heat to sensible heat to provide the energy of a TC ( Ooyama, 1982 ). Therefore, the maintenance of a TC apparently depends on the condensation of moisture into cloud. To address this issue, we further estimated the long-term statistical relationship between dust AOD and atmospheric moisture.

Fig. 3. The 40-year (1980—2019) average of cloud liquid water mixing ratio (CLWMR; at 700 hPa; units: 10?5 kg kg?1 ) (left-hand column), moist heating rate (MHR;at 700 hPa; units: 10?4 K s?1 ) (middle column), and LWP (units: kg m?2 ) (right-hand column). The green and red lines are the negative and positive correlation coefficient contours for the correlations between the above three parameters and dust AOD at each grid point.

3.2. Impact of dust on cloud water

3.2.1.

Correlation

analysis

The dust AOD in tropical eastern Pacific offshore regions between 5°N and 15°N is produced mainly by the westward transport of the Saharan dust plume that crosses the tropical Atlantic from West Africa.Attributed to dust radiative heating ( Carlson and Benjamin, 1980 ;Alpert et al., 1998 ; Zhu et al., 2007 ; Wong et al., 2009 ), the transport of the Saharan dust plume is accompanied by significant warming between 900 and 600 hPa ( Wilcox et al., 2010 ). Therefore, the atmosphere at the altitude of 700 hPa is chosen for analysis first. The climatology of cloud liquid water mixing ratio (at 700 hPa) and its interannual correlation with dust AOD (with 0.05 statistical significance) at each grid point during 1980—2019 are shown in Fig. 3 (left-hand column). Cloud liquid water (at 700 hPa) is concentrated in the intertropical convergence zone (ITCZ) at 10°N and presents an evident negative response to dust AOD in tropical eastern Pacific offshore regions. This negative response becomes stronger in August and September, with a correlation coeffi-cient of less than ? 0.5, appearing in the offshore regions between 5°N and 15°N. This result can be explained by the dust semi-direct effect.Dust aerosols absorb solar radiation, which promotes the evaporation and reduction of clouds. On the other hand, cloud liquid water (at 700 hPa) also presents a positive response to dust AOD, with a small and sparse impact area, which can be explained by the dust microphysical effect.

The solar radiation absorbed by dust aerosols can heat the atmosphere and enhance moisture evaporation, but this cannot be clearly displayed by the radiative heating rate because clouds can absorb solar radiation as well as heat the atmosphere ( IPCC, 2001 ). The decrease in water clouds resulting from the dust semi-direct effect weakens the heating of the atmosphere, which is opposite to the heating effect of dust aerosols. To accurately investigate the influence of dust aerosols on atmospheric moisture, the moist heating rate is applied in this study.Fig. 3 (middle column) shows the climatology of the moist heating rate(at 700 hPa) and the interannual correlation between the monthly dust AOD and moist heating rate at each grid point during 1980—2019. A positive moist heating rate indicates the condensation of atmospheric moisture into clouds and the release of latent heat, which heats the atmosphere. Climatologically, the moist heating rate (at 700 hPa) in the ITCZ at 10°N is larger than that in other marine regions, which coincides with the distribution of cloud liquid water. The correlation between dust AOD and the moist heating rate (at 700 hPa) shows that strengthening dust AOD suppresses the moist heating rate in tropical eastern Pacific offshore regions. Although there is a positive correlation between dust AOD and moist heating rate, the corresponding area is very small compared with the areas of negative correlation. Combined with the influence of dust on cloud liquid water, the evaporation of cloud liquid water caused by the dust semi-direct effect absorbs heat and produces a cooling effect on the atmosphere.

Fig. 4. The difference (piClim-2xdust simulation result minus piClim-control simulation result) in cloud liquid water mixing ratio (CLWMR; at 700 hPa; units:10?6 kg kg?1 ) (left-hand column), moist heating rate (MHR; at 700 hPa; units: 10?6 K s?1 ) (middle column), and LWP (units: 10?2 kg m?2 ) (right-hand column). All the simulation results are the 30-year average of 1850—79.

Fig. 3 (right-hand column) shows the 40-year (1980—2019) average of LWP during the summer. The most pronounced feature of the LWP distribution is the maximum appearing along the ITCZ between 130°W and 150°W and the minimum appearing offshore of Baja California. The LWP over the tropical eastern Pacific offshore regions consistently shows negative correlations with dust AOD from June to September, and this association of LWP with dust AOD becomes the strongest in August because marine areas with a significant negative correlation (less than ? 0.5) are larger in August than in other months. The above analysis reveals that the integration of cloud liquid water in the atmosphere presents a negative response to dust aerosols over the tropical eastern Pacific offshore regions. Obviously, this is not caused by the dust microphysical effect,which promotes condensation in the atmosphere and theoretically produces a positive response. It is also not caused by the weakened SST induced by the dust radiative forcing, because dust AOD has no obvious negative correlation with SST. The reasonable explanation for the negative response of LWP to dust AOD is the semi-direct effect of dust aerosols.

The dust loading over the tropical eastern Pacific is smaller than that over the North Atlantic. A lot of dust aerosols and warm and dry air mass over the Sahara Desert are transported westward from West Africa to the North Atlantic when easterly trade winds pass over the Sahara Desert. Heavy dust loading over the North Atlantic affects not only humidity, but also SST, and even the thermodynamics such as convection and vorticity ( Sun and Zhao, 2020 ).

3.2.2. GCM model simulations

To understand the effect of dust aerosols on atmospheric moisture and to verify the relationships between dust AOD and atmospheric moisture obtained from the analysis of MERRA-2 data, GCM simulations from CMIP6 are applied in this study. The Aerosol Chemistry Model Intercomparison Project (AerChemMIP) is one of the CMIP6-endorsed MIPs and focuses on understanding and quantifying the contributions of atmospheric composition to climate change ( Collins et al., 2017 ). We selected the piClim-control and piClim-2xdust experiments in AerChem-MIP and compared their simulation results. The piClim-control experiment is the pre-industrial control simulation during the period from 1850 to 1879, while the piClim-2xdust experiment parallels the piClimcontrol experiment except that the emission flux of dust aerosols is doubled ( Collins et al., 2017 ).

Fig. 4 presents the differences in atmospheric moisture over the tropical eastern Pacific between piClim-control and piClim-2xdust. Accompanied by the increase in dust aerosols, change in cloud liquid water (at 700 hPa), including an increase and reduction, occurs mainly along the ITCZ and in the tropical eastern Pacific offshore regions. An increase in cloud liquid water is associated with the dust microphysical effect, which promotes cloud condensation, while a reduction in cloud liquid water is associated with the dust semi-direct effect, which promotes cloud evaporation. Over the eastern Pacific (5°—15°N, 85°—110°W), where dust AOD is calculated ( Fig. 2 ), cloud liquid water (at 700 hPa) undergoes a general reduction, which is most pronounced in September.

The difference in moist heating rate (at 700 hPa) between piClimcontrol and piClim-2xdust is shown in Fig. 4 (middle column). The moist heating rate is calculated by using the heating rate due to stratiform cloud and boundary layer mixing, which includes sources and sinks from parameterized physics of the boundary layer and stratiform cloud condensation and evaporation. However, the altitude of 700 hPa is already above the boundary layer. The calculated moist heating rate (at 700 hPa) is caused mainly by the condensation and evaporation of stratiform cloud. With increasing dust aerosols, the moist heating rate (at 700 hPa) shows a reduction pattern over the tropical eastern Pacific off-shore regions, reflecting a cooling effect on the atmosphere due to the evaporation of cloud water, which absorbs heat.

For the integration of cloud liquid water in the atmosphere, the reduction and increase of LWP, along the ITCZ and in the tropical eastern Pacific offshore regions when dust aerosols increase, are similar to that of cloud liquid water (at 700 hPa), because water clouds exist mainly in the middle and lower troposphere.

The above features of atmospheric moisture corresponding to heavy dust loading are consistent with the conclusions from the MERRA-2 data analysis ( Fig. 3 ), verifying the impact of dust on atmospheric moisture through the semi-direct effect.

Fig. 5 shows the vertical profile of air temperature difference averaged over the tropical eastern Pacific (5°—15°N, 85°—110°W) between piClim-2xdust and piClim-control simulations. On average during the June—July—August—September season, the cool temperatures are associated with the strengthening dust loading. According to the analysis of moist heating rate ( Figs. 3 and 4 ), the cool temperature is attributable to the cloud evaporation induced by the dust semi-direct effect, which converts the solar radiation absorbed by dust into latent heat. The dust effect in the tropical eastern Pacific is different from that in the eastern tropical Atlantic where dust loading is very large and cloud water is very small. Therefore, cloud evaporation consumes only a small part of the solar radiation absorbed by dust, while its majority is used to heat the atmosphere, inducing a significant warming of the lower troposphere accompanied by the transport of Saharan dust ( Wilcox et al.,2010 ; Sun and Zhao, 2020 ).

Fig. 5. Vertical profile of air temperature difference (units: K) (piClim-2xdust simulation result minus piClim-control simulation result) averaged over the tropical eastern Pacific (5°—15°N, 85°—110°W). All the simulation results are the 30-year average of 1850—79.

4. Discussion and conclusions

The net heating obtained by the Carnot heating engine is used to balance surface layer frictional dissipation in a steady TC; therefore the TC’s wind is driven by the thermodynamic energy produced by the Carnot heating engine. Based on this theory, the maximum wind speed is determined by the following equation ( Kowaleski and Evans, 2016 ):

where | V MAX | represents the maximum wind speed; C K and C D are the dimensionless enthalpy transfer (controlling enthalpy fluxes from the ocean) and drag coefficients, respectively; and kand kare the enthalpies of the saturated air at the sea surface, and the air surrounding a TC and in the lower troposphere, respectively. The subscript M indicates that the calculation of the right-hand side of Eq. (1) should be near the radius of maximum winds; ε is the Carnot cycle efficiency and depends on the SST ( T) and the entropy-weighted mean outflow temperature( T o ), because of its definition:

The enthalpy per unit mass of air ( k ) is expressed as( Emanuel, 1995 )

where c pd is the heat capacity of dry air at constant pressure, c l is the heat capacity of liquid water, T is the absolute temperature, and Lis the latent heat of vaporization. The specific humidity q is defined as

where m d and m w are the masses of dry air and water vapor in a given volume, respectively, and ρa(bǔ)nd ρa(bǔ)re the densities of dry air and water vapor, respectively.

By combining Eq. (4) and the findings from Figs. 3 and 4 , we can conclude that dust aerosols can enhance the specific humidity ( q ) of the air in the lower troposphere by converting moisture into water vapor and strengthening the water vapor density ( ρ) through the semi-direct effect. On the other hand, enthalpy ( k ) can be expressed as a function of specific humidity ( q ) by rewriting Eq. (3) as

In Eq. (5) , the value of ( c? c) T is negative, and its order of magnitude is 100 if c pd and c l are in units of kJ kgK. The order of magnitude of L(kJ kg) is 1000. Therefore, the coefficient of q in Eq. (5) is positive, and k will linearly increase with q . In principle, Eq. (5) demonstrates that the dust-aerosol-driven enhancement of specific humidity( q ) could induce an increase in the enthalpy of near-surface air ( k).This could further cause the suppression of TC intensity and even decrease the number of TCs along the coast, as shown in Figs. 1 and 2 ,according to Eq. (1) derived from potential intensity theory.

Funding

This work was supported by National Natural Science Foundation of China [grant numbers 41675062 and 41375096 ], and the Research Grants Council of the Hong Kong Special Administrative Region, China[project numbers CityU 11306417 and 11335316 ].

Acknowledgments

The MERRA-2 data were obtained from the Goddard Earth Sciences Data and Information Services Center. The archive of CMIP6 output is accessible from https://esgf-node.llnl.gov/projects/cmip6/ .The eastern Pacific TC track data (19802018) used in this study are from a global TC dataset archived by the Massachusetts Institute of Technology as a related resource for the open course “Tropical Meteorology ”( ftp://texmex.mit.edu/pub/emanuel/HURR/tracks/ ). In this dataset, eastern Pacific TC track data were obtained from NOAA’s Tropical Prediction Center.