Possible impact of North Atlantic sea surface temperature on decadal variability of dust activity in Gobi Desert

Semiarid to arid regions located in north of China are known as one of the largest sources of dust aerosols in the globe. Previous studies demonstrated direct and indirect effects of atmospheric dust loading on climate. The surface and meteorological properties are known to be affecting dust entrainment process. In this study, we found decadal variations of dust occurrence in Gobi Desert with the period of ∼24 years, utilizing the data acquired at the surface meteorological stations. An analysis of reanalysis datasets shows significant correlations between February North Atlantic Sea Surface Temperature (NASST) and precipitation in Gobi Desert and Mongolia in the following summer, causing a decadal variation of dust event frequency in the subsequent spring over the region. Strong time-lagged correlation is shown to be linked to an atmospheric wave train pattern that initiated from the NASST region, affecting large-scale circulation, ultimately causing surface drought over Gobi Desert.


Introduction
Mineral dust aerosols are known for their substantial impact on climate.It directly scatters and absorbs solar and terrestrial radiation (Tegen et al 1996), potentially amplifying dust storms through feedback processes (Chen et al 2023).It also indirectly affects radiation by functioning as a cloud condensation and ice nuclei (Karydis et al 2011, Huang et al 2014).Large uncertainties still exist in quantifying the radiative effect of mineral dust aerosols (Chen et al 2017a; Intergovernmental Panel on Climate Change Sixth Assessment Report 2021).
East Asian deserts, including Gobi Desert, are one of the largest sources of dust aerosols.Dust from deserts located in northern China frequently experiences long-range transport (Huang et al 2008, Chen et al 2017b, 2017c, Liu et al 2019).Much attention has been paid to air quality hazards among nearby cities, including Beijing (Liu et al 2014), Seoul (Park et al 2011) and Tokyo (Mori et al 2003), associated with desert dusts from Gobi Desert, traveling in great distances.Accordingly, understanding the trend of dust event frequency in Gobi Desert show importance, especially under the global warming situation.According to Intergovernmental Panel on Climate Change (IPCC) report, the aridity will increase in some of desert areas (including Gobi Desert) with high confidence, under the climate change, increasing threat of dust storms (IPCC Special Report on Climate Change and Land 2019).
Many factors are important for the deflation of soil particles to the air, including strong wind speed, precipitation, and land surface properties (Liu et al 2004, Zou and Zhai 2004, Jugder et al 2011).Dust events are the most frequent in spring season.Outbreaks of dust are well coinciding with an annual cycle of frequency of strong wind events over the Gobi Desert region (Park andIn 2003, Shao andDong 2006).Whereas, soil moisture is negatively coinciding with the intra-seasonal variation of dust events.Summer, particularly August, is the least frequent dust event season.This is because summer has strong precipitation compared to other seasons, as large particles such as dust particles go through wet removal process and increased soil moisture increases the erosion threshold of wind speed by strengthening the soil cohesion (Park and In 2003).
Previous studies have shown that the interannual variability and trend of Gobi Desert dust activity is shaped by climatic factors (Park and In 2003;Liu et al 2004;Jugder et al 2011;Shang and Liu 2020, Liu et al 2022, Wu et al 2022).Of all the climatic factors, surface wind speed as well as total precipitation are considered as the most important meteorological factors controlling variability and trends of the dust activity (Liu et al 2004, Zou and Zhai 2004, Kurosaki and Mikami 2007, Wu et al 2022).Vegetation cover plays a secondary but locally important role especially in desert regions experiencing rapid change in land cover (Wang et al 2023).In contrast to the extensive knowledge on the interannual variability of dust activity, investigations into multidecadal variability within Gobi Desert have remained relatively scarce.Numerous studies have examined the long-term trends of dust storm frequencies over China with a focus on expansive drylands such as Taklimakan Desert and Tibetan Plateau, rather than exclusively on Gobi Desert (Qian et al 2002;Zhu et al 2008, Kang et al 2016;Meng et al 2020, Shang and Liu 2020, Liu et al 2022).
Goal in this paper is to learn the long-term trend of dust event frequency in Gobi Desert including recent years, and to suggest possible causes for the variability, considering factors ranging from local sources to teleconnections.The primary focus of this study is the long-term predictability of dust frequency in Gobi Desert through the early detection of changes in a key precursor.Rest of the paper is outlined as follows: Data and method are presented in section 2. In section 3, dust event frequency is evaluated against AErosol RObotic NETwork (AERONET) optical property data.Then remote and local variables affecting decadal variability of dust frequency over Gobi Desert are investigated.Section 4 describes conclusions.

Study region
In figures 1(a) and (b), climatological mean patterns of the growing season (April to October) precipitation and vegetation are depicted.The largest precipitation is observed in the eastern to southern part of China, under the influence of the East Asian Summer Monsoon.While, Mongolia and the northwest region of China show relatively low precipitation, which is generally classified as arid to semi-arid region.Corresponding areas with Normalized Difference Vegetation Index (NDVI) below 0.1 are located within the regions where the precipitation is below 200 mm.The specified Gobi Desert area is mainly covered with bare soil except for the lower east region.The location of Gobi Desert in this study is defined as 90-112 °E, 40-45.5 °N, using definition as Chen et al (2017bChen et al ( , 2017c) ) and Kim et al (2021).

Ground-based observations
Conventional records from the World Meteorological Organization (WMO) meteorological network remained to be the best accessible dataset that covers longest temporal evolution of dust.The WMO protocol classifies dust events into four categories: (1) dust-in-suspension, (2) blowing dust, (3) slight to moderate sand-dust storm, and (4) severe sand-dust storm.The classes are distinguished through varying codes, and more specific descriptions of these categories are listed in Supporting Information (table S1).Here, category 1 is excluded as visibility exceeding 10 km does not seem to represent regional dust emission (Mahowald et al 2007).Dust index is calculated as the sum of category 2, 3, and 4. 3-hourly records (1974-2022) from stations that are archived at https://www.ncei.noaa.gov/data/global-hourly/archive/csvare used.The record provides quality checks, and data that do not pass all quality checks are neglected here.Figures 1(a) and (b) (also in Supporting Information; table S1) shows the surface monitoring sites utilized in this study.Because Gobi Desert dust frequency is dominated by the spring dust events in the region, the impact of land surface processes on recent trend of spring dust frequency will be investigated in this study (Supporting Information figure S1).Although the dust event monitored at the WMO network is a variable that is the most frequently used as a proxy of dust (Sun et al 2001, Lee andSohn 2011), there are some subjectivities in a way that observers judge conditions and there can be changes in recording practices.

AERONET
AERONET is an efficient surface-based network that uses sky radiometer and sun photometer to observe the Aerosol Optical Depth (AOD) and the Angstrom Exponent (AE) (Holben et al 1998).The version 3, level 2 AOD and AE daily dataset (only observed during daytime, 8 A.M. to 5 P.M. LT) from Dalanzadgad site (1998 ∼ present, http://aeronet.gsfc.nasa.gov;see figures 1(c) and (d)) are used for study.The data are cloud-screened and quality-assured, with specific description provided in Giles et al (2019).AOD and AE data were missing for the years 2002,2008,2009,2011, and 2016 during spring season.Data have been considered and analyzed based on data availability.

Satellite-derived index, NDVI
NDVI is an index of amount and greenness of vegetation, distinguishing from dead and lifeless objects (Kriegler et al 1969).For the period 1982-2015, NDVI from dataset produced by the National Aeronautics and Space Administration (NASA) Global Inventory Monitoring and Modelling Systems (GIMMS) at 8 × 8 km 2 spatial and 15-day temporal resolutions obtained by the Advanced Very High Resolution Radiometer (AVHRR) is used.To avoid inaccuracy in NDVI on representing vegetation due to winter snow, only April to October monthly NDVI is used.

Reanalysis data 2.3.1. ERA5
The European Centre for Medium-range Weather Forecasts Reanalysis 5th generation (ERA-5) is created utilizing 4D-Var assimilated scheme with the CYcle41r2 (CY41R2) of the European Centre for Medium-range Weather Forecasts (ECMWF) model forecasting.The 4-D element of ERA-5 uses measurements from satellite instruments and conventional data records for assimilation with an update interval on 12 h for meteorological fields (Hersbach et al 2020, Muñoz-Sabater et al 2021).The hourly output has a 31 km horizontal resolution with vertical levels of 137 (spanning from ground surface to 0.01 hPa) capturing much improved details of atmospheric events compared to previous, coarser resolution, global reanalysis datasets.Here, assimilated variables wind, pressure, temperature, precipitation, relative humidity, and snow cover (from 1973 to 2021) are used for analysis.

GLDAS
The NASA Global Land Data Assimilation System (GLDAS) is generated to ingest surface and satellite observation data with advanced offline (not coupled to the atmosphere) land surface modelling and data assimilation systems.The offline land surface simulation system is provided globally at high resolutions, which combines the ground-based and satellite instrument data to produce high quality land surface variables (Rodell et al 2004).The mode National Centers for Environmental Prediction (NCEP) Oregon State University (Department of Atmospheric Sciences), the United States Air Force, Hydrologic Research Lab of National Weather Service Land Surface Model (NOAH LSM) of second version of GLDAS is used for this study, with spatial resolution of 0.25°× 0.25°, and the timeseries of monthly data ranging from 1948 to 2014 (data since 1973 used for this study), which is longer than for data provided by utilizing other land surface models.Here, soil moisture and soil temperature are used for analysis.

Climate indices: palmer drought severity index (PDSI)
The Palmer Drought Severity Index (PDSI) was the first procedure to quantify drought severity across different climate successfully.Variables used to calculate a climatic water balance principle are precipitation, temperature, and available soil water content.Severity of drought is expressed in PDSI when the water supply shows continually lower than normal in a region (Palmer 1965).PDSI is typically expressed in the range of −4 (extreme drought) to +4 (intense moist), although much extreme values are possible.A Self-Calibrated Palmer Drought Severity Index (SC-PDSI) was introduced to achieve consistent values at different locations, making comparison among the regions easy (Wells et al 2004).Herein, SC-PDSI data from 1973 to 2018 will be used for analysis.
In this study, in order to emphasize the decadal variability, we systematically remove long-term trends of meteorological variables before performing correlation and regression calculations.

Evaluating dust event frequency data
Dust frequency record from WMO stations is evaluated first against AERONET data.Properties of vertical air columns from AERONET can be used to validate regional dust activity in the source regions, as dust raised from the Gobi Desert can be lifted to an elevation of <3 km in many cases (Sun et al 2001).Besides, spring dust events in the region dominantly occur during the daytime (Guan et al 2017), indicating that daytime AERONET observations can represent dust activity in Gobi Desert.
For comparison, spring AERONET data are used to check regional dust events.Dust days with AERONET observations are selected when dust events from WMO station are observed.Figures 1(c) and (d) represent optical properties (AOD and AE) of both dusty and non-dusty days.AE less than 1 is indicative of the presence of coarse particles, such as dust and sea salt (Eck et al 1999).Dust days show high AOD and low AE, while nondusty days with low AOD and high AE.In addition, low and high standard deviations of AOD are observed in non-dusty and dusty days, respectively, explaining varying characteristics of each dust case.

Decadal change in dust frequency: possible linkage with oceanic variability
Figures 2(a) and (b) depict a normalized temporal variation and power-spectrum of the recent 49 years of spring dust event frequency from WMO observations in Gobi Desert, respectively.The dust index does not show a linear trend for 1974-2022.The spectrum of dust index is characterized by two high and significant variances at interannual and decadal time scales, with two peaks at 7 and 24 years (shortness of dust observation record causes a limitation in the spectral analysis).In addition to inter-annual variations, a clear multidecadal oscillating variation of dust frequency is detected.Figure 2 reveals a declining trend between ∼1980 and ∼1990, followed by an increasing trend between ∼1990 and ∼2010, and then another declining trend post ∼2010.
Although not many studies covered recent 50 years as in whole, the trends of dust reported show consistent results with dust occurrences in Gobi Desert presented in this study for a limited corresponding period (Lee 2011, Guan et al 2017, Liu et al 2022).Figure 1 One of important modes for decadal climate variability is the North Atlantic Oscillation (NAO), featured as a seesaw pattern of surface pressure between subtropical high and subpolar low.It is a dominant climate mode, which strongly affects Northern Hemisphere climate, through excitement of large-scale Rossby wave train stretching to Eurasia, thereby modifing regional climate variations.Persistent North Atlantic SST (NASST) winter anomalies associated with the NAO have significant impacts on winter to following spring climate in Eurasia (Yu and Zhou 2004).While, influence of the wintertime NAO on the following summer Eurasian climate is proposed (Sung et al 2006;Ding et al 2020).Figure 2(c) reveals a clear correlation between decadal dust activity over the Gobi Desert region and NASST region.The time-lagged correlation map between 7-year running mean spring dust index and ERA-5 Sea Surface Temperature (SST) north of equator from winter to summer highlights strong positive correlations over the North Atlantic (especially high in 50-58°N, 310-330°E), indicating NAO effect (Supporting Information, figure S3).Supporting Information figure S4 illustrates the lagged correlation between SST from January to next year February SST and spring dust frequency in Gobi Desert.NASST exhibits a strong correlation with spring dust frequency in Gobi Desert year-round, but the most pronounced response is observed in February (as denoted by the black lines in figure S4, correlation coefficient = 0.92).Therefore, NASST index is calculated using SST values in February (hereafter Feb. NASST) due to its particularly robust correlation in this study.Feb. NASST index is defined as a region-averaged normalized value.Positive (negative) Feb. NASST is associated with the stronger (weaker) spring dust activity of Gobi Desert.
NASST anomaly reaches its peak during winter, but lingers into the summer due to the substantial heat capacity of the ocean (Sung et al 2006; figure S4).Throughout the manuscript, we computed statistics concerning Feb. NASST.However, this does not imply that wintertime perturbations are directly conveyed into the summer season.Instead, a warm anomaly in NASST during the summer months triggers Rossby wave trains over the North Atlantic, which are then transmitted to Gobi Desert and the surrounding region, leading to high temperatures and drought conditions in this region (Supporting Information, figure S5).
As demonstrated by a recent study by Liu et al (2022), dust frequency over Gobi Desert can be influenced by sea surface temperature anomalies across various oceans.Our results similarly reveal the impacts of tropical Atlantic SST, tropical Pacific SST, and NASST (figure S4).Nevertheless, our study indicates that NASST exerts the most substantial influence on dust occurrences in Gobi Desert.If we employ the same NAO region as Liu et al (2022), tropical Atlantic SST generally exhibits the most significant impacts, followed by NASST, aligning with the findings of Liu et al (2022).Li et al (2020) and DeAngelis et al (2023) connected NASST and/or tropical SST anomaly to heat and drought events over Northeastern China and Siberia via Rossby wave trains and teleconnections related to anomalous tropical heating.

Main factors that influence multidecadal dust variations
To investigate a possible driver of the decadal scale variation in spring dust frequency in the Gobi Desert, trends of meteorological and land surface variables were compared against dust index.A linear trend is excluded for temperature and soil moisture/precipitation, as clear linear increasing and decreasing trend is shown over Mongolia (Gobi Desert region as well), respectively, since the late 1970 from previous paper (Zhang et al 2020).Regional variables show clear seasonality; therefore, time-lagged correlations are derived (Supporting Information figure S6).A multidecadal trend of spring dust frequency is associated with spring wind speed and spring soil moisture, not spring precipitation, showing significant correlations (figures 2(d), (e), and Supporting Information figure S6 and table 1).Snowmelt and soil moisture from the previous winter and fall precipitation do not seem to be related (insignificant correlations) to spring dust activity (table 1).The summer precipitation, soil moisture, temperature, and drought index, on the other hand, show significant correlations with spring dust, along with fall soil moisture (figures 2(e), (f) and table 1).This indicates that wind speed and soil wetness along with summer precipitation are important driving factors of multidecadal dust activity in spring.The correlation coefficient between dust storm frequency and temperature in June-July-August (JJA), September-October-November (SON), and March-April-May (MAM) is positive.However, in December-January-February (DJF), this correlation is negative.In wintertime of Gobi Desert, cold fronts pass through the region, and frequency of cold fronts passing is known to be affecting dust storm frequency in following spring (Park et al 2021).To see if dust-driving factors are modulated by Feb. NASST, normalized variables are time-lagged regressed against 7-year running mean Feb. NASST index (figure 3).Warm (cold) Feb. NASST anomaly is associated with stronger (weaker) spring wind speed at the surface in Gobi Desert region (figure 3(a)).While, summer precipitation and PDSI anomalies (figures 3(c) and (d)) show a reversed pattern over the region on average, especially in edges of the desert, with a summer temperature increasing pattern over the region (figure 3(b)).With the surrounding regions including upper Mongolia to Northeast China, lower (higher) soil moisture pattern showed with warm (cold) Feb. NASST (not shown).On top of that, spatial patterns of summer precipitation and drought index are consistent with spring soil moisture.Warm (cold) Feb. NASST anomaly is clearly associated with modulating regional spring climate of Gobi Desert, leading to more (less) frequent dust events.

Decadal large scale circulations patterns modulating dust occurrence
In figure 4, 7-year averaged geopotential height anomaly at 500 hPa during summer is time-lagged and regressed against Feb.NASST.On decadal scale, pressure anomaly at 500 hPa clearly reveals Rossby-wave train like pattern, supporting wave propagation and further imposing wave propagation starting from North Atlantic to Gobi Desert (figure 4(a)).Wave activity flux (Takaya and Nakamura 2001) supports wave propagation (Supporting Information figure S7).Warm (cold) anomaly over Feb. NASST induces prominent stagnant high (low)-pressure anomaly over Mongolia and Gobi Desert in summer.Winter NAO is shown to be modulating strength and location of upper-tropospheric westerlies over subtropical Asia, changing precipitation and temperature patterns in China (drier North and South, while wetter in between) through relocating East Asian Summer Monsoon (Yang et al 2004, Sung et al 2006).Teleconnection pattern revealed in figure 4 resembles superimposed pattern of polar-Eurasian and Silk Road patterns (Li et al 2020), different from that solely triggered by NAO.Blocking at the upper level caused air subsidence at the lower level, resulting in heatwave structure at lower level (figures 4(b) and (c)).As a result, a decadal warm (cold) Feb. NASST anomaly exhibited subsidence of much hotter (colder) and drier (wetter) air than climatology in Gobi Desert, indicating heatwavelike patterns that hinder precipitation and decrease soil moisture storage.A descending motion provides sunny weather, which enabled solar radiation to reach the land surface unimpeded.On top of that, in climatologically dry region, a positive feedback mechanism occurs between atmospheric heating and further drying of the soil.It was shown that soil moisture/temperature interactions increase summer temperature variability, resulting in extreme temperatures when soil moisture is low (Lorenz et al 2010).
3.5.Re-emergence of soil moisture in cold desert, gobi desert In recent 50 years, decadal shifts of summer precipitation anomalies are appeared in Gobi Desert (Supporting Information figure S6).Over Gobi Desert, higher (lower) summer precipitation leads to higher (lower) summer soil moisture content.In this study, we selected the years of driest (2002, 2006, 2011, 2010, 2001, 2007, 2005, 2009, and 1987; exceeding 0.6 standard deviation) and wettest (1995, 1996, 1998, 1993, 1988, 1992, 1990, 2008, and 1984; exceeding 0.6 standard deviation) summers during observational period for a composite analysis (figure 5).The analysis of 9 years of meteorological variables and dust frequency revealed notable spring differences (soil moisture 0.36/0.40,difference 0.04 and dust events 112/62, difference 50 for dry/wet) between the dry and wet cases, achieving statistical significance with a p-value of less than 0.01.In wetter years, larger soil moisture exhibited starting from July, led by larger summer precipitation.Soil temperature goes below 0 °C during November to February, indicating soil moisture freezing.Difference in soil moisture remained through fall and seems to be transferred to the next spring, as snowfalls do not largely disturb soil moisture throughout the winter.It can be indicated that soil moisture anomaly from summer is maintained throughout the winter to spring through freezing of soil moisture.Because soil moisture from summer is carried over to following spring, interannual relation exists between summer drought and frequency of spring dust events, possibly causing strong correlations.This observation analysis is acceptable, as soil moisture is an important parameter for dust particle entrainment into the atmosphere, as water particles enhance adhesion between soil particles, which in turn increase threshold wind speed (Zou andZhai 2004, Wu et al 2022).

Conclusion
In this study, a recent 50 years of trend in spring dust frequency in Gobi Desert is analyzed.Using multiple datasets, a clear decadal variability with a period of 24 years of spring dust activity of Gobi Desert is identified.(1995, 1996, 1998, 1993, 1988, 1992, 1990, 2008, 1984) and smallest (2002, 2006, 2011, 2010, 2001, 2007, 2005, 2009, 1987) soil moisture from July to August.Observed values from GLDAS, and GIMMS datasets are averaged regionally, while those from WMO are averaged among stations.
Dust events in the region are defined using WMO station observations.Dust events detected by the WMO network agreed reasonably with AERONET optical property data, low AE and high AOD in dusty days.However, it must be stated that dusty days with available AERONET observation only represent 10% of days with dust events, because of data availability issue in AERONET record.
There are considerable positive relations between the February NASST index and the spring dust frequency in Gobi Desert on multidecadal scale (correlation coefficient = 0.92).While the NASST anomaly peaks during winter, it persists through summer.Positive (negative) NASST anomaly is associated with the stronger (weaker) spring surface wind of Gobi Desert, while the drier (wetter) summer.Significant relations between NASST and spring dust-related factors reveal multidecadal variabilities of spring surface wind (correlation coefficient with spring dust activity, 0.85) and soil moisture content (correlation coefficient with spring dust activity, −0.48).Variability of spring soil moisture is shown to be correlated with previous summer precipitation, rather than concurrent precipitation.Through observation datasets, it is shown that soil moisture anomaly obtained from summer precipitation is maintained throughout the year, and re-emergence of soil moisture anomaly caused spring dust activity variability.Multidecadal variations in summer NASST are responsible for Rossby wave propagation, influencing whether the Gobi Desert area experiences dry or wet summers and subsequently more or less dusty conditions in the following spring.In the future, multidecadal mathematical modeling at global scale as hindcast would be crucial to better understand the physical mechanism underlining possible relations between NASST and spring dust frequency in Gobi Desert.Upon validation through modeling, this research holds the potential to forecast dust event frequency on a decadal scale.Such predictions would serve as a crucial element for air quality forecasting, not only in China, South Korea, and Japan, but also on a global scale.
Drylands area have expanded as a result of anthropogenic climate change (Burrell et al 2020).Despite conservation and restoration efforts, Chinese drylands, including Gobi Desert, remain vulnerable to desertification (Li et al 2021).Hence, it is probable that the frequency of dust storms will rise.Nevertheless, this study underscores the potential significance of the multi-decadal oscillatory patterns in dust storm frequencies over Gobi Desert, which exhibit a robust correlation with NASST.Our study emphasizes the importance of characterizing not only the linear trends but also the non-linear variability in dust storm frequency.
Variations in NASST are closely related to NAO, Arctic Oscillation (AO), and other long-term internal climate variability.In future studies, it would be beneficial to establish connections between dust frequency over Gobi Desert and various natural climate variabilities and relevant indices, along with their shifts in response to anthropogenic climate change.Anomalies in tropical Atlantic SST and associated dynamics can affect meteorology over Sahara Desert.It is crucial to evaluate changes in dust storm frequency over Sahara Desert, including assessing the impacts of Atlantic SST variations.In this context, the analysis should incorporate the varying vegetation and soil types found in Gobi and Sahara Deserts, along with their respective impacts on soil moisture retention.

Figure 1 .
Figure 1.Climatology of (a) GIMMS NDVI and (b) ERA-5 precipitation during growing season (April -October) over period 1982-2015.Gobi Desert is defined in black box with WMO and AERONET stations designated in red dots and magenta star, respectively.(c) Scatter diagram of daily MAM AERONET AOD verses AE over the period, 1998-2021 shown with dust days and nondust days in yellow and gray dots, respectively.(d) Averaged optical properties (AOD and AE) of dust (yellow box) and non-dust (grey box) days are described.One standard deviation is shown in whiskers.
(b)  inLiu et al (2022) highlighted the presence of multidecadal oscillating variability in dust frequency over Gobi Desert, aligning with the findings of our study.Trends beyond 2015 remain uncertain due to the relatively short duration of available time, both in Liu et al (2022) and our study.Uncertainty in assessing dust activity over Gobi Desert comes partly from the limited number of monitoring stations.We have included figure S2 in the Supporting Information, which illustrates a similar trend over a significantly larger area (90-126.5°E,40-49°N) based on data from an extensive number of monitoring stations.

Figure 2 .
Figure 2. (a) Temporal variability of Gobi Desert normalized dust event frequency (dust index) during the period, 1974-2022 shown.Interannual variability is shown in thin line with dot.While, 7-year running average is shown in a thick line.(b) Power-spectrum of dust index depicted for the same period.Interannual and decadal timescale shown in blue and yellow shades, respectively.A recent trend of normalized spring WMO dust event frequency is given in black (c)-(f).Normalized 7-year averaged (c) Feb. NASST, (d) spring ERA-5 surface wind speed, (e) GLDAS soil moisture, (f) PDSI, recent trends are shown in red, green, blue, and red, each.

Figure 3 .
Figure 3. Time-lagged regressions between 7-year running averaged Feb. NASST index against normalized 7-year running averaged (a) spring ERA-5 850hPa wind speed, (b) summer ERA-5 850hPa temperature, (c) summer ERA-5 precipitation, and (d) summer GLDAS soil moisture are shown.Values passing 95% significance t-test are depicted in dots.Gobi Desert is shown in black box.

Figure 4 .
Figure 4. 7-year running averaged summer ERA-5 geopotential height at (a) 500 hPa is regressed against 7-year running averaged Feb. NASST index.7-year running averaged summer (b) temperature (°C) and (c) relative humidity (%) along with (line contour) geopotential height (m) that are regressed against 7-year running averaged Feb. NASST index in vertical-zonal cross sections averaged among 90-112 °E.Dotted regions indicate significance levels of 5%.Only statistically significant (passing significance levels of 5%) values of geopotential height are shown in (b) and (c).Gobi Desert is shown in black box.Long-term linear trend of temperature is ignored, as strong global warming signal is shown.

Table 1 .
The time-lagged correlations of regionally variables against spring dust event frequency within the Gobi Desert region in decadal scale (7-year running average).Statistically significant (95% level) values are shown in bold with ** .