Aero sol Par ti cles from Dried Salt-Lakes and Sa line Soils Car ried on Dust Storms over Beijing

1 Cen ter for At mo spheric Chem is try Study, De part ment of En vi ron men tal Sci ence and En gi neer ing, Fudan Uni ver sity, Shang hai 200433, China 2 Key Lab o ra tory of Ra dio met ric Cal i bra tion and Val i da tion for En vi ron men tal Sat el lites, China Me te o ro log i cal Ad min is tra tion (LRCVES/CMA), Na tional Sat el lite Me te o ro log i cal Cen ter, China Me te o ro log i cal Administration, Beijing 100081, China 3 State Key Lab o ra tory of At mo spheric Bound ary Layer Phys ics and At mo spheric Chem is try (LAPC), the In sti tute of At mo spheric Phys ics, CAS, Beijing 100029, China 4 The Cen ter for At mo spheric En vi ron men tal Study, Beijing Nor mal Uni ver sity, Beijing 100875, China 5 Cen ter for At mo spheric Chem is try Stud ies, Grad u ate School of Ocean og ra phy, Uni ver sity of Rhode Is land, Narragansett, RI 02882-1197, USA 6 State key Lab o ra tory of Loess and Qua ter nary Ge ol ogy, In sti tute of Earth En vi ron ment, CAS, Xian 710075, China


IN TRO DUC TION
Dust storms, which arise in desert and loess ar eas of north west ern China and sweep over the ma jor cit ies of North ern China, such as Beijing, each spring, have be come in creas ingly in tense over the past few years (Cyranoski 2003).
Dusts driven by strong west erly winds ac com pa ny ing cold fronts are trans ported long dis tances to Ko rea, Ja pan, the Pa cific Ocean, and even as far as North Amer ica (Duce et al. 1983;Gao et al. 1992;Mori et al. 2002 etc.).About half of the dust from deserts in China and Asia is fi nally deposited in seas near China and the re mote Pa cific Ocean (Arimoto et al. 1996;Zhang et al. 1997).The loose loess in arid and semi-arid ar eas in north west ern China and In ner Mon go lia is the main source of dust in China's storms.Yuan et al. (2006) col lected soil sam ples at var i ous land sur faces in Cen tral In ner Mon go lia.These were re suspend ed us ing a dry pow der at om izer in an en clo sure chamber to iden tify the or i gin of Asian sands for trac ing them as they trav eled down wind.Zhang et al. (2003) and Zhuang et al. (2001Zhuang et al. ( , 2003) ) stud ied the physico che mical char ac ter istics of dust storms and their im pacts on global cli mate dur ing long-range trans port in China in 2000, 2001, and 2002. Wang et al. (2004) dis cussed trans port path ways and sources of PM10 pol lu tion in Beijing dur ing spring 2001, 2002, and 2003 and found that the path ways and sources of pol lut ing aero sols are as so ci ated with Asian dust emissions.Re cently Zhuang et al. (1992Zhuang et al. ( , 2003) ) fur ther dem onstrated the mech a nism of cou pling and feed back be tween iron and sul fur in air-sea ex change and con cluded that trans port in dust storms is an im por tant path in global biogeochemical cy cles.
Many sci en tists have paid a good deal of at ten tion to the char ac ter iza tion of bulk aero sols in dust storms (Zhuang et al. 2001;Sun et al. 2004, etc.), while few have car ried out indi vidu al par ti cle anal y sis, which can give much more information on het er o ge neous re ac tions, morphometrics, and trans for ma tion dur ing long-range trans port.A dif fi culty with anal y sis of bulk aero sols is that it does not dif fer ent parti cles from dif fer ent sources.Yet, in di vid ual par ti cle anal ysis can pro vide use ful in for ma tion on the orig i nal source and chem i cal trans for ma tion of chem i cal spe cies dur ing longrange trans port.Gao and An der son (2001) an a lyzed sin gle aero sol par ti cles (us ing au to mated SEM) col lected from Beijing, Qingdao, and Waliguan dur ing April 1999 and revealed that the par tic u late mat ter of each re gion are complex, het er o ge neous mix tures of soil dust with anthro pogenic par ti cles from var i ous sources.Us ing im age anal y sis Whittaker et al. (2003) found that par ti cles in dust storm sam ples of PM10 at Beijing are a mix ture of soot, smelter pol lut ants, loess and gyp sum.Shi et al. (2003) also dis tinguished par ti cles col lected over Beijing to be soot, fly ash, and min er als by us ing im age anal y sis.Zhang et al. (2000Zhang et al. ( , 2003) ) re ported that sin gle par ti cles from both dust storm and non dust storm sam ples col lected from Qingdao, a costal city, were coated by sul fate and ni trate.
Here we re port new find ings based on in di vid ual par ticle anal y sis in ad di tion to XPS, IC anal y sis for aero sol samples col lected from a super dust storm over Beijing.The results dem on strate that be sides deserts, par tic u late mat ter was gen er ated from the sur face soils of dry salt-lakes and the saline soils of arid and semi-arid areas.

EX PER I MEN TAL SEC TION 2.1 Sam pling of Aero sol Par ti cles
Sam ples were col lected on 20 March 2002 when a super dust storm (DS) cov ered Beijing (39.9°N, 116.4°E) ac compa nied by strong winds of ~10 m s -1 .Sam pling was conducted at be tween 15:30 -19:50, just af ter the dust peaked.Two TSP sam ples were also col lected on non dust storm days: one from Beijing (des ig nated as "NDS", sam pled at 8:30 -10:30 on 25 July 2002, clear weather), and the other from Duolun (des ig nated as "DL", 42.2°N, 116.4°E, sampled at 12:06 -17:40 on 19 April 2002, clear weather).DS and NDS were taken on the roof of the Sci ence and Tech nology Build ing of Beijing Nor mal Uni ver sity, twelve sto ries (~40 m) above the ground.DL was col lected on the roof of a three-story build ing at Duolun Mid dle School.Aero sol samples were col lected on Nuclepore polycarbonate mem brane fil ters of 0.6-mm pore size and 90-mm di am e ter with a medium-vol ume sam pler (model: TSP/PM 10 /PM 2.5 -2; flow rate: 77.59 L min -1 ) man u fac tured by Beijing Geo log i cal In strument-Dickel Co., Ltd.The aero sols sam ples for ion chromato graphic (IC) anal y sis were col lected on a Whatman4l fil ter mem brane (Whatman Co., UK).Sam ple fil ters were put in a poly eth yl ene plas tic bag im me di ately af ter sampling.Af ter weigh ing (Sar to rius 2004MP Elec tronic balance, 0.01 mg) un der con stant tem per a ture and hu mid ity, the sam ples were re frig er ated for later anal y sis.All pro cedures were strictly mon i tored to avoid any pos si ble con tamina tion.

In di vid ual Par ti cles Anal y sis
In di vid ual par ti cles were an a lyzed with a HITACHI X650 Scan ning Elec tron Mi cro scope (SEM) cou pled with an en ergy-dispersive X-ray microanalyzer (EDX) fol low ing EDX 9100's pro ce dures.The op er at ing con di tions were set to 20-kV (ac cel er at ing volt age) and 40-pA (beam cur rent) with spec tral ac qui si tion time of 60 -100 s.A 1-cm 2 piece of the fil ter was mounted on a stub and coated with a thin graph ite film by heat ing and sput ter ing.We ran domly selected sev eral ar eas on the fil ter with rel a tively low par ti cle den sity and an a lyzed all the par ti cles in these ar eas for 14 elements.Car bon and ox y gen were not con sid ered be cause they were too abun dant in the polycarbonate fil ters.The inten si ties of the char ac ter is tic X-ray lines of each el e ment were con verted into their cor re spond ing atomic frac tions by the stan dard ZAF cor rec tion method -The method cal cu lates the ef fects of atomic num ber (Z), ab sorp tion (A), and characteristic (and some times con tin uum) flu o res cence (F), separately.The abun dance of each ox ide in the par ti cles was de ter mined with the stan dard ized ox ide method for EDX, which as sumes that each el e ment (ex cept Cl) ex ists as an oxide and sums the 14 abun dances of the el e ments to 100%.Ge om e try ef fects that arise from the fi nite size of mi cro parti cles in tro duce a sys tem atic er ror, but this er ror has lit tle effect on the char ac ter iza tion of par ti cle types (Gao and Anderson 2001).In all, we mea sured 565 sin gle par ti cles from DS, 419 from NDS, and 498 from DL. De tails of the method used are given in Yuan et al. 2004.

Ion Chro mato graphic (IC) Anal y sis SO 4
2-and Cl -were an a lyzed for the sam ples col lected dur ing the dust storm with a Dionex 600 IC sys tem that consists of an an ion sep a ra tion col umn (Dionex Ionpac AS11), a guard col umn (Dionex Ionpac AG11), a self-re gen er at ing sup pressed con duc tiv ity de tec tor (Dionex Ionpac ED50), and a gra di ent pump (Dionex Ionpac GP50).Peaknet 6 software was used for chro mato graphic data pro cess ing.The filters were ex tracted in an ul tra sonic cleaner and fil tered through 0.45-mm microporous mem branes (F 25 mm), and the ex tracts were then in jected into the IC with poly propylene sy ringes (Yuan et al. 2003).

X-Ray Pho to elec tron Spec tros copy (XPS) Anal y sis
The sur face com po si tion and struc ture of the aero sols was mea sured us ing XPS (PHI 5000C ESCA Sys tem, 14 kV, 250 W, 93.9 eV, exciton, AlK a, scan ning range for ac quir ing spec tros copy, 0 -1200 eV, Perkin Elmer Com pany, USA).The min ute spectroscopies of 10 el e ments (C 1s , O 1s , Na 1s , Mg 2p , Al 2p , Si 2p , S 2p , Ca 2p , Mn 2p , and Fe 2p ) were ac quired.A spe cial soft ware (PHI-MATLAB) was used for data analysis.The bind ing en ergy of pol luted C1s (284.8 eV) was used in cal i bra tion to con firm the bind ing en ergy of all other el e ments.
The rel a tive con cen tra tions of at oms are cal cu lated by the fol low ing for mula (Boudevilie et al. 1979): (1) Where n is the atom num ber over the aero sol sur face, I is the in ten sity of XPS, s is the area of the el e ment in some en ergy level, and E k is the ki netic en ergy of the pho toelec tron.De tails of the method are given in Zhang et al. (2005).

PMF Fac tor Anal y sis
Pos i tive Ma trix Factorization (PMF) was de vel oped by P. Paatero of the De part ment of Phys ics, Uni ver sity of Helsinki, Fin land, and mar keted jointly by YP-Teekniika K Y Com pany and the uni ver sity's De part ment of Phys ics.It was first ap plied to en vi ron men tal data by Juntto and Paatero (1994).Many works have shown it to be an ef fec tive and novel method for source ap por tion ment of aero sol par ti cles (Huang et al. 1999).PMF is a pow er ful tool that can work with out orig i nal source in for ma tion, pro vide all the nonnegative load ings of the el e ments in ma trix, and op ti mize results by ad just ing the stan dard de vi a tion of each data point.Sup pose X is a ma trix with n ´ m, where n is the number of sam ples and m is the num ber of chem i cal com po nents, in clud ing all ions, el e ments and or gan isms.Then X can be di vided into GF + E, where G is the ma trix of n ´ p, and F is the ma trix of p ´ m with p be ing the num ber of main pol lution sources.E is a ma trix de fined as fol lows: (2) where S ij is the stan dard de vi a tion.The re stric tion con dition is that the el e ments in G and F are all nonnegative.E can ac quire op ti mized re sults and con firmed G and F by com par ing Q val ues.Usu ally F is con sid ered a load ma trix and G the fig ure ma trix of the sources.

RE SULTS AND DIS CUS SION 3.1 Gen eral De scrip tions
The mass frac tion and the num ber frac tion [the ra tio of the num ber of par ti cles, in which the el e ment of in ter est was de tected, to the to tal num ber of par ti cles mea sured, also called num ber fre quency (see be low)] of each el e ment in the sam ples of DS, DL, and NDS were re ported in Yuan et al. 2006.The re sults show that: (1) the ma jor el e ments, in cluding Al, Si, P, S, and Ca were abun dant in sin gle par ti cles of DS, DL, and NDS, whose num ber frac tions were close to 100%; (2) Both Si and Al are dom i nant in the par ti cles of DS and DL, and the sum of Si and Al ac counted for ~65% of the to tal; whereas, greater num bers of S, P, and Cl were found in NDS, the sum of these three el e ments con trib uted 35% of the to tal; (3) Na and Mg in NDS were high est among the three sam ples, in di cat ing that Na and Mg in NDS were partly from pol lu tion; whereas, K was greater in DS and DL than in NDS, in di cat ing that more K was from dust, as with Si and Al; (4) Cu in DL had an ex tremely high num ber fre quency (96.8%), re veal ing there must be a spe cial source of Cu nearby.Fan et al. (1996) re ported the num ber fre quen cies of Mg, K, Ca, and Fe in dust storm par ti cles col lected in April 1991 in Beijing were all above 62%, re sults which were similar to this study.How ever, in this study we found very high num ber fre quen cies of S (99.5%) and Cl (82.5%) in DS; these val ues were much higher than that of S (14%) and Cl (0%) in the dust storm of 1991.

Com po si tion and Size Dis tri bu tion of In di vid ual Par ti cles
The high est re corded TSP hourly con cen tra tion was 10.90 mg m -3 for the DS pe riod of 20 March 2002.This was 36 times higher than TSP con cen tra tion re corded dur ing the non-DS.The 12 hr-av er aged PM 2.5 con cen tra tions also reached 1.39 mg m -3 (Sun et al. 2004).The size dis tri bu tion and the mor phol ogy of the par ti cles (mor phol ogy var ies with par ti cle size) were mea sured with SEM im ages of sin gle parti cles.Par ti cle sizes ranged from ~0.3 to ~20 mm.The median sizes were ~1.3 mm for DS par ti cles, ~0.8 mm for NDS par ti cles, and 1.2 mm for DL par ti cles.For DS par ti cles, > 80% were < 2.5 mm, which is re spi ra ble.
The av er age rel a tive mass per cent ages of SO 3 and Cl in DS par ti cles were 6.42% and 0.57% with the high est pro portions of 97.39% and 11.43%, re spec tively.The per cent age of (S+Cl)-con tain ing par ti cles con sti tut ing the to tal par ti cles mea sured was 82.5%; the av er age rel a tive mass percents of SO 3 and Cl were 6.70% and 0.58%, re spec tively, which were higher than the over all (n = 565) av er ages (2.6% and 0.42%, re spec tively).The (S+Cl+Na)-con tain ing par ti cles rep re sented 62.0% of the to tal par ti cles mea sured; the rel ative mass per cent of Na was 6.10%, and 1.4 times that of the over all mean (4.36%).These re sults showed that the three el e ments Na, S, and Cl were en riched in some DS par ti cles.The pro por tion of Na-con tain ing par ti cles with rel a tive mass per cent of > 15% in the DS sam ple was 1.24% of the to tal, which is 5 times higher than that (0.24%) in the NDS sam ple from Beijing.The rel a tive mass per cent of Na in the DL parti cles was 1.54%, which is lower than that (4.36%) in the DS sam ple from Beijing.The size of those S-en riched par ti cles with a rel a tive mass per cent of 15% S was < 1.0 mm, in di cating that these S-en riched fine par ti cles were mostly from sec ond ary pol lut ants.In con trast, (S+Cl)-con tain ing par ti cle size was > 1.0 mm.This in di cates that those par ti cles were mostly pri mary and their sources were likely the same though not from deserts, in which Si dom i nates the coarse par ti cles.The PMF anal y sis showed that the ma jor com po nents in the (Na+S+Cl)-rich par ti cles were Na 2 SO 4 and NaCl, and the num ber of these par ti cles was 9% of to tal par ti cles.We have iden ti fied (Na+S+Cl)-con tain ing par ti cles from the spectrum of the EDX anal y sis and the SEM im ages, as shown in Fig. 1a.Ad di tion ally, the com mon par ti cles in the DS samples are alu mi no sili cate par ti cles en riched in Si, Al, and Ca, as in di cated in Fig. 1b.These re sults dem on strate that par ticles that were en riched by a com bi na tion of S, Cl, and Na dur ing DS were likely not from deserts.

Cor re la tion Anal y sis of the El e ments in In di vid ual Par ti cles
Ta ble 1 lists the re sult of cor re la tion anal y sis by STA -TISTICA®.Sur pris ingly, among the 14 el e ments mea sured   2 shows a marked pos i tive cor re la tion between Cl and S in in di vid ual dust par ti cles with an av er age ra tio Cl/SO 3 of 0.09, which is close to the world av er age ratio for soil [0.06, from Bowen (1966)].How ever, the ra tio of Cl/Al 2 O 3 was ten times that of world soil.This sole cor re lation of Cl with S and the high ra tio of Cl/Al 2 O 3 im ply that Cl and S might come from an iden ti cal source where the soil has been en riched with both S and Cl.It is well known that dry salty lakes and sa line soils in In ner Mon go lia, which con tain both Na 2 SO 4 and NaCl (Yang 1996), lay on the path of dust storms, thus, dust storms would very likely pick up these par ti cles here.The high cor re la tion be tween Cl and S in dust storm par ti cles sug gests that these sa line soils could be a source of particulates for dust storms (Yuan et al. 2006).This anal y sis can also ex plain why a num ber of dust par ticles con tain ing Na, S, and Cl of crustal or i gins were found over Qingdao be fore pass ing across the ocean (Zhang et al. 2003).

Ter nary Plots Anal y sis
Ter nary plots (Yuan et al. 2004) of three ma jor com ponents [clay and quartz (Al 2 O 3 , SiO 2 , etc.), cal cite (CaCO 3 , CaO), and salts (sul fate, phos phate, chlo ride)] of sin gle parti cles re vealed that Na 2 SO 4 and NaCl might be ex ter nally/inter nally mixed in the DS par ti cles.Yuan et al. (2006) proved that the CaCO 3 , CaO, CaSO 4 , and salts, i.e., sul fate and phos phate plus chlo ride, such as (NH 4 , Na) x (SO 4 , Cl, PO 4 ) y , mixed with clay and quartz in those par ti cles.

Com po si tions and Cor re la tions of the Sol u ble Ions
Fig ure 3  the TSP and PM 2.5 sam ples col lected dur ing DS.The concen tra tions of the SO 4 2-and Cl -tracked each other.The av erage con cen tra tions of SO 4 2-and Cl -dur ing DS are 1.95 and 0.48 mg g -1 , re spec tively.These re sults are 27 and 8 times those of DL (SO 4 2-0.07 mg g -1 , Cl -0.06 mg g -1 ).It has been re ported that con cen tra tions of SO 4 2-and Cl -in typ i cal salty soils in North China are 1.4 and 1.5 mg g -1 (Jiang and Zhang 2001).Com par ing all these re sults, it is clear that there are more S and Cl in the in di vid ual par ti cles of DS, im ply ing that there must be other sources than deserts gen er at ing these par ti cles, such as sa line soil en riched with S and Cl com po nents.Fig ure 4 il lus trates con cen tra tions and cor re la tion of in land city, and wind di rec tion on the day of the storm was from the north west or north at more than 10 m s -1 .Cl -could not have come from the sea, which is lo cated to east or southeast.Given the pos i tive re la tion ship be tween re sults for individual par ti cle anal y ses and IC anal y sis for S and Cl and the pre vail ing winds dur ing DS, it is likely that S and CL share a com mon source some where on the path of the dust storm to the north west or north of Beijing.

PMF Fac tor Anal y sis of the El e ments in In di vid ual Par ti cles
The re sults of the 565 in di vid ual par ti cles col lected during DS, 419 par ti cles of NDS, and 498 par ti cles of DL have been an a lyzed by PMF2, which has been suc cess ful in getting source ap point ments of aero sol par ti cles in Hong Kong, Thai land, and Spain, etc. (Huang et al. 1999;Lee et al. 1999;Predes et al. 1999;Jiang and Zhang 2001).The re sult of PMF anal y sis show that S and Na are of the same fac tors (55% SO 3 , 41% Na 2 O, and 36% P 2 O 5 ), in di cat ing that Na 2 SO 4 was a ma jor com po nent of sul fates in the dust aerosols.The re sults from graphic tech nique anal y sis based on the SEM data also dem on strate that Na 2 SO 4 is the main contrib u tor to this fac tor in dust aero sol par ti cles (Yuan et al. 2004).Ac cord ing to graphic anal y sis of in di vid ual par ti cles, (NH 4 )x SO 4 ex isted in dust aero sols, though in small numbers, thus giv ing Na 2 SO 4 as the ma jor sul fate.Cl, Na, and Mg are of the same fac tor for the sam ples of DS, in which the ra tio of Cl/Na is 5.This is quite dif fer ent to its ra tio in sea water (1.8), which is nor mally based on the ex is tence of both NaCl and MgCl 2 .As pre vi ously dis cussed northwesterlies or northlies pre vailed dur ing DS, mak ing it im pos si ble for sea salts to be a fac tor in the par tic u late mat ter of DS and sug gest ing that par ti cles en riched with Na and Cl were sourced from salt pans and sa line soils in north west and north ern China.This is fur ther ev i denced by the main compo nents of soils from salt pans and sa line soils be ing rock salt (NaCl) and Glau ber's salt (Na 2 SO 4 ) (Xu 1993).

Sur face Anal y sis by XPS
The XPS (X-ray pho to elec tron spec tros copy) anal y sis is an ex cel lent method for study ing par ti cle sur face struc ture and com po si tion.Ta ble 2 lists XPS re sults of sam ples collected dur ing DS.The rel a tive mass con cen tra tion per centage of Na on the sur face was 0.30% in par ti cles from DS, 0.27% for NDS and 0.18% for DL, in di cat ing greater Na enrich ment of DS par ti cles.This re sult again con firms that there must have been an al ter nate source for Na on the storm's path as it trav eled to ward Beijing.XPS re sults also re veal the rel a tive mass con cen tra tion per cent age of S on the sur face to be 0.80% for DS par ti cles, a re sult higher than that for both NDS and DL sam ples (0.39%).Sim i larly, these results along with those of SEM and PMF anal y ses in di cate an other source be sides the desert as be ing re spon si ble for S, Na, and CL lev els in DS par ti cles.

Source of Those Par ti cles En riched with S, Cl, and Na
Okada and Kai ( 2004) an a lyzed soil sam ples col lected at Qira in the Taklamakan Desert by EDX and found that 'Cl-rich' and 'Na-rich' par ti cles were de tected in ap prox imately 10% of to tal par ti cles.The back ward tra jec tory from 500 m above Qira at the time of sam pling showed that the air par cel did not ar rive from over oce anic ar eas within the previ ous five days.The pres ence of at mo spheric ha lite par ti cles in the Taklamakan Desert was re ported by Molnár et al. (1993) and Yabuki et al. (1998Yabuki et al. ( , 2002)).Hence, 'Cl-rich' and 'Na-rich' par ti cles were sug gested to be mainly com posed of ha lite.The 'Na-rich' par ti cles con tained S as the sec ond most abun dant el e ment, ex clud ing Cl.A 'S-dom i nant' particle was iso lated from the two sam ples.This par ti cle was evap o rated upon ir ra di a tion with an elec tron beam, lead ing the re search ers to con clude that it was com posed of sul fate not orig i nat ing from a min eral source; how ever, they did not ex plore fur ther the pos si ble sources of those par ti cles contain ing S, Cl, and Na.
For DS two pos si ble paths ex isted for the storm's approach: north and north west (Ren et al. 2003).Fig ure 5 shows that all kinds of chlo ride-sul phate sa line soils ex ist in the trans port path way of DS.Strong wind could have car ried these par ti cles en riched with S and Cl and trans ported them to Beijing.The sa line soils cover an area of about 763.01 10 4 hm 2 in In ner Mon go lia, these ac count for 21% of China's to tal (3630.53´ 10 4 hm 2 ) (Fan and Ma 2001).There are also a great num ber of salt lakes in In ner Mon go lia (Li et al. 1990;Xu 1993).The A'LaShan salt lake area to the north and north west of Beijing is one of four ma jor salt lake ar eas in In ner Mon go lia, and it lies on the dust trans port path to Beijing.Be cause of con tin ual drought in re cent years, there are many salt pans in these salt-lake ar eas.The main com ponents of the salt pans are rock salt (NaCl) and Glau ber's salt (Na 2 SO 4 • 10H 2 O) pro vid ing an abun dant source of S and Cl en riched par ti cles.An ear lier study (Wang et al. 1987) of aero sol com po si tion in the drylands of north ern China reported the com po si tion of aero sols col lected at a site (see Fig. 5) lo cated on Tianchi Moun tain, east of Urumqi, to be strongly en riched with both S and Cl.They at trib uted this en rich ment to up per most soil lay ers be ing en riched with S and Cl in this area.It's worth not ing that there are many saline soils around this sam pling site, 45°53'N, 88°07'E, such as I 2 , I 14 , and I 20 in Fig. 5.These soils are typ i cally en riched with sul fate and chlo ride.Thus, for this early study, strong wind could have raised sur face sa line soils dur ing their sampling pe riod giv ing such re sults.Thus, our study along and this ear lier study both show that be sides deserts, sur face soils of dry salt-lakes and sa line soils in arid and semi-arid ar eas can pro vide par tic u late mat ter for dust storms.

CON CLU SIONS
(1) In di vid ual par ti cle anal y sis of aero sols col lected from a super dust storm on 20 March 2002 in Beijing showed that among all the 14 el e ments mea sured, only S and Cl had re mark able pos i tive cor re la tion.(2) PMF anal y sis in di cated that NaCl and Na 2 SO 4 were major com po nents of these par ti cles, in which S and Cl showed sig nif i cant pos i tive cor re la tion.(3) The ev i dence that S and Cl in this dust storm were highly re lated in di cates that dried salt-lake and sa line soils over north ern and north west China might have con trib ute to the dust storm's or i gins.

Fig. 1 .
Fig. 1.Spec trum of a cer tain in di vid ual par ti cle of duststorm aero sol by EDX: (a) S-and Cl-rich par ti cle; (b) com mon dust par ti cle.
vid ual par ti cles in DS, only one pair of el e ments, S with Cl, show good cor re la tion (the cor re la tion co ef fi cient be ing 0.75, a = 0.05, p << 0.01).Cor re la tion co ef fi cients are only 0.53 in NDS sam ples and 0.07 in DL sam ples, re spectively.It is clear that aero sol par ti cles in DS are dif fer ent to those par ti cles of NDS and both DS and NDS par ti cles are dif fer ent again to the non-dust par ti cles of DL, al though the desert of Duolun, In ner Mon go lia was con sid ered to be a ma jor source of aero sols for dust storms over Beijing.It must be noted that in di vid ual par ti cle anal y sis is dif fer ent to bulk aero sol anal y sis.The con cen tra tions of crustal el ements, such as Si, Al, and Fe, nor mally show pos i tive cor rela tion in bulk aero sol sam ples col lected at dif fer ent times and sites; whereas, the re sults from in di vid ual par ti cle analysis do not show such a re la tion ship(Yuan et al. 2006).Indi vid ual par ti cle anal y sis gives in for ma tion re gard ing an indi vid ual par ti cle from a large num ber of par ti cles in an aerosol sam ple; this does not pres ent in for ma tion re gard ing bulk anal y sis.Fig ure Fig ure3shows the con cen tra tions of the SO 4 2-and Cl -in Cl -of six aero sol sam ples col lected se quen tially dur ing DS.The sam pling times were: 10:20 -12:20, 12:22 -14:22, 14:25 -16:25, 16:25 -18:25, 18:25 -20:25, and 20:25 -22:25.The fig ure shows a very pos i tive cor re la tion (R 2 = 0.9908) be tween SO 4 2-and Cl -in these sam ples.Beijing is an
This work was sup ported by the Beijing Ur ban Me te o rol ogy Re search Fund (No. UMRF 200704), the key pro jects from the 11 th five-year plan of national sci en tific and tech no log i cal (2008BAC34B04-2), the fund from State Key Lab o ra tory of At mo spheric Boun d ary Layer Phys ics and At mo spheric Chem is try (LAPC-KF-2008-11), Na tional Ba sic Re search Pro gram of China (Nos.2005CB422202 and 2006CB403702), National Na t u ral Sci ence Funds of China (Grant Nos.30230310, 20077004, and 20477004), and Beijing Nat ural Sci ence Funds (Grant Nos.8991002 and 8041003).REF ER ENCESArimoto, R., R. A. Duce, D. L. Savoie, J. M. Prospero, R. Talbot, J. D. Cul len, U. Tomza, N. F. Lewis, and B. J. Ray,

Fig. 5 .
Fig. 5. Dis tri bu tion of sa line soils over In ner Mon go lia, China.Note: 1. two ar eas with red line mean that two path (North-west path and North path) for dust to Beijing.2.I 1 , I 2 , and II 1 mean dif fer ent types sa line soil in China.