Tropospheric aerosol extinction coefficient profiles derived from scanning lidar measurements over Tsukuba, Japan, from 1990 to 1993

Tropospheric aerosol extinctioncoefficient profiles derived from scanning lidarmeasurements over Tsukuba, Japan, from 1990to 1993

Yasuhiro Sasano

Mie scattering lidar was used to observe aerosol extinction coefficient profiles in the troposphere overTsukuba ~140 E, 36 N!, Japan, for three years from March 1990 to February 1993, and data obtained infair weather were analyzed. The lidar measurements were made by a vertical scanning mode togenerate profiles of extinction coefficients from the lidar level to a 12-km altitude. The extinctioncoefficients were derived from the lidar signals using a two-component ~air molecule and aerosol! lidarequation, in which the ratio of aerosol extinction to backscattering was assumed to be constant. Sea-sonal average profiles were derived from individual profiles. Three-year average profiles were alsocalculated and modeled using mathematical expressions. The model profile assumed ~1! a constantextinction ratio in the atmospheric boundary layer ~ABL!, ~2! an exponentially decreasing extinction ratioabove the ABL, and ~3! a constant extinction ratio in the upper troposphere where the extinction ratio canbe defined as the ratio of the aerosol extinction coefficient to the air molecule extinction coefficient. Theextinction ratios both in the ABL and in the upper troposphere and the scale height that was used toexpress the exponential decrease were used as three unknown parameters. Seasonal variation of opticalthickness that was obtained by integrating extinction coefficients with height was also investigated.© 1996 Optical Society of America

Key words: Lidar, aerosol lidar, backscatter and extinction, aerosol, troposphere.

1. Introduction

Global climate changes caused by human activitiessuch as consumption of fossil fuels and emission ofanthropogenic chemicals have attracted worldwidescientific and socioeconomic attention. One such is-sue is global warming caused by greenhouse gasessuch as carbon dioxide, methane, water vapor, andchlorofluorocarbons, some of which are of natural or-igin and some of which are anthropogenic. Changesin the emission rates and distributions of these gasesmust be accurately monitored and investigated.Another important factor in global warming is par-

ticulate matter such as aerosols and clouds. Cloudscontribute differently to short-wave and long-waveradiation depending on their type, altitude, optical

The author is with the National Institute for EnvironmentalStudies, Tsukuba, Ibaraki 305, Japan.Received 3 December 1995; revised manuscript received 11

March 1996.0003-6935y96y244941-12$10.00y0© 1996 Optical Society of America

thickness, particle size, and so on. Thus it is quitedifficult to estimate the effects of clouds on warming.Cloud generation is affected by the general circula-tion of air and water. Modeling cloud formation ingeneral circulation models has not been successful.Aerosols have direct and indirect effects on climate.

The indirect effect is that they act as condensationnuclei that cause cloud formation. The direct effectis scattering and absorption of solar radiation.Aerosols help to cool the lower atmosphere caused bythe umbrella effect of scattering, whereas absorption-type aerosols help to heat the atmosphere. There-fore it is essential to understand the amount anddistribution of aerosols as well as their optical prop-erties.1,2 The lack of global data on aerosol distribu-tion, however, has limited accurate determination ofaerosol net effects on climate.3,4 Vertical distribu-tion information is required because light scatteringand absorption are altitude dependent, as are cloudproperties.4Aerosol measurements have been taken by in situ

sampling and chemical analysis, size distributioncounting by the optical scattering principle ~see, e.g.,

20 August 1996 y Vol. 35, No. 24 y APPLIED OPTICS 4941

Refs. 5 and 6!, and remote-sensing techniques such aslidars, which have been shown to be effective forstratospheric aerosol measurements. The increaseand decay of aerosols in the stratosphere after majorvolcanic eruptions such as El Chichon and Pinatubohave been clearly observed by lidars ~see, e.g., Refs.7–9!. Since lidars can be operated from the groundas long as weather conditions permit, a large amountof data has been accumulated frommany observationsites. Although tropospheric aerosols are also agood target of lidar measurements because of thefrequency of observation opportunities and altitudeavailability, few studies have been reported on tro-pospheric aerosol measurements with lidars.Kent et al.10 demonstrated the capability of satel-

liteborne solar occultation sensors to derive uppertropospheric aerosol characteristics from the Strato-spheric Aerosol and Gas Experiment ~SAGE! II.They showed a long-term variation of aerosol massloading and effective sizes as a function of altitude.I describe the lidar measurements of tropospheric

aerosols that were taken with Mie scattering lidar ata ground station at Tsukuba ~140 E, 36 N!, which isapproximately 60 km northeast of Tokyo, Japan, forthree years beginning in March 1990. I describe theaerosol extinction coefficient profiles and opticalthicknesses that were estimated from these measure-ments. Despite the fact that the measurementswere taken at a single station, the data could be usedfor input to climate models and validation data foraerosol assessment using satellite image data.Lidar measurements provide aerosol extinction ~or

backscattering! coefficient profiles, as described in de-tail later. Lidars with a single-wavelength lasercannot be used to provide any information on sizedistribution or composition of aerosols but can beused to provide information on altitude profiles ofextinction coefficients as well as optical thickness byintegrating them with altitude. In this analysis,only the data obtained in fair weather were analyzed,thus this was clear-sky aerosol climatology.In Section 2 I describe the instrument that was

used, the observation site, the observation method,and the conditions. In Section 3 I explain the dataanalysis procedure. The measurements were takenin a vertical scanningmode to obtain data to as low asground level ~lidar level!. Lidar signals in digitalformwere analyzed by using the solution proposed byFernald,11 in which he considers air molecules andaerosols in the lidar equation. In Section 4 I discussthe analytic results and the aerosol model profilesthat were created with mathematical expressions.

2. Measurements

A. Lidar Instrument and Observation Site

When laser pulses are emitted into the atmosphere,one can use a lidar to detect light backscattered byparticulates in the atmosphere and to record its mag-nitude as a function of time after pulse emission.Themagnitude of backscattered light is related to theconcentration of particulates whereas the time differ-

4942 APPLIED OPTICS y Vol. 35, No. 24 y 20 August 1996

ence between pulse emission and signal reception canbe used to determine the distance to the scatterers.Since scattering by particulates can be described bythe Mie scattering theory when the size of particu-lates is comparable to or greater than the wavelengthof light, the lidar used for detecting aerosols is calledMie lidar.The lidar that was used in this study is rotatable in

the azimuth and elevation directions with a receivingtelescope of 1.5-m diameter. As a light source, thesecond harmonic of a Nd:YAG laser was used with532-nm wavelength, 10-W average power, and 25yspulse repetition rate. The lidar was located on thetop floor of an eight-story building with an observa-tory dome at the top. The height of the lidar was34 m above ground. The roof was open when mea-surements were taken. Details of the lidar systemcan be found in Shimizu et al.12The lidar site is located in the central part of the

Kanto Plain, approximately 60 km northeast of thecenter of Tokyo. The second largest lake in Japan,Lake Kasumigaura, is approximately 10 km east ofthe site. Mt. Tsukuba, 876 m high, is approximately30 km to the north. The lidar site is surrounded bya geographically flat area that consists of an urbanarea with low buildings ~i.e., Tsukuba City with anarea of 260 km2 and a population of 156,000!, scat-tered villages, cultivated fields, and woodlands.Gamo13 described the characteristics of meteoro-

logical conditions and development of an atmosphericmixed layer over this area on the basis of meteoro-logical data for five years from 1977 to 1981. Ac-cording to his analysis, westerly winds prevail mostlyin January and February whereas the winds shiftfrom the west to the south in the daytime duringother months. The change in wind direction occursin the morning hours from May to October, whereasit occurs after 15:00 JST ~Japan standard time! inMarch, April, November, and December. Relativehumidity reaches its maximum between 1 and 2 hafter sunrise and gradually decreases until midday.The minimum relative humidity is 25–30% from Jan-uary to March, 55–60% from July to September, andhas values between 25% and 60% during the othermonths.

B. Observation Method and Conditions

The lidar was directed toward a position that was 72deg east from the north for each measurement andwas scanned in the elevation direction from a hori-zontal position to the zenith with a speed of 0.9 degys.Received signals were digitized every 100 ns with ananalog–digital converter, resulting in a 15-m rangeresolution. Signals were accumulated for 25 pulsesand then recorded on magnetic tapes. Signals fromas far away as 150 km were recorded.One cannot get enough information from a verti-

cally pointing lidar for ranges near the lidar becausethe field of view and the laser path do not overlapsufficiently ~the geometric form factor is less thanunity!. The current lidar measurements with verti-cal scanning were used to obtain information about

aerosol distribution to as low as ground level by em-ploying the analysis procedure described in Section 3.This procedure is advantageous because most aero-sols are distributed in the lower atmosphere at whichmeasurements cannot be made with conventional li-dar.Along with the lidar, measurements were made

with a Sun photometer ~500-nm wavelength!, a laserparticle counter, and an aureolemeter to obtain col-umn optical depth, aerosol size distribution informa-tion, and the refractive index of aerosols. Theresults have been reported by Takamura et al.14The measurements were taken routinely at ap-

proximately 10:00, 13:00, and 16:00 JST on the oper-ator’s duty days ~Monday through Friday except forholidays!. When the clouds were low in the mea-surement direction, measurements were not takenbecause difficulty with data analysis was expectedwhen horizontal inhomogeneity existed in the atmo-sphere. The data used in this analysis were ob-tained during the period from March 1990 toFebruary 1993.

3. Data Analysis Procedure

A solution was proposed by Fernald11 for the two-component lidar equation that takes into consider-ation air molecules and aerosols under theassumption of a constant extinction to backscatterratio ~hereafter referred to as S1 or scattering param-eter!. To use Fernald’s solution one must specify aboundary condition of a certain value for the extinc-tion coefficient at a designated range. It has beenshown that the boundary conditions at the far sidegive a better convergence when integration is madetoward the near side. Since the extinction value atthe boundary is usually unknown, it is often assumedthat an aerosol-free layer exists at a certain level,which is called the matching method to calibrate thelidar signal. With this method we assume that scat-tering occurs from air molecules only, which is con-sidered to be correct around the tropopause and inthe upper stratosphere. According to Kent et al.15who analyzed aerosol profiles from satellite sensorssuch as SAGE, SAGE II, and Stratospheric AerosolMeasurement II, themodal value of the ratio betweenaerosol extinction and air molecule extinction at 1.0mm is the minimum value at a level of approximately10 km.Data that were used for this study were all ob-

tained from measurements with a vertical scanningmode. When the elevation angle of the lidar direc-tion is large enough, a high-quality signal can beobtained to the altitude at which one can apply thematching method. Information near the groundwould not be available because the geometric formfactor is less than unity. On the other hand, mea-surements at low elevation angles provide informa-tion on aerosols near ground level whereas thematching method can not be used with those signalsbecause the assumption of an aerosol-free layer is notalways valid.To avoid these difficulties in lidar measurements, I

used the iterative procedure proposed by Sasano andNakane16 and also used by Takamura et al.14 for thedata obtained with a vertical scanning mode. Withthe iterative procedure, one can assume that horizon-tal distribution of aerosols is not homogeneousaround a constant value at each altitude level.Boundary conditions for each lidar signal obtainedthrough scanning are given as follows:First we define Rs as the shortest range where the

geometric form factor is regarded as unity and Re asthe longest range where the quality of the lidar signalis high enough to be analyzed in terms of signal-to-noise ratio ~see Fig. 1!. Xe and Ye are defined as thefar ends of the region to be analyzed in the verticaland horizontal directions, respectively. The lidarsignal that corresponds to each elevation angle ui canbe expressed as Pi~Rj!, where Rj 5 jDR with a rangeresolution of DR. The subscript i ~51, M! is theorder of elevation angle and j ~51, N! is the order ofrange bins. The iterative steps for processing eachmeasurement are as follows:

~1! Give zeros for the aerosol extinction coefficientas boundary conditions for all the lidar signals ~i 5 1,M! at R 5 Re. If the location of the boundary isoutside the rectangular box defined by Xe, Ye, and theorigin, the boundary is assigned to the farthest pointin the box, as shown in Fig. 1.

~2! Calculate extinction coefficients by integratingthe lidar solution from the boundary according toFernald’s formula.

~3! Define a grid in the box with spatial resolutionsof Dx and Dy in the vertical and horizontal directions,respectively. The extinction coefficients derived inthe previous step within the same subgrid box areaveraged to give a single value for each subgrid box.If there is no corresponding data in a subgrid box, thedata from the nearest four points can be averagedand substituted in the subgrid box. A vertical pro-file can be derived by horizontally averaging the two-dimensional grid data.

~4! The averaged extinction coefficient can be usedas a new boundary at each altitude level. If thereare any negative values in the averaged extinction

Fig. 1. Definition of the data analysis region and boundary ~seetext!.


coefficient, the altitude Xc where the minimum valueappears can be searched. In this case, the boundaryvalue of zero extinction coefficients is given at alti-tude Xc. For lidar signals not including altitude Xc,the same condition is given as for Step ~1!. The pro-cess from step ~2! to step ~4! can be repeated until thechange rate of averaged extinction coefficients be-tween the latest and previous calculations becomesless than 5%. Note that a constant scattering pa-rameter was used in the data-processing proceduredescribed above.

Extinction coefficients of air molecules were calcu-lated using aerological sonde data supplied by theJapan Meteorological Agency based on the Rayleighscattering theory. Information on tropopauseheights was also derived from sonde data.The extinction coefficient profiles used in the fol-

lowing analysis represent the final results of theabove procedure and the horizontally averaged val-ues. Standard deviations from average values werealso calculated. Variation coefficients defined as theratio of standard deviation to average value wereused as the index of horizontal homogeneity in orderto screen out good quality data. In the current anal-ysis, most of the data were analyzed with Rs 5 5 km,Re5 12 km,Xe5 12 km, Ye5 10 km, Dx5 50m, andDy 5 50 m. Errors were the result of ~1! underesti-mation by applying the matching method, ~2! a con-stant scattering parameter, and ~3! a representativevalue for the scattering parameter.Russell et al.17 discussed the errors that were

caused by applying the matching method, which al-ways underestimates the extinction coefficient.These are relatively large errors, especially in theregion with small extinction coefficients. When theFernald solution was applied for backward integra-tion, the solution tended to converge to true values asthe integration proceeded, even when the boundarycondition at the far side contained an error. Thisadvantage is also applicable to the current analysis.The error in the boundary condition is less importantin the lower atmosphere where optical thickness in-creases. It was assumed that the aerosols understudy have the same composition and size distribu-tion and therefore a constant scattering parameter.Aerosols in the atmosphere, however, are expected tohave various origins, compositions, size distributions,and physical and chemical properties. It is quiteprobable that the aerosols in a layered structure havedifferent characteristics.It is not easy to assess this effect because there has

been little data on physical and chemical properties ofaerosols in the atmosphere. Therefore we decided toassume a constant scattering parameter but not toapply the scattering parameter as a function of range.Sasano et al.18 made a theoretical study of the effectsof constant scattering parameters on lidar solutions.When aerosols are homogeneous and spherical, the

Mie scattering theory can be applied to estimate ascattering parameter, which is a function of size dis-tribution and refractive index. It is not an easy task,


however, to estimate even a constant scattering pa-rameter, because aerosol size distribution and therefractive index vary considerably depending on theirorigin and history. A lidar measurement itself can-not provide any information on scattering parame-ters.Takamura et al.14 proposed the use of total optical

thickness data obtained from direct Sun measure-ments with a Sun photometer along with lidar-de-rived extinction profile data in order to estimate thescattering parameter. They showed that the scat-tering parameters have values between approxi-mately 30 and 70. This technique requires a precisedetermination of tropospheric aerosol optical thick-ness from Sun photometer measurements that aremade simultaneously with lidar measurements.Since this was not the case for the current data set,we decided to use a representative value of 50 for thescattering parameter in our analysis.When we display two-dimensional extinction coef-

ficient data in X–Y grids, there are often cases inwhich isolated clouds exist aloft and a low signal-to-noise ratio exists because of high aerosol density nearthe ground. To make a data set for extinction coef-ficient profiles from good quality data, the followingcriteria were applied to the individual measurementdata in order to exclude low-quality data automati-cally.Let us denote the aerosol and air molecule extinc-

tion coefficients at the altitude level of zk as a1~zk! anda2~zk!, respectively, and the standard deviation foraerosol extinction coefficient as s1~zk!. Then we candefine err~zk! 5 s1~zk!y@a1~zk! 1 a2~zk!#. When ~1!err~zk! exceeds 2.0 at any level or ~2! err~zk! exceeds0.5 at ten or more levels, one can consider the data tobe of low quality and can thus be disregarded for theanalysis. With this screening, data from a total of158 days from March 1990 to February 1993 werefinally selected for further analysis. The number ofdays selected for each month is listed in Table 1.During the summer ~July, August, and September!,the amount of effective data is small because cumulusclouds often form during the daytime because ofstrong convection.All the data that we used were obtained on days

with no clouds, which resulted in a clear-sky aerosolclimatology.One profile was derived from one measurement.

A maximum of three profiles was obtained for one

Table 1. Number of Days of Lidar Observationsa

Year

Month

1 2 3 4 5 6 7 8 9 10 11 12

1990 — — 4 4 3 1 1 3 0 1 4 91991 14 8 3 10 2 3 1 3 3 1 8 51992 10 10 6 8 3 1 2 1 2 6 4 21993 5 7 — — — — — — — — — —

Total 29 25 13 22 8 5 4 7 5 8 16 16

aTotal of 158 days including nine Kosa days.

day. In the following, seasonal averages were cal-culated using one representative profile from one day.The measurement times were prioritized as 13:00,16:00, and 10:00 JST in descending order. Thehigher priority data were selected as a representativeprofile of the day to avoid any bias because of thenumber of profiles. In all, 70 profiles were selectedfrom the 13:00 JST measurements, 55 profiles from16:00 JST, and 35 from 10:00 JST.Seasonal average profiles were calculated by em-

ploying the representative profiles of each day.Months were arranged into seasons, that is, March,April, and May for spring; June, July, and August forsummer; September, October, and November for fall;December, January, and February for winter. Thewinter of 1990 consisted of December 1990, January1991 and February 1991, with similar definitions forthe winters of 1991 and 1992. A three-year averageprofile was also calculated for each season.Differences in measurement time affect the extinc-

tion coefficient profiles in the lowest atmosphericlayer just above the ground. At 10:00 JST, theheight of the atmospheric boundary layer ~ABL! isstill low, having relatively higher aerosol extinctioncoefficients. No substantial differences were foundin the extinction profiles above the ABL.Menzies et al.19 took the geometric mean of the

backscattering coefficient at each level because thefrequency distribution could be expressed with a log-normal probability function. Kent et al.15 analyzedaerosol profiles derived from satellite sensor data interms of a frequency distribution. We calculatedarithmetic means because aerosol extinction was con-sidered to have a linear relationship with radiativetransfer.

4. Results and Discussion

A. Vertical Profiles of Aerosol Extinction Coefficients

Figure 2 depicts the average profiles of the aerosolextinction coefficient for each season. The thincurve in the figure represents the extinction coeffi-cient for air molecules, for which the density profile ofair molecules for this calculation was taken from theU.S. Standard Atmosphere. Since the data qualitywas not so good in the spring and summer of 1990,the analysis was made to a 10-km altitude. To il-lustrate the variability of the profiles in each season,Fig. 3 shows the average profile and the magnitude ofstandard deviation for 1992 data. The relative mag-nitude of variability ~standard deviation! differs onlyslightly with altitude and season.Tropopause height varies depending on the season.

The average tropopause heights for the same periodas for the lidar data analysis are 11.7 km ~1.8 km! inspring, 15.3 km ~1.2 km! in summer, 14.4 km ~2.4 km!in fall, and 10.0 km ~2.0 km! in winter. The valuesin parentheses represent the standard deviations.Aerosol extinction coefficients decrease in altitude

from the lowest level to the highest level, getting closeto the value of air molecules or less at the 2–5-kmaltitude range, with the crossing level being depen-

dent on season and year. The crossing level is lowerin fall and winter than in spring and summer. Theterm crossing means that an aerosol extinction coef-ficient becomes equal to an air molecule extinctioncoefficient at a specific altitude. An increase in ex-tinction was seen above approximately 9 km after thewinter of 1991, which may be the result of aerosolsfrom the Pinatubo eruption that came down as thetropopause lowered in winter. Similar phenomenawere reported by Tratt and Menzies8 and Menziesand Tratt.20 Kent et al.10 also presented the strato-spheric material incursions into the upper tropo-sphere, which occur mostly in winter and spring.Larger extinction coefficients were found in the

midtroposphere from 4 to 8 km mainly in the springrather than in any other season. The variabilitywithin three years in this region ~in Fig. 3! showslarger fluctuations in summer and fall and smallerfluctuations in winter and spring. This may be dueto the small number of data samples for summer andfall ~see Table 1!.Three-year averages of the aerosol extinction coef-

ficient are represented by thin curves in Fig. 4.Again it is clear that spring has the largest extinctionover almost all the altitude ranges. The large valuesabove 8 km in winter are due to the Pinatubo aero-sols. The thickness of the ABL increases in the fol-lowing seasonal order: winter, spring, summer, andfall. The profiles shown in Fig. 4 are modeled withmathematical expressions for convenience. We usedthe following equations for the three-year averageprofiles:

a1*~z! 5 Aa2~z! for z # hb,

a1*~z! 5 ~A 2 B!a2~hb!exp@2~z 2 hb!yH#

1 Ba2~z! for z . hb,

where a represents the extinction coefficients, sub-scripts 1 and 2 represent aerosols and air molecules,respectively, hb is the height of the ABL, and H is ascale height that can be used to express the rate ofdecrease with altitude. These equations show thatthe aerosol extinction can be expressed by two terms:one is proportional to air molecule extinction and theother decreases exponentially with altitude. In thefirst equation, the aerosol extinction is expressed asproportional to air molecule extinction only, whichresults from the assumption that turbulent mixing islarge enough to cause aerosols to distribute homoge-neously in the ABL, resulting in a constant mixingratio in terms of the extinction coefficient.The unknown parameters A, B, H, and hb were

determined in the following way: First, the firstterm in the second equation was assumed to be muchsmaller than the second term above the midtropo-sphere when the effect of stratospheric aerosol intru-sion was negligible. Then parameter B was easilydetermined by the least-squares method and is thusconsidered as the extinction ratio, that is, the ratiobetween aerosol extinction and air molecule extinc-tion. Next, parameterAwas tentatively determined


Fig. 2. Seasonal mean aerosol extinction profiles: top left, spring; top right, summer; bottom left, fall; bottom right, winter.

by applying the least-squares method to the regionbelow hb, which was treated as a variable parameter.The second equation was then applied to the regionbetween hb and hu with parameters A and B to de-termine parameter H. Since the residuals for the


fitting were considered to be dependent on hb, thebest hb could be determined in order for the averagedresiduals in the region to be minimum. The hb var-ied from 0 to 2 km in 100-m intervals.The solid curves in Fig. 4 represent the modeled

Fig. 3. Seasonal mean aerosol extinction profiles with standard deviations for 1992: top left, spring; top right, summer; bottom left, fall;bottom right, winter.


Fig. 4. Three-year mean aerosol extinction profiles ~thin curves!, air molecular extinction ~medium thick curves!, and modeled profiles~bold curves!.


profiles. The parameter hu was arbitrarily given as5.6, 5.6, 4.0, and 5.0 km for spring, summer, fall, andwinter, respectively. To avoid the influence fromstratospheric aerosols, we set the upper altitude lim-its at 10.0, 10.0, 8.0, and 7.0 km for the respectiveseasons. These parameters are listed in Table 2 aswell as the results of parameter estimation for thecases with scattering parameters ~extinction-to-back-scattering ratio S1! of 30 and 70 in order to observethe effects of the scattering parameter. As discussedin Section 2, the extinction coefficient profiles dependon the scattering parameter. The aerosol extinctioncoefficients are proportional to the scattering param-eter in the region where attenuation of laser light isnegligible. On the other hand, dependency of theextinction coefficients on the scattering parameter issmall in the region where attenuation is large, suchas in the lower atmosphere.Table 2 indicates that the dependency of parame-

ters A, H, and hb on scattering parameter S1 is rel-atively small whereas parameter B dependssignificantly on S1, because B was determined fromthe data in the clear region in the upper troposphere.B shows values of one or less regardless of the season,which means the aerosol extinction is less than theair molecular extinction. B also shows larger valuesin spring than in any other seasonwith aminimum inthe winter.The heights of the atmospheric boundary layer hb

were determined as 1.0, 1.2, 1.7, and 0.3 km forspring, summer, fall, and winter, respectively, asshown in Table 2. The ABL is highest in the fall andlowest in the winter. The extinction ratio in theABL has a value of near 5.7 in the fall and around 9during the other seasons. The absolute value of ex-tinction coefficients that correspond to this extinctionratio is approximately 1024 m21, which can be con-verted to a visibility of 40 km.Scale height H for exponential decrease above the

ABL is 0.86 km in the summer, 0.38 km in the fall,and approximately 1.0 km in the winter and spring.Kent et al.10 reported a seasonal movement of mate-rial into the upper troposphere from below, indicatingwhat appears to be the tongue of high extinction aero-

Table 2. Parameters for Modeled Extinction Profiles

Season S1 A B H ~km! hb ~km! ByA

Spring 30 9.49 0.51 1.07 0.8 0.05450 9.82 0.83 0.97 1.0 0.08570 9.89 1.10 1.32 0.9 0.111

Summer 30 8.56 0.26 0.90 0.9 0.03050 8.91 0.41 0.86 1.2 0.04670 9.17 0.56 0.79 1.3 0.061

Fall 30 5.43 0.35 0.48 1.4 0.06450 5.68 0.52 0.38 1.7 0.09270 5.96 0.66 0.38 1.7 0.110

Winter 30 8.21 0.18 0.90 0.2 0.02250 8.03 0.28 1.05 0.3 0.03570 8.10 0.36 1.14 0.3 0.044

sol that extends upward from the 6-km level. Theseincursions show regular seasonal behavior withpeaks in the spring. The larger scale height in thewinter and spring might be relevant to what Kent etal. observed with the SAGE II data.Few observational studies have thus far provided

data on aerosol profiling in the troposphere. Basedon the reports of Jaenicke21 on scale heights for aero-sol distribution and background aerosols, mass pro-files for maritime aerosols have a scale height of900 m and the mass above 2400 m is a constant 0.07times of that near the ground. As for continentalaerosols, the scale height is 730 m and the ratio be-tween the mass above 2400 m and that near theground is 0.04. The extinction coefficient profilesthat we obtained are not necessarily the same as themass profiles, but the scale heights derived here andreported by Jaenicke are relatively the same. Theratio of aerosol extinction of background to that nearthe ground was calculated from ByA and is listed inTable 2 with the change from 0.02 to 0.11 and anaverage of 0.065, which is close to the Jaenicke value.Warneck22 described profiles of aerosol number den-sity and reported that the scale height for the lowertroposphere is approximately 1 km and the extinctionratio is constant with altitude for the upper tropo-sphere, which is similar to our results.

B. Optical Thickness

By integrating the extinction coefficient profile fromone level ~z1! to another ~z2!, optical thickness can bedefined as

t1~z1 2 z2! 5 *z1

z2a1~z9!dz9,

Figure 5 shows the seasonal variation of optical thick-ness calculated for three altitude regions, that is,0–3, 3–12, and 0–12 km, from daily representativeprofiles of aerosol extinction coefficients. The sea-sonal average and its standard deviation are alsocalculated.The three-year average optical thickness from

ground level to 12 km was approximately 0.2, withthe maximum in spring and the minimum in winter.Large fluctuations were found in spring and summer.The optical thickness between the ground and 3 kmhas the same tendency toward seasonal variation asthat between the ground and 12 km.The optical thickness between 3 and 12 km, which

is not affected by the lower ABL, shows little seasonalvariation, although there is a slightly larger value inspring. The optical thickness from the ground to 3km is approximately 0.8 times that from the groundto 12 km, with most of the tropospheric aerosols ex-isting in the lowest 3 km of the atmosphere.The optical thickness also depends on the scatter-

ing parameter S1. When the optical thickness issmall, the dependency on S1 is large and vice versa.Optical thicknesses were calculated with S1 5 30 and70 and compared with S1 5 50 in Fig. 6. The ordi-


Fig. 5. Seasonal changes in optical thickness for altitude ranges of 0–12 km ~bold curve!, 0–3 km ~medium thick curve!, 3–12 km ~dashedcurve!.

nate is a ratio of the optical thickness for S1 5 30 and70 to that for S1 5 50. The abscissa is the opticalthickness for S1 5 50. As the optical thickness de-creases, the ratios approach the values of the S1 ratio~1.4 and 0.6!. When the optical thickness for S1 5 50is larger than 0.2, dependency of the optical thicknesson S1 decreases and the ratio shows values within20% of unity. Although not shown here, the optical

Fig. 6. Dependence of optical thickness on S1.


thickness calculated with S1 5 30 and 70 for theatmosphere from the ground to 3 km is always within20% of that calculated with S1 5 50, even when theoptical thickness is small, showing less dependencyon S1.Since the atmospheric boundary layer is low in

winter, the optical thickness in winter is the smallestdespite the fact that the aerosol extinction coefficientin the region close to the ground shows similar valuesto those in other seasons. This means that it is notsufficient to infer optical thickness from visibilitymeasurements alone.The optical thickness from the ground to 3 km is

relatively large in spring and summer. One of thereasons for this seasonal variation ~Fig. 5!may be thecontribution from Kosa ~Asian dust! aerosols thatoriginate in the desert regions in the Asian continentand travel across Japan in the spring. The opticalthickness from 3 to 12 km also shows larger values inthe spring. Dust is often observed to be brought upto the middle troposphere in the source regions andtransported over long ranges.23 Polluted air is an-other reason for the increase of optical thickness insummer. The lidar measures scattering efficiency oflight, which might be affected by aerosol growth be-cause of high humidity during the summer.24Hofmann6 showed the seasonal variation of aerosol

number density obtained from a balloonborne opticalcounter measurement made over Laramie, Wyoming~41 N!. He showed that optically active aerosol hasamaximum in column number density between 5 and10 km in the spring. This is quite similar to the

seasonal variation of optical thickness in the3–12-km range shown in Fig. 5.Shiobara et al.25 inferred aerosol size distributions

in the column atmosphere from measurements ofsunlight taken by a scanning spectral radiometer atSendai, which is approximately 240 km north ofTsukuba along the Pacific coast. They estimatedseasonally averaged optical thickness for the tropo-sphere by subtracting the contribution of strato-spheric aerosol optical thickness. Their resultsshow that the average optical thickness was 0.28,0.33, 0.13 and 0.12 for spring, summer, fall, and win-ter, respectively. Although the location differencemakes direct comparison meaningless, the valuesthey obtained are close to ours and show similar sea-sonal variations. From the size distributions theyestimated that the soil aerosols of Kosa are dominant.Tanaka et al.26 measured aerosol size distributions

and optical thickness using an aureolemeter on dayswhen Kosa events were observed and on normal dayswhen Kosa events were not observed at Nagasaki inwestern Japan. The optical thickness for non-Kosadays ranged from 0.2 to 0.8 and that for Kosa daysfrom 0.5 to 1.0 at a 500-nm wavelength.We compared the aerosol extinction coefficient pro-

files for the non-Kosa and Kosa days in spring bydefining the Kosa days to include the day on whichthere were reports of Kosa somewhere in Japan, aswell as the two days immediately before and afterthat day. The comparison shows that larger extinc-tion coefficients were found in an average sense in theregion from 1.6 to 4 km for the Kosa days than for thenon-Kosa days. However, it is difficult to define thedifference precisely because fluctuation is larger thanthe difference, which may be the result of aerosollayers aloft even if no Kosa reports are made fromground-based observations.

5. Concluding Remarks

The aerosol extinction coefficient profile for each sea-son has beenmodeled. The data were obtained fromlidar measurements at Tsukuba, Japan, for the pe-riod from March 1990 to February 1993. Arith-metic, not geometric, means were made for the databecause the aerosol effects on radiation were consid-ered to be linear with aerosol extinction and opticalthickness. When we solved the lidar equation, theratio of extinction to backscattering coefficients ~re-ferred to as scattering parameter S1! was assumed tobe constant with distance and had a value of 50. Itwas also assumed that the aerosol was distributedalmost homogeneously along the horizontal direction~layered structure!.The data obtained at 13:00 and 16:00 JST on cloud-

less days, when a convective mixed layer was ex-pected to develop, were given higher priority, sincethey satisfied the condition of horizontal homogene-ity. Although there might be a problem in assuminga constant scattering parameter, this is the only tech-nique applicable to data obtained from Mie lidar. ARaman lidar technique allows for more quantitativeanalysis of aerosol distribution.

The author expresses sincere thanks to TamioTakamura of Chiba University and Tadahiro Ha-yasaka of Tohoku University for their useful discus-sions throughout this work. He is also grateful toIchiro Matsui of the National Institute for Environ-mental Studies, and Yoshihiro Sato and ManabuTakada of F. I. T. for their technical assistance withthe daily lidar measurements. This research wascarried out as part of the Observational Study onAerosols Distribution with Ground-Based Lidars un-der a research grant from the Science and TechnologyAgency of Japan.

References1. L. L. Stowe, R. Hitzenberger, and A. Deepak, eds., “Experts

Meeting on Space Observations of Tropospheric Aerosols andComplementary Measurements,” World Climate ResearchProgram Report WCRP-48 ~World Meteorological Organisa-tion, Geneva, 1990!.

2. S. G. Jennings, ed. Aerosol Effects on Climate ~University ofArizona Press, Tucson, Ariz., 1993!.

3. J. E. Hansen, and A. A. Lacis, “Sun and dust versus green-house gases: an assessment of their relative roles in globalclimate change,” Nature ~London! 346, 713–719 ~1990!.

4. R. J. Charlson, S. E. Schwartz, J. M. Hales, R. D. Cess, J. A.Coakley, Jr., J. E. Hansen, and D. J. Hofmann, “Climate forc-ing by anthropogenic aerosols,” Science 255, 423–430 ~1992!.

5. D. J. Hofmann, “Increase in the stratospheric background sul-furic acid aerosol mass in the past 10 years,” Science 248,996–1000 ~1990!.

6. D. J. Hofmann, “Twenty years of balloon-borne troposphericaerosol measurements at Laramie, Wyoming,” J. Geophys.Res. 98, 12,753–12,766 ~1993!.

7. D. M. Tratt and R. T. Menzies, “Evolution of the Pinatubovolcanic aerosol column above Pasadena, California, observedwith a midinfrared backscatter lidar,” Geophys. Res. Lett. 22,807–810 ~1995!.

8. M. T. Osborn, R. J. Decoursey, C. R. Trepte, D. M. Winker, andD. C. Wood, “Evolution of the Pinatubo volcanic cloud overHampton, Virginia,” Geophys. Res. Lett. 22, 1101–1104 ~1995!.

9. A. O. Langford, T. J. O’Leary, M. H. Proffitt, and M. H. Hitch-man, “Transport of the Pinatubo volcanic aerosol to a northernmid-latitude site,” J. Geophys. Res. 100, 9007–9016 ~1995!.

10. G. S. Kent, P.-H. Wang, M. P. McCormick, and K. M. Skeens,“Multiyear stratospheric aerosol and gas experiment. II. Mea-surements of upper tropospheric aerosol characters,” J. Geo-phys. Res. 100, 13,875–13,899 ~1995!.

11. F. G. Fernald, “Analysis of atmospheric lidar observations:some comments,” Appl. Opt. 23, 652–653 ~1984!.

12. H. Shimizu, Y. Sasano, H. Nakane, N. Sugimoto, I. Matsui,and N. Takeuchi, “Large-scale laser radar for measuring aero-sol distribution over a wide area,” Appl. Opt. 24, 617–647~1985!.

13. M. Gamo, “Diurnal variations of the mixed layer characteris-tics at Tsukuba,” J. Meteorol. Soc. Jpn. 66, 691–701 ~1988!.

14. T. Takamura, Y. Sasano, and T. Hayasaka, “Troposphericaerosol optical properties derived from lidar, sun photometer,and optical particle counter measurements,” Appl. Opt. 33,7132–7140 ~1994!.

15. G. S. Kent, M. P. McCormick, and S. K. Schaffner, “Globaloptical climatology of the free tropospheric aerosol from1.0-mm satellite occultation measurements,” J. Geophys. Res.96, 5249–5267 ~1991!.

16. Y. Sasano and H. Nakane, “Quantitative analysis of RHI lidardata by an iterative adjustment of the boundary conditionterm in the lidar solution,” Appl. Opt. 26, 615–616 ~1987!.


17. P. B. Russell, T. J. Swissler, and M. P. McCormick, “Method-ology for error analysis and simulation of lidar aerosol mea-surements,” Appl. Opt. 18, 3783–3797 ~1979!.

18. Y. Sasano, E. V. Browell, and S. Ismail, “Error caused by usinga constant extinctionybackscattering ratio in the lidar solu-tion,” Appl. Opt. 24, 3929–3932 ~1985!.

19. R. T. Menzies, G. M. Ancellet, D. M. Tratt, M. G. Wurtele, C.Wright, and W. Pi, “Altitude and seasonal characteristics ofaerosol backscatter at thermal infrared wavelengths using li-dar observations from coastal California,” J. Geophys. Res. 94,9897–9908 ~1989!.

20. R. T. Menzies and D. M. Tratt, “Evidence of seasonally depen-dent stratosphere–troposphere exchange and purging of lowerstratospheric aerosol from a multiyear lidar data set,” J. Geo-phys. Res. 100, 3139–3148 ~1995!.

21. R. Jaenicke, “Tropospheric aerosols,” inAerosol-Cloud-ClimateInteractions, P. V. Hobbs, ed. ~Academic, New York, 1993!, pp.1–32.


22. P. Warneck, Chemistry of the Natural Atmosphere ~Academic,New York, 1987!.

23. K. Kai, Y. Okada, O. Uchino, I. Tabata, H. Nakamura, T.Takasugi, and Y. Nikaiodou, “Lidar observation and numericalsimulation of a Kosa ~Asian Dust! over Tsukuba, Japan, dur-ing the spring of 1986,” J. Meteorol. Soc. Jpn. 66, 457–472~1988!.

24. G. Hanel, “The properties of atmospheric aerosol particles asfunctions of the relative humidity at thermodynamic equilib-rium with the surrounding moist air,” Adv. Geophys. 19, 73–188 ~1976!.

25. M. Shiobara, T. Hayasaka, T. Nakajima, and M. Tanaka,“Aerosol monitoring using a scanning spectral radiometer inSendai, Japan,” J. Meteorol. Soc. Jpn. 69, 57–70 ~1991!.

26. M. Tanaka, T. Hayasaka, and T. Nakajima, “Airborne mea-surements of optical properties of tropospheric aerosolsover an urban area,” J. Meteorol. Soc. Jpn. 68, 335–344~1990!.

Documents

Tropospheric aerosol extinction coefficient profiles derived from scanning lidar measurements over Tsukuba, Japan, from 1990 to 1993