Measurement of the wavelength dependence of atmospheric extinction due to scatter

Atmospheric Environment Perganton Press 1969. Vol. 3, pp. 551464.

Printedin GreatBritain.

MEASUREMENT OF THE WAVFJLENGTH DEPENDENCE OF ATMOSPHERIC EXT~CTION DUE TO SCATTER NC. AHLQUIST~II~ R.J. CWLSON Department of Civil Engiueering, University of Washington Seattle 98105, Washington (First received 11 Noem&

1968; and in $naf form 14 Febrwy

1969)

AbMu%--Atmospheric ha usually acts as a minus-blue filter for either solar radiation or for viewing objects at a distance. Dark objects are obscured by a blue hrus. The measurement of the actual wavelength dependence with a specially designed integrating nepheiometer will be described. Thre major points will be presented: (I) m siguificance of the wavelength dependence of atmospheric lit scatter; (2) The design of an instrument to measure the wavelength dependence of the extinction coe&ient due to scatter; (3) preliminary data to demonstrate the instrumental capability as well as to show the regularity of aerosol properties. INTRODUCTION

LXOHT scattering by aerosols in the atmosphere is responsible not only for visibility degradation but also for at least some of the color of objects viewed at a distance. Perhaps the best example of such a chromatic effect can be observed at sunset; if a large quantity of aerosol is present, the sun’s disk usually appears reddish in color while in cleaner air the sun maintains a more nearly white appearance. The presence of heavy pollution (e.g., in Los Angeles, California) at low humidity often results in a brown appearance, sometimes attributed to NOz. Distant white objects usually appear reddened in color while dark objects assume a blue or violet hue. Perhaps this latter phenomenon influenced the naming of the Blue Ridge and Oreat Smoky mountams of the Appalachian range of North America. ~isto~~Uy, the sky color and its magnificent variations posed an irresistable challenge to classical physicists. In a snmmary paper, NICHOLS(1908) discussed the then existing theories, some of which we now recognixe as being absurd even though they were proposed by such greats as Clausius and Rayleigh. Clausius once proposed water bubbles suspended in the air as the cause of color, while Rayleigh in his fist paper on sky color assumed the existence of the ether. The classical work of Rayleigh did, however predict correctly that the scattering of light by gases and sub-wavelength sized particles would depend on the inverse fourth power of the wavelength, henceforth called Rayleigh scatter. The study of the blueness of the sky was also approached ex~~ment~y; cyanometry was the name given to such obviations. A set of eight cards of varying shades of blue, called the Linke scale, was used in a manner similar to the Ringlemann smoke scale for judging sky color. The semi-quantitative description of colored haze has been available at least since ANG~STR(JM(1929). He concluded that the wavelength dependence could be empirically described in a manner analogous to that used for Rayleigh scatter by gases: b SCIt

"A'",

where b,,, is the extinction coefficient due to scatter, rZis wavelen~ 551

and Q is an

552

N.C. AHLQUIST~~~R.J.Oimwm

empirically determined expanent having a typical value of 1.3. Of course, in the absence of aerosol, ct = 4. The reddened appearance of high altitude, snow covered mount~ns (e.g., the Rocky fountains of North America) from high flying aircraft (CC. 10 km of altitude) at distances of several hundred km might well be an example of a color effect due to Rayieigh scattering if the intervening air is as aerosol-free as that observed by CHARLSON(~~~~) and RADKE and HOBBS (1969). The experimental measurement of either the wavelength dependence of the extinction coefficient or the color of objects viewed through air has usually been based on long path optical systems. For instance, MASAKI (1959, 1960) summarized three experimental approaches using photoelectric, photographic and visual telephotometry. MIDDLETON (1963)measured the wavelength dependence of “light haze” with a resultant bl = 2.09, BULLRICH'S(1964) value of about c( = 1 for a “heavy haze” is somewhat lower. MIRDLETON concludes that a value of 1.3 is appropriate for the typical haze. Smaller values are expected for fog where c( should approach zero. The theoretical approach to the wavelength dependence of extinction coefficient must necessarily be based on local properties of the scatterers in air (i.e., the aerosol) which are difficult to evaluate at a point much less over a long path. VAN DE HULST (1957), JUNGE(1963), and BULLIUCH (1964) developed a simple relationship between the size distribution of aerosol particles and the Angstrom exponent a. Junge”s power-law model was used to describe the size dist~bution over the range of sizes in the atmosphere : dN = Cr -p. d logr Here, w is the number of particles in the logarithmic radius increment d log r, C is a function of the total amount of aerosol and fl is an exponent with a value of about 3. Within certain assumptions, this derivation leads to a simple relationship between the exponent of the size dist~bution expression and the exponent of the wavelen~h dependence of extinction : a=@-2 Thus, information about a over a long atmospheric path would yield size ~s~bution information over the same long path and not at a point, as is usually used for direct measurement of aerosol properties. If it were possible to measure c1at a point, it would be possible to study the correlation with other aerosol characteristics, such as size ~~~bution, measured in the same air sample. The purpose of this paper is to present an instrumental approach to the measurement of a and to discuss the results of preliminary atmospheric mewre-

merits. THE

DIRECT

MEASUREMENT

Presuming that &gstrom’s expressed as :

0F THE

ANC~STRC)M

BXPoNENT

empirical expression is adequate, the expanent can be g=

-

d log Lt

dlogI

*

Therefore, the use of suitably narrow intervals of log A and log b,,,, would permit a

Measurement of the Wavelength Dependence of Atmospheric Extinction

553

direct measurement of c(.In reality, it is necessary to use finite log 1 intervals, thus, the instrumental design to follow is based on :

where Alog 2-0.05-0.1

is necessary to provide a measurable Alog b,,,,. THE

INSTRUMENT

FIGURE 1 is a functional diagram of the instrument based on this approach. The design and geometry of the lensless optical assembly is essentially the same as that of earlier instruments (AHLQUISTand CHARLSON, 1968). The unit is constructed in a 2 m

FIG. 1. Block diagram of multi-wavelength integrating nephelometer.

length of 15 cm aluminum pipe. Wavelength discrimination is provided by a set of four narrow band-pass (5-10 nm) interference filters mounted on a wheel in front of the multiplier phototube. The wheel is rotated by a 30 rpm motor through a flexible driveshaft. A two-beam system is used to remove the effect of variation of flashlamp intensity. The reference phototube, a type lP39 vacuum photodiode, is mounted in a box on the side of the main optical unit, and receives its light through a small opal-glass window in the side of the flashlamp housing. A second filter wheel is located between the reference phototube and the light source. This wheel is synchronized to the first filter wheel, and contains gelatin filters to roughly match the response of the reference tube to the multiplier phototube for each channel. The Ilash tube produces much higher light intensity than the one employed in earlier adaptation of the nephelometer. This is necessary because of the large amount of light lost in the optical filters and the low repetition rate for each channel (l/2 sec.). The lamp is a Sylvania type R4336 and is designed for service in airport runway approach lighting systems. Forced-air cooling is provided by a small, direct

554

Pi. c.

AEILWIST and R.

J.

t%ARLSON

centrifugal blower which also pulls the air sample through the sample chamber when the unit is operated at a fixed location. The eleetronies are built in two units to help isolate the signal circuitry from the qashfamp supply. The power supply unit contains a 12 V-1400 V square-wave converter and the flashlamp energy storage system. Resonant charging provides 240 V to an 8 MFD capacitor which delivers approximately 23 J/flash. This unit also contains the SCR for triggering the flashlamp, and supply 24 V d.c. for the bIower and 12 V 3OOOHz a.c. for power to the amplifiers. The me of 12-M V de. for primary power input has several advantages over 5.17V, 60 HZ operation. The power transformers are smaller and lighter, the high frequency of operation makes filtering easier, and for mobile operation no heavy inverters are needed When run in the laboratory, power is obtained from a 13 V regulated supply, making the instrument insensitive to line voltage fluctuations. Total power consumption at 13 V is about 90 W. The ampler unit co&&s the signal pressing circuits, the timing circuitry, the low voitage power supplies for the electronics, the high voltage supply for the multiplier phototube, and the inverter for the 30 rpm motor. The signal from the multiplier phototube, a current pulse of about 701.l~ duration, is integrated by an operational amplifier located next to the phototube socket. The integrating capacitor is shunted by a resistor to aBow the output to return to zero between flashes. The time constant is 1000 hsec. The reference phototube output is also integrated in a simple R-C integrator and is then fed to a logarithmic amplifier. A portion of the reference signal is subtracted from the sample signal to balance out the signal due to light scattered from the walls of the optical assembly. The signal pulse then drives another log amplifier. The log ampIif&s employ a transistor as a non-&rear feedback element around a~ operational amplifier. The transconductance of a hipolar transistor is one of the most predictable log elements known, a well-made device being accurate to 1 per cent over 5 or 6 decades of current, The log constant of this type of log generator is a function of the absolute temperature of the feedback transistor, so these transistors are located in an integrated circuit device (Fairchild 726 temperature stabilized pair) which contains an active tem~rature regulator that holds the chip temperature steady at X2@’C- Sti any zero offset in the feedback amplifiers could cause errors at low input Cur=% these amplifiers employ a form of chopper stabilization. The output of each an@ifier is sampled by a field-effect transistor gate just before each flash and the bias on the amplifier is adjusted for zero volts output. The required bias value is held in a capacitor between sampling periods so that the operational amphfier is always habancr-d. To prevent the log ampfifiers from pr~u~~g noise between &~&es when there 8fe no input signals, a current equivalent to about half scale output is fed to the inputs. These idling currents are cut off during the balancing and flash intervals by FIST gates. The overall accuracy of the log amplifiers is better thank 1 per cent over 3 decades of signal. Since the outputs of the amplifiers am equal to the logs of the aam& and reference signala, the ratio of these signals may be obtained by subtmrction. This k3aWm@M by another operational ampliGer which also raises the signal level to about 5 V for 2 decades of signal. The output of this difference amplifier is sampled by one of four FET gates and held in storage capacitors. The gates are timed ahout 200 Irrsiec,after the current

Measurementof the Wavelength Dependence of Atmospheric Extinction

555

flash so that the log amplifiers can settle to the correct values. The gates are open for about 100 hsec. The d.c. voltages are each passed through a single R-C time constant to four FET voltage followers which drive the recorder. The output is from zero to 5 V and the instrument is usually adjusted so that this corresponds to a range of 0.2-20x 10-4m-‘. A switch permits extending this range to 0.2-200x 10-4m-’ by reducing the gain of the difference amplifier. As many as seven outputs are possible, four log bscet (9 channels and three independent difference quantities or Angstrom exponents over separate wavelength ranges. The sensitivity of the multiplier phototube-filter-flashl~p combination must be the same for all channels, and the background balancing signal must be adjustable for each channel since the light scattered by the walls of the optical unit is not necessarily equal for all wavelengths. A background and a phototube high voltage (gain) adjustment are provided and are switched for each channel by FET and diode gates. The operation of these and the detector gates is synchronized to the filter wheels by mechanical contacts on the filter drive motor shaft. The exact timing of the log amplifier idling current gates, stabilizing choppers, the flashlamp trigger, and the detector gates is controlled by a multivibrator delay chain. The timing sequence is initiated by a photoresistor and a neon lamp shining through slots in the reference phototube filter wheel, so that the titers are properly aligned when a meas~ment is made. The high voltage supply for the phototube must be well regulated (0.01 per cent) to maintain adequate tracking between channels. The output of the supply must also be adjustable to four values switched by the channel selector switches. To avoid switching and regulating at high voltages, the supply employs a flyback type inverter which compares the current through a resistor from the high voltage output against the current through one of four gain controls from an 85 V regulator tube. The s~tching is performed by diode gates. An electronic calibrator is incorporated in the amplifier system to allow determination of the log constant for the particular recording system. A gate circuit produces a current pulse at the input of the sample log generator, The value of the pulse is determined by fixed resistors switched in a l-2-5 sequence from 0.2 to 200 ,uA, which normally corresponds to b,,, = 0.2-200 x 10-4m- ‘. The reference log generator idling current gate is left open to provide a steady reference signal. The filter wheel drive motor is a synchronous a.c. time motor and is powered by an inverter located in the amplifier unit. This inverter normally delivers 60 Hz a.c. but may be switched to 120 Hz output, allowing a faster instrument response time. An RCA type 6810A 14 stage multiplier phototube with an S-11 photocathode was used for the experiments outlined below. However, a more expensive S-20 photocathode would provide better signal to noise ratio in the red part of the spectrum and possibly operation at even longer wavelengths. INSTRUMENT

CALIBRATION

The primary standards of light scattering used in the calibration of the multiwavelength nephelometer were the light scattering of particle-free air, CO, and Freon-12. The values used were based on the same assumptions used earlier (CHARLSON, HORVATH and PUWCHEL, 1967). Although the accuracy of these values is not known absolutely, they are reproducible and therefore reliable as relative standards.

556

N. C. AHLQUISTand R. J.

CHAWSON

For this calibration, and for the preliminary data in the next section, the four channels were :

1. 2. 3. 4.

436 L-5 nm interference filter NO+ 5 nm interference filter ~600 130 nm combination of photographic filters Broad band neutral density filter, with a 420 run ultraviolet cutoff filter.

The first two wavelengths were chosen to provide a reasonable Alog b,,,, signal for the determination of the AngstrSm exponent, a. The 600 nm filter was broad enough to compensate for the decreased response of the S-l 1 photocathode while providing information in the red part of the spectrum. The last one was chosen to approximate the wavelength response of the earlier nephelometers. Other Wers could be used for other purposes. Several instrumental functions are illustrated in the calibration strip chart in FIG. 2.

0.2 ’

-

Time-----+

FIG. 2. Tracing of calibration chart. Abscissa is time, dm which four separate scattering tests were made, A-D. Qrdinate K&W are light scatteringeoefkient, logarithmicin units of lo-+m-‘, and the h#s* exponent a, from O-4. A. Clean air as scattering medium B. Clean CO, C. Clean Freon-12 D. Magnesium oxide surface Curves a-d are the different b,,.( outputs: a. 436mncham1el b. Broad bnndchme~ c. 5oonmchFlmlel d. 600~1 channel e. Angstr&ll exponent

Measurement of the Wavelength Dependence of Atmospheric Extinction

557

The abscissa was time, during which various functions were tested, listed under parts A-D. The two ordinate scales are bscat and CL.Parts A, B, C and D are scattering measurements with air, CO,, Freon-12 and a ma~esium oxide coated wire as the scatterer respectively. Curves a, b, c and d refer to the 436 nm, broad band, 500 nm and 600 nm channels respectively. Curve e (ol>is proportional to [fog bsoat(436 nm)log brcat (500 nm)] on a scale of zero to 4 units. Curve e under D serves as a zero for a while Freon-12 (e under C) provides an a reading of 4 for a Rayleigh scatterer. Although the absolute accuracy of the bsert calibration depends on calculated values for scatter by air, COZ and Freon-l& the relative accuracy of both bsfat and a records can be estimated. FIGURE3 shows a plot of the scale used for data reduction and the 30

/ o

20

Freon

12

A co2

i

m Air

/

6

0.2

1

0

40 20 60 80 % of Recorder Scale

IO0

FJCS.3. Results of CaIibmtionshown in Fro. 2. Scattering coeflicient is piotted as the logarithm vs. linear position on the recorder chart. The ordinate is light scattering coefficient (fO-4m-1), abscissa is per cent of recorder scale.

eight points for Freon-12, CO, and air. For values of b,,, larger than 1 x 10W4,the relative inaccuracy was less than 4 per cent. This amount of error could account for as much as about 0.03 in the value of Alog b,,,,, or about 0.5 in dlfor the present Alog 1 of 0.06, It is important to note that the m~s~ement of tr does not depend on a knowledge of the absolute magnitude of the Rayleigh scattering of air, COZ and Freon-12, but only on their relative ma~tude, their assumed ,Im4 wavelen~ dependence and the spectral whiteness of magnesium oxide, The relative magnitude of bscatfor the 3 gases was measured with an integrating nephelometer (C HARLSUN, HORVATHand PUESHEL, 1967). This measurement should be independent of the wavelength dependence of the broad band nephelometer since all three gases can be assumed to be Rayleigh scatterers. As a result, the accuracy of the ratios of light scattering coe5cient of the calibration gases depends only on the signal-to-noise ratio of the nephelometer.

558

N.C. AHLQUIST~II~R.J.CHARLSON

By refining this calibration procedure, it should be possible to improve the accuracy of m~surement of the Angstrom exponent. The limit of about 2 per cent (i.e., a signal-to-noise ratio of about SO)available in the measurement of b,,,, > 1 x 10V4m-” would yield an error of 0.02 in Alog b,,,, or about 0.3 in CLfor the present wavelength separation. The accuracy of both b,,,, and a measurements also depends on the geometry of the integrating nephelometer since it integrates over slightly less than the assumed O-n angular range. Further, due to limitations in real materials, the opal glass window probably deviates slightly from the ideal cosine scattering characteristic. Although the results below may be affected by these necessary imperfections, the errors incurred are probably of a systematic nature, thus allowing objective and reproduceable measurements. Further, any such errors are quite possibly small since the geometry deviates only slightly from ideal&y. Nonetheless, this point remains to be analyzed, an analysis that would be welcomed by the authors as a publication in the open literature. PRELIMINARY

DATA

Several weeks of continuous operation under both stable and unstable meteorological conditions have resulted in both general observations and a few specific events. For these measurements, the apparatus was located in the penthouse above More Hall on the University of Washington Campus. This site has been used extensively for urban

I

0.2 I I

0600

0700

1

0800

I

0900

Time, PDT, 3 Oct. 1968 FIG. 4. A typical sixth of mcwkd multi-wa-h li&t scwerh data. Also included is carbon mm~xide. TIC ~di~te soaleskrr mvm a, b and c arc 0.2-20 x lo-*m’ ‘. Curve d is a from0~;curvreisCOfromO_lOppan;a,bandcare436,UIO~~~~~~~~~Y~

Measurementof the Wavelength Dependence of Atmospheric Extinction

559

air chemistry measurements and does not seem to be dominated by local sources. The intake is about 20 m above the ground. The most obvious observation regarding the records of bScatas a function of time is that the wavelength dependence is nearly constant and that a is somewhat higher than expected. FIGURE4 is a copy of a section of strip chart including a rush hour CO peak of about 7 ppm and some variation of bscatwith almost no change in the Angstrom exponent. All three channels reflect even the short-term variations of the logarithm of light scattering. Further, as can be seen from FIG. 5, a log-log plot of b,,, VS.wavelength at the 3 wavelengths indicates that Angstrom’s power law approximation is

400

436

500 600 Wavelength, nm

700

Fg. 5. Several plots of b,,., versus wavelength, 1, both on logarithmic scales. One decade on the abscissa(d) is four times as long as a decade on the ordinate so that a Rayleigh scatterer (A-*) has unit slope. Abscissa units are in 11111, ordiik: in lo-+rn-‘. Days and times (PDT) are given, along with Freon-12 and air calibrations for slope reference.

valid, at least in this sort of cursory ex~ation. The record of the ~~~~rn exponent obtained from the difference of the logarithms of scatter at 436 and !I00 nm is nearly constant in time with a value of about 1.8. In cases where b,,, was dominated by aerosol scatter, a range 1.2< ax2.5 was observed. FIGURE6 includes two graphs from a period of meteorological stability followed by

0

I

l

“.

.

3

*.a

‘0

CY

.

l

_U.

9

i

: .

FIG. 6. The dependence

2

. . .

5

4

8 NUMBER OF CASES 6

0

I a

2

3

4

of the Angstrom exponent, a, on scattering coefficient, b,,,,, at 500 nm. for the periods 1600 1 October-1300 6 October and 1300 10 October-1600 11 October, 1968 (PDT). B. Histogram A. Hourly averages of a plotted versus b,,., of values of for values of b,,,, 1.7 x lo-4m-‘, i.e., for aersol scatter. The data periods were selected only to provide a wide range of values of b,,,,. The value for particle-free air and the error band are given on the abscissa of A.

0.6

0.8

1.0

2.0

6.0

8.0

10.0

Measurementof the Wavelength Dependence of AtmosphericExtinction

561

a frontal passage and good atmospheric mixing, FIGURE6A is a graph of CIas a function of b,,,, at 500 nm. The increase in a at low bseatcan be attributed to the increased importance of the Rayleigh scatter of air at these low values of scattering. Interestin~y, if L = > baayi+hP a appears to be almost independent of b,,,,. FIGURE 6B is a histogram of values of c( for b,,,, > 1.7 x 10-4m-‘, i.e., in those cases of FIG. 6A where the scattering by aerosol dominates the scattering by air molecules. About 90 per cent of the cases fall in a range 1.5 < c(e 2.2 with the distribution having a modal value of about 1.8. The data periods were chosen only to provide an adequate range of Lt ; however, no claim can be made for these data being typical of other atmospheric aerosols. It is important to consider the possibility of the influence of relative humidity. The instrument was located in a heated room (TN 15-WC) with ambient temperatures ranging from 0-20°C during the test period. As a result, small droplets might have been evaporated prior to measurement, changing both b,,,, and 8. A change in B would result in a change in a which could account for the somewhat high value of a since such evaporation would probably tend to deplete the larger size classes. Extensive experiments are planned both in Seattle and elsewhere to test the regularity of a and the effects of humidity. Several recognizable events occurred during this preliminary set of observations: at 1. Following a SW-NE wind shift at 0530 PDT on 4 October, 1968, the light ~tteri~~ffi~ent 436 run dropped from 6 x lo-*m- r to 0.9 x 10-4m-1 with a simultaneous increase in a from 1.9 to 2.7 as shown in FIG. 7. This increase in wavelength dependence was clearly due to the increased importance of Rayleigh scatter at the low levels of b,,.,. At 436 nm, the Rayleigh coefficient of air is about 0.5 x 10-4m-1, indicating that about half the scatter was due to air. half due to particles. Several peaks of light scattering occurred with no change in a. These were frequently associated with increased CO, indicating an automotive source. (See FIG. 4) This would imply that the sire distribution of automotive aerosols quickly becomes the same as other atmospheric aerosols or that the Seattle aerosoi is dominated by an automotive source. Occasional peaks or sharp variations in light scattering occurred with a simultaneous variation in a. Both increases and decmases in a could at times be observed to correlatewith the change in lit scatter. Wowever,no general trend in such correlations is yet apparent. Simultaneous measurement of the Aitken count with a General Electric condensation nucleus counter showed only isolated instances where a correlation existed either with light scatter or with a, SO* and a are apparently negativeIy correlated. On two occasions, (8 October, 1500-1800 PDT and 9 October, 08O&I100PDT) concentrations of SOI of about 0.1 ppm coincided with large valuesof b,,,, and an unusually low a of 1.2. RGURB 8 is a copy of one of the two light scattering and a records. Two possible explanations are that SO, has resulted ln an I&SO, aerosol or that the SO2 sourceis also a soof an aerosol, the aerosol in both cases having a different wa* length-scatterdenendencefrom that of urban haze. This result might be due to negative correlationofaandb ,Ca1,although the preliminary data OfFIG.6A do not so indicate.

2.

3.

4.

5.

APPLICATIONS

AND

CONCLUSIONS

The multi-wavelength adaptation of the integrating nephelometer makes possible the local m~surem~t of the wavelength dependence of the extinction coefficient due to scatter of atmospheric air. The major result of preliminary atmospheric measurements is that the wavelength dependence is relatively constant and in approximate agreement with krgstrtim’s empirical formula. These results also have application to several other problems.

E

A

562

N. C. AHLQUISTand R. J.CHARLSON

201 15IO: 86-

a

0.2’

I

0400

0500

0600 Time,

0700

0800

0900

PDT, 4 Oct. 1968

FIG. 7.Lightscattering variation during early morning wind shift. Scales are the same as Fig. 4.

20

1

15 1 49

IO:

86-

I

OJ

0.8-

0.3-1

I

I

0800

0900 Time,

PDT,

1 1000

II00

9 Oct. 1968

Fro. 8. Light scattering variation during pexiod with about 0.1 ppm Son. Fig. 7.

Scalesarcthesameas

The calibration of earlier integrating mpbelo~.~&~~(AHL,QU&W and CHARLSON, 1967, 1968; CHARLSON, JikBWATH and PLRWX&, 1967) was baaed on an assumed inmunental wavelength 4AamMistic. The eB42tive wavelcmg& was estimated to be 460 nm. The broad band channel results indicate that the effktive wavebgth of the

Measurement

of the Wavelength

Dependence

of Atmospheric

ExtMion

563

previous instruments was more nearly490-500 run for both Am4(Bayleigh) and AS2 (atmospheric) scatter. The Freon-12 part of the calibration (FIG. 2) wdti W@St fiat a correction factor of about 0.8 should be applied to the earlier rne~~men~ of bsEal.Besides this correction factor, it is possible to compensate for the mismatch between the human eye (maximum sensitivity at 550 mu) and the nephelometer. Another factor of 0.8 can be obtained by an interpolation of the record to 550 nm. Thus an empirical correction factor of around 0.64 appears to be in order for the interpretation of the earlier published values in terms of meteorologi~l =W?e, L,, via the Koschmieder formula: 3.91 L,==---. bscat

This refinement can be applied to the previously published estimates of the product of meteorological range and mass con~n~tion (CHARLSON, kbLQWST and HORVA’W 1968). The previous modal value of 1.2 gma2 is increased to 1.8, with a range from 0.9 to 3.6 for 90 per cent of the cases. However, since these results are based on a limited series of atmospheric measurements, it would seem appropriate to base any Gnal correction factors on more representative experiments. It must be emphasized that the broad band device does not measure exactly the same quantity sensed by the eye but that ~~~ti~ of aerosol properties permit a useful relationship to exist between the response of the eye and the instrument. The nearly constant value of a can be interpreted in terms of the previously mentioned Bullrich-Junge theory as indicating a constant-slope size distribution. However, this results in a value for the exponent, /?, of about 3.8 which is somewhat higher than that obtained by other means. The implied value of psr3.8 may be real since JUNGE and ABEL(1965) have given a range of 3 Q fi < 4 for a~o~h~c aerosols. The efkct of relative humidity may be important to a, but was not considered in this prelimmary run. Comparing the value of a = 1.8 to Middleton’s (1963)value of 1.3 observed in the 1930’swouId suggest that the Seattle aerosol in 1968 had a higher b than the earlier case. This is particularly understandable in view of the current air pollution control practice of removing only the larger particles from sources and in view of the increase in particles in smaller size classes due to the automotive source. The obvious answer to this question lies in the simultaneous measurement of a and /J and in improving the accuracy of the measurement of a. Nonetheless, the a% 1.8 and /I- 3.8 values seem high which will necessitate an investigation of gossible errors. The red-brown color of urban haxe might be attributable to a Am2characteristic. In comparison, the NO2 extinction coefficient has a AS2slope at 400 nm, Ae3 at 450 nm, and about Ae4 between 500 and 600 nm. Thus NO, would present a more vivid color than that due to wavelen8th dependent light scatter, but both would have the same minus-blue character. Experiments specifically designed to compare the color effects of aerosol and NO2 are possible and would settle this question. Finally, the wavelength dependence for this short period in Seattle appears to be Sweeney regular to permit the inference of scattering at any visible wavelen8th from measurement at one wavelength or over a broad band. This bears possible relevance to routine air pollution monitoring situations where so complex an instrument as this would be difkult to justify or maintain. Experiments are currently being planned to test the generality of this result in other locations and over longer periods.

564

N. C. AHLQUISTand R. J.

CHARLSON

Acknowledgement-This work was sponsored by the National Air Pollution Control Administration under Research Grant No. APOO336-05. The authors note the contribution of Dr. H. HORVATHwho emphasized the importance of a wavelength dependence. Both authors wish to thank Mrs. LIN AHLQUISTfor her assistance in drafting the figures. REFERENCES AHLQUISTN. C. and CHARISONR. J. (1967) “A New Instrument for Measuring The Visual Quality of Air,” J. Air Pollut. Control Ass. 17,467-469. AHL~IJISTN. C. and CHAarso~ R. J. (1968) “Measurement of the Vertical and Horizontal Profile of Aerosol Concentration in Urban Air with the Integrating Nephelometer”, Environmental Science and Tech. 2,363-366. hcjST~&d A. (1929) “On the Atmospheric Transmission of Sun Radiation and on Dust in the Air”, Geogr. Ad. 12,130. ~RKH K. (1964) “Scattered Radiation in the Atmosphere and the Natural Aerosol,” Adv. Geophys., 10,101-260. Academic Press, New York. CHARLSONR. J. (1968) “Atmospheric Aerosol Research at the University of Washington,” J. Air Polhd. Control Ass. 18,652-654. CURLSON R. J., AHLQ~~~TN. C. and HORVATHH. (1968) “On the Generality of Correlation of Atmospheric Aerosol Mass Concentration and Light Scatter,” Atmospheric Environment, 2, 455-w. CHARLWNR. J., HORVATHH. and Pusscxnr~ R. F. (1967) “The Direct Measurement of Atmospheric Light Scattering CoetBcient for Studies of Visibility and Pollution”, Atmospheric Environment, 1,469-478. JUNGEC. E. (1963) Air Chemistry andRadioactivity, Academic Press, N.Y. JUNGEC. E. and ABEL N. (1965) “Modification of Aerosol Size Distribution in the Atmosphere and Development of an Ion Counter of High Sensitivity,” Final Technical Report, U.S. Army Contract DA 91-591-EUC3484. Mm H. (1959) “Apparent Colors of Natural Objects,” Sci. Lt, Tokyo, 8,67-86. MASAKIH. (1960) “Apparent Colors of Natural Objects (II),” Sci. Lt, Tokyo, 9,39-54. MIDDLEIDN W. E. K. (1963) Vidon Through the Atmosphere, University of Toronto Press, Toronto. NICHOLSE. L. (1908) “Theories of the Color of the Sky”, Phys. Rev. 26,479-511. RADKBL. F. and HOBBSP. V. (1969) “Measurement of Cloud Condensation Nuclei, Light Scattering coefficient, Sodium containing particles and Aitken Nuclei in the Olympic Mountains of Washington.” J. Atmos. Sci. 26, 281-299. VANDE HULSTH. C. (1957) Light Scattering by SmallParticles, John Wiley, New York, 417-418.