Pyrheliometric determination of atmospheric turbidity in the harmattan season over Ile-Ife, Nigeria

RenewableEneryyVol. I, No. 3/4,pp. 555-566,1991

0960-1481/91 $3.00+.00 PergamonPressplc

Printedin GreatBritain.

TECHNICAL NOTE Pyrheliometric determination of atmospheric turbidity in the harmattan season over Ile-Ife, Nigeria Z . D , ADEYEFA a n d J. A . ADEDOKUN Physics Department, Obafemi Awolowo University, Ile-Ife, Nigeria

(Received 3 January 1990; accepted 2 February 1990) Abstract--Measurements of direct solar radiation intensity, using an Angstrrm compensation pyrheliometer carried out over three harmattan seasons (1985-1987) at Ile-Ife (7°29'N, 4°34'E), Nigeria, have been used to determine atmospheric turbidity based on five different models of turbidity viz: Schiiepp (B), Angstrrm (3), Kastrov (C), Unsworth (z~) and Linke (T). The five parameters indicate high aerosol loading of the atmosphere during the period and high correlation is established between them: (0.919 ~< r ~< 0.999). An inverse relationship has been noticed between horizontal visibility and atmospheric turbidity: ( - 0 . 8 0 < r ~< -0.76).

1. INTRODUCTION The determination of atmospheric turbidity parameters is very important in climatological studies and especially in atmospheric radiative model formulations [1]. In the West African region, the advent of the dust-laden harmattan winds is an annual event particularly between November and the following March. This period witnesses the invasion of the region by the north-easterly winds blowing from the Sahara, giving rise to high atmospheric turbidity and consequently, poor visibility. Many studies of the Saharan dust transport monitored over the Atlantic (e.g. [2, 3]), over Europe (e.g. [1, 4, 5]), around the Sahel region (e.g. [6-9]) and south of the region [10--16] have established the fact that the Sahara is a source region of millions of tons (per year) of atmospheric aerosol, mainly quartz, that have significant climatological effects on the solar radiation absorption over a wide area of the globe. The present study examines the variation in atmospheric turbidity due to the Saharan dust monitored through pyrheliometric measurement of direct solar radiation over Ile-Ife (7°29'N, 4°34'E), about 2000 km southwest of the source region, during the harmattan seasons of 1985-1987. A comprehensive examination of five different models of atmospheric turbidity : Schiiepp (B), Angstrrm (fl), Kastrov (C), Unsworth (zo) and Linke (T) show that the parameters were all unanimous in depicting high aerosol loading of the atmosphere over the region.

when the sun's disc was not covered by clouds. In particular, thin cirrus cloud if over the sun's disc, makes the readings unreliable as measurements so taken refer to a re-radiation or transmission of the sun's radiation by the cirrus intermediary. Three Schott filters: yellow (OG1), red (RG2) and dark red (RG8) were also used in the measurements whenever suitable conditions permitted. Each filter has a Davos certified reduction factor DR valid for filter temperature range I0 ° to 35°C. The measurements were carried out on a 20 m high building at Obafemi Awolowo University campus, Ile-Ife, Nigeria. Other conventional surface data taken before and after the measurements include dry bulb and wet bulb temperatures, horizontal visibility, cloud amount and cloud types. Apart from the fact that data were only taken when suitable weather conditions allowed, the unreliable power supply in the country (power cuts were not predictable) served as constraints on the amount of data obtainable. Most of the data were taken at 10 GMT (11 LST) although some special days existed when data were taken virtually round the dock. 3. TECHNIQUES FOR DERIVING ATMOSPHERIC TURBIDITY Solar radiation traversing the atmosphere experiences significant attenuation given by the Lambert-Beer-Bonguer law :

2. DATA COLLECTION An Angstrrm compensation pyrheliometer, calibrated against a standard pyrheliometer at the Swedish Meteorological and Hydrological Institute (SMHI), Norrkrping, Sweden (courtesy of Dr Lars Dahlgren) was loaned to us for the purpose of these measurements. The pyrheliometer was manually operated to make measurements of direct solar radiation during conditions

/ =

f:

I(2) d l = s-

;0

Io(2) exp (--ma(2)) d2

(1)

where I is the direct solar radiation intensity at the earth's surface, I(2) is the intensity of the normal incidence radiation at the observing station as a function of wavelength, 10(2) is the extraterrestrial normal incidence radiation intensity at the mean earth-sun distance similarly expressed as a function of wavelength, s = R2/R~ is the correction factor to allow 555

556

Technical Note

for the variation in the actual earth-sun distance R from the orbital mean, R,, and exp ( - ma(2)) is the transmission factor for solar radiation of wavelength 2. The total attenuation coefficient a(2) may be split into three separate components as follows : a(~) = a~ (4) + av(,D + a . , 0 )

(2)

where a s is attenuation o f solar radiation in clean dry air according to Rayleigh's theory o f scattering by air molecules, an is the attenuation by the atmospheric aerosol content and aw is the attenuation due to water vapour. Several models have been derived for parameterizing the turbidity o f the atmosphere. Five o f such models, described below, have been used in our computations : (a) Linke's turbidity coefficient T Linke's turbidity coefficient T, introduced by Linke [17], is a measure o f haze and water vapour content of the atmosphere. It is defined as : T = P(m))(log I o - - l o g / - - l o g S)

(3)

Although it has been shown that the coefficient C is not a unique value because o f its dependence on atmospheric air mass m and consequently on the height of the sun, Kondratyev [22] and others have noted that in the interval from m = 3 before noon to m = 3 in the afternoon, the coefficient C may be considered practically constant, independent of atmospheric air mass. Now, since most essential calculations of C are done within this interval when m ~< 3, the use of C for computing atmospheric turbidity is quite feasible.

(e) Unsworth turbidity coefficient % Unsworth [23] has suggested a measure o f atmospheric turbidity which is a function of atmospheric aerosol content only. This coefficient was defined by : % = --ml.' In (IS/l*)

(7)

where mh =cosecO is defined as the relative air mass, 0 is the apparent elevation of the sun above the horizon and I* is defined as the normal incidence irradiance (for mean solar distance) at the bottom o f an aerosol-free atmosphere which includes a specified amount of water vapour.

where

P(m) = (man(m) log e)- t

4. C O M P U T A T I O N O F TURBIDITY P A R A M E T E R S

and m is the absolute air mass number. (b) Angstrrm's turbidity coefficient [3 The attenuation o f solar radiation by aerosols has been represented by a turbidity coefficient [3 and a wavelength exponent • by Angstrrm [18, 19] thus:

ao(,D = [3,t-"

(4)

where 2 is wavelength in micrometres. The exponent c, is a measure of the particle size and it varies between ,, = 0 for very large particles (where scattering and absorption are independent of wavelength) and ~ = 4 for very small Rayleigh particles (gas particles), fl is a measure of the quantity o f haze suspended in the air. (c) Schiiepp turbidity coefficient B Sehiiepp [20] also developed a method which has the same theoretical basis as Angstr/~m's but he replaced [3 by B which refers to a decimal base. The relationship between [3 and B is given by the equation : exp-¢/a~ = 10-B/(2a)"

In W = 0.0279--0.6225td

or, if written in another form B = 82" log e.

(5)

Schiiepp's B refers to a wavelength 2 = 0.5 /am which represents the central part of the visible spectrum while Angstr6m's 8 refers to the wavelength 2 = 1.0 pro. (d) Kastrov turbidity coefficient C Kastrov [21] obtained a simple formula for the integral solar radiation at the surface level : S,, = S0/(l + Cm) (6)

where C is a quantitative characteristic o f atmospheric transparency, So is the extraterrestrial intensity of the normal incidence irradiance (solar constant) at mean earth-sun distance, and Am, the measured direct flux at the surface level.

(8)

where ta was expressed in °F. As confirmed by Adedokun [28] this technique is suitable for the harmattan season although a power law relationship has been found to suit the rainy season [28, 29]. Computation of Angstr6m's coefficient [30 was facilitated by application of Table 7 in the IGY instruction manual [24]. This table has been compiled for the shortwave component of radiation Ik(2 < 0.630 ,am) obtainable as the difference between the total direct radiation I, and the radiation ls measured after transmission through the red filter Schott RG2 thus :

1, = I,-DR21R

from which we obtain :

C ~ l/m(So/S,, - 1)

The measured intensities have been reduced to mean solar distance by the factor R 2/R~, using Table 3 in the appendix o f the IGY instruction manual [24] for the necessary percentage corrections. The Linke turbidity factor T was computed according to eq. (3) using values o f P(m) revised after Feussner and Dubois [25] and valid for 10 = 1380 W m - 2 available from Table 6 in the appendix o f the IGY instruction manual [24]. For the Schiiepp coefficients B, use was made o f the new tables compiled by Valko [26]. This is an extended table based on 2 < 630 nm which was applicable for values of B ,<.<0.700. The B values so computed were then used to obtain ,8 values from the relation in eq. (5) for days with B ~< 0.700. Strictly speaking, the table is valid for a mean precipitable water Wvalue o f 1 cm. (For other values o f W, a correction scheme attached to the table was followed.) The B values were based on a = 1.5. For our estimation of IV, in the absence o f radiosonde data, use was made o f the Ojo [27] technique o f deriving precipitable water from surface dew-point temperatures td:

(9)

where DR2 is the red filter reduction factor. Many occasions of//0 values that were higher than those obtainable from the said table were encountered. Derivation o f both//0 and B is followed by estimation of turbidity coefficients [3, 8ig,//,9 and 8,, facilitated by use of a table by Angstr6m [30]. Here 8, refers to the shortwave

557

Technical Note Table 1. Schiiepp and Angstr6m's turbidity parameters Month

Day

B

/3

to

oz

t,

/39

/3,

fl,g

fl//3o

Nov. 85 Jan. 86 Feb. 86

30 9 3 4 7 24 9 13 14 16 19 4 13

0.219 0.234 0.224 0.244 0.224 0.219 0.223 0.239 0.196 0.194 0.200 0.220 0.125

0.178 0.190 0.182 0.199 0.182 0.178 0.182 0.195 0.160 0.158 0.163 0.179 0.102

0.172 0.185 0.187 0.200 0.187 0.170 0.173 0.191 0.154 0.152 0.180 0.171 0.097

1.6 1.6 1.6 1.5 1.6 1.6 1.5 1.6 1.6 1.6 1.6 1.6 1.6

0.205 0.219 0.209 0.228 0.209 0.204 0.208 0.223 0.183 0.181 0,187 0.206 0.117

0.212 0.227 0.217 0.236 0.217 0.212 0.215 0.231 0.190 0.188 0.194 0.213 0.122

0.189 0.202 0.193 0.218 0.193 0.189 0.199 0.206 0.169 0.167 0.172 0.190 0.108

0.197 0.211 0.202 0.224 0.202 0.197 0.204 0.215 0.177 0.175 0.180 0.198 0.113

1.035 1.027 0.973 0.995 0.973 1.047 1.052 1.021 1.039 1.039 0.906 1.047 1.052

Mar. 86 Apr, 86

Jan. 87 Apr. 87

estimates made for Ile-lfe were compared with those obtained for Lagos radiosonde station some 200 km away and found to be realistic. Then, Unsworth's table of I* [23] as a function of both relative air mass rn, and the precipitable water content of the atmosphere was utilised for the calculation of ~a. For calculating the energy absorbed by water vapour, we have used McDonald's result [31] to obtain:

radiation derived from measurements with RG2 filter,/3a, to the short wave radiation given by measurement with the OG1 filter while/3,a and /3rr refer to bands defined as differences between RG2 and OG1 and RG8 and RG2 respectively. Table 1 gives the values obtained on some selected days. It is to be noted that although/30 derivation was based on a table built up under the assumption that ~ = 1.3, an extension of the table using ~ = 1.5 and 1.6 have been adapted for deriving B. We have here found this technique justified in that the ratio/3//30 is approximately unity in all cases treated in Table 1. Kastrov's turbidity coefficients, C, were calculated in accordance with eq. (6). The values of So were obtained from a table compiled by Kastrov for solar radiant fluxes as applicable to a dry and clean atmosphere [22]. The computation of Unsworth's turbidity coefficient, 3° requires the knowledge of precipitable water vapour IV. Hence, for reasons explained above, eq. (8) was used and

A = 106.26(Wm) °'3°

(10)

where W m is equivalent to the amount of precipitable water along the path of the sun's rays. 5. RESULTS AND DISCUSSIONS Shown in Figs 1, 2 and 3 are graphs of the daily variations of T, C, za indicating the pulsations in atmospheric turbidity and, consequently, harmattan intensity. The gaps evident on

/

0

/

m -J 7

5

3

I

I

15

23 Dec

I

31 ~

[

8

[

16 Jan.

I

i

24

I :~

i

9

I

17 Feb

I

L

25

5 :~

i

13 Mar.

[

21

I

29

I

6

I

14

Days

Fig. 1. Graph of daily variation of Linke's turbidity coefficient T for 1986/87 season.

I

22

1

30

[

8

558

Technical Note

i/~[1111¢//11~

4.C

/

j

•

~

/

/

>,

2.G

j

lI

1

15

:~

[

I

I

23 Dec.

8

31 :'.-

[

16 Jon.

I

I

24

/

I

I

I

I

9

~

1

17

I

25

Feb.

/

I

5

13 Mor.

:-~

I

2i

I

I

29 : ::

6

1

14 Apt.

I

22

I

I

30 8 = : : Moy !..I

Doys

Fig. 2. As for Fig. 1 but for Kastrov's turbidity coefficient C.

the diagrams are indicative of occasions when data could not be taken either due to power cuts or adverse weather conditions. Figure 4 shows the frequency distribution of B and p measured during the experiment. The mean and standard deviations of each of the parameters are as shown in the figure. To examine the comparison between the various turbidity models we have plotted in Figs 5-13 the graphs of

////

% against C, To against T and C against T as obtained over each of the seasons 1984/85, 1985/86 and 1986/87, respectively. It can be found that a high correlation exists between the various models indicating that they are unanimous in depicting a high aerosol loading of the atmosphere during the harmattan. In general the correlation coefficients (0.919 ~< r ~< 0.999) are all significant. We have shown the regression output for each pair of parameters below the figures.

/I

.~,/

,

I

t

E o

g

I 15 Dec.

I

I

I

I

I

I

I

23

31 =:-

8

16 Jan.

24

I ~11

9

I

I

17

25

Feb.

J 5 =::

. I 13 Mar.

I 21

I 29 ~:=

Doys

Fig. 3. As for Fig. 1 but for Unsworth turbidity coefficient ~a-

I

6

I

14 Apr.

I

22

I

I

30 8 =:~Moy--l-I

559

Technical Note 1.5 ¸

LC

~_.

B-0.411 o'B:O.151 ~ _

N

1.4 ¸

p. 1.3-

Brain = 0 . 1 2 5

12-

?

I

L

02

I

0.3

0.4

IZ'

I

I

0.5 8 -

0.6

#:o.s35 o-~=o.123 ,8,,0,:0.545

;o

Bm,o

]

8

:o,lo2

i

0.7

1.0

=

0.90.80.7

[

II

I I-3

13

I5

T

Fig. 6. As in Fig. 5 but for z, against T. Regression output : constant, -0.27567; Std err. of Y est., 0.009411 ; R squared, 0.997721 ; No. of observations, 23; Degrees of Freedom, 21 ; X coefficient(s), 0.099009; Std err. of coef., 0.001032.

4

5.1 Seasonal variation in turbidity 2

I

I

0.2

0.3

#

I

I

0.4

0.5

p

Fig. 4. A frequency distribution map of Schiiepp coefficient B and Angstr6m coefficient fl for the three seasons. The means, extreme values and standard deviations are as indicated. 1.5

p 1.5-

Monthly means (37) of the parameters C and T have been computed for each of the seasons as depicted in Table 2. Similarly computed were the standard deviation (a), the confidence interval (CI) and the confidence limits "g'LLand XvL (for the lower and upper confidence limits respectively). The 95% confidence level was employed, implying that 95% of the observations lie within the interval )?UL and XLL [32, 331. This procedure was repeated for parameters z, and the energy absorbed by water vapour A [eq. (10)] and the results shown in Table 3. Also in Table 4, we have similar calculations done for parameters c(o, B and ft.

42" 4.0 3.8 3.6 g

12'

3.4 3.2

~'° tl

3.02.8-

IO

262.4-

0.9,

2.220-

0,8

1816-

0712

14-

C Fig. 5. A plot of Unsworth turbidity coefficient z, against Kastrov turbidity coefficient, C for the 1984/85 season. The correlation coefficient r between the two and the regression parameters for the two are as shown. Regression output: constant, 0.437011 ; Std err. of Y est., 0.032326; R squared, 0.973119 ; No. of observations, 23 ; Degrees of freedom, 21 ; X coefficient(s), 0.257816; Std err. of coef., 0.009350.

121.0

i

i II

i

i 13

T

i

i 15

i

i 17

Fig. 7. As in Fig. 5 but for C against T. Regression output : constant, - 2.62696 ; Std err of Y est, 0.119556 ; R squared, 0.974884; No. of observations, 23 ; Degrees of freedom, 21 ; X coefficient(s), 0.374471 ; Std err. of coef., 0.013116.

Technical Note

560

1.70

4.C 1.5C 5.5 125

/

"1'~,,,,~ x ~ 0 0

i.OC

Q

0% o o _ x

×

3.0

o'~

/

c

x

xg

2.E

0.75 o ~ X ~ 2.( XO

0.50

l o

1.5 025

I 1.0

I 1.5

I 2.0

I 2,5

I 5.0

I 3.5

/

J 4.0

oo

C

8

6

Fig. 8. As in Fig. 5 but for Zo against C for the 1985/86 harmattan season. Regression output : constant, 0.311183 ; Std err. o f Y est., 0.076237; R squared, 0.958999; No. o f observations, 55; Degrees o f freedom, 53; X coefficient(s), 0.265159; Std err. of coef., 0.007531.

0

I0

12

14

16

le

20

T Fig. 10. As in Fig. 5 but for C against T for the 1985/86 harmattan season. Regression output: constant, -2.04498 ; Std err. of Y est., 0.271279; R squared, 0.961938; No. o f observations, 55 ; Degrees o f freedom, 53 ; X coefficient(s), 0.360526; Std err. of coef., 0.009850.

1.70 1.50

1.5(

,o.

A 1.2.~ --

1.25

~o L O 0 - 0.75

x~o o

I.OC

/

0.~ - -

/.

oo

0.5(

0.56

x

/ 025

O.2E

I

6

I

8

I

I0

I

12

I

T

14

I

16

l

18

I

20

Fig. 9. As in Fig. 5 but for ~° against T for the 1985/86 harmattan season. Regression output : constant, - 0.27583 ; Std err. of Y est., 0.014133; R squared, 0.998590; No. o f observations, 55 ; Degrees o f freedom, 53 ; X coefficient(s), 0.099461 ; Std err. o f coef., 0.000513.

I.O

I

1,5

I

2.0

I

2.5

I

3.0

I

3,5

J

4.0

C Fig. l 1. As in Fig. 5 but for Ta against C for the 1986/87 harmattan season. Regression output: constant, 0,261902; Std err. of Y est., 0.042068 ; R squared, 0.974680; No. of observations, 72 ; Degrees o f freedom, 70 ; X coefficient(s), 0.307671 ; Std err. of coef., 0.005926.

Technical Note

III

561

II +I

I~ ~

~

~

III

~

~'~ ~'~ ~,'~

--

--

I~ ~

I~'~ ~

~ - ~

o~T~ --

--

I~

+I

~ ~ -

+I

~ I ' ~ I ~'~I ~ ~'~I ~

o

I~

II +I

~m

+I-

+I

~~I"

~

I'~ ~ I'~ ~ ~ ~I- ~ ~

+I

~ ~ ~I" ~--I

--

--

+I-

~ ~ I~ ~

~ ~

I~I {'~I ~ ~ I ~ ~I~

+I

~ - ~,-i ~

i~

--

+I----

~

~

~

~.~i ~.% ~

+I

~i~ ~ .

~

¢~-~ ~

~

~

i~

o --

+I

--

~m

--

+I----

m

~ N 6 N 6 ~ N +I

•m. -

~

~ - ~ 6 - ~ 6 +I

. --. . . . . o. ~ -. . --

g~-~ N

~

o N 6 N - +I

.

. ~

+I

-H

. +I-

. --

i.

& +1

+1

+1

- H - -

+I-

[I --

--

8

+1

+1 m.

_=

~

~

I ~

--

~

+1

+I--

+I

+I

_ ~

0

,J

.1

.1

.1

~'~ ,.i

~

~

562

Technical Note

I I I

I I ~ , ~

III

II

~?,E~

<

+1

+1

+i

+1

+1

+1

+1

+1

+1

+1

+1

+1

+1

+1

..=

&

0 O.

o .,,.

0

( " I (",I r , l ~

0

.c~--.

..2 e~

._~

.~_

e0

2

Z

0

,J

,j

.J

.j

.a

Technical Note

563

+1

+1

I ~

~ o ~

~

o

~o

~

~

-H

+1

q-I

q-I

q-I

÷l

-H

q-I

-H

-I-I

q-I

+1

+1

q-I

-t-I

-H

q-I

+1

+1

+1

+1

o= ._=

-H

~

e~

o~

+1

L I

II

O

,ff

,.z

.~

,d

,~

,d

.d

564

Technical Note 1.7o

4.0

/

1.5o

x

,¢~0~~/x

1.25

,,~ ~ O y 5.0--

o

C z.5-

I.OC

/ o

0.75

x

OSC 0.2~

/

1.5--

/x I.G

I

6

I

I

8

i

I0

I

12

14

T

I

16

I

I

18

o.~

20

I

6

8

l0

12

T

I

14

i

16

I

18

I

20

Fig. 12. As in Fig. 5 but for r° against T for the 1986/87 harmattan season. Regression output : constant, --0.27003 ; Std err. of Y est., 0.016596; R squared, 0.996059; No. of observations, 72 ; Degrees of freedom, 70 ; X coefficient(s), 0.098402; Std err. of coef., 0.000739.

Fig. 13. As in Fig. 5 but for C against T for the 1986/87 harmattan season. Regression output : constant, - 1.63939 ; Std err. of Y est., 0.139107; R squared, 0.973112; No. of observations, 72; Degrees of freedom, 70; X coefficient(s), 0.312095; Std err. of coef., 0.006200.

The November and December columns in the 1984/85 harmattan season were left blank because our measurements commenced in January of that season. The data samples (N) were unavoidably small in some of the seasons due to the aforesaid observation constraints.

5.2 Atmospheric turbidity and visibility To investigate how turbidity varies with visibility, we have plotted the Linke's turbidity factor T against observed visibility in Figs 14 and 15 for the 1985/86 and 1986/87 seasons, respectively. An inverse relationship can be found between

oooT D,E:,O V(km)

A

-3o

15 -d/"

°\;[

A

:6 15

,,, A ~

P 20 ">

-\

I0

7 I-

I

I

15 23 Dec.

I

31 -::

]

I

8

16 don.

I

24

I

I /~1

I 9// 12 =',: Feb.,--.,,I1Doys

I

20 Mot,

I

28 ~::

I

5

I

13 Apr.

21 ~1

Fig. 14. The graph of the daily variation in T and the horizontal visibility for the 1985/86 harmattan season. A negative correlation is found to exist between them (r = -0.80).

Technical Note

565

GO o-o-o

T V (km)

55

I---

> L

/

l I I

i

40 ~ kT,

&

-J

9

I 20 3 I-

I

I

15

23 Dec.

I 31 • I•

I

I

L

I

8

16 Jan.

24

I ~: =

17 Feb,

25 iq~

I

I

I

5

13

21 Mor.

I 29 =:-

I 6

I 14 Apr.

I

I

22

30

8 = ; ~ ~,k3y,.-~

Doys

Fig. 15. As in Fig. 14 but for the 1986/87 harmattan season. Here also, a negative correlation (r = -0.76) has been found between T and visibility.

atmospheric turbidity and visibility ( - 0 . 8 0 ~< r ~< -0.76). The more turbid the atmosphere, the shorter the horizontal visibility. Poor visibility, often caused by thick harmattan haze, frequently results in disruption of aviation schedules and, sometimes, aviation accidents over the region.

6. CONCLUSIONS The time series of the atmospheric turbidity parameters, obtained here, clearly show the existence of harmattan 'spells' that occurred periodically during the season. This confirms the observations of Adebayo [34] and Adedokun [161. A strong association can be found between the various turbidity parameters (0.919 ~< r <~0.999), indicating that they all agree in depicting the nature and intensity of the harmattan. Not only do they indicate similar periods for the 'spell' occurrence but they also show that a high atmospheric loading characterizes the harmattan seasons. This agrees with the finding of Oluwafemi [12] who, using a Volz-type photometer, recorded a rise in atmospheric optical density over Lagos during the harmattan months of December 1977 and January 1978. In a similar manner, Brinkman and McGregor [8], using an Eppley normal incidence pyrheliometer, recorded high values of an integral turbidity z (first used by Unsworth and Monteith [35]) in their measurement at Zaria, in northern Nigeria. The mean value of the wavelength exponent, c(, which is a measure of particle size has been shown in Table 4 to be between 1.5 and 1.6, the latter being more predominant for many months. This value is an indication of the fact that the nature of the harmattan dust monitored over the station is predominantly small. This corroborates the results of Adedokun [16]. Our results have shown that a negative but high correlation exists between atmospheric turbidity and horizontal visibility

( - 0 . 8 0 <~ r ~< -0.76). This is in agreement with the results of D'Almeida [9] who obtained a negative but high correlation r ~ -0.96 between ~ and visibility in his measurement carried out over the Sahel region. The monthly and seasonal means of atmospheric turbidity showed that the 1984/85 harmattan season was the most turbid while the 1985/86 season was the least. The months of December and January were shown to be more turbid than the other months in the season while April was shown to be the least turbid. This is in line with usual expectation as the rains often begin again in April following the dry season. Although the scope of this study has been somewhat limited by data availability, we believe that the results have given some illumination on the nature and variability of atmospheric turbidity in the harmattan season in the region and its crucial effect on horizontal visibility.

Acknowledgements---The Angstr6m compensation pyrheliometer used for our measurements has been loaned to us by the Swedish Meteorological and Hydrological Institute (ISMHI) (courtesy Drs Lars Dahlgren and Bjorn Holmgren). We thank the International Science Programme Uppsala University for support in air-lifting the equipment to Nigeria. The assistance of the Nigerian Meteorological Services in supplying us some radiosonde data is acknowledged. We thank the director and Mr I. Idowu. Computer assistance of Mr T. Kpohraror and the cartographic aid of Mr Bisi Bayewu are similarly appreciated. The contributions of Messrs O. A. Olaniran and M. A. Olajire who assisted with the collection of data are duly acknowledged with thanks. One of us (JAA) thanks Prof. Abdus Salam of the International Centre for Theoretical Physics for hospitality at the centre, and SAREC for financing his second visit as an associate to the centre during which period work on this paper was completed.

566

Technical Note REFERENCES

1. C. Tomasi, F. Prodi, M. Sentimenti and G. Cesari, Multiwavelength sun-photometers for accurate measurements of atmospheric extinction in the visible and near-IR spectral range. Appl. Optics 22, 622 (1983). 2. L. Shutz, Long range transport of desert dust with special emphasis on the Sahara. Ann. N.Y. Acad. Sci. 338, 512-532 (1980). 3. L. Shutz, R. Jaenicke and H. Pietrek, Saharan dust transport over the North Atlantic Ocean. Spec. Pap. geol. Soc. Am. 186, 87-100 (1981). 4. F. Prodi and G. Fea, A case of transport and deposition of Saharan dust over the Italian peninsula and southern Europe, J. Geoph. Res. 84, 6951-6960 (1979). 5. F. Prodi and C. Tomasi, Sahara dust program--I : the Italian network of sun-photometers. Extinction models based on multimodal particle size distributions. J. AerosolSci. 14, 517-527 (1983). 6. A. E. Kalu, The African dust plume : its characteristics and propagation across West Africa, in: Saharan Dust (Edited by C. Morales), Chap. 5, pp. 95-118. Wiley, New York (1979). 7. G.H. McTainsh, Harmattan dust deposition in northern Nigeria. Nature 286, 587-588 (1980). 8. A. W. Brinkman and J. McGregor, Solar radiation in dense Saharan aerosol in northern Nigeria. Quart. J. Roy. Met. Soc. 109, 831-847 (1983). 9. G. A. D'Almeida, A model for Saharan dust transport. J. Clim. appl. Met. 25, 903-916 (1986). 10. R. A. Hamilton and J. W. Archbold, Meteorology of Nigeria and adjacent territory, Quart. J. Roy. Met. Soc. 71,231-265 (1945). 11. E. E. Balogun, The phenomenology of the atmosphere over West Africa. Proc. Ghana Scope's Conf. on Environment and Development in West Africa. Ghana Academy of Arts and Sciences, pp. 19-31 (1974). 12. C. O. Oluwafemi, Preliminary solar spectrophotometric measurements of aerosol optical density at Lagos, Nigeria. Atmos. Environ. 13, 1611-1615 (1979). 13. C. O. Oluwafemi, Some measurements of the extinction coefficients of solar radiation in Lagos. Pageoph. 118, 775-782 (1980). 14. C. O. Oluwafemi, On spectral extinction of solar radiation within a tropical atmospheric boundary layer. Pageoph. 119, 831-840 (1981). 15. J. A. Adepetu, O. I. Asubiojo, F. Y. Iskander and T. L. Bauer, Elemental composition of Nigerian harmattan dust. d. Radioanalyt. Nucl. Chem. Articles 121, 141-147 (1988). 16. J. A. Adedokun, W. O. Emofurieta and O. A. Adedeji, Physical, mineralogical and chemical properties of harmattan dust at Ile-Ife, Nigeria. Theor. appl. Climatology 40, 161-169 (1989).

17. F. Linke, Transmission koeffizient und Trubungsfactor. Beitr. Phy. Freien Atmos. 10, 91 (1922). 18. A. K. AngstrSm, On the atmospheric transmission of sun radiation and on dust in the air. Geog. Ann. 11, 156166 (1929). 19. A. K. AngstrSm, On the atmospheric transmission of sun radiation II. Geog. Ann. 12, 130-159 (1930). 20. W. Schiiepp, Die Bestimmung der atmosph~irischen Trubung ans aktinometer messungen. Arch. Met. Biocl. Geoph. Ser. B1, 257-346 (1949). 21. V. G. Kastrov, On the basic actinometric formula. Met. Bull. No. 7 (1928). 22. K. Y. Kondratyev, Radiation in the Atmosphere, pp. 277-280. Academic Press, New York (1969). 23. M. H. Unsworth, Variations in the short wave radiation climate of the U.K. Int. Solar Energy Soc. Conf. on U.K. Meteorological Data and Solar Energy Applications, Royal Institution, London (1975). 24. CSAGI (Comit6 Special de l'Ann~e Gtophysique Internationale), Radiation Instruments and Measurements, Part 4, IGY Instruction Manual, pp. 398-407, 459-466. Pergamon Press, Oxford (1958). 25. K. Feussner and P. Dubois, Trubungsfactor, precipitable water. Taub. Gerlands Beitr. Z. Geophys. 27 (1930). 26. P. Valko, New tables for computing the turbidity coefficient B from actinometric measurements in the spectral range 2 < 630 /zm. Atmospheric Radiation 551.521.3, No. 13 (1971). 27. O. Ojo, The distribution of mean monthly precipitable water vapour and annual precipitation efficiency in Nigeria. Arch. Met. Geoph. Biocl. Ser. B, 18, 221 (1970). 28. J. A. Adedokun, Surface humidity and precipitable water vapour linkage over West and Central Africa: further clarification and evaluation of existing models. Int. J. Climatology 9, 425-433 (1989). 29. J. A. Adedokun, On a relationship for estimating precipitable water vapour aloft from surface humidity over West Africa. J. Climatology 6, 161-172 (1986). 30. A. AngstrSm, The parameters of atmospheric turbidity. Tellus XVI, 1, 64-74 (1964). 31. J. E. McDonald, Absorption of solar radiation by atmospheric water vapour. J. Met. 17, 3 (1960). 32. H. L. Balsey, Introduction to Statistical Methods (lst edn), pp. 116-125, 170-186, 251-255. Littlefield, Adams & Co., Totowa, New Jersey (1972). 33. A. E. Waugh, Elements o f Statistical Methods. McGrawHill, New York (1952). 34. S. I. Adebayo, Pronounced haze spell over Nigeria: 211 March 1977. Proc. P R E - W A M E X Symp. on the West African Monsoon (Edited by D. O. Adefolalu), pp. 270276. Leo Express Printers, Lagos, Nigeria (1979). 35. M. H. Unsworth and J. L. Monteith, Aerosol and solar radiation in Britain. Quart. J. Roy. Met. Soc. 98, 778797 (1972).