INTERNATIONAL JOURNAL OF CLIMATOLOGY, VOL. 17, 725±744 (1997) WINTER DUST HAZE IN TROPICAL SOUTHERN AFRICA ANNICK DOUGUEDROIT1* AND FRANCËOIS BART 1 CAGEP (CNRS), Institut de GeÂographie, UniversiteÂ de Provence (Aix-Marseille I), 29 avenue Robert Schuman, 13621 Aix-en-Provence Cedex, France. 2 DYMSET (CNRS), Institut de GeÂographie, UniversiteÂ Michel de Montaigne (Bordeaux 3), Domaine Universitaire, 33405 Talence Cedex, France Received 6 June 1994 Revised 25 August 1996 Accepted 2 September 1996 ABSTRACT Dust haze spreads over most of tropical southern Africa in winter, from May to October, with a maximum in July and August over the ZaõÈre Basin. Its origin is primarily biomass burning and secondarily soil de¯ation, which greatly increases during the dry season. It is mobilized on the continent by the unstable atmospheric boundary layer up to the inversion in the Indian trade wind at about 2000 m altitude. Then it is stretched north-westward by the trade wind and accumulates in the mid-troposphere over the ZaõÈre Basin and the Congo Republic. It is carried away over the Atlantic Ocean from the coast of those countries. It begins to disappear only when synoptic-scale ¯ow patterns change in September. Measures of atmospheric pollution due to biomass burning over the African continent and of atmospheric aerosols above the Atlantic Ocean con®rm these results. # 1997 by the Royal Meteorological Society Int. J. Climatol. 17: 725±744, 1997. (No. of Figures: 10. KEY WORDS: No. of Tables: 1. No. of Refs: 52.) dust haze; biomass burning; atmospheric pollution; southern dry season; tropical southern Africa. 1. INTRODUCTION Dust haze has been de®ned according to Villeneuve (1980) as `Haze constituted with dust particles . . . which are dry and so tiny that it is not possible to feel them or to see them with the naked eye, but which, as a whole, make the air looking dim and opalescent'. Such an opalescence is due to the optical properties of light. Dust haze has been studied in several areas of the world, the best investigated being the Sahara (Buecher, 1986; Carlson, 1979; Cerf, 1980; Druilhet and Durand, 1984; Fouquart et al., 1987; Morales, 1979). Its presence has been noticed in a few zones of southern Africa (Crabbe, 1986; DougueÂdroit et al., 1990) but no general study has been conducted in that area. Atmospheric dust is indirectly concerned with recent research on biomass burning because smoke particles constitute a part of the dust aerosols. Produced simultaneously with trace gases they modify the chemical content of the atmosphere, contribute to its pollution and participate in the biochemical cycles (Goldammer, 1990; Andreae, 1991). Such gases due to biomass burning become indicators of the origin, at least partly, of dust haze. The distribution in space and the evolution in time of dust haze in tropical southern Africa from May to October, i.e during the dry season and the months transitioned to the wet season, are ®rst discussed. They are characterized, for 2 months or more, by a remarkable persistence, which has not been described before anywhere in the world. It is linked with the particular features of the seasonal synoptic-scale ¯ow patterns above the southern African continent. *Correspondence to: A. Douguedroit, CAGEP (CNRS), Institut de Geographie, Universite de Provence (Aix-Marseille I), 29 avenue Robert Schuman, 13621 Aix-en-Provence Cedex, France CCC 0899-8418/97/070725-20 $17.50 # 1997 by the Royal Meteorological Society 726 A. DOUGUEDROIT AND F. BART 2. DUST HAZE, A CLIMATIC CHARACTERISTIC OF THE SOUTHERN WINTER 2.1. Data The statistical basis of this study is the daily data which have been collected all over the world for the 18 months of the International Geophysical Year (IGY), from 1 July 1957 to 31 December 1958, and gathered on daily maps by the Deutscher Wetterdienst. The 80 stations or so scattered through southern tropical Africa have been used here (Figure 1). The data are derived from the symbols for `dust haze' collected from the daily weather maps, and conform to the usual standards of meteorological services. Of course, dust visibility is a semi-qualitative notion; moreover it concerns only low altitude dust clouds, which seems fortunately to be an usual case over the continents where the dust originates. These data are at the moment the only daily ones easily available for the whole southern tropical Africa. Remote sensing studies could provide daily data on the spatial extension of dust haze either through the visible channel, but without any information about the height of those clouds, or by processing the data of the thermal infra-red (IR) channel. Up to now, this method, when applied over continents, is operating only on such simple cases as the Sahara (Legrand et al., 1994; N'Doume and Legrand, 1995). The symbol `dust haze' is different from `smoke', which has been left out. The smoke columns that rise from the great ®res have not been taken into account. Their contribution to the physical and chemical characteristics of the atmosphere is another question, which will be introduced later. The number of stations with the symbol `dust haze' has been calculated for every day in the whole of southern tropical Africa, between 16 S and 4 N, which are the extreme latitudinal boundaries of the extension of the winter dust haze. In this way, 80 stations were checked systematically during both the winter dust haze periods, from 1 July 1957 to 30 November 1957 and from 1 May 1958 to 30 November 1958. These counts made it possible to carry out an elementary statistical analysis of the results and to produce some maps showing the spatial evolution of this phenomenon during winter. Figure 1. Location of the stations in southern tropical Africa WINTER DUST HAZE IN SOUTHERN AFRICA 727 2.2. Dust haze occurrence and dry season 2.2.1. Southern dry season. Dust haze occurrence is closely linked to the southern dry season (Figure 2): in most of the stations very little rain is found during the period stretching from May±June to September±October (Walter and Lieth, 1967; Grif®th, 1972). The only exception is the eastern coast, where rain continues falling but in smaller amounts than during the rest of the year. Elsewhere, the rainy season generally stops in autumn, either in April when precipitation falls under 40 or 20 mm or in May when it is nil, especially in the south (Figure 3). In Shaba (southern ZaõÈre), the complete cover of thin clouds without any rain, which is characteristic of April, quickly disappears in the beginning of May (Mbenza, 1982; Crabbe, 1986; N'tombi, 1988, 1990). The area without any rainfall increases until July; the period July±August corresponds to its greatest extent, then it diminishes in September (Figure 3). The winter dry season on average lasts from 3 to 6 months in the north and up to 9 or 10 in the south (Dubresson et al., 1994) (Figure 2). The same regime extends to the Great Lakes area in East Africa, but the dry season is generally shorterÐonly 3 months, from June to August (Douguedroit et al., 1990). On the northern limit of the ZaõÈre basin, winter is less dry but precipitation distinctly weakens; the same climatic regime continues north of the Equator, with a short dry season from June to August, corresponding with the northern summer (Leroux, 1983). 2.2.2. The southern dry season in 1957 and 1958. 1957 and 1958 are generally considered as humid years in southern Africa as well as in western Africa (Nicholson, 1986, 1988; Tyson, 1986). In southern tropical Africa it is possible to compare the average rain amount (1946±1987) and the rain in 1957 and 1958, during the southern dry season at four stations (Table I). From these rain amounts we may easily deduce the differences between the stations. Both years were below average in Lusaka, above average in Mahalapye, and close to average in Bulawayo. In Solwezi, the 1957 dry season was below average and 1958 was above average. In fact the main Figure 2. Length of dry season (after Dubresson et al., 1994): 1, less than 3 months; 2, 3±6 months; 3, 6±9 months; 4, more than 9 months; 5, alluvial high plains of southern tropical Africa 728 A. DOUGUEDROIT AND F. BART Figure 3. Extension of drought in April and October: 1, area with a rainfall average lower than 30 mm in April; 2, station with less than 30 mm in April 1958; 3, area with a rainfall average lower than 30 mm in October Table I. Rain amount (in mm) during the dry season Stations Solwezi (12 S) Lusaka (15 S) Bulawayo (20 S) Mahalapye (23 S) May±November June±August 1946±1987 1957 1958 1946±1987 1957 1958 241 108 147 126 148 42 138 182 333 90 121 162 1 0 4 7 0 0 0 52 0 0 0 1 interannual variability originates in the length of the dry season. From June to August, the weather is absolutely dry, so that the variability is equal to zero (Hulme, 1992). The only exception is July 1957 in Mahalapye, located on the fringe of the tropical zone, with the heaviest winter rainfall in the period 1946±1987 (41 mm). The beginning and the end of winter may be either quite rainy or quite dry, depending on both place and year. In 1958 for example, Mahalapye did not receive any rain during October and November (average 1946±1987: 101 mm), whereas Solwezi received more than usual (278 mm, average 230 mm). Therefore we may suggest that the interannual variability of the dry season concerns mainly the beginning of the extension phase and the end of the regression phase of the chronological development of dust haze. The latter will be studied for both 1957 and 1958. 2.3. Chronological development of the dust haze in 1958 Dust haze weather prevails during winter in most of southern tropical Africa, especially over the inner plateau. Considering the daily number of stations affected by dust haze throughout 1958 (the only year with data from the beginning to the end of the winter dust season), we can display four main chronological phases (Figure 4). WINTER DUST HAZE IN SOUTHERN AFRICA 729 Figure 4. Daily number of stations affected by dust haze: (a) 1 May±30 November 1958; (b) 1 July±30 November 1957 2.3.1. Extension phase. This stretches from 6 May to 22 June. During this 7-week period, the number of stations affected by dust haze increases quite regularly, in spite of some ¯uctuations: ®ve on 11 May, 10 on 13 May. From 12 June to 21 June, the number of stations with dust haze ranges from 17 to 26. June 22 is one of the highest peaks, with 34 stations involved. The average values for the successive ®ve 10-day periods (or decads) are 5, 10, 13, 16, and 24 stations. 2.3.2. Main phase. This takes place from 22 June to 30 August. It is characterized mainly by both a great number of stations affected daily by dust haze (minimum 16, maximum 36) and many ¯uctuations according to short and long cycles, so that we may differentiate decreasing phases (22 June to 2 July, 24 July to 11 August) and increasing phases (2 July to 24 July, 11 August to 19 August). The average values of successive decads are about 25 stations. 2.3.3. Regression phase. From the beginning of September to 20 October, the number of stations affected by dust haze clearly decreases (from 25 to 8), despite some ¯uctuations, especially at the end of September, with a secondary peak of 24 stations on 1 October. 2.3.4. Relic phase. From the second half of October, very few stations are affected, often less than ®ve. At that time, dust haze scarcely occurs; it completely disappears in December. 730 A. DOUGUEDROIT AND F. BART A Figure 5. Monthly spatial extension of dust haze in 1958 (May±November). (A) Monthly mapsÐnumber of hazy days for each month: 1, 28± 31; 2, 21±27; 3, 12±20; 4, 7±11; 5, 1±6. (B) Map for the whole period (May±November 1958)Ðnumber of hazy days: 1, more than 100; 2, 71± 100; 3, 51±70; 4, 31±50; 5, 11±30; 6, 1±10 2.4. Spatial extension of the dust haze in 1958 The areas affected by dust haze during the dry season cover most of southern tropical Africa. The various chronological phases are characterized by speci®c regional features. The ®rst stage of this study is based on 1958 monthly data: each map represents the spatial distribution of the stations affected by dust haze according to the number of hazy days for each monthÐfrom May to NovemberÐand for the whole period (Figure 5 (A and B)). 2.4.1. Geography of dust haze extension. From January to April 1958, there is only very occasional dust haze at a few stations in coastal south-western Africa and the southern ZaõÈre Basin. We can also note some occasional dust haze in northern parts of the ZaõÈre Basin and in the Lake Victoria region, associated with the northern hemisphere main dry season (Leroux, 1983). 731 WINTER DUST HAZE IN SOUTHERN AFRICA A B A Figure 5 (continued). From 6 May to 28 May, dust haze settles down mainly over the southern regions of the ZaõÈre Basin, on the northern boundary of the southern Africa plateau. For this whole month, the core of the dust haze area consists of the inner regions eastward from Luanda, along the Angola±ZaõÈre border: for example, Malanje (Angola) was continually subjected to dust haze from 9 May. Most of the stations scattered throughout Angola and ZaõÈre between the 5 S and 10 S parallels were affected during at least 12 days in the month. During the following days, the dust haze area spreads irregularly towards the upper Zambezi Basin from western Zambia to Shaba. From mid-June, there is a great northward and eastward extension: at the time of the solstice, the dust haze extends from the Atlantic coast to Lake Victoria through the rainforest regions and savanna plateau of ZaõÈre, Angola and western Zambia. 2.4.2. Main dry season and maximum dust haze area. For nearly three months, from the end of June to midSeptember, dust haze is very common in most of the inner regions of southern tropical Africa. 732 A. DOUGUEDROIT AND F. BART The core of the area is located mainly in the ZaõÈre Basin south of the Equator. In July it spreads from the mouth of the ZaõÈre (Matadi) eastward to the shores of Lake Tanganyika (Kalemie): this wide area becomes hazy for 30 or 31 days during the month. From there dust haze stretches in three main directions. It extends to the Great Lakes of eastern Africa, especially south of Lake Victoria: in Tanzania, Tabora is affected on 24 daysÐand continually from 9 JulyÐMoshi, close to the Kilimanjaro, on 18 days. It reaches the mountainous regions of the western Rift Valley (Kabale, in south-western Uganda, 26 days). It extends to northern Botswana (Maun, 24 days). Occasionally, the dust haze may extend beyond the Equator (Lisala, at the top of the ZaõÈre bend, 19 days), and even reaches the shore of the Indian Ocean in Kenya (Mombassa) and Tanzania (Dar es Salaam). It pushes toward the south-eastern inner regions, especially from Malawi (Lilongwe, 7 days) down to Zimbabwe (Harare, 7 days), without reaching the Mozambique coast. In August the outline of the dust haze area is quite similar; however, its area is a little wider whereas the number of stations affected during the whole month is fewer. In fact, a slight southward drift is noticeable; close to the Equator, dust haze begins to decrease: only 9 days in Kisangani (21 days in July), 16 days in Kabale (26 days in July), 20 days in Matadi (30 days in July); by contrast it increases on both sides of Zambezi (30 days in Maun, 13 days in Lusaka). Some very occasional thunderstorms occur in the ZaõÈre Basin in August. This evolution becomes more and more obvious in September, when the core of the dust haze moves from the southern part of the ZaõÈre Basin to the Zambezi Basin. By slow degrees, especially from the second decad of this month, rains and storms take the place of the dust haze north of the 10 S parallel. 2.4.3. South-eastward regression. The south-eastward regression takes place mainly from mid-September to November: dust haze progressively leaves the ZaõÈre Basin (October), becomes concentrated in the lower Zambezi Basin (November), and ®nally disappears (December). The September map thus displays transitional features: dust haze area spreads over both the Zambezi Basin and the plateau stretching from northern Angola to Lake Tanganyika. The maximum number of hazy days occurs in Maun (30 days), but in some other scattered towns there were 21±25 hazy days: Malawi±Zimbabwe area, Mwinilunga (northern Zambia, 11 S), Nova Lisboa (Angola, 12 S) and Kalemie (ZaõÈre, 6 S). At most of these stations, dust haze temporarily vanishes around mid-September when the ®rst stormy stage of the southern rainy season occurs. Closer to the Equator, it does not come back later on: 17 September in Kikwit (ZaõÈre, 5 S), 13 September in Luluaburg (ZaõÈre, 6 S), 12 September in Kabale (Uganda, 1 S). In contrast it comes for the ®rst time to Mzimba (Malawi, 12 S) on 10 September, at the end of the dry season. In October, dust haze withdraws from the ZaõÈre Basin: no more than 10 hazy days in Kamina (ZaõÈre, 9 S) or Nova Lisboa, 8 days in Kalemie, 7 days in Lubumbashi (ZaõÈre, 12 S), all before 15 October; less than 5 days elsewhere. Most of the stations with more than 10 hazy days are situated in the Zambezi Basin, south of 12 S. This area stretches from 20 E (Maun, 20 days, most of them before 20 October), to Lake Malawi (Mzimba, 28 days), through Zambia (Ndola, 16 days) and Zimbabwe (Harare, 16 days): sometimes dust haze may reach the Mozambique coast (Quelimane, 4 days). In November, the last spots of the southern dust haze may be observed close to Lake Malawi (Mzimba, 11 days until 12 November) and in the lower Zambezi valley (Tete, 12 days, until 26 November). At the same time, boreal dust haze is coming back in the Equator region. 2.5. Comparison 1957±1958 We found it interesting to compare the chronological and spatial processes of dust haze in 1957 and 1958, by using July±November daily data. The evolution of dust haze is globally similar in both years; only some slight differences may be detected, mainly in September and October, when for three decads, the number of stations affected by haze is much greater in 1958 (25, 17, 15) than in 1957 (18, 11, 11). 2.5.1. Chronological processes. The comparison of the second half-year data only allows the maximum extension and the regression phases of dust haze to be taken into account (Figure 4). WINTER DUST HAZE IN SOUTHERN AFRICA 733 Figure 6. Monthly spatial extension of dust haze in 1957 (July±November)Ðnumber of hazy days for each month: 1, 28±31; 2, 21±27; 3, 12±20; 4, 7±11; 5, 1±6 734 A. DOUGUEDROIT AND F. BART There is a great similarity of the July±November sequence of both 1957 and 1958: the maximum extension of dust haze takes place in July and August; from the beginning of September to mid-October, the number of hazy stations decreases; ®nally, there are only very few left, until the end of November. Only slight variations are obvious: from the last days of August, dust haze withdraws more quickly in 1957, especially during the ®rst half of September, and from 22 September to 10 October. The most noticeable difference can be discerned around 26 September, with half as many stations affected in 1957. Those variations are connected with some change in spatial processes. 2.5.2. Spatial processes. The monthly maps representing the dust haze area throughout the period July± November 1957 (Figure 6) displays both common features and differences between 1957 and 1958. In July 1957, the hazy area seems to be more compact than in July 1958. Particularly, it does not stretch eastward beyond Lake Victoria and Lake Tanganyika: Moshi, near to Kilimanjaro, is not affected at all (18 days in 1958), the same as the Indian Ocean coast. The core of the dust haze area is rather more southward orientated, from the ZaõÈre basin, to the upper Zambezi region; Serpa Pinto (Angola, 15 S) gets 30 hazy days (only 13 in 1958), Bulawayo (Zimbabwe, 20 S) 16 (none in 1958). The subsidiary nucleus centred on the Lake Malawi western region, which appears about 10 September 1958, is already obvious in July 1957: 16 hazy days in Mbeya (Tanzania, 9 S), 20 in Mzimba (Malawi, 12 S). There is no signi®cant evolution in August, so that the same differences between 1957 and 1958 may be pointed out: many more hazy days in the southern inner regionsÐSerpa Pinto (31 hazy days instead of 12 in 1958), Lusaka (24 instead of 13)Ðand near Lake MalawiÐMzimba (24 days, none in 1958); far fewer hazy days in some eastern Africa stationsÐTabora (6 instead of 20), Kasama (Zambia, 3 instead of 16), Harare (0 instead of 16). In the northern part of ZaõÈre Basin (Mbandaka and Kisangani, close to the Equator), dust haze has already quite vanished. In 1957, the regression phase (which starts mainly in September in both years) is quicker in the subequatorial regions: in Kindu for instance (ZaõÈre, 3 S), the last (and single) hazy day in 19 September, whereas in 1958, there are 13 hazy days until 29 September; in Tabora (Tanzania, 5 S), there is only one day with dust haze in September 1957, there are 16 in September of the next year. On the contrary, dust haze is stronger in some central regions following a south-west±north-east direction, from Serpa Pinto (24 days instead of only 2 in September 1958), to Kongolo (ZaõÈre, 22 days instead of 16). In the same way, the Lake Malawi nucleus, which extends southward to Zimbabwe in 1958, extends northward to Tanzania in 1957: thus, by comparing the number of hazy days in September 1957 and 1958, we notice respectively 3 and 25 days in Harare, but 15 and 0 in Mbeya; in Dodoma (Tanzania), September 1957 is the haziest month in the year (11 days, instead of zero in September 1958). In October, the dust haze area, centred on Lake Malawi in both years, is smaller in 1957 than in 1958. In most of the western stations, there are very few hazy days: in Mwinilunga only 2 (17 in 1958), in Nova Lisboa, 0 (10 in 1958). In fact dust haze leaves the Atlantic side and becomes more concentrated in the Zambezi±Malawi region: 31 days in Mzimba (28 in 1958), 27 days in Lilongwe (6 in 1958), 24 days in Mbeya (3 in 1958), 19 days in Lusaka (3 in 1958). In both years this is the last dust haze pocket, before the rainy season. Even if they display speci®c aspects due to the variability of all the climatic parameters, the two years 1957 and 1958 give a good description of the important place of dry haze during winter in southern tropical Africa. They con®rm regional or local studies (Rwanda, Douguedroit et al. (1990); ZaõÈre, Crabbe (1986)). 3. EMISSION AND TRANSPORT OF DUST HAZE The extent and the duration of winter dust haze in the tropical southern Africa atmosphere seem unique in the world. They indicate especially favourable conditions combining the existence of available aerosols and atmospheric dynamics to create a seasonal accumulation above the continent. WINTER DUST HAZE IN SOUTHERN AFRICA 735 3.1. Origin of the dust Dust haze is here made up of aerosols with two main origins, which are de¯ated soil dust and ®re debris. The other possible origins, such as sea salts, volcanic dust, and industrial emissions, are, in fact, out of the question. The dust making up the haze is partly eroded from dried lands which are no longer protected by vegetation. The geology of the substratum, combined with a dry season and the reducing savanna vegetation explains the availability of ®ne unbound particles on extensive unconsolidated surfaces. The ®rst particles torn away from the soils by turbulence in a desert are formed of clay (Pewe, 1981; Middleton et al., 1986), as is con®rmed in the case of the Sahara (Morales, 1979). Sources of clay in southern Africa are diverse. They may particularly originate from the old African Precambrian platform, which includes rocks such as schists suitable for producing thin dust, and from alluvium and colluvium of the Quaternary. Moreover, soils formed by weathering of crystalline rocks have a low silt/clay ration due to the decomposition of silicate mineral to provide several types of clay, as well as alluvial deposits and fans, playa sediments and river beds. It is dif®cult to determine the importance of the soil supply in suspension in the atmosphere. We only know that reports of aeolian dust present in the South-east trade winds from southern Africa along the eastern edge of the Atlantic Ocean have an average loading of 114 mg m73 that mineralogically re¯ects the detrital soil characteristics of the area (Chester et al., 1972). The second main origin of aerosols is biomass burning. The study of the chemical content of precipitation in the Congo Republic showed that the in¯uence of the particles and gaseous emissions by biomass burning from savanna is preponderant and estimated at about 80 per cent of the chemical content (Lacaux et al. 1991). As that study is based on the two dry seasons of the Congo Republic, only the southern part of which is affected by the same atmospheric circulation as the rest of the tropical southern Africa, it represents a good idea but not a precise percentage of the origin of the dust haze accumulating in that area (see Section 2). If four-®fths of the chemical content has a biomass burning origin, we can consider that about the same percentage can be applied to aerosols. The smoke from biomass burning contains particles of different sizes. The small dust particles are active in scattering solar radiation associated with he opalescent atmosphere. They also have a long life in the atmosphere unlike the heavy ones (Crutzen and Andreae, 1990). Biomass burning is widespread in southern Africa, especially in the dry season. Because the broad features of the major biomes are determined mainly by regional climate, the plateau between 10 and 20 S was covered by a natural vegetation made of tropical deciduous dry forests (miombo) and north of 10 S by tropical broad-leaved forests. Both have been widely replaced by the anthropogenic savannas and cultivated lands. The north of the ZaõÈre Basin is occupied by the equatorial broad-leaved forest and desert spreads along the west coast at the foot of the plateau escarpment (Dubresson et al., 1994) (Figure 7). Biomass burning occurs every year in the savannas. It is associated with several purposes: removal of dry savanna vegetation at the end of the dry season, shifting cultivation (chitimene), clearing of forest and bushes early in the dry season to promote regeneration of grass, etc. The very low population density (<5 km72) is associated with very extensive agro-pastoral processes (Dubresson et al., 1994). Much of the burning occurs in the dry season, especially the savanna ®res in tropical areas (Crutzen and Andreae, 1990; Goldammer, 1990; Andreae, 1991; Delmas et al., 1991). The location and the frequency of these savanna ®res are not as well known as in he northern part of the continent. The photographic images of the nighttime low-light satellite imagery of the Defense Meteorological Satellite Program (DMSP) of the USA have displayed, for 1986 and 1987, the biomass burnings in tropical Africa (Cahoon et al., 1991). They con®rm the importance of biomass burning in the southern part of the continent. The annual synthesis of all ®re locations roughly displays a coincidence with the savanna areas but differences with the dust haze area were detected in 1957 and 1958. There were nearly no ®res in June 1986 and 1987 in the equatorial broad-leaved forest and in north-eastern Angola, both areas that are affected by the dust haze in winter. The peak of burning activity coincides with the northern boundary of the tropical broad-leaved forest south of the equatorial one. At the end of the dry season in October, the greatest ®re activity is limited to southern counties, as is the dust haze. The importance and the locations of biomass burning displayed by the satellite images are consistent with the conclusions of the study of the distribution of CO2 emissions from biomass burning during the dry season in 736 A. DOUGUEDROIT AND F. BART Figure 7. Main vegetation areas in southern tropical Africa (after Dubresson et al., 1994): 1, desert; 2, open semi-arid vegetation; 3, thorn bushes, steppe; 4, savanna woodland (miombo); 5, forest±savanna mosaic; 6, rain forest; 7, alluvial high plains of southern tropical Africa tropical Africa based on the FAO statistics 1975±1980 (Wee Min Hao et al., 1990). It con®rms that the general distribution of biomass burning in winter in tropical southern Africa does not vary much from year to year. So, although dust haze areas cover biomass burning locations, they also have a greater spatial distribution. The redistribution of the haze over southern Africa is due to the atmospheric circulation. 3.2. Mobilization of the dust in the atmosphere The mobilization of the dust in the atmosphere develops rapidly when the dry season begins. It is connected with an important atmospheric change in tropical southern Africa, the main features of which are known (Findlater, 1971; Taljaard, 1972; Mills, 1979; Leroux, 1983; Tyson, 1986) (Figure 8). In May, the low atmosphere becomes favourable to turbulence, which takes advantage of dust availability for transporting it in the air. The Inter Tropical Convergence Zone (ITCZ) has crossed the Equator and the ZaõÈre Air Boundary (ZAB), which separates the trade winds from the Indian and Atlantic Oceans, has slipped northward on the continent. East of the ZAB, the low atmosphere has the extension of the trade winds from the Indian Ocean. Because the inversion of the trade wind is found at about 2±25 km along the Malagasy coast, a shallow layer under the inversion spreads over the eastern escarpment (Bigg, 1992). As early as April it has already reached most of tropical southern Africa (Hills, 1979; Leroux, 1983). As it has been humidi®ed by its crossing over the ocean, it gives rain on the Mozambique coast. Its humidity is quickly exhausted and rainfall is limited to the coast: the air coming from the east has lost most of its moisture when it reaches the inland plateau. Meanwhile, pressure below 700 hPa increases in the area (DougueÂdroit et al., 1990), giving rise to a mid-tropospheric high pressure both centred on western Zambia. It is due to polar air masses, stronger in winter, reaching the latitude of the continent in the Atlantic Ocean, and then being carried on eastward by the highs (Taljaard, 1972; Leroux, 1983; Tyson, 1986). It is linked with the general pressure increase in the north, which appears in 1958 when the highs become deep enough to get over the escarpment (DougueÂdroit et al., 1990). The wind blowing north- WINTER DUST HAZE IN SOUTHERN AFRICA 737 Figure 8. Extension of the dust haze and the corresponding characteristics of the weather through three days, in May (12th), June (15th) and July (13th): 1, area limits of the six classes of dew point temperatures represented on the ®gures; 2, isobar; 3, station with dust haze (small dot station without dust haze) 738 A. DOUGUEDROIT AND F. BART eastward from the highs is subsiding and dry. Add to this the temperature becoming warmer by midday. All these reasons explain that in May the water balance in the soil shows a de®cit in Shaba (Alexandre and Nzengu, 1976) as in the whole continental tropical southern Africa. Simultaneously it is harvest time, the savanna vegetation is dry and huge ®res develop. Soils become bare, their upper layer dries out and disintegrates. The high ground temperatures cause buoyancy in daytime, and the heat of the ®re from biomass burning is reradiated from the particles to the atmosphere. So particles are convected away from the ground and diffused upward in the boundary layer of the atmosphere. 3.3. Accumulation of the dust in the atmosphere in winter The location of the dust accumulation in the atmosphere in winter is due more to the features of the seasonal synoptic-scale ¯ow patterns above the continent than to the location of the sources of aerosols themselves on the southern edge of the main accumulation area of the dust haze. During the dry season, the wind blows from the south-east over the southern Africa plateau towards the ZaõÈre Basin and from the northwest and west in the western part of the continent, over the Atlantic escarpment (Leroux, 1983). The trade wind from the Indian Ocean spreads over the continent north of 20 S, after a dif¯uence near the east coast, with one part of the ¯ow blowing northward over the ocean (Taljaard, 1972; Mills, 1979; Bigg, 1992). It extends as far as the ZAB (Figure 9). West winds blowing from St Helena over the Atlantic Ocean and called `monsoon' sweep the lower part of the ZaõÈre Basin. There is also an inversion in the Indian trade wind. It is found at about 2±25 km over the Malagasy Republic and is supposed to pass above the eastern escarpment and spread over the continent (Taljaard, 1972; Mills, 1979; Bigg, 1992). So particles are diffused buoyantly upward under the inversion level of the trade wind, which prevents them from rising any higher. Then they drift horizontally on the prevailing winds north-westward to the ZaõÈre Basin and the Atlantic Coast. They spread over the continent beyond the aerosol sources, as far as the Equator, which, on average, coincides roughly with the boundary of the trade wind in the north at 2000 m altitude. The synopticscale ¯ow patterns explain the average concentration of the dust haze: aerosols are trapped by an inversion at about 2 km above the continent. The maximum concentration is located in the north-west of southern Africa, this side of the north edge of the area where south-east winds blow. Due to the location of the high at that altitude, those winds turning westward over the Atlantic Ocean carry away the dust in the same direction. Within such a Figure 9. Average monthly boundaries of wind ¯ows (from Leroux, 1983); 1, ZAB at 1000 m altitude (J July, Au August); 2, north boundary of the south-east trade at 2000 m altitude (M March, A April, My May, Jn June, J July, Au August, S September, O October); 3, ITCZ at 1000 m altitude. Arrow main direction of the ¯ow at 2000 m altitude over the Atlantic Ocean WINTER DUST HAZE IN SOUTHERN AFRICA 739 general outline, the precise situation changes from year to year as in 1957 and 1958, depending on the yearly features of the atmospheric circulation. Recent research concerning biomass burning con®rms in a roundabout way the above situation. Because biomass burning produces trace gases constituted mainly of carbon monoxide and hydrocarbons, in addition to nitrogen oxides which are related to the formation of ozone in the troposphere, the spatial distribution of these gases becomes important for our purpose (Crutzen and Andreae, 1990; Fishman and Larsen, 1987; Reichle et al., 1986). Soundings made in the Congo Republic during 3 years, from 1983 to 1986 and in 1990±91 have shown that the annual evolution of ozone concentration has a marked increase in the dry season, which coincides near Brazzaville (5 S) with the southern winter. They display similar ozone and Aitken nuclei pro®les from June to August with typical important values trapped at a level from 1±15 km to 2±5 km high between the top of the `monsoon' below and the trade wind inversion above. They originate from biomass burning (Cros et al., 1988, 1992). The long episode of important ozone concentration in October 1984 accompanied by haze was connected with south-east trades reaching Brazzaville (Cros et al., 1988). Field measurements made in the Congo Republic, near Brazzaville, in the Southern Hemisphere, display important conclusions: pollution from biomass burning reaches the country in winter. It is associated with the south-east wind and is trapped at about 2 km under its inversion. The situation which prevails above the tropical southern plateau extends as far as the Congo Republic, where it is an acknowledged fact that some dust haze episodes occur, although there is no mention of haze in the IGY data. The hypothesis that haze does not seem to be reported in that country is con®rmed. It explains why in 1957 and 1958 the haze extension looks as if it is stopped in the north-west by the ZaõÈre river, west of its bend. In fact, it co-incides with the border between ZaõÈre and Congo Republics. So dust haze is the most visible feature of the atmospheric pollution from biomass burning. Even, if it does not persist every day all the season long, the associated modi®cation in the chemistry of the atmosphere does. Highest ozone concentrations are also present in Angola in winter, between July and September (Fabian and Pruchniewicz, 1977). 3.4. Contraction of the dust haze area This contraction corresponds with an important change in the atmospheric ¯ows over the area: the southward movement of the ZAB associated with the Indian trade withdrawal, and the arrival of the Atlantic west wind above the continent. The ZAB begins receding in August. In September, it reaches, at ground level, a south-west±north-east line through ZaõÈre to Lake Tanganyika and by October has reached 20 S in the west (Leroux, 1983). North of the ZAB, Atlantic wind blows from the west, from the ground level up to high altitudes. There inversion ceases to exist. Humid air takes the place of the previous dry air coming from the east, so that important rainfalls occur (at least around 100 mm in October 1958). The conditions associated with the dust haze ®rst disappear in the north, later in the south. Moreover, even south of the ZAB, they become less favourable; in parts of the area, east winds blow up to levels above 500 hPa. This average evolution hides very important differences from day to day, as shown by the 1958 data (see Section 2). The daily ¯uctuations of the ZAB are very broad, as much as several hundreds of kilometres between two successive days. On 21 September 1958, the Indian trade blew almost as far as the sources of the ZaõÈre and the Zambezi in the low atmosphere, and dry subsiding air coming from the west had invaded south of 15 S, with a very dry core in the west and diverging winds around. West wind spread over all the ZaõÈre Basin eastward to Lake Victoria, with stormy showers prevailing over the Great Lakes region. The dust haze was con®rmed within a small area close to Lake Malawi (Figure 10(A)). One week later, on the 29 September, it again reached 10 S in the ZaõÈre Basin and the Equator in East Africa: the stations with dust haze are much more numerous than on 21 September (Figure 10(B)). In tropical southern Africa October is always the month of transition between the dry and the wet seasons. On 5 and 6 October 1984, as in October 1957 and 1958, (dry) haze was reported in most of the area, and smoke and rain with thunder could be observed at the same time (Connors et al., 1991). Smoke and dry haze have a 740 A. DOUGUEDROIT AND F. BART Figure 10. Withdrawing dust haze in September 1958: (A) weather characteristics on 21 September. (B) weather characteristics on 29 September: 1, wind direction (here from south to north); 2, station with dust haze; 3, station with stormy shower; 4, area with a dew point temperature higher than 10 C; 5, area with a dew point temperature lower than 10 C; 6, lines of equal dew point temperature (in C) relationship with biomass burning, which increases in October in eastern Africa (Cahoon et al., 1990). Results obtained by the MAPS experiment for the period 5±13 October 1984 concerning the carbon monoxide amounts for 5 latitude 6 5 longitude grid-squares display a maximum in the same area of southern Africa (Reichle et al., 1990). The soundings at Harare and Bulawayo on 6 October 1984 shows a vigorous convection associated with rain and thunderstorm. Such a deep convection only affects a small portion of the country according to Meteostat images, but it leads to a vertical transport of the CO during the growth stages of the cumulus clouds (Connors et al., 1991). At the end of the dry season, atmospheric conditions related to biomass burning and the trapping of the dust haze disappear in most of tropical southern Africa with the withdrawal towards the south of the ZAB. 3.5. Transport of the dust haze out of southern Africa A striking feature of the spatial distribution of the dust haze consists of a major concentration above the plateau. The coast is much less hazy except in a small portion towards the Atlantic Ocean. However, its location changes from month to month and, on average, in 1958 is just south of the ZaõÈre river mouth. WINTER DUST HAZE IN SOUTHERN AFRICA 741 Knowledge of the spatial distribution of dust haze above the Atlantic and Indian Oceans is a means to show whether the dust leaves the continental area and is transported far away. It is derived from seasonal average ship, space shuttle or satellite data. It is consistent with what has been discussed previously on the dust haze above the continent. In autumn, dust haze estimated from ship-data corresponds to a thin coastal fringe of low values with a maximum along South Namibia (McDonald, 1938, in Pye, 1987). This point highlights a small aerosol plume originating from the desert (Rao et al., 1988). Images from NOAA calculated in equivalent aerosol optical depth (EAOP) for the same season (average 1990 and 1991) display for tropical Africa a plume with its maxima in Nigeria decaying southward as far as Angola and eastward to the mid-Atlantic (Husar et al., pers. comm., 1996; Stowe et al., 1996). If we apply the model of aerosol plumes originating from the continent, with high values near the coast decreasing steadily with distance, we have to deal with haze coming from western Africa transported at altitude in the ITCZ and towards the south-east by the winds. This is consistent with the lack of noteworthy dust emission from tropical southern Africa in spring. Winter, from June to August, is the season showing, according to ship data, the longest dust plume towards the west, with two rather low maxima along the coasts of Angola and Namibia (Pye, 1987). Seasonal time charts of NOAA display for winter a plume originating near the ZaõÈre mouth from the Congo Republic to north Angola and fanning out towards the mid-Atlantic, where it joins the dust plume from western Africa (Husar et al., pers. comm., 1996; Stowe et al., 1996). These images are consistent with the spatial distribution of the dust haze above southern Africa as studied before (see Section 2). Dust haze reaches the Atlantic coast of the ZaõÈre and the Congo Republics and is carried westward over the ocean. Its omission in the IGY data of the latter country is con®rmed when its place is shown as important in the ZaõÈre Republic atmosphere. It is carried over the Atlantic by the wind blowing westward on the edge of the St Helena high at about 2000 m. In the middle atmosphere the southern dust haze does not join the northern one, which is spread over the ocean at the latitude of the Senegal-Mauritania coasts. In spring, from September to November, ship-data show the dust to be at the maximum of the four seasons near Namibia (McDonald, 1938, in Pye, 1987). Flights of 16±17 September 1989 of the CITE 3 experiment show ozone concentration, in air masses originating from southern Africa. It is rather high in the boundary layer (<2000 m altitude) where is it associated with smoke and haze, and high in the mid-level of the atmosphere from 2000 to 4000 m (Shipman et al., 1993; Andreae et al., 1994). Seasonal NOAA images for September to November 1990±1991 display a plume originating in mid-Angola and fanning out westward as far as the MidAtlantic. The maximum value between 027 and 03 EAOP (including the maximum monthly value of 032 in September) is lower than the winter maximum (Husar et al., pers. comm., 1996; Stowe et al., 1996). Spring is the only season when the Mozambique Channel (the annual average of EOAP (011) of which is signi®cantly low) presents on NOAA images a plume from south-eastern Africa. The EOAP has a strong maximum in October (021) (Husar et al., 1996; Stowe et al., 1996). All these results are consistent with the average atmospheric situation above the continent during the transition between the dry and the wet season; as the ZAB withdraws southward with successive episodes of local deep convection, biomass burning developed mainly in Eastern Africa induces transport in the free atmosphere of aerosols and associated gases. Dust and gases cease to be trapped in the low atmosphere. They are transported westward by winds, giving a spring aerosol plume from Angola above the Atlantic and eastward above the Mozambique Channel by local depressions. 3.6. Southern tropical Africa in the global biochemical cycles Dust haze can be considered as an indicator of biochemical cycles. References are usually made to the two better known areas, i.e. Amazonia and West Africa (Goldammer, 1990; Andreae, 1991; Delmas, 1991). Important differences in dust haze development and transport exist between northern and southern Africa. It plays a less important part north of the Equator than south for several reasons. It occurs as spells of aerosols originating from the desert in the Sahara itself, where dust storms have been much studied and classi®ed (Carlson, 1979; Morales, 1979; Legrand et al., 1985; Buecher, 1986). They are common in the Sahara, where they represent on the whole a 742 A. DOUGUEDROIT AND F. BART much greater number of spells than in the Kalahari. The very different surface of the deserts in the Sahara and Kalahari explains the more important role of aerosols of the former in the atmosphere. Biomass burning develops during the dry season in the savannas and produces aerosols and trace gases (Cahoon et al., 1990; Goldammer, 1990; Andreae, 1991; Delmas, 1991). The dry season lasts 5 o 6 months over most of tropical southern Africa, with fast transitions to the wet seasons, except in small areas. In northern Africa it decreases from 10 or 11 months on the southern edge of the Sahara to 1 or 2 months near the Guinea coast (Leroux, 1983). It produces a longer season of biomass burning in southern Africa. Savanna surfaces are larger in the Southern Hemisphere, where, according to satellite images, a greater number of ®res occur every year (Cahoon et al., 1990). If aerosol emissions are estimated by well-de®ned percentages of burnt surfaces with a low ratio for the deserts (Goldammer, 1990; Andreae, 1991; Delmas, 1991), tropical southern Africa should be a larger source of aerosols and gases than the northern Africa. Conditions of aerosol transport are very different in both cases. In tropical northern Africa, prevailing winds (which ensure its horizontal ¯ow) blow southward in the direction of the ITCZ, i.e the north-easterly tradewinds (Harmattan). The aerosols are trapped under the trade inversion. As deep convection linked with squall lines propagating westward develop in the ITCZ, the aerosols are driven upward to mid-troposphere, where they are taken over the Atlantic. They are transported mainly in altitude layers as high as 5±6 km. Those processes develop all year round over northern Africa, with only small differences in their intensity, due to the ITCZ sliding from near the Gulf of Guinea in the northern winter to about 15±20 N in summer and back again during the second half of the year. Indications of Saharan aerosols descending in the Caribbean con®rm the transport across the Atlantic (Morales, 1979), and estimations of the EAOP over northern tropical Atlantic obtained from NOAA are consistent with such an outline. They display long plumes of aerosols originating from the coast, with a decreasing concentration westward as far as America during all four seasons (Husar et al., pers. comm., 1996; Stowe et al., 1996). In winter, aerosols are also transported over the Mediterranean towards Europe in relation to the Mediterranean polar front (Morales, 1979). In contrast, transport out of southern Africa differs according to the seasons. Ships have only noticed a low dust concentration along the coast (McDonald, 1938, in Pye, 1987). An aerosol plume originates from the continent above the Atlantic only in southern winter and spring as discussed earlier, but none has been detected in summer and autumn (Husar et al., pers. comm. 1996; Stowe et al., 1996). So we can put together the particular conditions of production and transport of aerosols and associated gases in tropical southern Africa. The small surface of the desert and the comparatively uniform length of the dry season, the importance of savanna surfaces and the great number of ®res, explain the importance of aerosols and gases in the atmosphere during the winter dry season and their decrease in spring. The conditions of the atmospheric circulation explains the accumulation of dust and gases above the continent during the dry season and their transport above the west and central tropical southern Atlantic in winter and spring. Transport out of the continent is limited to two seasons in southern Africa instead of four in northern Africa, and is less important in the dry season whereas the ®res seem to be more numerous (Cahoon et al., 1991). The sinks of aerosols and gases are partly on the continent and partly in the eastern and central Atlantic Ocean, unlike northern Africa, which produces aerosols and gases transported, with a regular decrease in importance, across the Atlantic as far as America. The importance of aerosols, and consequently of associated gases, in the atmosphere of tropical southern Africa can be estimated only qualitatively in this research. Their seasonal evolution such as is described here displays an unquestionable role in atmospheric chemistry and climate, especially in winter and spring. Further studies have to emphasize the quantifying of southern Africa production of aerosols and gases and its part in changes of atmospheric chemistry, including their possible in¯uence on climate (Houghton, 1995). 4. CONCLUSION Dust haze, quite common in dry areas, presents one distinctive feature in tropical southern Africa: its wide extent during the whole austral winter. Anywhere else, it is composed of spells lasting a few days, as long as certain synoptic features contribute to mobilize dust and to transport it far away. In tropical southern Africa, data for 1957 and 1958 (con®rmed by data from more recent years) display that dust haze is mainly a seasonal WINTER DUST HAZE IN SOUTHERN AFRICA 743 phenomenon, lasting in winter from May to October, i.e the whole dry season. Its extent increases until June, reaches its peak in August and decreases afterwards. In the ZaõÈre Basin, this hazy weather is continuous during 2 or 3 months and intermittentÐat least one day out of twoÐin the previous and following months. Altogether, it lasts less but remains frequent. It is due to the synoptic-scale ¯ow patterns of the dry winter. The dust from soils and biomass burning is mobilized upward as high as 2500 m, where it is stopped by an inversion in the trade winds. Then it is spread northward as far as the Equator and westward by those trades prevailing over most of the continent at that level. Dust accumulates during the heart of the winter above the north-west part of southern Africa before escaping over the Atlantic Ocean at the latitude of the ZaõÈre and Congo Republics. Recent studies on the atmospheric chemistry of, and on atmosphere dust detected over, the Atlantic Ocean by satellites con®rm the place of gases from biomass burning, especially during the dry season and their transport across the Atlantic being limited primarily to winter and secondarily to spring, which correspond with the austral dry season and the transition to the wet season. During the months of transition, the aerosols are less important but, due to changing circulation conditions in the atmosphere, they are more easily spread above continent and contiguous oceans. This is consistent with the seasonal evolution of dust haze, i.e atmospheric aerosols, such as is shown here. Dust haze has not only a climatological importance, it is also an indicator of the in¯uence of biomass burning, which produces most of the aerosols modifying the atmospheric chemistry evolution. ACKNOWLEDGEMENTS The authors thank Dr L. L. Stowe (NOAA, Washington, DC) for giving them a copy of papers not yet published and to Dr. Brian Giles for improving the English. EDITORIAL NOTE The paper by Tyson et al. on `An air transport climatology for subtropical southern Africa', Int. J. Climat., 16, 265±291 was not available to the authors when this paper was being revised. REFERENCES Alexandre, J. and Nzengu, J. 1974. `Le reÂgime hydrique dans les sols de la reÂgion de Lubumbashi (Haut-Shaba, ZaõÈre)', PeÂdologie, 24 (1), 49± 63. Andreae, M. O. 1991. `Biomass burningÐits history, use and distribution and its impact on environmental quality and global climate', in Levine, J. S. (ed.), Global Biomass BurningÐAtmospheric, Climate and Biospheric Implications, Cambridge University Press, Cambridge pp. 1±21. Andreae, M. O., Anderson, B. E., Blake, D. R., Bradshaw, J. D., Collins, J. E., Gregory, G. L. Sachse, G. W. and Shipham, M. C. 1994. `In¯uence of plumes from biomass burning on atmospheric chemistry over the equatorial and tropical South Atlantic during CITE 3', J. Geophys. Res., 99 (D6), 12793±12808. Bigg, G. R. 1992. `Validation of trends in the surface wind ®elds over the Mozambique Channel', Int. J. Climatol., 12 (8), 829±838. Buecher, A. 1986. Recherches sur les poussieÁres mineÂrales d'origine saharienne, TheÁse doctorat d'Etat, UniversiteÂ de Reims, 220 pp. 50 pp. Cahoon, D. R., Stocks, B. J., Levine, J. S., Cofer, W. R. and O'Nell, K. P. 1990. `Seasonal distribution of African Savanna ®res', Nature, 359, 812±815. Carlson, T. N. 1979. `Atmospheric turbidity in Saharan dust outbreaks as determined by analysis of satellite brightness data', Mon. Wea. Rev., 322±335. Cerf, A. 1980. `CaracteÂristiques du trouble atmospheÂrique en Afrique de l'Ouest', La MeÂteÂorologie, 5±21. Chester, R., Eder®eld, J. J., Grif®n, J. J., Johnson, L. R. and Padgham, R. C. 1972. `Eolian dust along the eastern margins of the Atlantic Ocean', Mar. Geol., 13, 91±106. Connors, V. S., Cahoon, D. R., Jr., Reichle, H. G., Brunre, E. G., Garstang, M., Seiler, N. and Scheel, E. 1991. `Savanna burning and convective mixing in southern Africa: implications for CO emissions and transport', in Levine, S. J. (Ed.), Global Biomass Burning,Ð Atmospheric, Climatic and Biospheric Implications, Cambridge University Press, Cambridge, pp. 147±154. Crabbe, M. 1986. La neÂbulositeÂ et la visibiliteÂ au ZaõÈre, AcadeÂmie Royale des Sciences d'Outre-Mer, Bruxelles, 92 pp. + 93 cartes. Cros, B., Delmas, R., Nganga, D., Clairac, B. and Fontan, J. 1988. `Seasonal trends of ozone in equatorial Africa: experimental evidence of photochemical formation', J. Geophys. Res., 93 (D7), 8355±8366. Cros, B., Nganga, D., Minga, A., Fishman, J. and Brackett, W. 1992. `Distribution of tropospheric ozone at Brazzaville, Congo, determined from ozonesonde measurements', J. Geophys. Res., 97, 12869±12875. Crutzen, P. J. and Andreae, M. O. 1990. `Biomass burning in the tropics: impact on atmospheric chemistry and biogeochemical cycles', Science, 250, 1669±1678. 744 A. DOUGUEDROIT AND F. BART Delmas, R. A., Loudjane, P., Podaire, A. and Menaut, J. C. 1991. `Biomass burning in Africa: an assessment of annually burned biomass in Levine, J. S. (ed.), Global Biomass BurningÐAtmospheric, Climatic and Biospheric Implications, Cambridge University Press, Cambridge, pp. 126±132. Deutscher Wetterdienst (not dated). International Geophysical Year 1957±1958, World Weather Maps, Part II Tropical Zone (25 N±25 S), Seewetteramt, Hamburg. DougueÂdroit, A., Bart, F. and N'tombi, M. K. M. 1990. `Type de temps de grande saison seÁche dans l'Afrique des Grands Lacs', Publ. Assoc. Int. Climatol., 3, 115±128. Druilhet, A. and Durand, P. 1984. `Etudes de la couche limite convective saharienne en preÂsence de brume seÁche (expeÂrience Eclats)', Boundary-layer Meteorol., 51±77. Dubresson, A., Marchal, J.-Y. and Raison, J. -P. 1994. Les Afriques au Sud du Sahara, GeÂographie Universelle, Belin-Reclus, 480 pp. Fabian, P. and Pruchniewicz, P. D. 1977. `Meridional distribution of ozone in the troposphere and its seasonal variations', J. Geophys. Res., 82, 2063±2073. Findlater, J. 1971. `Mean monthly air¯ow at low levels over the western Indian Ocean', Geophys. Mem. 16 (115), 55 pp. Fishman, J. and Larsen, J. C. 1987. `Distribution of total ozone and stratospheric ozone in the tropics: implications for the distribution of tropospheric ozone', J. Geophys. Res., 92 (D6), 6627±6634. Fouquart, Y. et al. 1987. `Observations of saharan aerosols: results of Eclats ®eld experiment. Part 1: thicknesses and aerosol size distributions', J. Clim. App. Meteorol., 28±37. Goldammer, J. G. 1990. `Fire in the tropical biota, ecosystem processes and global challenges', Ecol. Stud. 84 (97), 97 pp. Grif®th, J.-F. 1972. `Eastern Africa', in Climate of Africa, World Survey of Climatology, Vol. 10, Elsevier, Amsterdam, 601 pp. Hills, R. C. 1979. `The structure of the Inter-Tropical Convergence Zone in equatorial Africa and its relationship to East African rainfall'. J. Inst. Brit. Geogr., 4 (3), 329±352. Houghton, J. T. 1995. Climate Change, 1994, Radiative Forcing of Climate Change and an Evaluation of the IPCC IS92 Emission Scenarios, Cambridge University Press, Cambridge, 339 pp. Husar, R. B., Stowe, L. L. and Prospero, J. 1996. `Patterns of tropospheric aerosols over the oceans'. Personal communication. Lacaux, J. P., Delmas, R. A., Cros, B., Lefeivre, B. and Andreae, M. O. 1991. `In¯uence of biomass burning emissions on precipitation chemistry in the equatorial forest of Africa', in Levine, S. J. (ed.) Global Biomass BurningÐAtmospheric, Climatic and Biospheric Implications, Cambridge University Press, Cambridge, pp. 167±173. Legrand, M., Bertrand, J. J. and Desbois, M. 1985. `Dust clouds over West Africa: a characterization by satellite data', Ann. Geophys., 3, 777± 783. Legrand, M., N'DoumeÂ, C. and Jankowiak, I. 1994. `Satellite-derived climatology of the Saharan aerosol', in Lynch, D. K. (ed.), Passive Infrared Remote Sensing of Clouds and the Atmosphere II, Proceedings SPIE 2309, pp. 127±135. Leroux, M. 1983. Le climat de l'Afrique tropicale, Champion, Paris, 633 pp., 1 atlas. Mbenza, M. 1982. `L'eÂvolution de la teneur en eau de l'air aÁ proximiteÂ du sol dans la reÂgion de Lubumbashi au cours de la peÂriode sans pluies continuelles', GeÂol. Ecol. Trop., 3, 191±199. Middleton, N. J., Goudie, A. S., and Wells, G. L. 1986. `The frequency and source areas of dust storms', in Nickling, W. G. (ed.), Aeolian Geomorphology, Allen and Unwin, Winchester, pp. 237±259. Morales, C. 1979. Saharan Dust. Mobilization, Transport, Deposition, Scope 14, Wiley, New York, 320 pp. N'Doume, C. and Legrand, M. 1995. `Optimisation du traitment et de l'imagerie Meteosat IR pour l'eÂtablissement d'une climatologie des aeÂrosols sahariens', Veille climatique satellitaire, 53, 66±80. Nicholson, S. E. 1986. `The spatial coherence of African rainfall anomalies: interhemispheric teleconnections', J. Clim. Appl. Meteorol., 25, 1365±1381. Nicholson, S. E., Jeeyoung, K. and Hoopingarner, J. 1988. Atlas of African Rainfall and its Interannual Variability, Florida State University, Tallahassee, Fl, 15 pp., maps. N'tombi, M. K. M. 1988. `Evolution de l'humiditeÂ atmospheÂrique au cours de l'anneÂe dans la partie infeÂrieure de la troposheÁre d'une reÂgion tropicale humide. Cas de Lubumbashi (Haut-Shaba, ZaõÈre)', in Actes JourneÂes Climatologie, Mont-Rigi, M. Erpicum ed., LieÁge, pp. 237± 252. N'tombi, M. K. M. 1990. Etude des sondages aeÂrologiques et des images satellitaires de MeÂteÂosat en vue de l'explication du climat de la reÂgion de Lubumbashi (Shaba meÂridional, ZaõÈre), TheÁse de doctorat, FaculteÂ des Sciences, UniversiteÂ de LieÁge, 377 pp. Pewe, T. L. 1981. 'Desert dust: origin, characteristics and effect on man', Geol. Soc. Am. Spec. Pap., 186 pp. Pye, K. 1987. Aeolian Dust and Dust Deposits, Academic Press, London, 334 pp. Rao, Cr. N., Stowe, L. L., McClain, E. P. and Sapper, J. 1988. `Development and application of aerosol remote sensing with AVHRR data from the NOAA satellites', in P. V. Hobbs, and McCormick, M. P. Aerosols and Climate. A. Deepak Publ., Hampton, VI, pp. 69±79. Reichle, H. G. Jr., Conners, V. S., Holland, A., Hypes, W. D., Wallio, H. A., Casas, J. C., Gormsen, B. B., Saylor, M. S. and Hasketh, W. D. 1986. `Middle and upper tropospheric carbon monoxide ratios as measured by a satellite-borne remote sensing during November 1981', J. Geophys. Res., 91 (10), 865±888. Shipman, M. C., Bachmeier, A. S. and Anderson, B. E. 1993. `CITE 3, meteorological highlights', J. Geophys. Res., 98 (D12), 23305±23324. Stowe, L. L., Ignatov, A. M. and Singh, R. R. 1996. `Development, validation and potential enhancements to the second generation operational aerosol product at NOAA/NESDIS', J. Geophys. Res., submitted. Taljaard, J.-J. 1972. `Synoptic meteorology of the Southern Hemisphere', in Newton, C. W. (ed.) Meteorology of the Southern Hemisphere, Meteorological Monographs, American Meteorological Society, Boston, MA, pp. 139±213. Tyson, P. D. 1986. Climatic Change and Variability in Southern Africa, Oxford University Press, Oxford, 220 pp. Villeneuve, G. O. 1980. Glossaire de meÂteÂorologie et de climatologie, 2eÁme eÂd., Presses de l'UniversiteÂ Laval, QueÂbec, 648 pp. Walter, H. and Lieth, H. 1967. `Afrique au sud de l'Equateur 5 N±22 S', in Klimadiagram Weltatlas, Gustav Fischer Verlag, Iena. Wei Min Hao, Mei-Huey Liu and Grutzen, J. P. 1990. `Estimates of annual and regional releases of CO2 and other trace gases to the atmosphere from ®res in the tropics, based on the FAO statistics for the period 1975±1980', Ecol. Stud., 84, 440±462.