Africa mainly comprises developing countries, many of which depend heavily on rain-fed agriculture to feed their populations. From an agriculture standpoint, precipitation is perhaps the most important variable in tropical climates. The recent years, the decrease in rainfall evolution resulted in low agricultural production and affected thousands of people, with serious degradation of the environment or ecosystems (coast littoral, alluvial plain). A better understanding of the interannual variations of the precipitation would be of obvious benefit. The Congo forest basin, located in west equatorial Africa has attracted much less studies than other parts of Africa. This study which focused on Congo-Brazzaville purports to fill this gap. Congo-Brazzaville is one of the countries making up the western part of the Congo forest basin, also called the Atlantic Equatorial Africa. It is located, between 5°S–4°N and 11°–18°E (Figure 1), and has a total area of about 342 000 km2. Two major ecosystems are distinguished: the forest zone and the savanna zone. The forest zone is estimated to account for 70% of the total area, and the savanna zone for 30%. The Congolese basin, the northwest Plateaus, the Chaillu massif, and the Mayombe mountain range are covered by tropical forest, while savannahs dominate the Niari valley, the Cataractes plateau, and the Bateke plateaus. The high plateaus and the chain of mountains are the main topographic features of the country, with elevation varying between 600 and 800 m, and some peaks over 800 and 1000 m. The Niari Valley extends to the southwest of the country, creating a discontinuity between the plateaus (Bateke and Cataractes) and the Mayombe mountain range. These vegetation and elevation features could influence climate conditions at local and regional scales.
In Congo-Brazzaville, as in most of the wet tropics, agriculture is mainly rainfall-dependent. Unganai (1996) in Zimbabwe show that agriculture and the general economic performance are highly dependent on rainfall. The decrease of rainfall could therefore affect the country's food security and development. In Congo-Brazzaville rainfall is a major climatic element; it has significant socio-economic impacts on the rural population which depend on rain-fed agriculture. Although the economy of Congo-Brazzaville does not strongly rely on agriculture (only 8% of the GDP), this economic activity provides a living for approximately 90% of the rural populations.
Numerous authors have studied rainfall variability in Africa since the 1970s (Hulme, 1992; Hastenrath and Lamb, 1977; Palmer, 1986; Wolter, 1989; Hastenrath, 1990; Nicholson and Nyenzi, 1990; Bigot, et al., 1997; Nicholson, 2000; Hulme, et al., 2001). There are many studies focused on Western Africa (Lamb, 1978, 1983; Nicholson, 1979; Farmer, 1988; Lamb and Peppler, 1992; Fontaine, et al., 1999; Fontaine and Philippon, 2000; Mahé, et al., 2001), Eastern Africa (Beltrando, 1990, Beltrando and Camberlin, 1993; Camberlin, 1993, 1997; Seleshi and Zanke, 2004) and southern Africa (Nicholson, 1986a; Mason et al., 1999; Richard et al., 2000; Fauchereau et al., 2003). Few studies are available for Atlantic Equatorial Africa (Maloba Makanga and Samba, 1997; Samba et al., 2008). However, changes in annual rainfall have been studied at the scale of Africa as a whole (e.g. Nicholson, 2000; Hulme, 2001). These analyses indicated that the second half of the 20th century suffered predominantly negative rainfall anomalies, corresponding to a decline of precipitations. The hydrological studies in the Basin of the Congo River show a decline of precipitation of about 10% during the last decades since 1980s (Mahé et al., 1990; Lienou et al., 2008).
The aim of this paper is to present the results of a study of the rainfall trends and variability in Congo-Brazzaville, in the Congo forest basin over the common period 1932–2007. This study seeks to determine the organisation of spatial and temporal precipitation anomalies in the country. We analyse annual and seasonal rainfall and also rainy days based on 12 key stations located in Congo-Brazzaville.
The organisation of this paper is as follows: the data originate from the national meteorological service; these data are presented in Section 2. The relevant aspects of the statistical methods are described in Section 3. Section 4 gives a brief description of rainfall generation mechanisms in the Congo forest basin. The results on the local and regional rainfall evolution over Congo-Brazzaville are in section 5. Finally, a discussion and a conclusion are presented in section 6.
The data used in this study include monthly rainfall amounts and number of rainy days for 13 synoptic stations provided by the Direction Nationale de la Météorologie of Congo-Brazzaville. The data were processed by the ‘Centre de Recherches et d’Etudes en Environnement (CREE)' of Marien Ngouabi University of Brazzaville. Only 12 stations have rainfall data with less than 8% missing values. The period 1932–2007 is taken as the common period over which good quality data is available at these 12 stations. In the equatorial climatic context of Congo-Brazzaville, we consider a rainy day as a day when the precipitation equals or exceeds 1 mm. Summary of statistics for the annual rainfall totals at each station are given in Table I.
Table I. Stations used in this paper, their altitude and statistical parameters (over the period 1932–2007)
Standard Dev (mm)
For each station, the monthly rainfall totals (Table II) have been standardized and plotted. The standardized annual and seasonal rainfall totals were used to study the variability and trends of rainfall in Congo-Brazzaville. The main rainfall regimes of Congo-Brazzaville (Figure 2) are displayed as standardized monthly precipitation in each station. All stations are characterized by a bimodal regime, with a majority of them showing peaks in the three-month periods of March–May (MAM) and September–November (SON). All stations situated south of Gamboma (2°S) display peaks centered on March or April and November, while the two rainfall peaks intervene in May and October for the stations situated in the north (Figure 2). Combined with the mean annual rainfall pattern, these results are indicative of two main types of climate regimes. The area situated north of 2°S has a humid equatorial climate, it is considered as the northern part of the Congo-Brazzaville. The area south of 2°S has a humid tropical climate, it is considered as the southern part of the Congo-Brazzaville (Figure 1). However, the network of observations used is characterized by the uneven distribution of stations, with an average density of a synoptic station for 26 308 km2. Nevertheless, the density of stations is lower in the northern part of Congo-Brazzaville.
Table II. Monthly and seasonal rainfall statistics (over the period 1932–2007)
3. Statistical methods
The annual and seasonal rainfall at each station is standardized to obtain the annual and seasonal rainfall anomalies at that station according to the expression
Where xij is the monthly or annual rainfall for the station i in the year j, xi is the mean monthly or annual rainfall for station i, and σi is the standard deviation of annual totals. Standardized time-series were used by many authors, including Kraus (1977), Nicholson (1986b), Osman and Shamseldin (2002), to document interannual rainfall variability. In this study, we also used this expression to depict the annual regime (Figure 2). We used this method to reduce the local climatic peculiarities. The normalized series are then averaged spatially.
The synoptic stations series were submitted to a non-parametric Mann-Kendall test to detect trends and abrupt changes in the time-series over the period 1932–2007, for both the annual and the seasonal rainfall totals. The principles of this test have been largely described by Sneyers (1975) and Vandiepenbeeck (1995), and it was used in several climate studies in Africa (Seleshi and Zanke, 2004) and elsewhere (Esteban-Parra et al., 1998; Lana et al., 2004).
In the Mann-Kendall test, for each element xi(i = 1, …, n) of a series yi of length n, ni is the number of elements j which precede i(i > j) such as xi > xj. The trend statistic t of the test is computed as follows:
The distribution of t, under the null hypothesis, is practically a normal distribution with the average and the variance given by the following expressions:
The reduced statistics of the test, given by |u(t)|, is thus compared to a normal distribution law.
The null hypothesis can, therefore, be rejected for high values of |u(t)|, this being the probability α1 of rejecting the null hypothesis when it is derived from a standard normal distribution table:
The Mann-Kendall test consists in calculating two series of statistical values, one from the beginning of the series, the second from the end. These series are shown in the form of two curves respectively called the direct curve (ui) and the retrograde curve (u′i). A trend is significant when the curve ui exceeds the 5% threshold, i.e. when |ui|> 1.96. Sneyers (1975) demonstrated the usefulness of this test, using its direct progressive and retrograde forms, for identifying the intervals in which trends are most pronounced, and trend turning points and/or climate shifts. The point which marks the beginning of the change corresponds to the intersection between the direct curve and the retrograde curve, u′i. Graphically, the retrograde and direct curves are often confused when there is no significant trend in the series. When values of u(t) are significant, one concludes to a rising or decreasing trend, for u(t)> 0 or u(t)< 0, respectively.
Different sub-periods from 1932 to 2007 were compared to examine whether there was any change in rainfall trends throughout the period of investigation. Whenever a change is detected and dated, a comparison between the two sub-periods is carried out using the t-test and where trends are indicated to be significant in the discussions of the results they are at the 5% level. It is only when the Student's t-test is significant that the shift in the average climate is validated. The Student's t-test is computed from the difference between the averages of two sub-samples, as follows:
Where x1 and x2 are the average of the first and second sub-samples, respectively; n1 and n2 the size of the first and second sub-samples; σ1 and σ2 the standard deviations of the first and second sub-samples.
4. Overview of rainfall generating mechanisms over Congo-Brazzaville
The general climatological situation over the Congo forest basin is driven by three permanent anticyclones located northwestwards (Azores), southwestwards (St Helena), and southeastwards (Mascarene). The air masses originating from these anticyclones converge along the Inter-Tropical Convergence Zone (ITCZ) which separates the southerly low-level winds from the northerly winds, and the Inter-Oceanic Confluence Zone (IOCZ), separating the westerly from the easterly winds in the southern part of Africa. A low-level pseudo-monsoon flux of moist air driven by the Ste Helena is trapped along the convergence zones by the high-altitude warm air originating from the north and the southeast. The warm monsoon flow produces a large zone of convective clouds bounded by the ITCZ and IOCZ. Therefore, the shift in the position of the ITCZ and IOCZ during the year determines the different rainfall regimes. The rainfall generating mechanisms are commanded by a zone of shallow depression systems in the Congo forest basin, accompanied by the north–south oscillation of the ITCZ. In the tropics, including the Congo forest Basin, the ITCZ is the major synoptic-scale system controlling seasonal rainfall (Leroux, 1975, 1980; Okoola and Ambenje, 2003). The ITCZ, over the Congo forest basin, tends to have a meridional orientation (running north–south relatively parallel to the meridians) due to the contribution of the westerly flow resulting from the recurved southeasterly flow off the cold Southeast Atlantic Ocean (Okoola and Ambenje, 2003). This meridional branch of the ITCZ affects the synoptic climatology of the region from September to April (Nicholson, 1986a). It enables the whole e Congo forest Basin to remain relatively wet between the time the overhead sun crosses the Congo forest Basin southwards (September) up to the time of its return in March. Therefore, the year in most of Congo-Brazzaville aeras is divided into two main periods: the dry period (June–August or December–January–February) and the wet period (September–October and March–May), which is not uniformly wet however.
A brief description of the mechanisms for rainfall formation for each season is given below. December–February (DJF) is the small dry season in the south and the main one in the north of Congo-Brazzaville. In surfaces the anticyclones have a situated relatively symmetric with regard to the equator and to Africa. In the Northern Hemisphere, the Azores anticyclone is welded with the cell Egypto-Libyan of thermic origin; in the Southern Hemisphere, the St. Helena anticyclone (in Atlantic Ocean) and the Mascarene anticyclone (in the Indian Ocean) strongly moved eastward. During this season the low-pressure intertropical convergence zone (ITCZ) associated to the shallow depression, reduce the activity of the mechanisms of the precipitations in the Congo basin. The Southeast air over Atlantic Ocean recurves towards the Congo Basin to arrive in the western parts of Atlantic Equatorial Africa as a weakly wet and stable westerly airstream contributing to the Congo precipitation. The dry season from June to August (JJA) is the small dry season in the north and the big one in the south of Congo-Brazzaville. During JJA, the country predominantly falls under the influence of cool southwesterly winds. These dry air masses originate from the St. Helena anticyclone. In JJA, most of the country is generally dry; the exception being the north extreme of Congo-Brazzaville. The ITCZ has an extreme north position with regard to the Congo Basin inhibiting any mechanisms of precipitations.
The MAM and SON are transition seasons between two dry seasons (DJF and JJA). Major rain-producing systems during seasons (MAM and SON) are: the meridian migration of the ITCZ (runs north–south or south–north), eastern flux air from the Congo basin and the convective precipitations. The ITCZ oscillations in Africa vary annually, causing most of the variability of season in the shallow depression over Congo-Brazzaville.
Congo-Brazzaville has two crop seasons related to the Potentially Useful Rainy Seasons (PURS) describe by Samba et al. (1999). The first PURS is corresponding to March–May (MAM) and the second season to September–November (SON). With dynamics atmospheric and according to Samba et al. (1999, 2008), Samba and Mpounza (2005), four seasons exist in Congo-Brazzaville. The first season from December to February (the small dry season in south and the big one in north Congo-Brazzaville), the second is the main rainy season from March to May (MAM), the third is the dry season from June to August (the small dry season in north and the big one in south Congo-Brazzaville), and the fourth is the small rainy season from September to November, known locally as Mwanga, Ndoolo, Kisiwu, and Ntombo, respectively. Samba et al. (1999), showed that in Congo the September start season SON or Ntombo known locally as Luanga or Bangala, when the period of the sun is maximum. March is the planting month for season MAM, and October and November are the planting months for season SON. A large majority of the rural people are subsistence farmers who depend on the 2% of Congo-Brazzaville's land that is suitable for producing crops for their own needs. Agricultural production is influenced by the significant spatial-temporal variations that occur in the rainfall.
5.1. Rainfall variability
It is known that rainfall over the equatorial and wet tropical areas are characterized by a relatively low interannual variability. In Congo-Brazzaville, all stations show a coefficient of variation (CV) of annual rainfall between 0.12 (Djambala) and 0.29 (Pointe-Noire, Figure 3). The northern stations of Impfondo, Ouesso and Djambala have the lowest variability (CV < 0.15). Annual rainfall over the southwestern stations is more variable. Considering SON rainfall variability (CV > 0.24), which is directly affecting agricultural production over most parts of Congo-Brazzaville, all of the stations exhibit a coefficient of variation above 0.16. Most stations, where SON rains constitute about 32% of the annual rainfall, the year-to-year variability is high (CV > 0.20). MAM rainfall over the stations is highly variable (CV > 0.30). The rainfall variability is more important at the seasonal scale than at the annual scale. Also, the south presents a stronger variability than the north.
Figure 3 shows standardized time series of annual rainfall totals at different stations. It shows again, time series plots of the annual rainfall evolution over the space Congo-Brazzaville. By linking anomalies and 5-year moving average we were able to determine the different periods of the annual rainfall evolution. Most of stations of Congo-Brazzaville recorded rainfall annual series with majority of negative anomalies during the 1930s and 1940s. The decade 1950 shows for the majority of positive anomalies. During the 1950s, the year 1958 knew a deficit of the precipitation generalized on all Congolese territory. The 1960s is characterized by precipitations excess on all the Congolese territory. The 1980s and 1990s knew the deficit the most important precipitation. This precipitation decrease is about 10–20%. The transition between positive and negative anomalies decades is not uniform on all the territory. It appears either at the end of the 1960s, or during the 1970s. The evolution of the precipitation at Brazzaville is particularly stable, although there are important positive abnormalities during the 1990s.
The southern regions show consecutive year periods with wet and dry years with apparent trend, whereas the northern regions have shown persistent below-average rainfall since the late 1970s. It is to be noted that the major drought years of 1958, 1983, and 2006 covered the whole of Congo-Brazzaville. During the 1983 drought, the annual rainfall totals in all regions were below average (below one standard deviation with very low totals (>1.5 standard deviations) Congo-Brazzaville.
The time series of the MAM seasonal rainfall (Figure 4) shows the same evolution as annual rainfall. The season rainfall evolution associated with every station is characterized by positive and negative anomalies respectively the decades 1950s and 1960s, and the 1970s and 1980s. However, we note some peculiarities in the evolution of the precipitation at Dolisie and at Brazzaville where positive anomalies appear during the 1980s. The SON seasonal rainfall evolutions show a predominance of the negative anomalies during the last two decades in most of the north stations (Figure 5). This evolution rests stable over the southern part. The rainfall analysis does not show a link between the annual evolutions on one hand and the seasonal evolutions on the other hand. With regard to what precedes, the evolution of the precipitation is recorded in different ways over Congolese territory. This evolution is marked by predominance of the positive anomalies for the stations of the south, with some exceptions. However, for the stations of the north, the evolution is marked by a decrease of precipitation.
5.2. Trend analysis on annual and seasonal rainfall amounts
The non-parametric Mann-Kendall trend test results, shows that there is a significant trend in the annual and seasonal rainfall totals in the north of Congo-Brazzaville over the period 1932–2007. Figure 6 related to annual rainfall totals southern stations points out stable evolution and northern (Gamboma, Impfondo) show significant decreases. Figure 6 gives the graphical result of the Mann-Kendall trend test, indicating a significant decline in rainfall since the 1980s. Most of southern stations show a stable evolution. So, in the northern part we can define two sub-periods (before and after the 1980s). The first one and the second respectively correspond to the wet and the dry period. The annual decrease rainfall between two sub-periods reaches 10% at the north. The March–May seasonal rainfall (Figure 7), the southern stations point out significant decrease (Pointe-Noire, Loubomo, Brazzaville), whereas the northern stations (Impfondo, Ouesso) shows increase (Figure 7).
From the student t-test the decline was compared between two sub-periods discovered by Mann-Kendall (Table III). The southern stations (e.g. Brazzaville) point out a significant decrease in the mean MAM seasonal rainfall, from 513 mm to 459 mm, i.e. by 10%. The mean MAM seasonal rainfall at Dolisie decreased by 17%, from 521 mm to 435 mm and at Pointe-Noire is decreased by 31%, from 514 mm to 353 mm. One notes an opposition of evolution of rains to the annual scale and at the level of the season MAM between the north and the south. This opposition of evolution of rains is not observed for the season SON (Figure 8). Indeed, the majority of the Congo-Brazzaville stations show a significant increase of rains during the season SON.
Table III. Student's t-test on sub-periods defined by Mann-Kendall test; corresponding unlike rainfall between two sub-periods (over the period 1932–2007)
5.3. Changes in rainy days
Daily rainfall analysis reveals that the annual mean rainy days of the synoptic stations point out two periods (Figure 9): one between 1930s and 1940s with weak declines in the annual number of rainy days, the second period (1950s and 1960s) shows an increase in the annual number of rainy days. However, the eastern synoptic stations have a significant decrease in the annual rainy days during the last decade. In the major drought years of 1958, 1983, and 1984, the annual number of annual rainy days was below average over Congo. In northern Congo-Brazzaville the annual number of rainy days increases from a mean of 107 rainy days to 144 days (increase by 34% e.g. Impfondo). In Impfondo the annual number of rainy days decreased the last decade from a mean of 144 rainy days to 122 rainy days (decrease by 15%). The annual number of rainy days at Ouesso increases from a mean of 1990 rainy days to 134 days (increase by 49%). In southern parts (e.g. Dolisie), the annual number of rainy days increased from a mean of 55 rainy days to 109 rainy days (increase by 97%). The decline in the annual number of rainy days in central Congo-Brazzaville is less pronounced and is about 20%. The strong variability of number of rainy days is more brought up during seasons MAM and SON (Table IV).
Table IV. Coefficient of variation of the annual, MAM and SON rainfall totals and rainy days over the period 1932–2007
Rainy days (number)
Rainy days (number)
Rainy days (number)
The Mann-Kendall test non-parametric applied among rainy days defines sub-periods for the various series of the stations of Congo-Brazzaville (Figure 10). In Table V is also contained a carry-forward of the report of the total of the precipitation (P) on the rain days number (N) in the year, a parameter little studied in the climatic analysis of the Congo forest basin. This report was used in western Africa by Carbonnel (1987) and Chaouche (1988) to detect changes in the precipitation. At the Congo-Brazzaville we note changes in the evolution of the precipitation (P) and the rain day number (N). We also note an increase of rain days in most part of stations. This increase of N can be accompanied by a decline or an increase of precipitation.
Table V. Student's t-test on sub-periods defined by Mann-Kendall test, and the rapport P (precipitation in mm) on N (days number of precipitation)
Sub-periods by N
Mean P (mm)
Figure 11 represents the center averages (A, B, or C) of the distributions of P and N of every sub-period (Table V). These averages A, B, or C of P and N, show a moving of center (black arrows), which indicates a modification between sub-periods. Three categories of stations are observed: in the first category (Impfondo, Pointe-Noire, Dolisie), P decreases and N increases; in the second (Brazzaville, Mouyondzi, Ouesso), P and N increases, and in the third category (Djambala) one observe at first an increase of N and decrease of P, followed by a decrease of P and N.
6. Discussion and conclusion
The obtained results are in agreement with the main conclusions of the other research works made on the region (Moron, 1994; Beltrando and Camberlin, 1995; Camberlin, 1995; Ogallo, 1988; Maloba Makanga and Samba, 1997; Nicholson, 2000; Hulme et al., 2001; Mahé et al., 2001). They show that the relatively rainy periods are situated between the 1950s and the end of the 1960s, and that the 1970s and 1980s were the least rainy. The 1930s and 1940s are characterized by a deficit of the annual rainfall. The evolution of the precipitation shows a decrease from 1968. The results of this work will allow one to appreciate the incidences in the local scale of a climatic evolution which takes place in a context of global change because the impacts in climate change have more local repercussions.
We have seen that the annual and seasonal rainfall totals over Congo-Brazzaville (northern) show recent low trend over the period 1932–2007. The annual and the seasonal rainfall in Congo-Brazzaville show significant recent declines since about the 1980s. The progressive Mann-Kendall test plot revealed a remarkable similarity between trends of annual rainfall total at Impfondo and Gamboma. A decrease in MAM seasonal rainfall over the southern area corresponds to a stationary in the northern. However, the SON seasonal rainfall points out the increase over Congo-Brazzaville. The effects of this seasonal rainfall evolution are a decrease in annual rainfall in the north and is stationary in the south. Furthermore, it appears that the southern part has low annual precipitations (e.g. Pointe-Noire) which is due to the long cold period of the South Atlantic Ocean over the period approximately from May to August.
The totals of the rains of the seasons MAM and the SON do not present the same evolution between two seasons. The evolutions between the annual and seasonal rainfall totals do not present similar evolution either. In some stations (e.g. Impfondo, Ouesso), one observes same evolutions between year and seasons. The weak concordance of the seasonal evolutions in the annual evolution can be connected to weak correlations or connections between both seasons (MAM and SON). Indeed, there are three rainy types in the space Congo: the rains of disturbances or lines of grains, convective rains and the maritime trade winds or pseudo-monsoon (Samba and Mpounza, 2005). The season MAM is dominated by convective rains (70%) with few rains of disturbances (20%), while the season SON is dominated by the disturbances (50%) and convections (30%).
Taking the whole of Africa, the North of Congo was touched by a precipitation deficit during the end of the 20th Century. If in western and eastern Africa, this deficit appears since the end of 1960s. In the north Congo, the deficit intervenes at the end of 1970s. This deficit is moved off decade in the north Congo. On the other hand the south Congo is marked by rainfall stability.
The works of Motha et al. (1980), Nicholson (1980), Seleshi and Damarée 1995, Janicot and Fontaine (1998), Fontaine et al. (1995) on the rainfall of western Africa showed the existence of strong meridian pressure gradient of the precipitations. The seasonal meridian movement of the ITF (Intertropical Front) or the ITCZ (Intertropical Convergence Zone) as well as its structure indeed lead to a quasi-zonal distribution of the annual isohyets with a strong diminution of the precipitations North (Janicot, 1992). Moron (1996) showed a good coherence of the spatiotemporal structures of the precipitations of the space. Eastern Africa also remains a region marked by the homogeneity with regard to the rainfall behaviour.
On the other hand, a strong variety of the rainfall regimes characterizes this part of the African continent. We note a very strong pressure gradient NW-NE with a spatial coherence of the distribution of the precipitations. On the other hand, Congo presents a very low spatial coherence of the precipitation. The Atlantic Ocean stream, because of its short trajectory ocean/continent, and also of the subsidence in the trade wind, remains stable. The major part of the precipitation is of origin convective. Cloud heaps or waves are the disturbances which bring rains to the synoptic scale or to the meso-scale. They are these aerologic and orographic factors, illustrated by Pointe-Noire stations (on the coastal plain), Dolisie (under the wind), Djambala (on plateau Bateke) and Impfondo (in the Congo basin) which contribute to the complexity of the spatial distribution of the precipitations in the Congo area.
This study has shown that:
1.There is trend in the annual, SON and MAM rainfall totals and rainy days over Congo-Brazzaville in the period 1932–2007.
2.The annual and the MAM seasonal rainfall in Congo-Brazzaville show a significant decline. For annual rainfall the decline concerns the northern part, and MAM seasonal rainfall the decrease affect the southern stations evolution.
3.The SON seasonal rainfall in Congo-Brazzaville show a significant increases for most of stations.
4.The rainy days in Congo-Brazzaville show two principal periods: one for low number before 1968, and the second for high after 1970.
The authors thank the directors and staff of the Direction Nationale de la météorologie of Congo-Brazzaville, for providing most of the rainfall data used in this study. The authors wish to thank Pierre Camberlin from Centre de Recherches de Climatologie (Université de Bourgogne -Dijon-France) for his advice. We also thank the anonymous reviewers for their constructive remarks that greatly improved the manuscript.