Global and Planetary Change 26 Ž2000. 137–158
www.elsevier.comrlocatergloplacha
The nature of rainfall variability over Africa on time scales of
decades to millenia
Sharon E. Nicholson)
Department of Meteorology, Florida State UniÕersity, Tallahassee, FL 32306, USA
Abstract
This paper begins with an overview of the African rainfall regime, noting in particular the contrast among various regions
of the continent, followed by a description of the nature of climatic Ži.e., rainfall. variability over Africa on time scales of
decades and centuries. The decadal scale is examined using modern data covering the twentieth century. The century scale is
examined using historical reconstructions of climate, based on a combination of geologic, geographic and historical
information Že.g., lake chronologies, landscape descriptions, archives and diaries..
The presentation includes some results of an analysis of a new historical semi-quantitative data set for Africa covering
the last two centuries. It was produced using a combination of historical information, nineteenth century rainfall records, and
statistical relationships among various sectors of Africa. Presented here are reconstructions of lake level fluctuations for
numerous lakes of eastern and southern Africa.
This overview of climatic fluctuations is utilized to uncover inherent spatial and temporal characteristics of the rainfall
variability. The dominance over time of various spatial modes is emphasized and the questions of synchroneity of the
hemispheres and the abruptness of change are considered. The contrast between the two hemispheres is also surveyed,
notably the different time scales of variability and potential causal factors in the variability. One of the most important
contrasts is the multi-decadal persistence of anomalies over most of northern Africa. This has implications for the causes of
long-term fluctuations, even those historical and paleo-time scales. q 2000 Elsevier Science B.V. All rights reserved.
Keywords: rainfall variability; decadal scale; century scale
1. Introduction
The climates of Africa during the late Pleistocene
and Holocene represent two extreme scenarios, with
the expansion of the Sahara at the peak of the last
glacial and the encroachment of the savanna and
steppe lands into the desert core after the glacial.
During modern times, however, remarkable fluctuations of rainfall have occurred over nearly the entire
)
Fax: q1-850-664-9642.
E-mail address: [email protected] ŽS.E. Nicholson..
continent. This paper presents an overview of the
nature of these fluctuations, emphasizing characteristics evident in the modern record and comparing
them with historical fluctuations derived from proxy
records.
The results indicate basic modes of variability that
occur on both short- and long-time scales and underscore the large magnitude of even recent fluctuations. They also suggest significant differences in the
causal mechanisms driving rainfall variability in
southern- and northern-hemisphere Africa, but a relative synchroneity of rainfall fluctuations in the two
0921-8181r00r$ - see front matter q 2000 Elsevier Science B.V. All rights reserved.
PII: S 0 9 2 1 - 8 1 8 1 Ž 0 0 . 0 0 0 4 0 - 0
138
S.E. Nicholsonr Global and Planetary Change 26 (2000) 137–158
hemispheres. Finally, information presented in this
article suggests that the magnitude of rainfall changes
associated with the extreme conditions of the late
Pleistocene and Holocene may not have been extremely different from those producing fluctuations
during recent historical times.
Fig.1. Schematic of the general patterns of winds, pressure and convergence over Africa Žfrom Nicholson, 1996.. Dotted lines indicate the
ITCZ, dashed lines, other convergence zones.
S.E. Nicholsonr Global and Planetary Change 26 (2000) 137–158
2. Climatological overview
Fig. 1 shows the prevailing circulation patterns
over Africa during JulyrAugust and January. These
correspond to the summer seasons of the northern
and southern hemispheres, respectively. In most of
Africa, however, the term summer is not strictly
139
appropriate, these months instead corresponding to
the Ahigh-sunB seasons of the two hemispheres. For
the sake of convenience, the term AsummerB will be
retained in this article, referring to the high-sun
seasons.
The two most apparent seasonal shifts ŽFig. 1. are
in the pressure over the Sahara and the location of
Fig. 2. Mean annual rainfall over Africa Žin mm.. Calculation is based on the entire length of record at each station and generally covers the
period 1925 to 1990 or earlier.
140
S.E. Nicholsonr Global and Planetary Change 26 (2000) 137–158
the convergence zones, particularly the Intertropical
Convergence Zone ŽITCZ.. In JulyrAugust, the low
pressure over the Sahara lies between the NE trades,
or Harmattan, and the humid ASWB monsoon flow.
The surface position of the ITCZ, which separates
these two wind systems, lies at about 188 to 208N. A
second convergence zone, the Zaire Air Boundary
ŽZAB. separates the flows off the Atlantic and In-
Fig. 3. Characteristics of rainfall seasonality over Africa. Top: month of maximum rainfall. Bottom: left, length of the rainy season
Žmonths.; right, seasonal character.
S.E. Nicholsonr Global and Planetary Change 26 (2000) 137–158
dian Oceans. High pressure prevails over southern
Africa. In January, the picture is much reversed, with
high pressure over the Sahara and a low over southern Africa. The convergence zones have moved
southward, with the ITCZ penetrating far into the
southern hemisphere.
The upper air patterns also shift seasonally. In
JulyrAugust, the upper-level flow is easterly over
most of the continent. Imbedded in the easterlies are
three jet streams, the African Easterly Jet ŽAEJ., the
Tropical Easterly Jet ŽTEJ. and a small easterly
maximum at about 108S. The AEJ and TEJ are
important for the development of the summer rainfall
regime over northern Africa. The AEJ, in particular,
provides energy for the development and maintenance of rain-bearing disturbances. Over the temperate extremes of the continent westerlies prevail. They
become more frequent in January, when those of the
northern hemisphere are displaced far southward.
The prevailing patterns of rainfall and its seasonality are a direct consequence of these circulation
patterns. In general, rainfall is associated with the
mid-latitude westerlies Ži.e., with the frontal systems
within them. and the convergence zones. Thus, the
longer the season during which one of these systems
dominates in a given region, the higher the mean
141
annual rainfall in that region. Thus, the rainiest
locations are the poleward extremes, where mean
annual rainfall is on the order of 800 to 1200 mm,
and the equatorial zone, where it is on the order of
1200 to 2000 mm ŽFig. 2.. Rainfall is also high over
the highland areas, such as those of eastern Africa,
Cameroon and Nigeria and coastal sectors of Liberia,
Sierra Leone and Guinea. The deserts lie in the
subtropical latitudes, where there is little influence of
either the westerlies or the convergence zones. There
are also desert areas in equatorial eastern Africa, a
consequence of more local effects ŽNicholson, 1996..
The desert regions tend to separate the summer
and winter rainfall regimes over Africa ŽFig. 3.. In
the northern hemisphere is a vast region where maximum rainfall occurs in August. Here, most of the
rainfall is associated with easterly waves and cloud
clusters; the energy and instability associated with
the AEJ are critical for the development of these
systems. This is roughly the Sahel–Soudan zone of
North Africa. Just to the south, over West Africa, is
an area with summer rains, but relatively dry conditions in August. Here, the rainfall is relatively localized in origin, rather than being linked to the largescale systems of the Sahel–Soudan. North of the
Sahara is an area of winter rainfall, brought by
Fig. 4. Development of Soudano–Saharan depressions. Plan view of 200 mb westerly subtropical jet and upper level trough superimpiosed
on tropical easterly flow at 850 mb Žfrom Nicholson, 1981a, based on Flohn.. This is a rough generalized schematic; latitude of the wind
systems fluctuates during the course of the year.
142
S.E. Nicholsonr Global and Planetary Change 26 (2000) 137–158
temperate frontal-type disturbances in the prevailing
mid-latitude westerlies.
In the Sahara itself, the rainfall regime is more
complex. The maximum rainfall generally occurs
during the transition seasons, but many areas have
secondary winter or summer influences. The disturbances that prevail during the transition seasons are
basically of extra-tropical origin ŽNicholson, 1981a.,
developing from elongated and usually diagonal
troughs in the mid-latitude westerlies ŽFig. 4.. However, the super-imposition of the tropical easterlies
beneath these troughs can be instrumental in the
development of the disturbances. The systems, such
as the one that brought catastrophic rains to Algeria
and Tunisia in September of 1969, can extend from
to Mediterranean coast to 108N. These occasionally
allow winter rainfall to penetrate into such southern
latitudes and they are responsible for most of the
significant rainfall events in the desert core.
In the southern hemisphere, the rainfall regimes
are also complex. Summer rains with a January
maximum are most common. However, large areas,
especially in the western portions of the sub-continent, experience transition-season maxima in rainfall. These may be linked to changes in the SST
patterns along the coasts. For example, the rainy
Fig. 5. Rainfall seasonality in individual years at four South African stations. The left-hand graphs are plots of monthly rainfall Žwith an
arbitrary magnitude. in each year with available data. The right-hand plots indicate the percent of rainfall occur during the cold season ŽMay
to October., with the horizontal axis ranging from 0% to 100%. Large blank areas indicate missing data.
S.E. Nicholsonr Global and Planetary Change 26 (2000) 137–158
143
Fig. 6. The most frequently occurring continental-scale rainfall anomaly types for the twentieth century Žfrom Nicholson, 1986; the values represent regionally averaged
standardized departures; positive values shaded..
144
S.E. Nicholsonr Global and Planetary Change 26 (2000) 137–158
season along the Benguela coast occurs when SSTs
reach their seasonal maximum. As in the northern
hemisphere, winter rains prevail along the poleward
extreme, brought by mid-latitude frontal systems
imbedded in the westerlies. Even in the summer
rainfall areas of the southern hemisphere, the westerlies appear to be important for the development of
rain-bearing disturbances, which take on a AhybridB
character and combine features of tropical and midlatitude systems.
Fig. 7. Decadally averaged rainfall anomalies for the 1950s, 1960s, 1970s and 1980s. These are expressed as a regionally averaged percent
of standard departure for 90 climatic regions. These patterns represent the two basic modes of rainfall variability over the continent:
anomalies of one sign prevailing, or anomalies of the opposite sign in equatorial and subtropical latitudes. The 1970s pattern is, however, an
unusual mode not seen at other times during the 20th century.
S.E. Nicholsonr Global and Planetary Change 26 (2000) 137–158
In the equatorial latitudes, there is a general pattern of two rainy seasons, both of which occur
during the transition season months. In the southern
equatorial latitudes, the second maximum shifts to
the summer season. The traditional explanation for
the seasonality is that the two maxima correspond to
the equatorial positions of the ITCZ during its northward and southward excursions. This explanation is
insufficient, however. For one, the ITCZ still lies
within parts of this region in summer. Also, over
East Africa the seasonality is generally linked to the
relative stability of the northeast and southerly trade
wind regimes in winter and summer, as well as to the
passage of the ITCZ. In general, little is known
about the synoptic systems bringing rain to equatorial Africa ŽNicholson, 1996.. There is some evidence of wave disturbances over equatorial East
Africa, but this is not unequivocal.
In the marginal areas between the winter and
summer rainfall regimes, the patterns of seasonality
are even more complicated. In some cases, such as
parts of South Africa, western Sahara and the Horn
of Africa, there are even three peaks in the seasonal
distribution. This reflects in part year-to-year changes
in the seasonal cycle, such as those illustrated in Fig.
5 for four relatively arid stations in western South
Africa. This figure shows both the seasonal distribution year-for-year at these stations and the percent of
rainfall falling during the cold season.
145
The four stations are in relatively close proximity,
as little as 28 of latitude or longitude apart. Nevertheless, two have summer rainfall ŽBeaufort West and
Pella., one has winter rainfall ŽSpringbok.; and at
one ŽCalvinia. rainfall can occur at any time. At all
of them, the seasonality of the rainfall switches from
year-to-year, including complete shifts between winter and summer rainfall. During the historical past,
prolonged shifts in the seasonality occurred. In the
1820s and 1830s, for example, the winter rainfall
regime appears to have penetrated further inland
ŽNicholson, 1981b..
In summary, the rainfall regimes over Africa are
quite complex and controlled by quite diverse
dynamic causes. It is hard to find a commonality
in the atmospheric dynamics producing the rainfall
that is implied by such a concept as the African
AmonsoonsB.
3. Nature of rainfall variability
3.1. Spatial modes of Õariability
Fig. 6 shows the principal spatial modes of rainfall variability over Africa. These were determined
ŽNicholson, 1986. using a correlation technique that
produces patterns analogous to principal components. The six most commonly occurring patterns
actually represent two basic modes of variability. In
Fig. 8. Rainfall fluctuations in the Sahel, eastern Africa and southern Africa from 1901 to 1994, expressed as regionally averaged
standardized departure. Regions represented are shown in the small map.
146
S.E. Nicholsonr Global and Planetary Change 26 (2000) 137–158
Fig. 9. Variance spectra for climatic parameters ŽNicholson and Entekhabi, 1986, 1987.. Ža. Rainfall in the Sahel and Soudan of West
Africa, in eastern Africa, and in southern Africa. Žb. SSTs along the Benguela Coast of the Atlantic and the Southern Oscillation Index. The
Southern Oscillation Index is a measure of the ENSO cycle.
S.E. Nicholsonr Global and Planetary Change 26 (2000) 137–158
147
Fig. 9 Ž continued ..
one, anomalies of the opposite sign prevail in equatorial and subtropical latitudes. In the other, rainfall
anomalies of the same sign prevail over most of the
continent. This mode generally appears as a reduction in rainfall throughout most of Africa.
The positive pole of the mode, above-average rain-
fall throughout the continent, occurs much less frequently.
These patterns, derived from an analysis of individual years, are also dominant on decadal and longer
time scales. The means for the 1950s, 1960s, 1970s
and 1980s decades are presented in Fig. 7. A strong
148
S.E. Nicholsonr Global and Planetary Change 26 (2000) 137–158
contrast is apparent between the 1950s, with abovenormal rainfall in subtropical latitudes and drier conditions in equatorial latitudes, and the 1960s, when
this spatial pattern was reversed. The changes were
both abrupt and large in magnitude Žsee Fig. 8.. In
the 1970s, an unusual pattern occurred: generally
below-average rainfall in the northern hemisphere,
but abnormally high rainfall throughout most of
southern-hemisphere Africa. Prior to the 1970s, this
pattern had occurred in only two or three individual
years since 1900. A speculative explanation for its
occurrence is given in Section 3.2. In the 1980s, the
spatial mode with AcontinentallyB uniform conditions occurred, with subnormal rainfall covering most
of Africa ŽNicholson, 1993..
These spatial patterns appear to be fundamental
modes of rainfall variability over Africa. They typify
the variations that occurred during recent historical
times ŽSection 4.. They are also evident in the
fluctuations characterizing the late Pleistocene and
Fig. 10. Time series of rainfall departures for the four individual regions of eastern africa and compared to the series for eastern Africa as a
whole Ždashed lines.. Values are expressed as a percent standard departure from the long-term mean Žfrom Nicholson, 1996.. The
seasonality of the regions represented is indicated by mean monthly rainfall Žbottom diagrams. for four representative stations. Region 31 is
north and west of Lake Victoria, region 32 is roughly centered on the lake, region 33 is to the northeast of the lake and region 38 is to the
south.
S.E. Nicholsonr Global and Planetary Change 26 (2000) 137–158
early to mid-Holocene ŽNicholson and Flohn, 1980..
These patterns indicate that despite regional contrasts
in the atmospheric dynamics producing the rainy
season, strong commonalities exist in the factors
governing the variability of rainfall on annual and
longer time scales. The coherence on a continental
scale indicates that these are large-scale factors linked
to the general atmospheric circulation. These basic
modes also indicate the existence of at least two
primary causal sets of factors.
3.2. Temporal characteristics of rainfall Õariability
These spatially coherent patterns indicate that the
interannual variability of rainfall over much of the
continent can be described using time series for a
few select regions. Those for the West African Sahel, eastern Africa and AsouthernB Africa Žapproximately, the latitudes from 258N to 258S. are shown
in Fig. 8. These are representative of the northernhemisphere subtropics, the equatorial region and the
southern-hemisphere subtropics. The trends in rainfall variability in the Sahel and southern Africa are
roughly parallel, with relatively dry conditions in the
1910s, the 1940s, and the 1980s and wetter condi-
149
tions in the 1950s. However, the contrast between
the two regions in the 1970s, as noted earlier, is also
clear. The trends over eastern Africa are roughly
out-of-phase with those in the other regions.
The most striking aspect of the series is the
multi-year persistence of rainfall anomalies in the
Sahel and the absence of this characteristic in the
other time series. This contrast is also apparent in the
variance spectra of rainfall over the continent. In
eastern and southern Africa ŽFig. 9a., the most
prominent time scales of variability are approximately 2.3, 3.5 and 5 to 6 years. The spectra of
rainfall variability in the West African regions of the
Sahel and Soudan show little high frequency variance. They are dominated by low frequency variance, with time scales of 7 years or greater accounting for over 50% of the variance in annual rainfall.
This characteristic is evident throughout nearly all of
the West African sector in which rainfall reaches a
maximum in August, but it is particularly strong in
the central regions relatively far from direct maritime
influences.
The rainfall fluctuations over eastern Africa illustrate another interesting aspect about variability, the
fact that the factors controlling rainfall variability
Fig. 11. Schematic of the distribution of significant spectral peaks with periods on the order of 2.2–2.4, 3.3–3.8 and 5–6 years, for rainfall
over the African continent and SSTs over the Atlantic and Indian Oceans Žbased on Nicholson and Entekhabi 1986; Nicholson and Nyenzi,
1990 and others.. For rainfall, vertical, horizontal and slanted lines indicate, respectively, significant peaks at about 3.5, 5–6 and 2.3 years.
For SSTs, the large bold number indicate the most significant peak; smaller numbers indicate secondary spectral peaks. The boldly outlined
squares are part of a tropical sector of strongly coherent SST variability on time scales of about 5 to 6 years.
150
S.E. Nicholsonr Global and Planetary Change 26 (2000) 137–158
may differ considerably from those that produce the
mean conditions. The diversity of climatic controls is
evident from the patterns of rainfall and its seasonality over eastern Africa. Mean annual rainfall ranges
from well over 1200 mm in Uganda and over the
highlands to less than 200 mm in the core of the
Kenyan deserts. Some areas having a single rainy
season during summer of the respective hemisphere,
others having two rainy seasons occurring during the
transition seasons, and some having a complex
regime that is a combination of these ŽFigs. 2 and 3..
Despite this diversity, interannual variability is remarkably coherent throughout the region, as the
similarity of the rainfall time series for four diverse
regional sectors shows ŽFig. 10.. Thus, the factors
governing variability in this region are much more
coherent throughout eastern Africa than those controlling mean climate. Interestingly, much of this
variability Žroughly 50%. is accounted for by conditions in October and November, which is not the
main rainy season in any of the sectors ŽNicholson,
1996..
3.3. Causes of rainfall Õariability oÕer Africa
The above discussion provides some general insights into factors governing rainfall variability. Specific factors that have been emphasized in studies of
variability include the El Nino
˜ phenomenon, seasurface temperatures ŽSSTs., and land-atmosphere
feedback. Each of these can alter the prevailing
atmospheric dynamics and circulation, such as the
Fig. 12. Time series of rainfall in the Sahel and southern Africa Žas in Fig. 7. and SST anomalies for equatorial sectors of the Atlantic and
Indian Oceans, 1946 to 1990. Units are standard departures for rainfall and degrees Centigrade for SSTs. SSTs represent an annual average
for the latitudinal sector from 108N to 108S.
S.E. Nicholsonr Global and Planetary Change 26 (2000) 137–158
Hadley and Walker circulations or upper level jet
streams Žsee Nicholson, 1989 for a review.. Although El Nino
˜ Žor ENSO. is centered on the Pacific,
it influences global atmospheric circulation and
world-wide sea-surface temperatures, which in turn
affect Africa.
The three timescales dominant in rainfall variability over eastern and southern Africa are evident both
in the ENSO cycle and in SSTs throughout the
tropical Atlantic and Indian Oceans ŽFigs. 9b and
11.. This suggests a possible Acause and effectB
relationship that has been confirmed in a number of
studies, including Nicholson and Entekhabi Ž1986,
1987., Nicholson and Kim Ž1997. and Nicholson
Ž1997a.. Overall, it appears that the Atlantic plays a
larger role than the Indian Ocean.
For ENSO, the relationship is strongest for interannual variability. It may be less important on
longer-time scales Ždecades to centuries. because of
its cyclic nature: in most areas, each cycle includes
periods of above-average and below-average rainfall.
SSTs, however, show low-frequency variations that
are linked to rainfall variability.
The contrast between the rainfall spectra of West
African rainfall and those of SSTs and ENSO would
initially suggest that these factors are not the main
ones governing rainfall variability. Nevertheless, numerous studies Že.g., Rowell et al., 1995. have suggested that they are a key mechanism for interannual
variability in West Africa. There is also some evidence that the interdecadal variability may be linked
to SSTs.
A possible association is suggested by the time
series in Fig. 12, depicting again rainfall in the Sahel
and southern Africa and comparing it with SSTs in
the equatorial Atlantic and Indian Oceans. There is a
clear contrast between abnormally low SSTs during
the wet 1950s and abnormally high SSTs during the
dry 1980s. This association is even larger in scale,
with the continental rainfall patterns of the 1950s
and 1980s being coincident, respectively, with cold
and warm SST anomalies throughout most of the
Atlantic and Indian Oceans from 408N to 408S. Also,
the warming trend in SSTs occurred in the 1960s,
when rainfall began to decline over much of Africa.
Nevertheless, caution is needed in concluding that
SSTs force decadal rainfall variability over Africa.
Rainfall changes may precede SSTs changes or may
151
lag them by several years. Also, SST patterns do not
consistently produce the expected anomaly in the
observational record. It is possible that the SSTs
produce a background state that favors either positive or negative rainfall anomalies, but that the more
complex mechanisms are required to produce the
stark decadal changes. This is particularly true for
the Sahel, where the overall character of variability
changes abruptly and dramatically from decade to
decade.
In the Sahel, these mechanisms may include feedback between the land surface and the atmosphere
ŽLare and Nicholson, 1994.. The hypothesis is that
the changes in soil moisture, vegetation cover, albedo
and so on produced by abnormally high or low
rainfall may in turn modify the large-scale atmosphere in such a way as to reinforce the patterns
producing the rainfall anomaly. Although this topic
is beyond the scope of this paper, it is relevant to
point out here that numerous papers, including that
by Harrison in this volume, have demonstrated that
in the Sahel such a mechanism probably serves to
enhance the rainfall variability produced by other
factors. It is quite possibly a cause of the strong
year-to-year persistence so characteristic of Sahel
rainfall variability but lacking elsewhere over Africa.
4. Rainfall variability during historical times
Various analyses of modern rainfall variability
over Africa identify common, underlying characteristics. Studies of rainfall fluctuations during the re-
Fig. 13. Plot of the total number of stations with rainfall data
during the period 1850–1899. Light shading: stations in South
Africa; no shading, stations in Algeria; dark shading; stations in
remaining countries in Africa.
152
S.E. Nicholsonr Global and Planetary Change 26 (2000) 137–158
Table 1
Examples of historical information useful for deriving past climates
ca. 1670–1760
ca. 1760–1820
1849
1850
1851
1852
1853
1854
1855
Lake Ngami was probably completely desiccated during this period. In about 1820, a very old man from the lake area
told Stigand that when he was a boy, old men of his tribe recalled the time when there was no lake in the area, but rather
a river Žthe Mokolane, or Thaoge., which ran through a wooded plain covered with tree species including Combretum
imberba and Acacia giraffae. Želsewhere said Mochwere, Mogotlo or camelthorn and Matshiara. The old men recalled
playing among these along the river’s banks. Confirmation of the story came when, during a recent desiccation of the
lake, stumps of these were found in areas formerly covered the lake. Campbell and Child state that the growth of
Combretum, a slow-growing species, requires that the lake had been dry for about 100 years or longer. Schwarz
independently suggests ca. 1670 as the first desiccation of the lake.
Lake Ngami was probably very extensive throughout this period, much deeper and more extensive than at any time in the
present century and than during most of the last century. Livingstone reports about 1850 that Old Magalakwe Ža Tswana
chief ca. 1830–1840. remembers when waves on the lake were frequently so high that hippos and fish were thrown to
shore. The Makarikari pans, which in Livingstone’s time were usually filled about half of the year, never dried up about
30 or 40 years previously Ži.e., ca. 1810 or 1820., according to older bushmen there, and the basin contained many fish,
hippos, and crocodiles. Chapman also heard such reports from natives and had been led to believe that the Botletle
became deep and broad in its lower course and that the Shua River was navigable; during his visits in the 1850s, he found
that these reports were no longer true. According to the people in the Makarikari area, the Ngami waters, which fed the
pans via the Botletle river suddenly dried up about 1820 or so; there is independent evidence that this must have been
toward 1820. Natives Žlake people and Makobas. variously reported that the lake had receded from 1 to 3 miles all
around within their lifetime, with a marked change within the 20 years before ca. 1850. Whites in the area also reported a
concurrent drying up of numerous fountains. Most natives remembered canoeing among the dark evergreens on Ngami’s
shores. Also toward 1800, the Mababe depression was a large lake and the Zambezi and Okavanga systems were linked
to Ngami such that peoples canoes along them to the lake. Also, the Bayei people canoed from the Mababe depression to
Lake Ngami via the Thamalakane River. Furthermore, it is apparent that in early 19th century, the Botletle consistently
flowed out of Lake Ngami, which is the only case when the lake retains a depth of about 10 ft or more. A bit further from
the Ngami, the Damaras and the Namaquas Awere loud in their claim that more rain fell half a century agoB Ži.e., in early
19th century..
Oswell, Livingstone and Murray rode along Ngami’s shores for 5 or 6 miles, estimating that it was about 14 miles across,
SSW to NNE, and about 8 miles wide. The Batuani natives told them that to walk around the lake, you travel about 50
miles Ž2 days. SW then 25 miles NW, thus the circumference is about 75 miles. Livingstone describes the Botletle as a
X
beautifully wooded river, lined with enormous baobabs, one of which was 76 in girth. A space on the west devoid of
trees indicated a recent higher stand of the lake. Oswell describes the lower end of the Zouga, indicating it reaches a
small lake Žprobably Lake Dow. and that it divides into two branches: one flowing eastwards to be lost in the salt pans
and the other NE and ENE toward the Matabele country. He describes natives poling mokoros, using 11 ft poles, along
the northeast end of the lake; indicates much higher than modern lake levels, as too shallow for such poles now. Oswell
also describes Botletle as flowing out of Ngami, a condition which requires a lake depth of 10 ft.
The lake and rivers flowing in or out of it were apparently rather full Žsee 1851..
Leyland found the Botletle River very low, with parts of it being very dry. Livingstone remarked that the river did not
X
rise within 3 of its 1850 level. Leyland was there in June.
Chapman reached the Botletle about 140 miles SE of Ngami Žnot far from the Ntwentwe Pan., where its bed contained
only pools of very salty water. This was in July.
In July, Anderson reached the western extreme of the lake, finding it receded and exceedingly shallow, with reeds lining
the shore. He estimated the circumference as 60 to 70 miles, with an average width of about 7 miles, and nowhere wider
than 9 miles. The northern shore was low and sandy with no vegetation of any kind for at least 1r2 miles from the shore,
but usually about a mile from the shore. Anderson canoed 13 days up the Botletle in August Žthe flood season. and found
X
the depth seldom less than 5 . Anderson described the lake level as low at the beginning of August but said it rose 3 ft
during their stay there. Chapman also reported reeds and rushes extending 1 to 2 miles along the lakeshore in late
November; he tested the lake’s depth and found it not to be more than 6 ft.. In October, he was at the Thamalakane and
wanted to go down by boat, but it was too shallow.
Chapman said in November 1853 that the plain of short prickly grass on saline soil surrounding lake had been formerly
covered with water, but during the last twenty years the lake’s waters had receded very much, generally for 2 or 3 miles
all around.
X
In January, Chapman found the swamps near Chapo’s Town to be dry, but near Shogotsa, the salt pans contained 2 of
water.
In June, Chapman sailed on Lake Ngami and found its depth to be generally about 3 1r2 ft. He boated on the Thaoge
River. In January, he found the Botletle in flood, overflowing its banks and flooding cornfields.
S.E. Nicholsonr Global and Planetary Change 26 (2000) 137–158
cent historical past can help to determine if these
characteristics are also apparent on longer time scales
and if they therefore represent fundamental features
of the rainfall regime over Africa. Such studies can
also serve to assess the magnitude of rainfall fluctuations required to produce major changes in the regional geography, e.g., vegetation patterns and surface hydrology.
Prior to 1900, few actual rainfall data are available except for the countries of South Africa and
Algeria ŽFig. 13.. The historical fluctuations must be
reconstructed instead using proxy indicators, such as
references to famine and drought, conditions of lakes
and rivers, agricultural practices, and so on. Further
details of the appropriate methodology are discussed
in Nicholson Ž1979..
An example of typical historical information is
given in Table 1. This example relates to general
time periods, but a vast amount of information is
also available for individual years. Although the
reliability of such information might be questioned,
if these Adata pointsB are used prudently and appropriately, credible reconstructions can be produced.
Fig. 14 shows a time series of rainfall variability in
central Namibia produced by such documentary evidence and compares it to actual rainfall measurements in the region. The agreement is excellent.
Figs. 15–19 present several historical reconstructions for Africa. Fig. 15 shows the continental patterns of rainfall anomalies, based on proxy data, that
Fig. 14. Time series of seasonal rainfall character for central
Namibia from 1850 to 1903. Units on the left indicate anomaly
classes: q3, q2, q1 correspond to extraordinarily wet, very wet
and wet years, 0 corresponds to normal conditions; and y1, y2
and y3 correspond to relatively dry, very dry, and severe drought,
respectively. Superimposed on this is the rainfall record for Rehoboth, a station in region 69; units Žon the right. are millimeters.
153
prevailed during three past periods: the 1820s and
1830s, ca. 1870–1895, and ca. 1895 to 1920. All
three of these periods stood out in records throughout the continent and represent rather rapid transitions to wet or dry states. Clearly, the continentally
coherent patterns derived from the modern record are
apparent here, as is the pattern of opposition between
the equatorial regions and subtropics.
Fig. 16 shows the historical reconstructions of the
levels of numerous African lakes; those for Victoria,
Malawi and Chilwa are shown in more detail in Figs.
17 and 18. Two particular features are evident from
these diagrams. First, is the relative synchroneity of
the fluctuations. This includes a relative synchroneity
between the two hemisphere, as illustrated by a
comparison between the long-term fluctuations of
Lakes Malawi, Chilwa and Chad ŽFig. 19.. Second,
the fluctuations are large in magnitude, with most
lakes having varied by at least 5 to 10 m over the
course of the 19th and 20th centuries. The high
stands of the 19th century were in most cases regained briefly in the 1960s or 1970s.
None of the historical indicators provides a precise measurement of rainfall, but rough estimates can
be gleaned from several sources. Grove Ž1972. shows
that the Nile discharge during the period ca. 1880–
1895 was about 35% greater than for the period
1910–1940. In West Africa, rainfall at Freetown,
Sierra Leone, declined by a comparable amount during the same period ŽNicholson, 1981a,b.. This is
comparable to the 15% to 35% increase that the
model of Hastenrath and Kutzbach Ž1983. indicated
would sustain the high early Holocene stands of the
East African lakes. Shaw Ž1985. concluded that the
nineteenth century high stands of Lake Ngami required that it receive only about 11% of the Okavango’s current total inflow, implying a rainfall increase of about this same order of magnitude. Owen
et al. Ž1990. calculated that a 120-m transgression of
Lake Chilwa could have been produced by a 30% to
50% increase in rainfall, compared to the present
century, if these conditions persisted for about a
century.
Anomalies of this magnitude are not unusual today. In Southern Africa, rainfall for the periods
1967–1969 and 1974–1978, two periods of high
lake levels ŽFig. 20., was on the order of 20–40%
above normal; the rapid transgression of the East
154
S.E. Nicholsonr Global and Planetary Change 26 (2000) 137–158
Fig. 15. African rainfall anomalies for three historical periods Žbased on material in Nicholson, 1978, 1980, 1981a,b, 1995.. Minus signs denote evidence of drier conditions; plus
signs, above-average conditions; small circles, near-average conditions; circled symbols, regional integrators such as lakes and rivers.
S.E. Nicholsonr Global and Planetary Change 26 (2000) 137–158
Fig. 16. Schematic of variations of the levels of African lakes in
the 19th and 20th centuries. Both the trends and the magnitudes of
the fluctuations are highly generalized. Sketch based on Sieger
Ž1887., Pike and Rimmington Ž1965., Nicholson Ž1976., Nicholson Ž1978, 1980., Vincent et al., Ž1979., Maley Ž1981., Crossley
et al. Ž1984., Shaw Ž1985., Owen et al. Ž1990., Hastenrath Ž1988.
and others.
African lakes in the early 1960s was also associated
with rainfall about 20–40% above the mean ŽFig.
16.. However, there were much larger increases in
1961, the year of the abrupt rise in lake levels. The
stations of Wajir, Eldama and Lokitaung received
612, 402 and 302 mm, respectively, in November,
compared with monthly means of 58, 48 and 39 mm
ŽNicholson, 1996.. Similar, but not quite so extreme,
Fig. 17. Fluctuations of Lake Victoria since 1800, based on
documentary evidence in earlier years and actual measurements as
of 1896 Žfrom Nicholson, 1997b..
155
Fig. 18. Fluctuations of Malawi and Chilwa since 1800, based on
documentary evidence in earlier years and actual measurements as
of 1896 for Malawi and 1949 for Chilwa Žfrom Nicholson,
1997c..
conditions occurred again in 1963. In the Sahel,
rainfall during the 1950s was about 25–40% above
the long-term mean ŽTable 2..
Fig. 19. Rough sketch of fluctuations of Lakes Victoria, Chilwa
and Chad since 800 A.D., based on Nicholson, 1997c and Maley,
1981..
156
S.E. Nicholsonr Global and Planetary Change 26 (2000) 137–158
Fig. 20. Maps of rainfall anomalies in Southern Africa for the periods 1961–1964, 1967–1969. and 1974–1978, expressed as a percent
above or below the long-term mean for 18 grid squares.
S.E. Nicholsonr Global and Planetary Change 26 (2000) 137–158
Table 2
Mean annual rainfall for two multi-year periods at select stations
in the Sahel and Soudan zones of West Africa
Mean rainfall Žmm.
BILMA
ATBARA
NOUAKCHOTT
KHARTOUM
AGADEZ
TIMBUKTOU
NEMA
DAKAR
BANJUL
1950–1959
1970–1984
20
92
172
178
210
241
381
609
1409
9
54
51
116
97
147
210
308
791
Thus, the long-term changes of environment during the recent historical past may represent primarily
a more persistent occurrence of conditions commonly characterizing briefer periods of the twentieth
century. This might also be the case for longer time
scales: the climatic conditions that sustained the
expanded Sahara of the late Pleistocene or the savanna conditions and deep lakes of the early
Holocene might not have been vastly different than
those occurring occasionally even today.
5. Summary and conclusions
Overall, our results have shown that the features
of African rainfall variability, evident in the modern
record, also characterise the historical past. These
characteristics appear to be fundamental features of
the rainfall regime, and hence can be useful in both
filling in and interpreting the paleoclimate record.
The major fluctuations occur roughly synchronously over most of the continent. Thus, rainfall
fluctuations are relatively synchronous in the southern and northern hemispheres.
On the other hand, the causes of variability are
probably significantly different in the two hemispheres. In semi-arid regions of the northern hemisphere, there is a strong inter-annual persistence of
rainfall anomalies that may be indicative of a landatmosphere feedback mechanism. This characteristic
is absent in southern-hemisphere sectors of Africa.
It is important to distinguish between the factors
producing the mean climate over Africa, which are
157
quite diverse from region to region, and the factors
controlling the temporal variability of rainfall. The
latter are much more uniform throughout the continent and must therefore be large-scale aspects of the
general atmospheric circulation, such as the Walker
or Hadley circulations or monsoon intensities, or
ocean influences, such as sea-surface temperatures.
The magnitude of changes needed to produce the
climatic conditions of the late Pleistocene and
Holocene may not have been greatly different than
those that occur at least occasionally during modern
times.
Acknowledgements
Recent work described in this article was sponsored by NSF Grants ATM-9024340 and ATM9417063 and NOAA Grant NA46GP0285.
References
Crossley, R., Davison-Hierschmann, S., Owen, R.B., Shaw, P.,
1984. Lake level fluctuations during the last 2000 years in
Malawi. In: Vogel, J.C. ŽEd.., Late Cainozoic Palaeoclimates
of the Southern Hemisphere ŽRotterdam.. pp. 305–316.
Grove, A.T., 1972. Climatic change in Africa in the last 20,000
years, Les Problemes de Developement du Sahara Septentrional Ouarghla Colloquium vol. 2, ŽAlgiers..
Hastenrath, S.L., 1988. Climate and Circulation of the Tropics.
Reidel, Dordrecht.
Hastenrath, S.L., Kutzbach, J.E., 1983. Paleoclimatic estimates
from water and energy budgets of East African lakes. Quat.
Res. 19, 141–153.
Lare, A.R., Nicholson, S.E., 1994. Contrasting conditions of
surface water balance in wet years and dry years as a possible
land surface–atmosphere feedback mechanism in the West
African Sahel. J. Clim. 7, 653–668.
Maley, J., 1981. Etudes palynologiques dans le Bassin du Tchad
et Paleoclimatologie de l’Afrique nord tropicale de 30,000 ans
a` l’epoque actuelle, Travaux et Documents de l’O.R.S.T.O.M.,
129, Paris.
Nicholson, S.E., 1978. Climatic variations in the Sahel and other
African regions during the past five centuries. J. Arid Environ.
1, 3–24.
Nicholson, S.E., 1979. The methodology of historical climate
reconstruction and its application to Africa. J. Afr. Hist. 20,
31–49.
Nicholson, S.E., 1980. Saharan climates in historic times. In:
Williams, M.A.J., Faure, H. ŽEds.., The Sahara and the Nile
ŽRotterdam.. pp. 173–200.
158
S.E. Nicholsonr Global and Planetary Change 26 (2000) 137–158
Nicholson, S.E., 1981a. Rainfall and atmospheric circulation during drought periods and wetter years in Africa. Mon. Weather
Rev. 109, 137–154.
Nicholson, S.E., 1981b. The historical climatology of Africa. In:
Wigley, T.M.L., Ingram, M.J., Farmer, G. ŽEds.., Climate and
History ŽCambridge.. pp. 249–270.
Nicholson, S.E., 1986. The spatial coherence of African rainfall
anomalies: inter-hemispheric teleconnections. J. Clim. Appl.
Meteorol. 25, 1365–1381.
Nicholson, S.E., 1989. African drought: characteristics, causal
theories and global teleconnections. In: Berger, A., Dickinson,
R.E., Kidson, J.W. ŽEds.., Understanding Climate Change
ŽAmerican Geophysical Union.. pp. 79–100.
Nicholson, S.E., 1993. An overview of African rainfall fluctuations of the last decade. J. Clim. 6, 1463–1466.
Nicholson, S.E., 1995. Environmental change within the historical
period. In: Goudie, A.S., Adams, W.M., Orme, A. ŽEds.., The
Physical Geography of Africa. Oxford Univ. Press, pp. 60–75.
Nicholson, S.E., 1996. A review of climate dynamics and climate
variability in eastern Africa. In: Johnson, T.C., Odada, E.
ŽEds.., The Limnology, Climatology and Paleoclimatology of
the East African Lakes. Gordon & Breach, pp. 25–56.
Nicholson, S.E., 1997a. An analysis of the ENSO signal in the
tropical Atlantic and western Indian Oceans. Int. J. Climatol.
17, 345–375.
Nicholson, S.E., 1997b. Historical fluctuations of Lake Victoria
and other lakes in the northern Rift Valley of East Africa. J.
Great Lakes Res., in press.
Nicholson, S.E., 1997c. Fluctuations of Rift Valley Lakes Malawi
and Chilwa during historical times: a synthesis of geological,
archaeological and historical information. J. Great Lakes Res.,
in press.
Nicholson, S.E., Entekhabi, D., 1986. The quasi-periodic behavior
of rainfall variability in Africa and its relationship to the
Southern Oscillations. Arch. Meteorol., Geophys. Bioklimatol.,
Ser. A 34, 311–348.
Nicholson, S.E., Entekhabi, D., 1987. Rainfall variability in equatorial and southern Africa: relationships with sea surface temperatures along the southwestern coast of Africa. J. Clim.
Appl. Meteorol. 26, 561–578.
Nicholson, S.E., Kim, J.Y., 1997. The relationship of the El Nino
˜
Southern Oscillation to African rainfall. Int. J. Climatol. 17,
117–135.
Nicholson, S.E., Nyenzi, 1990. Temporal and spatial variability of
SSTs in the Tropical Atlantic and Indian Oceans. Meteorol.
Atmos. Phys. 42, 1–17.
Owen, R., Crossley, R., Johnson, T., Tweddle, D., Kornfield, I.,
Davison, S., Eccles, D., Engstrom, D., 1990. Major low levels
of Lake Malawi and their implications for speciation rates in
cichlid fishes. Proc. R. Soc. London. pp. 519–553.
Pike, J.G., Rimmington, G.T., 1965. Malawi: a Geographical
Study. Oxford.
Rowell, D.P., Folland, C.K., Maskel, K., Ward, M.N., 1995.
Variability of summer rainfall over tropical north Africa
Ž1906–92.: observations and modelling. Q. J. R. Meteorol.
Soc. 121, 669–704.
Shaw, P., 1985. The desiccation of Lake Ngami: an historical
perspective. Geogr. J. 151, 318–326.
Sieger, R., 1887. Schwankungen der innerafrikanischen seen.
Jahresber., Ver. Geogr. ŽVienna.. pp. 41–60.
Vincent, C., Davies, T., Beresford, A., 1979. Recent changes in
the level of Lake Naivasha, Kenya, as an indicator of equatorial westerlies over East Africa. Clim. Change 2, 175–189.