A high-resolution geochemical record from Lake Edward, Uganda Congo and the timing and causes of tropical African drought during the late Holocene

A high-resolution geochemical record from Lake Edward, Uganda Congo and the timing and causes of tropical African drought during the late Holocene

ARTICLE IN PRESS Quaternary Science Reviews 24 (2005) 1375–1389 A high-resolution geochemical record from Lake Edward, Uganda Congo and the timing a...

523KB Sizes 0 Downloads 0 Views

ARTICLE IN PRESS

Quaternary Science Reviews 24 (2005) 1375–1389

A high-resolution geochemical record from Lake Edward, Uganda Congo and the timing and causes of tropical African drought during the late Holocene James M. Russella,b,, Thomas C. Johnsonb a

Limnological Research Center, University of Minnesota, 310 Pillsbury Dr. SE, Minneapolis, MN, 55455, USA Large Lakes Observatory, University of Minnesota Duluth, Research Lab Bldg., Duluth, MN, 55812-2496, USA

b

Received 9 June 2004; accepted 8 October 2004

Abstract High-resolution analyses of the elemental composition of calcite and biogenic silica (BSi) content in piston cores from Lake Edward, equatorial Africa, document complex interactions between climate variability and lacustrine geochemistry over the past 5400 years. Correlation of these records from Lake Edward to other climatically-forced geochemical and lake level records from Lakes Naivasha, Tanganyika, and Turkana allows us to develop a chronology of drought events in equatorial East Africa during the late Holocene. Major drought events of at least century-scale duration are recorded in lacustrine records at about 850, 1500, 2000, and 4100 cal year BP. Of these, the most severe event occurred between about 2050 and 1850 cal year BP, during which time Lake Edward stood about 15 m below its present level. Numerous additional droughts of less intensity and/or duration are present in the Lake Edward record, some of which may be correlated to other lacustrine climate records from equatorial East Africa. These events are superimposed on a long-term trend of increasingly arid conditions from 5400 to about 2000 cal year BP, followed by a shift toward wetter climates that may have resulted from an intensification of the winter Indian monsoon. Although the causes of decadeto century-scale climate variability in the East African tropics remain obscure, time-series spectral analysis suggests no direct linkage between solar output and regional rainfall. Rather, significant periods of 725, 125, 63–72, 31–25, and 19–16 years suggest a tight linkage between the Indian Ocean and African rainfall, and could result from coupled ocean-atmosphere variability inherent to the tropical monsoon system. r 2004 Elsevier Ltd. All rights reserved.

1. Introduction Paleoclimate records have shown that the Holocene climate of tropical East Africa was punctuated by numerous decade- to millennial-scale arid events (e.g. Lamb et al., 1995; Verschuren et al., 2000). There remains little consensus regarding the precise timing, geographic extent, and causes of these events, partly due to a paucity of well-resolved paleoclimate records from the East African tropics (Gasse, 2000). Most records lack high sampling resolution, or are of insufficient Corresponding author. Tel.: +1 612 626 7889; fax: +1 612 625 3819. E-mail address: [email protected] (J.M. Russell).

0277-3791/$ - see front matter r 2004 Elsevier Ltd. All rights reserved. doi:10.1016/j.quascirev.2004.10.003

length to place drought events within the context of millennial climate trends and rhythms. Resolving the patterns and causes of prolonged African drought is of critical importance to paleoclimatologists, not only because of the importance of tropical regions to the global water cycle but due to the socioeconomic disasters that can result from prolonged droughts in sub-Saharan Africa (Oldfield and Alverson, 2003). Here we present a detailed record of rainfall and drought based upon multi-proxy analyses of piston cores from Lake Edward, Uganda-Congo. This work builds upon previous studies of the paleoenvironmental history of Lake Edward, and provides evidence for numerous droughts during the late Holocene that affected most of the East African tropics. Our results

ARTICLE IN PRESS J.M. Russell, T.C. Johnson / Quaternary Science Reviews 24 (2005) 1375–1389

1376

highlight the geochemical complexity of lacustrine responses to climate variability, and implicate the tropical oceans as an important source for century-scale climate variability in tropical Africa.

open, draining to Lake Albert to the north via the Semliki River, but water loss by evaporation currently exceeds surface outflow by about 20% (Russell, 2004). As a result of high evaporation rates, Lake Edward’s water level and chemistry are extremely sensitive to rainfall variability (Lehman, 2002). Rainfall averages 900 mm/year at the elevation of Lake Edward (Viner and Smith, 1973), and comes primarily during two rainy seasons associated with the migration of the intertropical convergence zone (ITCZ) across the equator: the ‘long rains’ from March to May, and the ‘short rains’ from October to November. Instrumental weather records indicate that fluctuations in the short rains account for significantly more interannual variability than anomalies in the long rains (Nicholson, 1996). Sea surface temperature (SST) gradients in the Indian Ocean exert significant influence on rainfall in the region, and particularly on the intensity of the short rains (Nicholson, 1996; Birkett et al., 1999; Murtugudde et al., 2000). In addition, incursions of humid westerly flow derived from the Atlantic cause high rainfall in East Africa (Nicholson, 1996). El Nin˜o Southern Oscillation (ENSO) events are positively correlated with rainfall in the region (Nicholson, 1996), although the influence of ENSO events

2. Lake Edward, Uganda Congo: limnology and climate setting Lake Edward (01N, 301E, 912 m a.s.l.) is situated in a half-graben on the border between Uganda and the Democratic Republic of the Congo. Lake Edward is the smallest of the great rift lakes of East Africa, with a surface area of 2325 km2, a maximum depth of 117 m, and a mean depth of 37 m (Fig. 1) (Lehman, 2002). The lake has a conductivity of 900 mS/cm, an average pH of 9, and a chemistry dominated by Na+, K+, Mg2+, and HCO 3 (Talling and Talling, 1965; Hecky and Degens, 1972). Lake Edward is oligomictic and eutrophic (Hecky and Degens, 1972). The largest source of water to the lake is surface runoff from highland areas surrounding the lake: the Ruwenzori Mountains to the north, the Kigezi highlands to the east, and the Virunga Volcanoes to the south, as well as the Kazinga Channel which drains Lake George. Lake Edward is presently

30o E

29o 30' E

N 0

10 km

o

Core Site

E96-1P 30

E96-2P

E96-6M

E96-5M

90 70 50

10o N

0o 30' S 30 10

N

Nile River Congo River L. Albert Lake Edward L. Kivu

L. Turkana

L. Naivasha Indian Ocean

0o

L. Victoria L. Tanganyika L. Malawi 10o S 30o E

40o E

Fig. 1. Bathymetric Map of Lake Edward with coring sites, adapted from Russell et al. (2003a). Insets show the position of Lake Edward within the chain of East African Great Lakes and the location of the region within Africa.

ARTICLE IN PRESS J.M. Russell, T.C. Johnson / Quaternary Science Reviews 24 (2005) 1375–1389

Research activities spurred by an International Decade for East African Lakes (IDEAL) piston coring expedition to Lake Edward in 1996 have begun to unravel a rich paleoclimatic history from the lake (Lærdal et al., 2002; Russell et al., 2003a, b; Beuning and Russell, 2004). Lithostratigraphic data indicate that Lake Edward experienced a transition from a wet early Holocene to a more arid late Holocene by about 5200 cal year BP (Russell et al., 2003a), a shift noted at virtually all sites in equatorial and north Africa and likely driven by the precessional orbital cycle (Gasse, 2000). Lake Edward subsequently experienced a period of low lake levels beginning about 4000 cal year BP that culminated in a major lowstand at 2000 cal year BP, after which time lake levels rose to near their modern elevation (Russell et al., 2003a). High-resolution analyses of the stable isotopic and %Mg composition of inorganic calcite in piston cores provide evidence for a drought cycle in Lake Edward with a period of 725 years since 5400 cal year BP (Russell et al., 2003b). Based upon the similarity of this period to climate cycles in marine records from the Indian and western Pacific Oceans, Russell et al. (2003b) suggest that millennialscale variability in tropical East Africa during the late Holocene is strongly linked to fluctuations in the Indian monsoon. Here we present a new, high-resolution record of %Biogenic Silica (%BSi) from Lake Edward spanning the interval from 900 to 3500 cal year BP. Using geochemical data and cross-proxy correlation, we demonstrate that %BSi is a sensitive proxy for water balance in Lake Edward. We evaluate the timing and regional correlation of droughts observed in the Lake Edward during the past 5000 years, and discuss the causes of decade to century-scale drought in tropical Africa.

All cores for the present study are archived at the National Lacustrine Core Repository (LacCore) at the University of Minnesota (Schnurrenberger et al., 2001). Percentage of Mg data, and chronologies for these cores, based on AMS 14C dates on hand-picked terrestrial plant fragments and charcoal, have been

10 20 30 40 50 60 70

1

2

3

4

5

6

7

8

1

1

1.5

1.5

2

2

2.5

2.5

3

3.5

4. Materials and methods

Depth (m)

% Biogenic Silica 0

Age (cal Kyr BP)

3. Lake Edward, East Africa: previous paleoclimatic studies

published previously (Russell et al. 2003a, b; Beuning and Russell, 2004). All dates are reported in cal year BP (before 1950) unless otherwise noted. Samples for %BSi analysis were taken at 1-cm continuous intervals from core E96-1P, dated by six AMS 14C dates on charcoal that were used to develop an age model based upon linear interpolation through calibrated ages (Fig. 2; Russell et al., 2003a). These samples were freeze-dried and homogenized, then 20 mg subsamples were analyzed for %BSi at 10 cm intervals using multiple extraction of hot alkaline digestions at 85 1C in 0.5 M NaOH (modified after DeMaster, 1979). These data were used to determine digestion times and correction factors for Si released from detrital mineral matter for the single extraction method of Mortlock et al. (1989). All samples were then analyzed in duplicate using the single extraction technique with a 60-min extraction time and a correction of 1.1% silica to account for inorganic mineralderived Si. The average standard deviation of duplicate single-extraction analyses was 0.43% BSi, and the correlation coefficient between single and multiple extraction analyses was 0.98 (n ¼ 76). Russell et al. (2003a) noted the presence of nodules and concretions of opal (amorphous SiO2) within E961P. An unsampled half of core E96-1P, archived at LacCore, was X-rayed to examine the distribution of these nodules and its relationship with the %BSi data. The diameter of each nodule was measured directly from

Age (cal Kyr BP)

appears to be modulated by ENSO’s impacts on SST’s in the Indian and Atlantic oceans (Nicholson et al., 1997). Thus, rainfall in the Lake Edward region is strongly influenced by SSTs in the adjacent tropical oceans.

1377

3

0

5

10

15 20 25 30

35

0

3.5

Cumulative Opal-A′ Nodule Diameter (mm)

Fig. 2. Right: %BSi in E96-1P spanning the past 3500 years. Shaded horizontal bars indicate the cumulative diameter of opal-A0 nodules summed per cm of core (the lower x-axis). Note that high concentrations of nodules generally occur during periods of low %BSi. Left: Depth-Age model for E96-1P. The model is based on linear interpolation through six dates on fine-grained charcoal, as described in Russell et al. (2003a). Error bars represent 2-sigma calibrated age-ranges.

ARTICLE IN PRESS J.M. Russell, T.C. Johnson / Quaternary Science Reviews 24 (2005) 1375–1389

the X-ray image and its depth in the core was noted. Nodules and concretions were also sampled and characterized using scanning electron microscopy (SEM), energy dispersive spectrometry (EDS), and Xray diffraction (XRD) at the Center for Interfacial Engineering Characterization Facility, University of Minnesota.

180 140

Counts

1378

100 60 20

5. Results Percent of BSi measurements from E96-1P span the interval from 900 to 3450 cal year BP, and range from 2% to nearly 70% BSi (Fig. 2). Generally high values (average of 35%) from 3450 to 1500 cal year BP are interrupted by four periods of very low %BSi centered at 1900, 2750, 3100, and 3400 cal year BP, in addition to considerable decade-scale variations. %BSi values are generally lower from 900 to 1500 cal year BP (average 27%), with pronounced century-scale variation. X-ray images from core E96-1P indicate the presence of dense, round nodules with diameters ranging from 2 to nearly 30 mm. SEM-EDS and XRD analyses demonstrate that these nodules are composed of amorphous inorganic opal-A0 . This phase is indicated by the nodules’ elemental composition (495% Si by weight) and their XRD pattern: a broad peak at 4.1 A˚ with no secondary peaks that would indicate the presence of crystalline phases such as cristobalite or tridymite (Fig. 3, Jones and Segnit, 1971; Hein et al., 1978). Under SEM, the nodules were observed to be 2–5 mm ‘lepispheres’ (Weaver and Wise, 1972) often coalesced into cements, and were seen to encase and replace diatom frustules (Fig. 3). The sum of the diameter per centimeter of these nodules plotted against depth indicates that concentrated beds of these nodules occur at distinct depths generally associated with low %BSi values (Fig. 2). No opal concretions were found in sediments younger than 1400 cal year BP.

6. Interpretation Our approach combines lithostratigraphic and multiproxy geochemical methods to investigate the paleohydrology of Lake Edward during the late Holocene. This approach has yielded a detailed reconstruction of moisture balance from equatorial East Africa. 6.1. Interpretation of the %BSi record in E96-1P Percent of BSi is a first-order proxy for production by diatoms, single-celled algae that excrete an exoskeleton composed of amorphous hydrous silica (opal, SiO2  nH2O). Many studies of the African Great Lakes have

5

10 15 20 25 30 35 40 45 50

Two-Theta Angle (deg)

Fig. 3. Top left: X-ray diffractogram of an opal-A0 nodule from core E96-1P section VI 11 cm depth. Right: X-radiography of core E96-1P section IX, scale at right is in cm. Note the presence of dense nodules composed of inorganic opal. Bottom: SEM images of nodules from 1P section VII 27 cm depth (left) and section VI 10 cm depth (right). Note the weak cementation in the former and dense cementation in the latter.

interpreted %BSi to reflect variations in wind intensity and lake upwelling (e.g. Stager and Johnson, 2000; Johnson et al., 2002). Upwelling transfers dissolved SiO2 and phosphate to the surface mixed layer where they are utilized by algae, causing increased diatom production and high-sedimentary BSi accumulation (Johnson, 2002). This interpretation of %BSi is based on studies of algal succession in relation to the seasonal onset of upwelling in the large rift lakes Malawi and Tanganyika (Hecky and Kling, 1987), and may not explain BSi sedimentation in all of the African Great Lakes at all time-scales. Other studies have interpreted %BSi fluctuations to reflect variations in rainfall or hydrologic balance due to climatic effects of the geochemical mass balance of silica (e.g. Hu et al., 2003). Measurements of dissolved SiO2 concentrations ([SiO2]aq) in the water column of Lakes Malawi (Johnson, 2002) and Edward (Fish, 1953; Russell, 2004) suggest that long-term variability in wind intensity and upwelling are not likely to control diatom burial on long time-scales. Water column profiles of dissolved Si in Edward are typical of African rift lakes, and show elevated [SiO2] in the hypolimnion that result from diatom dissolution at depth (Fig. 4). However, the size of the hypolimnion relative to the epilimnion in Lake Edward is quite small, and it cannot hold large masses of dissolved nutrients relative to those already present in the epilimnion (Fig. 4). Thus, it is unlikely that changes in water-column stratification can support large, decadeto century-scale changes in %BSi such as those observed

ARTICLE IN PRESS J.M. Russell, T.C. Johnson / Quaternary Science Reviews 24 (2005) 1375–1389 ● [SiO2]

0

200



0



▼ ●

20

Water Depth (m)

rates are high. This is exactly what one would predict if hydrologic inputs of dissolved SiO2 control the %BSi in sediments in Lake Edward.

µM

100

1379





Epilimnion

6.2. Diagenetic effects on %BSi

40 Hypolimnion

● ▼

60 ● ▼

80



100 ▼

▼ Moles

4.0

3.0

2.0

1.0

0

120

Si x 109

Fig. 4. Filled circles: dissolved SiO2 concentrations averaged from Fish (1953) and 1996 measurements from Lake Edward’s surface to 80 m water depth. Filled triangles: total dissolved SiO2 hypsograph calculated from SiO2 concentration data and bathymetric data in Lærdal (2000) at water depth intervals of 0–20, 20–40, 40–70, 70–90, and 90–110 m. The SiO2 concentration for the latter interval was extrapolated from the trend in SiO2 concentration measured at 60 and 80 m water depth. Note that despite much higher concentrations in the hypolimnion the total amount of dissolved SiO2 is considerably lower in the hypolimnion that in the epilimnion.

in E96-1P, as periods of intense upwelling will quickly deplete hypolimnetic [SiO2]aq (Fig. 2). The residence time of dissolved SiO2 in Lake Edward relative to river inputs is approximately 4 year, so longterm changes in river inputs of Si can substantially alter BSi burial rates in Lake Edward. The mechanism linking burial of diatoms silica to river input of SiO2 is likely related to coupled variability in Si inputs, Si burial, and biogenic Si dissolution (Johnson, 2002). If diatom burial is high relative to river inputs of silica, [SiO2]aq will fall, promoting the dissolution of BSi and limiting rates of BSi burial. Conversely, during wetter periods, river delivery of Si to the lake is high, promoting higher [SiO2]aq in the water and high rates of BSi burial (Johnson, 2002). Support for our interpretation of %BSi as a proxy for water balance comes from a strong negative correlation between the %Mg in inorganic calcite and %BSi in core E96-1P (Fig. 5). Both limnologic and geologic studies of Lake Edward have shown that the %Mg in calcite is negatively correlated with water balance in Lake Edward (Lehman, 2002; Russell et al., 2003b), as is common of carbonates in lakes (Curtis and Hodell, 1993). Profiles of %Mg in calcite and %BSi have a correlation coefficient r ¼ 0:771 (Fig. 5), indicating that during periods of drought %BSi and diatom burial rates are low, while during wetter periods diatom burial

Whole core X-ray studies show that discrete beds of sand to gravel-size nodules and concretions of opal-A0 are present during several periods of low %BSi values in E96-1P (Figs. 2 and 3). Studies of hypersaline alkaline brines in the USA and Africa have documented numerous water bodies that inorganically precipitate opal (Jones et al., 1967). [SiO2]aq in Lake Edward ranges from about 80 mM in surface waters to 260 mM in the hypolimnion (Fig. 4). Neither value is close to saturation (2000 mM) with respect to amorphous silica (Jones et al., 1967), but the presence of inorganic opal nodules implies much higher [SiO2]aq in Lake Edward, at least in surface sediments, in the past. Jones et al. (1967) found that many alkaline brines had SiO2 concentrations near opal saturation at the sediment–water interface, and much lower [SiO2]aq in both the overlying water column and deeper in the sediments. Jones et al. interpreted these patterns to indicate silicate dissolution at the sediment–water interface, and opal precipitation just beneath the sediment– water interface. Today Lake Edward is mildly alkaline and has a pH of 9, but the solubility of silica rises rapidly at pH49.2. During lake lowstands, evaporative concentration of dissolved carbonate could raise the alkalinity and pH of Lake Edward substantially (Lehman, 2002), conditions that would favor the release of silica from diatom frustules in surface sediments. The dissolved silica could then precipitate as opal in shallow surface sediments at pH gradients caused by organic matter decomposition in Lake Edward’s sediments, similar to the trends observed by Jones et al. (1967) (see also Hesse, 1988). The presence of diagenetic products of diatom dissolution somewhat complicates our interpretation of %BSi profiles in Lake Edward. However, we note that both opal- A0 precipitation and reduced diatom burial rates in response to lowered [SiO2]aq cause low %BSi values during periods of drought. Thus, both opal precipitation and silica mass balance have similar effects on %BSi in relation to climate, allowing us to use %BSi as a qualitative indicator of drought. Due to its chemistry and short residence time, we suggest that %BSi is a sensitive proxy for drought in Lake Edward. All but two of the nodules are found in sediments that predate 1800 cal year BP (Fig. 2), during which time we infer reduced lake levels in Edward based upon sediment lithology. Taken together, this evidence suggests that from about 4000 to 1800 cal year BP, Lake Edward was a more shallow, alkaline, saline lake

ARTICLE IN PRESS J.M. Russell, T.C. Johnson / Quaternary Science Reviews 24 (2005) 1375–1389

1380

% Biogenic Silica 0 10 20 30 40 50 60 70 70

1000

% Biogenic Silica

60

Age (cal yr BP)

1500

50 40 30 20 10

r = -0.771

2000 0 10

14 18 22 % Mg in Calcite

2500

3000

3500 10 12 14 16 18 20 22 24 % Mg in Calcite Fig. 5. Left: %BSi and %Mg in calcite data resampled at a 20-year time-step using software of Paillard et al. (1996) to place the data on the same age scale. This was necessary as %BSi and %Mg data are not sampled at the same depths. Circles indicate %BSi data and triangles indicate %Mg data. Arrows at the right of this plot indicate sediments that contain opal-A0 concretions. Right: correlation between %BSi and %Mg in the resampled datasets. Although the correlation is not as visually obvious in the raw data, there are still large positive shifts in %Mg during each period of low %BSi.

than it is today. It is the timing of short-term fluctuations within this period that we will now discuss.

7. A chronology of century-scale climate events in tropical Africa The %Mg record from Lake Edward is based on stacked %Mg records from three cores, and provides evidence for a series of major droughts indicated by rising %Mg values centered at about 400, 900, 1500, 1950, 2700, 3600, 4200, and 4900 cal year BP. Additional minor %Mg peaks occur at about 1200, 2200, and 3100 cal year BP (Fig. 6; Russell et al., 2003b). These fluctuations are superimposed on a long-term positive trend in %Mg that indicates progressively more arid conditions from 5000 to about 1800 cal year BP, followed by a slight trend toward negative %Mg values indicating wetter climates. This long-term geochemical trend is coherent with lithologic data that suggests lower lake-levels in Lake Edward from about 4000 to

1900 cal year BP. Furthermore, a 7% shift in %Mg between about 2050 and 1850 cal year BP supports previous estimates of the timing of a major lake lowstand at about 2000 cal year BP based upon piston core lithology from Lake Edward (Russell et al., 2003a). Low %BSi values during all of the major and minor %Mg fluctuations within the 900–3500 cal year BP interval supports our interpretation of %Mg increases as periods of drought (Fig. 6). These include very low %BSi values and abundant opal-A0 concretions at about 1900 cal year BP (Fig. 2). In addition, the %BSi record provides evidence for numerous decade to century-scale droughts that are not clearly resolved in the lowerresolution %Mg data, particularly after 1500 cal year BP (Fig. 2 and 6). Rainfall variability at the interannual level is strongly coherent throughout tropical East Africa, because rainfall anomalies in the region are controlled by large-scale variations in the tropical oceans and atmosphere (Nicholson, 1996). Therefore, it seems likely that the decade- to century scale variability in moisture

ARTICLE IN PRESS J.M. Russell, T.C. Johnson / Quaternary Science Reviews 24 (2005) 1375–1389 5M Mg (mol %) 0

0

10

20

Stacked Mg (mol %)

30

Detrended Mg (mol %)

1381

% Biogenic Silica

Age (cal yr BP)

1000 2000 3000 4000 5000 0

10

20

1P % Mg (mol %)

0

10

20

0

10

20

-3

1

5 10 30 50 70

6M Mg (mol %)

Fig. 6. Left: Raw %Mg data from cores E96-1P, E96-5, and E96-6 M. Open triangles at right of each cores’ %Mg profile indicate the position of AMS dates used in core age-models. These three cores %Mg records were stacked and detrended as described in Russell et al. (2003b). Briefly, the %Mg profile of each core was resampled at a 20-year step using Analyseries (Paillard et al., 1996). A mean, unweighted average %Mg value was then calculated at each 20-year interval from each core that contained sediments spanning that interval. No adjustments or tuning of core chronology were made. Data from core E96-6 M spanning the interval from 2900 to 1400 cal year BP was excluded from this stack, as this interval in this core is poorly dated and lithologic data indicate that during this time the Kazinga Channel debouched into Lake Edward very close to the core site, altering the water chemistry at the site (Russell et al., 2003a, b). The detrended %Mg record (right) correlates inversely with the %BSi curve from E96-1P (right), and provides evidence for numerous droughts over the past 5400 years. Closed triangles at right indicate intervals with high opal-A0 nodule abundance.

7.1. African drought: 0– 1000 cal year BP Based on rising values of %Mg in calcite, Russell et al. (2003b) posited generally dry conditions between about 300 and 500 cal year BP (Fig. 7). This may correlate with a period of falling lake levels from 250–450 cal year BP inferred from ostracode assemblages at Lake Tanganyika (Fig. 7; Alin and Cohen, 2003). However, these records contradict evidence for very wet conditions at Lake Naivasha during this time (Verschuren et al., 2000; Verschuren, 2004). Unlike the rest of our record, sediments from Lake Edward

Edward % Mg

-4 -2 2

60

4

40

0

20 10

0.4

0

0 High -0.4

Tanganyika Level

Naivasha Lake Level

20

30

Edward % BSi

0

40

Turkana Lake Level

balance observed at Lake Edward reflects climate variations experienced by much of tropical East Africa. There are no other records of drought from tropical Africa with comparable resolution that span this entire interval. Lake level records from Lake Tanganyika (Alin and Cohen, 2003; Cohen et al. 1997) and Naivasha (Verschuren et al., 2000) cover the past ca. 2000 years, and geochemical data from Lake Turkana (Ricketts and Johnson, 1996; Halfman et al., 1994; Johnson et al., 1991) cover the past 5500 years. In addition, an annually resolved record of diatom burial from Lake Tanganyika spans the interval from 3800 to 1100 cal year BP (Cohen et al., submitted). Together, these sites form an E–W and N–S transect that covers virtually all of the East African plateau, and, with other records from Africa and the adjacent oceans, shed light on the geographic extent and causes of events recorded at Lake Edward.

Low 0

500

1000

1500

2000

Cal yr BP Fig. 7. Lake Edward detrended %Mg and %BSi data, Lake Naivasha water level (Verschuren et al., 2000), Lake Tanganyika water level (Alin and Cohen, 2003), Lake Turkana water level inferred from stacked records of piston core carbonate content (Halfman et al., 1994). Note strong evidence for drought throughout East Africa at about 900 and 2000 cal year BP, in addition to considerable variability at about 1500 cal year BP.

spanning the interval from 0 to 700 cal year BP have not been analyzed at high resolution and are not welldated (Russell et al., 2003b), so we cannot offer new insight into the timing of these recent East African climate events.

ARTICLE IN PRESS J.M. Russell, T.C. Johnson / Quaternary Science Reviews 24 (2005) 1375–1389

A large positive %Mg event dated by four AMS14C dates in Lake Edward indicates a prolonged drought between 760 and 960 cal year BP (about 1000–1200 AD). This drought corresponds almost precisely in timing with a period of low lake levels noted at Lake Naivasha, Kenya between about 750 and 990 cal year BP (Fig. 7; Verschuren et al., 2000), termed Naivasha Drought 5 (ND5, Verschuren, 2004). Evidence for this drought has also been noted at Lake Tanganyika (Alin and Cohen, 2003), Lake Turkana (Halfman et al., 1994), and possibly Lake Victoria (Stager et al., 2003). Verschuren (2004) posits that this drought affected nearly the entire African continent; however, drought in the Sahel appears to have preceded ND5 by 100–200 years (Street-Perrott et al., 2000), suggesting at least some spatial climate variability within Africa during this time. If the Sahel chronology is correct, drought in the Sahel appears correlated in timing with arid events associated with the collapse of Mayan civilizations in Central America (Hodell et al., 2001) as opposed to ND5 in equatorial East Africa. Virtually all of these authors have discussed correlation between hydrologic events in Africa at about 900 cal year BP and generally warm conditions in Europe, the so-called ‘Medieval Warm Period.’ Neither global nor northern hemispheric temperature reconstructions of Mann et al. (2003) show elevated temperatures that can be precisely correlated in timing with drought at Lake Edward. Although above average temperatures during the ‘High Medieval’ (750–850 cal year BP) fall within the interval of drought observed in tropical East Africa (Bradley et al., 2003), drought in tropical Africa appears to have begun about 1 century before European warming. We conclude that drought affected virtually all of tropical East Africa between about 750 and 950 cal year BP, and that this drought does not appear synchronous with warming in Europe or drought in South America. Based on the similarity of the well-dated Edward and Naivasha records, we concur with Verschuren (2004) that this event should receive its own designation: ND5.

occurs during the period of low %BSi nodules at about 1900 cal year BP (Fig. 2). These trends are compatible with a major drought, corroborating the suggestion of Russell et al. (2003a) that Lake Edward reached its lowest water level of the mid- to late Holocene, about 15 m below present lake level, at about 2000 cal year BP. There is considerable evidence for drought throughout tropical East Africa at 2000 cal year BP. Lake Tanganyika, to the south of Lake Edward, experienced its lowest lake level of the past 2500 years at 2000 cal year BP (Fig. 7; Alin and Cohen, 2003). The Crescent Island Crater of Lake Naivasha was completely desiccated until about 1840 cal year BP, almost precisely in phase with our interpolated age for the end of the arid event in Lake Edward of 1850 cal year BP (Fig. 7). Aridity has also been noted at this time at Lake Turkana to the north (Figs. 7 and 8) (Halfman et al., 1994; Ricketts and Johnson, 1996). These three records, each located 500–1000 km from Lake Edward, show strong evidence for a period of severe drought centered at about 2000 cal year BP. We suggest that a major drought at about 2000 cal year BP affected nearly all of the East African plateau, and is one of, in not the, most severe droughts that affected the region during the midto late Holocene. Following this drought, %Mg, %BSi, and lithologic data suggest a wet period and rising lake levels in Lake Edward, until a second drought commenced at 1550 cal year BP. The geochemical record of this event differs from other events in that following an abrupt shift at 1550 cal year BP, both %Mg and %BSi data exhibit sustained high and low values, respectively. Both records, but particularly %BSi, indicate a series of century-scale droughts within this ‘geochemical plateau,’ and century-scale variations in %BSi appear to continue until the top of the E96-1P core at 925 cal year BP, including a rapid decline correlating to the initiation of ND5 at the core top. -0.3 -0.1

7.2. African drought: 1000– 2000 cal year BP The %Mg record indicates two pronounced events between 1000 and 2000 cal year BP in Lake Edward. First, %Mg sharply rises between 1550 and 1480 cal year BP, and then oscillates at high values until beginning to fall gradually at 1150 cal year BP (Fig. 7). This trend is confirmed by the %BSi record, which falls abruptly between 1540 and 1460 cal year BP, then oscillates at a roughly century-scale period until a final shift toward low %BSi occurs at the beginning of ND5 at 950 cal year BP. Secondly, %Mg rises sharply between 2050 and 1850 cal year BP, during which time %BSi falls. A substantial peak in opal nodule abundance

Edward % Mg

0.1

Turkana δ18O

1382

0.3

-4 -2 0 2 4 0

1000

2000 3000 Age (cal yr BP)

4000

5000

Fig. 8. Detrended %Mg in Lake Edward and detrended d18O in calcite from Lake Turkana (Ricketts and Johnson, 1996) smoothed with a 5-point running average. Both records indicate periods of drought, highlighted by gray shading.

ARTICLE IN PRESS J.M. Russell, T.C. Johnson / Quaternary Science Reviews 24 (2005) 1375–1389

Correlations to other paleoclimate records between 1500 and 1000 cal year BP appear tenuous. Stable isotopic data indicate negative water balance at Lake Turkana (Ricketts and Johnson, 1996) at about 1500 cal year BP, while carbonate flux data indicate higher lake levels (Halfman et al., 1994). Ostracode data indicate that lake levels at Lake Tanganyika were generally rising from about 1800 to 1000 cal year BP. Minor shifts toward lower levels occurred at about 1100 and 1550 cal year BP, roughly the same timing as two of the events in Lake Edward (Fig. 7; Alin and Cohen, 2003). However, stromatolite data at Lake Tanganyika indicate a highstand centered at about 1500 cal year BP, followed by a period of generally low lake levels until about 700 cal year BP (Cohen et al., 1997). Lake Tanganyika laminae thickness data indicates a pronounced wet interval from 1750 to 1450 cal year BP, followed by a pronounced drought. Verschuren (2004) suggests that following a period during which Lake Naivasha filled after 1840 cal year BP, lower and fluctuating lake levels occurred until ND5 at 950 cal year BP, confirming the general pattern observed at Lake Edward and Tanganyika. The lack of coherence between these records during this period could be due to insufficient sampling resolution or climate proxy sensitivity, particularly in light of the considerable short-term climate variability indicated by the Lake Edward %BSi data during the time period. Nevertheless, these records generally confirm a pronounced wet period beginning at about 1850 cal year BP, during which time lakes in tropical East Africa rebounded from extremely low levels experienced at 2000 cal year BP. This wet phase lasted until a period of more arid conditions commenced at about 1500 cal year BP. Aridity after 1500 cal year BP was not as severe as at 2000 cal year BP, and numerous, multidecadal periods of aridity occurred during this latter interval. The Lake Edward %BSi data indicates that droughts occurred at about1360, 1230, and 1070 cal year BP, with events of shorter duration at 1140 and 1310, and 980 cal year BP (Fig. 5). These events were followed by pronounced aridity during ND5, from 950–750 cal year BP. 7.3. African drought: 2000– 5500 cal year BP High-resolution, well-dated records of drought from tropical Africa spanning the entire mid- to late Holocene are few. Of the available records, only the stable isotopic record from Lake Turkana, Kenya, has a similar length and resolution to the Lake Edward record (Johnson et al., 1991; Ricketts and Johnson, 1996). Changes in d18O should not perfectly track changes in %Mg in calcite due to inherently non-linear responses of both climate proxies to climatic perturbations (Ricketts and Johnson, 1996). Nevertheless, rising values of both

1383

%Mg and d18O should reflect periods of negative water balance in the two lakes. Within the limits of radiocarbon age dating and variable proxy response for these records, nearly all of the major droughts recorded in Lake Edward %Mg have events of comparable timing and duration in Lake Turkana (Fig. 8). Both records show evidence for droughts, indicated by rising %Mg or d18O values, centered on about 1500, 2100, 2800, 3600, 4100, 4400, and 4900 cal year BP. The Lake Tanganyika laminae thickness record generally confirms this synchrony, displaying evidence for century-scale arid conditions centered at about 3550, 2750, and 2050 cal year BP (Cohen et al., submitted), as well as events at 2100 and 2300 cal year BP that correlate with small shifts in %Mg and %BSi in Lake Edward. Among these events, numerous paleoclimatic records from East Africa show evidence for aridity at 4000 cal year BP, including lake level records (reviewed by Gasse, 2000), paleoceanographic data (deMenocal et al., 2000) and ice core data from Mt. Kilimanjaro (Thompson et al., 2002). However, the event recorded at Lake Edward at 4000 cal year BP does not appear as intense as drought at this time in North Africa and the Arabian peninsula (Gasse and Van Campo, 1994; Gasse, 2000), suggesting some regional variability in this event. In summary, the correlation between drought events recorded at Lakes Turkana, Tanganyika, Naivasha and Edward suggests that many, if not all, of the major drought events recorded at Lake Edward affected most of East Africa. These include major drought events centered at about 850, 2000, and about 4000 cal year BP that can be clearly correlated throughout tropical Africa. Additional, long high-resolution, well-dated records are needed to test these hypotheses and to test for correlation at shorter time-scales between 5500 and 2000 cal year BP.

8. Paleoclimatic implications That the Lake Edward drought record appears representative of rainfall variability over much of tropical East Africa suggests that the causes of centuryto millennial scale variability in the region operate at large spatial scales. This is supported by clear links between the Lake Edward record and proxies for Indian Ocean Monsoon variability. Major drought events in the Lake Edward record at 4000, 2000 and 900 cal year BP have all been suggested as periods of reduced Indian monsoon intensity (Gasse and Van Campo, 1994; Von Rad et al., 1999; Lu¨ckge et al., 2001). Furthermore, evidence for a major intensification of the Indian winter monsoon after 2000 cal year BP has recently been suggested (Lu¨ckge et al., 2001), and could have caused a generally positive water balance in East

ARTICLE IN PRESS J.M. Russell, T.C. Johnson / Quaternary Science Reviews 24 (2005) 1375–1389

African lakes following the arid period at about 2000 cal year BP. A tight linkage between past variations in African rainfall and the Indian Ocean Monsoon is supported by a strong 725-year cycle in both Lake Edward and Indian Ocean records (Russell et al., 2003b). What, then, is the cause of this variability? 8.1. Causes of centennial- to decadal scale climate variations in tropical East Africa

100

Raw Power

1384

125

10

72 33 29

1

26

19

16

0.1

0.01 0

The causes of century- to millennial-scale climate variability in tropical Africa are poorly understood. Numerous authors have postulated direct links between Holocene arid events in Africa and cold events at the northern high latitudes, particularly aridity at 4100 and 8200 cal year BP (Lamb et al., 1995; DeMenocal et al., 2000; Gasse, 2000). Other authors have suggested direct links between decade- to century-scale variability in solar irradiance and East African rainfall (e.g. Verschuren et al., 2000). Others have suggested that climate variability in tropical Africa is driven by dynamical coupling in the tropical ocean-atmosphere system (Gasse and Van Campo, 1994; Russell et al., 2003b). Variability in many of these components of the Earth’s climate system operate at unique periods, so we performed a suite of power spectral analyses to further elucidate the relationship between these climate forcings and African drought. The highly significant 725-year cycle in %Mg in calcite from Lake Edward (Russell et al., 2003b) and from the Arabian and South China Seas (von Rad et al., 1999; Wang et al., 1999) indicates a link between the Indian monsoon and African rainfall at millennial time scales. The causes of this cycle are unknown. Wang et al. (1999) and von Rad et al. (1999) suggest that this cycle is a harmonic of the high latitude 1500-year cycle (e.g. Bond et al., 1997), implying that high-latitude variability in ocean circulation drives both cycles. Russell et al. (2003b) instead suggest that the 725-year cycle may originate from internal variability within the tropical monsoons. Whether the 1500 year cycle is caused by changes in thermohaline convection at the high latitudes or variability in tropical SSTs is currently a matter of intense debate (Broecker, 2003). If late Holocene variability at a 725 year period in the tropical monsoons is connected to the high-latitude 1500-year cycle, the absence of strong ice-sheet feedbacks and pronounced thermohaline reorganizations during the late Holocene could lend credence to the suggestion of a tropical driver for the 1500 year cycle. At present, we offer no additional insight into this cycle, but will focus on shorter-term climate variability documented by the Lake Edward data. In addition to the725-year cycle, multi-taper method spectral analysis (Mann and Lees, 1996) of E96-1P %BSi reveals drought periods significant at the 99%

AR1 spectrum 99% Conf. Int. Raw spectrum

0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 Frequency (1/yr)

0.1

Fig. 9. Multi-taper method (MTM) power spectral output from %BSi in core E96-1P. Due to variability in sedimentation rate in E96-1P, the sampling interval was not constant in time. The average sampling interval was 3.6 years (Nyquist frequency of 7.2 years). Results are plotted to a frequency of 0.1 year, as no significant frequencies higher than 0.1 were observed. To allow MTM analysis on a variably sampled dataset, %BSi data were resampled at a 2-year step with linear detrending using Analyseries (Paillard et al., 1996). MTM spectral analyses were performed with 25% lag and confidence intervals were calculated based upon median smoothing of the best-fit spectrum (Mann and Lees, 1996). The best-fit spectrum and 99% confidence interval are shown in the periodogram; harmonic periodicities significant above the 99% level are labeled. To further analyze highfrequency variation in our data, we removed all periods 4125 years from the %BSi dataset by subtracting a Gaussian filtered series set at appropriate frequencies from the %BSi data. The raw power of the MTM analysis of this filtered dataset was virtually identical to the unfiltered %BSi data, with a correlation coefficient of 0.97 between the two power spectra.

confidence level of 125, 72, 33, 29, 26, 19, and 16 years (Fig. 9). 125-year periods have been observed in speleothem geochemical data from South Africa (LeeThorp et al., 2000), and periods of 76, 33, 29, and 18.5 years have been observed in Nile River flood gauge records (De Putter et al., 1998). The Lake Tanganyika laminae thickness record indicates robust periods of 67–75 and 25–31 years (Cohen et al., submitted). Time series spectral analysis of % Carbonate data from Lake Turkana indicates significant periods at 76, 32, and 18.6 years (Halfman et al., 1994), and varve thickness studies from the Arabian Sea indicate significant periodicities at 125, 29–31, and 26 years, in addition to numerous others (von Rad et al., 1999). The combination of age errors of these records and bandwidth error on frequency of the spectral analysis may alter reported periods by up to about 10%, so many of the climate periods reported by these authors could be the same as those present in Lake Edward %BSi. In particular, periods of 120, 76–72, 33–29, and 19–18 years appear in many records from Africa and the Indian Ocean. Verschuren et al. (2000) report a strong correlation between solar variability over the past 1000 years and drought at Lake Naivasha. In their record, periods of low solar output correlate to periods of increased East African rainfall, and vice versa. Among these droughts

ARTICLE IN PRESS J.M. Russell, T.C. Johnson / Quaternary Science Reviews 24 (2005) 1375–1389 Cal yr BP

-2 0 2

0.1

4

0 -0.1 -0.2 -0.3 -0.4

ND5

Insolation (W/m2) relative to 1950

Edward % Mg Anomaly

-4

-0.5 0

1000

2000

3000

4000

5000

Fig. 10. A comparison of solar irradiance forcing (Weber et al., 2004; Crowley, 2000) and Lake Edward %Mg for the past 5500 years. The solar irradiance data was generated through splicing of the decadal atmospheric D14C record of Stuiver et al. (1998) to modern irradiance data (Crowley, 2000). Note the general lack of correlation between the two records, and that the phasing between drought in Lake Edward and solar output is not constant through time as highlighted by the shaded events.

is ND5, from 950 to 750 cal year BP, also noted at Lake Edward. Low amplitude periods of 26 and 126 years are present in atmospheric D14C, a proxy for past solar output (Stuiver et al., 1998; Stuiver and Braziunas, 1993) and are present in the Lake Edward data, but whether these periods reflect variations in solar output or oceanic effects on D14C is unclear (Stuiver and Braziunas, 1993). Regardless, no proxy data from Lake Edward exhibits significant periodicity at periods that dominate the solar spectrum, such as 11, 22, 88, and 206 years. Neither tests of correlation, nor cross-spectral analysis, reveals any coherence at any period between atmospheric D14C and proxy record for drought from Lake Edward. Moreover, visual inspection of the correlation between global radiative forcing based on atmospheric D14C (Crowley, 2000, Weber et al., 2004) and the Lake Edward %Mg drought record indicate little, if any, significant correlation exists. While ND5 lies on a period of high solar output, major droughts at about 2900 and 4900 in Lake Edward lie on periods of reduced solar output, and other events have no solar correlate at all (Fig. 10). We interpret this to indicate that, although solar forcing may influence African climate, solar forcing is not the dominant pace-maker of decade- to century-scale climate variability in equatorial East Africa during the mid- to late Holocene. Many of the multi-decadal periods observed in Lake Edward occur in Indian Ocean records and have been attributed to the influence of lunar and planetary effects on tidal cycles and sedimentation (Currie and Fairbridge, 1985; Loutre et al., 1992; Berger and von Rad, 2002). However, there is no known mechanism linking astronomical alignment nor lunar position to African rainfall, nor a suggestion of such a linkage in modern

1385

rainfall data from Africa. Rather, we attribute the climate variability observed in these regions to coupled ocean-atmosphere dynamics within the tropical monsoons. Folland et al. (1998) applied empirical orthogonal function (EOF) analysis to 140-year long global SST and rainfall datasets. Although the length of their weather records limits the resolution of periodicity in their analysis, Folland et al. (1998) reveal several patterns in global sea-surface temperatures that operate at interdecadal to centennial time-scales and are caused by coupled ocean-atmosphere variability. Folland et al.’s second EOF corresponds to interhemispheric SST contrasts that are strongly linked to the African monsoon. They suggest that this mode operates on a 60–70 year timescale, in agreement with previous observations of 50–80 year oscillatory behavior in global temperature datasets (Schlesinger and Ramankutty, 1994). The third EOF of Folland et al. (1998) captures ENSO-like SST structures in the Indian and Pacific Oceans and is similar to the ‘Pacific Decadal Oscillation’ of Mantua et al. (1997). This mode fluctuates at an approximately 30-year period, and bears significant correlation with rainfall variability in the western Indian Ocean and tropical East Africa (Cole et al., 2000). Although the periodically varying tropical SST modes described by Folland et al. (1998) cannot be matched with any certainty to specific events in the Edward record, Folland et al.’s (1998) analysis demonstrates that tropical SSTs can vary at 20–70 year time-scales due to internal coupled ocean-atmosphere feedbacks. The lack of strong evidence for solar forcing in the Lake Edward record, and the significant influence that the modes of Folland et al. (1998) exert on the tropical monsoons suggests that variability internal to the monsoon system may account for much of the variation observed at 20–100 year periods in the Lake Edward record. 8.2. Patterns and causes of long-term climate change at Lake Edward Virtually all prior studies of long-term climate change in tropical Africa have noted an abrupt transition from wet to dry conditions at about 5500 cal year BP at sites north of about 101S (DeMenocal et al., 2001; Gasse, 2000). This shift is a response to slow, precessionally driven changes in seasonal insolation (Kutzbach and Street-Perrott, 1985). The abruptness of this shift in northern Africa is primarily due to positive feedbacks between rainfall, vegetation, and sea-surface temperatures in the Sahara and Sahel (Ganopolski et al., 1998; Claussen et al., 1999). Russell et al. (2003a) noted an abrupt shift from dark gray clays to reddish carbonate muds at about 5200 cal year BP that likely reflect a stage in this transition toward aridity. However, they did not discuss the causes of the ensuing long-term shift in %Mg in calcite.

ARTICLE IN PRESS 1386

J.M. Russell, T.C. Johnson / Quaternary Science Reviews 24 (2005) 1375–1389

Paleorecords from the subtropical limit of the seasonal migration of the ITCZ have noted gradual reductions in rainfall beginning by about 6000 cal year BP (Haug et al., 2001; Fleitmann et al., 2003). These changes were attributed to decreased northward penetration of the ITCZ into the subtropics as well as declining monsoon ‘intensity’ in response to the changes in seasonal insolation driven by the precessional orbital cycle. However, these studies have difficulty discriminating between changes in ITCZ position and monsoon intensity due to their location at the limits of the ITCZ’s seasonal migration. Lake Edward, situated directly on the equator, is ideally situated to record changes in monsoon intensity as opposed to position. We posit that the long-term increase in %Mg between 5500 and 2000 cal year BP reflects a gradual reduction in monsoon intensity. We ascribe this trend to orbitally induced changes in land-sea contrast as well as gradual reductions in tropical SSTs (Sonzogni et al., 1998) over the course of the mid-Holocene (Claussen et al., 1999). If so, the abruptness of the the mid-Holocene shift noted in the Sahara and Sahel is not typical of all of tropical Africa. After about 2000 cal year BP, %Mg begins to decrease, indicating a shift toward generally wetter conditions. This inference is supported by lithologic data: shallow water cores from Lake Edward contain no obvious lowstand deposits that date younger than 1800 cal year BP (Russell et al., 2003a). A similar pattern has been noted at several other sites within the African tropics (Ricketts and Johnson, 1996; Wirrmann et al., 2001). Recent studies of the Indian monsoon suggest an abrupt shift toward increased winter rainfall and winter monsoon intensity in the Arabian Sea at 2000 cal year BP (Lu¨ckge et al., 2001), as well as generally rising sea-surface temperatures in the Arabian Sea since about 2300 cal year BP (Doose-Rolinski et al., 2001). None of these authors discuss the mechanism driving this event, but suggest that it stemmed from an intensification of the winter monsoon system. At present, the cause of this change is not known. It is possible that interactions between the ENSO and insolation caused the observed changes. ENSO events are positively correlated to rainfall in East Africa, and both model and lake-based data suggest that ENSO events reached their peak amplitude between 1000 and 2000 cal year BP (Moy et al., 2002). However, Moy et al. (2002) also suggest declining ENSO intensity since about 1000 cal year BP, and a shift toward more arid conditions in East Africa during this time has not been observed. Several authors have ascribed changes in monsoon activity during the past 3000 years to slight shifts in seasonal insolation (e.g. Abbott et al., 1997). However, records of the South American monsoon indicate continued declines in average rainfall from about 6000 cal year BP to present without any change

toward wetter conditions at 2000 cal year BP (Haug et al., 2001). This would suggest that if the increase in rainfall experienced by Lake Edward after 2000 cal year BP were driven by insolation, the effects were felt only by the Indian and not by the Atlantic Ocean monsoon systems. This would seem at odds with the global-scale effects of changing insolation (Short et al., 1991). Although we cannot explain the observed change at 2000 cal year BP in Lake Edward, we suggest that it arises from non-linear interactions between the tropical Indian Ocean and orbitally forced changes in insolation.

9. Conclusions Rainfall in tropical Africa over the past 5000 years has varied considerably at all time-scales. At Lake Edward, these paleoclimatic variations are recorded by complex geochemical signals that include variations in carbonate and silica geochemistry and biogenic silicate diagenesis. Although it cannot be unequivocally demonstrated that decade- to century scale rainfall variations are synchronous through tropical East Africa, several well-dated records from the East Africa plateau show evidence for coherent climate variations in tropical Africa. Among the events recorded at Lake Edward are a well-known drought event at 4000 cal year BP, as well as major periods of drought that occurred at about 900 and 2000 cal year BP. Aridity at 2000 cal year BP resulted in a major lowstand in Lake Edward, and is likely one of the most arid intervals during the entire Holocene in tropical East Africa. The causes of rainfall variations in tropical East Africa remain obscure. There is no clear evidence from Lake Edward linking rainfall variations in equatorial East Africa directly to solar variability. Rather, the large regional extent of African drought and links between both individual paleoclimatic events and drought periodicity suggests that East African climate is linked strongly to variability in the tropical Indian Ocean. Decade- to century-scale variations in Lake Edward’s climate may be driven by coupled ocean–atmosphere variation that affect tropical SSTs, monsoonal circulation, and continental rainfall. Finally, we would note that today’s %Mg values in calcite are about 15%, compared to an average value of about 20% for the last 2000 years. Although these values cannot be interpreted quantitatively, it appears that the past several hundred years have seen generally wetter climates, free from the severe, centuries-long droughts that plagued tropical Africa during much of the late Holocene. We cannot at the present time predict when such droughts may occur in the future, but given the dependence of local economies on freshwater resources in the region, our data highlight the need for the development of sustainable water resource

ARTICLE IN PRESS J.M. Russell, T.C. Johnson / Quaternary Science Reviews 24 (2005) 1375–1389

management practices in light of the severe droughts that have occurred in the past.

Acknowledgements Yvonne Chan is thanked for laboratory assistance with %BSi analyses of E96-1P. Amy Myrbo and Doug Schnurrenberger are thanked for assistance with SEM images of opal nodules, and the Center for Interfacial Engineering, U of MN, is acknowledged for assistance with SEM and XRD work on Lake Edward sediments. We thank Peter DeMenocal for insight and thoughtful review of an earlier version of this manuscript. This research was supported by NSF-ATM # 0314832 and # 9805293. This paper is IDEAL contribution # 155.

References Abbott, M.B., Seltzer, G., Kelts, K.R., Southon, J.R., 1997. Holocene paleohydrology of the tropical Andes from Lake Records. Quaternary Research 47, 70–80. Alin, S.R., Cohen, A.S., 2003. Lake-level history of Lake Tanganyika, East Africa, for the past 2500 years based on ostracode-inferred water-depth reconstruction. Palaeogeography, Palaeoclimatology, Palaeoecology 199, 31–49. Beuning, K.R.M., Russell, J.M., 2004. Vegetation and sedimentation in the Lake Edward basin, Uganda-Congo during the late Pleistocene and early Holocene. Journal of Paleolimnology 32, 1–18. Birkett, C., Murtugudde, R., Allan, T., 1999. Indian Ocean climate event brings floods to East Africa’s lakes and the Sudd marsh. Geophysical Research Letters 26, 1031–1034. Bond, G.C., Showers, W., Cheseby, M., Lotti, R., Almasi, P., DeMenocal, P., Priore, P., Cullen, H., Hadjas, I., Bonani, G., 1997. A pervasive millennial-scale cycle in North Atlantic Holocene and glacial climates. Science 278, 1257–1266. Bradley, R.S., Hughes, M.K., Diaz, H.F., 2003. Climate in Medieval Time. Science 302, 404–405. Broecker, W.S., 2003. Does the trigger for abrupt climate change reside in the ocean or the atmosphere? Science 300, 1519–1522. Claussen, M., Kubatzki, C., Brovkin, V., Ganopolski, A., 1999. Simulation or an abrupt change in Saharan vegetation in the midHolocene. Geophysical Research Letters 26, 2037–2040. Cohen, A.S., Lezzar, K.E., Cole, J., Dettman, D., Ellis, G.S., Eagle, M., Chororoka, K., Plisnier, P.D., Langenberg, V., Michelo, V., Zilifi, D. Late Holocene linkages between decade-century scale climate variability and productivity at Lake Tanganyika, Africa. In review for The Holocene. Cohen, A.S., Talbot, M.R., Awramik, S.M., Dettman, D.L., Abell, P., 1997. Lake level and paleoenvironmental history of Lake Tanganyika, Africa, as inferred from late Holocene and modern stromatolites. GSA Bulletin 109, 444–460. Cole, J.E., Dunbar, R.B., McClanahan, T.R., Muthiga, N., 2000. Tropical Pacific forcing of decadal variability in the western Indian Ocean over the past two centuries. Science 287, 617–619. Crowley, T.J., 2000. Causes of climate change over the past 1000 years. Science 289, 270–277. Currie, R.G., Fairbridge, R.W., 1985. Periodic 18.6-year and cyclic 11year induced drought and flood in northwestern China and some global implications. Quaternary Science Reviews 4, 109–134.

1387

Curtis, J., Hodell, D., 1993. An isotopic and trace element study of ostracods from Lake Miragoane, Haiti: a 10,500 year record of paleosalinity and paleotemperature changes in the Caribbean. In: Swart, P.K., Lohmann, K.C., McKenzie, J., Savin, S. (Eds.), Climate Change in Continental Isotopic Records. Geophysical Monograph 78. American Geophysical Union, pp. 135–152. De Putter, T., Loutre, M.F., Wansard, G., 1998. Decadal periodicities on Nile River historical discharge (A. D. 622–1470) and climatic implications. Geophysical Research Letters 25, 3193–3196. DeMaster, D.J., 1979. The marine budgets of silica and Si32. Ph.D. Thesis, Yale University, New Haven. DeMenocal, P., Ortiz, J., Guilderson, T., Sarnthein, M., 2000. Coherent high- and low-latitude climate variability during the Holocene warm period. Science 288, 2198–2202. Fish, G.R., 1953. Appendices A1-A6. Annual Report for 1953. East African Fisheries Research Organization. East African High Commission, Jinja, Uganda. Fleitmann, D., Burns, S.J., Mudelsee, M., Neff, U., Kramers, J., Mangini, A., Matter, A., 2003. Holocene forcing of the Indian Monsoon recorded in a stalagmite from Southern Oman. Science 300, 1737–1739. Folland, C.K., Parker, D.E., Colman, A.W., Washington, R., 1998. Large scale modes of ocean surface temperature variation since the late nineteenth century. In: Navarra, A. (Ed.), Beyond El Nin˜o: Decadal and Interdecadal Climate Variability. Springer, Berlin, pp. 73–102. Ganopolski, A., Kubatzki, C., Claussen, M., Brovkin, V., Petouhkov, B., 1998. The influence of vegetation-atmosphere-ocean interaction of climate during the mid-Holocene. Science 280, 1916–1919. Gasse, F., 2000. Hydrological changes in the African tropics since the Last Glacial Maximum. Quaternary Science Reviews 19, 189–211. Gasse, F., Van Campo, E., 1994. Abrupt post-glacial climate events in West Asia and North Africa monsoon domains. Earth and Planetary Science Letters 126, 435–456. Halfman, J.D., Johnson, T.C., Finney, B., 1994. New AMS dates, stratigraphic correlations and decadal climate cycles for the past 4 ka at Lake Turkana, Kenya. Palaeogeography, Palaeoclimatology, Palaeoecology 111, 83–98. Haug, G.H., Hughen, K.A., Sigman, D.M., Peterson, L.C., Ro¨hl, U., 2001. Southward migration of the intertropical convergence zone through the Holocene. Science 293, 1304–1307. Hecky, R.E., Degens, E.T., 1972. Late Pleistoncene-Holocene chemical stratigraphy and paleolimnology of the Rift Valley Lakes of central Africa, Wood’s Hole Oceanographic Institution Technical Reports. Wood’s Hole Oceanographic Institute, Wood’s Hole, p. 93. Hecky, R.E., Kling, H.J., 1987. Phytoplankton ecology in the great lakes of the rift valleys of central Africa. Archiv fur Hydrobiologie Ergebnisse Limnologie 25, 3–34. Hein, J.R., Scholl, D.W., Barron, J.A., Jones, M.F., Miller, J., 1978. Diagenesis of late Cenozoic diatomaceous deposits and formation of the bottom simulating reflector in the southern Bering Sea. Sedimentology 25, 155–181. Hesse, R., 1988. Diagenesis #13. Origin of chert: diagenesis of biogenic siliceous sediments. Geoscience Canada 15, 171–192. Hodell, D., Brenner, M., Curtis, J., Guilderson, T., 2001. Solar forcing of drought in the Maya lowlands. Science 292, 1367–1370. Hu, F.S., Kaufman, D., Yoneji, S., Nelson, D., Shemesh, A., Huang, Y., Tian, J., Bond, G., Clegg, B., Brown, T., 2003. Cyclic Variation and Solar Forcing of Holocene Climate in the Alaskan Subarctic. Science 301, 1890–1893. Johnson, T.C., 2002. Biogenic silica profiles in the sediments of large tropical lakes: examples from East Africa. In: Renault, R.W., Ashley, G.M. (Eds.), Sedimentation in Continental Rifts. SEPM Special Publication Number 73, pp. 247–255. Johnson, T.C., Halfman, J.D., Showers, W., 1991. Paleoclimate of the past 4000 years at Lake Turkana, Kenya, based on the isotopic

ARTICLE IN PRESS 1388

J.M. Russell, T.C. Johnson / Quaternary Science Reviews 24 (2005) 1375–1389

composition of authigenic calcite. Palaeogeography, Palaeoclimatology, Palaeoecology 85, 189–198. Jones, B.F., Rettig, S.L., Eugster, H.P., 1967. Silica in alkaline brines. Science 158, 1310–1314. Jones, J.B., Segnit, E.R., 1971. The nature of opal I: nomenclature and constituent phases. Journal of the Geological Society of Australia 18, 57–68. Kutzbach, J.E., Street-Perrott, F.A., 1985. Milankovitch forcing of fluctuations in the level of tropical lakes from 18 to 0 kyr BP. Nature 317, 134–137. Lærdal, T., Talbot, M.R., Russell, J.M., 2002. Late Quaternary sedimentation and climate in the Lakes Edward and George area, Uganda-Congo. In: Odada, E.O., Olago, D. (Eds.), The East African Great Lakes: Limnology, Palaeolimnology, Biodiversity. Kluwer Academic Publishers, Dordrecht, The Netherlands, pp. 429–470. Lamb, H.F., Gasse, F., Benkaddour, A., El Hamouti, N., van der Kaars, S., Perkins, W.T., Pearce, N.J., Roberts, C.N., 1995. Relation between century-scale Holocene arid intervals in tropical and temperate zones. Nature 373, 134–137. Lee-Thorp, J.A., Holmgren, K., Lauritzen, S.E., Linge, H., Moberg, A., Partidge, T.C., Stevenson, C., Tyson, P.D., 2000. Rapid climate shifts in the southern African interior throughout the mid to late Holocene. Geophysical Research Letters 28, 4507–4510. Lehman, J., 2002. Application of AVHRR to water balance, mixing dynamics, and water chemistry of Lake Edward, East Africa. In: Odada, E.O., Olago, D. (Eds.), The East African Great Lakes: Limnology, Palaeolimnology, Biodiversity. Kluwer Academic Publishers, Dordrecht, The Netherlands, pp. 235–260. Loutre, M.F., Berger, A., Bretagnon, P., Blanc, P.L., 1992. Astronomical frequencies for climate research at the decadal to century time scale. Climate Dynamics 7, 181–194. Lu¨ckge, A., Doose-Rolinski, H., Khan, A.A., Schulz, H., von Rad, U., 2001. Monsoonal variability in the northeastern Arabian Sea during the past 5000 years: geochemical evidence from laminated sediments. Palaeogeography, Palaeoclimatology, Palaeoecology 167, 273–286. Mann, M.E., Lees, J.M., 1996. Robust estimation of background noise and signal detection in climatic time series. Climatic Change 33, 409–445. Mann, M., Jones, P.D., 2003. Global surface temperatures over the past two millennia. Geophysical Research Letters 30, 5-1–5-4. Mantua, N.J., Hare, S.R., Zhang, Y., Wallace, J.M., Francis, R.C., 1997. A Pacific interdecadal climate oscillation with impacts on salmon production. Bulletin of the American Meteorological Society 78, 1069–1079. Mortlock, R.A., Froelich, P.N., Philip, N., 1989. A simple method for the rapid determination of biogenic opal in pelagic marine sediments. Deep Sea Research Part A: Oceanographic Research Papers 36, 1415–1426. Moy, C.M., Seltzer, G., Rodbell, D.T., Anderson, D.M., 2002. Variability of El Nin˜o/Southern Oscillation activity at millennial timescales during the Holocene epoch. Nature 404, 162–165. Murtugudde, R., McCreary, J., Busalacchi, A.J., 2000. Oceanic processes associated with anomalous events in the Indian Ocean with relevance to 1997–1998. Journal of Geophysical Research 105, 3295–3306. Nicholson, S.E., 1996. A review of climate dynamics and climate variability in Eastern Africa. In: Johnson, T.C., Odada, E.O. (Eds.), The Limnology, Climatology, and Paleoclimatology of the East African Lakes. Gordon and Breach Publishers, Amsterdamn, pp. 25–56. Nicholson, S.E., 1997. An analysis of the ENSO signal in the tropical Atlantic and western Indian Oceans. International Journal of Climatology 17, 345–375.

Oldfield, F., Alverson, K., 2003. The Societal Relevance of Paleoenvironmental Research. In: Alverson, K., Bradley, R.S., Pedersen, T.F. (Eds.), Paleoclimate, Global Change, and the Future. Springer, Berlin, pp. 1–10. Paillard, D., Labeyrie, L., Yiou, P., 1996. Macintosh program performs time-series analysis. EOS Transactions, AGU 77, 379. Ricketts, R.D., Johnson, T.C., 1996. Climate change in the Turkana basin as deduced from a 4000 year l.ong d18O record. Earth and Planetary Science Letters 142, 7–17. Russell, 2004. The Holocene paleolimnology and paleoclimatology of Lake Edward, Uganda-Congo. Ph.D. Thesis, University of Minnesota, Minneapolis, MN, USA. 217pp. Russell, J.M., Johnson, T.C., Kelts, K.R., Lærdal, T., Talbot, M.R., 2003a. An 11,000-year lithostratigraphic and paleohydrologic record from equatorial Africa: Lake Edward, UgandaCongo. Palaeogeography, Palaeoclimatology, Palaeoecology 193, 25–49. Russell, J.M., Johnson, T.C., Talbot, M.R., 2003b. A 725-year cycle in the Indian and African monsoons. Geology 31, 677–680. Schlesinger, M.E., Ramankutty, N., 1994. An oscillation in the global climate system of period 65–70 years. Nature 367, 723–726. Schnurrenberger, D.W., Kelts, K.R., Johnson, T.C., Shane, L.C.K., Ito, E., 2001. National lacustrine core repository (LacCore). Journal of Paleolimnology 25, 123–127. Sonzogni, C., Bard, E., Rostek, F., 1998. Tropical sea-surface temperatures during the last glacial period: a view based on alkenones in Indian Ocean sediments. Quaternary Science Reviews 17, 1185–1201. Stager, J.C., Johnson, T.C., 2000. A 12,400 14C yr offshore diatom record from east central Lake Victoria, East Africa. Journal of Paleolimnology 23, 373–383. Stager, J.C., Cumming, B.F., Meeker, L.D., 2003. A 10,000 year highresolution diatom record from Pilkington Bay, Lake Victoria, East Africa. Quaternary Research 59, 172–181. Street-Perrott, F.A., Holmes, J.A., Waller, M.P., Allen, M.J., Barber, N.G.H., Fothergill, P.A., Harkness, D.D., Ivanovich, M., Kroon, D., Perrott, R.A., 2000. Drought and dust deposition in the west African Sahel: a 5500-year record from Kajemarum Oasis, northeastern Nigeria. The Holocene 10, 293–303. Stuiver, M., Braziunas, T.F., 1993. Sun, ocean, climate, and atmospheric 14CO2: an evaluation of causal and spectral relationships. The Holocene 3, 289–305. Stuiver, M., Reimer, P.J., Bard, E., Beck, J.W., Burr, G.S., Hughen, K.A., Kromer, B., McCormac, F.G., van der Plicht, J., Spurk, M., 1998. INTCAL98 radiocarbon age calibration, 24,000-0 cal BP. Radiocarbon 40, 1041–1083. Talling, J.F., Talling, I.B., 1965. The chemical composition of African lake waters. Internationale Revue Gesamten Hydrobiologie 50, 421–463. Thompson, L.G., Mosely Thompson, E., Davis, M.E., Henderson, K.A., Brecher, H.H., Zagorodnov, V.S., Mashiotta, T.A., Lin, P., Mikhalenko, V.N., Hardy, D.R., Beer, J., 2002. Kilimanjaro ice core records: evidence of Holocene climate change in tropical Africa. Science 298, 589–593. Verschuren, D., 2004. Decadal and century-scale climate variability in tropical Africa during the past 2,000 years. In: Battarbee, R.W., Gasse, F., Stickley, C.E. (Eds.), Past Climate Variability Through Europe and Africa. Developments in Paleoenvironmental Research, vol. 6. Kluwer Academic Publishers, Dordrecht, The Netherlands Chapter 8. Verschuren, D., Laird, K.R., Cumming, B., 2000. Rainfall and drought in equatorial east Africa during the past 1100 years. Nature 403, 410–414. Viner, A.B., Smith, I.B., 1973. Geographical, physical, and historical aspects of Lake George. Proceedings of the Royal Society of London, Series B 184, 235–270.

ARTICLE IN PRESS J.M. Russell, T.C. Johnson / Quaternary Science Reviews 24 (2005) 1375–1389 von Rad, U., Schaaf, M., Michels, K.H., Schulz, H., Berger, W.H., Sirocko, F., 1999. A 5000-yr record of climate changes in varved sediments from the oxygen minimum zone off Pakistan, northeastern Arabian Sea. Quaternary Research 51, 39–53. Wang, L., Sarnthein, M., Erlenkeuser, H., Grimalt, J., Grootes, P., Heilig, S., Ivanova, E., Kienast, M., Pelejer, C., Pflaumann, U., 1999. East Asian monsoon climate during the late Pleistocene: high resolution sediment records from the South China Sea. Marine Geology 156, 245–284.

1389

Weaver, F.M., Wise, S.W., 1972. Ultramorphology of deep sea cristobalitic chert. Nature 237, 56–57. Weber, S.L., Crowley, T.J., van der Schrier, G., 2004. Solar irradiance forcing of centennial climate variability during the Holocene. Climate Dynamics 22, 539–553. Wirrmann, D., Bertaux, J., Kossoni, A., 2001. Late Holocene paleoclimatic changes in western central Africa inferred from mineral abundance in dated sediments from Lake Ossa (Southwest Cameroon). Quaternary Research 56, 275–287.