Potential relation between equatorial sea surface temperatures and historic water level variability for Lake Turkana, Kenya

Potential relation between equatorial sea surface temperatures and historic water level variability for Lake Turkana, Kenya

Accepted Manuscript Potential relation between equatorial sea surface temperatures and historic water level variability for Lake Turkana, Kenya Chris ...

484KB Sizes 0 Downloads 57 Views

Accepted Manuscript Potential relation between equatorial sea surface temperatures and historic water level variability for Lake Turkana, Kenya Chris Bloszies, Steven L. Forman PII: DOI: Reference:

S0022-1694(14)00781-1 http://dx.doi.org/10.1016/j.jhydrol.2014.10.001 HYDROL 19950

To appear in:

Journal of Hydrology

Received Date: Revised Date: Accepted Date:

2 December 2013 19 June 2014 1 October 2014

Please cite this article as: Bloszies, C., Forman, S.L., Potential relation between equatorial sea surface temperatures and historic water level variability for Lake Turkana, Kenya, Journal of Hydrology (2014), doi: http://dx.doi.org/ 10.1016/j.jhydrol.2014.10.001

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

1 2

Potential relation between equatorial sea surface temperatures and historic water level variability for Lake Turkana, Kenya

3 4 5

Chris Bloszies*, Dept. of Earth and Environmental Sciences, University of Illinois at Chicago, 845 W. Taylor St., Chicago, IL 60607, [email protected]

6 7

Steven L. Forman, Dept. of Geology, Baylor University, One Bear Place #97354 Waco, TX 76798-7354

8 9

*Corresponding Author

10 11 12 13

Keywords: Lake Turkana, historic lake level, monsoon variability, Indian Ocean, Atlantic Ocean, East Africa

14 15

Abstract Water level in Lake Turkana, Kenya in the past ca. 150 years is controlled primarily from the biannual passage

16

of the East and West African monsoon, with rainfall volume related partially to sea surface temperatures (SSTs) in

17

the Western Indian and East Atlantic oceans. Empirical orthogonal function analyses show significant correlation

18

between Eastern Atlantic or Western Indian SSTs and lake level anomalies, with the first mode accounting for 66%

19

and 55% of the variability. The primary geographic loadings are consistent with a Gulf of Guinea moisture source

20

and positive Indian Ocean Dipole (IOD) state. The second mode explains 10% of variability, and reflects the

21

westward extension of an Indian Ocean cool pool, potentially indicative of a normal to a negative IOD state. There

22

is significant spatial correlation between basin rainfall anomalies associated with Eastern Atlantic SSTs and a low in

23

the continental divide between the Kenyan and the Ethiopian highlands, which is a passage for moisture from the

24

Congo Basin. Linear regression analysis with Bootstrap sampling and Monte Carlo simulations define numeric

25

relations between Western Indian and Eastern Atlantic SSTs and lake level change for AD 1992 to 2013. The

26

monthly and yearly lake level reconstructions based on this numeric analysis capture the decadal-scale variability

27

and the 15 m drop in water level in the early 20th century. Meter-scale variability in lake level since ca. AD 1930 is

28

associated with precipitation sourced from the Western Indian Ocean with IOD variability, whereas the 15 m drop in

29

water level in the early 20th century may reflect a profound decrease in moisture from Atlantic/Congo Basin source.

30

These numerical solutions are poised to reconstruct water level variations in the past ca. 300 years for Lake Turkana

31

with new proxy records of SSTs from the Western Indian Ocean and the Gulf of Guinea.

32

33

1. Introduction

34

A critical system for understanding climate dynamics in the 21st century is monsoonal circulation on global and

35

meso-scales (Fig. 1). Uncertainty remains on how this complex system will respond to subdecadal oscillations (e.g.

36

the Indian Ocean Dipole) in sea surface temperatures (SSTs), anthropogenic atmospheric and oceanic warming, and

37

land-use changes in the tropics (e.g. Ackerley et al., 2011; Christensen, 2007; Giannini, 2010; Nicholson, 2009;

38

Wang et al., 2012). Global climate models predict a substantial increase in rainfall for equatorial regions with

39

elevated greenhouse gases in the 21st century, often associated with an intensified and a spatially extended monsoon

40

(Cherchi et al., 2011; Hsu et al., 2012). These models depict warmer SSTs in the equatorial oceans in response to a

41

rise in global air temperatures, which may increase the severity of the West African and Indian monsoons (Ashok

42

and Saji, 2007; Biasutti, 2013), but may have limited effect on the East African Monsoon (Shongwe et al., 2011;

43

Williams et al., 2012). The predicted distribution of precipitation over equatorial Africa is spatially diverse (Cook

44

and Vizy, 2012), with drought forecast to deepen in the western Sahel, an expected rainfall deficit for southern

45

Africa, and a concomitant increase in rainfall across Equatorial Africa (Biasutti, 2013; Hoerling et al., 2006). A

46

possible wetter 21st century monsoon is reinforced by a mesoscale climate model for East Africa which predicts a

47

19% and 22% increase in precipitation for the boreal fall and boreal winter, respectively for the Lake Turkana Basin,

48

Kenya (Shongwe et al., 2011). However, since the 1970s, annual rainfall has decreased substantially across East

49

Africa (Funk et al., 2005), with a 14% drop for Ethiopia (Viste et al., 2013), and a ~20% fall for southern Sudan

50

(Funk et al., 2011). Also, seasonal precipitation over East Africa for the boreal winter has declined ~15% since

51

1999, punctuated by severe droughts in AD 2004 to 2005, 2009 and 2010 to 2011 (Lyon and DeWitt, 2012). These

52

20th and 21st century droughts are associated with a precipitous drop in crop production and decreased food security,

53

leading to famine and political unrest in Ethiopia, northwest Kenya and southern Sudan (Funk et al., 2005).

54

A number of mechanisms have been proposed to explain 20th and 21st century drought in West and Central

55

Africa, particularly the Sahel (Giannini, 2010; Giannini et al., 2003; Nicholson, 2013). One hypothesized cause of

56

sustained drought for the 1960s to the 1980s is desertification from intensified human land use and animal

57

overgrazing (Kucharski et al., 2013; Nicholson et al., 1998). However, with increased precipitation and “re-

58

greening” in the Sahel in the late 20th century broader climate controls are inferred, with warming of the equatorial

59

Atlantic and the Indian oceans and related shifts in moisture transport and upper air flow (e.g. Folland et al., 1986;

60

Hastenrath et al., 1993). Subsequent climate modeling tuned to SST variability captures broadly the timing and

61

extent of these droughts and are associated with changes in the intensity of the West and East African monsoons

62

(Behera et al., 2005; Goddard and Graham, 1999; Hoerling et al., 2006; Riddle and Cook, 2008; Vizy and Cook,

63

2001). Higher resolution reconstructions (< 2.5° x 2.5°), with input of vetted historical climate data (e.g.

64

NCEP/NCAR; Kucharski et al., 2013), yield increased fidelity on the footprint and severity of these African

65

droughts when forced with a combination of land-use changes, interannual SST variability and basin-wide increase

66

of interdecadal SSTs; the later attributed to the global rise in air and ocean temperatures in the 20th and 21st centuries

67

(Biasutti, 2013; Giannini et al., 2008; Giannini et al., 2003). Recently, increased SSTs in the southern tropical Indian

68

Ocean associated with enhanced convection, precipitation and heightened subsidence over north Africa is invoked

69

for decreased moisture transport from Atlantic-derived sources to the Horn of Africa in the past 30 years (Du et al.,

70

2013; Lyon and DeWitt, 2012; Ummenhofer et al., 2009).

71

Interannual variability in East African rainfall in the past 30 years is hypothesized to be broadly linked to warm

72

pool temperatures in the Western Indian Ocean. A variety of statistical analyses indicate a robust relation between

73

Indian Ocean SSTs and rainfall intensity for East Africa (Black et al., 2003; Mutai et al., 2012; Omondi et al., 2012).

74

Particularly, floods during the boreal winter in 1997, 2008 and 2010 are attributed to positive Indian Ocean Dipole

75

(IOD) states (Fig. 1a; Behera et al., 2005; Birkett et al., 1999; Hastenrath et al., 2010). These extreme rainfall events

76

appear coeval with a meter-scale rise in water level for many East African lakes in 1997 and 2006 to 2007 (Becker

77

et al., 2010; Ricko et al., 2011b). Wet conditions in 1998, 2008 and 2010 are also associated with a strong zonal

78

West African Monsoon with anomalously high SSTs in the Eastern Equatorial Atlantic Ocean (Fig. 1b; Williams et

79

al., 2012). Intensification of East and West African monsoons with warming SSTs may be reflected in peak water

80

levels for Lake Turkana (Ricko et al., 2011a; Velpuri et al., 2012), which include a rise of >4 m in 1994, ~2 m in

81

1997 and 2008 and ~1 m in 2010 (Velpuri et al., 2012). The floods following the anomalous rainfall in 1997/1998

82

are implicated in the increased incidence of Malaria across East Africa, particularly in drier upland areas

83

(Hashizume et al., 2012; Linthicum et al., 1999).

84

Prior analyses of climate dynamics indicate that there is a plausible, but non-linear link between SSTs in the

85

equatorial oceans and the strength of the East and West African monsoons, which appears to modulate interannual to

86

interdecadal rainfall variability in equatorial East Africa (Black, 2005; Goddard and Graham, 1999; Omondi et al.,

87

2012). This study further examines the potential links between variable SSTs for the Indian and Atlantic oceans and

88

changes in East African precipitation, reflected in water level of Lake Turkana in the past ca. 150 years.

89

Specifically, this empirically-based statistical analysis attempts to partition by season, the moisture from Atlantic-

90

and Indian-derived sources, associated with the passage of the East and West African monsoons (Fig. 1c). This

91

analysis evaluates if there is a plausible link between changes in SSTs for the Atlantic and Indian oceans from AD

92

1992 to 2013 and ultimately variations in water level for Lake Turkana. In turn, the fidelity of this numeric relation

93

between SSTs and lake level changes is tested against a longer water level record from ca. AD 1857 to 1992 for

94

Lake Turkana. Finally, this study attempts to present a systematic approach to quantify pre-historical changes in

95

hydrologic balance for the Lake Turkana Basin with input of a single proxy from two oceanic proxy sources. This

96

approach is capable of reproducing water levels prior to AD 1888, a period for which our present understanding of

97

East African hydroclimate is largely inferred with at best decadal to multi-decadal resolution (e.g.) (Nicholson,

98

1988; Verschuren et al., 2000).

99

2. Lake Turkana hydrology and climatology

100

Lake Turkana is the largest lake in an arid environment (Sombroek et al., 1982) and is located within the East

101

African Rift Valley. Water levels are sustained mostly by discharge from the Omo River, sourced in the Ethiopian

102

Highlands, and the Kerio and Turkwel rivers which drain the Kenyan Highlands (Fig. 2). The Omo River catchment

103

(~73,000 km2) occupies about 50% of the area for the Lake Turkana Basin, and yields ~90% of the annual water

104

contribution (Avery, 2010), equivalent to 2.3 ± 0.6 m in lake level. The lake presently has no outlets and the only

105

significant loss of water is by evaporation from the lake surface. The annual estimated evaporation rate is 2.63 m/yr,

106

and is associated with an annual drop in lake level of ~0.5 m in the boreal winter (Avery, 2010; Hopson, 1982). The

107

remaining ~10% of water flow to Lake Turkana is assumed to be from the discharge of the Kerio and Turkwel rivers

108

and ephemeral streams and is equivalent to 0.13 m/yr of lake level (Avery, 2010).

109

Approximately >80% of the rainfall into the watershed of Lake Turkana during the 20th century occurs between

110

March and November with the biannual passage of the east and the west African monsoons, referred to as the

111

“short” and “long” rains (Avery, 2010). The “long” rains occur from early March to early June and reflect the

112

northward expansion of the Intertropical Convergence Zone (ITCZ) over East Africa. The “long” rains reach a

113

northward limit in late July and early August coeval with maximum discharge for the Omo River (Cheung et al.,

114

2008; Viste et al., 2013). The “short” rains are associated with the southward passage of the ITCZ, which may

115

deliver less rainfall than the “long” rains, though there is significant interannual variability (Camberlin and Okoola,

116

2003). These rains usually last from late September to early November, though the onset, and termination are

117

variable with rainfall persisting at times into early January (Black et al., 2003; Diro et al., 2011). To account for this

118

intraseasonal variability, the “long” rains are defined as the period between March and June, with the “short” rains

119

between October and January.

120

Anomalous SSTs in the Western Indian Ocean in the 20th and 21st centuries may modulate atmospheric

121

convection, the availability of precipitable water for East Africa, and effect the strength and the duration of passage

122

of the East African Monsoon (Black et al., 2003; Goddard and Graham, 1999; Saji and Yamagata, 2003). Warm

123

SSTs (~28 to 29 °C) adjacent to East Africa may enhance atmospheric convergence and result in increased “short”

124

rain season precipitation across Kenya, Ethiopia and Somalia. On balance, cooler SSTs (~ 24 to 25 °C) result in a

125

strengthened dry Turkana Jet (Kinuthia and Asnani, 1982; Nicholson, 1996), associated with appreciably less vapor

126

transport, often below the threshold for precipitable water (cf. Marchant et al., 2007; Nicholson, 1996). Typically,

127

the East African Monsoon in the boreal fall is associated with a relatively cool SSTs in the Western Indian Ocean

128

(~26 to 27 °C), with SSTs in the Eastern Indian Ocean comparatively ~2 to 3 °C warmer. This gradient in ocean

129

surface temperature reflects strong, westerly zonal winds across the Indian Ocean with equatorial convergence

130

which induces west to east surface currents associated with the Wyrtki Jet (Fig. 1b; Hastenrath et al., 1993; Wyrtki,

131

1973). Variations in the strength of the westerly winds in the boreal autumn modulates the west to east SST gradient

132

in the Indian Ocean with weak equatorial westerlies associated with a diminished Wyrtki Jet (Fig. 1a; Hastenrath et

133

al., 1993; Wyrtki, 1973). A weakened Wyrtki Jet may result in warm water to pool against the Kenyan coast (Xie et

134

al., 2002), with suppressed upwelling (Murtugudde and Busalacchi, 1999). A zone of low pressure often strengthens

135

over this warmer sea surface, which may reverse Walker Circulation over the Indian Ocean, further slowing the

136

Wyrtki Jet and sustaining warm SSTs along coastal Kenya (Hastenrath et al., 2010; Webster et al., 1999). Surface

137

convergence of moist air farther inland and air mass ascension over the Ethiopian and the Kenyan highlands

138

augments the intensity of the southward passage of the ITCZ, which can result in increased precipitation with the

139

“short” rains (Camberlin and Philippon, 2002).

140

Recent studies have linked anomalous SSTs for the Western Indian Ocean to changes in the equatorial zonal

141

SSTs gradient, expressed as the Indian Ocean Dipole (IOD) and quantified as the Dipole Mode Index (DMI; Abram

142

et al., 2008; Saji et al., 1999). The DMI is derived from a 140 year record of SSTs from the western (Seychelles) and

143

eastern (Sumatra) Indian Ocean (Fig. 1c; Saji et al., 1999) to quantify changes in the SST gradient across the Indian

144

Ocean. IOD events are defined by >1-σ deviations in the DMI record, with a ‘positive index’ event generally

145

reflected by warm Western Indian Ocean temperatures and a cooler Indo-Pacific Warm Pool (Saji et al., 1999).

146

Coral records spanning part of the past 7 ka from the Indian Ocean indicate the persistence of this ocean-atmosphere

147

dipole mechanism on decadal timescales (Abram et al., 2009). Coral records with subannual resolution from the

148

warm pool in the Western Indian Ocean show a significant correlation with rainfall and rainfall proxies for East

149

Africa (Damassa et al., 2006; Kayanne et al., 2006). Satellite observations of water mass changes (GRACE) for

150

lakes Turkana, Victoria, Tanganyika, and Malawi indicate that meter-scale variability in lake level changes for 2006

151

to 2007 are significantly correlated to zonal changes in the SST gradient of the Indian Ocean, typical of IOD events

152

(Becker et al., 2010).

153

Another potential significant source of moisture to Lake Turkana is from the West African Monsoon, with the

154

zonal advection of Atlantic-derived moisture from the Congo Basin (Williams et al., 2012). Rainfall over the Lake

155

Turkana Basin may be associated with a zone of convergence between Indian and Atlantic oceans derived-air

156

masses; an extension of the Congo Air Boundary (CAB; see Fig. 1c), which is often coincident with a precipitation

157

maximum over central Africa (Nicholson, 2000). The West African Monsoon occurs with the ITCZ passage across

158

west and central Africa with precipitation amount apparently modulated by SSTs in the Eastern Indian Ocean (EIO)

159

(Giannini et al., 2003; Nicholson, 2008). Interannual SST fluctuations in the Eastern Atlantic Ocean (EAO),

160

specifically for the Gulf of Guinea, appear to be spatially isolated, associated with the zonal Tropical Atlantic

161

variability (Diro et al., 2011; Lamb, 1978; Xie and Carton, 2004). The West African Monsoon is bimodal with two

162

annual precipitation peaks, concurrent with the “long” and “short” rains for East Africa, and with meridional

163

passage of the tropical rainfall belt from ~10 °N to ~15 °S (Nicholson, 2008).

164

3. Materials

165

This study analyzed monthly mean SST data from the Indian and Atlantic oceans for January, 1992 to April,

166

2013 and time series of similar duration of water level variability for Lake Turkana derived from satellite

167

measurements. The record of SSTs for the Western Indian and Eastern Atlantic oceans is derived from the NOAA

168

Optimum Interpolation global gridded dataset (OISST V2;

169

http://www.esrl.noaa.gov/psd/data/gridded/data.noaa.oisst.v2.html; Reynolds et al., 2002). These data are either

170

direct measurements of SSTs or satellite-derived SST estimates (1° by 1° resolution) and are averaged to yield

171

monthly mean SSTs (see Fig. 1c) for the Western Indian Ocean (WIO; Fig. 3a), the Eastern Indian Ocean (EIO; Fig.

172

3b), and the Eastern Atlantic Ocean (EAO), specifically the Gulf of Guinea (Fig. 3d). These records of SSTs are

173

presented as monthly mean anomalies from the 21-yr average (SSTAs). Further, a measure of zonal SST gradient is

174

derived from the monthly difference of SSTAs between WIO and EIO (hereafter ∆SST), which is significantly

175

correlated (r = 0.93; p<0.005) with the DMI record (Fig. 3c) (Saji et al., 1999). A range of SST variability is

176

represented in the time series of SSTAs for the WIO, with positive IOD events during 1997, 2006 and 2009, which

177

are also expressed in the DMI record and the ∆SST time series. Additionally, monthly mean anomalies in SSTs are

178

extracted from the NOAA Extended reconstructed SST dataset (ERSST v3b;

179

http://www.esrl.noaa.gov/psd/data/gridded/data.noaa.ersst.html; Smith et al., 2008) from 1854 to present for the

180

Western Indian, the Eastern Indian and the Eastern Atlantic oceans. The extended SST records for the Indian and

181

Atlantic oceans show a noticeable and sustained rise since 1950 and have been linearly de-trended for that period.

182

Further, SSTs data prior to 1950 have increasing error and uncertainty because of sparseness of data and

183

interpolations, which mutes variability (Smith et al., 2008). Also, anomalously warm SSTs between AD 1939 and

184

1941 (Kennedy et al., 2011; Reynolds et al., 2002) and cool SSTs in AD 1945 (Thompson et al., 2008) reflect

185

significant data artefacts, with changes in situ measurement of water temperatures. Thus, SSTs between 1936 and

186

1946 are normalized against a 2-yr moving average to remove interannual anomalies, yet preserving annual

187

fluctuations.

188

The record of lake level from November 1992 to May 2013 is remotely derived from measured lake surface

189

elevation every ~10 days with the passage of the TOPEX/Poseidon, Jason-1 and Jason-2/OSTM satellites over Lake

190

Turkana (http://www.pecad.fas.usda.gov/lakes/images/lake0093.TPJO.2.txt). The satellite measurements of lake

191

surface elevation is the basis for a time series of 30-day changes in water level and is presented as a record of

192

monthly lake level anomaly (Fig. 3e). An important time series for this analysis is a synthesized lake level record

193

between AD 1857 and 2012 (Avery, 2010; Johnson and Malala, 2009) which has variable resolution from annual to

194

multi-decadal reflecting constraining data from geomorphic observations (Nicholson, 1988), historical accounts

195

(Butzer, 1971), direct gauging (Hopson, 1982) and post-1992 satellite measurements (see Fig. 10a). This historic

196

lake record reflects interpolations between ~30 annual estimates of lake level, most prior to AD 1970 (Butzer,

197

1971). These measurements are discontinuous and the magnitude of peak lake level at ca. AD 1880 and 1900 is

198

based principally on observations of the elevation of delta surfaces for the Omo River. Many Inferred lake high and

199

low stands between AD 1900 to AD 1950 are constrained by a single observation (Butzer, 1971). Lake level

200

variability between AD 1950 and AD 1961 was measured from a gauge stationed in Ferguson’s Gulf, which drains

201

when lake level falls below 362.3 m and thus, is insensitive to lower lake levels [Avery, 2010].

202

Rainfall amounts for the Turkana Basin are extracted from the global 1° x 1° gridded time series of precipitation

203

(GPCC v6; http://www.esrl.noaa.gov/psd/data/gridded/data.gpcc.html). The GPCC v6 dataset combines

204

precipitation monitoring product for the years 1901 to 2010, and the “first guess” product for 2010 to April 2013.

205

The data used in this study is from January 1992 to April 2013 to closely match the water level record for Lake

206

Turkana, and then decomposed into monthly anomalies for the Lake Turkana Basin.

207

4. Results

208

4.1 The potential delay between changes in SSTs and Lake Turkana water level anomalies

209

Recent studies of monsoon variability indicate a significant relation to SST fluctuations and often with a

210

predictable delay associated with a moisture trajectory from an oceanic evaporative source to precipitation on the

211

adjacent landmass. This delay is associated with low level (850 hPa) transport of atmospheric moisture derived from

212

the WIO warm pool associated with modification to the zonal Walker Circulation and the passage of the ITCZ

213

(Mutai and Ward, 2000; Yamagata et al., 2004). Climate modeling of the seasonal variability of the IOD indicates a

214

significant correlation between Indian Ocean SSTs and anomalous precipitation for East Africa, with a lag of ~3-

215

months (Behera et al., 2005). However, other attempts to predict East African rainfall cite a shorter delay (~15 to 10

216

days) for 850 hPa wind anomalies as a proxy for heightened equatorial SSTs in the WIO (Mutai and Ward, 2000).

217

Similarly, rainfall events in East Africa are linked to a ‘westerly surge’ of Atlantic-derived moisture in the 850 hPa

218

wind field, which occurs ~20 days prior, resulting from SST variability in the Gulf of Guinea (Mutai and Ward,

219

2000; Tazalika and Jury, 2008). A hydrologic model infers a ~55 to ~100 day delay between precipitation in the

220

Turkana catchment and subsequent lake level response (Ricko et al., 2011b), although this lag is larger than assumed

221

(3 to 43 days) for a hydrologic model based on quantifying net-basin supply components (Velpuri et al., 2012).

222

The potential temporal offset between SSTAs and lake level change is assessed through lagged correlation

223

analyses of the respective monthly anomalies. These analyses directly compare monthly anomalies for a lag between

224

0 and 6 months, with the highest Pearson coefficient indicating the most probable lag (Fig. 4). These monthly time

225

series which comprise the years AD 1993 to AD 2013 represent a dataset with n = 252, and when pared down by

226

season these datasets have n = 80. The most significant correlation is associated with a 1-month lag between SSTAs

227

in the Western Indian (r=0.33; p<0.005) and Eastern Atlantic (r=0.16; p<0.05) oceans and the corresponding lake

228

level anomaly. Further, significant correlations are apparent seasonally, specifically between October to January

229

SSTAs in the Western Indian (r=0.59; p<0.005) and Eastern Atlantic (r=0.34; p<0.005) oceans and lake level

230

anomaly from November to February, which indicates a potential delay of one month. However, the highest

231

apparent correlation is associated with a delay of less than a month (r=0.37; p<0.005) between SSTAs (March to

232

June) in the WIO and lake level change. Surprisingly, the correlation coefficient is the highest (r=0.18; p≈0.11) for a

233

3 month lag between Eastern Atlantic SSTAs (March to June), and lake level anomaly (June to September), though

234

of low significance. Thus, in this study we assume a 30 day or less delay between changes in SSTs in the WIO and

235

resultant water level response for Lake Turkana, similar to previous studies (Mutai and Ward, 2000; Tazalika and

236

Jury, 2008; Velpuri et al., 2012). However, a delay of 90 days is applied between changes in SSTs (March to June)

237

in the Gulf of Guinea and lake level anomalies, reflecting far traveled and recycled moisture from the Congo Basin

238

(Cook and Vizy, 2012; Levin et al., 2009; Vizy and Cook, 2001; Williams et al., 2012).

239

4.2 Empirical orthogonal function analysis of SSTs and Lake Turkana Basin rainfall variability

240

Empirical orthogonal function (EOF) analysis is used to deduce temporal and spatial signatures of descending

241

modes of climatological variability, which facilitates multidimensional comparison between potentially associated

242

SSTAs and rainfall variability (e.g. Goddard and Graham, 1999; Mutai and Ward, 2000). Prior studies have

243

employed EOF analysis to deduce the geographic impact of IOD-type variability on rainfall anomalies for East

244

Africa (Bowden and Semazzi, 2007; Omondi et al., 2012) and changes in water volume for the East African Great

245

Lakes (Becker et al., 2010). Thus, an anomalous rise in water level for Lake Turkana may result from SST

246

anomalies, like with IOD-type variability (Avery, 2010; Becker et al., 2010), though this hydroclimatic connection

247

has yet to quantified statistically. This analysis relates the first EOF modes between time series of SSTAs and lake

248

level anomaly and between Turkana Basin rainfall and lake level changes. Significant correlations exist between

249

primary SSTA modes for the Eastern Atlantic (r=0.21, p<0.005) and the Western Indian oceans (r=0.35, p<0.005)

250

and lake level anomaly, which explains 66% and 55% of the variability, respectively (Fig. 5a and 5c). As, expected

251

there is a high significance between lake level anomaly and rainfall in the Lake Turkana Basin (r=0.73; p<0.005)

252

with this association reflecting about 50% of total variability (Fig. 5b). The primary geographic loading (EOF1) for

253

the Eastern Atlantic and Western Indian oceans are consistent with the spatial signature of precipitation with tropical

254

Atlantic SSTs variability (Fig. 5d) and positive IOD events (Fig. 5f), respectively (Saji et al., 1999; Vizy and Cook,

255

2001; Yamagata et al., 2004). The EOF2 explains 10.7% of variability and is broadly associated with the negative

256

mode of the tropical Atlantic variability, reflected by the prominent equatorial tongue of low SSTs extending

257

westward which occurs with easterly zonal wind shear and upwelling events in the Gulf of Guinea (Polo et al., 2013;

258

Vizy and Cook, 2002). The second mode (EOF2) for the Western Indian Ocean explains 10% of variability, and

259

reflects the westward extension of the equatorial cool water in the Indian Ocean, potentially indicative of a normal to

260

a negative IOD state (Saji et al., 1999; Williams and Funk, 2011).

261

Spatial correlation analysis between the principal temporal mode of SSTAs and rainfall anomaly for the Lake

262

Turkana Basin is used to deduce areas of the catchment associated with moisture derived from Western Indian and

263

Eastern Atlantic oceanic sources (Fig. 6). These principal temporal modes have been smoothed using a 3-month

264

moving average to emphasize areas of high statistical association on a yearly timescale, and smoothed for potential

265

varying atmospheric delays. The geographic distribution of correlations between rainfall anomalies and primary

266

mode of Western Indian SSTAs indicate a spatially diverse signature (Fig. 6b) potentially increasing in significance

267

with proximity to the Kenyan coast. The spatial correlation between basin rainfall anomalies and the primary mode

268

of variability for Eastern Atlantic SSTAs indicate a region of high significance in the lower Omo River catchment

269

(Fig. 6a). This area of higher loadings reflects a topographic saddle between the Kenyan and Ethiopian highlands,

270

with the lowest elevations (~450 masl) for the ~6000 km long continental divide (Indeje et al., 2001; Kinuthia and

271

Asnani, 1982). The spatial significance of this low continental divide is recognized previously as an orographic

272

pathways for Atlantic/Congo-derived moisture (e.g. Diro et al., 2011; Viste et al., 2013; Vizy and Cook, 2001).

273

4.3 Is there a correlation between equatorial SST and Lake level anomalies?

274

The influence of anomalous equatorial SSTs on rainfall for East Africa has been inferred on annual (Goddard

275

and Graham, 1999) and interannual timescales (Manatsa et al., 2012; Omondi et al., 2012), and for the “short” and

276

“long” rain seasons (Black et al., 2003; Camberlin and Okoola, 2003). Many studies indicate a potentially non-linear

277

response of East African rainfall to sea surface warming (Black et al., 2003; Ummenhofer et al., 2009), where

278

elevated seasonal SSTs in the WIO are associated with excess rainfall. Black et al. (2003) analyzed late 20th century

279

SSTs in the Indian Ocean, and rainfall anomalies for East Africa and demonstrated that excessive precipitation in

280

1997 is associated with elevated SSTs in the WIO with a >1.5 °C rise of the monthly mean SST. In turn, a δ 18O/16O

281

time series for coral records spanning the late 20th century from off of the Kenyan coast significantly correlates (r2 =

282

0.76; p<0.005) SST variability in the WIO with rainfall variations for East Africa (Kayanne et al., 2006). Recent

283

studies suggest that moisture transport into East Africa as a result of IOD-type events is enhanced by the zonal SST

284

gradient in the Indian Ocean, and the associated 850 hPa wind anomalies (Nicholson, 2013; Ummenhofer et al.,

285

2009). Our analysis expands on these observations by investigating the intraseasonal relation between anomalous

286

equatorial SSTs, during the short and long rains for East Africa and a possible response of water levels for Lake

287

Turkana. Specifically, a 21 year time series of SSTAs for the Eastern Atlantic and the Western Indian oceans, and

288

the SST gradient across the equatorial Indian Ocean, is partitioned by season and by SST state. Then the statistical

289

relation is evaluated between SST and lake level anomalies. Linear regression and associated correlation statistics

290

are augmented with Bootstrap sampling, and a function with errors is computed from 2000 Monte Carlo simulations

291

(Efron and Tibshirani, 1993; Zoubir, 1993).

292

This analysis focused on SSTAs for the Indian Ocean during the long and short rain seasons. The SST state for

293

the Indian Ocean is indicated by the averaged seasonal ∆SST, which reflects the zonal SST gradient for the

294

equatorial Indian Ocean. A positive SST state for the zonal Indian Ocean is recognized when the seasonal average

295

∆SST is >0.25σ, with a negative SST gradient defined by a seasonal averaged ∆SST <0.25σ. This is consistent with

296

the DMI record (Saji et al. (1999), with a positive SST gradient in the Indian Ocean often indicative of a positive

297

IOD event (Black et al., 2003; Cai and Qiu, 2013; Saji et al., 1999). Another regression analysis is conducted with a

298

positive SST state for the WIO warm pool is identified by a seasonally averaged SSTA of >0.25σ; negative SST

299

state is defined by a seasonally averaged SSTA <0.25σ. This ranking captures the anomalous increase in SSTs off

300

the Kenyan Coast associated with increased rainfall for East Africa (e.g. Goddard and Graham, 1999), while

301

isolating these SSTAs from processes which reflect basin-scale anomalies (cf. Black et al., 2003; Ummenhofer et al.,

302

2009). In the Eastern Atlantic (Gulf of Guinea), for all subsequent regression analyses, a positive SST state is

303

indicated by a seasonal averaged SSTA >0.25σ, whereas a negative SST state is defined as <0.25σ than the seasonal

304

mean SSTA.

305

Input time series consist of monthly SSTA or ∆SST values, systematically separated by the average seasonal

306

SSTA, and regressed against the corresponding month value for lake level anomaly, with the appropriate month

307

delay according to the lagged correlation analysis. For example, for the Western Indian Ocean during the “short

308

rains”, an average of October to January SSTAs above 0.25σ will select these four monthly values such that the

309

October SSTA value is paired with the lake level anomaly for November for one month lag. The value of 0.25σ is

310

chosen to systematically separate monthly SSTA or ∆SST values that result from a seasonal climate anomaly like

311

the IOD, while preserving a large sample size to ensure statistical robustness.

312

Statistical metrics for linear regression analyses indicate that there are significant, positive correlations between

313

Indian and Atlantic oceans SSTAs and changes in Lake Turkana water level (Figs. 7 and 8). The strongest apparent

314

correlation (r=0.78; p<0.005) is between SSTAs in the WIO from October to January, for a positive SST state (i.e.

315

seasonal averaged SSTA>0.25σ), and lake level anomaly from November to February (Fig. 7a). In contrast, for the

316

negative SST state (i.e. seasonal averaged SSTA<0.25σ) in the Indian Ocean, there is no significant association

317

between SSTAs and lake level changes (Fig. 7b). Further, the shift to a positive statistical correlation with an

318

increase in seasonal SSTA state of <0.25σ to >0.25σ may reflect a threshold-type response in seasonal precipitation

319

in East Africa, as inferred previously (Black et al. (2003). There is a significant apparent correlation (r=0.36;

320

p<0.05) between positive SSTAs for the EAO from October to January and lake level anomaly (Fig. 7e); with low

321

associated significance for the negative SST state (Fig. 7f). The only apparent significant correlation is for the long

322

rain season (March to June) between negative anomalies of SSTAs for the WIO and lake level (r=0.28; p<0.05; Fig.

323

8b), with low significance for the corresponding positive state of SSTAs (Fig. 8a). There appears to be low or no

324

appreciable statistical correlation between ∆SST (Figs. 8c and 8d), SSTAs in the EAO (Figs. 8e and 8f) and lake

325

level anomaly from March to June.

326

4.4 What is the validity of modeling Lake Turkana water level variability with changes in Western Indian and

327

Eastern Atlantic SSTs?

328

The EOF and linear regression analysis indicates a varied response of Lake Turkana water level to monthly

329

changes in SSTs for the Atlantic and Indian oceans, during the “long” and “short” rain seasons. We underscore that

330

the statistical relations between SSTs and lake level are apparent, reflecting the limited time span of data (21 years)

331

and the simplistic assumptions of linear regression analyses, though augmented by Bootstrap analysis and Monte

332

Carlo simulations. There are documented climatic dynamics which associate SST variability, particularly in the

333

Indian Ocean that appears to control precipitation receipts and lake level for East Africa (Becker et al., 2010; Birkett

334

et al., 1999; Black et al., 2003; Ummenhofer et al., 2009). Further, there are less significant relations during the

335

“long” rains between SSTAs in the Indian and Atlantic oceans and lake level anomaly which may reflect complex

336

and multiple sources of moisture, including the Gulf of Guinea, recycled air masses from the Congo Basin, and the

337

WIO with the northward migration of the ITCZ (Vizy and Cook, 2002; Polo et al., 2013).

338

To evaluate the fidelity between SST anomalies in the WIO and EAO and water level change for Lake Turkana

339

the results of the regression analysis are used to constrain these variables (Table 1) and tested against a longer record

340

of lake level (see Fig. 10a). The numeric solutions are derived from the regression analysis of the 21-year time series

341

of SSTAs and corresponding water level variability for Lake Turkana. These four equations describe lake level

342

variations for input of SSTAs from warm pools in the EAO and WIO (Fig. 3). Elevated WIO warm pool

343

temperatures are closely associated with increased rainfall for East Africa, rather than ∆SST across the equatorial

344

Indian Ocean (Ummenhofer et al., 2009). The four regression equations for positive SST states bound model output.

345

We quantitatively adjust the model output to match variations revealed in the satellite derived record of water level

346

variations for Lake Turkana. Also, to compensate for evaporation an annual lake fall of ~0.5 m/yr is subtracted,

347

consistent with the gauged lake level record (Avery, 2010; Ferguson and Harbott, 1982). Basin precipitation (see

348

Fig. 9) is inferred by partitioning the water volume for modelled water level changes for a lake with surface area of

349

7,334 km2 over a catchment area of ~148,600 km2 (cf. Forman et al., 2014). Finally, the model output is

350

parameterized to reflect the empirical seasonal rainfall distribution between 1992 and 2013; associated weightings

351

suggest long rain contribution is twice that of the short rains with seasonal moisture contributions from the Indian

352

Ocean comprising roughly three times that of the Atlantic Ocean (Table 1).

353

There are significant correlations (r = 0.85; p<0.005) between modeled rain fall and the empirical values for

354

annual rainfall contributions from the short rain season (Fig.9c). In contrast, the annual precipitation associated with

355

the long rain season is usually underestimated (Fig. 9b) and the associated correlation is not significant though there

356

are notable exceptions in AD 1998 and 2008-2009. The modeled total rainfall contribution is significantly correlated

357

(r=0.48; p<0.005) with the lake-level derived record of rainfall (Fig. 9a). The model calibration assumes a dominant

358

contribution from the WIO warm pool for both the “long” and “short” rains and simulations reflect a minimal, and

359

relatively consistent Atlantic-derived rainfall contribution of ~30 mm/yr (Figs. 10a and 10b). Monthly analysis

360

indicates the highest correlation coefficients and significance occurs during positive IOD events in 1997, 2006 and

361

2008-2011, when moisture is predominately sourced from the WIO (Cai et al., 2009; Saji et al., 1999; Ummenhofer

362

et al., 2009).

363

The functions defined from the 21-yr record of SST anomalies and calibrated lake level response (Table 1) are

364

tested against a longer record of SST anomalies from AD 1857 to 1992 (Fig. 10a) and evaluated against

365

corresponding change in water level for Lake Turkana (Fig. 10b). An implicit and untested assumption is that the

366

apparent SSTs to lake level relationship on an interannual to annual timescale for AD 1992 to 2013 is applicable for

367

a longer simulation of water level variability for Lake Turkana. Simulations of lake level are hind-cast from an

368

average water plain for AD 2012 at +1.2 m. The corresponding uncertainty for lake levels are calculated from the 1-

369

σ error associated with the Monte Carlo/bootstrap analysis for linear regression statistics (see Figs. 7 and 8) and is

370

summed cumulatively (Fig. 10b). The solutions to the model equations are weighted relative to the associated

371

correlation coefficients, identical to model lake level in the late 20th and early 21st centuries (Table 1). Also

372

incorporate seasonal variations in evaporative loss (equivalent to 0.5 m/yr), with scaled output defined in the 21-yr

373

calibration period (Fig. 10b). Further, the coefficients of these four functions, which quantify monthly lake

374

variations based on SSTAs, are weighted according to the associated correlations and resolve two functions which

375

describe annual lake level changes for input of annual SSTAs for the WIO and the EAO (Table 1).

376

There is an apparent significant correlation between the hind-cast lake level, based on monthly input SSTAs,

377

and the empirically-derived lake level for AD 1857 to 2013 (r = 0.90; p < 0.005; Fig. 10b). A simplified model

378

calculation with input of annually resolved SSTAs captures the multi-decadal lake level variability, with

379

considerable significance (r=0.85; p< 0.005; Fig. 10b). The monthly-based simulation captures well the 17 m fall in

380

lake level between Ad 1890 and 1940. However, on annual to decadal timescales there are a number of low

381

amplitude (< 2 m) discontinuities between the simulated and empirical-based lake level prior to AD 1950. Further,

382

pronounced ~3 m variability which includes the modeled low water level of -2.5 m between AD 1957 and 1940 is

383

not represented in measured lake level record, though there is a noticeable gap in this record between AD 1938 and

384

1948. Single observations that constrain high and low lake stands between AD 1920 and 1938 are inconsistently

385

replicated, with discrepancies in magnitude and timing of peaks. There is an inconsistency with the timing of

386

simulated lake level fluctuations between AD 1857 and AD 1910 with one broad and sustained peak, which differs

387

from the distinct double peak in lake level in the empirical record, though this couplet is not well constrained

388

chronological These second order discrepancies between simulated and historic lake levels pre AD 1955 may reflect

389

discontinuous and relative observations of water levels (Avery, 2010; Butzer, 1971; Johnson and Malala, 2009;

390

Nicholson, 1988) or degraded resolution of the extended SST record (ERSST), which exhibits increased uncertainty

391

prior to 1950 (Smith et al., 2008).

392

5. Discussion

393

The EOF analysis (Figs. 5a-5c) and associated correlations (Fig. 5d-5f) indicate a significant relation amongst

394

changes in basin rainfall and lake level, and SST variability for the WIO warm pool. The principal mode of Lake

395

Turkana Basin rainfall, as a 3-month moving average, is significantly correlated with lake level anomaly (r=0.73;

396

p<0.005; Fig. 5b). In contrast, there is lower statistical association between Atlantic SSTAs and lake level change,

397

though the EOF spatial signature is centered on the Gulf of Guinea, an important source area for Atlantic-derived

398

moisture (Cook and Vizy, 2012; Levin et al., 2009). The highest spatial correlation between Eastern Atlantic SSTAs

399

and lake level variability is for the Kenyan Highlands which may be linked to low level westerly wind anomalies

400

(Williams et al., 2012) and a low topographic divide, which may be a preferential pathway for Atlantic/Congo Basin

401

derived moisture (Fig. 6a) (Hession and Moore, 2011; Indeje et al., 2001; Kinuthia and Asnani, 1982).

402

There are significant spatial correlations between Western Indian Ocean SSTAs and rainfall for the Ethiopian

403

Highlands and the southernmost catchment of the Turkana Basin (Fig. 6b). The first EOF mode, associated with

404

58% of the rainfall variability in the Basin, has a similar footprint indicating the primacy of moisture sources from

405

the southern Ethiopian Highlands and catchments of the Turkwel and Kerio rivers (Fig. 6c). The second EOF

406

explains 16% of the rainfall variability with a distinct spatial signature for the northern part of the catchment (Fig.

407

6d), potentially reflecting advected moisture from an Indian Ocean source. There also appears to be a spatial

408

precipitation dipole in East African possibly linked to SST variability; Kenyan and eastern Ethiopian rainfall is

409

associated with Indian Ocean-derived moisture and varies in anti-phase with western Ethiopia and Uganda,

410

associated with Atlantic Ocean-derived moisture (Bowden and Semazzi, 2007; Omondi et al., 2012; Tierney et al.,

411

2013). The spatial pattern for correlations between Turkana Basin rainfall anomalies and the first temporal mode of

412

variability for SSTAs appear to reflect this dipole, with positive significance for the WIO associated with the

413

southwestern portion of the Lake Turkana Basin (Fig. 6b).

414

The state of the SSTs in the Western Indian Ocean (WIO) has an inferred effect on precipitation variability for

415

the Turkana Basin and thus, lake level. Elevated SSTAs (>0.25 σ) have an apparent significant relation to lake level

416

change, particularly during the “short” rains (Fig 7a). Lower, yet significant correlations exist between the Indian

417

Ocean SST gradient (∆SST) and lake level gains with the “short” rains (Fig. 7c). There is an apparent less

418

significant relation between Eastern Atlantic SSTAs and lake level change for the same period (Fig. 7e). In contrast,

419

lower SSTAs (<0.25 σ) in the Indian and Atlantic oceans have no apparent statistical significance (p ≥ 0.22) with

420

changes in lake volume during the “short” rains (Figs. 7b, 7d and 7f). There are weak to insignificant relations

421

between SSTAs or ∆SST for the Indian and Atlantic oceans and registered changes in lake volume during the “long”

422

rains (Fig. 8). Though the time series is short (21 years), this analysis indicates a potential threshold-like response

423

with SSTAs ~>0.25σ for the WIO warm pool for above average rainfall and associated rise in lake level (Black et

424

al., 2003).

425

Our analyses show consistently low statistical significance between SSTAs for the EAO and WIO and lake

426

level changes during the “long” rain season. Other parameters such as cloudiness, wind vectors and sea level

427

pressure differences in the EAO yield significant relations with precipitation in the Greater Horn of Africa (Williams

428

et al., 2011). There is a complex derivation of moisture during the “long” rains which involves the northward

429

movement of ITCZ with variable sources from cyclonic activity born in the Gulf Guinea, recycled moisture from the

430

Sudd and Congo Basin, and advection from the Indian Ocean (Levin et al., 2009; Williams et al., 2012). Isotopic

431

analyses of meteoric rainfall in the Ethiopian Highlands reveal variable moisture sources with mostly positive δ18O

432

values (~-0.2 ‰) relative to Kenyan rainfall (~-2.5 ‰) (Levin et al., 2009). These distinct isotopic differences

433

indicate that rain fall in Kenya is mostly derived from the Indian Ocean; whereas Omo River discharge sourced from

434

the Ethiopian Plateau reflects variable moisture contributions from the Congo Basin, and other Atlantic-derived

435

sources. Further, one tree-ring time series for Juniperus procera from the Ethiopian Highlands for the period AD

436

1915 to AD 2010 show a >2 ‰ decline in cellulose δ18O values since ca. AD 1990, associated with decreased

437

moisture contributions from the Congo Basin, which is implicated in pervasive drought conditions in the past decade

438

(Williams et al., 2012).

439

Recent statistical analyses indicate IOD events and associated SST variance may have increased in the past

440

century (Manatsa et al., 2012), and is a likely causative factor for increased precipitation variability in East Africa,

441

with potential linkages to broader-scale warming (Manatsa and Behera, 2013). Aridity in the 21st century for Central

442

and East Africa is associated with an increase in vertical energy flux over the southern tropical Indian Ocean that

443

leads to dry static energy exported to the East Africa, increasing atmospheric stability and subsidence (Williams et

444

al., 2012). Thus, the pronounced >15m drop in lake level through the 20th century may reflect decreasing Congo

445

Basin derived-moisture (Williams et al., 2012) with meter-scale variability in lake level attributed to increase in SST

446

variance for WIO warm pool (Manatsa et al., 2012). An earlier pluvial period in East Africa ca. AD 1680 to AD

447

1765 is also inferred to reflect positive in SSTs anomalies in the western Indian Ocean (Tierney et al., 2013;

448

Ummenhofer et al., 2009).

449

The inferred water level variability for Lake Turkana in the past ca. 150 years is consistent with other

450

hydrologic records from East Africa, particularly from Lake Naivasha in southeast Kenya (Fig. 10c). Instrumental

451

measurements, historic observations and proxy records for many lakes in East Africa indicate peak water levels at

452

ca. AD 1870 and 1900 (e.g. Nicholson, 2000; Verschuren et al., 2001), the latter coincident with an inferred

453

historical high stand for Lake Turkana (Butzer, 1971). Our model, based on monthly SSTA, replicates a number of

454

meter-scale rises in lake level between AD 1920 and 1990 (Fig. 10b). However, significant peaks in lake level at AD

455

1916 and 1899 are not fully resolved, though a broad high stand reconstructed between AD 1860 and 1890 is to

456

similar heights (within 2 sigma errors) of the record high stand (Fig. 10b). Lake level reconstructions based on

457

annually averaged SSTA show multi-decadal scale variability consistent with the historical record for Lake Turkana

458

(Fig. 10).

459

This study indicates that a robust approach to modeling water level variability for Lake Turkana is dependent on

460

monthly SSTA input for the Western Indian Ocean and the Gulf of Guinea. However, proxies such as coral bands,

461

are typically resolved annually (Cole et al., 2000; Damassa et al., 2006; Nakamura et al., 2011), though seasonal

462

resolution may be possible (Kayanne et al., 2006). Thus, we have developed an annual SSTA-lake level model based

463

on corresponding monthly parameterization which can hind-cast past lake level variations, with a highly significant

464

correlation (r=0.95; p<0.005). However, a clear deficiency of this model remains the inability to hind-cast lake

465

levels beyond the temporal limit of instrumental SST data (AD 1854). Further, we assume a dominance of the

466

western Indian Ocean as the prime moisture source, rather than Atlantic/Congo Basin derived sources. One tree-ring

467

time series from the Ethiopian plateau indicates that the flux of Atlantic-derived moisture has progressively

468

diminished in the past 100 years (Williams et al., 2012). This approach to lake level reconstructions is constrained

469

by the short instrumental record of high quality SSTs measurements in WIO and EAO. Coral records from the WIO

470

are important proxy records that could extend temporally this lake level reconstruction exercise. Currently, the

471

longest coral record for the WIO warm pool is a discontinuous time series of inferred changes in SSTs for the much

472

of the 17th and 20th centuries (Damassa et al., 2006) and continuous coral records for the Gulf of Guinea are limited

473

to the 20th century (Swart et al., 1998). A critical datum deduced from early observations and our modeling is that

474

water level for Lake Turkana was 5 ± 5 m at AD 1857 and is a “starting” water level for further hind-cast

475

simulations with proxy SST input. Recent field research of relict beaches around the shores of Lake Turkana have

476

deduced decadal and subdecadal water level changes in the past 1000 years (Forman et al., 2014) and in tandem with

477

SST-based modeling of lake level change should yield new insights on water supply and climate variability for East

478

Africa.

479

6. Conclusion

480

There appears to be a significant spatial and temporal relation between variability in SSTs in the WIO and EAO

481

and water level changes for Lake Turkana in the past ca. 21 years. EOF analyses indicate that significant

482

correlations exist between primary SSTA modes for the Eastern Atlantic and the Western Indian oceans and lake

483

level anomaly, which explain 66% and 55% of the variability, respectively. The primary geographic loading for the

484

Eastern Atlantic and Western Indian oceans are consistent with the spatial signature of precipitation with tropical

485

Atlantic SSTs variability and positive IOD events (Ruiz-Barradas et al., 2003; Saji et al., 1999; Vizy and Cook,

486

2001; Yamagata et al., 2004). The second mode explains 10% of variability, and reflects the westward extension of

487

the equatorial cool water in the Indian Ocean, potentially indicative of a normal to a negative IOD state (Saji et al.,

488

1999; Williams and Funk, 2011). There is significant spatial correlation between basin rainfall anomalies associated

489

with Eastern Atlantic SSTAs and a low in the continental divide between the Kenyan and Ethiopian highlands,

490

which is a passage for moisture from the Congo Basin.

491

Linear regression analysis augmented by Bootstrap sampling and Monte Carlo simulations define numeric

492

relations between WIO and EAO SSTs and lake level change for AD 1992 to 2013. The robustness of this numeric

493

solution to reconstructed lake level is evaluated against the historic record from 1857 to 1992. The monthly and

494

yearly lake level reconstructions capture the decadal-scale variability and the 15 m drop in water level in the early

495

20th century. This model assumes the primacy of the WIO as a source of moisture, particularly during the East

496

African Monsoon, which is consistent with decreasing influence of Atlantic-derived moisture on East Africa

497

hydrology through the 20th century (cf. Williams et al., 2012). Thus, meter-scale variability in lake level may be

498

related to variable precipitation sourced from the WIO with IOD variability, whereas the 15 m drop in early 20th

499

century may reflect a profound decrease in moisture from Atlantic/Congo Basin sources.

500

7. References

501 502

Abram, N.J., Gagan, M.K., Cole, J.E., Hantoro, W.S., Mudelsee, M., 2008. Recent intensification of tropical climate variability in the Indian Ocean. Nature Geosci 1, 849-853.

503 504

Abram, N.J., McGregor, H.V., Gagan, M.K., Hantoro, W.S., Suwargadi, B.W., 2009. Oscillations in the southern extent of the Indo-Pacific Warm Pool during the mid-Holocene. Quaternary Science Reviews 28, 2794-2803.

505 506

Ackerley, D., Booth, B.B.B., Knight, S.H.E., Highwood, E.J., Frame, D.J., Allen, M.R., Rowell, D.P., 2011. Sensitivity of Twentieth-Century Sahel Rainfall to Sulfate Aerosol and CO2 Forcing. Journal of Climate 24, 4999-5014.

507 508

Ashok, K., Saji, N., 2007. On the impacts of ENSO and Indian Ocean dipole events on sub-regional Indian summer monsoon rainfall. Natural Hazards 42, 273-285.

509 510

Avery, S., 2010. Hydrological impacts of Ethiopia's Omo basin on Kenya's lake Turkana water levels and fisheries. African Development Bank, Tunis, p. 146.

511 512

Becker, M., LLovel, W., Cazenave, A., Güntner, A., Crétaux, J.-F., 2010. Recent hydrological behavior of the East African great lakes region inferred from GRACE, satellite altimetry and rainfall observations. Comptes Rendus Geoscience 342, 223-233.

513 514

Behera, S.K., Luo, J.J., Masson, S., Delecluse, P., Gualdi, S., Navarra, A., Yamagata, T., 2005. Paramount impact of the Indian Ocean dipole on the East African short rains: A CGCM study. Journal of Climate 18, 4514-4530.

515

Biasutti, M., 2013. Forced Sahel rainfall trends in the CMIP5 archive. Journal of Geophysical Research: Atmospheres.

516 517

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.

518 519

Black, E., 2005. The relationship between Indian Ocean sea–surface temperature and East African rainfall. Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences 363, 43-47.

520 521

Black, E., Slingo, J., Sperber, K.R., 2003. An Observational Study of the Relationship between Excessively Strong Short Rains in Coastal East Africa and Indian Ocean SST. Monthly Weather Review 131, 74-94.

522 523

Bowden, J.H., Semazzi, F.H., 2007. Empirical analysis of intraseasonal climate variability over the Greater Horn of Africa. Journal of Climate 20, 5715-5731.

524 525

Butzer, K.W., 1971. Recent history of an Ethiopian delta: the Omo River and the level of Lake Rudolf. University of Chicago, Dept. of Geography, Chicago.

526 527

Cai, W., Qiu, Y., 2013. An Observation-Based Assessment of Nonlinear Feedback Processes Associated with the Indian Ocean Dipole. Journal of Climate 26, 2880-2890.

528 529

Cai, W., Sullivan, A., Cowan, T., 2009. How rare are the 2006-2008 positive Indian Ocean Dipole events? An IPCC AR4 climate model perspective. Geophysical Research Letters 36.

530 531

Camberlin, P., Okoola, R.E., 2003. The onset and cessation of the "long rains" in eastern Africa and their interannual variability. Theoretical and Applied Climatology 75, 43-54.

532 533

Camberlin, P., Philippon, N., 2002. The East African March-May rainy season: Associated atmospheric dynamics and predictability over the 1968-97 period. Journal of Climate 15, 1002-1019.

534 535

Cherchi, A., Alessandri, A., Masina, S., Navarra, A., 2011. Effects of increased CO2 levels on monsoons. Climate dynamics 37, 83-101.

536 537

Cheung, W.H., Senay, G.B., Singh, A., 2008. Trends and spatial distribution of annual and seasonal rainfall in Ethiopia. International Journal of Climatology 28, 1723-1734.

538 539 540 541 542

Christensen, J.H., B. Hewitson, A. Busuioc, A. Chen, X. Gao, I. Held, R. Jones, R.K. Kolli, W.-T. Kwon, R. Laprise, V. Magaña Rueda, L. Mearns, C.G. Menéndez, J. Räisänen, A. Rinke, A. Sarr and P. Whetton, 2007. Regional Climate Projections, In: Solomon, S., D. Qin, M. Manning, Z. Chen, M. Marquis, K.B. Averyt, M. Tignor, H.L. Miller (Ed.), Climate Change, 2007: The Physical Science Basis. Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change. University Press, Cambridge, pp. 847-940.

543 544

Cole, J.E., Dunbar, R.B., McClanahan, T.R., Muthiga, N.A., 2000. Tropical Pacific forcing of decadal SST variability in the western Indian Ocean over the past two centuries. Science 287, 617-619.

545 546

Cook, K.H., Vizy, E.K., 2012. Impact of climate change on mid-twenty-first century growing seasons in Africa. Climate Dynamics 39, 2937-2955.

547 548

Damassa, T.D., Cole, J.E., Barnett, H.R., Ault, T.R., McClanahan, T.R., 2006. Enhanced multidecadal climate variability in the seventeenth century from coral isotope records in the western Indian Ocean. Paleoceanography 21.

549 550

Diro, G., Grimes, D.I.F., Black, E., 2011. Teleconnections between Ethiopian summer rainfall and sea surface temperature: part I—observation and modelling. Climate dynamics 37, 103-119.

551

Du, Y., Cai, W., Wu, Y., 2013. A New Type of the Indian Ocean Dipole since the Mid-1970s. Journal of Climate 26, 959-972.

552

Efron, B., Tibshirani, R., 1993. An introduction to the bootstrap. CRC press.

553 554

Ferguson, A., Harbott, B., 1982. Geographical, physical and chemical aspects of Lake Turkana. Lake Turkana: A report of the findings of the Lake Turkana Project 1975, 1-107.

555 556

Folland, C.K., Palmer, T.N., Parker, D.E., 1986. SAHEL RAINFALL AND WORLDWIDE SEA TEMPERATURES, 1901-85. Nature 320, 602-607.

557 558

Forman, S.L., Wright, D.K., Bloszies, C., 2014. Variations in water level for Lake Turkana in the past 8500 years near Mt. Porr, Kenya and the transition from the African Humid Period to Holocene aridity. Quaternary Science Reviews 97, 84-101.

559 560

Funk, C., Eilerts, G., Verdin, J., Rowland, J., Marshall, M., 2011. A Climate Trend Analysis of Sudan. U.S. Geological Survey Fact Sheet 2011-3072, 6 p.

561 562 563

Funk, C., Senay, G., Asfaw, A., Verdin, J., Rowland, J., Michaelson, J., Eilerts, G., Korecha, D., Choularton, R., 2005. Recent drought tendencies in Ethiopia and equatorial-subtropical eastern Africa. Famine Early Warning System Network, USAID, Washington, DC.

564 565

Giannini, A., 2010. Mechanisms of Climate Change in the Semiarid African Sahel: The Local View. Journal of Climate 23, 743756.

566 567

Giannini, A., Biasutti, M., Verstraete, M.M., 2008. A climate model-based review of drought in the Sahel: Desertification, the regreening and climate change. Global and planetary Change 64, 119-128.

568 569

Giannini, A., Saravanan, R., Chang, P., 2003. Oceanic forcing of Sahel rainfall on interannual to interdecadal time scales. Science 302, 1027-1030.

570 571

Goddard, L., Graham, N.E., 1999. Importance of the Indian Ocean for simulating rainfall. Journal of Geophysical Research 104, 19,099-019,116.

572 573

Hashizume, M., Chaves, L.F., Minakawa, N., 2012. Indian Ocean Dipole drives malaria resurgence in East African highlands. Scientific Reports 2.

574 575

Hastenrath, S., Nicklis, A., Greischar, L., 1993. Atmospheric-hydrospheric mechanisms of climate anomalies in the western equatorial Indian Ocean. Journal of Geophysical Research: Oceans 98, 20219-20235.

576 577

Hastenrath, S., Polzin, D., Mutai, C., 2010. Diagnosing the Droughts and Floods in Equatorial East Africa during Boreal Autumn 2005-08. Journal of Climate 23, 813-817.

578 579

Hession, S., Moore, N., 2011. A spatial regression analysis of the influence of topography on monthly rainfall in East Africa. International Journal of Climatology 31, 1440-1456.

580 581

Hoerling, M., Hurrell, J., Eischeid, J., Phillips, A., 2006. Detection and attribution of twentieth-century northern and southern African rainfall change. Journal of Climate 19, 3989-4008.

582 583

Hopson, A.J., 1982. Lake Turkana: a report on the findings of the Lake Turkana Project, 1972-1975. Overseas Development Administration London.

584 585

Hsu, P.-c., Li, T., Luo, J.-J., Murakami, H., Kitoh, A., Zhao, M., 2012. Increase of global monsoon area and precipitation under global warming: A robust signal? Geophysical Research Letters 39, L06701.

586 587

Indeje, M., Semazzi, F.H., Xie, L., Ogallo, L.J., 2001. Mechanistic model simulations of the East African climate using NCAR regional climate model: influence of large-scale orography on the Turkana low-level jet. Journal of climate 14, 2710-2724.

588

Johnson, T.C., Malala, J.O., 2009. Lake Turkana and its link to the Nile. The Nile, 287-304.

589 590

Kayanne, H., Iijima, H., Nakamura, N., McClanahan, T.R., Behera, S., Yamagata, T., 2006. Indian Ocean Dipole index recorded in Kenyan coral annual density bands. Geophysical Research Letters 33, L19709.

591 592 593

Kennedy, J., Rayner, N., Smith, R., Parker, D., Saunby, M., 2011. Reassessing biases and other uncertainties in sea surface temperature observations measured in situ since 1850: 2. Biases and homogenization. Journal of Geophysical Research: Atmospheres (1984–2012) 116.

594

Kinuthia, J., Asnani, G., 1982. A newly found jet in North Kenya (Turkana Channel). Monthly Weather Review 110, 1722-1728.

595 596

Kucharski, F., Zeng, N., Kalnay, E., 2013. A further assessment of vegetation feedback on decadal Sahel rainfall variability. Climate Dynamics 40, 1453-1466.

597 598

Lamb, P.J., 1978. Large-scale Tropical Atlantic surface circulation patterns associated with Subsaharan weather anomalies. Tellus 30, 240-251.

599 600

Levin, N.E., Zipser, E.J., Cerling, T.E., 2009. Isotopic composition of waters from Ethiopia and Kenya: insights into moisture sources for eastern Africa. Journal of Geophysical Research: Atmospheres (1984–2012) 114.

601 602

Linthicum, K.J., Anyamba, A., Tucker, C.J., Kelley, P.W., Myers, M.F., Peters, C.J., 1999. Climate and satellite indicators to forecast Rift Valley fever epidemics in Kenya. Science 285, 397-400.

603 604

Lyon, B., DeWitt, D.G., 2012. A recent and abrupt decline in the East African long rains. Geophysical Research Letters 39, L02702.

605 606

Manatsa, D., Behera, S.K., 2013. On the epochal strengthening in the relationship between rainfall of East Africa and IOD. Journal of Climate.

607 608

Manatsa, D., Chipindu, B., Behera, S.K., 2012. Shifts in IOD and their impacts on association with East Africa rainfall. Theoretical and Applied Climatology 110, 115-128.

609 610

Marchant, R., Mumbi, C., Behera, S., Yamagata, T., 2007. The Indian Ocean dipole–the unsung driver of climatic variability in East Africa. African Journal of Ecology 45, 4-16.

611 612

Murtugudde, R., Busalacchi, A.J., 1999. Interannual Variability of the Dynamics and Thermodynamics of the Tropical Indian Ocean. Journal of Climate 12, 2300-2326.

613 614

Mutai, C., Polzin, D., Hastenrath, S., 2012. Diagnosing Kenya Rainfall in Boreal Autumn: Further Exploration. Journal of Climate 25, 4323-4329.

615 616

Mutai, C.C., Ward, M.N., 2000. East African Rainfall and the Tropical Circulation/Convection on Intraseasonal to Interannual Timescales. Journal of Climate 13, 3915-3939.

617 618

Nakamura, N., Kayanne, H., Iijima, H., McClanahan, T.R., Behera, S.K., Yamagata, T., 2011. Footprints of IOD and ENSO in the Kenyan coral record. Geophys. Res. Lett. 38, L24708.

619

Nicholson, S., 2000. Land surface processes and Sahel climate. Reviews of Geophysics 38, 117-139.

620 621

Nicholson, S.E., 1988. Historical fluctuations of Lake Victoria and other lakes in the northern Rift Valley of East Africa. MONOGRAPHIAE BIOLOGICAE, 7-36.

622 623

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, pp. 25-56.

624 625

Nicholson, S.E., 2008. The intensity, location and structure of the tropical rainbelt over west Africa as factors in interannual variability. International Journal of Climatology 28, 1775-1785.

626 627

Nicholson, S.E., 2009. A revised picture of the structure of the “monsoon” and land ITCZ over West Africa. Climate Dynamics 32, 1155-1171.

628 629

Nicholson, S.E., 2013. The West African Sahel: A Review of Recent Studies on the Rainfall Regime and Its Interannual Variability. ISRN Meteorology 2013, 32.

630 631

Nicholson, S.E., Tucker, C.J., Ba, M., 1998. Desertification, drought, and surface vegetation: An example from the West African Sahel. Bulletin of the American Meteorological Society 79, 815-830.

632 633

Omondi, P., Awange, J.L., Ogallo, L.A., Okoola, R.A., Forootan, E., 2012. Decadal rainfall variability modes in observed rainfall records over East Africa and their relations to historical sea surface temperature changes. Journal of Hydrology 464, 140-156.

634 635

Polo, I., Dong, B.W., Sutton, R.T., 2013. Changes in tropical Atlantic interannual variability from a substantial weakening of the meridional overturning circulation. Climate Dynamics, 1-20.

636 637

Reynolds, R.W., Rayner, N.A., Smith, T.M., Stokes, D.C., Wang, W., 2002. An improved in situ and satellite SST analysis for climate. Journal of climate 15, 1609-1625.

638 639

Ricko, M., Carton, J.A., Birkett, C., 2011a. Climatic Effects on Lake Basins. Part I: Modeling Tropical Lake Levels. Journal of Climate 24.

640 641

Ricko, M., Carton, J.A., Birkett, C., 2011b. Climatic effects on lake basins. Part I: modeling tropical lake levels. Journal of Climate 24, 2983-2999.

642 643

Riddle, E.E., Cook, K.H., 2008. Abrupt rainfall transitions over the Greater Horn of Africa: Observations and regional model simulations. Journal of Geophysical Research 113, D15109.

644 645

Ruiz-Barradas, A., Carton, J.A., Nigam, S., 2003. Role of the atmosphere in climate variability of the tropical Atlantic. Journal of climate 16, 2052-2065.

646 647

Saji, N., Yamagata, T., 2003. Possible impacts of Indian Ocean dipole mode events on global climate. Climate Research 25, 151169.

648 649

Saji, N.H., Goswami, B.N., Vinayachandran, P.N., Yamagata, T., 1999. A dipole mode in the tropical Indian Ocean. Nature 401, 360-363.

650 651

Shongwe, M.E., van Oldenborgh, G.J., van den Hurk, B., van Aalst, M., 2011. Projected changes in mean and extreme precipitation in Africa under global warming. Part II: East Africa. Journal of Climate 24, 3718-3733.

652 653

Smith, T.M., Reynolds, R.W., Peterson, T.C., Lawrimore, J., 2008. Improvements to NOAA's historical merged land-ocean surface temperature analysis (1880-2006). Journal of Climate 21, 2283-2296.

654 655

Sombroek, W.G., Braun, H., Pouw, B.J.A., 1982. Exploratory soil map and agro-climatic zone map of Kenya, 1980. Scale 1: 1,000,000. Kenya Soil Survey.

656 657 658

Swart, P.K., White, K.S., Enfield, D., Dodge, R.E., Milne, P., 1998. Stable oxygen isotopic composition of corals from the Gulf of Guinea as indicators of periods of extreme precipitation conditions in the sub‐Sahara. Journal of Geophysical Research: Oceans (1978–2012) 103, 27885-27891.

659 660

Tazalika, L., Jury, M., 2008. Intra-seasonal rainfall oscillations over central Africa: space-time character and evolution. Theoretical and Applied Climatology 94, 67-80.

661 662

Thompson, D.W., Kennedy, J.J., Wallace, J.M., Jones, P.D., 2008. A large discontinuity in the mid-twentieth century in observed global-mean surface temperature. Nature 453, 646-649.

663 664

Tierney, J.E., Smerdon, J.E., Anchukaitis, K.J., Seager, R., 2013. Multidecadal variability in East African hydroclimate controlled by the Indian Ocean. Nature 493, 389-392.

665 666

Ummenhofer, C.C., Sen Gupta, A., England, M.H., Reason, C.J., 2009. Contributions of Indian Ocean sea surface temperatures to enhanced East African rainfall. Journal of Climate 22, 993-1013.

667 668

Velpuri, N.M., Senay, G.B., Asante, K.O., 2012. A multi-source satellite data approach for modelling Lake Turkana water level: calibration and validation using satellite altimetry data. Hydrol. Earth Syst. Sci. 16, 1-18.

669 670

Verschuren, D., Laird, K.R., Cumming, B.F., 2000. Rainfall and drought in equatorial east Africa during the past 1,100 years. Nature 403, 410-414.

671 672

Viste, E., Korecha, D., Sorteberg, A., 2013. Recent drought and precipitation tendencies in Ethiopia. Theoretical and Applied Climatology 112, 535-551.

673 674

Vizy, E.K., Cook, K.H., 2001. Mechanisms by which Gulf of Guinea and eastern North Atlantic sea surface temperature anomalies can influence African rainfall. Journal of Climate 14, 795-821.

675 676 677

Vizy, E.K., Cook, K.H., 2002. Development and application of a mesoscale climate model for the tropics: Influence of sea surface temperature anomalies on the West African monsoon. Journal of Geophysical Research: Atmospheres (1984–2012) 107, ACL 2-1-ACL 2-22.

678 679

Wang, B., Liu, J., Kim, H.-J., Webster, P.J., Yim, S.-Y., 2012. Recent change of the global monsoon precipitation (1979–2008). Climate dynamics 39, 1123-1135.

680 681

Webster, P.J., Moore, A.M., Loschnigg, J.P., Leben, R.R., 1999. Coupled ocean–atmosphere dynamics in the Indian Ocean during 1997–98. Nature 401, 356-360.

682 683

Williams, A.P., Funk, C., 2011. A westward extension of the warm pool leads to a westward extension of the Walker circulation, drying eastern Africa. Climate Dynamics 37, 2417-2435.

684 685 686

Williams, A.P., Funk, C., Michaelsen, J., Rauscher, S.A., Robertson, I., Wils, T.H., Koprowski, M., Eshetu, Z., Loader, N.J., 2012. Recent summer precipitation trends in the Greater Horn of Africa and the emerging role of Indian Ocean sea surface temperature. Climate Dynamics, 1-22.

687

Wyrtki, K., 1973. An Equatorial Jet in the Indian Ocean. Science 181, 262-264.

688 689

Xie, S.-P., Carton, J.A., 2004. Tropical Atlantic variability: Patterns, mechanisms, and impacts. Geophysical Monograph Series 147, 121-142.

690 691

Xie, S.P., Annamalai, H., Schott, F.A., McCreary Jr, J.P., 2002. Structure and mechanisms of south indian ocean climate variability*. Journal of Climate 15, 864-878.

692 693

Yamagata, T., Behera, S.K., Luo, J.J., Masson, S., Jury, M.R., Rao, S.A., 2004. Coupled ocean‐atmosphere variability in the tropical Indian Ocean. Earth's Climate, 189-211.

694 695 696

Zoubir, A.M., 1993. Bootstrap: theory and applications, pp. 216-235.

697 698 699

700

Figure 1: Central Africa, Indian and Atlantic Ocean basins and associated zonal climate patterns. Arrows signify

701

zonal and vertical wind vectors. (a) Walker Circulation associated with Indian Ocean Dipole (IOD) events, and

702

enhanced East African Monsoon precipitation. (b) Zonal circulation patterns for the West African Monsoon. (c)

703

Distribution of climate zones included in this analysis. Red boxes delineate Eastern Atlantic (EAO; 5°W-10°E by

704

5°N-10°S), Western Indian (WIO; 50-70°E by 10°N-10°S) and Eastern Indian oceans (EIO; 90°E-110°E by 0°-

705

10°S), from (Saji et al., 1999). The Lake Turkana Basin represented by red border. Intertropical Inter-Convergence

706

Zone (ITCZ) and Congo Air Boundary (CAB) represented by brown lines. Adapted from (Nicholson, 1996; Webster

707

et al., 1999; Williams and Funk, 2011).

708 709

Figure 2: Elevation and watershed limits of the Lake Turkana Basin (thick red lines) and discussed locations. Major

710

rivers are delimited (blue lines), sub-basins (thin red lines) and zones of limited drainage (hatched fill). Drainage

711

boundaries and elevation calculated from Global Multi-resolution Terrain Elevation Data 2010 (GMTED2010).

712 713

Figure 3: Monthly time series used in statistical analyses between November 1993 and April 2013 for (a) Western

714

Indian (WIO), (b) Eastern Indian (EIO), (c) Monthly difference between Western and Eastern Indian Ocean (∆SST;

715

green line), which correlates significantly (r=0.93) with the Dipole Mode Index (DMI; orange line), (d) Eastern

716

Atlantic (EAO) sea surface temperature anomalies (SSTA), (e) monthly anomaly of lake level changes.

717 718

Figure 4: Lagged correlation analysis between (a) Western Indian Ocean and (b) Eastern Indian Ocean SSTAs and

719

lake level anomaly. The “short” rain and “long” rain periods refer to the SSTAs from October to January and March

720

to June, respectively, with the lake level time series appropriately lagged.

721 722

Figure 5: Results of Empirical Orthogonal Function (EOF) analysis of equatorial SST and rainfall anomalies for the

723

Lake Turkana Basin, and comparison with the time-series of lake level variability. Correlation of PCA 1 and lake

724

level anomaly, delayed by one month (see Fig. 4) for (a) the EAO, (b) Turkana Basin rainfall, and (c) the WIO;

725

spatial distribution of EOF 1, (d) for the EAO, (e) Turkana Basin rainfall and (f) the WIO.

726

727

Figure 6: Spatial correlation between gridded time series of rainfall anomaly for the Lake Turkana Basin and PCA1

728

of SSTAs for (a) the EAO and (b) the WIO. These time series have been smoothed with a 3-month moving average

729

to reveal annual associations. Significance shown by dotted black lines. For comparison, (c) primary and (d)

730

secondary spatial EOF modes are shown. Black lines for EOFs signify zero variability.

731 732

Figure 7: Linear regressions for the “short” rain season (October to January) between SSTAs and lake level

733

anomaly. Red diamonds represent average seasonal SSTA>0.25σ and blue diamonds are average seasonal

734

SSTA<0.25σ. Regression parameters and associated statistical metrics reflect Monte Carlo bootstrap sampling

735

(n=2000), mean and 1-σ error.

736 737

Figure 8: As in Figure 7, but for the “long” rain season, from March to June.

738 739

Figure 9: Comparison of model-derived and empirical rainfall for the Lake Turkana Basin by season and source

740

between 1993 and 2012. (a) Rainfall contribution from the “short” and “long” rain seasons as a percentage of yearly

741

total, for model-derived vs. empirical rainfall. Note reverse scale for “long” rains. (b) Absolute “long” rain

742

precipitation by year. (c) Absolute “short” rain precipitation by year. (d) Modeled rainfall associated with the WIO

743

(solid line) and the EAO (dotted line). (e) Correlation between empirical and modeled rainfall with monthly

744

resolution on a yearly and seasonal basis.

745 746

Figure 10: Comparison of instrumental records of SSTs and lake level variations for AD 1857 to 2012. (a) Input

747

SSTAs for the EAO (blue line) and WIO (red line) monthly model simulation. Grey boxes reflect periods of

748

adjustment of the SSTA timeseries: (i) between AD 1936 and 1946 the 2 year moving average is removed to

749

eliminate significant data artefacts (see text), and (ii) post 1955 linear adjustment removes a pronounced warming

750

trend. The thin black line reflects the 0.25-σ threshold. (b) Historic and simulated water levels for Lake Turkana.

751

Periods of lake level record denoted by numerals: (I) pre-AD 1888 from Nicholson (1988), (II) between 1888 and

752

1970 data and lake record from Butzer (1971), (III) between 1971 and 1992 reflects annually gauging until 1975,

753

intermittent measures until monthly gauging began in 1988 (Avery, 2010). (IV) post-November 1992, monthly data

754

from remote satellite measurement. Dashed lines represent periods of uncertain water level measures. Period IV

755

reflects our model calibration period. Red line based on input of monthly SSTAs and associated 1-σ error. Blue line

756

reflects input of annual SSTAs, offset by 10 m for clarity. Water levels are relative to the 2008 datum of 362 masl.

757

(c) Lake Naivasha water level record for AD 1857 to 2012. Combined instrumental record (black line; Stoof-

758

Leichsenring et al., 2011), lake core record (black dashed line; Verschuren, 2001) and inference from vegetated

759

shorelines (grey dotted line; Nicholson, 1988).

760 761 762

Table 1: Linear regression statistics used in modeling Input SSTs

Input SST period

W. Indian SSTAs (WIO)

"Short" rains (October to January) "Long" rains (March to June)

E. Atlantic SSTAs (EAO)

Delay in lake level response (months) 1 0

"Short" rains (October to January)

1

"Long" rains (March to June)

3

W. Indian (Annual) E. Atlantic (Annual) a Statistically significant at the 99% level (p < 0.005) b Statistically significant at the 90% level (p < 0.05) Annual SSTAs

763 764 765 766

Criteria for SST state

equation of best fit

r

R2

Relative model weighting

Positive Negative Positive Negative Positive Negative Positive Negative Positive Positive

Y = 1.3488*X – 0.8292 Y = 0.0623*X – 0.2958 Y = 0.2745*X – 0.0889 Y = 0.3517*X – 0.0787 Y = 0.707*X – 0.2736 Y = –0.0124*X – 0.3192 Y = 0.4231*X – 0.1593 Y = 0.1521*X – 0.0103 Y = 0.0902*X – 0.051 Y = 0.0314*X – 0.012

0.78a 0.08 0.18 0.28b 0.37b 0.02 0.17 0.18 -

0.61a ~0.01 ~0.03 0.08b 0.14b ~0 ~0.03 ~0.03 -

25.0% 50.0% 7.5% 17.5% 75.0% 25.0%

767 768 769 770 771 772

• • • • •

SST anomalies appear to correlate significantly with Lake Turkana water level Regressions of SST anomalies and lake level change augmented by Bootstrap sampling. Increases in lake level linked to positive Indian Ocean SST anomalies in boreal fall. Post ca. 1930 meter-scale variability is linked to variations in Indian Ocean SST. Historical drop of 15m may reflect reduced role of Atlantic-derived sources.