Content uploaded by Bastien Dieppois
Author content
All content in this area was uploaded by Bastien Dieppois on Jan 25, 2016
Content may be subject to copyright.
1 23
Theoretical and Applied Climatology
ISSN 0177-798X
Volume 121
Combined 1-2
Theor Appl Climatol (2015) 121:139-155
DOI 10.1007/s00704-014-1229-5
Low-frequency variability and zonal
contrast in Sahel rainfall and Atlantic sea
surface temperature teleconnections during
the last century
B.Dieppois, A.Durand, M.Fournier,
A.Diedhiou, B.Fontaine, N.Massei,
Z.Nouaceur & D.Sebag
1 23
Your article is protected by copyright and
all rights are held exclusively by Springer-
Verlag Wien. This e-offprint is for personal
use only and shall not be self-archived
in electronic repositories. If you wish to
self-archive your article, please use the
accepted manuscript version for posting on
your own website. You may further deposit
the accepted manuscript version in any
repository, provided it is only made publicly
available 12 months after official publication
or later and provided acknowledgement is
given to the original source of publication
and a link is inserted to the published article
on Springer's website. The link must be
accompanied by the following text: "The final
publication is available at link.springer.com”.
ORIGINAL PAPER
Low-frequency variability and zonal contrast in Sahel rainfall
and Atlantic sea surface temperature teleconnections
during the last century
B. Dieppois &A. Durand &M. Fournier &A. Diedhiou &
B. Fontaine &N. Massei &Z. Nouaceur &D. Sebag
Received: 3 February 2014 /Accepted: 13 July 2014 /Published online: 23 July 2014
#Springer-Verlag Wien 2014
Abstract This study systematically examines teleconnections
between Atlantic sea surface temperature (SST) and the west–
east distribution of Sahel rainfall throughout the twentieth
century, taking nonstationarity into account. Sahel rainfall
variability of six selected rain gauges displays three dominant
time scales: multi-decadal (>20 years), quasi-decadal (8–18
years) and interannual (2–8 years). Regarding their patterns
of low-frequency scales, three coherent Sahelian subregions
can be identified: the Atlantic Coast (Dakar), western–central
Sahel (Nioro and Mopti) and eastern Sahel (Niamey, Maradi,
Maine-Soroa). Cross-analyses combining spectral and multi-
variate analyses of 20 station-based data and West-African
gridded rainfall data statistically confirm dissimilarities
between the western and eastern Sahel. Western and eastern
Sahel rainfall data are correlated with SSTs from different
regions of the Atlantic Ocean, especially in the North and
tropical South Atlantic. As determined by wavelet coherence
and phase, in-phase relationship with North Atlantic SSTs
only occurs in wet periods and at the multi- and quasi-
decadal scales. This teleconnection depends on the time peri-
od and the time scale, displaying a NW–SE pattern, which
suggests nonuniform modulations of meridional displace-
ments of the Intertropical Convergence Zone (ITCZ). Tropical
South Atlantic SST variability is often related to opposite
patterns between the Gulf of Guinean Coast (in phase) and
Sahel region (out of phase).
1 Introduction
During the 1970s and 1980s, West Africa experienced a
pronounced precipitation deficit, with the development of
persistent droughts. This dry period was evident throughout
the Sahel region (10° N–18° N, 17° W–30° E), but also
extended to the Gulf of Guinea coast (Lebel et al. 2000). Le
Barbé et al. (2002) and Lebel and Ali (2009)showedthatthe
annual rainfall deficit of the 1970–1989 period, compared to
the previous 1950–1969 period, was about 190 mm year
−1
.
These changes appeared to be homogeneous throughout the
Sahel, but zonal contrasts have been clearly evident over the
last 25 years (Dai et al. 2004;Nicholson2005; Lebel and Ali
2009; Fontaine et al. 2011a), as well as in the projections of
future climate changes (Christensen et al. 2007; Fontaine et al.
2011b; Monerie et al. 2012). A west–east pattern of Sahel
rainfall variability therefore seems to be superimposed on the
more widely known meridional gradient of the annual Sahel
rainfall, in agreement with Janicot (1992) and Moron (1994).
Sea surface temperature (SST) forcing can be considered
the dominant driver of West-African rainfall variability (e.g.
Folland et al. 1986;Palmer1986;Gianninietal.2003;Luand
Delworth 2005), but few studies have discussed zonal con-
trasts of Sahel rainfall variability and their teleconnections
with SSTs. Ward (1998) was the first author to highlight
B. Dieppois (*):A. Durand :M. Fournier :N. Massei :D. Sebag
Laboratoire Morphodynamique Continentale et Côtière (M2C),
Université de Rouen, CNRS, FED 4116 SCALE, Mont-Saint
Aignan, France
e-mail: bastien.dieppois@gmail.com
A. Diedhiou
Laboratoire d’étude des Transferts en Hydrologie et Environnement
(LTHE), Université de Grenoble, IRD, Grenoble, France
B. Fontaine
Centre de Recherches en Climatologie (CRC), Université de
Bourgogne, CNRS, Dijon, France
Z. Nouaceur
Laboratoire Identité et Différenciation de l’Espace,
de l’Environnement et des Sociétés (IDEES), Université de Rouen,
CNRS, FED 4116 SCALE, Mont-Saint Aignan, France
D. Sebag
Laboratoire HydroSciences Montpellier (HSM),
Université de Montpellier 2, IRD, Montpellier, France
Theor Appl Climatol (2015) 121:139–155
DOI 10.1007/s00704-014-1229-5
Author's personal copy
stronger Indo-Pacific connections with eastern Sahel and
stronger North and tropical Atlantic connections with the
western Sahel. Subsequent studies highlight the role of Indian
and tropical Pacific SSTs (Bader and Latif 2003)orwarmer
Mediterranean SSTs in the west–east contrast in Sahel rainfall
(Fontaine et al. 2011c,2011d). The moisture supply of the
Sahel is, however, dependent on horizontal convergence of
moisture evaporated from the adjacent oceans, particularly the
tropical south Atlantic (Hagos and Cook 2008). Meanwhile,
Nieto et al. (2006) also highlight a band along the North
Atlantic from Sahel latitudes to the Iberian coasts as a major
source of Sahel moisture. This raises the question on the role
of Atlantic SSTs on Sahel rainfall zonal variability.
Atlantic SST–Sahel rainfall teleconnections may vary on
specific time scales, which in addition can be expressed during
specific time periods. Numerous studies (e.g. Folland et al.
1986;Fontaineetal.1998;Gianninietal.2003) attribute the
decadal drying phase in the Sahel in 1970s to an inter-
hemispheric temperature gradient in the global ocean basins
with a particularly strong signal in the Atlantic Ocean. As a
consequence of the latter, these inter-hemispheric SST fluctu-
ations are related to the Atlantic Multi-decadal Oscillation
(AMO), a coherent pattern of multi-decadal variability in
SSTs centred on the North Atlantic Ocean. The AMO has
been linked to the occurrence of Sahel droughts (Sutton and
Hodson 2005) and the recent partial rainfall recovery in the
Sahel (Hagos and Cook 2008; Fontaine et al. 2011d;Mohino
et al. 2011).
At the interannual time scale, the equatorial SST pattern,
characterised by SST anomalies in the cold tongue region (6°
S–2° N, 20° W–5° E), is considered to have a significant
impact on the West-African monsoon system. Colder SSTs
in the tropical South Atlantic are associated with wetter con-
ditions in the Sahelian regions and dryer conditions in the
Guinean coastal regions (Lamb 1978;Janicot1992;Rowell
et al. 1995; Ward 1998). This is due to a northward displace-
ment of the Intertropical Convergence Zone (ITCZ) (Polo
et al. 2008; Losada et al. 2010). Such SST anomalies over
the tropical South Atlantic have been linked to the Atlantic
Niño mode, which sometimes can be induced by El Niño–
Southern Oscillation (ENSO) (Servain et al. 1982; Latif and
Grotzner 2000). This teleconnection between Atlantic and
Pacific tropical SSTs is, however, nonstationary, which can
lead to changes in the response over West Africa (e.g. Janicot
et al. 2001;Poloetal.2008;Wangetal.2009; Rodriguez-
Fonseca et al. 2011). According to the latter, since the late
1970s, Pacific Niño events determined the Sahelian rainfall
anomaly, overriding the Sahel rainfall enhancement signal of
the concurrent Atlantic Niño event. Thus, rainfall anomalies
of the same sign occur in the Sahel and along the Gulf of
Guinea Coast due to the adjacent cold Atlantic waters.
The aim of this work is to explore the changing character-
istics of Sahel rainfall variability and their relationships with
Atlantic SST from eastern to western Sahelian regions. Using
a time-space approach based on spectral analysis of both
station-based and gridded hydroclimatic data, we examine
the potential nonstationary behaviour of such relationships
throughout the twentieth century. In Section 2,wediscuss
data and methods, after which we analyse time-scale patterns
of Sahel rainfall variability from eastern to western regions in
Section 3. These analyses, which were first performed on 20
rain gauges, are summarised using six selected rain gauges
which appear to be less sensitive to instrumental and digitiza-
tion errors or to local-scale forcing (cf. Section 3.1). The
robustness of results obtained, however, is validated using a
larger sampling of rain gauges or using homogeneous gridded
data. Following the same cross-checked process in Section 4,
we statistically investigate the potential influence of Atlantic
SSTs in driving zonal contrasts in Sahel rainfall variability
using composite analysis and bivariate wavelet analysis. Fi-
nally, the results are summarised and discussed in Section 5.
2 Data and methods
2.1 Data
The monthly precipitation amounts used were obtained from
the NOAA/NCDC Global Historical Climate Network data-
base (GHCN-2). After the 1990s, the GHCN-2 database was
completed with database of the National Meteorological De-
partment of Senegal, Mali and Niger. To investigate local
evolutions of low-frequency time scales of rainfall variability
from the Atlantic Coast to Lake Chad, only six selected
Sahelian weather stations were selected (Table 1): Dakar
(14.7° N, 17.5° W, 24 m), Nioro du Sahel (15.2° N, 9.4° W,
237 m), Mopti (14.5° N, 4.2° W, 271 m), Niamey (13.5° N,
2.1° E, 234 m), Maradi (13.5° N, 7.1° E, 368 m) and Maine-
Soroa (13.2° N, 12.0° E, 338 m). We note that this selection
has been retained following a preliminary study examining the
spectral contents of 20 rain gauges over the Sahel region
(Table 1). The spatial coherence of the selected rain gauges
nevertheless will be examined in Section 3.1. In addition,
high-resolution grids from the CRU TS 3.10.1 monthly data
sets (1901–2009) were selected to compare and validate the
results from station-based time series through spatial analysis
of Sahel rainfall variability (more detailed explanations on the
CRU TS 3.10.1 data are available at badc.nerc.ac.uk/view).
The SST gridded data (2° latitude–longitude grid) is ex-
tracted over the Atlantic basin (60° S–70° N; 65° W–30° E)
from the extended reconstructed sea surface temperature
(ERSST-V3b) of the National Climatic Data Centre. The
ERSST-V3b is generated using in situ SSTs (ICOADS, Inter-
national Comprehensive Ocean-Atmosphere Data Set) and
improved statistical methods that allow stable reconstruction
(Smith et al. 2008). Two selected key SST indices were
140 B. Dieppois et al.
Author's personal copy
computed using the ERSST-V3b data set: the Atlantic Multi-
decadal Oscillation (AMO) index derived from van
Oldenborgh et al. (2009) and the tropical South Atlantic
(TSA) index as computed by Enfield et al. (1999). The
AMO index is defined as the SST averaged over 25°–60° N,
7°–70° W. The TSA index is the average of the monthly SST
from 0–20° S and 10° E–30° W.
2.2 Methodology
2.2.1 Rainfall analysis
Any rain gauge contains some stochastic noise due to either
instrumental and digitization errors and/or local-scale forcing,
but can be viewed as a real measure of rainfall measurements
at a given station. By contrast, any gridded rainfall field, even
based on rain gauge observations, results from statistical
methods and field analyses performed on the available net-
work, and, in these regions, is often performed on a limited
number of stations. Discrete measurements in space therefore
tend to create artificial spatial coherence of regional or conti-
nental extent, depending on gauge density and the statistical
methods used.
To assess the spatial coherence of each Sahelian time
series, a preliminary study is therefore performed in
Section 3.1. The spatial coherence of interannual precipitation
anomalies is first estimated empirically using pointwise cor-
relation with the CRU TS 3.10.1 rainfall fields. The pvalues at
the 95 % confidence level (i.e. p=0.05) are computed accord-
ing to the Student’sttest after recalculating the degrees of
freedom with the estimated decorrelation scales. The fraction
of maps with p<0.05 is estimated over the Sahel region (grey
dashed lines in Fig. 1). The stochastic noise can also disturb
the phase of any time scale of the regional-scale variability.
Therefore, we have computed the correlation between each
station-based time series and spatial indices (of ~222 km
2
)
centred on each rain gauge (grey boxes in Fig. 1) with a
significance level based on a random-phase test (Ebisuzaki
1997). Briefly, for each pair of time series to be correlated,
1,000 random time series having the same power spectrum as
the original one but with random phases are correlated.
In Section 3.2, time-scale exploration is pursued using
continuous wavelet transform (CWT) in order to visualise
potential changes in the spectral content of the analysed sig-
nal. By decomposing the time series into time-frequency
space, it can be determined which frequencies are dominant
variability modes affecting Sahel rainfall at certain time inter-
vals. Such decomposition of the monthly rainfall signals is
performed through a non-orthogonal set of wavelets (Morlet
wavelet, order 6) to produce the local wavelet spectra. De-
tailed explanations of CWT methodology and its application
to climatic signals are now widely documented in the literature
(Torrence and Compo 1998;Maraun2006). The significance
test of the wavelet spectrum assumes a red-noise background
spectrum for the null hypothesis. The wavelet spectrum is
tested for every point in time and scale to check whether the
power exceeded a certain critical value determined by 1,000
Monte Carlo simulations of first-order auto-regressive
(AR[1]) processes.
To objectively and statistically identify coherent spatial
patterns of time-scale fluctuations in Sahelian rain gauges, a
multivariate analysis of wavelet spectra is performed (Rouyer
et al. 2008). This method, which is based on the covariance of
each pair of wavelet spectrum, allows us to quantify and
summarise the dissimilarities between eastern and western
rain gauges by hierarchical clustering. Meanwhile, a red-
noise test is performed for defining the best number of clus-
ters. Such red-noise test at p=0.05 and 0.1 is determined
through the dissimilarity between two sets of random time-
series having the same one-order autocorrelation, mean and
Tabl e 1 Detailed information and geographical location of Sahelian rain gauges studied
Sahelian rain gauges
Stations Country Coordinates Record period Stations Country Coordinates Record period
Dakar Senegal 14.7 N, 17.5 W, 24 m 1897–2010 Hombori Mali 15.9 N, 1.7 W, 287 m 1930–2002
Mbour Senegal 14.4 N, 17 W, 0 m 1931–2009 Tillaberi Niger 14.2 N, 1.5 E, 209 m 1923–1997
St Louis Senegal 16.1 N, 16.5 W, 4 m 1854–2010 Niamey Niger 13.5 N, 2.1 E, 234 m 1905–2010
Kaolack Senegal 14.9 N, 16.07 W, 25 m 1921–2003 Tahoua Niger 14.9 N, 5.3 E, 386 m 1922–2000
Linguère Senegal 15.4 N, 15.1 W, 21 m 1934–2003 Maradi Niger 13.5 N, 7.1 E, 368 m 1931–2005
Podor Senegal 16.6 N, 14.9 W, 7 m 1918–2003 Magaria Niger 13 N, 8.9 E, 360 m 1938–1998
Matam Senegal 15.6 N, 13.3 W, 15 m 1922–2003 Zinder Niger 13.8 N, 9 E, 451 m 1905–2002
Nioro Mali 15.2 N, 9.4 W, 237 m 1919–2001 Goure Niger 14 N, 10.3 E, 460 m 1936–1996
Segou Mali 13.4 N, 6.2 W, 288 m 1933–2000 Maine-Soroa Niger 13.2 N, 12 E, 338 m 1936–2005
Mopti Mali 14.5 N, 4.2 W, 271 m 1921–2010 Maiduguri Nigeria 11.9 N, 13.1 E, 354 m 1909–1995
Weather stations in italics indicate the six selected rain gauges to summarise the results
Low-frequency variability and zonal contrast in Sahel rainfall 141
Author's personal copy
variance as western and eastern Sahel rainfall time series. We
then check whether the best number of clusters is similar using
6or20raingauges.
In Section 3.2, the results based on station-based time
series are then checked through the CRU TS. 3.10.1 field,
which is less sensitive to stochastic noise, using band-pass
filters according to the significant time scale of the CWT. Any
spatial differences in mean and variance for all points (x)inthe
rainfall field were removed by the following standardisation: x
−mean(x)/sd(x). First, power Hovmöller diagrams (time-
latitude plot) allow us to examine the spatiotemporal variabil-
ity of West-African Sahel rainfall (Torrence and Compo
1998). Here, the power Hovmöller diagram is a scale average
of the wavelet power spectra at multiple longitudes around
12–15° N. Second, to broaden our understanding on a wider
scale, we calculate the difference in variance distribution of
each time scale of rainfall variability between the wet (1948–
1965) and dry (1972–1989) periods throughout West Africa.
2.2.2 Atlantic sea surface temperature and Sahel rainfall
relationship
In Section 4.1, composite analysis is performed to construct
multi-decadal, quasi-decadal and interannual typical states of
Atlantic SSTs conditional on the rainfall value time series
from the Atlantic Coast to Lake Chad. In order to discuss
the findings at several time scales, all data sets were filtered
over multi-decadal (>20 years), quasi-decadal (9–19 years)
and interannual (2–8 years) scales by fast Fourier transform
(FFT). The pvalues at the 90 % confidence level are comput-
ed assuming normal distributions after recalculating the de-
grees of freedom with the estimated decorrelation scales.
In Sections 4.2 and 4.3, the time-frequency stability of the
relationships between Sahel rainfall time series and gridded
CRU TS 3.10.1 data sets with selected key Atlantic SST indices
(AMO and TSA) is examined using wavelet coherence and
phase. In brief, wavelet coherence assesses the linear relation-
ship between two temporal signals x(t)andy(t)aroundtimet
i
and scale (Maraun 2006), with the phase diagram describing the
time lag between the signal at any time-scale location. Similar
procedures to those computed for CWT are conducted to test
covarying power (i.e. 1,000 Monte Carlo simulations of two
independent AR[1] processes). This approach has already been
successfully applied in many climate analyses (Torrence and
Web ster 1999; Maraun and Kurths 2004). The results based on
station-based time series are then checked through the CRU TS.
3.10.1 field following the same procedure as for CWT analysis.
3 Sahel rainfalls during the last century
3.1 Spatial coherence of Sahel rainfall time series
Our first hypothesis is that the year-to-year fluctuations of
monthly rainfall amounts at each station should be integrated
into a larger-scale anomaly pattern over the Sahel. This is to
assess the robustness of time series in addressing local-scale
forcing, instrumental and digitization errors. Figure 1displays
pointwise correlation between annual rainfall anomalies at each
station with the CRU TS 3.10.1 field focused on West Africa.
Statistically, there is almost a significant correlation in the maps
(pfield< 0.1 %; Fig. 1). The year-to-year variations of all rain
gauges are significantly correlated with a broad band over the
Sahel (p<0.05 % between 69.68 and 94.92 %; Fig. 1). The
Fig. 1 a–fPointwise correlations between annual rainfall anomalies of
each Sahelian rain gauge and of CRU TS 3.10.1 rainfall field over West
Africa. Only correlations significant at the 95 % confidence level
(p<5 %) are displayed. The field significance (in black), the fraction of
Sahel regions (grey dashed lines)withp< 5 % and the correlation of each
Sahel rainfall time series with spatial indices (~222 km
2
) centred on each
rain gauges (light grey boxes) based on random-phase test (in light grey)
are indicated on the lower-right corner
142 B. Dieppois et al.
Author's personal copy
highest values of correlations are observed in the surrounding
areas, which are shifted eastward from Dakar to Maine-Soroa
(Fig. 1). Secondly, correlations (r) with spatial indices of
~222 km
2
centred on each time-series using a random-phase
test are always greater than 0.69 and significantly above the
one-sided level of significance (p) (Fig. 1). Therefore, the
phases of fluctuations in each Sahel rainfall time series are
consistent with nearby regional climatic fluctuations registered
in the gridded data set. We, however, note that the weakest
spatial coherences of annual rainfall time series are detected at
Nioro and Mopti. High spatial autocorrelation of the CRU TS
3.10.1 field is also clearly observed in Fig. 1.
These six Sahelian rain gauges therefore can be used to
investigate the Sahel rainfall variability at the regional scale
(i.e. not too local or large).
3.2 Time-scale fluctuations of Sahel rainfall
3.2.1 Analysing rainfall time series
Several significant time scales of Sahel rainfall variability are
identified by CWT: seasonal (between 6 months and 1 year),
interannual (between 2–3and5–8 years), quasi-decadal (be-
tween 8–14 and 12–18 years) and multi-decadal (between 19–
30 and 16–36 years) (Fig. 2). Seasonal cycle exhibits an
annual contrast between dry (Nov–Mar) and wet (Apr–Oct)
seasons. During dry years, this contrast is therefore lower,
henceadecreaseinvarianceofseasonalscales(Fig.2). The
annual cycle accounts for 41.7 and 56.1 % of the average total
monthly variance, while all the low-frequency time scales of
variability are between 1.7 and 3.6 % (Table 2). Annual data
shows that the interannual scale of Sahel rainfall variability is
dominant (40.5–77.1 %, Table 2). A nonstationary behaviour
nevertheless is clearly identified at the multi-decadal to inter-
annual scales (Fig. 2).
To statistically identify coherent spatial patterns in the
low-frequency fluctuations (>2 years) of Sahel rainfall var-
iability, each pair of wavelet spectra is compared by using
hierarchical clustering (Fig. 3a). Only two clusters are sig-
nificantly above the dissimilarity between two sets of red
noise at p=0.05. The wavelet spectrum of Dakar rainfall is
significantly dissociated from those of Nioro and Mopti,
reducing the statistical significance at p=0.1. Thus, based
on six selected rain gauges, three patterns of time-scale
Fig. 2 Wavelet power spectrum of the monthly Sahel rainfall amount (a
Dakar, bNioro du Sahel, cMopti, dNiamey, eMaradi and fMaine-
Soroa). Bold line (the so-called cone of influence) delineates the area
under which power can be underestimated as a consequence of edge
effects, wraparound effects and zero padding; thin and dashed contour
lines show the 90 % confidence limits based on 1,000 Monte Carlo
simulations of the red-noise background spectrum before and after re-
moval of the seasonal cycle, respectively
Low-frequency variability and zonal contrast in Sahel rainfall 143
Author's personal copy
fluctuations are located over three Sahel subregions (Figs. 2
and 3a): the Atlantic Coast (Dakar), western–central Sahel
(Nioro and Mopti) and eastern Sahel (Niamey, Maradi,
Maine-Soroa). These can be depicted as follows:
(i) Over the Atlantic Coast (Dakar, Fig. 2a), a high variance
of interannual scales (2–3to5–8 years) of variability is
significantly detected during wet anomalies, which is
marked by the high amplitude of seasonal cycles. During
dry anomalies, only the multi-decadal time scale (19–30
years) is identified, being significantly above the red-
noise background spectrum.
(ii) Over western–central Sahel (Nioro and Mopti,
Fig. 2b, c), overall low-frequency scales of variability
(interannual, between 2–3and5–8 years; quasi-decadal,
10–16 years; multi-decadal, 16–36 years) show high
variance, which is significantly above the red-noise
background, during wet anomalies.
Tabl e 2 Contribution of variability scales detected in monthly (annual) Sahel rainfalls, expressed as a percentage of total variance captured by the FFT
filter
% Average contribution of the total variance
Variability scale Dakar Nioro du Sahel Mopti Niamey Maradi Maine-Soroa
Seasonal, monthly (1 year) 41.7 51.5 56.1 56 52.6 45.1
Interannual, monthly (annual) (2–8 years) 2.3 (73.9) 1.4 (40.5) 1.8 (77.1) 1.7 (64.5) 1.3 (73) 2.2 (73)
Quasi-decadal, monthly (annual) (9–19 years) 0.4 (12.1) 0.3 (7.9) 0.2 (9.5) 0.2 (8.7) 0.3 (8.3) 0.2 (9.6)
Multi-decadal, monthly (annual) (>20 years) 0.1 (4.5) 0.9 (22.8) 0.2 (7.9) 0.2 (6.1) 0 (1.8)
a
0(1.8)
a
Low-frequency, monthly (>2 years) 3 3.6 2.2 2.6 1.7 2.6
a
Time s erie s too sh ort
Fig. 3 a Cluster tree of the
wavelet spectra for six Sahel
rainfall time series. bThe same
analysis for 20 Sahel rainfall time
series. The cluster tree was
obtained by using the
dissimilarity matrix on which
flexible clustering was applied.
Red dashed line displays the
statistical significance at the 95
and 90 % confidence limits
determined through the
dissimilarity between two sets of
random time series having the
same one-order autocorrelation,
mean and variance as western and
eastern Sahel rainfall time series
144 B. Dieppois et al.
Author's personal copy
(iii) Over eastern Sahel (Niamey, Maradi, Maine-Soroa,
Fig. 2d–f), the interannual variability (between 2–3
and 5–8 years) significantly increases during wet anom-
alies. During dry anomalies, only the quasi-decadal time
scales (8 or 12–18 years) is significantly identified. We
note that the multi-decadal time scales (16–36 years)
can only be detected in Niamey rainfall (other time
series being too short).
Cross-analysis based on wavelet clustering of 20 rain
gauges, however, only shows the differences between western
and eastern Sahel (Fig. 3b). Nevertheless, examining the
individual wavelet spectra of 20 rain gauges, the patterns of
time-scale fluctuations observed at Dakar seem only detected
in the coastal stations, i.e. Mbour and St Louis (not shown).
3.2.2 Multi- to interannual time scales in the West-African
rainfall field
To compare with the results from six rainfall time series, the
time-space fluctuations of each scale of variability are ex-
plored through a standardised rainfall field using band-pass
filtering optimising the captured variance of each sector: at the
multi-decadal (>20 years), quasi-decadal (8–18 years) and
interannual (2–8years)timescales.
At the multi-decadal time scale, the power Hovmöller
diagram around 12–15° N displays stronger multi-decadal
variability between 17.5° and 8° W (Fig. 4(a.1)). This scale,
which is identified throughout the Sahel, shows a decrease
since the mid-twentieth century (Fig. 4(a.1)). As determined
by the difference of multi-decadal variance between wet and
dry periods (Fig. 4(a.2)), the decrease of the multi-decadal
variability is almost generalised over West Africa.
At the quasi-decadal time scale, according to the power
Hovmöller diagram, 5° W–0° longitude seems to separate the
western and eastern Sahel, especially between the 1950s wet
anomalies and the 1970s–1980s dry anomalies (Fig. 4(b.1)).
The difference of quasi-decadal variance between wet and dry
periods clearly shows E–W separations (Fig. 4(b.2)). Through-
out West Africa, we note a decrease of quasi-decadal variability
on the western part (including Sudano-Guinean zone), while an
increase is detected over the eastern Sahel (Fig. 4(b.2)).
The interannual variability is more important in the western
Sahel (Fig. 4(c.1)). The 1950s wet anomaly is marked by a
longitudinal expansion of this variability, which therefore is
detected over the eastern Sahel (Fig. 4(c.1)). As for the multi-
decadal scale, interannual variability decreases throughout
West Africa between wet and dry periods (Fig. 4(c.2)).
According to the wavelet hierarchical clustering of 20 rain
gauges, patterns of variability in the West-African field dis-
play contrasts between the eastern and western regions. Thus,
only the difference between western and eastern Sahel rainfall
time series will be discussed in the following sections.
4 Multi-scale relationships between Atlantic SSTs
and E–W variability of Sahel rainfall
4.1 Atlantic SST composite fields
Typical states of Atlantic SSTs are constructed using compos-
ite analysis based on fluctuations of rainfall time series located
over the eastern to western Sahel from the multi-decadal to
interannual time scales (Fig. 5).
At the multi-decadal scales (>20 years), the inter-
hemispheric SST anomalies are observed (Fig. 5(a)).
Multi-decadal wet anomalies are associated with warm
SSTs in the North Atlantic and cold temperatures in the
South Atlantic (Fig. 5(a)). The statistical relationship
between Sahel rainfall and South Atlantic SSTs is less
robust in Niamey and Maine-Soroa (Figs. 5(a.5) and 6).
Most of North Atlantic SST composite fields display a
cold core in south-east Newfoundland surrounded by
warmer SSTs in the subtropics and the eastern Atlantic
basin (Fig. 5(a)).
The quasi-decadal composite fields (9–19 years) are
more complex (Fig. 5(b)). Generally, warmer North Atlan-
tic SSTs and colder tropical South Atlantic (TSA) SSTs are
identified (Fig. 5(b)). The statistical relationship between
the tropical South Atlantic SSTs and Sahel rainfall is more
robust over eastern Sahel (Fig. 5(b)). Moreover, the TSA-
Sahel rainfall teleconnection patterns are different in eastern
and western regions (Fig. 5(b)). The quasi-decadal signals
of TSA SSTs are detected over the Gulf of Guinea for
Nioro and Mopti rainfall in western Sahel (Fig. 5(b.2–3)),
while it is located over the Brazilian coast for Niamey,
Maradi and Maine-Soroa rainfall over the eastern Sahel
(Fig. 5(b.4–6)).
At the interannual scale (2–8 years), Sahel wet anom-
alies are associated with (Fig. 5(c)): (i) patterns of colder
TSA SSTs, which refer to the equatorial SST pattern (or
Atlantic Niño) and (ii) North Atlantic SST anomalies.
Statistical links with North Atlantic SST, however, are
weak.
Relationships between rainfall measurements at gauges
located in West-African Sahel rainfall and Atlantic SSTs
are different for each time scale. Sahel Atlantic
teleconnections are sensitive to time-scale fluctuations and
to the east–west gradient. Therefore, Atlantic SST variabil-
ity may play, at least partially, an influence on zonal
contrasts in Sahel rainfall measured at the station. These
“typical”states of Atlantic SSTs, however, could also vary
in time, and thus may independently be detected in the
northern and southern regions of the Atlantic basin. We
note, however, that differences in the number of assimilate
data between the North and South Atlantic regions might
affect the results (more detailed explanations at http://
icoads.noaa.gov/r2.html).
Low-frequency variability and zonal contrast in Sahel rainfall 145
Author's personal copy
4.2 Eastern and western Sahel vs. AMO index
According to earlier studies (Knight et al. 2006; Zhang and
Delworth 2006;Mohinoetal.2011), increasing rainfall over
the Sahel is related to the positive phase of the AMO, due to a
northward shift of the ITCZ. In other words, the relationship
between Sahel rainfall and North Atlantic SSTs is in phase.
The time-frequency stability and its variability from western
to eastern Sahel are evaluated in the following section using
wavelet coherence and phase.
4.2.1 Analysing rainfall time series
The AMO index and Sahel rainfall are coherent at the multi-
and quasi-decadal time scales from western Sahel up to
Niamey (Fig. 6(a.1–d.1)). Conversely, the spectral coherence
Fig. 4 Left:(1) Power Hovmöller diagram over the Sahel regions (12–
15° N) at the multi-decadal (a), quasi-decadal (b), and interannual (c)
scales of rainfall variability. Right:(2) Variance (power) difference
between dry (1972–1989) and wet (1948–1965) periods for each scale
of Sahel rainfall variability (a–c)
146 B. Dieppois et al.
Author's personal copy
is generally low: (i) at the interannual time scale and (ii)
using the easternmost Sahelian rain gauges, i.e. in Maradi
and Maine-Soroa (Fig. 6(e.1–f.1)). The phase (i.e. time lag)
of the AMO-Sahel rainfall teleconnection does not appear to
be sensitive to its time scales and is most often in phase (i.e.
more rainfall when the North Atlantic SSTs are warmer;
Fig. 6).
The multi-decadal variability of Dakar rainfall (19–30
years) is coherent and in opposition of phase with the AMO
index since the 1970s (i.e. more rainfall when the North
Atlantic SSTs are colder; Fig. 6(a)). Nioro, Mopti and Niamey
rainfall displays significant coherence with the AMO index
before 1970s, but the phase is difficult to interpret from just
three rain gauges (Fig. 6(b–d)).
At the quasi-decadal scale, Dakar rainfall is significantly
coherent and in phase with the AMO index before the 1920s
and since the 1940s (Fig. 6(a)). We note that this second
period is highly sensitive to red noise and does not appear
significant in the CWT of Dakar rainfall (Fig. 2a). Quasi-
decadal coherences with the AMO index are also significantly
detected between 1930 and 1980 in Nioro, Mopti and Niamey
rainfall (Fig. 6(b–d)).
Our first time-scale exploration of AMO index-Sahel rain-
fall teleconnection through six time series is consistent with
east–west contrasts. The statistical links to the AMO index
appear to decrease from the Atlantic Coast to the easternmost
regions of West African Sahel.
4.2.2 Analysing the West-African rainfall field
Wavelet coherence and phase between AMO index and the
West-African rainfall field at the multi- (>20 years) and quasi-
decadal (8–18 years) time scales are displayed in Fig. 7.
At the multi-decadal time scale, the AMO index is signif-
icantly coherent with the Sahelian rainfall field during two
periods (Fig. 7(a.1)): (i) before 1970 over the western regions
and (ii) after 1990 over the eastern regions. These multi-
decadal coherence patterns are often in phase over the Sahel
according to the time-series analysis. The time lag is high and
often appears in quadrature of phase (white shaded;
Fig. 7(a.1)). This teleconnection is not restricted to the Sahel
(Fig. 7(a.2–3)). Before 1950, it occurs throughout West Afri-
ca, from the western Sahel regions between 18.5° and 6° Wup
to Cameroon (Fig. 7(a.2)). After 1970, the multi-decadal
coherence pattern of AMO, which is always distributed ac-
cording to a NW–SE plan, is significantly extended to eastern
Sahel (Fig. 7(a.3)). Moreover, during this period, the multi-
decadal coherence patterns of the AMO index display a dipole
Fig. 5 (a) Multi-decadal (>20 years) Atlantic SST composite fields
associated with positive anomalies (above standard deviation) of Sahel
rainfall multi-decadal variability from the Atlantic Coast to Lake Chad
basin (1, Dakar; 2, Nioro du Sahel; 3,Mopti;4,Niamey;5,Maradi;6,
Maine-Soroa) using FFT filtering. (b–c) Idem for the quasi-decadal (9–18
years) and interannual (2–8 years) scales, respectively. Grey dashed
contour lines denote differences significant at the 95 % confidence level
according to the Student’sttest after recalculating the degrees of freedom
with the estimated decorrelation scales
Low-frequency variability and zonal contrast in Sahel rainfall 147
Author's personal copy
structure marked by an in-phase and out-of-phase relationship
with rainfall, respectively, over the eastern Sahel and the
western part of West Africa (Fig. 7(a.3)). We note that the
occurrence of the out-of-phase relationship over the western
part of West Africa is consistent with the results of AMO-
Dakar wavelet phase spectrum (Figs. 6(a.2) and 7(a.3)).
Quasi-decadal variability of the AMO index is significantly
coherent and in phase with the Sahelian rainfall field during
three periods (Fig. 7(b.1)): (i) at the beginning of the twentieth
century (i.e. 1901–1920), (ii) between 1940 and 1960 over the
central Sahel and (iii) up to the eastern Sahel in the 1970s.
Observed quasi-decadal coherence pattern between 1940 and
1960 over the central Sahel is in good agreement with the
time-series analysis (Figs. 6(b–d) and 7(b1)). As for the multi-
decadal time scale, this teleconnection is detected over a large
part of West Africa, according to a NE/SW plan, from the
western Sahel regions between 18.5° and 2° E up to Camer-
oon (Fig. 7(b.2)). In the 1970s, the quasi-decadal coherence
pattern appears restricted to the Sahelian region, but is rarely
significantly detected (Fig. 7(b.3)).
In summary, teleconnections between Sahel rainfall and
North Atlantic SSTs (AMO index) are significantly detected
at the multi- and quasi-decadal time scales. During the 1950s
wet anomaly, fluctuations of North Atlantic SSTs contribute to
an increase of multi- and quasi-decadal variability of western
Sahel rainfall (up to 6° W/2° E), as well as for the western part
of West Africa. This statistical relationship appears weaker
during the 1970s dry anomalies. Since the 1990s, the NE/SW
teleconnection pattern seems to be shifted northward and is
detected over the eastern Sahel. According to Mohino et al.
Fig. 6 Squared wavelet coherence (1) and phase (2) between monthly
Sahel rainfall amount (aDakar, bNioro du Sahel, cMopti, dNiamey, e
Maradi, and fMaine-Soroa) and the AMO index (van Oldenborgh et al.,
2009:25°–60° N; 7°–70° W). Bold line delineates the area under which
power can be underestimated as a consequence of edge effects, wrap-
around effects and zero padding; thin contour lines show the 90 %
confidence limits based on 1,000 Monte Carlo simulations of two inde-
pendent AR[1] processes
148 B. Dieppois et al.
Author's personal copy
(2011), the multi-decadal variability of North Atlantic SSTs
can contribute to the recent partial recovery of eastern Sahel
rainfall through meridional displacements of the ITCZ. Nev-
ertheless, our findings suggest that the thermal influence of
North Atlantic SSTs is associated with tilted displacements
along a NE/SW plan, which may contribute to zonal contrasts
in Sahel rainfall variability.
4.3 Eastern and western Sahel rainfall vs. TSA index
The interannual relationship between tropical South Atlantic
SST variability (TSA index) and Sahel rainfall is nonstation-
ary due to increased influence of ENSO variability since the
1970s (Janicot et al. 2001). According to Losada et al. (2010),
SST anomalies restricted to the tropical South Atlantic result
in rainfall anomalies of the same sign over the Gulf of Guinea
Coast and of the opposite sign over the Sahel. This
teleconnection, therefore, is characterised by an out-of-phase
relationship over the Sahel. Accordingly, we have evaluated
its time-frequency stability along an E–Wgradient.
4.3.1 Analysing rainfall time series
Figure 8displays squared wavelet coherence and phase be-
tween TSA index and Sahel rainfall time series. At the multi-
decadal time scale, Dakar (19–30 years; Fig. 8(a)), Nioro and
Mopti (16–36 years; Fig. 8(b, c)) rainfall and TSA index show
significant out-of-phase coherence patches before 1950s (i.e.
more western Sahel rainfall when the TSA SSTs are colder).
Multi-decadal variability of Niamey, Maradi and Maine-Soroa
rainfall often is not coherent to the TSA index and is in
quadrature of phase (i.e. large time lag; Fig. 8(d–f)). These
results are consistent with the multi-decadal Atlantic SST
composite fields and show less robust statistical links with
South Atlantic SSTs over the eastern Sahel (Fig. 5(a)).
At the quasi-decadal scale, the 12–18-year variability of
eastern Sahel rainfall (Niamey, Maradi, Maine-Soroa) display
out-of-phase coherence patches with the TSA index since the
1970s (Fig. 8(d–f)). Dakar, Nioro and Mopti rainfall variabil-
ity (8–14 and 10–16 years, respectively) does not present
coherence patches at the quasi-decadal scale, except at the
beginning of the twentieth century (Fig. 8(a–c)). The differ-
ences of Atlantic SST composite fields between gauges locat-
ed in western and eastern Sahel (Fig. 5(b)) therefore could
reflect changes in time.
At the interannual scale (2–8 years), Sahel rainfall displays
several out-of-phase coherence patches with the TSA index
(Fig. 8). This relationship, however, appears too intermittent,
and no zonal difference is detected from the analysis of time
series.
Fig. 7 To p:(1) Wavelet coherence and phase between AMO index and
Sahel rainfall field using Hovmöller diagrams over the Sahel regions (12–
15° N) at the multi- (>20 years; a) and quasi-decadal (9–18 years; b)
scales of co-variability. Middle and bottom:(2–3) Wavelet coherence and
phase between AMO index and West-African rainfall when multi- (a)and
quasi-decadal (b) coherence patterns are significantly detected over the
Sahel (20-year averages of this period is performed). For each scale, the
right and left columns display the wavelet coherence level using colour
scale (from blue to red) and wavelet phase (blue, in-phase relationship;
red, out-of-phase relationship), respectively. Thin contour lines refer to
significant coherence performed by Monte Carlo simulations at the 90 %
confidence limits
Low-frequency variability and zonal contrast in Sahel rainfall 149
Author's personal copy
Statistical relationships between TSA index and Sahel
rainfall reveal differences between gauge estimates in western
and eastern Sahel. These contrasts appear to result from time-
space changes, which cannot accurately be investigated from
six rain gauges.
4.3.2 Analysing West-African rainfall f ield
The wavelet coherence and phase of West-African rainfall are
investigated at the multi- (>20 years) to interannual (2–8
years) scales through teleconnections between the TSA index
(Fig. 9). This extended analysis is also an opportunity to check
the findings of Losada et al. (2010) who found that TSA SST
anomalies are associated with rainfall anomalies of the same
sign over the Gulf of Guinea Coast and of the opposite sign
over the Sahel.
At the multi-decadal scale, the TSA index is significantly
coherent with Sahelian rainfall during two periods
(Fig. 9(a.1)): (i) at the beginning of the twentieth century (with
more important time lag) over the western Sahel and (ii)
between the 1970s and 1980s over the eastern Sahel. Before
1950, observed out-of-phase coherence patches over the west-
ern Sahel are consistent with analysis of Sahel rainfall time
series (i.e. more western Sahel rainfall when the TSA SSTs are
colder; Fig. 8(a–c)). Furthermore, this out-of-phase relation-
ship occurs throughout the western part of West Africa
(Fig. 9(a.2)). During the 1960s and 1970s, this out-of-phase
relationship is shifted eastward and is restricted to the Sahelian
region (Fig. 9(a.3)). Moreover, according to the findings of
Fig. 8 Squared wavelet coherence (1) and phase (2) between monthly
Sahel rainfall amount (aDakar, bNioro du Sahel, cMopti, dNiamey, e
Maradi and fMaine-Soroa) and the TSA index (Enfield et al. 1999:0–20°
S, 10° E–30° W). Bold line delineates the area under which power can be
underestimated as a consequence of edge effects, wraparound effects and
zero padding; thin contour lines show the 90 % confidence limits based
on 1,000 Monte Carlo simulations of two independent AR[1] processes
150 B. Dieppois et al.
Author's personal copy
Losada et al. (2010), warmer TSA SSTs are associated with
more rainfall over the Gulf of Guinea Coast (Fig. 9(a.3)). We
note that this multi-decadal relationship was not identified
using rainfall time series probably because of its extremely
localised detection.
At the quasi-decadal time scale, the TSA index also dis-
plays out-of-phase coherence patches with Sahelian rainfall
during two periods (Fig. 9(b.1)): (i) at the beginning of the
twentieth century over the western Sahel and (ii) in the 1970s
and 1980s over the central and eastern Sahel. Thus, multi- and
quasi-decadal relationships between TSA SSTs and Sahel
rainfall are characterised by an eastward shift in the second
haft of the twentieth century (Fig. 9(a, b)). During both pe-
riods, this teleconnection is restricted to the Sahel and displays
Fig. 9 To p:(1) Wavelet coherence and phase between TSA index and
Sahel rainfall field using Hovmöller diagrams over the Sahel regions (12–
15° N) at the multi-decadal (>20 years; a), quasi-decadal (9–18 years; b)
and interannual (2–8years;c)scalesofco-variability.Middle and bottom:
(2–3) Wavelet coherence and phase between TSA index and West-Afri-
can rainfall when multi-decadal (a), quasi-decadal (b) and interannual (c)
coherence patterns are significantly detected over the Sahel (20-year
averages of this period is performed). For each scale, the right and left
columns display the wavelet coherence level using colour scale (from
blue to red) and wavelet phase (blue, in-phase relationship; red,out-of-
phase relationship), respectively. Thin contour lines refer to significant
coherence performed by Monte Carlo simulations at the 90 % confidence
limits
Low-frequency variability and zonal contrast in Sahel rainfall 151
Author's personal copy
an insignificant in-phase relationship over the Gulf of Guinea
Coast (Fig. 9(a, b)).
At the interannual time scale, the high nonstationary
time-scale fluctuations (Fig. 8) are likely the reason of low
coherence levels captured by our 2–8-year band-pass filter-
ing (Fig. 10(c)). Analysing the Hovmöller diagrams leads us
to some clarifications (Fig. 9(c.1)): (i) out-of-phase interan-
nual coherence patches are not detected after 1980; (ii) after
1980, the relationship between TSA index and Sahel rainfall
is in quadrature phase (i.e. large time lag of about 3 years);
and (iii) before (after) 1950s, out-of-phase relationships are
detected over the eastern (western) Sahel. This latter result is
reversed compared to the multi- and quasi-decadal
teleconnection changes. Before 1950s, interannual relation-
ship between rainfall and TSA index is restricted to the
Sahel regions, but an in-phase relationship (not significant)
over the Gulf of Guinea coastal regions (Fig. 9(c.2)). After
1980s, relationships of the same sign are detected
throughout West Africa (Fig. 9(c.3)). Thus, such differences
are in accordance with the sensitivity experiments of Polo
et al. (2008) and Wang et al. (2009) who have shown that
since the 1970s, changes in the TSA SST teleconnections
with Sahel rainfall could be due to ENSO-like patterns in the
Pacific.
In summary, colder TSA SSTs most often are associated
with wetter conditions in the Sahel and dryer conditions in the
Gulf of Guinea Coast. When the TSA SSTs are warmer, the
opposite is true. This is not detected before 1950s, when the
multi-decadal variability of the CRU TS 3.10.1 field could be
spatially autocorrelated due to few numbers of available rain
gauges. Based on the findings of Losada et al. (2010), we
show that the multi- to interannual SST anomalies restricted to
the TSA region are associated to meridional displacements of
the ITCZ. Nevertheless, these meridional displacements of the
ITCZ are sometimes not uniform depending on their time
scales and time periods.
Fig. 10 (a–c) Reconstructions of multi-decadal (>20 years), quasi-de-
cadal (9–18 years) and interannual (2–8 years) scales of Sahel rainfall
variability using FFT filtering (1,top) and their associated Atlantic SST
teleconnections (2,bottom). Atlantic Coast, central or western Sahel, and
eastern Sahel regions are plotted using light grey solid lines,dashed lines
and bold lines, respectively. Only the Dakar, Mopti and Maine-Soroa rain
gauges are used
152 B. Dieppois et al.
Author's personal copy
5 Discussion
Using both station-based and gridded data, North and tropical
South Atlantic SST teleconnections with Sahel rainfall occur
throughout the twentieth century. Since superposition of var-
ious Atlantic SST teleconnections modulates Sahel rainfall
variability in driving wet or dry anomalies even at the local
scale, Fig. 10 displays a detailed review of these results using
fast Fourier transform band-pass filtering of multi-decadal to
interannual variability of Dakar, Mopti and Maine-Soroa
rainfall.
During the wet period (1950s and 1960s), increasing multi-
decadal variability of the inter-hemispheric SST gradient
(warmer and colder anomalies of North and tropical South
Atlantic SSTs, respectively), as well as warmer quasi-decadal
anomalies of North Atlantic SSTs, are mainly associated with
wet conditions at Dakar and Mopti (Fig. 10(a, b)). More
generally, the North Atlantic SSTsignal is only coherent with
Sahelian rainfall during wet or “wetter”(cf. 1990s) periods, as
shown by Hodson et al. (2010). According to earlier studies,
such a teleconnection pattern could involve northwestward
shifts of the ITCZ. At the interannual time scale, a relationship
between TSA SSTs and Maine-Soroa rainfall is also detected
(Fig. 10(c)). According to Losada et al. (2010), this
teleconnection, which is associated with rainfall anomalies
of opposite signs between the Gulf of Guinea Coast and the
Sahel, might drive a northeastward displacement of the ITCZ.
By contrast, during the dry 1970s and 1980s, the TSA SST
variability seems generally dominant in driving Sahel rainfall
fluctuations (Fig. 10). These fluctuations are mainly detected
at the multi- and quasi-decadal scales, while interannual var-
iability strongly decreases (Fig. 10). At the multi-decadal
scale, warmer TSA SSTs are likely associated with observed
drier conditions at stations over the eastern Sahel and a part of
the central Sahel (Fig. 10(a)). Northeastward latitudinal mod-
ulations of the ITCZ might be involved in agreement with
Losada et al. (2010) and Dieppois et al. (2013). At the quasi-
decadal scale, variations of the inter-hemispheric SST gradient
(i.e. synchronous teleconnections with AMO and TSA index)
dominated in the 1970s and 1980s (Fig. 10(b)). This favoured
latitudinal modulations in the ITCZ and therefore the exis-
tence of successive drier (1969–1975 and 1981–1990) and
wetter (1976–1980) conditions at Mopti, eastern Sahel
(Fig. 10(b)). This is consistent with the findings of Dieppois
et al. (2013), showing that a strengthening of cross-equatorial
Atlantic SST and pressure gradients is related to an increase of
monsoon flows and a northward shift of the ITCZ over the
eastern Sahel.
Since the 1990s, as proposed by Mohino et al. (2011),
multi-decadal variability of North Atlantic SSTs could con-
tribute to the partial recovery of eastern Sahel rainfall (Lebel
and Ali 2009; Fontaine et al. 2011b;Fig.10(c)). At the
interannual scale, however, our results do not display robust
statistical connections between Sahel rainfall and Atlantic
SSTs during this period (Fig. 10(a, c)). This is probably due
to a very intermittent interannual variability signal in the
Sahelian rain gauges. Extra-Atlantic relationship—beyond
the scope of this study—should be considered (Fig. 10(c)),
in particular the rising influence of Pacific Ocean in the Sahel-
SST teleconnection (Janicot et al. 2001; Polo et al. 2008;
Wang et al. 2009). Interestingly, Joly and Voldoire (2010)
proposed that the collapse of the correlations between TSA
SSTs and Sahel rainfall after the 1970s might be due to the
counteracting effects of the Pacific and Atlantic basins during
these decades. Moreover, eastern Mediterranean SSTs (Polo
et al. 2008; Fontaine et al. 2009, 2010c), as well as the thermal
gradient between Indian and Mediterranean basins (Fontaine
et al. 2011c,2011d), which present an increase of statistical
links since 1980s, could also be involved.
6 Conclusions
Based on a time-space approach using both station-based and
gridded hydroclimatic data, this study examined (i) the
nonstationarity of interannual to multi-decadal Sahel rainfall
variability from the western to eastern Sahelian regions and
(ii) Atlantic SST–Sahel rainfall teleconnection variations
throughout the twentieth century. Finer time-scale explora-
tions allowed us to identify dissimilarity between western
and eastern Sahel rainfall variability. Analysing the selected
six rain gauges, three Sahel subregions can be identified: the
Atlantic Coast (Dakar), western-central Sahel (Nioro and
Mopti) and eastern Sahel (Niamey, Maradi, Maine-Soroa).
Only two clusters, i.e. western and eastern Sahel, however,
are significantly detected by extending wavelet clustering to
20 rain gauges or analysing the West-African field.
Western and eastern Sahel rainfall is correlated with SSTs
from different regions of Atlantic Ocean, especially in the
North and tropical South Atlantic. The North Atlantic SST
variability is only coherent with Sahelian rainfall during wet
or “wetter”(cf. 1990s) periods. It seems to modulate the
meridional thermal contrast and, at the multi- and quasi-
decadal scales, the northward displacements of the ITCZ,
hence explaining the observed in-phase relationship. The ex-
istence, however, of nonuniform northward displacements of
the ITCZ, tilted along NW–SE plan, also impact rainfall
amounts. In the 1950s, fluctuations of North Atlantic SSTs
favoured multi- and quasi-decadal rainfall variability over the
western part of West Africa including the western Sahel up to
6° W/2° E. Since 1990, a similar North Atlantic SST
teleconnection pattern, but shifted northward, has been ob-
served at the multi-decadal time scales, in accordance with the
recent rainfall recovery observed over the eastern Sahel.
The occurrence of opposite teleconnection patterns be-
tween the Gulf of Guinea Coast (in phase) and the Sahel
Low-frequency variability and zonal contrast in Sahel rainfall 153
Author's personal copy
(out of phase) indicates that TSA SST variability is associated
with meridional displacements of the ITCZ and hence with
Sahel rainfall. This teleconnection is, however, highly sensi-
tive to the considered time periods and scales. Such induced
northward ITCZ and monsoon flux displacements occurred
over the western or eastern Sahel regions.
Thus, North Atlantic variability and the TSA SST variabil-
ity appear to contribute, at least partially, to driving contrasts
between western and Sahel rainfall. The Atlantic forcing,
however, is not the only SST forcing that may affect such
dissimilarity. Pacific SSTs are likely to drive counteracting
effects with the Atlantic SST variability at both interannual
and quasi-decadal time scales, while eastern Mediterranean
and Indian SST fluctuations might modify meridional entropy
gradient. Upcoming investigations will be based on sensitivity
experiments focusing on multiple oceanic basins and conduct-
ed with several atmospheric global climate models forced by
SSTs.
Acknowledgments This study was conducted as part of the CORUS2
project entitled “Impact de la pression anthropique et du Changement
Global sur les flux sédimentaires en zone sahélienne”(grant no. 6116)
supported by the French Institute of Research for Development (Institut
de Recherche pour le Développement, IRD). This topic has been a
contribution of the AMEDE action (Analyse Multi-Echelle de la
Dynamique Eolienne au Sahel) based on the FED 4116 SCALE (TEQQ
project) supported by the Upper Normandy region (France). This study
also contributes in the FRIEND-AOC research program (Flow Regimes
from International Experimental and Network Data in Central and West
Africa), which is a UNESCO initiative within its International Hydrolog-
ical Programme (IHP).
References
Bader J, Latif M (2003) The impact of decadal-scale Indian Ocean sea
surface temperature anomalies on Sahelian rainfall and the North
Atlantic Oscillation. Geophys Res Lett 30:2169. doi:10.1029/
2003GL018426
Christensen JH, Hewitson B, Busuioc A, Chen A, Gao X, Held I, Jones
R, Kolli RK, Kwon WT, Laprise R, Mangaña Rueda V, Meams L,
Menéndez CG, Räisänen J, Rinke A, Sarr A, Whetton P (2007)
Regional climate projections. In: Solomon S, Qin D, Manning M,
Chen Z, Marquis M, Averyt KB, Tignor M, Miller HL (eds) Climate
change 2007: the physical sciences basis. Contribution of Working
Group I to the fourth assessment report of the Intergovernmental
Panel on Climate Change. Cambridge University Press, Cambridge
Dai A, Lamb PJ, Trenberth KE, Hulme M, Jones PD, Xie P (2004) The
recent Sahel drought is real. Int J Climatol 24:1323–1331
Dieppois B, Diedhiou A, Durand A, Fournier M, Massei N, Sebag D, Xue
Y, Fontaine B (2013) Quasi-decadal signals of Sahel rainfall and
West African monsoon since the mid-twentieth century. J Geophys
Res Atmos 118:12587–12599
Ebisuzaki W (1997) A method to estimate the statistical significance of a
correlation when the data are serially correlated. J Clim 10:2147–
2153
Enfield DB, Mestas-Nuñez AM, Mayer DA, Cid-Serrano L (1999) How
ubiquitous is the dipole relationship in the tropical Atlantic sea
surface temperatures? J Geophys Res 104:7841–7848
Folland CK, Palmer TN, Parker DE (1986) Sahel rainfall and worldwide
sea temperatures, 1901-85. Nature 320:602–606
Fontaine B, Trzaska S, Janicot S (1998) Evolution of the relationship
between nearglobal and Atlantic SST modes and the rainy season in
West Africa: statistical analyses and sensitivity experiments. Clim
Dyn 414:353–368
Fontaine B, Garcia-Serrano J, Roucou P, Rodriguez-Fonseca B, Losada
T, Chauvin F, Gervois S, Sivarajan S, Ruti P, Janicot S (2009)
Impacts of warm and cold situations in the Mediterranean Basins
on the West African Monsoon: observed connection patterns (1979-
2006) and climate simulations. Clim Dyn 35:95–114
Fontaine B, Roucou P, Gaetani M, Marteau R (2011a) Recent changes in
precipitations, ITCZ convection and northern tropical circulation
over North Africa (1979–2007). Int J Climatol 31:633–648
Fontaine B, Roucou P, Monerie PA (2011b) Changes in the African
monsoon region at medium-term time horizon using 12 AR4
coupled models under the A1b emissions scenario. Atmos Sci Lett
12:83–88
Fontaine B, Monerie PA, Gaetani M, Roucou P (2011c) Climate adjust-
ments over the African-Indian monsoon regions accompanying
Mediterranean Sea thermal variability. J Geophys Res 116:
D23122. doi:10.1029/2011JD016273
Fontaine B,Gaetani M, Ullmann A, Roucou P (2011d) Time evolution of
observed July–September sea surface temperature-Sahel climate
teleconnection with removed quasi-global effect (1900–2008). J
Geophys Res 116:D04105. doi:10.1029/2010JD014843
Giannini A, Saravannan RP, Chang P (2003) Ocean forcing of Sahel
rainfall on interannual to interdecadal time scales. Science 302:
1027–1030
Hagos SM, Cook KH (2008) Ocean warming and late-twentieth-century
sahel drought and recovery. J Clim 21:3797–3814
Hodson DLR, Sutton RT, Cassou C, Keenlyside N, Okumura Y, Zhou T
(2010) Climate impact of recent multidecadal changes in Atlantic
Ocean Sea Surface Temperature: a multimode comparison. Clim
Dyn 34:1041–1058
Janicot S (1992) Spatio-temporal variability of West African rainfall. Part
I: regionalizations and typings. J Clim 5:489–497
Janicot S, Trzaska S, Poccard I (2001) Summer Sahel-ENSO
teleconnection and decadal time-scale SST variations. Clim Dyn
18:303–320
Joly M, Voldoire A (2010) Role of the Gulf of Guinea in the inter-annual
variability of the West African monsoon: what do we learn from
CMIP3 coupled simulations? Int J Climatol 30:1843–1856
Knight JR, Folland CK, Scaife AA (2006) Climate impacts of the Atlantic
multidecadal oscillation. Geophys Res Lett 33:L17706. doi:10.
1029/2006GL026242
Lamb PJ (1978) Case studies of tropical Atlantic surface circulation
patterns during recent sub-Saharan weather anomalies: 1967 and
1968. Mon Weather Rev 106:482–491
Latif M, Grotzner A (2000) On the equatorial Atlantic oscillation and its
response to ENSO. Clim Dyn 16:213–218
Le Barbé L, Lebel T, Tapsoba D (2002) Rainfall Variability in West
Africa during the years 1950–90. J Clim 15:187–202
Lebel T, Ali A (2009) Recent trends in the Central and Western Sahel
rainfall regime 1990–1997. J Hydrol 375:52–64
Lebel T, Delclaux F, Le Barbé L, Polcher J (2000) From MCG scales to
hydrological scales: Rainfall variability in West Africa. Stoch Env
Res Risk Ass 14:275–295
Losada T, Rodriguez-Fonseca B, Janicot S, Gervois S, Chauvin F, Ruti P
(2010) A multimodel approach to the Atlantic equatorial mode.
Impact on the West African monsoon. Clim Dyn 35:29–43
Lu J, Delworth TL (2005) Ocean forcing of the late 20th century Sahel
drought. Geophys Res Lett 32:L22706. doi:10.1029/
2005GL023316
Maraun D (2006) What can we learn from climate data? Methods for
fluctuation time/scale and phase analysis. Universitât Postdam
154 B. Dieppois et al.
Author's personal copy
Maraun D, Kurths J (2004) Cross wavelet analysis: significance testing
and pitfalls. Nonlinear Process Geophys 11:505–514
Mohino E, Janicot S, Bader J (2011) Sahel rainfall and decadal to multi-
decadal sea surface temperature variability. Clim Dyn 37:419–440
Monerie PA, Roucou P, Fontaine B (2012)Mid-century effects of Climate
Change on African monsoon dynamics using A1B emission scenar-
io. Int J Climatol 33:881–896
Moron V (1994) Guinean and Sahelian rainfall anomaly indices at annual
and monthly time scales (1933-1990). Int J Climatol 14:325–341
Nicholson SE (2005) On the question of the ‘recovery’of the rains in the
West African Sahel. J Arid Environ 63:615–641
Nieto R, Gimeno L, Trigo R (2006) A Lagrangian identification of major
sources of Sahel moisture. Geophys Res Lett 33:L18707. doi:10.
1029/2006GL027232
Palmer TN (1986) The influence of the Atlantic, Pacific and Indian
Oceans on Sahel rainfall. Nature 322:251–253
Polo I, Rodriguez-Fonseca B, Losada T, Garcia-Serrano J (2008) Tropical
Atlantic variability modes (1979–2002). Part I: time-evolving SST
modes related to West African rainfall. J Clim 21:6457–6475
Rodriguez-Fonseca B, Janicot S, Mohino E, Losada T, Bader J,
Caminade C, Chauvin F, Fontaine B, Garcia-Serrano J, Gervois S,
Joly M, Polo I, Ruti P, Roucou P, Voldoire A (2011) Interannual and
decadal SST-forced responses of the West African monsoon. Atmos
Sci Lett 12:67–74
Rouyer T, Fromentin JM, Stenseth NC, Cazelles B (2008) Analysing
multiple time series and extending significance testing in wavelet
analysis. Mar Ecol Prog Ser 359:11–23
Rowell DP, Folland CK, Maskell K, Ward MN (1995) Variability of
summer rainfall over tropical North Africa (1906–1992): observa-
tions and modelling. Quart J Roy Met Soc 121:669–704
Servain J, Picaut J, Merle J (1982) Evidence of remote forcing in the
equatorial Atlantic Ocean. J Phys Oceanogr 12:457–463
Smith TM, Reynolds RW, Peterson TC, Lawrimore J (2008)
Improvements to NOAA’s historical merged land-ocean surface
temperature analysis (1880–2006). J Clim 21:2283–2296
Sutton RT, Hodson DLR (2005) Atlantic Ocean forcing of North
American European summer climate. Science 309:115–118
Torrence C, Compo GP (1998) A practical guide to wavelet analysis. B
Am Soc Met Soc 79:61–78
Torrence C, Webster P (1999) Interdecadal Changes in the ENSO-
Monsoon System. J Clim 12:2679–2690
van Oldenborgh GJ, te Raa LA, Dijkstra HA, Philip SY (2009)
Frequency- or amplitude-dependent effects of the Atlantic me-
ridional overturning on the tropical Pacific Ocean. Ocean Sci 5:
293–301
Wang C, Kucharski F, Barimalala R, Bracco A (2009) Teleconnections of
the Tropical Atlantic to the Tropical Indian and Pacific Oceans a
review of recent findings. Spec Issue Meteorol Z 18:445–454
Ward NM (1998) Diagnosis and short-lead time prediction of summer
rainfall in tropical North Africa at interannual and multidecadal
timescales. J Clim 11:3167–3191
Zhang R, Delworth TL (2006) Impact of Atlantic multidecadal oscilla-
tions on India/Sahel rainfall and Atlantic hurricanes. Geophys Res
Lett 33:L17712. doi:10.1029/2006GL026267
Low-frequency variability and zonal contrast in Sahel rainfall 155
Author's personal copy
