JOURNAL OF GEOPHYSICAL RESEARCH: BIOGEOSCIENCES, VOL. 118, 1–12, doi:10.1002/jgrg.20115, 2013
In the hot seat: Insolation, ENSO, and vegetation
in the African tropics
Sarah J. Ivory,1 Joellen Russell,1 and Andrew S. Cohen 1
Received 27 December 2012; revised 19 June 2013; accepted 5 September 2013.
[1] African climate is changing at rates unprecedented in the Late Holocene with profound
implications for tropical ecosystems and the global hydrologic cycle. Understanding the
specific climate drivers behind tropical ecosystem change is critical for both future and
paleomodeling efforts. However, linkages between climate and vegetation in the tropics
have been extremely controversial. The Normalized Difference Vegetation Index (NDVI) is
a satellite-derived index of vegetation productivity with a high spatial and temporal
resolution. Here we use regression analysis to show that NDVI variability in Africa is
primarily correlated with the interannual extent of the Intertropical Convergence Zone
(ITCZ). Our results indicate that interannual variability of the ITCZ, rather than sea surface
temperatures or teleconnections to middle/high latitudes, drives patterns in African
vegetation resulting from the effects of insolation anomalies and El Niño–Southern
Oscillation (ENSO) events on atmospheric circulation. Global controls on tropical
atmospheric circulation allow for spatially coherent reconstruction of interannual vegetation
variability throughout Africa on many time scales through regulation of dry season length
and moisture convergence, rather than precipitation amount.
Citation: Ivory, S. J., J. Russell, and A. S. Cohen (2013), In the hot seat: Insolation, ENSO, and vegetation in the African
tropics, J. Geophys. Res. Biogeosci., 118, doi:10.1002/jgrg.20115.
1.
Niño–Southern Oscillation (ENSO), upper level perturbations, sea surface temperature (SSTs), Hadley Cell intensity,
and Intertropical Convergence Zone (ITCZ) excursions
[Rossignol-Strick, 1983; DeMenocal and Rind, 1993;
Nicholson and Selato, 2000; Hulme et al., 2001; Mercier
et al., 2002; New et al., 2003; Nicholson and Grist, 2003;
Anyah and Semazzi, 2004; Hoerling et al., 2006; Giannini
et al., 2008, Chiang, 2009]. Therefore, studies of both past
and future climate change create the need for quantitative
constraints on the modern relationship between climate
controls and vegetation.
[4] Rainfall seasonality in tropical Africa is determined by
the annual migration of large-scale circulations like the ITCZ
and Congo Air Boundary [Nicholson, 1996]. Summer rainfall occurs in southern and northern tropical Africa, and an
equatorial bimodal regime lies in between. This simplistic
model notwithstanding, the spatial heterogeneity of African
vegetation indeed mostly results from regional controls on
rainfall and dry season length, in addition to local factors
such as topography, lake effects, continentality, and SSTs
of adjacent oceans [Nicholson, 1996]. Rainfall varies over
kilometers from <200 to >2000 mm/year and dry season
length from 0 to >6 months. While biomes range from tropical
rainforest to open desert, Africa is dominated by semi-arid
or other seasonal vegetation that is very sensitive to hydrologic variability [White, 1983]. As a result of the mosaic
of influences, it is unclear whether vegetation variability
is influenced more by large-scale forcing or local/regional
changes and how variability of rainfall amount and dry
season length may independently affect vegetation [Hély
et al., 2006].
Introduction
[2] Ecosystems and rainfed agriculture provide essential
resources to millions of Africans. Threats to tropical vegetation related to rapidly changing climate make this region
extremely vulnerable [Intergovernmental Panel on Climate
Change (IPCC), 2007] and have dangerous, costly implications for daily economy and subsistence [W.W. Forum, 2000;
Few et al., 2004]. Despite these concerns, it remains uncertain
how natural vegetation communities and cropland productivity
will respond to altered atmospheric and oceanic circulation in
the future and little infrastructure exists to mitigate these
changes [Boko et al., 2007; Brown and deBeurs, 2008].
[3] Modern observational data and paleoclimate records are
both powerful tools for understanding biosphere dynamics
[Bradley, 1985]. Interpreting both types of records, however, is complicated by much uncertainty surrounding the
question of what climatic mechanisms regulate African
vegetation change today. Such changes have been attributed
to insolation, low-latitude teleconnections to high latitude
or midlatitude, vegetation/soil moisture feedbacks, El
Additional supporting information may be found in the online version of
this article.
1
Department of Geosciences, University of Arizona, Tucson, Arizona,
USA.
Corresponding author: S. J. Ivory, Department of Geosciences,
University of Arizona, 1040 E 4th St., Tucson, AZ 85721, USA.
(ivorysj@email.arizona.edu)
©2013. American Geophysical Union. All Rights Reserved.
2169-8953/13/10.1002/jgrg.20115
1
IVORY ET AL.: AFRICAN CLIMATE AND VEGETATION
Saharan Africa. African biome NDVI values range from 0.1
(grassland, bushland) to 0.7 (evergreen rainforest) [Davenport
and Nicholson, 1993; Hély et al., 2006]. At both extremes
of vegetation density (very high and very low values of leaf
area index), NDVI saturates. In order to account for this,
values above 0.7 or below 0.1 are not considered. Following
the methods of Russell and Wallace [2004], 3 month mean
values of NDVI fields were used to represent the seasonal
mean vegetation and were regressed on time series of climatic
variables that represent variability in the tropical atmosphereocean system.
[9] Regression coefficients between NDVI and each
climate variable are displayed on the resultant regression
maps. The patterns displayed in these maps represent
regions of negative (brown) and positive (green) correlation
to each climate variable, while grey represents regions of no
relationship between NDVI and the climate variable used.
Regression, instead of correlation, was conducted for this
analysis, because regression coefficients produced and
displayed in the maps in each grid cell have the added
advantage of representing the quantitative relationship
between climate and NDVI. These displayed values in each
map represent the NDVI anomaly for one standard deviation of each climate time series. The purpose of the study
is to evaluate the variability of NDVI over the study period
rather than any possible trends in the data, so we chose a
data set with a time range that has been used in similar
studies linking vegetation to climate [Xu et al., 2010;
Meng et al., 2011; Hountondji et al., 2009; Fabricante
et al., 2009; Revadekar et al., 2012] and was also evaluated
by Beck et al. [2011] and found to have the least error associated with temporal change.
[10] The climate variables used for the analysis are important components within the tropical atmosphere-ocean system which are strongly correlated to rainfall both seasonally
and from year to year [Nicholson, 1986; Camberlin et al.,
2001; Hoerling et al., 2006; Seidel et al., 2008]. These are
time series of SSTs and indicators of ITCZ position and
intensity that have linkages to rainfall timing and amount
and thus are likely to influence vegetation. We examine
SST variability from within three regions of known importance for African rainfall: the Southern Atlantic, (0–20°S,
10°E–10°W), the North Indian Ocean, (15–25°N, 55–75°E),
and the South Indian Ocean (16–30°S, 36–48°E) [Hoerling
et al., 2006]. ITCZ indicators were chosen based on Seidel
et al. [2008] and are outgoing long-wave radiation (OLR),
200 mb geopotential height (troposphere height), and incoming solar radiation at the top of the atmosphere (insolation).
Within the region of the ITCZ, regions of low pressure at the
surface and high pressure aloft follow the sun seasonally and
result in the lofting of air masses deep into the troposphere
causing clouds and rainfall. Thus, the extent of the ITCZ
region can be tracked by evaluating changes in insolation as
well as high troposphere heights and low OLR from extensive
cloud cover.
[11] To investigate the influence of climate variability on
vegetation in the summer season of each hemisphere,
December-January-February (austral summer) and JuneJuly-August (boreal summer), time series were constructed
for each variable for the 1980–2000 study period. SSTs are
latitude-weighted mean values from the 1° × 1° HadISST1
data set [http://climexp.knmi.nl; Rayner et al., 2003]. ITCZ
[5] Satellite data provide continuous, high-resolution data
sets of vegetation productivity and can help identify the
direct relationship between vegetation and climate on global
to local scales [Russell and Wallace, 2004; Brown and
deBeurs, 2008; Brown et al., 2010]. For this study, a 21 year
8 km resolution Normalized Difference Vegetation Index
(NDVI) record (1980–2000) from NOAA’s Advanced Very
High Resolution Radiometer (AVHRR) has been used to
reconstruct patterns of interannual vegetation variability
throughout tropical Africa [Tucker et al., 2005]. Whereas
studies in the tropics have linked NDVI to SST and precipitation variability, the relative effects of atmospheric and
oceanic variability at the continental scale has not been
addressed [Anyamba and Eastman, 1996; Anyamba et al.,
2002; Martiny et al., 2006; Brown et al., 2010]. In this study,
we have regressed NDVI fields on climatic variables previously proposed as possible regulatory mechanisms for
Afrotropical vegetation as well as assessing the relative importance of these variables in the tropical atmosphere-ocean
system [Rossignol-Strick, 1983; DeMenocal and Rind, 1993;
Nicholson and Selato, 2000; Hulme et al., 2001; Mercier
et al., 2002; New et al., 2003; Nicholson and Grist, 2003;
Anyah and Semazzi, 2004; Hoerling et al., 2006; Giannini
et al., 2008, Chiang, 2009]. Our results provide a new perspective on tropical climate-vegetation interactions and modern
vegetation variability within the African tropics.
2.
Methods
[6] Vegetation across the sub-Saharan African continent
varies greatly in terms of its structure and composition
[White, 1983]. Near the equator in West Central Africa,
dense Guineo-Congolan rainforests and semi-deciduous forests are prevalent. In the rift valley of East Africa, even near
the equator, vegetation is much more open with large areas of
semi-arid vegetation consisting largely of Acacia dominated
bushland, open woodland, and wooded grassland. To the
north and south of the equator, respectively, Sudanian and
Zambezian woodlands are located over extensive zones
nearly from coast to coast. Further north and south, into the
subtropics, scrublands and steppe of the Sahel and Karoo
are replaced in some areas by absolute desert.
[7] Although many studies have looked at controls on vegetation variability on local scales within a single vegetation
type, few studies have looked at climatic influences on vegetation across broad spatial scales and across vegetation types
to examine large-scale regional coherency of variability. This
type of study is integral to understanding the relationship of
vegetation to large-scale climate phenomenon that control
regional hydrology and identify areas particularly sensitive
to change that are likely to be greatly altered in the future.
[8] Normalized Difference Vegetation Index (NDVI), a
satellite-derived index of absorbed red versus scattered
near-infrared light, is highly correlated with vegetation net
primary productivity [Tucker et al., 2005]. NDVI has been
shown to be representative of African vegetation types, and
thus is useful in quantitatively linking vegetation variability
to climate factors that influence productivity on broad spatial
scales [Davenport and Nicholson, 1993; Hély et al., 2006;
Brown et al., 2010]. In order to better understand climatic
controls across Africa, we used the 1980–2000 GIMMS
AVHRR data set of Tucker et al. [2005] throughout sub2
IVORY ET AL.: AFRICAN CLIMATE AND VEGETATION
Figure 1. Leading EOFs used to produce time series for (a) December-January-February and JuneJuly-August Insolation, (b) December-January-February and June-July-August OLR, and (c) DecemberJanuary-February and June-July-August 200 mb geopotential height (troposphere height).
indicator time series are the first empirical orthogonal
function (EOF) of troposphere height, OLR, and insolation
constructed using NCEP/NCAR Reanalysis data [http://
www.esrl.noaa.gov/psd/, Kalnay et al., 1996] within tropical
Africa (23°N–23°S, 0–40°E; Figure 1). EOFs are a statistical
method allowing us to extract a single time series that represents the most important variability over a broad spatial
scale. This technique was used for the ITCZ indicator time
series as these variables were reconstructed over the whole
of the tropics, including areas north and south of the
equator. Unlike the SST time series, a simple average would
not be representative of variability in both hemispheres at
once. Plots of all time series used in this analysis are presented
in Figure 2.
3
IVORY ET AL.: AFRICAN CLIMATE AND VEGETATION
Figure 2. Plots of all time series used in analysis. (a) Time series used for regression analysis of
atmospheric and oceanic variability, (b) time series of leading PCs with percent variance explained, and
(c) diagnostic time series of NDVI patterns to large-scale climatic phenomena.
[12] Significance of individual regression patterns was
determined by a Monte Carlo test. First, the latitude-weighted
mean covariance was calculated for each pattern. This was
then compared with mean covariance produced by 10,000
regressions of random standardized time series and takes into
account time series autocorrelation.
[13] Finally, in order to evaluate the main components of
variability in the tropical atmosphere-ocean system and the
relationships between the climatic variables, we performed
a separate principal components analysis (PCA) on all
climatic variables discussed above. PCA is an analysis used
commonly in climatological and ecological studies which
allows for the reduction of dimensionality in large data sets
and is used here in order to extract the main components of
variability within a system [Jolliffe, 1990]. Loadings for
the three significant axes are presented in the supporting
information Table S1, and plots of all time series can be
found in Figure 2.
(Table 1). This suggests that interannual vegetation variability
is strongly controlled by atmospheric circulation.
[15] The patterns produced by regressing NDVI on insolation
within the same season are shown in Figures 3a and 3b. The
December-January-February pattern shows a strong positive
anomaly in southeast Africa with a negative anomaly in all
other regions, especially marked in North and South Africa
(Figure 3a). The June-July-August pattern is similar, but with
signs reversed, showing a negative anomaly in East Africa
and a strong positive anomaly persisting in North Africa
(Figure 3b). Additionally, another positive anomaly stretches
from the Indian to the Atlantic coast in southern Africa.
[16] The pattern produced by regressing NDVI on JuneJuly-August OLR within the same season is shown in
Figure 4. Here we see a band of strong negative anomaly
running zonally from 10°N to 15°N, a second weaker negative anomaly in South Africa, and a positive anomaly in the
equatorial region.
3.
3.1. Principal Components Analysis
of Climatic Variables
[17] In order to extract the main components of variability
within the tropical climate system and interrelation of these
climatic variables, a PCA was performed. The PCA resulted
in three significant axes, explaining 60.0% of the variance
(Table 1). A biplot with loadings of the first two axes shows
that nearly half of the variance is explained by the first two
PCs which load strongly on the ITCZ indicators (Figure 5;
Table S1).
Results
[14] When each climate time series is regressed on NDVI,
the ITCZ indicators consistently show consistently higher
values than the SST time series (Table 1). Both the JuneJuly-August and the December-January-February insolation
time series show significant covariance with NDVI from
the same season. Additionally, June-July-August OLR yields
the highest covariance of all of the time series when regressed
on NDVI within the same season. In contrast, covariance
produced by SSTs is low and never reaches significant values
4
IVORY ET AL.: AFRICAN CLIMATE AND VEGETATION
Table 1. Climate Variable Regressionsa
NDVI
December-January-February
January-February-March
February-March-April
March-April-May
April-May-June
May-June-July
June-July-August
July-August-September
August-September-October
September-October-November
October-November-December
November-December-January
NDVI
December-January-February
January-February-March
February-March-April
March-April-May
April-May-June
May-June-July
June-July-August
July-August-September
August-September-October
September-October-November
October-November-December
November-December-January
DJF Troposphere
Height
JJA Troposphere
Height
DJF OLR
JJA OLR
DJF
Insolation
JJA
Insolation
0.0095
0.0093
0.0104
0.0122
0.0120
0.0118
0.0082
0.0076
0.0081
0.0081
0.0085
0.0082
0.0128
0.0130d
0.0131d
0.0121
0.0120
0.0116
0.0111
0.0108
0.0128
0.0129d
0.0133d
0.0132d
0.0069
0.0069
0.0068
0.0085
0.0098
0.0113
0.0132d
0.0098
0.0098
0.0103
0.0111
0.0102
0.0101
0.0085
0.0094
0.0097
0.0082
0.0080
0.0183b,e
0.0118
0.0134d
0.0135d
0.0135d
0.0116
0.0155c,e
0.0117
0.0105
0.0104
0.0099
0.0104
0.0127
0.0113
0.0106
0.0121
0.0130d
0.0131d
0.0106
0.0079
0.0083
0.0091
0.0098
0.0110
0.0141c,e
0.0111
0.0084
0.0089
0.0091
0.0123
DJF Atlantic
JJA Atlantic
DJF S. Indian
JJA S. Indian
DJF N. Indian
JJA N. Indian
0.0083
0.0083
0.0083
0.0091
0.0094
0.0108
0.0121
0.0126
0.0130d
0.0118
0.0107
0.0104
0.0108
0.0080
0.0083
0.0096
0.0103
0.0111
0.0095
0.0091
0.0093
0.0100
0.0107
0.0110
0.0083
0.0082
0.0082
0.0086
0.0084
0.0090
0.0092
0.0085
0.0086
0.0092
0.0113
0.0103
0.0083
0.0075
0.0076
0.0079
0.0080
0.0083
0.0105
0.0104
0.0094
0.0085
0.0094
0.0108
0.0107
0.0094
0.0084
0.0091
0.0088
0.0088
0.0106
0.0098
0.0088
0.0100
0.0101
0.0094
0.0089
0.0083
0.0084
0.0091
0.0098
0.0107
0.0139d
0.0120
0.0107
0.0098
0.0097
0.0114
a
Mean regression coefficients based on regression of 3 month means of African NDVI on all climatic variables.
Values are significant to 99%.
c
Values are significant to 95%.
d
Values are significant to 90%.
e
Bold values are regression patterns used in subsequent figures.
b
[19] PC2 (20.4% variance) loads strongly on troposphere
height in both seasons and contrasts with both seasons of
OLR (Figure 5). The biplot supports the notion that both
troposphere height and OLR are positively correlated
between seasons (troposphere height: r = 0.82, p < 0.001 ;
OLR: r = 0.43, p = 0.05) and negatively correlated with each
other (December-January-February: r = 0.631, p = 0.001;
June-July-August: r = 0.262, p = 0.13). OLR is a metric of
atmospheric water vapor content, with high values indicating
cloudless days. In contrast, high values of troposphere height
indicate deep tropospheric convection producing tropical
[18] PC1 (23.8% variance) loads strongly on June-JulyAugust insolation and contrasts with December-JanuaryFebruary insolation (Figure 5). In this 21 year record,
insolation changes are the result of variations in solar irradiance, which show a strong negative correlation between
seasons (r = 0.812; p value < 0.001). This illustrates the
antiphased nature of summer insolation anomalies and
suggests the existence of an interhemispheric gradient such
that, for a given year, high summer insolation in one
hemisphere corresponds to lower summer insolation in the
opposite hemisphere.
Figure 3. NDVI regressions on insolation with mean covariance (COV), as displayed in Table 1, at the
bottom right of each panel. (a) Regression map of December-January-February NDVI on DecemberJanuary-February insolation, (b) regression map of June-July-August NDVI on June-July-August insolation.
Double asterisks indicate 99% confidence and single asterisk represents 95% confidence.
5
IVORY ET AL.: AFRICAN CLIMATE AND VEGETATION
Figure 4. NDVI regressions on OLR with mean covariance (COV), as displayed in Table 1, at the bottom
right of each panel. Regression map of June-July-August NDVI on June-July-August OLR. Double asterisks indicate 99% confidence and single asterisk represents 95% confidence.
variability in December-January-February and June-JulyAugust (Table S1).
[21] SST variability in these regions is often considered
important for rainfall patterns throughout Africa [Hoerling
et al., 2006]. The PCA agrees with this, suggesting that
SSTs play a strong role in driving the atmosphere-ocean
system. However, despite the influence of SSTs on rainfall,
no strong relationship exists between SST and NDVI
(Table 1). It is possible that our study does not capture the
relationship between SSTs and NDVI due to a problem of
inappropriate temporal or spatial scale that does not capture
SST variability. However, other studies using shorter
time series or lower spatial resolution have used similar
storms, and low values indicate a more stable, stratified
troposphere [Seidel et al, 2008]. Thus, PC2 suggests that higher
troposphere height, or deep convection within the tropics,
is coincident with lower OLR and increased cloudiness
throughout the tropics.
[20] All three significant PCs show an important contribution of variance from the SST time series (Table S1). In
addition to insolation and OLR, PC1 loads on DecemberJanuary-February and June-July-August North Indian
Ocean SSTs and June-July-August South Indian Ocean
SSTs, while PC2 shows a very high contribution from
December-January-February South Indian Ocean SSTs.
PC3 (15.8%) loads most strongly on Atlantic Ocean
Figure 5. Axes 1 and 2 for PCA of all standardized climate time series with loadings of all variables. Red
dots represent the loadings of individual years with years of warm ENSO events during (1988, 1999),
marked in italics.
6
IVORY ET AL.: AFRICAN CLIMATE AND VEGETATION
Figure 6. African mean vertical velocity fields and correlations on climate variables. (a) mean 1980–2000
December-January-February 500 mb vertical velocity field, (b) December-January-February 500 mb vertical
velocity field correlated with December-January-February insolation, (c) mean 1980–2000 JuneJuly-August 500 mb vertical velocity field, (d) June-July-August 500 mb vertical velocity field correlated
with June-July-August insolation, (e) mean 1980–2000 June-July-August 500 mb vertical velocity field,
(f) June-July-August 500 mb vertical velocity field correlated with June-July-August OLR. Bold contours
on correlation plots are significant to 95%.
atmosphere (Figure 3). The patterns produced in both seasons
are similar; however, the anomalies are of opposing sign due
to the antiphased relationship between the insolation time
series from each season. This means that when JuneJuly-August insolation is high, December-January-February
insolation is low that same year. Also, this suggests that a
reversal of the insolation gradient produces a vegetation
response of opposing sign.
[23] As the ITCZ tracks the sun north and south throughout
the year, variable insolation within the tropics may affect
vegetation through two mechanisms: enhanced rainfall or
techniques to define significant relationships between SST
and rainfall [Nicholson, 1986; Camberlin et al., 2001;
Hoerling et al., 2006].
4.
Discussion
4.1. Dry Season and Drought Stress
[22] The NDVI regression patterns produced by DecemberJanuary-February and June-July-August insolation illustrate
the response of vegetation across broad spatial scales and vegetation types to large-scale circulation within the tropical
7
IVORY ET AL.: AFRICAN CLIMATE AND VEGETATION
Figure 7. NDVI regressions on dry days with mean covariance (COV), as displayed in Table 1, at the bottom
right of each panel. (a) Regression map of December-January-February NDVI on southern hemisphere
dry days, (b) regression map of June-July-August NDVI on northern hemisphere dry days. Double
asterisks indicate 99% confidence and single asterisk represents 95% confidence.
[27] In December-January-February, a band of negative
correlation extends southeastward from 10°S to 40°S, south
of the mean ITCZ position shown in Figure 6a, suggesting
a more southerly position of convection when DecemberJanuary-February insolation is high (Figure 6b). In this
season, the positive NDVI anomaly in southeast Africa agrees
with a southerly ITCZ which brings more rainfall and
lengthens the growing season; however, a strong negative
NDVI anomaly persists south of 20°S suggesting drier
conditions (Figure 3a).
[28] In East Africa, a meridional component of the ITCZ
develops in December-January-February, the Congo Air
Boundary, extending northeastward from Namibia. The
Congo Air Boundary is a region of convergence of dry air
originating from the northern Indian Ocean with moister air
masses from the Congo basin and southern Atlantic Ocean.
This eastward moisture penetration from the Atlantic is
important for driving rainfall in East Africa during the southern
hemisphere rainy season [McHugh, 2004]. The correlation of
December-January-February insolation with 500 mb vertical
velocity suggests that a southward shift of the ITCZ is mirrored
by a southward shift of the semi-permanent South Atlantic anticyclone, an area of positive correlation west of southern Africa
(Figure 6b). This southward retreat of the anti-cyclone results
in reduced convection on the western side of the Congo
Air Boundary leading to greater eastward Atlantic moisture
penetration during austral summer and promoting a wetting
in southeast and drying in southern Africa consistent with the
pattern observed in the December-January-February NDVI
regression pattern (Figure 3a). In contrast, when DecemberJanuary-February insolation is low, a northward ITCZ and
South Atlantic anti-cyclone block Atlantic moisture from
penetrating further eastward.
[29] In June-July-August, negative correlation between
insolation and vertical velocity is observed beginning near
the mean ITCZ position centered around 7°N, but extends
much farther northward (Figure 6d). This suggests that rising
air associated with rainfall is stronger and more northward
when June-July-August insolation is high. Additionally, this
band of negative correlation extends southward as well along
the West African coast. This may be the result of a northward
enhanced evapotranspiration. For both of the insolation
regression patterns in Figures 3a and 3b, the positive anomalies occur in the summer hemisphere. Higher vegetation
productivity during a rainy season when insolation is high
suggests that it is unlikely that the insolation regression
patterns are created by greater evaporation. These patterns
instead seem linked to summer rains, whose position and
intensity are determined by insolation [Nicholson, 1986].
[24] One possible mechanism could be an intensification of
Hadley circulation leading to greater summer rainfall.
Although this may explain positive anomalies for both JuneJuly-August and December-January-February in much of the
summer hemisphere, a strong negative anomaly in southern
Africa during austral summer would be contradictory
(Figure 3a). Thus, it seems that greater summer precipitation
alone cannot fully explain this pattern.
[25] Instead, we propose that the insolation regression
patterns are related to the vegetation response to changes in
rainfall due to excursions northward or southward of the
ITCZ mean position. In this scenario, with high insolation
in June-July-August, positive anomalies would be expected
throughout North Africa, as a more northward ITCZ causes
the growing season to begin early and rainfall to extend
northward [Brown and deBeurs, 2008]. With high insolation
in December-January-February, the situation would be
reversed, and the ITCZ would be southward.
[26] In order to investigate how the position of ITCZrelated convection may be related to the vegetation regression pattern, 500 mb vertical velocity fields were correlated
with the December-January-February and June-July-August
insolation time series. Vertical velocity is a metric of vertical
motion in the atmosphere with negative values indicating
rising air masses. Thus, the mean ITCZ positions for
December-January-February and June-July-August are
represented in Figures 6a and 6c by an area of minimum
vertical velocity. Changes in this mean position of rising
air associated with the ITCZ are therefore represented by
areas of negative correlation with the insolation time
series, indicating more lofting air masses, and areas of
positive correlations indicating reduced lofting with respect
to insolation.
8
IVORY ET AL.: AFRICAN CLIMATE AND VEGETATION
Figure 8. NDVI regressions on the Southern Oscillation Index with mean covariance (COV), as
displayed in Table 1, at the bottom right of each panel. Regression map of June-July-August NDVI on
December-January-February Southern Oscillation Index. Double asterisks indicate 99% confidence and
single asterisk represents 95% confidence.
extension of the South Atlantic anti-cyclone mirroring the
northward ITCZ position as indicated by a strong area
positive correlation in the Southern Atlantic.
[30] Tropical vegetation, particularly trees in areas with
prolonged dry seasons, store starch during the growing season.
Changes in ITCZ position alter the length of the dry season
and growing season when deciduous trees hold their leaves
and when vegetation is productive [Brown and deBeurs,
2008]. A time series of southern hemisphere (10–20°S;
0–40°E) and northern hemisphere (10–20°N; 0–40°E) dry
days (days/year with < 0.5 mm rain) was created as a
metric for the effect of length on vegetation. Figures 7a and 7b
show the regression maps yielded by regressing DecemberJanuary-February and June-July-August NDVI on southern
and northern hemisphere dry days, respectively. Both
patterns are significantly similar, but of opposing sign, to
the December-January-February and June-July-August insolation regression patterns (Figures 3a and 3b). This suggests that
higher insolation displaces the ITCZ toward the summer
hemisphere and produces a shorter dry season and a longer
growing season leading to increased vegetation productivity
(Figure 8a).
[31] Studies of hydrology typically focus around mean
annual precipitation (MAP); however, vegetation productivity
is a function of effective moisture [Nicholson, 1986]. We
propose that insolation controls vegetation productivity
through two mechanisms when summer insolation is high:
greater rainfall intensity and shorter dry season length due to
a displacement of the ITCZ and subtropical anti-cyclones
toward the summer hemisphere. Although enhanced convective precipitation associated with high summer insolation
may help drive these vegetation anomalies, the relationship
connecting rainy season productivity to the length of the
preceding dry season should not be ignored. This is consistent
with modeling results of Hély et al. [2006], demonstrating
increased sensitivity of semi-deciduous and deciduous biome
types to changes in dry season length.
4.2. Hadley Circulation and ENSO
[32] The OLR time series produces the highest value of
covariance of all the time series when regressed on NDVI
for June-July-August (Table 1). In addition, the EOF plot
implies that anomalies of this variable are greater toward
the subtropics (Figure 1). Thus, the June-July-August OLR
pattern when regressed on NDVI results from changes in
atmospheric circulation of the same sign throughout the
tropics (Figures 1 and 4). Furthermore, although the troposphere height time series never yields covariance values with
greater than 90% significance, the PCA biplot suggests an
important antiphased relationship between these two variables
(Figure 5). When troposphere height is high, OLR is low.
[33] The ITCZ can vary in intensity, mean position, and
width [Nicholson, 1996]. The gradient of OLR from the
equator to the subtropics, as observed by the EOF, suggests
that the vegetation response in June-July-August is related to
tropical belt width (Figure 1). Thus, the negative anomalies
near the subtropics and positive anomalies near the equator
are due to a narrower ITCZ during summers when OLR is
high (Figure 4).
[34] This is supported by correlation of June-July-August
OLR with June-July-August 500 mb vertical velocity.
Figure 6f shows a positive correlation, indicating reduced
convection, covering much of northern and central Africa.
This circulation pattern would dry the subtropics and wet
the equatorial region, as the ITCZ does not reach as far north.
[35] This is further supported by positive NDVI anomalies
persisting even in the southern hemisphere during the JuneJuly-August dry season, suggesting that enhanced precipitation
near the equator influences vegetation into the dry season
(Figure 4). This is perhaps the result of enhanced starch storage,
common in southern African woody plants in years with a
wetter, more productive rainy season [Rutherford, 1984].
Although the regression yields a 95% significant value only
in northern hemisphere summer, numerous 90% significant
values for regression of troposphere height and OLR
9
IVORY ET AL.: AFRICAN CLIMATE AND VEGETATION
Figure 9. Location of ITCZ (cloud position) and generalized patterns of vegetation anomalies
(green = positive, brown = negative). These patterns are created by (a) insolation variability producing
vegetation anomalies which are antiphased between the north and south hemispheres due to ITCZ excursions,
and (b) OLR/troposphere height variability producing vegetation anomalies which are symmetrical about the
equator due to ITCZ width changes.
on interannual vegetation variability throughout Africa.
Rainy season moisture convergence and dry season length
have a profound effect on vegetation in both northern and
southern tropical Africa. In contrast, commonly cited mechanisms, like SSTs, do not exhibit a strong connection.
[39] ITCZ variability and the African hydrologic budget
are tightly connected to ENSO and insolation on short
and long time scales (Figure 9). High summer insolation
produces in ITCZ excursions resulting in a shorter dry season
and positive vegetation anomalies in the summer hemisphere
(Figure 9a). Higher atmospheric heights and a wider tropical
belt during warm ENSO events result in a symmetrical
vegetation pattern centered about the equator (Figure 9b).
[40] Although we support the suggestion that ENSO is the
dominant mode of variability in the African tropics, our study
points to a strong, direct link between atmospheric circulation,
rather than SSTs, and vegetation productivity [Schreck and
Semazzi, 2004]. One third of the African population lives in
semiarid regions dependent on natural and rainfed agricultural
resources especially sensitive to variability [IPCC, 2007]. It is
yet unknown how ENSO frequency and amplitude will change
with future warming [Collins et al., 2005; Yamaguchi and
Noda, 2006]; however, it will almost certainly have a dramatic
effect on African vegetation. Although the latest generation of
climate models accurately simulate many aspects of atmospheric dynamics, inaccuracies of ENSO behavior and the
ITCZ will hinder accurate predictions of future vegetation
change, effects of tropical vegetation on the carbon and hydrologic cycles, and simulations of past vegetation dynamics.
[41] Despite the small amplitude changes over the time
period studied (~0.25 W/m2 or 0.018%) [Shindell et al., 2006],
insolation has a marked effect on tropical atmospheric circulation and vegetation productivity. The effect of insolation on
short time scales has been little investigated despite suggestions that very small solar irradiance changes may influence
tropical rainfall patterns [Shindell et al., 2006]. On longer
time scales, high-amplitude insolation variability is likely
the primary mechanism of Afrotropical vegetation change.
suggest that this relationship may also exist for the short
and long rainy seasons (March-April-May and OctoberNovember-December) as well as in December-JanuaryFebruary (Table 1).
[36] Such changes in atmospheric circulation throughout the
tropics may suggest a link to climate modes on a global scale.
Regression of the Southern Oscillation Index on June-JulyAugust NDVI produces a pattern that is significantly similar
to June-July-August OLR (Figure 8). This supports many
studies linking ENSO to vegetation and rainfall anomalies
throughout Africa [Anyamba and Eastman, 1996; Nicholson
and Selato, 2000]. Higher troposphere heights throughout
the tropics during ENSO events is consistent with results
of our PCA, showing an antiphased relationship between
OLR and troposphere height [Camberlin et al., 2001].
Furthermore, ENSO warm event years (1988, 1999) on the
PCA biplot occur near the positive end of the second axis near
both seasons of troposphere height (Figure 5). Cold events,
associated with higher OLR and lower troposphere height,
would produce a pattern similar to Figure 4. During warm
events, the relationship is reversed when OLR is lower.
[37] Our study points to a strong, direct link between atmospheric circulation during ENSO events and African vegetation variability [Schreck and Semazzi, 2004]. Since the
Southern Oscillation Index is calculated based on surface
pressure, rather than SST anomalies, we observed that the
immediate effect of ENSO on African vegetation is not
related directly to SSTs but rather atmospheric height anomalies. The north-south symmetry of the OLR and Southern
Oscillation Index patterns in NDVI is consistent with studies
illustrating the effect of cold events on the ITCZ that
promotes reduced convergence at the northern and southern
extremes during rainy seasons [Janicot, 1997].
5.
Conclusions
[38] Despite heterogeneous MAP and vegetation, global
atmospheric controls on the ITCZ have a strong influence
10
IVORY ET AL.: AFRICAN CLIMATE AND VEGETATION
Camberlin, P., S. Janicot, and I. Poccard (2001), Seasonality and atmospheric dynamics of the teleconnection between African rainfall and
tropical sea-surface temperature: Atlantic vs. ENSO, Int. J. Climatol.,
21, 973–1005, doi:10.1002/joc.673.
Chiang, J. (2009), The tropics in paleoclimate, Annu. Rev. Earth Planet. Sci.,
37, 263–297, doi:10.1146/annurev.earth.031208.100217.
Cohen, A., et al. (2007), Ecological consequences of early Late Pleistocene
megadroughts in tropical Africa, Proc. Natl. Acad. Sci. U.S.A., 104,
16,422–16,427, doi:10.1073/pnas.0703873104.
Collins, M., C. M. I. P. The, and M. Groups (2005), El Niño- or La Niña-like
climate change, Clim. Dyn., 24, 89–104, doi:10.1007/s00382-004-0478-x.
Davenport, M., and S. Nicholson (1993), On the relation between rainfall
and the Normalized Difference Vegetation Index for diverse vegetation
types in East Africa, Int. J. Remote Sens., 14, 2369–2389, doi:10.1080/
01431169308954042.
DeMenocal, P., and D. Rind (1993), Sensitivity of Asian and African climate to
variations in seasonal insolation, glacial ice cover, sea-surface temperature, and
Asian orography, J. Geophys. Res., 98, 7265–7287, doi:10.1029/92JD02924.
Fabricante, I., M. Oesterheld, and J. M. Paruelo (2009), Annual and seasonal
variation of NDVI explained by current and previous precipitation across
Northern Patagonia, J. Arid Environ., 73, 745–753, doi:10.1007/s10265009-0302-0.
Few R., M. Ahern, F. Matthies, and S. Kovats (2004), Floods. health and
climate change: A strategic review, in Working Paper 63, Tyndall Centre
for Climate Change Research, University of East Anglia, Norwich, pp. 138.
W. W. Forum (2000), The Africa Water Vision for 2025: Equitable and
sustainable use of water for socioeconomic development. UN Water/Africa,
pp. 34.
Giannini, A., M. Biasutti, I. Held, and A. Sobel (2008), A global perspective on
Africa climate, Clim. Change, 90, 359–383, doi:10.1007/s10584-008-9396-y.
Hély, C., L. Bremond, S. Alleaume, B. Smith, M. Sykes, and J. Guiot (2006),
Sensitivity of African biomes to changes in the precipitation regime, Global
Ecol. Biogeogr., 15, 258–270, doi:10.1111/j.1466-822x.2006.00235.x.
Hoerling, M., J. Hurrell, J. Eischeid, and A. Phillips (2006), Detection and
attribution of twentieth-century northern and southern African rainfall
change, J. Clim., 19, 3989–4008, doi:10.1175/JCLI3842.1.
Hountondji, Y. C., N. Sokpon, J. Nicolas, and P. Ozer (2009), Ongoing
desertification processes in the sahelian belt of West Africa: An evidence
from the rain-use efficiency, in Recent Advances in Remote Sensing and
Geoinformation Processing for Land Degradation Assessment, Int. Soc.
for Photogramm. and Remote Sens., edited by R. Achim and H. Joachim,
pp. 173–186, Taylor and Francis, London, U. K.
Hulme, M., R. Doherty, and T. Ngara (2001), African climate change:
1900–2100, Clim. Res., 17, 145–168, doi:10.3354/cr017145.
Intergovernmental Panel on Climate Change (IPCC) (2007), Climate
Change 2007: The Physical Science Basis and Impacts, Adaptation, and
Vulnerability. Contribution of Working Groups I and II to the Fourth
Assessment Report of the Intergovernmental Panel on Climate Change,
Cambridge Univ. Press, Cambridge, U. K.
Janicot, S. (1997), Impact of warm ENSO events on atmospheric circulation
and convection over the tropical Atlantic and West Africa, Ann. Geophys.,
15, 471–475, doi:10.1007/s00585-997-0471-x.
Jolliffe, I. T. (1990), Principal component analysis: A beginner’s guide — I.
Introduction and application, Weather, 45, 375–382, doi:10.1002/j.14778696.1990.tb05558.x.
Kalnay, E., et al. (1996), The NCEP/NCAR 40-year reanalysis project, Bull.
Am. Meteorol. Soc., 77, 437–470, doi:10.1175/1520-0477(1996)
077<0437:TNYRP>2.0.CO;2.
Martiny, N., P. Camberlin, Y. Richard, and N. Philippon (2006), Compared
regimes of NDVI and rainfall in semi-arid regions of Africa, Int. J. Remote
Sens., 27, 5201–5223, doi:10.1080/01431160600567787.
McCabe, G., M. Clark, and M. Serreze (2001), Trends in Northern
Hemisphere surface cyclone frequency and intensity, J. Clim., 14,
2763–2768, doi:10.1175/1520-0442.
McClean, C., J. Lovett, W. Küper, L. Hannah, J. Sommer, W. Barthlott,
M. Termansen, G. Smith, S. Tokumine, and J. Taplin (2005), African plant
diversity and climate change, Ann. Mo. Bot. Gard., 92, 139–152.
McHugh, M. (2004), Near-surface zonal flow and east African precipitation
receipt during austral summer, J. Clim., 17, 4070–4079, doi:10.1175/
1520-0442.
Meng, M., J. Ni, and M. Zong (2011), Impacts of changes in climate variability
on regional vegetation in China: NDVI-based analysis from 1982 to 2000,
Ecol. Res., 26, 421–428, doi:10.1007/s11284-011-0801-z.
Mercier, F., A. Cazenave, and C. Maheu (2002), Interannual lake level
fluctuations (1993–1999) in Africa from Topex/Poseidon: Connections with
ocean-atmosphere interactions over the Indian Ocean, Global Planet.
Change, 32, 141–163, doi:10.1016/S0921-8181(01)00139-4.
New, M., B. Hewitson, C. Jack, and R. Washington (2003), Sensitivity of
southern African rainfall to soil moisture, CLIVAR Exchanges, 27,
45–47.
This supports the many studies showing large-scale ecological
and hydrologic changes in Africa that follow the beat of
insolation [Rossignol-Strick, 1983; Partridge et al., 1997;
Cohen et al., 2007; Vershuren et al., 2009].
[42] Our analyses suggest that today ITCZ width is mostly
controlled by ENSO variability, but future warming may also
affect ITCZ width. A warmer atmosphere from anthropogenic
forcing results in higher troposphere height and wider migrations, wetting the outskirts of the tropics and drying the equatorial region. Seidel et al. (2008) have already documented this
widening of the tropics. This has particular implications, such
as crop failure and desertification, for semi-arid vegetation
that is extremely sensitive to changes in dry season length
[Hély et al., 2006; McClean et al., 2005].
[43] In addition, southern Africa depends on the placement
of mid-latitude westerlies and semi-permanent anti-cyclones
determining the eastward penetration of moisture into
sensitive semi-arid areas like East Africa [McHugh, 2004].
Recent studies suggesting a poleward westerlies displacement as the atmosphere warms imply that these regions could
see a marked drying as moisture needed to drive convergence
may no longer be available [McCabe et al., 2001].
[44] In much of Africa, nothing is more important to human
lives and livelihoods than the seasonal rains following the sun
each year and the plants sustained by this rainfall. Although
the human relationship with plants has changed over time
in Africa, the overarching climate mechanisms controlling
landscape distribution and productivity have not. The quantitative constraints we provide here may give some insight
concerning how climate change will and did affect the
distribution of important ecosystems and the climate regimes
under which crops thrive.
[45] Acknowledgments. We thank the National Science Foundation
for a Graduate Research Fellowship (2009078688); Paul Goodman and
Michael McGlue at the University of Arizona for technical help and critical
discussions; and Compton Tucker and Jorge Pinzon at NASA GIMMS for
access to the AVHRR NDVI data set.
References
Anyah, R., and F. Semazzi (2004), Simulation of the sensitivity of Lake
Victoria basin climate to lake surface temperatures, Theor. Appl.
Climatol., 79, 55–69, doi:10.1007/s00704-004-0057-4.
Anyamba, A., and J. Eastman (1996), Interannual variability of NDVI over
Africa and its relation to El Niño/Southern Oscillation, Int. J. Remote
Sens., 17, 2533–2548, doi:10.1080/01431169608949091.
Anyamba, A., C. Tucker, and R. Mahoney (2002), From El Niño to La Niña:
Vegetation response patterns over east and southern Africa during the
1997–2000 period, J. Clim., 15, 3096–3103, doi:10.1175/1520-0442
(2002)015<3096:FENOTL>2.0.CO;2.
Beck, H. E., T. R. McVicar, A. van Dijk, J. Schellekens, R. de Jeu, and
L. Bruijnzeel (2011), Global evaluation of four AVHRR–NDVI data sets:
Intercomparison and assessment against Landsat imagery, Remote Sens.
Environ., 115, 2547–2563, doi:10.5194/hess-16-3835-2012.
Boko, M., I. Niang, A. Nyong, C. Vogel, A. Githeko, M. Medany,
B. Osman-Elasha, R. Tabo, and P. Yanda (2007), Africa, in Climate Change
2007: Impacts, Adaptation and Vulnerability. Contribution of Working
Group II to the Fourth Assessment Report of the Intergovernmental Panel
on Climate Change, edited by M. L. Parry et al., pp. 433–467, Cambridge
Univ. Press, Cambridge, U. K.
Bradley, R. S. (1985), Quaternary Palaeoclimatology: Methods of
Palaeoclimatic Reconstruction, pp. 472, Unwin Hyman, London.
Brown, M., and K. deBeurs (2008), Evaluation of multi-sensor semi-arid
crop season parameters based on NDVI and rainfall, Remote Sens.
Environ., 112, 2261–2271, doi:10.1016/j.rse.2007.10.008.
Brown, M., K. deBeurs, and A. Vrieling (2010), The response of African
land surface phenology to large scale climate oscillations, Remote Sens.
Environ., 114, 2286–2296, doi:10.1016/j.rse.2010.05.005.
11
IVORY ET AL.: AFRICAN CLIMATE AND VEGETATION
Rutherford, M. (1984), Relative allocation and seasonal phasing of growth of
woody plant components in a South African savanna, Progress in
Biometeorology, 3, 200–221.
Schreck, C., III, and F. Semazzi (2004), Variability of the recent climate of
eastern Africa, Int. J. Climatol., 24, 681–701, doi:10.1002/joc.1019.
Seidel, D., Q. Fu, W. Randel, and T. Reichler (2008), Widening of the
tropical belt in a changing climate, Nat. Geosci., 1, 21–24, doi:10.1038/
ngeo.2007.38.
Shindell, D., G. Faluvegi, R. Miller, G. Schmidt, J. Hansen, and S. Sun
(2006), Solar and anthropogenic forcing of tropical hydrology, Geophys.
Res. Lett., 33, L24706, doi:10.1029/2006GL027468.
Tucker, C., J. Pinzon, M. Brown, D. Slayback, E. Pak, R. Mahoney,
E. Vermote, and N. Saleous (2005), An extended AVHRR 8 km NDVI
dataset compatible with MODIS and SPOT vegetation NDVI data, Int. J.
Remote Sens., 26, 4485–4498, doi:10.1080/01431160500168686.
Vershuren, D., J. S. Damsté, J. Moernaut, I. Kristen, M. Blaauw, M. Fagot,
G. Haug, and CHALLACEA project (2009), Half-precessional dynamics
of monsoon rainfall near the East African Equator, Nature, 462,
637–641, doi:10.1038/nature08520.
White, F. (1983), Vegetation of Africa: A descriptive memoir to accompany
the Unesco/AETFAT/UNSO vegetation map of Africa; Natural Resources
Research Report XX; UNESCO, Paris, France, 356 pp.
Xu, D. Y., X. W. Kang, D. F. Zhuang, and J. J. Pan (2010), Multi-scale quantitative assessment of the relative roles of climate change and human activities in desertification—A case study of the Ordos Plateau, China, J. Arid
Environ., 74, 498–507, doi:10.1016/j.jaridenv.2009.09.030.
Yamaguchi, K., and A. Noda (2006), Global warming patterns over the
North Pacific: ENSO versus AO, J. Meteorol. Soc. Japan, 84, 221–241.
Nicholson, S. (1986), The spatial coherence of African rainfall anomalies:
Interhemispheric teleconnections, J. Climate Appl. Meteorol., 25,
1365–1381, doi:10.1175/1520-0450.
Nicholson, S. (1996), A review of climate dynamics and climate variability
in eastern Africa, in The Limnology, Climatology, and Paleoclimatology
of the East African Lakes, edited by T. C. Johnson and E. O. Odada,
pp. 664, Gordon and Breach, London.
Nicholson, S. E., and J. P. Grist (2003), The seasonal evolution of the atmospheric circulation over West Africa and Equatorial Africa, J. Clim., 16,
1013–1030, doi:10.1175/1520-0442(2003)016<1013:TSEOTA>2.0.CO;2.
Nicholson, S., and J. Selato (2000), The influence of La Niña on African
rainfall, Int. J. Climatol., 20, 1761–1776, doi:10.1002/1097-0088
(20001130)20:14.
Partridge, T., et al. (1997), Orbital forcing of climate over South Africa: A
200,000-year rainfall record from the Pretoria Saltpan, Quat. Sci. Rev., 16,
1125–1133, doi:10.1016/S0277-3791(97)00005-X.
Rayner, N., D. Parker, E. Horton, C. Folland, L. Alexander, D. Rowell,
E. Kent, and A. Kaplan (2003) Global analyses of sea surface temperature,
sea ice, and night marine air temperature since the late nineteenth century,
J. Geophys. Res., 108(D14), 4407, doi:10.1029/2002JD002670.
Revadekar, J. V., Y. K. Tiwari, and K. Ravi Kumar (2012), Impact of climate
variability on NDVI over the Indian region during 1981–2010, Int. J.
Remote Sens., 33, 7132–7150, doi:10.1080/01431161.2012.697642.
Rossignol-Strick, M. (1983), African monsoons, an immediate climate
response to orbital insolation, Nature, 304, 46–49, doi:10.1038/304046a0.
Russell, J., and J. Wallace (2004), Annual carbon dioxide drawdown and the
Northern Annular Mode, Global Biogeochem. Cycles 18, GB1012,
doi:10.1029/2003GB002044.
12