A proxy rainfall record for northeastern South Africa based on carbon isotope analysis of
four baobab (Adansonia digitata L.) trees shows centennial and decadal scale variability
over the last 1,000 years. The record is in good agreement with a 200-year tree ring record
from Zimbabwe, and it indicates the existence of a rainfall dipole between the summer and
winter rainfall areas of South Africa. The wettest period was c. AD 1075 in the Medieval
Warm Period, and the driest periods were c. AD 1635, c. AD 1695 and c. AD1805 during the
Little Ice Age. Decadal-scale variability suggests that the rainfall forcing mechanisms are a
complex interaction between proximal and distal factors. Periods of higher rainfall are significantly
associated with lower sea-surface temperatures in the Agulhas Current core region
and a negative Dipole Moment Index in the Indian Ocean. The correlation between rainfall
and the El Niño/Southern Oscillation Index is non-static. Wetter conditions are associated
with predominantly El Niño conditions over most of the record, but since about AD 1970 this
relationship inverted and wet conditions are currently associated with la Nina conditions.
The effect of both proximal and distal oceanic influences are insufficient to explain the rainfall
regime shift between the Medieval Warm Period and the Little Ice Age, and the evidence
suggests that this was the result of a northward shift of the subtropical westerlies rather than
a southward shift of the Intertropical Convergence Zone.
S1 Fig. Age models for baobab trees. (A) The ages assigned to isotope samples from core samples
taken from baobab trees were determined from linear interpolations of core length (xaxis)
with AMS radiocarbon dates. The 1-sigma radiocarbon error ranges are portrayed as vertical
lines or as crosses (for bomb-carbon dates). (B) For trees that yielded ring structures the
ring count (x-axis) matched the 1-simga AMS radiocarbon ages (vertical lines or crosses) with
a 1:1 except where a buttress forms in one of the trees.
S2 Fig. The baobab isotope record and rainfall. Instrumental rainfall from the Pafuri station
(orange, left axis) and the CRU3.20 rainfall for the region (red, left axis) cannot be correlated
with the baobab δ13C record (blue, right axis) because of systemic errors in the age model.
S3 Fig. Comparison of ENSO proxy datasets. The ENSO proxy datasets of Li et al.  (dark
blue), Cook et al.  (red), Stahle and Cleavland  (pale bue) and Braganza et al.  (purple)
are coherent with the Niño3.4 index (http://www.cpc.ncep.noaa.gov/data/indices/sstoi.
indices) (black). Only a 150-year section of the record is portrayed for clarity. Note that the indices
of Briganza et al. and Stahle et al. are plotted on the inverted right axis because of the
manner in which they are formulated.
S4 Fig. Oceanic sea-surface temperature trends. SST reconstructions for the Makassar Stait
of Oppo et al.  (black, left axis) and Newton et al.  (blue, left axis) match the timing and
relative temperature changes that took place in the Agulhas Current core region [55, 56] (red,
right axis). The Ifaty record is shown along with the biweight mean value.
S1 Table. Correlation of baobab δ13C with environmental parameters.