Late Quaternary vegetation history and climatic change in the eastern Orange Free State, South Africa

S.Afr.J. Bot. , 1989 , 55(1): 107- 116

107

Late Quaternary vegetation history and climatic change in the eastern Orange Free State, South Africa L. Scott Department of Botany, University of the Orange Free State, P.O. Box 339, Bloemfontein, 9300 Republic of South Africa

Accepted 25 July 1988 Sixty-nine pollen spectra from three sites with alluvial and swamp deposits and 12 spectra representing the modern vegetation in the eastern Orange Free State and Drakensberg areas, provide evidence about the Late Quaternary environmental history of Clarens. Factor analysis is used to compare pollen data between sites. Stepwise changes in the vegetation at Clarens are indicated, relating to changes in climatic variables during the Late Pleistocene and Holocene. The earliest palaeo-environmental indications recorded immediately north of Clarens are of humid grassland with strong fynbos and swampy conditions ca. 23000 years BP, followed by progressively cooler, drier conditions ca. 22 600 to 19000 years BP. Grass in the vegetation became more prominent between ca. 18 000 and 17 000 years BP. Shrubby grassland with some fynbos elements occurred in the Little Caledon Basin ca. 13000 years BP, under slightly warmer conditions which were relatively moist, probably also during winters. Podocarpus forests developed in distant mountain ravines but together with fynbos elements, declined before 10 700 years BP, possibly under the influence of slightly warmer conditions and growing seasonal contrasts. The Holocene record is incomplete, starting apparently with grassland indications for warm conditions, and ending with subhumid, markedly seasonal conditions and local swamp development as found along the Little Caledon River. Nege-en-sestig stuifmeelspektra van alluviale en vleiafsettings van drie plekke en 12 spektra van die moderne plantegroei in die Oos-Vrystaat en die Drakensberggebied, gee bewyse vir die omgewingsgeskiedenis van Clarens. Die stuifmeeldata van die verskillende plekke word vergelyk deur faktoranalise. Die plantegroei by Clarens het stapsgewys verander onder die invloed van wisselings in klimaat gedurende die Laat Pleistoseen en Holoseen. Die vroegste paleo-omgewingsaanwysings, afkomstig van net noord van Clarens, is van vogtige grasveld met fynbos en moerastoestande, ca. 23 000 jaar voor hede, gevolg deur progressief koeler, droer toestande, ca. 22 600 tot 19 000 jaar voor hede. Gras in die vegetasie was meer prominent tussen ca. 18 000 en 17 000 jaar voor hede. Bossie-ryke grasveld met fynbos onder effens warmer toestande, wat relatief vogtig was, ook gedurende winters, het teen ca. 13000 jaar voor hede in die Klein Caledon-Vallei voorgekom. Podocarpus-woude het in afgelee klowe ontwikkel maar het saam met fynbos elemente begin kwyn voor 10 700 jaar voor hede, moontlik as gevolg van effens warmer toestande en groeiende seisoenale kontraste. Die Holoseen-rekord is onvolledig, maar begin waarskynlik met grasveld en toon indikasies vir warm toestande; dit eindig met half-vogtige, sterk seisoenale toestande en lokale moerasvorming so os die langs die Klein Caledon-Rivier. Keywords: Late Quaternary, palaeoclimate, pollen analysis, South Africa, vegetation change

Introduction The vegetation in the mountainous Clarens area is mainly grassland , experiencing heavy frost in winter. A shrubby ericoid component (termed 'fynbos' for the purpose of this paper) occurs on mountain tops and slopes at relatively high altitudes. In protected ravines and especially on some south-facing slopes along mountains in the Clarens area, montane woodland with dense growths of shrubs and trees are found. Podocarpus forests, however , only occur ca. 50 km further eastwards in ravines of the wetter Drakensberg . The annual precipitation in the Clarens area is ca. 700 mm occurring mainly in summer , but increases sharply towards the east and south-east to more than 1 400 mm within 50 km, along the Natal Drakensberg escarpment. Towards the west of the study area the rainfall declines gradually to ca. 560 mm around Bloemfontein, which is about 250 km away . The topography in the Clarens area is controlled by bedrock consisting largely of sandstone of the Clarens Formation underlying peaks of basaltic lava of the

Drakensberg Group. Both units overlie mudstones and sandstones of the Elliot Formation, which are exposed along slopes of deep valleys in which Quaternary alluvium accumulated . According to Visser & van Riet Lowe (1955) who studied the recent geology and Stone Age archaeology in the Little Caledon Valley (Figure 1), the nature of alluvial deposits can be ascribed to changing precipitation patterns and climatic fluctuations during the Pleistocene in the area. Palaeoclimatic changes have provisionally been indicated during the Late Quaternary by pollen in sediments exposed at two sites (Craigrosse and Cornelia, Figure 1), along stream and erosional channels in the Clarens area (Scott 1986) . Some problems were, however, encountered with the interpretation of these palynological data partly as a result of anomalous radiocarbon dates. Visser et al. (1986) studied the environmental context of a vivianite occurrence on the farm Elim (Figure 1) north of the town . More pollen spectra in swamp deposits from this site were studied and dated by radiocarbon, and the results are presented in this paper and compared with

108

S.-Afr.Tydskr. Plantk., 1989 , 55(1)

IOOKM

• JOHANNESBURG

I

TRANSVAAL

~ ~~~ ,•..

...

ORANGE FREE STATE

:Y2000m LESOTHO

NATAL

30"

40 ALIWAL NORTH

-.----

Figure I

Map of study area showing sediment sites and localities of surface pollen samples (1-12).

previous results. As a control, 12 modern surface pollen spectra from various vegetation types in the Clarens and Drakensberg areas are described. A revision of the earlier palynological interpretations of Scott (1986) is made and on the basis of the new results a more detailed Late Quaternary palaeo-enviromental chronology for the eastern Orange Free State can be reconstructed. Methods Samples for pollen analysis and radiocarbon dating were

collected from previously cleaned exposures of alluvial and swamp deposits at the three sites mentioned. Modern pollen samples consist of one hyrax dung sample and 11 surface scrapings of soils at different localities between ca. 1 700 and 2 700 m altitude, in the Clarens , Natal and Lesotho areas (Figure 1, Table 1). Pollen was extracted from sediments by means of standard laboratory techniques and the residues were mounted in glycerine jelly for microscope analysis and pollen counts. Samples for radiocarbon dating were submitted to Dr

109

S.Afr.J. Bot., 1989,55(1)

Table 1 sample

Surface pollen samples. 9

Hyrax dung

Vegetation

Table 2 area

Geology

Depth (cm)

No. Alt (m)

Locality

I 2 3 4

Lesotho

Upland fynbos vegetation

Basalt

Craigross ie

Natal

Upland grassland with fynbos

Basalt

Organic loam

Natal

Upland grassland with fynbos

Basalt

Brown silt

Natal

Upland grassland with Protea savanna

Basalt

Brown silt

2700 2560 2 120 2000

Brown silt

Natal

Small Podocarpus forest

Sandstone

Brown silt

Golden Gate O.F.S.

Grassland on steep north-facing slope

Basalt

Brown silt

1 950

C larens O.F.S.

Upland grassland with open montane woodland

Sandstone

8

1920

Clarens O.F.S.

Upland grassland and shrubland

Sandstone

9

I 890

Clarens O.F.S.

Montane woodland and shrubland

Sandstone

10

I 750

Little Ca ledon Valley, O.F.S.

Grassland

Siltstone

11

1 735

Little Calcdon River, O.F.S.

Swamp vegetation with Phragmites

Siltstone

12

I 730

Little Ca ledon River,O.F.S.

Stream terrace vegetation with Artemisia

Siltstone

5 6

I 950 2100

7

Surface pollen samples The locations and descriptions of the 12 surface pollen samples (1-12) are shown in Figure 1 and Table 1 respectively. The results of pollen analysis of these samples are shown in Figure 2. High peaks of especially Gramineae (Poaceae), Cyperaceae, Ericaceae and Podocarpus pollen at the different sites reflect vegetation at these localities. Generally the Gramineae make up the most important component of the pollen spectra. Ericaceae and other 'fynbos' elements such as Passerina seem to be relatively important in high-altitude sites. The only exception is that of the high-lying sample (number 6 near Golden Gate) from a steep north-facing slope dominated by Gramineae. Highest values for Cyperaceae were obtained for swamp sites near Craigrossie (samples 11 and 12) while relatively prominent values of this pollen above 2 000 m are probably derived from inconspicuous species of Cyperaceae occuring on slopes. Compositae plants are prominent on some slopes, but the high peak at 1 920 m near Clarens (sample 8), is probably attributable to local overgrazing. Although surface pollen spectra are useful in palynological studies, they have limitations. Pollen spectra from a dated hyrax midden sequence resting on sandstone cliffs at the Clarens village (same location as surface sample 9), show that vegetation in the area has changed considerably during the last 20 years, and this is supported by a series of five aerial photographs taken at different times between 1947 and 1984 (Scott 1988a and unpublished data). During 20 years the number and density of trees and shrubs increased markedly, while the composition of the vegetation understorey also

BOTANY-H

Dark grey si lt

14C date year (BP)

Pta 3842

660 ± 60 10700 ± 130 9460±110 10600 ± 120 10400 ± 140 10 600 ± 100

Pta 3389 Pta 3686 Pta 3675 Pta 3383 Pta 3682

-330 -380

Pta 3845 Pta 3674

12600±110 20000 ± 300

Elim Black silt Black silt Black silt

Vogel, National Physical Research Laboratory, CSIR, Pretoria.

-81 to 88 -171 to 181 - 176 to 186 -296 to 306 -466 to 476 -489 to 496

Lab. no

Cornelia Black silt

J.e.

Radiocarbon dates from the Clarens

Black silt

-350 -365 -540 -580

Pta 4111 Pta 4357 Pta 4374 Pta 4359

20400 21 500 22 600 23400

± ± ± ±

620 300 240 340

changed. These findings suggest that caution is necessary in the use of surface pollen spectra as a control for interpretations of fossil spectra, especially in areas where farming is practised. The Elim sequence The Elim site (Figure I) at the upper reaches of the As River in the Rooiberge, was studied by Visser et al. (1986), in order to record vivianite in swamp deposits at this locality. Two distinct sediment units were recognized in the described sequence, situated on a northfacing slope of more than 20°, viz.: (1) a basal unit, :'S 6 m thick, consisting mostly of dark-coloured silts with lighter streaks of sandy material, and (2) an upper channel fill of light-coloured coarse-grained sediment of ca. 1,3 m thickness. Pollen samples from the lower unit (section C, Figure 3) studied in this report, cover a wider lateral and temporal interval than the locally truncated section described by Visser et al. (1986). Four radiocarbon dates of organic silts from the bottom 2,5 m cover the time interval between ca. 23 400 years BP and 20400 years BP (Table 2). By extrapolation, the 18 000year BP level, which is presumed to be the time of the coldest period of the last glacial maximum in some parts of the world, is likely to be located in the upper 3,5 m. Visser et al. (1986) described three 'cycles' in the lower unit, each consisting of a basal sandy part overlain by dark grey to black silt layers. Vivianite occurs in the third cycle in a distinct dark band at the ca. 2,5-m level of section e. No unconformity is visible between the top of the third cycle and the rest of the upper part of section C, which consist of more light and dark sand layers and which is unconformably overlain by modern light sandy soil. The Holocene sandy channel fill described by Visser et al. (1986) is only represented by one pollen sample in the present report (section A, Figures 3 & 6), as other levels in the unit do not contain pollen. The studied sample

S.-Afr.Tydskr. Plantk., 1989,55(1)

110 UJ n. ~ > UJ fUJ I « ~ « UJ en UJ U en ..J UJ « a: ::la: n. a: ::l U n.

UJ 0 ::l f-

en;:!

NATALI LESOTHO AREA

CLARENS AREA

I

2700 I

2

2560 I

3

2120 •

4

2000. I 1950 •

6

2100

7

1950

8

1920

9 10

1890 1750

"

1735

12

I

I

UJ

:!

«UJf-

a:

5

_..J~

!.1

>

g..J

UJ d::l IOU

"

~ UJ

~

UJ«

~d

Z

UJ « UJ

~ ~ ~ « I enz (ii ::lUJ a: ~ ~ > ~ffi :i! ~ ;:!

•

.,

I

I

••

••

" 0 U

2700_

2

2560_

3

2120

4

2000 •

5

1950 I

6 7

2100 • 1950 _

8 9

1920 1890 _

10

1750_

"

1730 I

17"35

UJ « UJ «

(ii

U ::> fU

~

UJ f0::

:3

«

•

a: UJ n.

> fI

~g

UJ UJ

10

Ia: f-I

f2

~~

(/)

I

UJ n.

:i! ::l :i!

UJ n.

0 (5

-

I

~UJ

..J~

~

~

- -- -« a: (!)

« a: UJ z z ::l (!)

::lz

~~ x

« I n.

a: UJ n. > U

~ ~z ::l ::If-

>

a:~

f-

a:

)-

«

F

~

I

I

I

• • •

en

U

~

cr UJ

(/)

~

U

•

• • I

UJ « UJ

crUJ

a:

u.. :J

I

I

« z

-I

•

I

•

•••

1730

~

UJ « UJ z

«

UJ «UJ UJ«

UJ « UJ

•

••

I

(ii

12

x

0 0 n.

«

I

UJ

CLARENS AREA

g

• • •• ;::

NATALI LESOTHO AREA

~Ci

::l I f-

I

5

•

UJ U

a:

g

5« ~~

en

~

en ~

fUJ « UJ

3 N

«

• • I I I

UJ « UJ

UJ « UJ U «

0::> ZID UJ_ I 0:: Uf-

UJ « UJ U UJ « UJ « « UJ « ..J UJ « UJ « ::l ..J « « iIi (/) ..J ~ ::> UJ UJ U UJ a: 0 (/) U 0I iIi :i! « « n. )0:: « « :J ::> U U UJ a: I « :J UJ OJ (/) (!) U f- :i!

I

I

~

Den

~3

UJ « UJ U «

cr

it z i ? en

~

UJ « UJ

• ~

UJ ;:! « a: ~ ..J )UJ ..J u.. > In. :J I 0 ..J n. 0 UJ >10 0:: UJ :i! « I ::> U n.

cr

UJ n. )UJ f;" ;:! ~ « «UJ (/) ~ ~ ..J ..J U 0 n.9 UJ ~ o(!) Z 9 0 6 a: - « UJI ..J I n . UJ n.o en

E

• I

• • • •••

I I

I I

1r g

I

I

I

I

Z

::>

• • •

I

I

I

I

I

I I

0

I

1 I

•

I

Z

UJ

~~

I

•

UJ

ffi

I

1 I

0

i1: F

I

I

I

I

•• •• I

I

Figure 2 Pollen percentage diagram of modern surface samples in the eastern Orange Free State and Drakensberg areas. 250 pollen grains were counted per sample.

comes from ca. 80 cm below the top. Two additional pollen samples from the opposite gully wall are presented in section B . The lower one is derived from pale sand stratigraphically above section C and presumably of terminal Pleistocene to Holocene age. The upper one belongs to a brown silty channel fill, of ca. 7S cm depth, containing bones of domesticated goat or sheep (identified by James Brink), and which is therefore of Late Holocene age. The Cornelia sequence The polleniferous swamp deposits at the Cornelia site directly north of Clarens occur near the top of a drainage line with a slope of ca. ISO (Figure 1) . The site is situated within a channel which is steeply incised into Pleistocene alluvial sediments forming a 20 to 2S-m terrace. The sediment studied forms part of a +3,S-m terrace within this channel (Scott 1986) . About 30 cm of

organic-rich grey silt are exposed at the stream-bed level (Figure 4). Radiocarbon dates show that the organic sediments range between ca. 20 000 and 12 600 years BP (Table 2). A pale brown sand which forms the upper part of the + 3,S-m terrace has no potential for pollen analysis. However, undated, presumably Holocene, dark organic sands with pollen (Figure 4) occur within similar sediments lower down the stream. The Craigrossie sequence The Craigrossie site is situated about 20 km west of the source of the Little Caledon River (Figure 1). The upper drainage basin of the river includes slopes up to ca. 2 SOO m altitude. The sediments contain pollen derived from local swamps and potentially from flood or slack water. Therefore a source area larger than those at Elim and Cornelia should be considered (Scott 1986). The vegetation of the catchment includes types similar to the ones

111

S.Afr.l. Bot., 1989 , 55(1) 0:

0: (D 0:

,. "'aa

:J SECTION A

c:::::3

SECTION B

B

SECTION C

~

0

I f-

U

W 0

t:;

...J

I f-

~

~ W

"

!,!

n.

&

4:~

" ""'

§~§~
1i't'j'jll:d",

0-

iiiw I"

"a:,.

-,

f-

,

1

--

w

"Erz

f-

- ,

---'d

,.wn.

w W Z ~ 0:

~t'j

Il' !::

"u "
W

w

(f)

(f)

1

.

~

N

1

":;;; ~ "u w w

:;;;

~gL5

tj~3~ "0 ««:J:J 1i'a z

W I

U

IW ZZ O

l-ucr

~g~~

U~<:.!X


" "' 0:

wW

"

w

,.Il'

"::J

f-

Z

W

W

...J

~ili '!l 0:

:J

(f)

w

z a

~::!

f-

W

w

...J

w!,!

Er

U

(f)O:

~

" ...J

w

fZ

Z

(f)

...JU

0

N

:J

a

...J

f-

U

0:

Il' z

W

...J ...J

1i'

,

,

50000

50100 150 200-

I

250 -

fI

300-,

Ir

350400-

- -CYCLE 2

, ~{€ • :'::. =-==-

~.;., ~

c

I

22 600 ! 240 550-,

I

123400 ! 340 6 00-

I

,

o ~~~i;lbN§~ND o g~~~~~~~ND Figure 3

E2

II

I

450500-

CYCLE I [

101

IIiI DARK GREY SILT "

AND SAND

I

t

I

101

I

1

II [9

,

,

I

I

ALTERNATING SAND AND SILT LAYERS

EI

m '

DARK SILT

~

DARK BROWN CHANNE L-FILL WITH BONES AND BOULDERS

fB -

SANDY SOIL

Pollen percentage diagram of the Elim swamp sediment sequence.

deposition of sediments, the rate of which cannot be estimated accurately by the available dating, or by contamination by microscopic root or other structures. The latter possibility seems to be the most likely. If it is assumed that the radiocarbon dates of the organic silts are minimum ages as a result of contamination, the top of the silt is probably not younger than ca. 10 700 years BP, on the basis of the 'oldest' determination at the top of this unit (Pta 3389). The age of grey and pale sand (between ca. 90 and 160 em) corresponding to the top part of pollen zone CR3 and to CR4 at Craigrossie, is probably Holocene (Figure 5) . It seems to represent a stage of sand movement and erosion which took place any time between ca. 10 000 and 1 000 years BP. Unconformities are possible at the

surrounding the site and also some higher-lying types with more fynbos elements . The polleniferous exposure is an anomalously wet, organic sequence of sediments belonging to a + lO-m terrace on the north side of the stream . The sediments, saturated by spring seepage, do not seem to represent reworked organic material from upstream deposits, as these are generally inorganic. The Craigrossie sequence can broadly be subdivided into eight sedimentary units as shown in Figure 5. Radiocarbon dating of brown organic silt levels below 350 em is problematical since the five dates which range between ca. 10 700 and 9 640 years BP are not in stratigraphic order (Table 2). If the anomalies in the radiocarbon dating were not caused by reworked organic matter, they could be explained either by rapid

~

« 0:: C>

Cl

'"Z ~

>

§

::;)

0

o

U

:I:

~

~

:J

DARK SAND~;:::_=_-_-=-_-_~

100 340

I' I I

200001:300 1

400

Figure 4

250

1I

12600! 110

360

380

•

HOLOCENE

Pollen percentage diagram of the Cornelia swamp sediment sequence.

250 250 36 250 250 173 280 310

0::

'" Z ...J '" ~

Il.

112

S.-Afr.Tydskr. Plantk. , 1989 , 55(1)

SURFACE GREY SAND

ORGANIC LOAM

I

--

I

r

RED SAND

50

-

I

I"

CR5

250 ,--->-----
I PALE SAND

II

660 t 60

100

I

I

245 CR4

I I

,

GREY SAND

250

100

CR.

___ _!_~~O~.!.IQ _ BROWN SILT

248 250 250

250 250

--- - 11070ofI30200 CR.

250 CR2

250 249 201

~

265

II060O:!:120

250

BROWN SAND

.50

275

250 250

BROWN SILT

I 110400!'40

I

II

I I

275 CRI

250 250

I

250 250

110600!100

~

280 294

II Figure 5

250 251

Pollen percentage diagram of the Craigrossie alluvial sediment sequence.

upper and lower contacts of the pale sand. A marked shift in pollen composition (described below) occurs above this sand , between pollen zone CR4 and the bottom of CR5 , corresponding to an organic loam. The top of zone CR5 corresponds to red and grey colluvial sand units . Pinus pollen is present in the sands which therefore date to historical times with agricultural disturbance . The lower part of CR5 is of prehistoric age with a radiocarbon date of 660±60 years BP (Pta 3842, Table 2, Figure 5). This date in the upper dark soil may also be 'too young ' due to possible root contamination from the present swamp vegetation. Pollen analysis of the sediment sequences On the basis of changes in the pollen curves in section C of the Elim sequence , it can be subdivided into three units El , E2 and E3 , with zone E2 subdivided into subunits a and b (Figure 3). The deepest zone represents a well-developed swamp as is shown by the high pollen concentration consisting largely of Cyperaceae . In the second zone (E2) the 'fynbos ' elements (Ericaceae, Passerina, Stoebe, C/iffortia , R estionaceae , etc.) are especially well developed , while in E3 grass pollen is more prominent and diversity of pollen types appears to be low. The three polleniferous layers of apparently Holocene age (sections A and B) from Elim , provide an incomplete pollen record for this period. Most notable is the high proportion of Proteaceae pollen in the pale sand (section B) , suggesting that proteas which are presently restricted to protected slopes in the Rooiberge were

formerly more widespread in the area. The pollen in zone C1 and C2 at Cornelia (Figure 4) shows relatively high numbers derived from fynbos plants. Essentially the spectrum represents a compressed equivalent of the Elim sequence. The high proportion of bryophyte spores (Anthoceros and Riccia) in zone C1, is probably indicative of local moisture. At Craigrossie, slightly more persistent long-distance Podocarpus pollen in zone CR1 (Figure 5) indicates either forest expansion in distant mountain ravines , or relatively low influx of local pollen , resulting in overrepresentation of long-distance pollen. Zone CR2 shows increased percentages of Chenopodiaceae which have previously been attributed to dry conditions (Scott 1986) . The results of factor analysis (described below), however, allow an alternative possibility. Zone CR3 contains relatively high Gramineae and probably local Artemisia pollen , but the latter only peaks in zone CR4. Zone CR5 comprises high numbers of Cyperaceae and semi-aquatic Gunnera perpensa pollen, which suggests a marked change in the local environment. Factor analysis of pollen data

In order to summarize the complex pollen results and to allow better correlation of data between sites, factor analysis (d. principal components analysis of a correlation matrix, Nie et al . 1975) on percentage values of the 28 most prominent pollen taxa was carried out. Exotic Pinus and taxa which were rarely recorded (in fewe r than 5 samples and generally remaining below

113

S.Afr.l. Bot., 1989 , 55(1)

1%) , were excluded from the analysis. Information on the first three of 28 factors (F1, F2 and F3) which emerged from factor analysis (with no rotation) on all 81 pollen spectra , are presented here . These three factors describe groups of co-varying pollen taxa which can be related to environmental parameters. Although Fl, F2 and F3 are the factors with the highest degree of variance they only describe 31,6% of the total variance . Additional factor analysis (with and without rotation) on the data of individual sites (Elim , Craigrossie, Cornelia and surface localities) produced related factors , often with slightly higher degrees of variance; however, the order of importance of these factors varied from site to site . For simplicity and better intersite comparisons only the results of factor analysis on all the data lumped together are presented here . Palaeo-environmental conclusions reached are generally consistent with those of individual site analyses, and this may suggest that the strongest patterns of change revealed by factor analysis of the total data set are not merely related to intersite differences but also to regional palaeo-environmental fluctuations . Factor 1 which accounts for 11 ,9% of the variance, has highly positive loadings on pollen types belonging to trees and shrubs and strongly negative loadings on cryptogamic spores of Riccia and Anthoceros, and pollen of Aizoaceae-type and ' fynbos' taxa (Table 3) . Since tree and shrub taxa occur mainly on low-lying, relatively warm mountain slopes, and the 'fynbos' types mainly in cool upland areas above the tree line of ca. 2000 m, it is reasonable to associate FI with changes in temperature conditions. The use of Aizoaceae-type pollen (including Mesembryanthemaceae) as an indicator for cool temperature conditions may seem unjustified, as parent plants of this group are often associated with succulents of warm dry areas and not with the cool wet parts of the Drakensberg. However , I observed numerous dwarf succulents belonging to the Aizoaceae at high altitudes (above 2000 m) in the relatively dry Rooiberge, on rocky outcrops with no, or very shallow soils . This may explain the association between pollen derived from the plant group and upland fynbos. The 12 surface pollen samples are not enough to strongly support the contention that Fl is associated with temperature conditions , but factor scores on FI (% values x loadings for taxa, summed) for the different surface samples (Figure 6) , show a weak negative correlation with altitude (correlation co-efficient = -0,428, R2 = 18,29%). On the basis of the assumption that altitude can be related to temperature, the Fl score curves for the three pollen sequences can tentatively be applied as a crude palaeotemperature index (Figure 6). If more surface pollen samples are analysed in future and a stronger correlation is obtained between Fl scores and altitude, Fl scores could probably be transformed into temperature values. Factor 2 which accounts for 10,7% of the variance, has high positive loadings on pollen of regional slope vegetation, and strongly negative loadings on that of local swamp and sandy stream terrace vegetation (Table 3). It can be assumed that this factor is associated with the

Table 3 and 3

Loading on factors 1, 2

FI

F2

F3 -0 , 15

Podocarpus

0, 14

0,03

Proteaceae

0,11

0 ,04

0, 11

Oleaceae

O,OR

0 , 14

-0 , 17

Rhus

0 .66

0 ,31

0,25

Ebenaceae

0,56

0 ,50

0, 14

Tarchonanthus

0,45

0 ,38

0, 11

Gramineae

-0,10

0,32

-0,13

Cyperaceae

-0,10

-0,50

-0,19

D,26

-O,D8

-0 , 19

D,35

D, II

D,38

D,16

-O,4D

-0,23

-O ,D2

D,29

D,48

D,3D

-0,29

D,D8

-O,ID

D,03

D,69

D,33

D,51

-0 ,01 -0,21

Typha Ranunculaceae Gunnera Compositae Arlemisia Stoehe type A nlhospermum Mohria Clifforlia Passerina

-0 ,36

D,27

-0 ,38

0 ,36

-D,31

-0,25

0,62

-0 , 10

E ricaceae

-0 ,31

-0,66

-0,25

Aizoaceae type

-0,58

0,49

-0 , 19

Chenopodiaceae

0 ,21

-0,28

D,12

Liliaceae

0, 19

0,19

--0,20

Umbe lliferae Caryophyllaceae Pteridophyta Anthoceros Riccia Trihulus

0,21

0 ,25

0,34

-0,51

--0,06

0 ,37

D,35

D,26

D, 12

--0,53

--0,02

0,62

-0,53

-O,D2

D,6D

0,25

-0 ,05

0, 16

local 'swampy ness' of the sites . Swamp configuration at a site influence fossil pollen spectra and reflect local rather than regional conditions . It is, however, reasonable to relate more swamp pollen to generally increased precipitation at some sites, since swamps can be expected to grow during moist phases. It is therefore possible that F2 provides an index of past moisture conditions at the swamps , Elim and Cornelia, occurring on slopes north of Clarens. Low swamp and high regional pollen values may indicate either smailer swamp size or deposition in submerged conditions. Lake formation at Elim and Cornelia is not indicated in view of steep slopes, but deposition of more regional pollen at Craigrossie may at least be partly due to trapping in flood or slack water. Therefore positive F2 score values for Elim and Cornelia (Figure 6) may be indicating local dryness with deposition of relatively more regional pollen rather than swamp pollen, but for Craigrossie they may be indicating dry conditions and/or stream deposition. Negative scores for F2 can only be associated with relatively wet local conditions in all three sites. A weak positive correlation between F2 scores of surface pollen spectra and altitude is evident (Figure 6) , but this can be attributed to sampling bias as swampy pollen spectra were all derived from relatively low-lying valleys . No surface spectra for high-lying valleys were available.

S.-Afr.Tydskr. Plantk. , 1989 , 55(1)

114 ALTITUDE AND DEPTH (M)

F3 (S.9%VARJ

F2 (l0.7%VARJ

FI (11.9 % VARJ

.1

2700

SURFACE POLLEN SAMPLES

.2 2~00

2300 r

2100

2~ = 18,29%

~6

1900

0

.4

.7 .S

.~

-2

-I

o

.8

•7

9

10 "2 .11

1700 CRAIGROSSIE SEQUENCE (l73~M) 660t60.

~~

.3 .4

II .01 2

3

-2

o

-I

2

3

-2

-I

0

2 EXPLANATION

CR~

'NATAL/LESOTHO AREA

CR4--------------

9460!IIO __ _ l0700! 130' _ =

0::

l0600! 120.

• CLARENS AREA

-------- - ---

2

CR3

:7' CLOSE - INTERVAL SAMPLES

3

CR2

x WIDE-INTERVAL SAMPLES

4 l0400! 140. ,0600!,00.

o SAND AND SILT

CRI

[I ORGANIC LAYERS • 14C DATES (YR BP) 6 -2

CORNELIA SEQUENCE (lBOOM)

-I

o

o

2

-2

o

-I

2

-2

-I

o

2

EI.CI.CRI. ETC.-POLLEN ZONES

HOLOCENE P ~ ~ !)

2 12600!II0. 20000!300.

tm1

3 4

@f-n---2

ELIM SEQUENCE (lSSOM)

-I

0

2

-2

-I

0

2

-2

-I

0

2

-2

-I

0

2

j.

HOLOCEiEI

C

--------- - ---- -------

- ------------- ----

0 E3

2

20400t 6201 21~00!300

t

22600 240. 23400!340 •

3

E2 (b)

4 ~

(0)

ET- - - - - -

6 -2

-I

0

2

-2

-I

0

2

Figure 6 Score values for the first three factors in factor analysis of pollen data from sediment sequences near Clarens and from surface pollen samples in the region.

The third factor (F3) , accounting for 8,9% of the variance, has high positive loadings on Stoebe-type and other Compositae pollen , and on cryptogamic spores of Riccia and Anthoceros , and high negative values on Cliffortia and a wide range of other pollen types including Ericaceae and Gunnera (Table 3). Although the correlation between F3 scores of surface pollen samples and altitude is similar to that of Fl, it is unlikely to be related to temperature conditions alone, since shrubby Compositae vegetation (including the Stoebe group) is not confined to warm low-lying areas . The significance of this factor is not clear; it contrasts karaid indicators against mesic fynbos and possibly points to an inter-relationship between moisture and temperature

conditions , or seasonality aspects , of the past environments . Positive values may imply low species diversity with un favourably dry conditions during summer, limiting the growth of grassy and fynbos communities , and favourable (relatively warm) conditions during winter, allowing hardy Compositae to survive . The prominence of cryptogamic spores in F3 may be due to their relative importance in relation to other local swamp elements during unfavourable conditions . The positive values of this factor on modern surface samples 7 and 8 may indicate impoverishment with Compositae shrub encroachment due to poor soil qualities and recent overgrazing (Figure 6).

115

S.Afr.J . Bot. , 1989 , 55(1)

Discussion The pollen data from the Clarens area provide a provisional palaeo-environmental record for the last 23 000 years in the eastern Orange Free State as summarized in Table 4. As yet this record is incomplete but the potential of finding more polleniferous deposits in the area is good and future work may help to fill time gaps. On the basis of extrapolation of the four radiocarbon dates in section C of the Elim sequence , it is estimated to cover the time interval between ca . 23 000 and ca . 17 000 years BP . The resolution at the Cornelia site which seems to cover part of this interval and that between ca. 17 000 and ca. 13 000 years BP , is not good enough to allow any reliable deduction about the environment during the phase . The pollen data from Elim are important in that they are some of the very few sources of information about the vegetation in southern Africa from 23 000 to 17 000 years BP. The coolest conditions are indicated for a time period after 20 000 years BP and before ca . 18 000 years

Table 4 Summarized interpretation of main palaeoenvironmental stages recorded in the Clarens area Palaeo-environmental Site zones

Age

Craigrossie

Late Holocene

indications Grassland. Sub-humid with summer rains as at present. Soil formation. The phase ends

CR5

with agricultural disturbance. Craigrossie

Early Holoce ne

Grassland with fynbos . Relatively warm and dry (slightly more winter rains?)

Terminal Pleistocene and Holocene ,

Grassland . Intermediate to warm temperatures. Fairly humid with torrential summer rains?

CR4 Craigrossie CR2 , CR3 Cornelia C2?

Craigrossie CRI

starting before 10 700 years BP. Part of the late

Glacial , starting ca. 13 000 years BP Cornelia C2?

Grassland characterized by fynbos . Podocarpus forests in distant ravines. I ntermediate temperature and moisture conditions with relatively dry summers.

Elim E3

ca. 18 000 to

High-altitude grassland

17 000 years BP

characterized by fynbos elements. Relatively dry

ca. 22600 years BP to 19 000 years BP

Maximum fynbos development with cold dry conditions ca.

becoming more humid. Elim E2 Cornelia Cl

19 000 years BP. Wetter with swampy events locally , ca. 20 000 and 22 000 years BP (with a shift to winter-rain conditions in between?) Elim E3

ca. 23 000 years BP

Humid upland grassland with fynbos and strong local swamp development

BP. Relatively dry conditions for this phase are indicated and are comparable to pollen indications for th e coolest phase in the Transvaal at roughly the same time (Scott & Thackeray 1987). Due to apparently anomalous radiocarbon dates in the Craigrossie sequence the start of this record is not possible to date . However , with a minimum date of ca. 10 700 years BP near the top of the lower organic silts (pollen zone CR3, 171-181 cm) , the base (ca . 560 cm) is unlikely to be older than 13 000 years, if a sedimentation rate comparable to that at Elim is assumed . Even though the Craigrossie sequence is incomplete , it provides an informative record of terminal Pleistocene and Holocene conditions in the area. Before 10 700 years BP , intermediate temperatures and relatively wet conditions are indicated by pollen in this sequence . Scott (1986) suggested that zone CR2 formed under dry conditions on the basis of high numbers of Chenopodiaceae pollen and a reduction in fynbos and long distance Podocarpus pollen in the zone. Grouping of taxa in F2 of this study, however , suggests an alternative possibility of relatively moist conditions for this zone, implying that the Chenopodiaceae community flourished locally as a result of stream disturbance and strong evaporation under warmer conditions, and not necessarily as a result of dry conditions . Correlation of the lower section of the Craigrossie sequence with moisture cycles suggested by pollen spectra of apparently similar age, starting ca. 13 000 years BP , from Aliwal North (at 1 300 m altitude, ca . 300 km to the south-west ; Coetzee 1967), remains difficult due to poor dating control and local differences between the sites. However, the initial strong presence of Stoebe-type pollen in both sites may be related to regional conditions affecting both areas at the end of the Pleistocene . Factor 1 vaguely suggests warm temperatures after 9 460 years BP at Craigrossie, when conditions were relatively dry , but this is not certain, being based on few spectra. Although the lack of dating does not provide support , this may correspond to indications for a temperature optimum found during the early Holocene (ca . 6500 years BP) at palynological sites in the South African woodland areas (Scott 1988b). The relatively moist conditions during formation of pollen zones CR5 correspond with interpretations of wetness during the Late Holocene in South Africa . These conditions became optimal after 6 500 years BP in the Transvaal (Scott 1988b) , and remained relatively wet during the rest of the Holocene. An approach for the assessment of past seasonal patterns in palynological data in the eastern Orange Free State highland area has been suggested by Scott (1986) . The possibility of detecting past seasonal effects by pollen analysis (F3 in the present study) is promising, but needs further study. If factor 3 indeed describes seasonal changes, the fluctuations indicated can be compared with past seasonal patterns simulated for the Southern Hemisphere on the basis of the earth's orbital fluctuations (Kutzbach & Guetter 1986). These patterns suggest, for example, slightly warmer winters and

S.-Afr.Tydskr. Plantk., 1989,55(1)

116 cooler, dry summers ca. 9 000 years BP , but the pollen data as well as the simulations are not detailed enough at this stage to allow meaningful comparisons. The data from Elim provide support for the contention that the coldest spell , ca . 19 000 years BP , was accompanied by relatively dry condlitions, and that warming, ca. 13 000 years BP, was accompanied by relatively moist conditions . These findings are in conflict with moderately moist conditions simulated for the Southern Hemisphere land areas (0-30 C) between 18 000 and 15 000 years BP, and drier simulated conditions during the Late Glacial at 12 000 years BP (Kutzbach & Guetter 1986). However , data from a closer grid of sites from southern Africa are required for more meaningful comparison with the model. The main change which took place in the vegetation of the area at ca . 2 000 m altitude comprises the disappearance in the Holocene, of a marked upland fynbos element which characterized the environment before ca. 10 700 years BP. This fynbos is believed to have spread to altitudes as low as 1 000 m in areas which were humid enough. Similar results were recorded in Transvaal sites between 450 and 650 km to the north (Scott 1982, 1987a). Towards the west , however, the spread of upland fynbos during these times was limited by drought , assuming the rainfall gradient remained similar to today's pattern . This is suggested by the increase of shrubby grassland with Stoebel Ely tropapp us types and other Compositae , but with no Ericaceae, from Equus Cave, at ca. 1 200 m altitude and ca . 400 km to the west of Clarens (Scott 1987b) . A comparable dry , shrubby vegetation, but with very little grass apparently occurred ca . 430 km towards the south-west , ca. 8 000 years BP and earlier, at ca . 1 700 m altitude in the Karoo near Noupoort , and was later replaced by karoid grassland in the Holocene (Bousman et al. 1988). It can be concluded that the vegetation in the Clarens area was subjected to cyclic changes during the last 23 000 years and that these changes were probably controlled by climatic parameters. Relatively cool conditions are indicated for the early part of the sequence (after ca. 23 000 years BP), becoming most intense ca . 19 000 years BP, while warmer conditions developed at the end of the Pleistocene and during the early Holocene (Figure 6). Relatively moist conditions occurred ca. 23 000 years BP and again ca. 13 000 years BP and then during the latter part of the Holocene , while a relatively dry climate prevailed ca. 20 000 to 17 000 years BP. High Compositae, including Stoebe type, values in pollen spectra of ca. 13 000 years BP, may be indicative of a different, less seasonal rainfall distribution. Tentative correlations between events in this chronology and sequences from other sites in South Africa are possible (Scott 1988b). D

Acknowledgements Research in the eastern

Orange Free State was

supported by grants from the Foundation for Research Development , Council for Scientific and Industrial Research, Pretoria. Dr J.c. Vogel of the CSIR provided the radiocarbon dates. The studied deposits were located by field observations of Martin Wessels, C1arens. Technical assistance and help with the preparation of this manuscript by Glenys van Aswegen is highly appreciated. James Brink is thanked for identifying bones from a sediment exposure and Britt Bousman for constructive critisism.

References BOUSMAN , C.B. , PARTRIDGE , T.e. , SCOTT, L., METCALFE , S.E. , VOGEL , J .e., SEAMAN , M. & BRINK, J.S. 1988. Palaeoenvironmental implications of Late Pleistocene and Holocene valley fills in Blydefontein Basin, Noupoort , e.P. Palaeoecol. of Afr. 19 (in press). COETZEE, J.A . 1967. Pollen analytical studies in East and southern Africa. Palaeoecoi. of Afr. 3: 1- 146. KUTZBACH , J.E. & GUETTER , P.J . 1986. The influence of changing orbital parameters and surface boundary conditions on climate simulations for the past 18 000 years. 1. Atmos. Sci. 43: 1726-1759. NIE, N.H., HADLAI HULL , c., JENKINS , J.G., STEINBRENNER , K. & BENT, D.H . 1975. Statistical package for the social sciences. McGraw-Hili , New York. SCOTT, L. 1982. Late Quaternary pollen record from the Transvaal bushveld, South Africa. Quaternary Res. 17: 339- 370. SCOTT, L. 1986. Pollen analysis and pal aeoenvironmental interpretation of Late Quaternary sediment exposures in the eastern Orange Free State , South Africa. Palaeoecolof Afr. 17: 113- 122. SCOTT, L. 1987a. Late Quaternary forest history in Venda , southern Africa. Rev. Palaeobot. Palynol. 53: 1- 10. SCOTT, L. 1987b. Pollen analysis of hyena coprolites and sediments from Equus Cave , Taung , southern Kalahari (South Africa). Quaternary Res. 28: 144- 156. SCOTT, L. 1988a. Hyrax (Procaviidae) and dassi e rat (Petromuridae) middens in paleoenvironmental studies in Africa. In: Fossil packrat middens: the last 40 000 years of biotic change. University of Arizona Press , Tucson (in press). SCOTT, L. 1988b. Climatic conditions in southern Africa since the last glacial maximum , inferred from pollen analysis. Palaeogeogr. Palaeoclimatol. Palaeoecol. (in press). SCOTT, L. & THACKERAY , J.F. 1987. Multivariate analysis of late Pleistocene and Holocene pollen spectra from Wonderkrater, Transvaal , South Africa . S. Afr. 1. Sci. 83: 93-98. VISSER, J.N.J. , BEUKES , G.1. & SCOTT, L. 1986. Vivianite in late Pleistocene swamp deposits near Clarens, S. Africa . Trans. Geoi. Soc. S. Afr. 89: 395-400. VISSER, D.J.L. & VAN RIET LOWE , e. 1955. The geology and archaeology of the Little Caledon River Valley. Mem . Geoi. Surv. S. Afr. 37: 1-46.