(School of Geography and the Environment and St John S College) University of Oxford, Oxon

Last Glacial Maximum conditions in southern Africa: are we any closer to understanding the climate of this time period?

Abi E C Stone*

(School of Geography and the Environment and St John’s College) University of Oxford, Oxon,

Abi Stone, School of Geography and the Environment, Oxford University Centre for the Environment, University of Oxford, South Parks Road, Oxford, OX1 3QY.

Email:

*Current address

School of Environment, Education and Development, University of Manchester, M13 9PL

Abstract

The Last Glacial Maximum (LGM) (21±2 ka) is an important period for which to understand past climatic and environmental conditions. In particular it is a key time-slice for evaluating the performance of numerical climate model simulations of glacial palaeoclimatesusing palaeoenvironmental data sets. However, our palaeoenvironmental datasets and reconstructions of climatic conditions at the LGM are still debated in certain regions. This is the case for southern Africa, despite more than half a century of research since early conceptual models of palaeoclimate were proposed. The greatestdebates are about the spatial patterning of relatively wetter and drier conditions than present and the position of the mid-latitude westerlies at the LGM. Different patterns emerge from separate syntheses of palaeoenvironmental proxies, from different numerical model simulations and from comparisons of the two. In this review of the progress over half a century of research in southern Africa: (i) a brief historical review of key conceptual models is given, (ii) key points of conflict that emerge in synthesis of palaeoenvironmental proxy records are outlined and (iii)numerical model simulations are considered. From these some points for future progress are suggested.

Keywords

Last Glacial Maximum (LGM), Quaternary, Climate, Southern Africa, Palaeoenvironmental proxy, Climate modelling

I Introduction

The climatic and environmental conditions of the low-latitudes and subtropics at times of high global ice volume (glacial, episodes) have been a focus of research and debate for many decades, with a particular focus on the last glacial maximum (LGM) (Clark et al., 2009). Debates centreon both hydroclimaticvariables and temperature changes. These are not only of interest in themselves, but also because understanding the response of hydrological systems to large-scale climate forcing between glacial and interglacial conditions informs our understanding of present-day and future hydrological changes at these latitudes.The LGM (21 ± 2 ka using the EPILOG programme definition (Mix et al., 2001; Clark and Mix, 2002)) is a valuable period for evaluating the response of the climate system to forcing factors usingnumerical climate models (Jansen et al., 2007). LGM radiative forcing is of a similar magnitude, but opposite direction, to that for the year 2000, and the LGM hadsubstantially different boundary conditions (of sea level and land ice). These characteristics offer useful conditions under which to assessmodel skill (a relative measure of model performance in hindcasting or forecasting) (Braconnot et al., 2007a; Jansen et al., 2007; Otto-Bliesner et al., 2009; Hargreaves et al., 2013). The PMIP (Paleoclimatic Modelling Intercomparison Project) (e.g. Braconnot et al., 2007a,b) and CMIP5 (5th phase of the Coupled Model Intercomparison Project) take the lead in evaluations, in which model performance is assessed using syntheses of palaeoclimaticdata (Braconnot et al., 2012), such as pollen-based and sea surface temperature data syntheses (e.g. Bartlein et al., 2011; MARGO, 2009). The LGM has a greater abundance of preserved proxy evidence than earlier glacial periods and has been the focus of long-standing investigations, such as the CLIMAP (Climate Long range Investigation, Mapping and Prediction) (1981) programme.

We have a variety of terrestrial and marine proxies that yield information about variations in temperature, hydroclimate, and windiness in dryland regions (see Table 1 that focusses on proxies from southern Africa, which span the LGM). However, our understanding of climatic and environmental conditions at the LGM is still debated, particularly for current and former dryland regions where preserved proxy data is spatially patchy and temporally discontinuous (Chase and Meadows, 2007; Gasse et al., 2008). This is the case for southern Africa (which we treat here as that south of 15 oS, following the sub-regions used by Gasse et al. (2008)) despite over 50 years of detailed research using proxy palaeoenvironmental data and numerical climate modelling. It would be sensible to expect our understanding to have improved, but conversely, current reading of the existing reviews highlights some large differences in opinion and as Thomas and Burrough (2012. p10) reflect ‘interpretations have become less, rather than more, focussed and clear’. Therefore, the motivation for reviewing the understanding of southern African palaeoclimates is in no small part to introducesome of these debatesto an audience less familiar with the specific debates for this region. It is important to reflect on the multiple working hypotheses that exist and encourage continued evaluation of our models (conceptual and numerical) and proxy datasets. One complicating aspect for understanding past climates in this region are the diverse nature of the available proxy record types (e.g. geoproxies, dune records and lake shorelines; speleothems and other chemical precipitates; pollen records from wetlands, lakes and hyrax middens; stable isotopesin groundwater, and at higher resolution in hyrax middens) and the related diversity in the inferences aboutpalaeoenvironments and palaeoclimatesthat it is possible to be made from them.This makes a regionalsyntheseschallenging because proxies respond to different forcing factors (for example responding to seasonal maximum or minimum temperatures, rather than annual mean values), with different time lagsand contain different resolution data (in terms of their sensitivity to climatic forcing, their accumulation rate and our ability to provide chronologies for them). Most terrestrial records are discontinuous and have ‘floating’ and relatively imprecise chronologies, compared to high resolution, continuous records such as ice-cores. Nonetheless, lower resolution records are still useful to investigate responses to gradual climatic forcing, such as orbital-driven glacial-interglacial shifts. Gaining a representative spatial picture is also made difficult by the large geographical distances between preserved record, which we can either leave as gaps or attempt to interpolate between sites. Interpolation may result in spatial patterns that are an artefact of a data; the lower the spatial coverage of our datasets the higher the chance that the spatial pattern we reconstruct is not a good approximation of the true spatial pattern.Overall, despite the challenges with, and the noise within, palaeoproxy data, they are robust enough to indicate that numerical climate model simulations need a lot of refinement in order to be able to reconstruct first-order spatial patterns of past climate for southern Africa, particularly for hydroclimate (precipitation).

This paper aims to: (a) give a brief but not exhaustive historical review of ideas, reflecting on the key conceptual models of climatic circulation patterns that have been proposed for the LGM; (b) highlight the value as well as the points of contradiction in syntheses of palaeoenvironmental proxy records for the region and (c) consider the range of results produced bynumerical simulations. Within these broad sections, I will highlight some overarching reasons why the question of what happened to the climate and environmental conditions of southern Africa at the LGM remains a difficult puzzle to solve. The conclusion will summarise what we do know and the potential steps forward in refining our understanding. The focus of this review is the subtropical latitudes of Africa south of 15oS, corresponding to Gasse et al’s (2008) region IIand region IV within their review of climatic patterns from 30-10 ka in equatorial and southern Africa; and the spatial coverage of Chase and Meadows (2007) review of the dynamics of the winter rainfall zone (Figure 1(b)).

II Major aspects of the modern day climatology of southern Africa

Major atmospheric circulation features that influence the distribution of precipitation in southern Africa include: the descending limb of the Hadley cell (in Austral winter); the ITCZ (Intertropical Convergence Zone); the East African monsoon system (with flows from the South Indian Anticyclone (SIA) in Austral winter and from high pressure of Arabia during Austral summer); the CAB (Congo Air Boundary), where low-level air from the Indian Ocean and the Congo basin and the Atlantic Ocean converge; westerly moisture flux from the tropical southeast Atlantic, which feeds into the Angola low pressure; the Atlantic Walker circulation, and the southern hemisphere (SH) mid-latitude westerlies that carry embedded cyclonic and anticylonic disturbances (associated with the passage of frontal systems) and reach the southwestern tip of the continent in Austral winter (Tyson, 1986; Tyson and Preston-Whyte, 2000; Wang, 2005; Reason et al., 2006). Figure 1 illustrates positions of key features in Austral winter and summer (Figure 1(c)), and indicates the division of the subcontinent into a summer rainfall zone (SRZ), year-round rainfall zone (YRZ) and winter rainfall zone (WRZ) (Figure 1 (a)). These atmospheric features are influenced by topographic and oceanic features, with marked differences in the west and east(note that only in the west is the ocean connected to high northern hemisphere (NH) latitudes).The cold Benguela Current is instrumental in reducing rainfall (by reducing water vapour flux) in the Namib coastal desert on the west coastand the warm Agulhas Current enhances moisture flux from the Indian Ocean (to the esat) into the subcontinent.

The waxing and waning of large polar ice caps and responses to other orbital forcing parameters at the global, hemispheric or regional level,will have had a profound influence on these oceanic and atmospheric circulation features, as will any more rapid shifts in global or regional climate. These changes to circulation features result in changes in palaeoenvironmental conditions in the subcontinent (e.g. Tyson et al., 2001). The precise manifestation of these changes is still contested, with knowledge progressing as an interplay between insights from the growing volume of proxy evidence and conceptual and numerical climate models.

Figure 1.Modern day climatology in southern Africa. (a) Mean annual rainfall and rainfall regimes and major oceanic surface currents (modified from Gasse et al., 2008). ABF, Angola-Benguela Front; DRM, Double Rainfall Maximum; SRZ, Summer Rainfall Zone; YRZ, Year-round Rainfall Zone; WRZ, Winter Rainfall Zone. (b) Subdivision of Southern Africa using the regions of Gasse et al. (2008) (I is Western Africa north of the ABF, II is Benguela Coastal Current and Aghulas Zone, III is equatorial and tropical East Africa, IV is the summer rainfall domain) and the three axis used byChase and Meadows (2007) and showing country outlines). Schematic representation of major atmospheric circulation features over southern Africa in (c) austral winter and (d) austral summer (RH side) (modified from Tyson, 1999 and Gasse et al., 2008).

Archive and Proxy Type / Example dataset / Dating method / Time period / Key reference(s)
Terrestrial
Lacustrine
Minerology &
sedimentology
Shoreline
Shoreline & sediment
Stomatolite
Sediment geochemistry / Lake Tritrivakely
Lake Chilwa
Lake Ngami
Mababe Basin
Makgadikgadi
Etosha Pan
Etosha Pan
Tswaing Crater / 14C
OSL
OSL
OSL
OSL
OSL
14C
14C, TIC-t3 / 40.8 – 2.9 kyr BP
43.7 – 8.5 ka
140.3 – 0.2 ka
37.7 – 5.2 ka
288 – 8.4 ka
>58 – 0.27 ka
40.1- 29.3 14C kyr BP (c1)
19.0 -2.5 14C kyr BP (o2)
200 ka / Williamson et al. (1998)
Thomas et al. (2009)
Burrough et al. (2007)
Burrough and Thomas (2008)
Burrough et al. (2009)
Hipondoka et al. (2014)
Brook et al (2013)
Kristen et al. (2007)
Dune / Northern & Eastern
Kalahari dunefield
South west Kalahari
dunefield / OSL
OSL / 164 – 3.9 ka
> 186 – 0.2 ka / Thomas et al. (2000), Stokes et al. (1997)
Stone and Thomas (2008) Telfer and Thomas (2007)
Fluvial sediments / Khumib River
Hoarusib
Tsauchab / OSL
OSL
OSL / 27.7 – 8.1 ka
44.0 – 20.3 ka
24.6 – 0.4 ka / Srivastava et al. (2004)
Srivastava et al. (2005)
Brook et al. (2006)
Speleothem (submerged)
Speleothem (growth age)
Stalagmite δ18O and δ13C
Stalagmite δ18O, δ13C and
pollen / Otavi Hills
Drotsky’s Cave
Lobatse Cave
Makapansgat
Wonderwerk Cave / U-series
U-series
U-series
U-series
14C / 129.9 – 7.5 ka
132.9 – 1.5 ka
103.3 – 15.7 ka
25.5 – 0.1 ka
34.8 – 0 calka BP
last 20 calka BP / Brook et al. (1999)
Brook et al. (1998)
Brook et al. (1998)
Homgren et al. (2003)
Brook et al. (2010)
Truc et al. (2013)
Groundwater δ18O and
noble gas / Stampriet Aquifer / 14C / 37 – 0 / 14C kyr / Stute and Talma (1998)
Hyrax midden, pollen,
δ13C, δ15N / Pakhuis Pass
De-Rif rock / 14C
14C / 23 – 0 calkyr BP
19.5 – 7.3 calkyr BP / Scott and Woodborne (2007)
Chase et al. (2011)
Marine
Dinoflagellate cyst &
pollen
Planktonic forams
Terrigenous sediment
Foram δ18O, alkenones
pollen
Foram abundance &
piMg/Ca
Alkenone, MAR-Q
δ13C of plant leaf wax / GeoB 1023-5
GeoB 1706
MD96 2094
GeoB 1711
ODP 1048B
MD96 2087
Transect of cores / 14C
δ18O-t4
δ18O-t4
14C,
δ18O-t4
14C
14C, corr5
14C / 21.1 – 1.5 calkyr BP
160 – 0 ka
300 – 0 ka
135 – 0 ka
24 – 0 calkyr BP
200 – 0 calkyr BP
23 – 0 calkyr BP / Shi et al. (2000)
Little et al. (1997)
Stuut et al. (2002)
Shi et al. (2001)
Farmer et al. (2005)
Pichevin et al. (2005)
Collins et al. (2011)

Table 1.Terrestrial and marine archives and proxies in southern Africa that cover the Last Glacial Maximum, giving some key example datasets

III Brief historical overview of early conceptual models of LGM conditions

Early textbook ideas of ice-age pluvial conditions in the tropics, including many current-day deserts, (Charlesworth, 1957; Flint, 1971) started to be revisedover thirty years ago by suggestions of increased aridity from the tropics to mid-latitudes and an expansion of desert regions at the LGM (e.g. Street and Grove, 1976; Sarnthein, 1978).A number of conceptual models of palaeoclimatic circulation features over southern Africa duringthe LGM have been proposed, based on early compilations of relatively small amounts of proxy data. Within these there is some vagueness about what precise period the LGMcovers. However,most make explicit reference to the effectsof higher (maximum) global ice volume and are therefore relevant to this discussion. The models can be categorised as: (i) a latitudinal contraction and migration of circulation features, (ii) a symmetrical circulation features between hemispheres,(iii)an increased circulation strength (without migration) and (iv) a longitudinal displacement of circulation features (Table 2) (modified from Deacon and Lancaster, 1988). Although perhaps out-dated these model types are still prominent in the current literature (e.g. 7 citations of model (iv) during 2013).

Model (i)proposesa latitudinal contraction of climatic belts, driven by an equatorward expansion of polar high pressure (van Zinderen Bakker, 1967; 1976; 1982) (Figure 2(a)). The pattern ismore marked pattern in the NH than SH, reflecting the larger relative expansion of ice in the NH (cf. Flohn, 1964). In southern Africa, strengthened high pressure cellsleads to drying, although a slight northern migration of the SAA (South Atlantic Anticyclone) allows westerly cyclonic rainsto extend into the southern Kalahari and southern Namib Desert (in winter months)(Figure 2(a)). A revised version of model (i)elucidates temperature reductions and suggests a less southerly extension of the ITCZ in summer, but northward expansion of the westerly cyclonic rain is still emphasised for winter (van Zinderen Bakker, 1976) (Table 1)(Figure 2b). The model also suggestsa spatial division into a cool, dry north and cold south with rainy winters (Figure 2(b)). Model (ii) is somewhat similar in that it emphasises a more marked NH than SH response, but with a focus on hemispheric symmetry of circulation features (Nicholson and Flohn, 1980). In this model the average ITCZ position is at the equator and there is a less marked equatorward movement of the midlatitude westerlies (and associated cyclonic rains) (Figure 2 (c)).

By contrast, a third group of models (iii) stress a strengthening, rather than migration, of features. Butzer et al. (1978) suggests the mid-latitude westerly belt was found at the same southerly latitude as present,that summer rains in the southern Kalahari increased (owing to subtropical disturbances and not the tropical monsoonal rainfall belt, which was contracted) and that there was no increase in winter cyclonic rains. Heine (1982) proposed a similar model, but with more frequent winter rainfall reaching the southern Kalahari. He proposed this was driven by the strengthened circulation systems in contrast to an equatorward migration of the cyclonic winter rains suggested in model (i). The implication is that there is greater humidity in the southern Kalahari and east of South Africa in Heine’s (1982) version of model (ii) compared to Butzer et al.’s (1978) version. The idea of a humid east, whilst the west remains arid (Figure 2(d)) of Heine (1982) (model (ii)) contrasts with the north-south moisture gradient proposed by van Zinderen Bakker’s (1976) (model (i)).

The final model (model iv) stresses longitudinal migrations of high pressure cells and the Walker Circulation, in addition to an equatorward migration of those features, and is tied to a modern analogue model for wet and dry spells developed by Tyson (1986) (Cockcroft et al., 1986). At the LGM (and throughout glacial periods) the region experiences dry spell conditions, with: a weakened Walker Circulation that has its ascending limb further east toward the Indian Ocean; a reduction in tropical-temperate interactions and a northward shift of westerly storm tracks. As a result the SRZ gets drier but receives a greater proportion of winter rain and the WRZ gets wetter(Figure 2(f)), so that the centre, north and northeast are drier and the southwest Cape is wetter.

In accounting for the differences between models it is useful to highlight that: (1) they were based on different amounts and types of palaeoenvironmental proxy data, for example, the Nicholson and Flohn (1980) model has a NH and equatorial evidence base, without much coverage of southern Africa (Table 1); (2) different collections of proxy data result in different apparent spatial patterns, for example the east-west versus north-south patterns, and (3) authors had different conceptions of how the climatic circulation features operated and were forced by LGM boundary conditions. Authors sometimes drew on numerical climate model simulations to back up aspects of their hypotheses. For example, Nicholson and Flohn (1980) state thatGates’s (1976) model accords with an equatorward displacement of circulation features and Butzer et al. (1978)find support for subtropical disturbances in the Kalahari from Williams et al. (1974). In addition, authors often revised their ideas in light of new evidence and ideas. For example, Butzer (1984) revised his ideas in light of the fact that the idea of increased summer rainfall at the LGM was losing favour.

More recent palaeoclimate interpretationsfit, even if not precisely, within the context of these early models and make reference to them. For example, Collins et al. (2011) stress hemispheric symmetry (model type ii) of expansions and contractions of the tropical African rainbelt around the LGM, whilst Chase and Meadows (2007) provide support for a northward expansion of the WRZ (emphasised most strongly in models ii and iv). This and other points of conflict arising from palaeoenvironmental proxies are discussed in the next section.