Orbital cycles

The influence of changes to Earth’s orbit on glacial–interglacial cycles of the Quaternary, linked predominantly to the cryosphere in the Northern Hemisphere (NH), is well established (e.g. Milankovitch, Reference Milankovitch1920; Hays et al., Reference Hays, Imbrie and Shackleton1976; Ruddiman, Reference Ruddiman2006). The varying eccentricity of the elliptical-shaped orbit influences total solar insolation receipt with quasi-regular cycles of ~400 ka and ~100 ka. The other two cycles relate to seasonality of the receipt of insolation: (i) the change in the degree of the title of the Earth (obliquity) influences the intensity of season in both hemispheres (~41 ka), whilst (ii) the direction the Earth’s axis points in space, combined with the rotation of the elliptical orbit around the Sun, influences the relative strength of seasons in each hemisphere via the precession of the equinoxes (~23 ka). The pacing of global glacial–interglacial cycles, inferred from marine oxygen isotope stratigraphy (MIS), starts as ~41 ka cyclicity (e.g. Pisias and Moore, Reference Pisias and Moore1981; Ruddiman et al., Reference Ruddiman, Raymo and McIntyre1986; Raymo et al., Reference Raymo, Rind and Ruddiman1990), transitioning during 1.25–0.7 Ma to ~100 ka cyclicity. In MIS, even numbers are glacials and odd numbers are interglacials (Figure 2A). The ~41 ka cyclicity is thought to relate to direct obliquity forcing, involving feedbacks with atmospheric greenhouse gases, whilst the ~100 ka cyclicity cannot be accounted for by the small total insolation variations related to eccentricity (Ruddiman, Reference Ruddiman2006). Notwithstanding the debates about the nature of the forcing-cryosphere response, and the role of internal forcing and ocean-atmospheric reorganisations in the climate system (e.g. Broecker and Denton, Reference Broecker and Denton1989; Berger, Reference Berger2013), the large change to the Earth’s climatic state during these glacial–interglacial cycles is likely to have influenced the terrestrial dryland regions.

The way conditions in drylands have been conceptualised over global glacial–interglacial periods has gone through large changes in thought. There was a widespread shift in the idea of dryland pluvials during glacial periods to dryland aridity, aided by the advent of chronological control for some proxies (e.g. Williams, Reference Williams1975; Sarnthein, Reference Sarnthein1977). This ‘glacial aridity paradigm’ was lent support by the timing of higher volumes of dust volumes in the Antarctic ice cores such as Vostok during glacials, and larger particles in the LGM layers (Petit et al., Reference Petit, Jouzel, Raynaud, Barkov, Barnola, Basile, Bender, Chappellaz, Davis, Delaygue, Delmotte, Kotlyakov, Legrand, Lipenkov, Lorius, Pepin, Ritz, Saltzman and Stiévenard1999), and back to 800 ka in EPICA DomeC (Lambert et al., Reference Lambert, Bigler, Steffensen, Hutterli and Fischer2012) (Figure 2B), and marine sediments (Figure 1, s1, Figure 2C). However, higher dust concentrations do not necessarily require an expansion of drylands, being wind-driven (more turbulent atmospheric circulation) and because of a potential reduction in wet-deposition of dust owing to a globally averaged lower moisture content during cooler glacials. The marine archives also represent an average of terrestrial material received from a wide spatial area, so that spatial heterogeneity within dryland regions will not be recorded. Therefore, whilst the dryland-derived dust proxy in ice core and marine sediment archives remains an appealing record for terrestrial drylands because they cover a long timescale, capturing multiple glacial–interglacial cycles, they are insufficient for reconstructing conditions in the drylands themselves. This requires terrestrial archives and proxies.

Figure 1. Map of current global dryland distribution, alongside the locations of the major proxy records explored in the three key examples in this paper and an inset of the TraCE-21 ka simulation for the LGM from Liu et al. (Reference Liu, Jiang and Lang2021) demonstrating which dryland regions are simulated to be wetter-than-present, or drier-than-present. Site numbers (s) are referred to in the text, and (1) North Atlantic dust records, ODP659 (deMenocal, Reference deMenocal1995), MD03–2705 and ODP659 (Skonieczny et al., Reference Skonieczny, McGee, Winckler, Bory, Bradtmiller, Kinsley, Polissar, De Pol-Holz, Rossignol and Malaize2019), (2) Mukalla Cave (Nicholson et al., Reference Nicholson, Pike, Hosffeld, Roberts, Sahy, Woodhead, Cheng, Edwards, Affolter and Leuenberger2020), (3) Hoti Cave (Nicholson et al., Reference Nicholson, Pike, Hosffeld, Roberts, Sahy, Woodhead, Cheng, Edwards, Affolter and Leuenberger2020), (4) North Arabian Sea dust record ODP721/722 (deMencoal, 1995), (5) Red Sea dust record K11 (Ehrmann et al., Reference Ehrmann, Wilson, Arz, Shulz and Schmiedl2024), (6) Gulf of Aden RC09–166 (Tierney et al., Reference Tierney, deMenocal and Zander2017), (7) Wadi Dabsa tufa record (Stone et al., Reference Stone, Inglis, Candy, Sahy, Jourdan, Barfod and Alsharekh2023), (8) Soreq Cave (Bar-Matthews et al., Reference Bar-Matthews, Ayaloin and Kaufman1997), (9) Egyptian tufa growth (Kele et al., Reference Kele, Sallam, Capezzuoli, Rogerson, Wanas, Shen, Lone, Yu, Schauer and Huntington2021), (10) Rössing Cave, Namibia (Geyh and Heine, Reference Geyh and Heine2014), (11) Tswaing Crater (Partridge et al., Reference Partridge, deMenocal, Lorentz, Paiker and Vogel1997), (12) CD15410–06 (Simon et al., Reference Simon, Ziegler, Bosmans, Barker, Reason and Hall2015), (13) Naracoorte Caves (Weij et al., Reference Weij, Sniderman, Woodhead, Hellstrom, Brown, Drysdale, Reed, Bourne and Gordon2024), (14) Leeuwin-Naturaliste-region caves (Weij et al., Reference Weij, Sniderman, Woodhead, Hellstrom, Brown, Drysdale, Reed, Bourne and Gordon2024), (15–17) speleothems on the northwest seaboard of Australia (Denniston et al., Reference Denniston, Polyak, Wanamaker, Ummenhofer, Humphreys, Cugley, Woods and Lucker2017), (18) Kati Thanda-Lake Eyre Basin, (19) Murray Darling Basin, (20) Tasman Sea dust record (Hesse, Reference Hesse1994), (21) Native Companion Lagoon, eastern Australia (Petherick et al., Reference Petherick, McGowan and Moss2008), (22) Lake Chilwa, Malawi (Thomas et al., Reference Thomas, Bailey, Shaw, Durcan and Singarayer2009), (23) Pella hydrax midden (Chase et al., Reference Chase, Niedermeyer, Boom, Carr, Chevalier, He, Meadows, Ogle and Reimer2019), (24) Zizou hyrax midden (Chase et al., Reference Chase, Niedermeyer, Boom, Carr, Chevalier, He, Meadows, Ogle and Reimer2019), (25) Spitzkoppe midden (Chase et al., Reference Chase, Niedermeyer, Boom, Carr, Chevalier, He, Meadows, Ogle and Reimer2019), (26) De Rif hyrax midden (Chase et al., Reference Chase, Quick, Meadows, Scott, Thomas and Reimer2011).

Figure 2. Key proxy records globally and for drylands, where: (A) the LR04 oxygen isotope record, with Marine Oxygen Isotope Stages (MIS) above (Lisiecki and Raymo, Reference Lisiecki and Raymo2005), (B) Antarctica Vostok (Petit et al., Reference Petit, Jouzel, Raynaud, Barkov, Barnola, Basile, Bender, Chappellaz, Davis, Delaygue, Delmotte, Kotlyakov, Legrand, Lipenkov, Lorius, Pepin, Ritz, Saltzman and Stiévenard1999) and EPICA DomeC (Lambert et al., Reference Lambert, Bigler, Steffensen, Hutterli and Fischer2012) dust records, (C) North Atlantic dust records, ODP659 (deMenocal, Reference deMenocal1995) and ²³⁰Th corrected MD03–2705 and ODP659 (Skonieczny et al., Reference Skonieczny, McGee, Winckler, Bory, Bradtmiller, Kinsley, Polissar, De Pol-Holz, Rossignol and Malaize2019) (sites near 1), (D) North Arabian Sea dust record ODP721/722 (deMencoal, 1995) (site 4), (E) Red Sea dust record K11 (Ehrmann et al., Reference Ehrmann, Wilson, Arz, Shulz and Schmiedl2024) (site 5), (F) Gulf of Aden RC09–166 terrestrial leaf wax δD (Tierney et al., Reference Tierney, deMenocal and Zander2017) (site 6), (G) speleothem-growth record derived South Arabian Humid Periods (SAHPS) (Nicholson et al., Reference Nicholson, Pike, Hosffeld, Roberts, Sahy, Woodhead, Cheng, Edwards, Affolter and Leuenberger2020) (sites 2 and 3), (H) Soreq Cave δ¹⁸O record (Bar-Matthews et al., Reference Bar-Matthews, Ayaloin and Kaufman1997) (site 8), (I) Tufa growth records from Wadi Dabsa, SW Arabia (Stone et al., Reference Stone, Inglis, Candy, Sahy, Jourdan, Barfod and Alsharekh2023) (site 7) and N Africa (summarised in Kele et al., Reference Kele, Sallam, Capezzuoli, Rogerson, Wanas, Shen, Lone, Yu, Schauer and Huntington2021) (within box 9), (J) Rössing Cave speleothem growth record (Geyh and Heine, Reference Geyh and Heine2014) (site 10), (K) Tswaing Crater rainfall proxy (Partridge et al., Reference Partridge, deMenocal, Lorentz, Paiker and Vogel1997) (site 11), (L) CD154–10-06P southwest Africa humidity record derived from Fe/K ratios for highly eroded terrestrial sediment (Simon et al., Reference Simon, Ziegler, Bosmans, Barker, Reason and Hall2015) (site 12), (M) KDE (Kernel Density Estimates) for speleothem U-Th ages using 5 ka bandwidth for Naracoorte (east) (site 13) and Leeuwin-Naturliste region (west) (site 14) speleothems in southern Australia (Weij et al., Reference Weij, Sniderman, Woodhead, Hellstrom, Brown, Drysdale, Reed, Bourne and Gordon2024), and (N) insolation curves for summer in each hemisphere to demonstrate precession-paced forcing, where solid orange line is 30°N June and dashed red line is 30°S December (from Laskar et al., Reference Laskar, Robutel, Joutel, Gastineau, Correia and Levrard2004).

Variations to the global monsoon system (Wang and Ding, Reference Wang and Ding2008) on ~23 ka (precession) cyclicity are also well established (e.g. Kutzbach, Reference Kutzbach1981; Cheng et al., Reference Cheng, Li, Sha, Sinha, Shi, Yin, Lu, Zhao, Cai, Hu, Hao, Tian, Kathayat, Cong, Zhoa and Zhang2022). This global-scale system includes the Asian Monsoon, North African Monsoon, North American Monsoon, the Indonesian-Australian Monsoon, South African Monsoon and South American Monsoon (Wang et al., Reference Wang, Wang, Cheng, Fasullo, Gui, Kiefer and Liu2017), with a strong influence on the hydroclimate of neighbouring dryland regions. For example, the well-established influence of the North African Monsoon on the hydroclimatic of the Sahara, including recent ‘greening’ during the North African Humid Period (see deMenocal and Tierney (Reference deMenocal and Tierney2012) for a brief overview). The global-scale influence of ~23 ka (precession) forcing of insolation is complicated by its occurring in antiphase between Earth’s two hemispheres. This owes to the hemisphere tilting towards the Earth (experiencing summer) moving in a cycle from a short pass to a long pass around the Sun in Earth’s elliptical orbit, such that for any year with a NH summer at short pass, the SH summer is at long pass, and vice versa. Therefore, insolation forcing is calculated and plotted for the hemisphere from which the monsoon-related response is being considered (Figure 2N shows a NH and SH summer insolation curve).

Millennial-scale variability

Climatic variability with millennial-timescale pacing was initially detected in the ice core proxies covering the last glacial (Dansgaard et al., Reference Dansgaard, Johnsen, Clausen and Langway1972, Reference Dansgaard, Clausen, Gundestrup, Hammer, Johnsen, Kristinsdottir and Reeh1982, Reference Dansgaard, Johnsen, Clausen, Dahl-Jensen, Gundestrup, Hammer, Hvidberg, Steffense, Sveinbjörnsdottir, Jouzel and Bond1993) as well as ice-rafted detritus in marine sediment cores in the North Atlantic (e.g. Heinrich, Reference Heinrich1988), with the noticeable coincidence in timing set out by Bond et al. (Reference Bond, Broecker, Johnsen, McManus, Labeyrie, Jouzel and Bonani1993) and Broecker (Reference Broecker1994). Hemming (Reference Hemming2004) considered evidence for a global climate imprint of these events and started to consider whether a connection (or teleconnection) between events had a physical basis, via ocean and atmospheric circulation. The 21 warming events from 87 to 12 ka have a quasi-periodic spacing of <2,000 years and last for 50–100 years (Wolff et al., Reference Wolff, Chappellaz, Blunier, Rasmussen and Svensson2010; Rousseau et al., Reference Rousseau, Boers, Sima, Svensson, Bigler, Lagroix, Taylor and Antoine2017). Millennial events with a similar 1470 ± 500-year cyclicity have also been observed in the Holocene (11.7 ka) in the North Atlantic (Bond et al., Reference Bond, Showers, Cheseby, Lotti, Almasi, deMenocal, Priore, Cullen, Hajdas and Bonani1997), although the magnitude of the variations is higher with larger ice sheets. The ‘agitation’ to the climate system provided by millennial-scale variability may have a role in glacial–interglacial transitions (e.g. Hodell and Channell, Reference Hodell and Channell2016). In order to explore potential teleconnections between NH high-latitude driven millennial events and low-latitude dryland climatic responses, numerical climate model simulations can be used.

Speleothem and tufa growth in drylands are good indicators for reductions in aridity (Table 1), with uranium-series dating allowing scientists to constrain their growth records over multiple 100 ka cycles. This makes them a useful archive to explore both glacial–interglacial cycles and precession forcing of the monsoon as drivers of Quaternary climate change in drylands.

On the Arabian Peninsula, the discontinuous Mukalla and Hoti Cave speleothems (14.917°N, 48.590°E and 23.083°N, 57.350°E) (Figure 1, s2,3) cover 1.1 Ma, with growth phases only occurring during peak interglacial and warm sub-stages during interglacials, and no growth during glacials (Nicholson et al., Reference Nicholson, Pike, Hosffeld, Roberts, Sahy, Woodhead, Cheng, Edwards, Affolter and Leuenberger2020). The record is used to delimit South Arabian Humid Periods (SAHPS) (Figure 2G), which show correspondence with other proxy records for humidity in the region (Nicholson et al., Reference Nicholson, Pike, Hosffeld, Roberts, Sahy, Woodhead, Cheng, Edwards, Affolter and Leuenberger2020) (Figure 1, s4–6, Figure 2D–F), and MIS 7 and MIS 5 interglacial tufa deposition at Wadi Dabsa (18.307°N, 41.564°) (Stone et al., Reference Stone, Inglis, Candy, Sahy, Jourdan, Barfod and Alsharekh2023) (Figure 1, s7, Figure 2I). The conceptual explanation for the absence of speleothem growth during glacial periods is that glacial boundary conditions ‘dampen’ monsoon strength. More specifically, large NH high-latitude ice sheets influenced atmospheric circulation, preventing on-land penetration of the summer monsoon (Burns et al., Reference Burns, Fleitmann, Matter, Neff and Mangini2001; Fleitmann et al., Reference Fleitmann, Burns, Pekala, Mangini, Al-Subbary, Al-Aowah, Kramers and Matter2011). A role for precession-forcing seems less conclusive – there is not a SAHP for every peak in insolation (at 15°N) even during an interglacial (the non-dampened state) (Nicholson et al., Reference Nicholson, Pike, Hosffeld, Roberts, Sahy, Woodhead, Cheng, Edwards, Affolter and Leuenberger2020).

Records of speleothem growth in the Negev Desert, adjacent to the Arabian Peninsula, do not align strictly with interglacials, suggesting glacial–interglacial forcing of aridity-humidity may not be widespread across NH drylands. The ~200 ka Soreq Cave (31.754°N, 35.021°E) δ¹⁸O record (Bar-Matthews et al., Reference Bar-Matthews, Ayaloin and Kaufman1997) (Figure 1, s8, Figure 2H), along with dated fragments across seven other caves, indicates four Negev Humid Periods (NHPs) (Vaks et al., Reference Vaks, Bar-Matthews, Matthews, Ayalon and Frumkin2010), which do not align strictly with interglacials: NHP-4 (MIS 10 into MIS 9), NHP-3 (MIS 8), NHP-2 (MIS 7), and NHP-1 (late MIS 6 into MIS 5). In examining a role for precession-paced forcing, Vaks et al. (Reference Vaks, Bar-Matthews, Matthews, Ayalon and Frumkin2010) find NHPs lead NH insolation peaks by several 1,000 years, with no clear explanation. The conceptual mechanism to explain monsoon-forcing of NHPs is complicated, involving a Mediterranean Sea moisture source from synoptic cyclones and an Atlantic moisture source. For the latter, a sea-surface temperature-driven reduction of the Azores High pressure system might allow the North African Summer Monsoon (Atlantic-source) to reach as far east as the Negev (deMenocal, Reference deMenocal1995, Reference deMenocal2004; Bar Matthews et al., Reference Bar-Matthews, Ayaloin and Kaufman1997).

Further west into northeast Africa, Egyptian tufa growth records also indicate deposition in glacial periods (MIS 2 to MIS 12), in addition to peak interglacials, but not with precession-pacing (Kele et al., Reference Kele, Sallam, Capezzuoli, Rogerson, Wanas, Shen, Lone, Yu, Schauer and Huntington2021) (Figure 1, s9, Figure 2I). Kele et al. (Reference Kele, Sallam, Capezzuoli, Rogerson, Wanas, Shen, Lone, Yu, Schauer and Huntington2021) propose a mechanism for ‘glacial enhancement’ of moisture in Egypt rather than the ‘dampening’ experienced in southern Arabia. When the Sunda and Sahul shelves (Indonesia and north of Australia) are exposed as landmass during glacial lowering of sea level, reduced oceanic Indonesian throughflow leads to atmospheric circulation changes that increase rainfall over East Africa. However, whilst Kele et al. (Reference Kele, Sallam, Capezzuoli, Rogerson, Wanas, Shen, Lone, Yu, Schauer and Huntington2021) draw on the CESM1 simulation of response to that glacial sea level change to support this idea (Di Nezio et al., Reference Di Nezio and Timmermann2016), the simulation only demonstrates an increase in rainfall over eastern equatorial Africa, not over Egypt.

In the SH, stalagmite and flowstone growth records from Rössing Cave, Namibia (22.531°S, 14.798°) (Figure 1, s10) date to interglacials MIS 5e, MIS 7 and MIS 11, but also extend through the MIS 10 glacial (Geyh and Heine, Reference Geyh and Heine2014) (Figure 2J). The proposed moisture source is an expansion of summer rainfall over to the west coast of southern Africa, bringing Indian Ocean Monsoon moisture, although it is not clear why Geyh and Heine (Reference Geyh and Heine2014) do not consider precession-paced forcing. Two key long proxy records for eastern southern Africa do reveal some precession-pacing: (i) Tswaing Crater sedimentary proxy record (25.416°S, 28.101°E) (Figure 1 s11, Figure 2K) (Partridge et al., Reference Partridge, deMenocal, Lorentz, Paiker and Vogel1997) and (ii) CD15410–06 (31.177°S, 32.159°E) (Figure 1, s12) humidity proxy based on Fe/K ratios for highly-weathered terrestrial sediment delivered to the ocean (Figure 2L) (Simon et al., Reference Simon, Ziegler, Bosmans, Barker, Reason and Hall2015). This example clearly illustrates the practice of different authors with respect to using the frameworks for Quaternary climatic change – Geyh and Heine (Reference Geyh and Heine2014, p. 381) frame their study by asking ‘whether the speleothems of the cave formed during glacial or interglacial periods.’

In Australia, speleothem growth records covering ~350 ka for Naracoorte Caves (~36.95°S, 140.75°E) (Figure 1, s13) and Leeuwin–Naturaliste region caves (~33.54°S, 115.02°E) (Figure 1, s14) in the southern dryland margin, are derived from speleothem ‘rubble’ pieces combined into an age-frequency plot smoothed with a kernel density estimator (KDE) (Weij et al., Reference Weij, Sniderman, Woodhead, Hellstrom, Brown, Drysdale, Reed, Bourne and Gordon2024) (Figure 2M). The KDE plots demonstrate a precessional rhythm of speleothem deposition for both caves, with good coherence with the 30°S summer (Dec 21st) insolation curve (Weij et al., Reference Weij, Sniderman, Woodhead, Hellstrom, Brown, Drysdale, Reed, Bourne and Gordon2024). The absence of growth peaks in the warmest part of the last three interglacials interrupts the precession-paced pattern, which leads Weij et al. (Reference Weij, Sniderman, Woodhead, Hellstrom, Brown, Drysdale, Reed, Bourne and Gordon2024) to suggest that a threshold of warmth might be crossed during peak interglacials, leading to a drop in moisture availability. We could term this an ‘interglacial dampening’ of moisture availability, in contrast to the ‘glacial dampening’ observed in the southern Arabian Peninsula. However, there is an absence of large KDE peaks in some of the glacial periods (e.g. during the two precession peaks later in MIS 8 and the final precession peak in late MIS 6 for Naracoorte and all of the precession peaks in MIS 8 for Leeuwin-Naturaliste) (Figure 2M), which requires further explanation.

Overall, the speleothem and tufa records from drylands and their margins do not straightforwardly show an anti-phased NH and SH moisture availability response predicted from precession-paced forcing of monsoons. Feasible mediating roles have emerged for either glacial or interglacial, end-member boundary conditions, with: a glacial dampening in southern Arabia, a glacial enhancement in north-east Africa and an interglacial dampening in Australia.

The Australian dryland margin speleothem record (Weij et al., Reference Weij, Sniderman, Woodhead, Hellstrom, Brown, Drysdale, Reed, Bourne and Gordon2024) is one line of evidence that challenges the ‘glacial aridity paradigm’ for the continent (e.g. Hesse, Reference Hesse1994; Miller et al., Reference Miller, Magee and Jull1997; Magee and Miller, Reference Magee and Miller1998). Alongside other speleothems on the northwest seaboard (Denniston et al., Reference Denniston, Polyak, Wanamaker, Ummenhofer, Humphreys, Cugley, Woods and Lucker2017) (Figure 1, s15–17), and proxy evidence in the Kati Thanda-Lake Eyre and Murray Darling Basins (LEB Figure 1, s18 and MDB Figure 1, s19), a continental-wide picture of an LGM without a substantial decrease in moisture balance emerges (Cadd et al., Reference Cadd, Williams, Saktura, Cohen, Mooney, He, Otto-Bliesner and Turney2024). Earlier LGM proxy data syntheses have variably excluded the arid interior in their treatment (e.g. Petherick et al.’s (Reference Petherick, Bostock, Cohen, Fitzsimmons, Tibby, Fletcher, Moss, Reeves, Mooney, Barrows, Kemp, Jansen, Nanson and Dosseto2013) temperate zone review), included it (e.g. De Deckker et al., Reference De Deckker, Moros, Perner, Blanz, Wacker, Schneider, Barrows, O’Loingsigh and Jansen2020), or focussed solely on it (e.g. Fitzsimmons et al., Reference Fitzsimmons, Cohen, Hesse, Jansen, Nanson, May, Barrows, Haberlah, Hilgers, Kelly, Larsen, Lomax and Treble2013), demonstrating different conceptualizations of spatial heterogeneity from the point of synthesis design. Key records from which arid (and cool) conditions have been inferred include: (i) low lake levels (e.g. Harrison, Reference Harrison1993), (ii) wind-derived terrestrial sediment in the Tasman Sea (Hesse, Reference Hesse1994) (Figure 1, s20) and in lagoonal sediments (Petherick et al., Reference Petherick, McGowan and Moss2008) (Figure 1, s21), (iii) an increased rate of dune accumulation across major desert dunefields (Hesse, Reference Hesse2016) and (iv) shifts in vegetation from trees to herbs and grasses (e.g. Dodson, Reference Dodson1975; Harle, Reference Harle1997; Colhoun et al., Reference Colhoun, Pola, Barton and Heijnis1999). The emerging consensus for a positive moisture balance during the LGM (De Deckker et al., Reference De Deckker, Moros, Perner, Blanz, Wacker, Schneider, Barrows, O’Loingsigh and Jansen2020; Weij et al., Reference Weij, Sniderman, Woodhead, Hellstrom, Brown, Drysdale, Reed, Bourne and Gordon2024; Cadd et al., Reference Cadd, Williams, Saktura, Cohen, Mooney, He, Otto-Bliesner and Turney2024) is motivation to re-examine these proxies and their interpretative frameworks.

Terrestrial dust has been critiqued as an aridity indicator (see Section title “Introduction”). For the LGM dune accumulation record, even before questioning whether a definitive link to aridity (and not primarily to windiness), there is a notable lack of spatial homogeneity. Whilst LGM accumulation rates are increased the Malee (Lomax et al., Reference Lomax, Hilgers and Radtke2011), the wider Simpson-Strzelecki contains mixed rates, suggesting “a mosaic of bare patches of sand… and large areas of vegetated, stable surface” (Hesse, Reference Hesse2016, p. 26). In the case of the vegetation records, there is a growing argument that lower CO₂ during the LGM had a greater influence than any inferred changes in moisture availability (e.g. Prentice et al., Reference Prentice, Villegas-Diaz and Harrison2022; Scheff et al., Reference Scheff, Seager, Liu and Coats2017). The lake and fluvial geomorphic records remain more enigmatic. First, there are potential spatial contrasts. In the LEB region, shorelines and basin floor sediment from Lake Callabonna-Frome and Kati Thanda-Lake Eyre indicate low or desiccated conditions, respectively, in contrast to oscillating-or-full lake conditions in the MDB region (see review by Fitzsimmons et al., Reference Fitzsimmons, Cohen, Hesse, Jansen, Nanson, May, Barrows, Haberlah, Hilgers, Kelly, Larsen, Lomax and Treble2013). Furthermore, LEB fluvial proxies (dated channel sediments and overbank sediments) yield a different hydrological record than the lake shorelines, with no discernible decrease in hydrological activity during the LGM observed in a KDE probability density curve (see Cadd et al., Reference Cadd, Williams, Saktura, Cohen, Mooney, He, Otto-Bliesner and Turney2024). It is, however, sensible to apply caution to using probability density estimates on age databases from discontinuous sedimentary archives, particularly when they are not sampled using the same vertical resolution across space (e.g. Stone and Larsen, Reference Stone and Larsen2011). Cadd et al. (Reference Cadd, Williams, Saktura, Cohen, Mooney, He, Otto-Bliesner and Turney2024) also explore precipitation minus evaporation in the transient iTRACE simulation (21 to 11 ka), which shows LGM conditions had at least as great (or greater) moisture availability than the Holocene across the whole continent. This simulation contrasts with the Fast Ocean Atmospheric Model (FOAM) time-slice simulation of reduced strength Indo-Australian summer monsoon during the LGM (Marshall and Lynch, Reference Marshall and Lynch2006) but is similar to the transient TraCE-21 ka simulation of a wetter LGM (~30% reduction in the aridity index) at the continental scale (Liu et al., Reference Liu, Jiang and Lang2021).

The ‘glacial aridity paradigm’ is also challenged in southern Africa. Two proxy syntheses indicate an area wetter than present west of ~27°E and south of ~17°S (covering the Namib Desert and most of the Kalahari), whilst the rest of the subcontinent was drier, including the eastern fringes of the Kalahari (Chase and Meadows, Reference Chase and Meadows2007; Gasse et al., Reference Gasse, Chalie, Vincens, Williams and Williamson2008). There is no clear peak in Kalahari dune building at the LGM for any part of the last ~120 ka (Stone and Thomas, Reference Stone and Thomas2008), and southern Africa is the testing ground for a strong critique of dune accumulation as a proxy for inferring dryland aridity (Chase, Reference Chase2009). Model simulations of LGM precipitation over southern Africa are variable and indicate either drier or wetter conditions (e.g. see Stone’s (Reference Stone2014) discussion of outputs from HadAM3, UGAMP and PMIP2 ensemble mean and NCAR-CCM3). TraCE-21 ka simulates an ~8% higher LGM aridity index than pre-industrial conditions for Southern Africa (Liu et al., Reference Liu, Jiang and Lang2021). In contrast, LOVECLIM transient simulations reveals a spatial contrast, with wetter-than-present conditions in the southwest (30°S, 17°E) and drier-than-present conditions in the central-Kalahari (20°S, 22°E), which is similar in nature, but not spatial boundaries, to the proxy syntheses.

The TraCE-21 ka model provides a global-scale perspective on dryland LGM conditions, simulating wetter conditions (16% lower aridity index) compared to present, as driven by lower gas concentrations (GHG) (Liu et al., Reference Liu, Jiang and Lang2021). However, there is spatial heterogeneity in the simulations: (i) wetter-than present in northern American drylands, The Mediterranean, inner Asia, northeast Brazil, southern South America and Australia, and (ii) drier-than-present in eastern Africa, the Sahel, (very slightly in north Africa), west and south Asia and southern Africa (Figure 1 inset). By separating out driving factors (orbitally-driven insolation (ORB), GHG, global ice-sheets (ICE) and meltwater forcing) using multiple simulation runs (Liu et al., Reference Liu, Jiang and Lang2021), explanations for the regional trends can be extracted: (1) NH mid-latitude drylands (northern America, inner Asia, the Mediterranean) respond most strongly to GHG and ICE; (2) NH subtropics respond mostly to ORB (the Sahel, north Africa, west and south Asia), with LGM-dryness roughly one precession-cycle ago; (3) SH mid-latitudes (Australia, southern South America) are in tandem with the NH mid-latitudes and only increase in aridity post-LGM with reduced ICE and (4) tropical drylands (eastern Africa, northeast Brazil) are also largely driven by ICE, with lower sensitivity to ORB. Southern Africa stands out from the rest of the SH in terms of a drier LGM, and with no dominant driver. Liu et al.’s (Reference Liu, Jiang and Lang2021) study is also a good demonstration of our current limitations in palaeomodel-proxy data comparisons. There are just 17 available hydroclimate proxies for (within, or near) 9 of the 11 dryland regions, with a maximum of two per region, and none for Australia. This greatly limits the conclusions we can draw from model-data comparisons and masks the spatial heterogeneity that is revealed in more complete proxy-data syntheses (e.g. Thomas and Burrough, Reference Thomas and Burrough2012; Stone, Reference Stone2021).

Overall, this example demonstrates a revision of the LGM aridity paradigm for both Australia and southern Africa back, although this is not necessarily a reversion to the idea of pluvial conditions, but rather conditions as wet, or slightly wetter than present. It also reveals complications and contradictions within the proxy record and between model simulations. Quaternary scientists approach these contradictions in different ways. For some, this represents a motivation to re-examine how some of the proxy records are interpreted in terms of the abductive reasoning back to climatic conditions (e.g. Chase and Meadows, Reference Chase and Meadows2007; Cadd et al., Reference Cadd, Williams, Saktura, Cohen, Mooney, He, Otto-Bliesner and Turney2024). For others, the conclusion is that the LGM was spatially heterogeneous within dryland regions (e.g. Thomas and Burrough, Reference Thomas and Burrough2012; Fitzsimmons et al., Reference Fitzsimmons, Cohen, Hesse, Jansen, Nanson, May, Barrows, Haberlah, Hilgers, Kelly, Larsen, Lomax and Treble2013; Stone, Reference Stone2014, Reference Stone2021).

Luminescence-dated shorelines of Lake Chilwa in the savannah to semi-arid climatic region of Malawi, ~15°30′S (Figure 1, s22), revealed lake highstands at 38.4–35.5, 24.3–22.3, 16.2–15.1, 13.5–12.7, 11.01 ± 0.76 and 8.52 ± 0.56 ka, (Thomas et al., Reference Thomas, Bailey, Shaw, Durcan and Singarayer2009) (Figure 3A). The first three have a coincidence of timing with North Atlantic Heinrich Events/Stadials (HS) 4, 2 and 1, and the ~11 ka highstand with the Younger Dryas (Figure 3A). A HadCM3 climate model simulation was used to explore a potential teleconnection. It showed that a reduction of the Atlantic meridional overturning circulation in response to a North Atlantic Heinrich iceberg discharge event has atmospheric teleconnections, with increases the moisture balance at Lake Chilwa (Thomas et al., Reference Thomas, Bailey, Shaw, Durcan and Singarayer2009) (Figure 3B). The spatial pattern in the simulation is for increased moisture balance over much of southern Africa (south of 15°S) but drier conditions north of the equator (Figure 3B). This invites future model-proxy comparisons for the African continent in response to millennial-scale climatic forcing.

Figure 3. (A) Timing of highstands of Lake Chilwa from luminescence dating of shorelines, compared to the timing of Heinrich Events, (B) HadCM3 simulation of terrestrial moisture balance over Africa caused by a ‘freshwater hosing’ 50–70°N to simulate a North Atlantic Heinrich Event, displayed as an anomaly (modelled years 275–300 minus the average of 0–100 years), where the inset shows modelled reduction in Atlantic Meridional Overturning Circulation (modified from Thomas et al., Reference Thomas, Bailey, Shaw, Durcan and Singarayer2009).

Millennial-timescale change can also be examined using high-resolution (~20–30 μm/y) banded hyraceum (fossilised urine) in hyrax middens on the western margin of southern Africa. The midden record composite δ¹⁵N from three dryland-zone middens (Pella, South Africa: 29.00°S, 19.14°E; Zizou: 24.07°S, 15.97°E and Spitzkoppe: 21.83°S, 15.20°E, Namibia) (Figure 1, s23–25) contains evidence of phases of increased humidity, which Chase et al. (Reference Chase, Niedermeyer, Boom, Carr, Chevalier, He, Meadows, Ogle and Reimer2019) correlate with HS3, 2 and 1. De Rif midden, just south of the current dryland zone (32.446°S, 19.221°E) (Figure 1, s26), is also characterised by humid conditions during HS1 (Chase et al., Reference Chase, Quick, Meadows, Scott, Thomas and Reimer2011). Although not considered using a simulation experiment, the driving mechanism explored is an oceanic teleconnection with NH millennial events, via their influence on the southern Atlantic (Benguela upwelling zone) sea-surface temperatures (SST), and in turn the influence SSTs have on rainfall over Southern Africa’s western margin. This draws on the good temporal correspondence between SST records in the South Atlantic Benguela upwelling zone (Farmer et al., Reference Farmer, deMenocal and Marchitto2005; Kim and Schneider, Reference Kim and Schneider2003) and the hyrax aridity proxies (Chase et al., Reference Chase, Quick, Meadows, Scott, Thomas and Reimer2011). However, unlike for the Chilwa Lake example, the hydroclimatic response of the De Rif midden during the YD (which also involves a North Atlantic meltwater event) is markedly different than for the HS events, with an abrupt drying signal. Chase et al. (Reference Chase, Quick, Meadows, Scott, Thomas and Reimer2011) invoke an immediate atmospheric interhemispheric teleconnection, as opposed to the oceanic teleconnection with Benguela SST-driven hydroclimatic control for the HS events. These nuances are complex, and a paucity of continuous and high-resolution terrestrial proxies in African drylands restricts the possibility to test and refine these interpretations, although transient numerical model simulations for discrete spatial locations may provide further insights.