1. INSTAAR, Univ. Colorado, Boulder, Colorado, USA gmiller@colorado.edu
2. Dept. of Geology, ANU
3. Geophysical Laboratory, Carnegie Inst. Washington, Wash. D.C., USA
4. RSES, ANU

Based on our studies at Lake Eyre and Lake Victoria, we hypothesize that systematic burning by early human colonizers led to rapid ecosystem change across semi-arid northern Australia. This change in vegetation structure and density led directly to the extinction of the Australian megafauna, and through a reduced transfer of moisture from biosphere to atmosphere, to a weakening of monsoon moisture penetration into the continental interior. The primary evidence on which this hypothesis is based is 1) the presence of a deep-water phase in Lake Eyre 60 ka ago, when monsoon forcing was modest, and the lack of a deepwater lake in the Holocene when monsoon forcing was stronger (Magee, 1997), 2) evidence that monsoon-dependent grasses were abundant around Lake Eyre 60 ka ago, and have been much less abundant through the Holocene (Johnson et al., 1999), and 3) our dates on the timing of megafauna extinction (Miller et al., 1999). The concept is supported by GCM simulations suggesting that the sensitivity of monsoon moisture penetration to vegetation density over northern Australia is high (2 to 4 mm day^-1 during the wet season). This hypothesis has specific predictions for monsoonal Australia: it predicts that the wettest phase of the last 150 ka occurred early in the last interglacial, with lesser wet phases later in stage 5, about 60 ka ago, and the Holocene. The Holocene monsoon is predicted to be most intense in the early Holocene. Finally, we predict that much of the northern semi-arid zone carried a C3-dominated shrub-tree vegetation 60 ka ago, changing to C4 desert scrub by ca. 50 ka ago and remaining so through to the present.

Previous work at Lake Gregory has shown that a megalake phase more than 10 times the modern lake occurred within the past 300 ka (Wyrwoll et al., 2000). Stratigraphic and geomorphic studies across monsoonal Australia have suggested that the summer monsoon was re-invigorated in the Holocene after a period of quiescence during the Last Glacial Maximum (LGM). However, it has been difficult to date the megalake phase and to find records that span the entire Holocene to test when the monsoon was reestablished.

Lake Gregory: Interbedded littoral sands and carbonates mark a distinctive shoreline during the megalake phase of Lake Gregory. We surveyed the shoreline at 295 m asl, 25 m above the current lake. We refined earlier estimates of the megalake shoreline by superimposing the leveled shoreline elevation on a regional DEM and verified the simulated lake outline in the field by identifying lacustine sediments in augered holes (up to 10-m-deep) in the major basins (Fig. 1). To test the age of the megalake phase we collected shoreline ironstones for U/Th dating, and sampled beach and dune sands below the shoreline for OSL dating (dating in progress at RSES, ANU). Carbonate deposited at the 295-m level during the last highstand has a ¹⁴C age ³40,600 (AA-33400). Based on changes in dune morphology, we suggest that dunes dated more than 200 ka (J. Bowler and K.-H. Wyrwoll, pers. comm., 1998) may have been partially reworked by subsequent high lake events, hence may provide minimum dates on the megalake phase. Dunes from late stage 5 have not been reworked.

A 90-cm-thick section of post-LGM lacustrine mud at Gilwah Waterhole, nearly 8 m above the floor of modern Lake Gregory, contains common limnic gastropods and bivalves at the base of the section. One gastropod has been dated by AMS ¹⁴C at 12,505±80 BP. Large (30 cm) calcite rhyzoliths at about the same elevation that formed when regional precipitation was high and Ca⁺² was mobilized, have a date of 11,145±75 BP. These two dates support our prediction that the summer monsoon was reinvigorated at the end of the LGM during a Southern Hemisphere summer insolation minimum, consistent with our contention that Northern Hemisphere conditions exert a dominate control on the Australian Monsoon.

A 6-m-thick sequence of interbedded lacustrine, fluvial, and eolian sediment is exposed along lower Sturt Ck; augering allowed us to recover a continuous section more than 9 m deep. At least two, and possibly four eolian cycles are represented in the section, representing >100 ka of episodic sedimentation. Carbon isotopes in the bulk organic matter preserved in this section are primarily of terrestrial origin. They indicate dominance by C3 vegetation except in the upper meter of section, where there is a rapid shift to much heavier ¹³C values, suggesting a replacement of regional C3 vegetation by C4 grasses. Although the dating is still uncertain, the pattern is consistent with our hypothesized environmental changes.

Wolfe Creek Crater is a roughly circular impact crater formed in late Precambrian quartzitic sandstone; the floor is 25 m below the surrounding landscape. The crater’s geometry precludes excavation of sediment by eolian processes, and the crater is expected to contain a continuous sedimentary record since impact. Clastic sediment input comes from crater wall erosion and from desert dust off the surrounding dune fields, the latter likely to be enhanced during arid phases. The crater also acts as a ground-water window. We recovered two short (4 m) and one long (10 m) augered sections. The physical stratigraphy documents alternating periods of dominance by red desert dust, lacustrine sedimentation, and periods of landscape stability and pedogenesis. Groundwater gypsum dominates the upper 4.5 m, punctuated by intervals of clastic sediment input. Magnetic susceptibility (MS; Fig. 2) tracks the flux of desert dust (oxidized Fe-rich clays coating fine-grained quartz grains) as spikes in MS, as well as the general increase in magnetic minerals below 4.5 m. Oogonia, which occur at several levels, have been ¹⁴C dated at 75 cm (4970±65 BP) and 368 cm (9820±90 BP) depths. Dates on humins appear less reliable. We tentatively estimate that the gypsum-dominated upper 4.5 m of section represents relatively high water tables of the last 10 to 12 ka, and that the dominantly minerogenic sediment of the deeper levels was deposited much more slowly, representing at least the previous 60 ka of record. Bulk carbon in these sediments is dominantly of aquatic origin; pollen is not preserved.

Although secure conclusions await the results of ongoing dating and measurements of specific environmental proxies, the preliminary results currently available support several predictions of our hypothesis derived from more southerly sites.