Catchment vegetation and erosion controls soil carbon cycling in 1 south-eastern Australia during the last two Glacial-Interglacial 2 cycles

Francke, A.1-3* , Dosseto, A. 1,2,4, Forbes, M. 4,5, Cadd, H.4,6, Short, J.7,8, Sherborne-Higgins, B.2,4, 5 Constantine, M. 6, Tyler, J.3, Tibby, J.7,8, Marx, S.2,4, Dodson, J.9, Mooney, S. 6, and Cohen, T.J.2,4 6 7 8 1 Wollongong Isotope Geochronology Laboratory, School of Earth, Atmospheric and Life Sciences, 9 University of Wollongong, Australia 10 2 GeoQuEST Research Centre, School of Earth, Atmospheric and Life Sciences, University of 11 Wollongong, Australia 12 3 Department of Earth Sciences, University of Adelaide, Australia 13 4 ARC Centre of Excellence for Biodiversity and Heritage, School of Earth, Atmospheric and Life 14 Sciences, University of Wollongong, Australia 15 5 KCB Australasia pyt ltd, Brisbane, Australia 16 6 School of Biological Earth and Environmental Sciences, University of New South Wales, 17 Australia 18 7 Department of Geography, Environment and Population, University of Adelaide, Australia 19 8 Sprigg Geobiology Centre, University of Adelaide, Australia 20 9 Institute for Earth Environment, Chinese Academy of Science, Xi’an, Shaanxi, China 21 22 * corresponding author: alexander.francke@adelaide.edu.au 23


Introduction
of national greenhouse gas emissions in Australia, for example, has suggested that not considering 48 cropland soil erosion overestimated the nation's net carbon flux into the atmosphere by 40% 49 (Chappell et al., 2015), explained by SOC lost by erosion and subsequently buried in sedimentary sinks 50 rather than being re-oxidised (Chappell et al., 2015). This has led to an ongoing debate as to whether 51 soil-carbon is a net atmospheric source or sink in the global carbon cycle (Chappell et al., 2015;52 Doetterl et al., 2016;Lugato et al., 2018), primarily related to gaps in our understanding of how lateral 53 soil fluxes connect terrestrial and aquatic carbon cycling (Luo et al., 2016). These uncertainties 54 become greater when constraining carbon fluxes on geological timescales, where "land use will be provided once the manuscript has been accepted. We encourage feedback to the authors.
separately. Preferential leaching can promote lower ( 234 U/ 238 U) activity ratios in detrital grains that 175 is not related to recoil-loss of 234 Th. Different scenarios for pre-and post-depositional preferential 176 leaching of 234 U and different scenarios for (reduced) loss of 234 Th by recoil after final deposition were 177 tested (see Supplement). Palaeo-sediment residences times were calculated for LC2 using Monte 178 Carlo simulations (10,000 simulations) to account for uncertainties in our input parameters using an 179 R64 script (available upon request to the authors). 180 Partial Least Square Regression (PLSR) analysis was performed on previously published pollen relative 181 abundance data (as predictor variables) and palaeo-sediment residence times (as response variables) 182 to consider relationships between vegetation change and catchment erosion. Aquatic and semi-183 aquatic pollen taxa (such as Cyperaceae) were excluded from the total pollen sum and relative 184 abundances were re-calculated for terrestrial taxa (Forbes et al., 2021, supplement). 185 Chenopodiaceae pollen was excluded from the terrestrial pollen sum as it occurred at very low 186 percentages, likely indicating it is derived by windblown transportation from outside the catchment 187 (Dodson, 1983;Williams et al., 2006). Re-calculated relative abundances of Myrtaceae + 188 Casuarinaceae, Acacia (genus), Asteraceae (Asteroideae or Tubuliflorae), and Poaceae and palaeo-189 sediment residence times were normalized (mean = 0, standard deviation = 1) prior to PLSR analyses. 190 PLSR analyses was undertaken for current and last glacial-interglacial sediments separately. 191 in 234 U (Fig. 3). Uranium isotope analyses of the 50 cm deep soil pit WBRC1 located on the ridge crest will be provided once the manuscript has been accepted. We encourage feedback to the authors. of neighbouring Lake Werri Berri yielded ( 234 U/ 238 U) activity ratios between 0.856 and 0. 892, thus 198 showing expected recoil-induced depletion of 234 U in fine-grained (<63 µm) detrital matter. The 199 activity ratios increased with greater soil depth (Fig. 3). 200 K/Ti ratios of bulk-detrital soil samples between 0 cm and 35 cm depth in WBRC1 are low and steady 201 (K/Ti = 0.4 to 0.5, Fig. 3). High ratios occur in the sample taken at 50 cm depth and in the underlying 202 saprolite (K/Ti = 2.2). SOC, as inferred from WBRC1 soil-TOC, is between ~1 and 3.9 %, with values 203 >1.5% only found in the upper 20 cm. The OM-rich topsoil layer overlayed a thick sandy horizon with 204 poor vertical soil stratification, classifying WBRC1 as skeletal soil. 205

206
Monte-Carlo modelled palaeo-sediment residence times (i.e. time elapsed between comminution 207 and final deposition reported in kyr, Fig. 2) ranged from 15 to 70 kyrs in sediments deposited between 208 133 ka to 130 ka (broadly equivalent to MIS 6), between 115 ka and 107.6 ka (broadly equivalent to 209 MIS5d), and between 17.8 cal ka BP and 11.6 cal ka BP (Late Glacial), respectively. Longer residence 210 times, between 70 kyrs and 124 kyrs, were evident between depositional ages of 130 ka and 115 ka 211 (broadly equivalent to MIS5e), between 17.8 cal ka BP and 16 cal ka BP, and during the Holocene 212 (11.6 ka to present day). 213 Terrestrial pollen taxa abundance indicate Lake Couridjah's catchment vegetation was composed of 214 sclerophyll trees and shrubs (Myrtaceae + Casuarinaceae between 44% and 86%) during both the last 215 (133.5 ka to 107.6 ka) and current (17.8 cal ka BP to present day) glacial-interglacial (Fig. 4). The 216 pollen of Acacia (0.6% to 3.5%) genus occurs at low abundance during both climate cycles. Herb and 217 grassland vegetation cover, comprising of Asteraceae (Asteroideae or Tubuliflorae, 1% to 48%) and 218 Poaceae (2.5% to 22.5%) contributed substantially to the vegetation composition in the Thirlmere 219 catchment during both glacial-interglacial cycles. Broadly, higher proportions of arboreal taxa 220 (Myrtaceae + Casuarinaceae and-or Acacia) and lower proportions of understorey taxa (Asteraceae will be provided once the manuscript has been accepted. We encourage feedback to the authors. and Poaceae) occurred during warmer and wetter intervals (130 ka and 115 ka and during Holocene, 222 Fig. 4). Peaks in macroscopic charcoal area (mm 2 /cm 3 /yr) between 130 ka and 115 ka and during Last 223 Glacial corresponded to declines in Myrtaceae + Casuarinaceae and-or Acacia (Fig. 4). Stable 224 Polycyclic Aromatic Carbon (SPAC) abundance was higher between 130 ka and 115 ka, 106 ka and 225 103 ka, and during the Holocene (Fig. 4). 226 PLSR analyses of terrestrial pollen taxa (Myrtaceae + Casuarinaceae, Acacia, Asteraceae, Poaceae) 227 and palaeo-sediment residence times were preformed separately for the last and current glacial-228 interglacial cycle (Fig. 5). The statistical analysis reveals strong positive loading on predictor axis 1 for 229 Myrtaceae + Casuarinaceae and Acacia during the last glacial-interglacial (Fig. 5). Only Myrtaceae + 230 Casuarinaceae shows strong positive loadings on predictor axis 1 for the current glacial-interglacial 231 (between 17.8 cal yr BP and present). Asteraceae and Poaceae have strong negative loadings on PLSR 232 predictor axis 1 in the last glacial-interglacial (133 ka to 107 ka). Strong negative loadings for 233 Asteraceae and Poaceae and weak negative loadings for Acacia occur on PLSR predictor axis 1 for the 234 current glacial-interglacial. PLSR predictor axis 1 is consequently indicative for the catchment 235 vegetation structure. Monte-Carlo modelled palaeo-sediment residence times showed strong 236 positive loadings on response axis 1 for both the last and the current glacial-interglacial cycle (Fig. 5). 237 A significant correlation is identified between the PLSR scores from predictor axis 1 (combined results 238 for the last and current glacial-interglacial cycle) and catchment sediment residence times (Fig. 7A, 239 R 2 = 0.48, p <0.005). No significant correlation is observed between K/Ti and palaeo-sediment 240 residence times (Fig. 7G, R 2 = 0.19, p >0.005), but a significant negative correlation is found between 241 K/Ti and Myrtaceae + Casuarinaceae abundance (Fig. 7H, R 2 = 0.52, p <0.005). 242 The accumulation of organic carbon in the sediments of Lake Couridjah, as inferred from TOCacc, 243 resembles the variability recorded in palaeo-sediment residence times (Fig. 6). Overall higher TOCacc 244 occured in both peak interglacials (130 ka to 115 ka, and the Holocene). Somewhat higher will be provided once the manuscript has been accepted. We encourage feedback to the authors. abundances of aerophilic + epiphytic diatoms (up to 77%) were recorded between 133.5 ka and 115 246 ka and during the Holocene (Fig. 6), while lower abundance (less than 45%) occurred between 115 247 ka and 107.6 ka, when planktonic diatoms were abundant. Planktonic diatom abundance was low 248 and variable between 133.5 ka and 115 ka, low and steady during the Last Glacial, and almost absent 249 during the Holocene (Fig. 6). 250 In soil pit WBRC1, higher ( 234 U/ 238 U) activity ratios occur at greater soil depth, close to the bedrock-261 weathering horizon interface, and lower ( 234 U/ 238 U) activity ratios occur in topsoil layer (Fig. 3). This where temporary storage in the fluvial system is considered shorter than 10,000 years (Li et al., 2018; will be provided once the manuscript has been accepted. We encourage feedback to the authors. there is no depth-dependent variability in K/Ti, unless saprolite is mobilised by deep erosion (Fig. 3). 291

Discussion
This manuscript is a non-peer reviewed EarthArXiv pre-print. A DOI for the peer-reviewed version will be provided once the manuscript has been accepted. We encourage feedback to the authors. predictor axis 1 (summarized as canopy and mid-storey cover), PLSR response axis 1 (vegetation-314 dependent soil erodibility), and modelled palaeo-sediment residence times (Fig. 6, 7A). will be provided once the manuscript has been accepted. We encourage feedback to the authors. A meta-analysis of 24 Late Pleistocene to Holocene pollen records has recently shown that vegetation 331 turnover and richness in SE Australia is mainly controlled by moisture (via tropical and westerly wind 332 systems) and sea-level change (controlling oceanic climates, Adeleye et al., 2020). This is consistent 333 with the expansion of herb and grass vegetation during periods of reduced regional precipitation in 334 the Thirlmere catchment between 133.5 ka and 130 ka and between 17.8 cal ka BP and 11.6 cal ka 335 BP . This supports that moisture has negative feedback on catchment-wide 336 erosion at Thirlmere, with drier climates not promoting slower erosion, but rather faster and deeper 337 erosion, due to the reduction of canopy and mid-storey cover as the vegetation underwent structural 338 change. These findings are also supported from the Murrumbidgee River palaeochannel (300 km SE will be provided once the manuscript has been accepted. We encourage feedback to the authors.

The last and current glacial-interglacial cycle
of Lake Couridjah), which has been shown to have palaeo-sediment residence times an order of 340 magnitude lower during glacial compared to interglacial periods (Dosseto et al., 2010). 341 Although occurring at low values only, any occurrence of Acacia pollen (between 0.6 and 3.5 %, pollen 342 count generally < 6, Fig. 4) has been demonstrated to indicate the presence of these species in the 343 mid-to upper strata of sclerophyllous vegetation communities (Dodson, 1983 Asteraceae (Fig. 4, 8). This trend is, however, not mirrored in Acacia and Poaceae, with both taxa 357 appear broadly anticorrelated during this time interval. The high PLSR-derived index of canopy and 358 mid-storey cover between 127 ka to 115 ka is therefore (statistically) mainly controlled by Acacia 359 (Fig. 4), which might imply a relatively high importance of the mid-story vegetation patterns and-or 360 upper canopy strata in the vicinity of the lake for the prediction of vegetation-dependent soil 361 erodibility during the last interglacial cooling (Fig. 6). 362 This manuscript is a non-peer reviewed EarthArXiv pre-print. A DOI for the peer-reviewed version will be provided once the manuscript has been accepted. We encourage feedback to the authors.
High and stable palaeo-sediment residence between 127 and 123 ka, and between 120 to 115 ka 363 imply slower and shallower erosion during periods of increased fire activity and fire-mediated 364 vegetation disturbance (high charcoal surface area > 1 mm 2 /cm 3 /yr, high sedimentary SPAC content). 365 Frequent disturbance by fire is probably also indicated by highly variable Acacia, and, to some degree 366 Poaceae, which both respond rapidly to fire disturbance, throughout the Late Glacial and Holocene. 367 The high variability in Acacia may (statistically) explain the weak negative loading on PLSR predictor 368 axis 1 during the current glacial-interglacial, implying that Acacia was less significant in controlling 369 soil erosion during the current, compared to the last, glacial-interglacial cycle (where Acacia shows a 370 strong positive loading on predictor axis 1). Frequent disturbance of Acacia may be of particular 371 importance during the Late Glacial, where macroscopic charcoal values are high (Fig. 6). This manuscript is a non-peer reviewed EarthArXiv pre-print. A DOI for the peer-reviewed version will be provided once the manuscript has been accepted. We encourage feedback to the authors. and-or (ii) by post-fire rainfall characteristics that may control weak post-fire erosion event only 387 (Tomkins et al., 2008). 388 Overall short and variable palaeo-sediment residence times (< 50 kyrs) between 115 ka and 107.6 ka 389 show very similar patters to canopy and mid-storey indicators in the pollen record (Myrtaceae + 390 Casuarinaceae, PLSR-derived predictor axis 1), and predicted vegetation-dependent soil erodibility 391 (Fig. 6). Short  parts of the lake via increased runoff and/or wave action, material that may have previously spent 408 an extended period stored in the catchment. This is also consistent with Forbes et al. (2021)'s will be provided once the manuscript has been accepted. We encourage feedback to the authors.
interpretation that the major hiatus between 107 ka and 17.38 cal ka BP likely resulted from a 410 combination of slow catchment erosion and aeolian deflation. 411

Soil development 412
The lower K/Ti in WBRC's soil samples compared to saprolite samples from the same location implies 413 K/Ti ratios can be used as an indicator for the degree of chemical weathering and soil development 414 in the lake's catchment. The absence of correspondence between K/Ti and palaeo-sediment 415 residence times for the last and current glacial-interglacial cycle implies a de-coupling between soil 416 development and catchment erosion (Fig. 7H). 417 Warmer and wetter climates as well as tree and forest growth is thought to represent the most 418 important processes affecting chemical weathering (Sverdrup, 2009). A strong control of root-soil 419 structure interactions and warmer and wetter climates on soil development in the Thirlmere 420 catchment is also supported by the moderate strong, statistically significant negative relationship 421 between K/Ti and Myrtaceae + Casuarinaceae (Fig. 7H). 422 Limited soil development between 115 and 110 ka is inferred from high K/Ti and corresponds to the 423 transition to relatively open grass and herb vegetation cover, low root-soil structure interactions (low 424 Myrtaceae + Casuarinaceae), and to faster and deeper erosion (low palaeo-sediment residence time, 425 Fig. 4, 7). This indicates a negative feedback between vegetation cover, catchment erosion, and soil 426 development during the late last interglacial cooling phase. 427

Catchment-wide carbon cycling 428
Overall faster erosion of thinner soils (short palaeo-sediment residence times, high PLSR vegetation-429 dependent soil erodibility) during colder and drier intervals (133.5 ka to 130 ka, 115 ka to 17.6 ka, 430 Late Glacial) could have resulted in high SOC erosion rates, rapidly degrading the relatively thin OM 431 rich topsoil layer described for the modern Thirlmere catchment (section 3.1, Fig. 8). Deeper and 432 faster erosion could have reduced soil-microbial respiration of OM rich topsoil, since material is 433 This manuscript is a non-peer reviewed EarthArXiv pre-print. A DOI for the peer-reviewed version will be provided once the manuscript has been accepted. We encourage feedback to the authors. (macrophytes, sedges) are inferred from the low abundance of aerophilic + epiphytic diatom taxa 451 . Higher planktonic diatom abundances indicate higher lake levels, and the 452 expansion of phytoplankton habitats at this time (Fig. 6). This may indicate that carbon sequestration 453 in Lake Couridjah was more controlled by aquatic productivity of phytoplankton, particularly 454 between 115 ka and 107.6 ka, where planktonic diatom abundance was high (Fig. 6). However, the 455 low TOCacc, despite the high planktonic diatom abundance (Fig. 6), may indicate that these high lake 456 level phases have a reduced capacity for OM-biomass accumulation, compared to intervals with high 457 This manuscript is a non-peer reviewed EarthArXiv pre-print. A DOI for the peer-reviewed version will be provided once the manuscript has been accepted. We encourage feedback to the authors.
TOCacc and high aerophilic and epiphytic diatom abundance. This is probably due to the low amount 458 of swamp vegetation (providing the substrate for the epiphytic diatoms) being a strong contributor 459 to the lake TOCacc . In summary, we infer a lower atmospheric carbon 460 sequestration in the Thirlmere catchment between 133.5 ka and 130 ka, between 115 ka to 107.6 ka, 461 and during the Late Glacial (compared to the wetter and warmer intervals). Climates were overall 462 colder and drier, resulting in low catchment productivity, deeper and faster erosion, re-463 mineralisation of old carbon stored at greater soil depth, limited nutrient supply to the lake, and 464 limited primary productivity by phytoplankton and aquatic to semi-aquatic plants living in the lake 465 ( Fig. 8). have also fostered longer and deeper SOC storage via bioturbation by roots (Fig. 8). 473 High TOCacc implies high net-carbon accumulation in Lake Couridjah between 130 ka and 115 ka and 474 during Holocene. High TOCacc is attributed to higher primary productivity in the lake basin, as is 475 observable today (Fig. 1). Significantly higher aerophilic and epiphytic diatom abundance implies the 476 increase in TOCacc was mainly related to aquatic to semi-aquatic (swamp) vegetation in the lake. 477 Somewhat lower planktonic diatom abundance consequently implies a reduction in planktonic 478 habitats, and lower lake levels. Aquatic and semi-aquatic productivity may be fostered by higher 479 nutrient supply from the catchment during the overall warmer and wetter climates of these periods. 480 This manuscript is a non-peer reviewed EarthArXiv pre-print. A DOI for the peer-reviewed version will be provided once the manuscript has been accepted. We encourage feedback to the authors.
Significant contributions of terrestrial OM to the lake carbon-pool may have originated from 481 relatively weakly decomposed topsoil SOC mobilised by shallower erosion (section 3.1, Fig. 8). 482 Additionally, more terrestrial OM supply during warmer and wetter intervals despite reduced 483 erosion, could probably be explained by an expansion of a canopy and mid-storey cover (Fi. 6). A 484 canopy and mid-storey cover produces significant amounts of easily transportable, loose leaf litter in 485 the catchment (Gordon et al., 2018), as also observed at the present day (Fig. 1), while a more closed 486 canopy and roots prevents deeper soil erosion despite wetter conditions. In summary, we infer high 487 atmospheric carbon sequestration both in the catchment and the wetland during the overall warmer 488 and wetter periods between 130 ka and 115 ka and during the Holocene (Fig. 8). 489 This manuscript is a non-peer reviewed EarthArXiv pre-print. A DOI for the peer-reviewed version will be provided once the manuscript has been accepted. We encourage feedback to the authors.

Conclusions
Erosion was high during colder and drier periods (133.5  Scheme (Project Number CE170100015). Special thanks also goes to Emily Barber, Andres Zamora, This manuscript is a non-peer reviewed EarthArXiv pre-print. A DOI for the peer-reviewed version will be provided once the manuscript has been accepted. We encourage feedback to the authors.  This manuscript is a non-peer reviewed EarthArXiv pre-print. A DOI for the peer-reviewed version will be provided once the manuscript has been accepted. We encourage feedback to the authors. coefficients and probabilities for correlations between isotope and elemental data are reported in 547 Mytraceae + Casuarinaceae, and K/Ti ratios of core LC2. The data are presented for the entire core will be provided once the manuscript has been accepted. We encourage feedback to the authors. LC2, i.e. statistical analyses presented in Fig. 7 were not carried out separately for the current and 570 last glacial to interglacial cycle as conducted for PLSR analyses. 571 572 Fig. 8: Conceptual model of vegetation change, catchment erosion, SOC-mobilisation, and lake-573 productivity in the Thirlmere catchment during warmer and wetter (peak-last interglacial, Holocene) 574 and colder and drier (penultimate glacial, late-last interglacial and Late Glacial) periods. Letters A to 575 D mark different landscapes in the Thirlmere catchment and are the same as for Fig. 1D  This manuscript is a non-peer reviewed EarthArXiv pre-print. A DOI for the peer-reviewed version will be provided once the manuscript has been accepted. We encourage feedback to the authors.