Late Pleistocene emergence of an anthropogenic fire regime in Australia’s tropical savannahs

The study investigates when Australia’s tropical savannahs shifted from a natural, lightning-driven fire regime to one dominated or modulated by human (Indigenous) management. Fire is a pervasive ecological force and human tool, with deliberate use dating back ~1 million years and habitual use by 250–350 ka. In Australia, Indigenous fire management (“firestick farming”) was widespread at European arrival and is considered key to landscape and biodiversity patterns. However, distinguishing prehistoric anthropogenic influences on fire regimes from climatic controls is challenging, especially in savannahs where fire is frequent and vegetation is fire-adapted. Prior records are often too short to characterize natural variability. The authors aim to use a continuous, 150,000-year sediment record from Girraween Lagoon (Northern Territory, Australia) and multiple fire and vegetation proxies to test whether, and when, Indigenous fire management became a dominant control on fire regime and ecosystem dynamics, hypothesizing a detectable transition from less-frequent, more-intense (natural) to more-frequent, less-intense (anthropogenic) fires.

The paper situates its research within several strands: (1) Global and Australian wildfire trends increasingly influenced by human activity and climate change, with catastrophic events documented. (2) Indigenous fire management as a long-standing land-use practice that shaped biodiversity and ecosystem structure, including in northern Australia. (3) The difficulty of detecting anthropogenic fire signatures in prehistoric records, especially in tropical savannahs with frequent natural fires. (4) Previous Australian records, including the debated Lynch’s Crater micro-charcoal increase ~40 ka, which meta-analyses did not support as a continent-wide human impact. (5) Studies elsewhere often detect hunter-gatherer fire influence mainly from the mid- to late Holocene (Australia, southern Africa, North America). The authors argue that few long records exist to define a natural baseline and that integrating multiple proxies over long timescales is necessary to separate climatic from human drivers.

Study site and core: Girraween Lagoon (12°31′3.6″S, 131°04′50.7″E; 25 m a.s.l.) near Darwin, NT, Australia. Monsoonal climate. A 19.4 m sediment core was collected from 4.5 m water depth using a floating platform and hydraulic corer. The sequence comprises alternating peats (wet periods, high lake level) and clays (dry periods, low lake level). A total of 235 samples were taken at 5–10 cm intervals. Chronology: Established using 12 radiocarbon dates (with hydrogen pyrolysis pretreatment) and 24 optically stimulated luminescence (OSL) ages on single quartz grains. Bayesian age-depth modeling was performed in R using rbacon with SHCal20 calibration, combining 14C and OSL ages. Age model outputs were interpolated at 250-year intervals (average raw resolution ~550 years). Proxies and laboratory analyses:

• Total organic carbon (TOC) used as a wetness indicator and to assess intervals where micro-charcoal particle accumulation rate (PARchar) is reliable (TOC > ~2–5%).
• Fire incidence/intensity proxies: (1) PARchar from counts of microscopic charcoal particles (>10 μm) in pollen slides, converted to accumulation rates using dry bulk density and modeled sedimentation rates; (2) Mass accumulation rate of stable polycyclic aromatic carbon determined by hydrogen pyrolysis (MARSPAC), a geochemical indicator of high-temperature pyrogenic carbon (≥7 condensed aromatic rings); (3) δ13C of SPAC (δ13CSPAC) as a relative indicator of the proportion of grass (mainly C4) vs woody (C3) biomass preserved in char, also modulated by fire intensity (higher-intensity fires preferentially combust fine grassy fuels, lowering δ13CSPAC).
• Vegetation proxy: Grass pollen as a percentage of total dryland pollen (percentage of C3 pollen reported; higher values indicate more wooded conditions). Pollen identification used regional libraries and the Australasian Pollen and Spore Atlas. Lycopodium spike added for concentration estimates; sieving at 125 μm and 7 μm. Standard chemical pretreatments applied. Hydrogen pyrolysis and isotope analyses: SPAC isolated via HYPy with Mo catalyst under high H2 pressure; carbon abundance and δ13C measured via EA-IRMS at James Cook University. Calibration against international standards; typical δ13C uncertainty ±0.2‰; SPAC percentages corrected for in situ production. Data processing and statistical analysis: Time series interpolated every 250 years. To mitigate bias from char–clay interactions that comminute micro-charcoal below 10 μm in clay-rich intervals, analyses comparing PARchar and MARSPAC were filtered by TOC thresholds (tested from >10% down to >1%; primary thresholds >5.1% and >2%). The record was split at an initial temporal threshold of 30 ka to represent pre-human natural vs anthropogenic fire regimes; sensitivity to the temporal split was tested from 65 to 1 ka. Relationships examined: (a) MARSPAC vs PARchar (both log10-transformed) with the expectation of a stronger positive relationship under the natural, higher-intensity regime; (b) percentage of C3 pollen vs δ13CSPAC with the expectation of a more pronounced positive relationship under the anthropogenic, lower-intensity regime. Statistical significance assessed via resampling: for each TOC/temporal threshold combination, 10,000 iterations resampled x–y pairs from the natural period to match the sample size of the anthropogenic period and compared Spearman’s ρ values to produce probabilities that observed differences could occur by chance. Sensitivity of results was plotted versus TOC threshold and temporal split. Additional qualitative comparisons were made across Marine Isotope Stages (MIS) and insolation peaks.

• A robust shift in fire regime is detected with high statistical certainty by at least 11 ka BP: from less-frequent, more-intense (natural) fires to more-frequent, less-intense (anthropogenic) fires associated with Indigenous management.
• Holocene mean PARchar was 42,200 particles cm−2 yr−1, more than double the previous two highest multi-millennial peaks over the past 150 ka (16,700 and 10,500 particles cm−2 yr−1 near 114 and 92 ka), indicating higher fire incidence in the Holocene.
• MARSPAC ranges over three orders of magnitude (1–131 mg cm−2 yr−1); consistently high during interglacial MIS 5 (mean 30.2 mg cm−2 yr−1) and consistently low during glacial MIS 2 (mean 6.5 mg cm−2 yr−1). MARSPAC in MIS 2 (post-human arrival) was about one-third of that in MIS 6 (pre-human arrival), consistent with lower average fire intensity after human arrival under similar glacial contexts.
• δ13CSPAC spans −27.3 to −13.0‰; relationships with vegetation differ between regimes in ways consistent with intensity-driven preservation biases.
• Statistical tests (filtered to reliable PARchar intervals): Histograms of Spearman’s ρ show low probabilities that observed differences arose by chance for TOC >5.1% and a 30 ka split: P = 0.0071 for MARSPAC vs PARchar (natural regime showing stronger positive relationship) and P < 0.0002 for C3 pollen vs δ13CSPAC (anthropogenic regime showing more pronounced positive relationship).
• Sensitivity analyses: Results are insensitive to the TOC threshold for TOC >2% and to the temporal split for splits older than 10 ka. The combined indicators converge to high confidence in an anthropogenic regime after ~11 ka.
• Qualitative comparisons support these findings: PARchar in interglacial MIS 1 (post-human arrival) exceeds any time in MIS 5; high δ13CSPAC around 15 ka during low C3 pollen and low MARSPAC during MIS 2 further suggest early, progressive human influence preceding 11 ka.
• The inferred natural fire regime had fewer, higher-intensity, larger late dry-season fires (higher SPAC proportion, lower δ13CSPAC), whereas the anthropogenic regime involved frequent, low-intensity, small mosaic burns (lower SPAC proportion in char, higher δ13CSPAC).

The study demonstrates that a natural fire regime dominated Australia’s tropical savannahs for most of the last 150,000 years, characterized by fewer but more intense, larger late dry-season fires likely driven primarily by lightning and climate. After human arrival (~65 ka), there is evidence that Indigenous fire management progressively influenced fire characteristics, with a clear transition to an anthropogenic fire regime by at least 11 ka. The anthropogenic regime—frequent, low-intensity, small, often early dry-season burns—reduced fuel loads and connectivity, thereby limiting high-intensity fires, shaping vegetation structure, and promoting biodiversity. By integrating multiple independent proxies (PARchar, MARSPAC, δ13CSPAC, and pollen) and controlling for sedimentary artifacts (clay-related charcoal comminution), the authors separate human from climatic controls. The relationships between fire proxies and vegetation differ markedly before and after the inferred transition, and statistical resampling shows low probabilities that these differences emerged by chance. Sensitivity analyses confirm robustness to choices of organic carbon thresholds and temporal splits. The results imply that human agency has modulated savannah fire regimes throughout the Holocene. Climate still ultimately governs variability in fire regimes, but human management can substantially alter fire intensity, frequency, and ecological outcomes. The findings align with ethnographic records of Indigenous management across northern Australia and suggest that similar early control over fire regimes by low-density hunter-gatherers might have occurred elsewhere globally.

Using a 150,000-year multiproxy lacustrine record from Girraween Lagoon, the study identifies a transition from a natural, higher-intensity, lower-frequency fire regime to an anthropogenic, lower-intensity, higher-frequency regime by at least 11 ka. This shift is attributed to Indigenous fire management in northern Australia and is supported by convergent evidence from micro-charcoal accumulation, geochemical pyrogenic carbon (SPAC) metrics, isotopic composition, and pollen, with robust statistical tests and sensitivity analyses. The work provides a long-term baseline of natural fire dynamics and demonstrates sustained human modulation of fire across the Holocene, helping explain observed biodiversity patterns and carbon dynamics. It highlights that modern increases in large, high-intensity fires following the disruption of Indigenous practices are reminiscent of the natural regime, with adverse consequences for biodiversity and greenhouse gas emissions. Re-implementing Indigenous fire regimes in tropical savannahs can reverse these trends, with additional carbon sequestration benefits, and similar strategies could mitigate catastrophic fire risks in other regions. The authors suggest that Australia is likely not unique, and earlier-than-recognized control over fire regimes by hunter-gatherers elsewhere should be investigated.

• Micro-charcoal particle accumulation rates (PARchar) are unreliable in clay-rich intervals because char–clay interactions comminute particles below the counting threshold (>10 μm), necessitating TOC filtering and reliance on geochemical measures (MARSPAC) for those periods.
• The exact timing of the onset of anthropogenic influence cannot be pinpointed; the initial 30 ka split is arbitrary for analysis, though sensitivity tests and multiple proxies indicate a transition by at least 11 ka.
• There is considerable inter-sample variation, and low-intensity fires could occur stochastically from lightning even under a natural regime, potentially blurring distinctions at short timescales.
• Charcoal and pollen transport and depositional biases exist (e.g., grass-derived char disperses more widely than wood-derived), which can influence terrestrial vs marine records and local representativeness.
• Climate remains a dominant driver of fire regime; disentangling climate- from human-driven variability requires assumptions tested here via proxy combinations and thresholds but cannot eliminate all confounding.