Relationships between low-temperature fires, climate and vegetation during the last 430 kyrs in northeastern Siberia reconstructed from monosaccharide anhydrides in Lake El’gygytgyn sediments

Landscapes in high northern latitudes are assumed to be highly sensitive to future global change, but the rates and long-term trajectories of changes are rather uncertain. In the boreal zone, fires are an important factor in climate–vegetation–interactions and biogeochemical cycles. Fire regimes are characterized by small, frequent, 25 low-intensity fires within summergreen boreal forests dominated by larch, whereas evergreen boreal forests dominated by spruce and pine burn large areas less frequently, but at higher intensities. Here, we explore the potential of the monosaccharide anhydrides (MA) levoglucosan, mannosan, and galactosan to serve as proxies of low-intensity biomass burning in glacial-to-interglacial lake sediments from the high northern latitudes. We use sediments from Lake El’gygytgyn (cores PG 1351 and ICDP 5011-1), located in the far north-east of Russia, and 30 study glacial and interglacial samples of the last 430 kyrs (marine isotope stages 5e, 6, 7e, 8, 11c, 12) that had different climate and biome configurations. Combined with pollen and non-pollen palynomorph records from the same samples, we assess past relationships between fire, climate, and vegetation on orbital to centennial time scales. We find that MAs were well-preserved in up to 430 kyrs old sediments with higher influxes from lowintensity biomass burning in interglacials compared to glacials. MA influxes significantly increase when 35 summergreen boreal forest spreads closer to the lake, whereas they decrease when tundra-steppe environments and, especially, Sphagnum peatlands spread. This suggests that low-temperature fires are a typical property of Siberian larch forest on long timescales. The results also suggest that low-intensity fires would be reduced by https://doi.org/10.5194/cp-2019-103 Preprint. Discussion started: 2 September 2019 c © Author(s) 2019. CC BY 4.0 License.


Background consideration
In recent decades, high northern latitudes have been warming at more than twice the rate of other regions on Earth (Serreze and Barry, 2011;IPCC, 2014). This Arctic amplification has widespread impacts on the Earth system, such as the hydrological cycle and carbon budgets (Schuur et al., 2015;Linderholm et al., 2018). The likelihood 50 for wildfire is increasing due to warmer temperatures, more lightening-induced ignition and potential shifts towards more flammable vegetation (Soja et al., 2007;Hu et al., 2010;Tautenhahn et al., 2016;Veraverbeke et al., 2017;Nitze et al., 2018). However, the rates and directions of fire regime and vegetation change are poorly constrained (Abbott et al., 2016).
Climate and vegetation types shape regional fire regimes (characterized by fire frequency, intensity, severity, 55 seasonality, area, type and amount of biomass burned; Harris et al. (2016)). Climate influences biomass availability and flammability via the length and conditions of the growing season, the type of biomass available to burn, and weather conditions that affect local soil conditions and fuel dryness affecting fire spread (Westerling et al., 2006). Accordingly, biomass burning is related to temperature variability on centennial to orbital time scales (Daniau et al., 2012;Marlon et al., 2013).

60
Vegetation drives fires by developing individual plant traits, which determine flammability and regeneration strategies (Rogers et al., 2015;Feurdean et al., 2019) and lead to tight internal fire-vegetation feedbacks (Krawchuk and Moritz, 2011;Pausas et al., 2017). Modern fire regimes differ strongly in the high-northern biomes (Fig. 1), such as between tundra and Siberian summergreen and evergreen boreal forests (Wirth, 2005;Rogers et al., 2015), with less than 1 % of Arctic and Subarctic tundra being affected by fires compared to c. 8 % of eastern Siberian 65 boreal forest (Nitze et al., 2018).
In the tundra, grasses and shrubs burn rarely and at low fire intensities, due to soil and organic matter drying-up during the short summer seasons and limited fuel availability (Krawchuk and Moritz, 2011). Fire intensity refers here to fire temperatures, combustion efficiencies and fire radiative power (Keeley, 2009). In Siberian summergreen boreal forests, dominated by larch (mainly Larix gmelenii and L. cajanderi) and shrub alder (Alnus 70 fruticosa) (Isaev et al., 2010), frequent low-intensity surface fires burn the understorey during dry summers (Kharuk et al., 2011). These fires characterize a "pyrome" of rare, cool and small events (Archibald et al., 2013) with fires that are mainly non-stand-replacing and of incomplete combustion (Rogers et al., 2015;van der Werf et al., 2017;Chen and Loboda, 2018). Fire is suspected to support the existence and regeneration of larch by, for example, selectively reducing regrowth and within-species competition (Kharuk et al., 2011; deciduous foliage that limits fire spreading to the crowns (Wirth, 2005). In contrast, Siberian evergreen boreal forests are dominated by Siberian pine (Pinus sibirica), spruce (Picea obovata), and fir (Abies sibirica), which are fire avoiders that rarely burn and hardly regenerate after fire, and that suppress fires due to a rather wet understorey, whereas Pinus sylvestris stands are highly flammable and can resist infrequent fires (Furyaev et al., 2001;Wirth, 80 2005;Isaev et al., 2010;Rogers et al., 2015;Tautenhahn et al., 2016). Hence, wildfires are rare, but once ignited they burn large areas at high intensities (Archibald et al., 2013) because fires can spread quickly to the crowns via the resin-rich conifer needles (Rogers et al., 2015;van der Werf et al., 2017).

Fig.1
While it is well-recognized that alternative stable states of biome configuration can be driven by fire (Lasslop et 85 al., 2016) and can characterize the boreal forest biomes (Scheffer et al., 2012;Rogers et al., 2015), the causes, feedbacks and thresholds that lead to shifts between stable states are still debated (Tchebakova et al., 2009;Gonzalez et al., 2010;Loranty et al., 2014;Abbott et al., 2016). Recently, Herzschuh et al. (2016) proposed that fire plays an important role in long-term climate-vegetation interactions and internal system feedbacks that determine alternative stable states in high northern biomes. However, knowledge of past fire regimes and 90 associated natural feedbacks in the high latitudes on long, centennial to orbital time scales is scarce. Previous interglacials provide analogues for a warming world (Yin and Berger, 2015) beyond human influence in contrast to the Holocene, when lightning was not the only source of ignition (Buchholz et al., 2003;Marlon et al., 2013;Veraverbeke et al., 2017;Dietze et al., 2018). Sediment cores of Lake El'gygytgyn provide a continuous Pliocene-Pleistocene environmental record in the Arctic 95 that suggest strong climate and vegetation shifts during the past 3.6 Myrs Tarasov et al., 2013;Andreev et al., 2014;Andreev et al., 2016;. Lake El'gygytgyn is a meteorite impact crater lake of 110 km², a diameter of ~12 km, and maximum water depth of 170 m formed about 3.6 Myrs ago  and reference therein). The 293 km² large catchment and wider region of NE Siberia is dominated by volcanic and metamorphic rocks and permanently frozen Quaternary deposits below 100 an active layer of up to 80 cm depth (Schwamborn et al., 2006). Climate conditions are cold, dry and windy, with a mean annual air temperature of c. -10°C and an annual precipitation of c. 180 mm (in 2002), mainly falling as snow (Nolan and Brigham-Grette, 2007). Current treeless herb tundra vegetation is composed of lichens, herbs, and grasses (mostly Poaceae, Cyperaceae) and a few dispersed dwarf shrubs of Salix and Betula near the lake and of Pinus pumila and Alnus in the surrounding Chukchi uplands . The treeline towards the 105 summergreen boreal forest is located c. 100-150 km to the south-west (Fig. 1a). In the period 2000 to 2018, no fires have occurred in the El'gygytgyn catchment and only a few in the estimated pollen source area covering several hundreds of kilometres (based on remote sensing data, after Nitze et al. (2018)).
While Simoneit et al. (1999) and references therein suggest that MAs form at burn temperatures > 300°C, several studies that have analysed the influence of various combustion conditions in natural samples indicated that MAs are thermal dehydration products at burning temperatures < 350°C, mainly under smouldering as opposed to flaming conditions (Pastorova et al., 1993;Gao et al., 2003;Engling et al., 2006;Kuo et al., 2008;Kuo et al., 2011).

120
In Holocene lake sediments, MAs provide complementary fire proxies to sedimentary charcoal (Elias et al., 2001;Schüpbach et al., 2015;Battistel et al., 2017;Schreuder et al., 2019;Dietze et al., in review). While LVG is preserved in marine sediment for at least the last 130 kyrs (Lopes dos Santos et al., 2013), to our knowledge, sedimentary MAs not been analysed in either high-latitude or in interglacial lake sediments.
Here, we study late glacial to interglacial fire histories from the last 430 kyrs using sedimentary MA from Lake 125 El'gygytgyn and assess long-term relationships between low-temperature fires and regional vegetation in the Russian Far East (Fig. 1). We focus on three late glacial-to-interglacial periods, which include marine isotope stages (MIS) 12-11c, 8-7e, and 6-5e. The pollen records from El'gygytgyn sediments have been used to reconstruct biome types and climate conditions over time Tarasov et al., 2013). The so-called "superinterglacial" MIS 11c (c. 420-380 kyrs ago) has been described as warmer and wetter compared to today, 130 which was supported by biomarker based temperature reconstructions (D'Anjou et al., 2013). The presence of spruce (Picea) pollen in El'gygytgyn sediments suggests that the evergreen boreal forest has been much closer to the lake than today Tarasov et al., 2013;Lozhkin et al., 2017). In contrast, MIS 7e (c. 240-220 kyrs ago) was cooler than today and only a few coniferous pollen grains were found. Birch and alder pollen suggest that shrub tundra prevailed during this interglacial . MIS 5e interglacial (c.

135
130-110 kyrs ago) was slightly warmer than today (Tarasov et al., 2013). Larch and alder pollen were found in El'gygytgyn sediments suggesting that a summergreen boreal forest existed close to the lake, whereas spruce pollen was absent . These differences in regional vegetation and climate conditions have been explained by global ice volume, insolation changes (see Yin and Berger (2012), Table 1, for interglacial astronomical and GHG characteristics), and interhemispheric ice sheet-ocean-atmosphere feedback mechanisms 140 Lozhkin et al., 2017).
Here, we aim to answer 1) whether sedimentary MAs are suitable proxies to reconstruct low-temperature fires even in interglacial Arctic lake sediments; 2) whether more biomass burning occurs during interglacials compared to glacials due to climate-driven changes in biomass availability, since wildfires are expected to be fuel-limited in cold environments (Krawchuk and Moritz, 2011;Daniau et al., 2012), and 3) as different vegetation composition 145 has been reconstructed from past interglacials Andreev et al., 2014), we consider whether more low-temperature biomass burning occurs in summergreen boreal forest compared to tundra or evergreen boreal forest during interglacials on centennial to millennial timescales.

150
We sampled sediment from two cores (Fig. 1b). A 12.7 m long sediment core -PG1351 -was recovered in 1998 and covers the last 270 kyrs, according to 14 C-and luminescence dates, as well as magnetostratigraphy (Nowaczyk   . A 318 m long composite core from ICDP site 5011-1 comprises three parallel sediment cores that cover the time period , dated by magnetostratigraphy, paleoclimatic and orbital tuning with more than 600 age tie points .

155
For MA analyses, we freeze-dried and homogenized 44 samples of c. 0.7-1.8 g dry sediment from core PG1351 covering MIS 8 to MIS 5e, integrating sediment of 1 cm core depth. Temporal resolution of these samples ranges from 140 to 960 years per sample. For the period between 430 and 405 kyrs ago (end of MIS 12 to MIS 11c), 13 samples of 0.5-1.3 g of dry sediment from ICDP core 5011-1 were taken for MA analyses, integrating sediment of 2 cm core depth. Eight of these 13 samples are from the same core depths as were previously analysed for pollen 160 . Temporal resolution of these samples varies between 200 and 970 years per sample comparable to core PG1351. Across all samples, temporal resolution is 333 ± 273 years per sample, giving centennial-scale averages.
We extracted the polar lipids of all MA samples using a Dionex Accelerated Solvent Extraction system (ASE 350, ThermoFisher Scientific) at 100°C, 103 bar pressure and two extraction cycles (20 min static time) with 100 % 165 methanol, after an ASE cycle with 100 % dichloromethane. For every sample sequence (n=13-18), we extracted a blank ASE cell and included it in all further steps. We added 60 ng of deuterated levoglucosan (C6H3D7O5; dLVG; Th. Geyer GmbH & Co. KG) as internal standard, and filtered the extract over a PTFE filter using acetonitrile and 5 % HPLC-grade water. We analysed the extracts with an Ultimate 3000 RS ultra-high performance liquid chromatograph (U-HPLC) with thermostated autosampler and column oven coupled to a Q

170
Exactive Plus Orbitrap mass spectrometer (Quadrupole-Orbitrap MS; ThermoFisher Scientific) with heated electrospray injection (HESI) probe at GFZ Potsdam, using measurement conditions adapted from earlier studies (Hopmans et al., 2013;Schreuder et al., 2018;Dietze et al., in review). Briefly, separation was achieved on two Xbridge BEH amide columns in series (2.1 x 150 mm, 3.5 µm particle size) fitted with a 50 mm pre-column of the same material (Waters). The compounds were eluted (flow rate 0.2 mL min -1 ) with 100 % A for 15 minutes, 175 followed by column cleaning with 100 % B for 15 min, and re-equilibration to starting conditions for 25 min.
Eluent A was acetonitrile:water:triethylamine (92.5:7.5:0.01) and eluent B acetonitrile:water:triethylamine (70:30:0.01). HESI settings were as follows: sheath gas (N2) pressure 20 (arbitrary units), auxiliary gas (N2) pressure 3 (arbitrary units), auxiliary gas (N2) temperature of 50 ˚C, spray voltage -2.9 kV (negative ion mode), capillary temperature 300 °C, S-Lens 50 V. Detection was achieved by monitoring m/z 150-200 with a resolution 180 of 280,000 ppm. Targeted data dependent MS 2 (normalized collision energy 13 V) was performed on any signal within 10 ppm of m/z 161.0445 (calculated exact mass of deprotonated levoglucosan and its isomers) or m/z 168.0884 (calculated exact mass of deprotonated dLVG) with an isolation window of 0.4 m/z. The detection limit was 2.5 pg on column, based on injections of 0.5 to 5000 pg on column of authentic standards of LVG, MAN, and GAL (Santa Cruz Biotechnology) and dLVG.

185
Integrations were performed on mass chromatograms within 3 ppm mass accuracy and corrected for relative response factors to dLVG (1.08 ± 0.10, 0.76 ± 0.10 and 0.24 ± 0.05 for LVG, MAN, and GAL, respectively), according to known authentic standard mixes injected before and after every measurement sequence and supported by characteristic isomer-specific MS² data. All samples were corrected by subtracting the maximum MA concentrations in the blank duplicates of each ASE sequence. To account for biases due to sediment properties and 190 sedimentation rates, MA influxes (mass accumulation rates in ng cm -2 yr -1 ) were calculated by multiplying the https://doi.org/10.5194/cp-2019-103 Preprint. Discussion started: 2 September 2019 c Author(s) 2019. CC BY 4.0 License. concentrations (ng g -1 ) with the sample-specific dry bulk densities , and the sample's sedimentation rates (cm yr -1 ) using the age-depth models presented by Nowaczyk et al. (2007) for the PG1351 core and  for the ICDP-5011-1 cores.
Pollen from sediment core PG1351 was analysed by Lozhkin et al. (2007), but there was no sediment left to sample 195 the same core depths for MA analyses. Therefore, we sampled 19 of the 44 MA sample's core depths for parallel pollen analyses to enable a direct comparison of MA and pollen records without age bias. These new pollen samples were prepared using standard pollen preparation procedures as have been used previously for the El'gygytgyn sediments . In addition to pollen and spores, non-pollen-palynomorphs (NPPs) such as algae remains and coprophilous fungi spores were counted. Existing pollen data  from 200 the same depths as MA samples of MIS 12 and 11c were harmonized with the new pollen samples of core PG1351 to compare the same taxa in percentages.

Analyses of source areas
MAs are transported attached to aerosols in the atmosphere (Sang et al., 2016;Schreuder et al., 2018). To discuss potential source areas of MAs, we calculated exemplary backward trajectory ensembles of two days towards Lake

210
MA influxes and ratios were correlated with the six most indicative pollen and NPP types excluding samples with < 0.5 % of a certain taxon to stabilize the signal-to-noise ratio (Prentice et al., 1996). Evergreen boreal forest is represented by Picea pollen and summergreen boreal forest (SGB) by the sum of Larix, Alnus, and Populus pollen percentages. The sum of Poaceae, Artemisia, Chenopodiaceae, Caryophyllaceae, Cichoriaceae, and Thalictrum pollen represent typical taxa of cold tundra-steppe environments. Pinus s.g. Haploxylon-type pollen can be 215 produced by a shrub stone pine (Pinus pumila), which does not survive fires, or by P. sibirica, a Siberian pine that can survive but is not adapted to fires (Ma et al., 2008;Keeley, 2012). Sphagnum spores reflect wet habitats (peatlands), whereas Selaginella rupestris spores from a ledge or rock spike-moss are indicative of extremely dry and cold habitats.
To quantify the relationships between low-temperature fires and vegetation, we log-transformed the MA ratios 220 and pollen percentages to correct for their skewed distributions. Then, we correlated MA records with the major land cover types using the Kendall's τ rank correlation coefficient that is robust against outliers, small sample sizes, and against spurious correlations, in contrast to Pearson's correlation coefficient (Aitchison, 1986;Jackson and Somers, 1991;Arndt et al., 1999). Kendall's τ and associated p-values were calculated using the function corr.test of the R package psych (Revelle, 2018) using pairwise complete observations. Kendall's τ is only provided

Results
MAs are detected well in all samples with LVG, MAN, and GAL concentrations of 77 ± 35, 109 ± 58, and 204 ± 129 ng g -1 (mean ± standard deviation, Fig. 2a), respectively. The standard instrumental errors from duplicate 230 measurements are 4.9 ± 2.9, 4.8 ± 4.1, and 7.1 ± 5.2 % for LVG, MAN, and GAL, respectively. Blanks contained 7.6 ± 4.7, 2.1 ± 0.7, and 2.1 ± 0.6 % of the respective mean LVG, MAN, and GAL concentrations in the samples, derived from carryover within the ASE preparation step, and are subtracted from the concentrations of the respective sample batch. focus here on relative changes in MA influxes, which are consistently higher during interglacials compared to the latter part of their preceding glacials, according to the boxplots (Fig. 2b). Among interglacials, influxes are highest during MIS 5e (e.g., LVGmedian: 2.2 ng cm -2 yr -1 , GALmedian: 4.2 ng cm -2 yr -1 ) and lowest during MIS 7e (LVGmedian:  Despite few parallel samples with n > 4 and pollen and spore amounts higher than 0.5 %, some linkages between
As both MA ratios show similar temporal trends, we used only LVG (MAN+GAL) -1 in the correlation analysis. during MIS 12-11c, but also when considering all samples (Fig. 4b).

Preservation of monosaccharide anhydrides in high-latitude lake sediments
The abundance of LVG and its isomers in smoke in absolute (influxes) and relative (ratios) terms depends on the amount and type of biomass being burnt and on burning conditions (Engling et al., 2006;Kuo et al., 2008;Fabbri 280 et al., 2009;Kuo et al., 2011). Yet it is unknown, whether sedimentary MAs are suitable biomass burning proxies in Arctic lake sediments on centennial to orbital timescales. Issues such as analytical uncertainties, the source of MAs (i.e. geographic source area, type of biomass burnt, burning conditions) and degradation during transport, deposition and after deposition need to be addressed in order use the proxies further back in time. We discuss the future potential of MA records in long-term high latitude fire reconstructions and present some first ideas on past despite absolute age uncertainties that are several hundreds of years .

300
We find a trend towards lower average MA influxes in older sediments (Fig. 2b), which is not as prominent when comparing the trajectories of MA influxes in time (Fig. 3a). This could indicate either a true signal of past fire activity or a certain degree of post-depositional degradation, or a combination of both aspects. As there is no temporal trend visible in MA ratios (Fig. 2c), the strong positive relationship between MA isomers (Fig. 4a) suggests that either they derive from the same source and/or have been degraded in a similar way during transport 305 and after deposition.

Potential source areas and degradation pathways from source to sink
We expect higher influxes when fires happen close to the lake because under atmospheric conditions, several degradation pathways limit chemical stability of MAs during aeolian transport to a few hours to days ( Hence, the potential source area could have been even larger during past interglacials, but still be located in the vegetated realm of eastern Siberia. We assume that shifts in geographic source areas associated with shifts in atmospheric circulation would affect the variability of MA influxes within rather than between interglacials. Monitoring and sediment properties of Lake El'gygytgyn suggest mixed and well-oxygenated bottom waters during summers and past warm periods, whereas during glacial periods long-term lake stratification led to rather anoxic bottom water conditions Nolan and Brigham-Grette, 2007). If we assume a constant influx of MAs and high degradation at the lake bottom, we would expect higher preservation during glacials than 330 interglacials -but we find higher MA influxes in interglacial sediments (Fig. 2b), similar to total organic carbon percentages . Hence, we assume that MA degradation was limited even over longer time, when occluded within or adsorbed to a mineral matrix or iron oxides ( El'gygytgyn sediment-based MA influxes are only marginally affected by degradation on centennial to orbital timescales, and, instead, represent relative changes in biomass burnt from a similar source during low-intensity fires.

Long-term relationships between low-intensity fires, climate, and vegetation
4.2.1 Orbital-scale fire occurrence and the role of biomass and fuel availability, insolation and climate

340
More low-intensity biomass burning occurs during interglacials compared to glacials, as we find higher MA influxes during interglacials compared to the preceding late glacials. Considering source areas and transport pathways (Sect. 4.1.2), we suggest that during interglacials more biomass has burnt closer to Lake El'gygytgyn ( Fig. 2b). Absolute biomass availability and land cover change cannot be assessed quantitatively from pollen percentage data due to (partly) unknown taxa-specific pollen and NPP dispersal, productivity and preservation 345 (Sugita, 2007) and our low sample resolution prevents the use of new quantitative land cover reconstructions (Theuerkauf and Couwenberg, 2018). Yet, more biomass availability is suggested by significantly higher abundance of tree and shrub pollen in interglacial sediments compared to glacial cold tundra-steppe assemblages dominated by herbs and grasses (Fig. 5) Tarasov et al., 2013). Accordingly, our rank correlation analysis shows that MA influxes tend to increase with higher amounts of tree and shrub versus tundra-350 steppe pollen (Figs. 3c, 4b). Reasons could be that during glacials, low temperatures and CO2 levels limit biomass (fuel availability) and fire spread, which has be shown in previous mid-to high-latitude reconstructions and model simulations (Thonicke et al., 2005;Krawchuk and Moritz, 2011;Daniau et al., 2012;Martin Calvo et al., 2014;Kappenberg et al., 2019). Thus, El'gygytgyn's glacial MA influxes represent a background signal from remote source areas, when fires associated with high-productivity biomes have shifted southwards.

355
We find also evidence for more biomass burning close to the lake in times of high summer insolation, in agreement with previous mid-to high latitude studies (Daniau et al., 2012;Remy et al., 2017;Dietze et al., 2018;Kappenberg et al., 2019). This is visible during the period of MIS 8-5e, when MA influxes not only peaked during MIS 7e and 5e, but also during maximum summer insolation of the preceding MIS 8 and 6 late glacial stages, despite pollen data suggesting rather low tree and shrub presence during these intervals (Fig. 3a, c). In addition, lower mean MA 360 influxes during MIS 11c compared to MIS 5e (Fig. 2b) might be linked to the rather moderate summer insolation during MIS 11 compared to MIS 5e (Yin and Berger, 2012). Overall, increased high latitude summer insolation seems to drive more biomass productivity and fires, for example, by the length of the growing season, but also by affecting land-atmosphere feedbacks and potential interhemispheric ice-ocean-land feedbacks that alter regional precipitation-evaporation patterns (Yin and Berger, 2012;Martin Calvo et al., 2014).

365
While it is difficult to fully disentangle biomass availability and climate conditions, we suggest that climate was the main driver of boreal forest fires on orbital timescales. This confirms previous studies from Holocene reconstructions, where, depending on regional biome configurations, wildfires have increased with increasing temperature and reduced precipitation patterns (Anderson et al., 2006;Remy et al., 2017;Molinari et al., 2018). The differences in MA influxes among interglacials seem to reflect centennial to millennial-scale temperature and moisture changes, as we find the lowest interglacial MA influxes during MIS 7e, known to be the coolest of the three interglacials considered here, when pollen of a birch and willow shrub tundra prevailed with little abundance of summergreen boreal and pine pollen   (Figs. 2b, 3a, 5). Yet, insolation and temperature control alone cannot explain the lower MA influxes during the "super-interglacial" MIS 11c 375 compared to MIS 5e.

Centennial
Regional moisture availability and biome configuration, hence fuel composition, seem to affect low-temperature fire occurrence on centennial-to millennial scales for two reasons. First, we find a negative relationship of MA influxes with Sphagnum spores across all samples and especially during MIS 12-11c (Fig. 4b), when Sphagnum spore abundance was highest (Fig. 5), indicating widespread peatlands. The presence of spruce pollen also 380 confirmed that MIS 11c was wetter than MIS 5e , suggesting that fuel wetness limits fires on centennial time scales.
Second, most biomass was burnt (i.e. highest MA influxes) during the warm and dry MIS 5e, which was dominated by summergreen boreal (larch) and pine forest . Our rank correlation analysis suggests a significant positive correlation of MA influxes with summergreen boreal tree pollen during 385 MIS 8-5e (Fig. 4b). In addition, during the late MIS 12 and early MIS 11c, MA influxes peaked at the time of high summergreen boreal pollen, before spruce pollen reached their maximum and after the maximum in Sphagnum spore abundance (Figs. 3a, c). As there was also no relationship between spruce pollen and MA influxes ( Fig. 4b), we suggest that evergreen spruce forest was not an important source of MAs on long timescales, in contrast to summergreen boreal forest.

390
There is also a tendency towards more MA influxes with increasing pine pollen abundance during MIS 8-5e (Fig   3 a, c), whereas during MIS 12-11c Pinus s.g. Haploxylon-type pollen was not or weakly negatively related to MA influxes (τ not significant, Fig. 4b), maybe because different pine species with different fire-related traits could have produced the P. s.g. Haploxylon-type pollen during the two periods. The shrubby P. pumila that dominates in modern treeline ecotones does not survive frequent low-intensity surface fires, whereas P. sibirica,

395
occurring in evergreen and high-elevation boreal forest, is not adapted to is but able to survive fires (Ma et al., 2008;Keeley, 2012). Pinus s.g. Diploxylon-type pollen, derived from P. sylvestris, are almost not present in Lake El'gygytgyn sediments. However, as independent high-resolution climate proxy data is lacking for our samples, we cannot fully disentangle the role of vegetation composition and centennial-to millennial-scale climate conditions. The relationships between MA influxes, summergreen-and evergreen boreal taxa and Sphagnum are independently confirmed by MA ratios that depend on the type of biomass burnt and on burning conditions such as duration (Engling et al., 2006;Kuo et al., 2008;Fabbri et al., 2009;Kuo et al., 2011). Before relating the MA influx-based with the MA ratio-based evidence, we have to state that, surprisingly, El'gygytgyn MAN and GAL

405
influxes were up to ten times higher than LVG influxes during all periods, and LVG MAN -1 and LVG (MAN+GAL) -1 ratios were much lower than 3 and 1, respectively (Fig. 2c). This is in contrast to previous observations and experimental burnings (Fig. 2d) that report MA ratios of about an order of magnitude higher than https://doi.org/10.5194/cp-2019-103 Preprint. Discussion started: 2 September 2019 c Author(s) 2019. CC BY 4.0 License.
Only Oros and Simoneit (2001b) and Otto et al. (2006) report similarly low MA ratios from burning of a mixed bark, needle, cone and wood sample of a temperate pine (Pinus monticola) and from a charred pine cone (P. banksiana), respectively. Yet, these softwood tree species are not present in Siberia and there was no relation between MA ratios and pine pollen during MIS 8-5e, and a tendency towards higher ratios LVG (MAN+GAL) -1 420 when pine increased during MIS 12-11c -which cannot explain the low El'gygytgyn MA ratios. There was also no relationship between MA ratios and spruce pollen in our samples, supporting the suggestion from MA influxbased evidence that evergreen conifers do not significantly influence El'gygytgyn MA records.
In the studied MIS 12-11c samples, Sphagnum is significantly positively related to LVG (MAN+GAL) -1 , with highest ratios in times of highest Sphagnum abundance, but also in times of very low, background MA influxes 425 (Fig. 3b, c; 4b). LVG MAN -1 ratios in modern-day aerosols of peatland fires in Russia were found to be > 7 (Fujii et al., 2014), that is, higher than softwood-derived MA ratios, but lower than those from hardwoods and grasses ( Fig. 2d) -which cannot explain the low El'gygytgyn MA ratios.
Instead, we find a significant negative correlation between El'gygytgyn MA ratios and the summergreen boreal pollen sum during MIS 12-11c, confirming reports that burning of larch wood produce relatively higher levels of 430 GAL (= lower LVG (MAN+GAL) -1 ratios) than other softwoods (Schmidl et al., 2008). The reported larch LVG MAN -1 ratios, however, are still about 1.8 to 5 times higher than ours. In addition, we find low ratios across all samples, mainly independent of high or low influxes. Hence, we hypothesise that an additional biomass source and/or specific burning conditions have contributed to the low sedimentary MA ratios found in Lake El'gygytgyn sediments.

435
One additional biomass source could be dense moss-lichen mats within the summergreen boreal forest. There, light can penetrate more easily through its open canopy compared to evergreen boreal forest, enabling the understorey to dry up quickly during summer droughts. The more open the canopy is, the denser the moss-lichen mats in the understorey become, providing several percent of the total biomass and carbon stored in NE Siberian larch forests and helping to insulate the permafrost (Isaev et al., 2010;Loranty et al., 2018). Together with larch overall MA yields at temperatures higher than 350°C (Engling et al., 2006;Kuo et al., 2008;Kuo et al., 450 2011;Knicker et al., 2013). Although average burn temperatures are unknown from low-intensity surface fires in larch forests, the suggested high speed of surface fires under open canopies (Sofronov et al., 2004) would decrease burn durations, with absolute MA yields increasing during shorter burn durations (Engling et al., 2006;Kuo et al., 2011). After burning of the moss-lichen layer, rejuvenation of larch and tree growth is promoted by active-layer thickening and nutrient release for several decades post fire, until moss-lichen mats recover (Loranty et al., 2018).

455
Overall, all samples and sedimentary MA influxes and ratios reported here are integrated over centennial to millennial timescales. Together with the MA influx-based evidence (i.e. no significant differences and very low MA ratios across glacial-interglacial periods, Fig. 2c), we propose that during all times there was a high contribution of MAs from low-temperature surface fires in summergreen larch forest presumably including the burning of moss-lichen mats of currently unknown MA emissions ratios. In periods of low MA influxes 460 (background influxes), higher MA ratios probably included more burning residues from remote grass, peatland and forest fires, whereas higher influxes suggest that summergreen boreal fires happened closer to Lake El'gygytgyn -with the northward spread of larch forest being well-documented in MIS 11c and 5e pollen records Lozhkin et al., 2017).
Considering a future warming and wetting of the high northern latitudes (Hoegh-Guldberg et al., 2018), we would 465 expect an increase in the availability of flammable biomass on long timescales, but low-intensity fire might decrease when fuel moisture exceeds a certain threshold -either by transition from a stable forest state to peatlands or by shifts from summergreen to evergreen boreal forest -despite potentially increasing fire ignitions (Veraverbeke et al., 2017).

470
Molecular proxies are increasingly being used in palaeoenvironmental studies, providing insights into past biogeochemical cycles during periods that provide natural analogues of the expected future regional change. Here, we have shown the potential of MA influxes and ratios in high northern lake sediments as proxies for the amount and type of low-intensity biomass burning. Although limited in samples, we can deduce first fire-climatevegetation relationships in north-eastern Siberia on long timescales.

475
• MAs can be measured well above the detection limit using UHPLC-HRMS on sediment samples of Lake El'gygytgyn for three previous glacial-to-interglacial periods and seem little affected by degradation. A declining trend in MA influxes with time is thought to represent the past amount of biomass burnt during lowtemperature fires that can be related to climate, regional biomass availability and biomass composition on orbital and centennial-to millennial-timescales. Siberian larch forests (Kharuk et al., 2011;Rogers et al., 2015;Chen and Loboda, 2018): a relationship that seems to hold on centennial to millennial timescales during past interglacials.
• Relatively higher MA influxes during interglacials and times of high summer insolation suggest that lowtemperature fires are closely linked to biomass availability and climate conditions that favour fuel dryness on 485 orbital timescales. Differences between interglacials are revealed by higher MA influxes when summergreen boreal forest has spread closer towards Lake El'gygytgyn, although there is no clear relationship to evergreen coniferous taxa.
• Acknowledgements. We dedicate this study to Sophie Stehle †, who assisted in the lab and left us too early. We 505 thank also Denise Dorhout for lab assistance, Norbert Nowaczyk, Stefanie Poetz, Niels Hovius, Georg Schwamborn and Stefan Kruse for discussions and Cathy Jenks for English editing.