Evidence that modern fires may be unprecedented during the last 3400 years in permafrost zone of Central Siberia, Russia

Recent climate change in Siberia is increasing the probability of dangerous forest fires. The development of effective measures to mitigate and prevent fires is impossible without an understanding of long-term fire dynamics. This paper presents the first multi-site palaeo-fire reconstruction based on macroscopic charcoal data from peat and lake sediment cores located in different landscapes across the permafrost area of central Siberia. The obtained results show similar temporal patterns of charcoal accumulation rates in the cores under study, and near synchronous changes in fire regimes. The paleo-fire record revealed moderate biomass burning between 3.4 and 2.6 ka BP, followed by the period of lower burning occurring from 2.6 to 1.7 ka BP that coincided with regional climate cooling and moistening. Minimal fire activity was also observed during the Little Ice Age (0.7–0.25 ka BP). Fire frequencies increased during the interval from 1.7 to 0.7 ka BP and appears to be partly synchronous with climate warming during the Medieval Climate Anomaly. Regional reconstructions of long-term fire history show that recent fires are unprecedented during the late Holocene, with modern high biomass burning lying outside millennial and centennial variability of the last 3400 years.


Introduction
Wildfire is one of the most important disturbance agents in forest ecosystems worldwide and has a great impact on forest resources, air pollution and ecological succession (Bond and Keeley 2005). During the last few decades, the extent and severity of fires in arctic and boreal regions is increasing due to climate changes and human activity (Soja et al 2007, Turetsky et al 2011, Wang et al 2021. A number of studies on fire regimes in the high latitude area of Eurasia and North America demonstrated that fires have long affected permafrost thaw and the function of forest and tundra ecosystems (e.g. Schuur  . However, recent global warming in the circumpolar Arctic is increasing fire frequencies (Goetz et al 2007, Krylov et al 2014, Feurdean et al 2020, Fonti and Büntgen 2020. In permafrost area of central Siberia, where larch forests are the dominant forest type, fire drives forest succession and vegetation patterns (Sofronov and Volokitina 2010). According to dendrochronological studies in central Siberia, the fire return interval (FRI) for the period spanning the 17th-20th centuries varies from 77 ± 20 in larch taiga to 320 ± 50 years at the northern forest limit, depending on topography and landscape features of the area (Kharuk et al 2011(Kharuk et al , 2021. Fire scar analysis on larch species in northern central Siberia demonstrated that the FRI tended to decrease during the last century (Kharuk et al 2011, Knorre et al 2019, Kirdyanov et al 2020, Wang et al 2021. Since in 1996 several severe fires caused extensive damage in different regions of Siberia (Kukavskaya et al 2016). Over the same time period exceptionally large fires (mega-fires) covering more than 10 000 ha became a prominent characteristic of the Siberian fire regime (Natole et al 2021).
In light of these recent and future projected changes in fire regime, it is important to improve our understanding of fire dynamics in these sensitive ecosystems. Long-term records of past fire regimes can be used to better understand natural fire variability, the role of fires in forcing land cover change, and subsequently to reduce uncertainties in the potential effects of future fire damage. The most commonly used method for reconstructing Holocene fire frequency and intensity is to study charcoal concentrations in peat and lacustrine sediments (Power et al 2008, Carcaillet et al 2009, Whitlock et al 2010, Carcaillet and Blarquez 2017, Feurdean et al 2020. Often these methods are used in combination with other paleoecological approaches when reconstructing past environmental changes, including plant macrofossil and pollen analysis. Studies in Yakutia in Siberia have shown the ability of charcoal records to indicate fire impact on vegetation and permafrost dynamics (Katamura et al 2009).
In this paper we present new charcoal data from four key areas located in a remote and poorly investigated region of central Siberia. Our objectives are (a) to reconstruct late Holocene biomass burning in four areas of permafrost, (b) to reveal the influence of climate change on fire dynamics and (c) to compare anthropogenically-driven changes in the modern fire regime.

Regional setting
Charcoal records were obtained from three peat and two lacustrine sediment cores across different landscape regions of central Siberia (figure 1 and table 1). Two lacustrine and one peat records are situated in the northwestern part of the Putorana Plateau and are represented by the low elevation peatland site Gervi and the high elevation lake sites Gluchoye and PTHE. One peatland record (Igarka) is located at the western margin of central Siberia near Yenisei River valley, and one peat core (Tura) was obtained from the Central Siberian Plateau in the Lower Tunguska River basin (the right tributary of the Yenisei River).
The climate of the study area, according to the Köppen-Geiger climate classification, is sub-Arctic (or subpolar) with long, usually very cold winters, and short, cool summers (Beck et al 2018). The mean annual air temperature varies from −7.8 • C in Igarka to −8.8 • C in Tura, although this is likely to be lower at the higher altitude sites in the Putorana Plateau. The mean annual precipitation in the western margin of central Siberian Plateau and at foothills of Putorana Plateau is about 650 mm yr −1 and its value decreases considerably to 370 mm yr −1 in the central part of central Siberia (Tura area). The region under study belongs to a zone of continuous permafrost. The vegetation cover is dominated by larch forests (Larix gmelinii). Mountain slopes above 400 m asl are occupied by woodlands (Alnus fruticosa, Betula nana, B. tortuosa, B. cajander) and alpine tundra cover the uppermost part of mountain ridges (Flora of Putorana 1976). Larix sibirica, Picea obovata and Pinus sibirica occur in the study area at Igarka (table 1).

Materials and methods
Coring and sediment description were carried out in July 2019 (Gervi, Tura) and August 2020 (Igarka). Cores from the seasonally thawed layer of permafrost were extracted using a Russian peat corer (inner chamber length of 50 cm and diameter of 5 cm). Sediment cores from Gluchoye Lake and PTHE Lake were collected in July 2006 as part of an earlier study by Self et al (2015). Cores were obtained from the deepest point of each lake using a 7 cm diameter HON-Kajak corer with a 50 cm Perspex coring tube. Samples for the macroscopic charcoal analysis were taken continuously with the 1 cm interval from all cores under study. Probes were prepared following Mooney and Tinner (2011). Charcoal particles with a size >125 µm were counted under a stereomicroscope MOTIC SMZ-171 at 40-fold magnification. Charcoal concentrations or charcoal counts (pieces cm −3 ) were transformed into charcoal accumulation rates or influx (CHAR, pieces cm −2 yr −1 ; figure 2) by multiplying charcoal concentration by sediment accumulation rates (cm yr −1 ) using the CharAnalysis software (Higuera 2009) adapted for the R language environment (R Core Team 2020). The CharAnalysis user independently determines a filtering method for interpolated charcoal accumulation rate (C int ) and a smoothing window, which describes a reconstruction of fire history. Table 2 shows the CharAnalysis parameters used in the present study. Sediment accumulation rates were calculated based on age-depth models constrained by radiocarbon dates (see details in supplementary materials 1, table suppl. 1 available online at stacks.iop.org/ERL/17/025004/mmedia).
The reconstruction of past fire events using Char-Analysis includes the decomposition of the charcoal record into a background signal (C back ), together with a high frequency peak component. The smoothing method and window width were defined by the ensemble-member strategy for separating charcoal peaks from background charcoal data (Blarquez et al 2013). The charcoal peaks (i.e. local fire episodes) were defined as a residual (C peak = C int −C back ). Significant charcoal peaks are defined based on a minimum-count peak screening test. FRIs in each charcoal sequences were determined as the time between two consecutive charcoal peaks (Whitlock et al 2010). The distributions of fire-free intervals were compared for each site using the nonparametric Kruskal-Wallis (for fire series >2) and the Wilcoxon signed-rank tests.
We used the Paleofire R package version 1.2.4 (Blarquez et al 2014) as the standardization technique of the obtained data to compare the local fire reconstructions from the study area. We used three stages of data transformation in the Paleofire package: the min-max rescaling of the CHAR values, a boxcox transformation to homogenize within-charcoal record variance and a final transformation to Z-score (Power et al 2008). Confidence intervals were calculated using a bootstrap resampling of the binned charcoal series and calculating the mean of each 1000 times. Zero Z-score values correspond to the mean charcoal influx over the base period (0.1-2.0 ka BP), and positive/negative Z-score values represent greater-than mean/lower-than-mean charcoal influx over the base period. We have limited our reconstruction to 3.4 ka BP because only two sediment cores under study covered a longer time interval.

Charcoal accumulation
Charcoal accumulation showed similar temporal dynamics, but CHAR values differ by an order of magnitude between sites as C int at Tura ranged from 3 to 37 pieces cm −2 yr −1 while that in PTHE Lake varied between 0.01 and 0.1 pieces cm −2 yr −1 (figure 2(c)). Charcoal influx at Gluchoye Lake amounted 0.02-0.22 pieces cm −2 yr −1 (figure 2(b)) and CHAR values at Igarka fluctuated between 0.1 and 4 pieces cm −2 yr −1 (figure 2(d)). This difference  could be a result of the lower charcoal productivity and flammability of mountain vegetation composed mainly of B. nana, Salix and A. fruticosa (Gluchoye Lake) and high altitude sparse vegetation, including prostrate willows and lichens (PTHE Lake) compared to Larix forests (Higuera et al 2009, Carcaillet andBlarquez 2017) at Tura and Gervi. The macroscopic charcoal analysis revealed the relatively high charcoal accumulation in the period from 3.4 to 2.7 ka BP at Tura and Gluchoye (figures 2(b)-(e)). In Tura C int ranged from 8 to 12 pieces cm −2 yr −1 , reaching the maximum value at 3.1 ka BP, while at Gluchoye C int amounted 0.02-0.05 pieces cm −2 yr −1 with the maximum value of 0.07 pieces cm −2 yr −1 . At PTHE the period of the high charcoal input occurred later (figure 2(c)). Between 3.4 and 3.0 ka BP, CHAR values at this sediment core ranged between 0.01 and 0.02 pieces cm −2 yr −1 and from 3.0 to 2.0 BP C int increased to 0.04 pieces cm −2 yr −1 .
The charcoal accumulation rate declined notably since 2.7 ka BP at Tura and Gluchoye and since 2.0 ka BP at PTHE. Between 2.7 and 0.17 ka BP the CHAR influx at Tura reduced to a range of 3-8 pieces cm −2 yr −1 . No charcoal pieces were recorded in sediment cores from Putorana lakes except for samples formed in time intervals 1.9-1.7, 1.35-1.0 and 0.5-0.25 ka BP, where C int was about 0.25 pieces cm −2 yr −1 at Gluchoye and 0.01 pieces cm −2 yr −1 at PTHE.
Charcoal accumulation at Igarka was low for the entire Late Holocene ( figure 2(d)). Between 4.7 and 0.5 ka BP, the CHAR influx did not exceed the value of 0.1 pieces cm −2 yr −1 with exception of a distinct peak at 3.6-3.5 ka BP. C int formed several maximums with the highest value at 2.9 pieces cm −2 yr −1 . Apparently, this period of the high charcoal input suggested a relatively long fire episode lasting about 100 years that included several forest fires. The next peak of charcoal accumulation detected at 0.85 ka BP when C int reached 0.4-0.5 pieces cm −2 yr −1 .
The macroscopic charcoal analysis of Gervi shows the relatively high charcoal accumulation in the period from 1.3 to 0.9 ka BP (figure 2(a)), when CHAR values ranged from 7 to 15 pieces cm −2 yr −1 . In the subsequent period 0.9-0.15 ka BP, C int decreased significantly and did not exceed 3 pieces cm −2 yr −1 .
The last 250 years was characterized by a sharp increase of the charcoal accumulation rate in all sediment cores under study achieving their highest values throughout the past 3400 years (figures 2(a)-(e)).
The CHAR values at Grvi increased from 3 to 50 pieces cm −2 yr −1 , with a maximum value of 70 pieces cm −2 yr −1 . At Tura CHAR influx raised from 8 to 21 pieces cm −2 yr −1 , with C int reaching maximum values of the entire sediment core at 25 and 38 pieces cm −2 yr −1 . In lakes Gluchoye and PTHE charcoal accumulation gradually increased since about 0.5 ka BP. A sharp rise of the charcoal input occurred during the last 170 years. C int grew from 0.01 to 0.1 pieces cm −2 yr −1 at PTHE and from 0.02 to 0.31 pieces cm −2 yr −1 at Gluchoye. A sharp rise of the charcoal input at Igarka occurred earlier than in other regions. CHAR influx grew explosively from 0-0.1 to 4 pieces cm −2 yr −1 at 0.5 ka BP. During the last 150 years charcoal accumulation in this site gradually declined to minimum values.

Fire return intervals (FRIs)
The FRI over the last 3400 years was generally of a long duration at Igarka and PTHE and reached (mean ± SE) 344 ± 127 and 386 ± 70 years respectively. The FRI at Gluchoye and Tura was shorter and determined as 182 ± 96 and 180 ± 21 years. The mean FRI at Gervi over a period of 1300 years was 189 ± 120 years. The average FRI at the regional scale derived by integrating the data from the five sites during the study period was 344 ± 34 years (figure 3). Table 3 shows the temporal distribution of the FRI between the different sites. Comparison of the FRI distributions in two periods 3.0-1.6 ka BP and 1.6 ka BP-present indicated that all sites experienced very similar fire regimes likely triggered by the same processes during the first and second time intervals (table 2 in suppl.; Wilcoxon test, P > 0.05).
The median FRI, ranging between 360 and 60 years (figure 4). The one-way analysis of variance shows no significant differences of FRIs (p = 0.17) along the elevation gradient from higher to lower  altitudes (tables 3 and 4 in suppl.). The longer FRI at Igarka could be explained by the difference of the local topography and environmental conditions between study sites. The peatland Igarka is located within poor drained and wet lowland where ignition and fire spread weaker compared to dissected plateaus at Tura, Gervi and Putorana lakes. Although, the FRIs are site-specific, temporal variability of the fire frequency (figure 3) showed that trends of FRI changes are homogeneous.

Biomass burning
Biomass burning was moderate in the period between 3.4 and 2.6 ka BP with fluctuation of the Z-score value around zero ( figure 3(b)). During the period 2.6-1.7 ka BP, biomass burning decreased significantly in comparison to the earlier time interval. The composite curve indicates a negative Z-score value, representing lower-than-mean charcoal influx. The next periods were characterized by a positive Z-score value indicating a high fire activity and increased biomass burning between 1.7 and 1.4 ka BP followed by a decline of the Z-score value during the time interval 1.4-1.3 ka BP. The moderate biomass burning with around zero Z-score value occurred between 1.3 and 0.7 ka BP. Since 0.7 ka BP the biomass burning gradually declined, with the minimum Z-score value occurring around 0.5 ka BP. The last 300 years were characterized by notable rise of biomass burning and the highest Z-score value for the entire period.

Discussion
Analysis of macroscopic charcoal in lacustrine and peat sediments from four study areas in central Siberia has allowed us to determine four main phases of fire regime variability (figure 3) that is clearly traceable in both the regional trend in biomass burning and individual charcoal records. Despite differences in FRIs and charcoal accumulation between sites, the trends and the time of changes were rather synchronous over the last 3400 years. Biomass burning and fire activity was moderate between 3.4 and 2.6 BP. (phase 1, figure 3). The charcoal records from Tura and Gluchoye Lake revealed an increase of CHAR index during this time. While the extremely low charcoal values were recorded from Igarka except for a high sharp peak of CHAR around 3.6-3.5 ka BP that suggested a severe fire episode in vicinity of the peatland. The FRI varied from 135 ± 22 years in Tura to 333 ± 188 years in Igarka which is longer than modern times. According to fire scar analysis in larch stems, the current FRI is estimated at 82 ± 7 years in Tura area (Kharuk et al 2011) increasing to 320 ± 50 years at the northern treeline, which lies 50-70 km north to the Igarka peatland (Shahgedanova et al 2002).
Biomass burning and fire activity decreased significantly during the period 2.6-1.7 ka BP (phase 2, figure 3). All charcoal records indicated a decline in charcoal accumulation with exception of the PTHE where the CHAR index was higher from 3.0-2.1 ka BP (figure 2). A decrease in charcoal influx and low biomass burning was recorded in the taiga area of West Siberia at Plotnikovo Mire, located on the eastern margins of the Great Vasyugan Mire system (Feurdean et al 2020).
A decrease in fire activity in Siberia since 2.6 ka BP (phase 2, figure 3) could have been caused by climate cooling and increase in humidity which has been widely recorded at this time, not only at the regional, but also at the global scale. Climatic reconstructions based on chironomid assemblages from lakes Gluchoye and PTHE revealed a noticeable decrease in summer temperature (Self et al 2015), which paralleled evidence of July temperature decline inferred from pollen and diatom data from Lama Lake (Andreev et al 2004, Kumke et al 2004. In addition, pollen data from lakes in northern Siberia (Klemm et al 2016) and tree ring inferred climate reconstructions in Yamal and Taimyr (Hantemirov andShiyatov 2002, Naurzbaev et al 2002) also indicated cooler conditions. Climate cooling in Siberia also coincided with an increase in K + deposition in the GISP2 ice core, indicating the strength of Siberian High, which may suggest changes in atmospheric circulation occurred (Meeker and Mayewski 2002).
The period between 1.7 and 0.7 ka BP is characterized by variable biomass burning and FRI (phase 3, figure 3). The time interval from 1.7-1.4 ka BP was marked by heightened biomass burning but extended FRIs, while the subsequent phases revealed low and moderate charcoal flux together with clear tendency of FRI to decrease. The divergence between charcoal flux and fire frequency could be a result of the probable high occurrence of low intensive fires. According to Kharuk et al (2021) about 90% of modern forest fires in Siberia are surface fires which spread along the forest floor, damaging vegetation litter and understory. The shortest FRIs for most of the areas under study were detected in the interval 1.3-0.7 ka BP (figure 3). The FRI at Gervi and Tura were determined as 72 ± 12 and 126 ± 22 years, respectively, indicating an increase fire frequency compared to the previous phase. Signs of a fire at 0.85 ka BP were identified at Igarka. This fire episode was preceded by a 2500 year long interval when forest fires were not detected. The reduction of FRI about 1.0 ka BP were also recorded at Plotnikovo Mire in western Siberia (Feurdean et al 2020).
The period of increased fire frequency from 1.7-0.7 ka BP included the climate warming of the Medieval Climate Anomaly (MCA, 800-1300 AD (1.15-0.65 ka BP); Mann et al 2009). An increase in summer temperatures at this time has been traced in tree-ring chronologies from northern Siberia (Hellmann et al 2016). Reconstructions of temperature and moisture conditions in Eurasia indicate a significant warming during the MCA at a global scale and a summer drought in several regions (PAGES 2k Consortium 2013). Studies in western Siberia (Feurdean et al 2019) and central Asia (Chen et al 2010) revealed an alternation of relatively wet and dry phases during the MCA.
The next phase of relatively low fire activity that occurred between 0.7 and 0.25 ka BP (phase 4, figure 3) may be synchronous with the climate cooling of the Little Ice Age (LIA, 1300-1850 AD (0.65-0.1 ka BP), PAGES 2k Consortium 2013). A climate deterioration during the LIA was observed in tree ring based paleoclimate reconstructions from eastern Taimyr, Yamal Peninsula and north-eastern Yakutia (Naurzbaev et al 2002;Hellmann et al 2016) and in multi-proxy records from a great number of regions in northern Eurasia (PAGES 2k Consortium 2013). According to the results of dendrochronological studies, the fire frequency in larch forests in the permafrost zone of central Siberia during the LIA was significantly lower than in previous and subsequent epochs (Kharuk et al 2011). During the shorter and possibly colder summers of the LIA, evaporation was reduced, which could lead to an increase in ground moisture in many habitats (Molinari et al 2018), creating an unfavorable environment for the occurrence and spread of forest fires. Our macroscopic charcoal records indicated rare fire events during the LIA across the study areas.
During the last 300 years, charcoal accumulation increased significantly in all sediment cores under study, as biomass burning reached its maximum values for the entire Late Holocene (phase 5, figure 3). Heightened CHAR and peak magnitude values detected in the peat and lake sediments suggest that fires became larger and more severe. Charcoal accumulation during the last two centuries exceeded its average values by an order of magnitude for all sediment cores (figure 2).
Over the last century, increases in temperature and the frequency of dry summer conditions across Siberia apparently enhanced fire activity (Groisman et al 2012). Human impact on fire activity was low in the Putorana Plateau. The areas around Gervi peatland, and the lakes at Gluchoye and PTHE are inhabited, but human influence is limited to rare tourist groups. The settlements of Tura and Igarka appeared in the 1920s. Modern observations in Siberia show that despite fire suppression measures, fires often occur during dry summers around settlements and can spread over large areas (Kukavskaya et al 2016). Therefore, we suggest that both human induced and wildfire may have occurred during the last two centuries in these study areas.
The increase of the CHAR at Igarka occurred earlier than in the other cores. Charcoal accumulation started to rise at the end of 14th-beginning 15th centuries and then charcoal input declined. We supposed that human induced fires led to the biomass burning. According to historical data the Russian colonization of this part of Siberia began in the 17th century (Toschev 2013), but the first Russian settlers already found a local hunter population in the area of Igarka. The settlement lies on the banks of the Yenisei River, historically one of the main trading routes in Siberia. During the 20th century wood-using industry and a river harbor were developed in Igarka, was obviously accompanied by active deforestation that led to some reducing the forest fire occurrence.
Fire regime studies in northern central Siberia based on the analysis of fire scars in larch wood showed that the number of fires increased during the 20th century, and that FRI declined in several regions (Kharuk et al 2021). According to this data, the most severe fires around Tura occurred in 1846, 1896-1899and 1914-1916(Knorre et al 2019. The large and devastating fire in 1899 covered an area of about 250 km 2 (Kirdyanov et al 2020), including the coring site. We hypothesize that this fire corresponds to the sharp peak in CHAR at a depth of 5 cm (∼150 BP) in the sediment core at Tura.

Conclusions
This study provides the first evidence for a late Holocene fire regime in the permafrost area of central Siberia. The data obtained from four study areas located in different landscapes revealed relatively low biomass burning between 3.4 and 1.7 ka BP, with minimum fire activity occurring from 2.6 to 1.7 ka BP and during regional climate cooling and moistening of the LIA (0.7-0.15 ka BP). Fire activity increased during the interval 1.7-0.7 ka BP, which appears to be partly synchronous with climate warming during the MCA. Reconstructions of fire history based on macroscopic charcoal records have shown that recent fires may be unprecedented during the late Holocene, and that the modern fire regime lies outside a 3400 year trend in fire activity. The charcoal accumulation rate during the modern era is in an order of magnitude greater than the average values for each cores. Extraordinary high peak magnitude detected in the last 200 years in peat and lake sediment suggests that fires during this period became larger and likely more severe compared with the long-term past. This trend was observed even at sites remote from human disturbance, and suggests that modern climate change was the main cause, consistent with predictions of increased fire risk across Siberia as a result of anthropogenic greenhouse warming (Groisman et al 2012).

Data availability statement
The data that support the findings of this study are available upon reasonable request from the authors.