Chronologies were established using radiocarbon dating via Accelerator Mass Spectrometry (AMS), applied to a combination of plant macrofossils and bulk peat samples, supplemented by 210Pb dating for recent sediments (Table ). Detailed chronology for Toolik I and II (TFSI and II) are published in Gałka et al. (2018), for GaI and GaII in Gałka et al. (2023), and for NBb and Erin (ER) are presented in Gałka et al. (2025). Age-depths models for all sites are presented in Appendix , along with details on chronology for the three unpublished sites, SG, RH, and DL.

We reconstructed changes in local-scale fire activity based on macroscopic charcoal analysis of 1–2 cm3 sample volumes taken at 1 cm contiguous intervals across cores from all sites. Sample preparation involved overnight bleaching using sodium hypochlorite (domestic bleach) and wet sieving through a 150 µm mesh. Charcoal identification and categorization of morphotypes (the latter conducted for 6 of the 8 records) followed the protocols of Feurdean et al. (2017) and Feurdean (2021). Charcoal morphologies were then grouped into non-woody (graminoids and forbs) and woody to determine the dominant fuel type. We also separate the charcoal fractions (5 out of 8 records) into two size classes: 150–500 µm and > 500 µm, to distinguish more local from more regional fires (Adolf et al., 2018). Additionally, we searched for large macrocharcoal fragments in samples used for plant macrofossil analysis. For this, 12 cm3 contiguous sediment samples were washed and sieved using a 200 µm mesh without applying bleaching, following Gałka et al. (2018).

Charcoal accumulation rate (CHAR particlescm-2yr-1) was calculated by dividing the particle concentration (particlescm-3) by the deposition time derived from the age-depth models (yrcm-1). To account for changes in sedimentation rates, which increased markedly in the upper sections of the records, CHAR values were interpolated to a median temporal resolution for each record.

We extracted shrub abundance from existing plant macrofossil and pollen records from TFSI, TFSII (Gałka et al., 2018), GaI, GaII (Gałka et al., 2023), NBb, and Erin (Gałka et al., 2025) to determine the local and regional changes in woody vegetation (Table ). Total shrub pollen percentages represent the combined abundance of Betula nana, Alnus, Salix, Ericaceae, Vaccinium, Rubus chamaemorus, and Dryas octopetala. For plant macrofossils, above-ground remains (seeds, leaves, and fruits) of Andromeda polifolia, Betula nana, Ledum groenlandicum, Vaccinium, and Salix spp. were quantified as counts. In contrast, the abundance of unidentified shrub roots was expressed as a volume percentage relative to the total vascular plant and moss remains.

Changes in peatland hydrology were assessed using existing testate amoeba depth-to-water level (DTW) reconstructions from the TFSI, TFSII (Taylor et al., 2019), GaI, GaII (Gałka et al., 2023), and ER (Gałka et al., 2025). All cores were collected from similar peatland types, specifically small hummock microforms. Quantitative reconstructions of water table depth were derived using the transfer function developed by Taylor et al. (2019).

To assess fire history across the study area, we developed a composite record of charcoal accumulation rate by combining all individual interpolated charcoal records (n=9). We generated a composite CHAR record following the protocol of Blarquez et al. (2014), which involves MinMax re-scaling, a Box–Cox transformation to homogenize variance across records, and Z-score standardization using a baseline period of last 2000 years. The transformed time series was smoothed using a LOWESS smoother with a 200 year window. Confidence intervals were estimated from the distribution of 1000 bootstrapped replicates.

To determine local and regional changes in woody vegetation, we created composite records of shrub abundance derived from individual pollen and plant macrofossil datasets (n=5). For plant macrofossils, two separate composite records were constructed, one for above-ground remains (seeds, leaves, and fruits) and one for root remains. Composite pollen and macrofossil time series were generated by applying a LOWESS smoothing with a 200 year window.

Hydrological changes were reconstructed by creating a composite DTW record (n=5). This was developed by combining existing individual records of DTW, using a Z-score transforming the DTW data (Swindles et al., 2019), and applying a LOWESS smoothing with a 200 year window.

To determine recent trends in fire activity, evaluate our understanding of charcoal source areas, and the amount of charcoal produced in tundra fire, we downloaded two datasets available from the Alaska Interagency Coordination Center (AICC) website: Alaska Fire History Location Points (fire points) and Alaska Fire History Perimeter Polygons (fire polygons; Miller et al., 2023). In case of fire polygons, the area burned was based on the size of the polygon; non-vegetated surfaces were not removed. In case of fire point data, the reported incident size was estimated from field observations or calculated using GIS from perimeter polygons; therefore, the recovered size of fire from fire point data should be treated with caution. Spatial analysis was conducted using ArcGIS v.3.4.3 with the coordinate system set to the Alaska Albers Equal Area Conic (EPSG: 3338) projection. We identified those fire points (n=35) and fire polygons (n=9) that fell within 2, 25, and 70 km from the Dalton Highway, north of the Brooks Range (Fig. ). We calculated the distance from each fire point and fire polygon from the Dalton Highway as well as the distance from the nearest five fire points to each of the fossil record. For the fire point data, an estimate of the area burned is provided in hectares (Table S1 in the Supplement). For the polygon data, the size of the fire is reported as area in hectares, determined by the ArcGIS' Calculate Geometry Attributes tool (Table S2 in the Supplement).

Of the nine macrocharcoal records analysed, seven span the last one to two millennia, while only two extend as far back as 1000 BCE albeit at low temporal resolution (Fig. ). CHAR values are consistently low across all sites, ranging from 0 to 1.07 particlescm-2yr-1 (mean = 0.04 ± 0.11 cm-2yr-1). The highest mean CHAR values were observed at SG (0.05 particlescm-2yr-1) and NBb (0.09 particlescm-2yr-1). Interpolated individual and composite CHAR records indicate consistently low biomass burning (negative Z-scores) between ∼ 1000 BCE and 1000 CE (Figs.  and ). A temporary increase in composite CHAR is observed after ∼ 1000 CE, peaking around 1200 CE (positive Z-scores), followed by a decline to negative Z-scores; however, values remained higher than those recorded prior to 1000 CE. Elevated CHAR values around 1200 CE are particularly visible at TFSII, NBb, ER and RH. A continuous rise in composite CHAR begins around 1900 CE, reaching its maximum between 1950 and 2015 CE (positive Z-scores), a trend consistently reflected across all individual records.

The charcoal morphological record indicates an almost equal representation of herbaceous (e.g., grass leaves and stems of grasses and forbs) and woody charcoal morphologies, with higher woody influx towards recent times (Fig. ). Small-size charcoal particles (150–500 µm) dominate over the larger fraction (> 500 µm) (Appendix ). No large macrocharcoal fragments (> 1 mm) were detected in plant macrofossil records.

Our satellite imagery record of fire points within a 70 km buffer along the Dalton Highway shows that 36 fires occurred between 1969 and 2023 (Table S1). Except for an extremely large fire of 103 896 ha occurring in 2007, most recorded fires were of small size (mean = 46 ha; median = 2 ha). Fire polygon data indicate nine fires with a mean size of 11825 ha (median = 115 ha; Figs a and a; Table S2). There were 10 recorded fires (0.5 peryear) between 1969 and 1990, followed by a fire-free period from 1990 to 2001 (Fig. a and b). Fire activity increased markedly between 2001 and 2017, with 23 recorded fires (1.5 peryear), before declining again, with only one fire observed in 2023.

On a temporal scale, we found that the nearest fires to the sampling points (range 3–29 km; mean 12 km) occurred in years 1971, 1979, 1990, and 2007, and the nearest sampling points to these fires were TFSII (3 km), DL (6.5 km), and RH (7 km) (Figs a, b and b; Table S3 in the Supplement). The second closest fires from the charcoal sampling (range 15–31 km; mean 20 km) occurred in years 1969, 1970, 1971, 1979, 1983, 1990, 2004, 2007, 2017. Given the inherent chronological uncertainties in sedimentary records and the very low amount of charcoal in our data, the slight increases in interpolated CHAR centred around the years 1965–1975, 1985–1990, and 2010–2015, may correspond to documented fire events occurring during seasons 1969–1983, 1990 and 2004–2017 (Figs a, b and a; Table S3). However, the CHAR record suggests greater burning activity around the 1970s. In contrast, satellite data indicate maximum fire size and frequency between 2001 and 2017 CE (Fig. a).

The pollen-based composite vegetation record, representing shrub abundance at local to regional scales, indicates consistently low shrub abundance (∼ 12 %) between 1000 and 500 BCE (Fig. b). A slight increase is observed from 250 CE (∼ 15 %) and again between 500 and 1000 CE (∼ 20 %). A marked rise in shrub abundance (∼ 40 %) occurs from approximately 1400 CE to the present, interrupted by brief declines between 1800 and 1945 CE and in the most recent samples.

The composite plant macrofossil record, reflecting local-scale shrub abundance, shows very low percentages of above-ground remains (∼ 2 %) from 1000 BCE to 1950 CE (Fig. b). A sharp increase (∼ 15 %) occurs after 1950 CE, continuing to the present. In contrast, the below-ground macrofossil record (roots) shows an increase beginning around 1900 CE, and after 1950 CE. A slightly elevated root abundance is also evident between ∼ 750 and 250 BCE (Fig. b).

The composite water table record indicates predominantly low (deep) water levels (positive Z-scores) between ∼ 250 and 500 CE and again between ∼ 1250 and 1600 CE, with persistently low levels over the past ∼ 50 years (Fig. c). The pronounced deepening around 250 CE is primarily driven by lower water levels at two of the three available sites: TFSII (Taylor et al., 2019) and NBb (Gałka et al., 2025). In contrast, shallow water levels (negative Z-scores) are primarily recorded between ∼ 500 and 1250 CE.

Our analysis of charcoal size classes reveals that most particles fall within the 150–500 µm range, with only a small proportion exceeding 500 µm, and no macrocharcoal fragments (> 1 mm) are identified (Fig. ; Appendix ). The dominance of smaller particles, together with evidence that the nearest recent fires occurred at a mean distance of ∼ 12 km away, suggests that the charcoal source areas of our tundra fires encompass broader landscapes (Figs.  and c). Whilst typical source areas for macrocharcoal are estimated to be a few kilometers, particularly in peatlands, regional source areas extending to 20 km and beyond, i.e., up to 100 km, have also been reported (Higuera et al., 2005, 2007; Adolf et al., 2018; Vachula et al., 2020). This is particularly true for smaller, lighter, and more aerodynamic particles, often of herbaceous origin, or under the specific wind and fire behaviour conditions (Vachula et al., 2020; Vachula and Rehn, 2023).

The timing of elevated charcoal accumulation rates in isolated but mostly in several consecutive samples around 1965–1975, 1985–1990, 2005, and 2010–2015, corresponds reasonably well with periods of frequent fires documented by satellite records (Figs a, b and a). The high CHAR values observed in 2005 and between 2010 and 2015 (Fig. a), particularly at sites 1 to 7 (Fig. ), may have been derived from the large fires, such as the Anaktuvuk River fire (2007), and others that occurred in 2007 (Fig. a and b), suggesting a link with fire size. The occurrence of a few small fires is also noted in 2000s (white dots) and 2010s (yellow dots; Fig. b) near most sites. In contrast, the elevated CHAR values in the 1970s (Fig. a), particularly evident at sites 5 to 9 (Fig. ), correspond to a few small fires near these sites (Fig. b, pink dots), implying a link to the proximity of fire. However, because the closest documented fires occurred at considerable distances (3–29 km) from our sampling sites, and fire source areas often overlapped, it remains difficult to conclude whether the elevated CHAR values primarily reflect fire size and proximity, or a combination of both.

An intriguing outcome of the charcoal–satellite comparison is that the most recent large fires (2001–2015 CE) are associated with lower-amplitude charcoal signals. This feature contrasts with the occurrence of less frequent, smaller fires between 1969 and 1985 CE and higher charcoal levels (Fig. ). This discrepancy may relate to differences in the type of biomass burned and fire intensity. Graminoids produce substantially less charred material per unit biomass than forbs and shrubs, particularly ericaceous species. High fire temperatures can lead to near-complete biomass consumption, resulting in the production of ash (Pereboom et al., 2020; Feurdean, 2021). Peatland dryness has intensified in recent times, potentially favouring higher fire temperatures and complete biomass combustion and may therefore explain the low charcoal values observed during recent large fires. These findings highlight the complexity of interpreting charcoal records, as charcoal abundance reflects not only fire size and proximity, but also fuel amount and type, as well as combustion conditions (see Sect. ).

Our satellite-based fire database from the northern slopes of the Brooks Range along the Dalton Highway agrees with previous findings that many tundra ecoregions of Alaska have experienced a marked increase in fire frequency and size over the past 5 decades (Higuera et al., 2008; Vachula et al., 2020; Miller et al., 2023; Scholten et al., 2024). Specifically, our data show that lightning-ignited fires occurred at an average rate of 0.5 peryear between 1969 and 1990, followed by a fire-free period from 1990 to 2001 CE. Fire activity surged again between 2001 and 2017 CE, with an average of one fire every 1.5 years and larger burned areas, before declining in the most recent years (Fig. a and b). Historical evidence from 1948 aerial imagery further documents tundra fires that burned sometime between 1880 and 1920 on the northern Brooks Range, i.e, near the Ketik and Meade Rivers, Shivugak, and Starfish Bluff, with additional anecdotal records mentioning fires in 1952 and 1959 on the Arctic Slope (Miller et al., 2023).

Our individual and composite charcoal records spanning the past three millennia indicate very low fire activity in Arctic tundra from ∼ 1000 BCE to 1000 CE (Figs.  and ). A slight increase occurred after 1000 CE, peaking around 1200 CE, but overall burning remained low until ∼ 1950 CE, when both individual and composite records show an abrupt and unprecedented rise in fire activity. This recent increase contrasts sharply with the rare and long fire-free intervals of the Holocene and supports findings from other tundra ecoregions in Arctic Alaska in documenting unusually frequent fires in these cold, with low-above ground biomass ecosystems (Higuera et al., 2008, 2011; Chipman et al., 2015; Hu et al., 2015; Vachula et al., 2020; Hoecker et al., 2020; Frank-DePue and Chipman, 2025). Holocene charcoal records from the northern slope of Alaska further indicate that fire return intervals were even longer prior to 3000 years ago. However, shorter fire cycles, like those observed in modern tundra fires, were also detected (Frank-DePue and Chipman, 2025). Another finding of our study is that individual charcoal records show a heterogeneous pattern in fire occurrence before 1950 CE, and a more homogeneous one thereafter, consistent with the emergence of more regularly occurring fires in recent decades. Lower temporal resolution and larger age control errors in the older part of the sediments could have also contributed to this apparent heterogeneity. However, the more elevated values in biomass burning around 1200 and 1600 CE, visible at multiple sites, suggests that homogenous fire occurrence across sites in the past could also be reconstructed.

Recent increases in temperature and lightning frequency are often cited as key drivers of enhanced fire activity across Arctic tundra regions (Hu et al., 2010, 2015; Hoecker et al., 2020; Sim et. al., 2023; Scholten et al., 2024). Our independent reconstruction of peatland hydrology reveals that the lowest water table levels (positive Z-score), indicating the driest conditions of the late Holocene, occurred in the past few decades, coinciding with an intense fire period (Fig. ). This recent deepening of water table was attributed to warming-induced permafrost thaw (Taylor et al., 2019; Gałka et al., 2023; Cleary et al., 2024). Nevertheless, we identified two earlier episodes of lowered water tables: a pronounced drying between ∼ 250 and 500 CE and a lower amplitude drying between ∼ 1250 and 1600 CE (Fig. ). The ∼ 250–500 CE dry period was primarily determined by low water levels at TFSII (Taylor et al., 2019) and NBb (Gałka et al., 2025), but only TFSII shows a corresponding increase in charcoal accumulation rates (Figs.  and ), suggesting that peatland drying did not consistently translate into higher fire activity during this period. In contrast, the ∼ 1250–1600 CE dry period coincides with greater but overall pronounced fluctuations in composite biomass burning, however, peak biomass burning occurs between 1000–1200 CE, thus before the onset of water level deepening. Arctic Alaska experienced cooler temperatures and decreased moisture during the late Holocene (Taylor et al., 2019; Frank-DePue and Chipman, 2025) The reduced summer precipitation and lower temperatures of the Little Ice Age (∼ 1250–1600 CE) are documented to have deepened the peatland water tables (Taylor et al., 2019). Reduced precipitation and drier peatland conditions likely promote the desiccation of surface fuels, increasing their flammability and likelihood of ignition and burning (Kettridge et al., 2015; Feurdean et al., 2022; Fran-DePue and Chipman, 2025).

Our multiproxy records provide a unique opportunity also to examine local feedbacks between fuel type and availability and fire, particularly in areas prone to pronounced drying, such as hummocks. Plant macrofossil data show the highest abundance and diversity of shrub communities, Betula nana, undifferentiated Ericaceae, Ledum, Andromeda, Vaccinium, and Salix, after 1900 CE, with a notable increase after 1950 CE (Fig. ). The pollen record, reflecting landscape-scale vegetation composition, shows a pattern of shrub expansion around 500–700 CE, 1200–1600 CE, and after 1950 CE (Fig. ). Thus, similarly to biomass burning, shrub abundance (particularly shrubs pollen) has generally increased over the last millennium and peaked in recent decades. The visual correlation between this increase in fire activity and expanded shrub cover suggests that greater woody biomass, thus fuel quantity promoted fire. Although this relationship could not be confirmed statistically (not shown), our results suggest a complex role for moisture–vegetation feedbacks in shaping fire regimes in Arctic tundra ecosystems.

Pollen and macrofossil evidence also indicate a compositional change in the shrub and their behaviour to fire, that may have affected the fire-vegetation relationship. There is a shift from Alnus and Salix, which were dominant before 1850 CE, to an increased abundance of Ericaceous shrubs and, to a lesser extent, Betula nana in recent decades (Gałka et al., 2018, 2023, 2025). Salix and Alnus typically grow on dry mineral soils, whereas Ericaceous shrubs are characteristic of peaty soils, which have become drier in recent decades. Ericaceous shrubs are also highly flammable and re-establish post-fire, while Betula species resprout vigorously following fire (de Groot and Wein, 1999; de Groot et al., 2013). A strong link between the expansion of resprouting taxa, such as Betula glandulosa, and elevated fire activity has been documented in north-central Alaska during the Lateglacial and early Holocene periods (Higuera et al., 2008). The satellite-derived vegetation maps reveals that the many recent fires in the study area occurred in tussock sedge–dwarf shrub and dwarf shrub communities (Raynolds et al., 2006, Appendix ).

Finally, temporal mismatches among our fire, vegetation, and moisture records could be attributed to taphonomic processes. Plant assemblages were dominated by graminoids (Cyperaceae) and mosses, primarily Sphagnum, with shrubs becoming abundant only in the past ∼ 500 years (pollen record) and especially over the past decades (plant macrofossils). Because graminoids and mosses produce little charcoal per unit biomass and combust rapidly (Pereboom et al., 2020; Feurdean, 2021), low charcoal accumulation during earlier dry periods may also reflect this herbaceous–moss dominance. Further, differences in the temporal and spatial resolution of proxies and speed of ecosystem responses may have contributed to temporal offsets among them. Pollen and macrocharcoal (especially particles < 500 µm) integrate regional to local vegetation and fire signals, whereas plant macrofossils and testate amoebae primarily reflect local and extra-local peatland conditions. Consistent with these differences, pollen-inferred shrub expansion appears to show a better temporal correspondence with increased charcoal-inferred biomass burning than do plant macrofossils or water-table reconstructions.