An Upper Pleistocene and Holocene Black Carbon‐Related Fire Record From the SW Balkans

Lake sediments are unique archives of human environment interactions. Lake Prespa is one of the oldest lakes in Europe, lying in the southwestern Balkans and thus on a possible dispersal route of anatomically modern humans from Africa. In this study, we investigated the effects of climate, vegetation and human activity on fire over the last 92,000 years in this region. Sediment samples were taken from Lake Prespa, and nearby Lake Ohrid for comparison of the regional relevance, and analyzed for benzene polycarboxylic acids as markers for burned organic carbon residues (black carbon, BC). Peaking contents of BC (up to 1 g BC kg−1 sediment) coincided with warm and humid phases, when forests expanded at marine isotope stages (MIS) 5 and 1. During the colder and more arid climates of MIS 4 and 2, BC contents were lowered by a factor of 10, with a distinct minimum during the Last Glacial Maximum (0.07 g BC kg−1). The ratio of pentacarboxylic acid to mellitic acid (B5CA/B6CA) declined from 1.2 at MIS 5 to values of 0.3 at MIS 2, confirming a change in fire regime. Overall, BC contents peaked at cycles of solar radiation, vegetation composition and fuel availability, and thus correspond to the BC signal of other environmental archives. However, in the Late Holocene and as a result of human sedentary settlement, BC production increased, independent of decreasing insolation forcing.

settlement, agriculture and forest clearing (Eastwood, 2004;Willis, 1994).First traces of anatomically modern humans reaching the Balkan Peninsula date back to more than 210 thousand years (ka) ago (Harvati et al., 2019), as the geographical position between the Asian and European continents made it a key area for the spread of modern humans (Groucutt et al., 2015).However, only Neolithic occupation eventually left a notable impact on pollen and microcharcoal records (e.g., Fouache et al., 2010;Marinova & Ntinou, 2018;Ntinou & Badal, 2000;Panagiotopoulos et al., 2013).
Palynological analysis of microscopic charcoal particles may miss some important fire residues due to their small size (e.g., <10 μm).In contrast, the geochemical analysis of black carbon (BC) promises information on the full spectrum of combustion residues, that is, fire tracing does not depend on visibility and thus the size of particles (Glaser et al., 1998;Goldberg, 1985;Keiluweit et al., 2010).For that matter we have relied on benzene polycarboxylic acids (BPCAs) as marker for molecular sized BC remains, providing both information on BC quantity and quality (Brodowski et al., 2005;Glaser et al., 1998;Kappenberg et al., 2016).However, similar to conventional charcoal counting, the mere presence of fire residues does not inform about the frequency of fires, because once charcoal is formed and remains on the soil surface, it can be reburned in subsequent fire events (Czimszik & Masiello, 2007).The more biomass available for burning, the more remains are likely to be blown into lakes; in this regard, the sum of BPCAs is an indirect biomarker of the amount of biomass that was available to fires.In addition, changes in fire temperature can also be detected with the BPCA technique.Hotter fires produce more mellitic acid (B6CA) compared to, for example, pentacarboxylic acid (B5CA); therefore, the B5CA/B6CA ratio decreases with increasing fire temperature (Schneider et al., 2010;Wolf et al., 2013).
When tracing BC in soil, it should be noted that BC can be altered by oxidation and biodegradation processes in addition to reburning (Rodionov et al., 2006).In contrast, BC is highly resistant to decomposition when buried under anaerobic conditions of lacustrine sediments (M.W. I. Schmidt & Noack, 2000).Therefore, the analysis of BC in such archives seems promising to complete information about past burnings and the associated environmental conditions of a region, such as the catchment area of a lake.
In this study, we investigated the fire history preserved in the Lake Prespa archive using a geochemical BC method.The Dessarate Lake (Prespa-Ohrid) region provides exceptional and continuous vegetation records and served as a floristic refugium during successive glacial-interglacial cycles of the Quaternary (e.g., Donders et al., 2021;Panagiotopoulos et al., 2014Panagiotopoulos et al., , 2020)).Our aim was (a) to identify the relationships between fire residue input and composition and environmental changes (climate, vegetation, human activity), and (b) to link the Prespa fire history with regional and global archives.We hypothesized that climate forcing controlled BC influx, but that lake sediments exhibit changes in fire regime with the establishment of Neolithic settlement in the region.Given the size and morphology of Lake Prespa and based on previous microcharcoal studies from the region, we assumed that the BC inferred fire signal has captured changes at a regional scale.

Study Area, Environmental, and Climatic Settings
For this study, two sediment cores from Lake Prespa and neighboring Ohrid in the southwestern Balkans were used, with the main sampling related to Lake Prespa.Lake Prespa is situated at an altitude of 849 m a.s.l., shared between the Republics of North Macedonia, Albania, and Greece (Figure 1), and has an approximate surface area of 253.6 km 2 and a catchment area of about 1,300 km 2 (Hollis & Stevenson, 1997).Lake Prespa has no surface outflow, but water is flowing through karst aquifers of the Galičica mountain range to nearby Lake Ohrid 15 km northwest (Figure 1).Lake Ohrid is shared between the Republics of North Macedonia and Albania at an altitude of 693 m a.s.l., and has a surface area of 358 km 2 and a catchment area of 1,310 km 2 .The location of the lakes at a transitional climate zone in combination with the diverse topography and their long existence is responsible for the high biodiversity and refugial character of the region (Albrecht & Wilke, 2008;Panagiotopoulos et al., 2013Panagiotopoulos et al., , 2014)).To date, multi-proxy approaches including sedimentological, geochemical and palynological analyses were performed on samples from these lakes to explore past environmental and climatic conditions of the region (e.g., Matzinger et al., 2006;Panagiotopoulos et al., 2020;Sadori et al., 2016;Wagner et al., 2010).Ten local pollen assemblage zones (PAZ; P1-P10) were determined based on cluster analyses and three major phases of vegetation development were distinguished over the last glacial-interglacial cycle (Panagiotopoulos et al., 2014): forested phases in marine isotope stage (MIS) 5 and MIS 1, dominated by deciduous trees suggesting higher temperatures and moisture availability, an open landscape in MIS 3 with patches of temperate trees, and 10.1029/2022PA004579 3 of 14 open landscapes in MIS 4 and MIS 2, when lower temperatures and humidity prevailed and vegetation was dominated by pines.

Sampling
Core composite Co1215 was collected from the northern part of Lake Prespa (40°57ʹ50ʺN, 20°58ʹ41ʺE) at 14.5 m water depth in November 2009 and June 2011, and comprises a total composite length of 17.76 m (Panagiotopoulos et al., 2013(Panagiotopoulos et al., , 2014;;Wagner et al., 2012).Isotopic, hydrological, lithological and tephrostratigraphical studies revealed the sensitivity of this site to climatic and environmental variability (Leng et al., 2013;Wagner et al., 2012), and indicated undisturbed sediment accumulations throughout the last 92 ka cal BP according to the age-depth model (Figure 2), which is based on radiocarbon dating, tephrochronology and electron spin resonance dating (Damaschke et al., 2013).Individual samples were taken in 2-cm intervals throughout the core composite and stored in plastic vials after freeze-drying.For our study, a total of 53 samples were analyzed for BC.
To test the hypothesis that fire residues at Lake Prespa reflect a regional signal, a low resolution sample set of a sediment succession from adjacent Lake Ohrid was analyzed for BC.This sediment succession was recovered within the scope of the International Continental Scientific Drilling Program (ICDP; Wagner et al., 2017) at the DEEP site (Figure 1; 41°02ʹ57ʺN, 20°43ʹ54ʺE, 243 m water depth; Francke et al., 2016) in 2013.Ten samples with approximate intervals of 10 ka between 89.5 and 5.8 ka BP were analyzed, corresponding to sediment depths between 36 and 2 m according to the age-depth model presented in Wagner et al. (2017).

Black Carbon Analysis via Benzene Polycarboxylic Acids
Pre-treatment of the samples included drying (40°C), sieving (<2 mm) and milling.The oxidation of BC to BPCA was performed according to the method of Glaser et al. (1998) with modifications by Brodowski et al. (2005) and Kappenberg et al. (2016).A limit of 5 mg organic carbon per sample was strictly maintained.Metal elimination was achieved by hydrolysis with trifluoroacetic acid (4 hr, 105°C).The residue was oxidized with 65% nitric acid (8 hr, 170°C), dried and BPCAs were purified with a cation exchange column (Dowex 50WX8,Fluka,Steinheim,Germany).Individual BPCAs were converted to trimethylsilyl ester by derivatization and measured using a gas chromatograph equipped with a flame ionization detector (Packard 6,890 gas chromatograph, Hewlett Packard GmbH, Waldbronn, Germany) and a HP-5 capillary column (30 m × 0.32 mm i.d., 0.25 mm film thickness, Macherey-Nagel, Düren, Germany; for oven program see Brodowski et al. (2005)).Citric acid was used as the first and biphenylene-dicarboxylic acid as the second internal standard to quantify the recovery of citric acid (average recovery: >80%).As proposed by Glaser et al. (1998), BPCA yields were corrected by a factor of 2.27 for CO 2 loss and insufficient conversion of BC to BPCAs, representing a conservative, minimum estimate of total BC in soil (Brodowski et al., 2005).To avoid a contribution of BPCAs from non-pyrogenic matter, the evaluation was limited to five and six times carboxylated BPCAs (B5CA and B6CA), that is, three and four times carboxylated BPCAs (B3CA and B4CA) were excluded (Kappenberg et al., 2016).Since the accuracy of BPCA analyses is usually little better than 10% of the mean value, we refrained from interpreting each peak of the BPCA contents and focused on the signals that stood out clearly from their respective backgrounds.

Black Carbon at Lake Prespa
Fire residue input, as depicted from BC content (g BC per kg sediment), was found in all samples of Lake Prespa with different contents.BC content was generally low in periods of cold and dry climates with open steppe vegetation and low arboreal pollen (AP) percentages (MIS 4 and 2; corresponding to PAZ7, 6, 3, and 2; Figure 3).The minimum BC value (0.07 g C kg −1 at ∼23.7 ka BP) was measured in the period of the last glacial maximum (LGM), and coincided with the lowest percentages of Quercus pollen of the record (Figure 3).The BC content was elevated during warm and humid phases with deciduous forests (MIS 5 and 1; PAZ10, 9, 8, and 1) with a maximum at ∼1.1 ka BP (1.06 g C kg −1 ), about 15 times higher than the minimum during the LGM.BC contents slightly increased during MIS 3, when oaks and other deciduous trees became more abundant than in MIS 4 and 2 (Figure 3).The BC contents correlated significantly with the concentrations of AP (r 2 = 0.47; p < 0.001) and non-arboreal pollen (NAP; r 2 = 0.56; p < 0.001); also with the percentages of AP-excl.pines (r 2 = 0.74; p < 0.001), temperate trees (r 2 = 0.75; p < 0.001) and Quercus (r 2 = 0.67; p < 0.001).The correlation with the percentages of pine pollen (r 2 = −0.67;p < 0.001) was negative.2013); for details see cited publication.Marine isotope stages (MIS) 1-5 (Bassinot et al., 1994) are also shown.

10.1029/2022PA004579
5 of 14 Some coincidences between BC and July solar insolation were observed throughout the archive; most clearly in MIS 5, where BC content was closely associated with the changes in insolation (July insolation shown as red line; Figure 3), but there were also exceptions in MIS 3 and the Late Holocene, where the input increased despite decreasing solar radiation.
Fire intensity was inferred from the ratio of B5CA to B6CA.The relative proportion of BPCA with five to six carboxyl groups depends on the combustion temperature and aerosol formation.At Lake Prespa, B5CA/B6CA ratio values declined from MIS 5 (ratio 1) to 16.3 ka BP in MIS 2 (ratio 0.4; Figure 3).Between ∼16 and ∼10 ka BP, the B5CA/B6CA ratio values seemed to follow changes in AP closely and reached a new minimum (maximum burning temperature) with the onset of the Neolithic.Overall, B5CA/B6CA ratio values correlated significantly with BC contents (r 2 = 0.55; p < 0.001).10.1029/2022PA004579 6 of 14

Black Carbon and Microcharcoal
Contents of BC and microcharcoal accumulation rates showed a similar pattern throughout the Prespa archive with increased values during MIS 5 and 1 and lowered values in MIS 4 and 2 (Figure 3).In general, the increase in BC contents was comparatively larger than that of microcharcoal, at least during MIS 5, 4, and 3.In turn, fluctuations observed in microcharcoal values, especially during MIS 2 and 1, were not detected with BC analysis, possibly due to the difference in sampling resolution.Overall, BC and microcharcoal yielded a significant though moderate correlation (r 2 = 0.33; p < 0.05).Note that absolute differences between BC contents and charcoal counts increased with increasing age of the samples (Figure 3).

Black Carbon at Lake Ohrid
The comparison of BC records from Lake Prespa with those from Lake Ohrid yielded coinciding results and showed that both BC content and ratio values follow a similar pattern (Figure 4).Despite the relatively low resolution of the BC record from Lake Ohrid, pollen records from the two lakes indicate similar vegetation history trends in the two adjacent catchments (Lake Prespa in Panagiotopoulos et al., 2014;Lake Ohrid in Sadori et al., 2016).

Black Carbon Input
The sediment core of Lake Prespa spans the last 92 ka and covers different vegetation and climate phases (Panagiotopoulos et al., 2014).We found that fire residue input varied in accordance with vegetation changes and especially between warmer and colder phases.The BC contents were largest during the forested phases of MIS 5 and 1, when a warmer and more humid climate prevailed in the region, while the contents were lower during the open-steppe vegetation of MIS 4 and 2 with colder and less humid conditions (Panagiotopoulos et al., 2014).Accordingly, BC contents correlated significantly with AP (minus Pinus) percentages throughout the complete sediment record, and in particular with that of deciduous oak percentages, which were higher during warm/humid periods (MIS 5 and 1; Figure 3).Although the deciduous oak percentage curve suggests their continuous presence within the Lake Prespa catchment for the entire study interval, they became a dominant vegetation component mainly within interglacial intervals (Panagiotopoulos et al., 2014).Hence, BC peaked during the presence of deciduous trees and not when, for example, pine trees were most abundant.
Usually there is a low flammability of deciduous oaks (Curt et al., 2011).However, several evergreen oaks in the Mediterranean are sclerophyllous and thus prone to the occurrence of fires; respective pollen in the study area may reach 15% during the Holocene, suggesting an increasing seasonal drought during this interval (Panagiotopoulos et al., 2014).In either case, the spread of different deciduous oak trees during periods with high BC accumulation likely indicates that fuel availability is critical to fire events, and thus depends on climate-related, edaphic factors that control fire activity and facilitate burning, particularly under warm/humid conditions.
In general, our data thus support earlier evidence that increased burning events are associated with warmer periods (Pyne et al., 1996;Thonicke et al., 2010;Wang et al., 2012).Specifically, for the Mediterranean region, fires have been reported to correlate with drought events and to be closely linked to seasonal climate (Bernhardt et al., 2012;Colombaroli & Tinner, 2013;Keeley, 2012).This is also true for Lake Prespa, where periods with increased mean annual temperatures and moisture availability promoted biomass production and thus maximum fuel availability within the catchment (Panagiotopoulos et al., 2013).As fuel moisture can hinder ignition and fire spread (Blackmarr, 1972;Wotton et al., 2010), annual summer drought (here indicated by high July insolation; Laskar et al., 2004) in overall humid periods ensures that the fuel is dry enough to burn (Scott et al., 2014), as also reported for other archives (Bistinas et al., 2014;Daniau et al., 2013;Kappenberg, Lehndorff, et al., 2019).We also assume that thunderstorms may have played a crucial role in BC production in the Prespa region.A number of studies have discussed that lightning activity is most prevalent when warm humid air masses are present, which is particularly the case during interglacials in summer.Also most overland thunderstorms in the Mediterranean occur in spring and summer (Galanaki et al., 2018;Kotroni and Lagouvardos, 2016).Mild and wet winter/spring seasons can also create broad areas of contiguous fuels allowing fire to be widespread during the dry season (Keeley, 2012).These seasonal conditions have already influenced several regions worldwide since the Oligocene and early Miocene, creating the five mediterranean-type climate regions of the world (Lamont and He, 2017;Pausas and Paula, 2012;Rundel et al., 2018).A recent study from Lake Ohrid during the Early Pleistocene showed that microscopic charcoal concentration peaked during phases of increased seasonality confirming that this pattern observed in Late Quaternary records occurred also within obliquity-paced climate cycles (Panagiotopoulos et al., 2020).
During periods of low solar radiation in summer, BC contents were comparatively low, while relatively high pollen concentrations indicate increased plant biomass production and fuel availability (Figure 3).Most likely, fuel drying and thus BC production was rather limited during these periods (e.g., in P10 and at the beginning of P9; Figure 3).As deciduous forests are rather unlikely to burn frequently on a broad spatial scale, fire development may be restricted by the density and moisture content of a forest environment (Marlon et al., 2013;Thonicke et al., 2010;Zhang et al., 2015).Overall, we found a negative correlation of BC and pine pollen percentages that dominated AP during intervals when open-steppe vegetation (P7-6, P4-2) prevailed in the Prespa region.However, low AP concentrations during these periods point to a rather open landscape with isolated pine stands (Panagiotopoulos et al., 2014; Figure 3), which could hamper fire spread.Furthermore, pines are often over-represented in pollen percentage diagrams, because they produce a lot of pollen relative to other taxa, and their pollen grains can travel long distances in an open landscape (e.g., Donders et al., 2021;Sadori et al., 2016).
The negative correlation of BC to non-arboreal pollen percentages is in contrast to other studies from different biogeographical regions, which showed increased BC contents in tundra-like, grass dominated environments during cold and dry climates (Wang et al., 2005(Wang et al., , 2012)).Also, Lehndorff et al. (2015) found elevated BC contents during periods with tundra-steppe vegetation, and decreased contents in tree-dominated periods in Central European lake archives.Apart from the fact that a minimum amount of fuel is required for fire start and spread, the limited fuel availability at Lake Prespa may have also been responsible for more frequent reburning of BC at the surface with less dense vegetation, as also suggested as a possible loss mechanism by Czimczik and Masiello (2007) and Kane et al. (2010).However, no direct evidence for this has been found so far; rather recent studies suggest that reburning is likely a minor factor in fire residue loss (Saiz et al., 2014;Santín et al., 2013Santín et al., , 2016)).

Relation of BC to Microcharcoal Counts
BC contents appear to be in agreement with microcharcoal accumulation rates, which also peaked during periods with maximum ecosystem productivity and high July insolation (Figure 3).In principle, it is possible that microcharcoal is degraded to finer particles that are not visible to the human eye.The larger BC than microcharcoal contents in MIS 5 to 3 could therefore indicate a larger range of fire residue sizes detected with the BPCA analyses.Degradation of recalcitrant BC structures likely result in a depolymerization of the condensed aromatic moieties and thus in lower portions of B6CA over time (e.g., Rodionov et al., 2010).Yet, such evidence for BC transformations only applies for aerobic conditions but not for anaerobic ones.Hence, the general larger BC values from MIS 5 to 3 likely indicate an increased input of fire residues from long-distance transport, which would have affected mainly non-visible, submicron-scale BC particles and would therefore only be detectable by BC analysis (Clark & Patterson, 1997;Kaal et al., 2008).In MIS 2 and 1, contents of BC and microcharcoal were similar on average, suggesting that fire residues at Lake Prespa during this period were predominantly from regional and/or local fires.

Black Carbon Quality
At Lake Prespa, B5CA/B6CA ratios decreased from MIS 5 to 2 (Figure 3).Declining ratios can be interpreted as indicators of a change to higher combustion temperatures (Wolf et al., 2013) and, related to this, decreasing atmospheric inputs of fire aerosols into environmental archives (Kappenberg, Braun, et al., 2019).Even though no exact combustion temperatures can be derived from the ratio values, Wolf et al. (2013) differentiated typical fire regimes such as forest floor fires, grass fires and domestic fires by their characteristic BPCA pattern.Therefore, it was reasonable to deduce a change in the B5CA/B6CA ratio in line with vegetation changes and thus fuel availability.Interestingly, our data do not support this assumption, and especially the ratios in MIS 3 and 2 were unusually low, more typical of domestic fires fueled with hardwood and abundant oxygen supply, resulting in high combustion temperatures (cf.Lehndorff et al., 2015;Wolf et al., 2013;Wöstehoff et al., 2022).However, the changes in the ratio from MIS 5 to 2 followed the offset between microcharcoal counts and BC from geochemical analyses (Figure 3).In principle, fire condensates are formed from the gas phase, and can be transported as aerosols over long distances through the atmosphere due to their small size (Keiluweit et al., 2010;T. C. Schmidt et al., 2004).Since these condensates generally have higher B5CA/B6CA ratios than charcoal (Kappenberg, Braun, et al., 2019), increased condensate input with increased opening of the landscape during MIS 4 and 2 might have influenced the ratio values found as also suggested by Lehndorff et al. (2015) in the context of forest opening.Here, however, it is not possible to determine the contribution of fire condensates to the fire signal.
As in the comparison with microcharcoal (Section 4.1.2),the increase of the B5CA/B6CA ratio with age could be due to the degradation of recalcitrant BC structures resulting in depolymerisation of the condensed aromatic moieties, and thus in lower portions of B6CA, as assumed, for example, by Rodionov et al. (2010) to explain increasing B5CA/B6CA ratios with soil depth and radiocarbon age.Additionally, the ratio could also have been altered due to a re-combustion of fire residues during MIS 4 and 2. Overall, it should be kept in mind that the ratio values may not reflect actual fire temperature if they are derived from very low BC contents, as was the case with some samples from MIS 2.

Lake Ohrid
The fire signal as here based on BC content and quality revealed a similar pattern during the Last Pleistocene for both Lake Prespa and Lake Ohrid (Figure 4).The warmer mean annual temperatures and elevated moisture availability, as prevailing in MIS 5 and 1, led to an increase in BC production, while colder temperatures and less humid conditions between MIS 4 and 2 decreased BC input.Two conclusions may be drawn from this: First, the results of the BC analyses can be considered reliable, and second, the fire residues preserved in the lakes seem to represent the regional rather than the local fire history as the two are located within a distance of 15 km to each other (Figure 1).Besides BC content, which as expected was elevated in both lake archives during the warmer periods, it is noteworthy that BC quality (i.e., fire temperature) also showed a similar trend in both archives.However, as Lake Ohrid is partly fed by underground hydraulic connections from Lake Prespa (Anovski et al., 1980;Eftimi & Zoto, 1997), BC particles could theoretically have been transported from one lake to the other.We consider this input to be negligible, as the water from the karst is extremely clear and does not contain any or only very few particles such as Prespa algae.

Central Europe and Anatolia
There are other lacustrine archives where fire residues have been analyzed, but only few cover the time span of the Prespa archive while providing good temporal resolution.The comparison of our results with lake sediment cores from Central Europe and Anatolia confirmed that BC (and microcharcoal) generally peaked during forested phases with warmer climatic conditions (Figure 5).Former studies have elaborated a global control of fire activity by climate (e.g., Inoue et al., 2018;Kappenberg, Lehndorff, et al., 2019).Here, no general conformity regarding the timing of fire residue peaks could be found.The forested phases did not occur simultaneously in all archives, which is related to differences in climate regimes and possible uncertainties in chronology.This underlines the dependence of fire activity on the respective climatic conditions in a region, as fire residue input followed forest development at least since the beginning of the Upper Pleistocene (Figure 5).2014); (iii) Lake Van (Turkey): BC input and pollen data after Kappenberg, Lehndorff, et al. (2019).All data is plotted against age in ka BP and marine isotope stages.
The BC production was generally elevated in the forested phases, although forest types varied by study area.In Central Europe, in the area of the maar lakes, boreal forests predominated and fire residues correlated most strongly with spruce pollen, whereas in the Mediterranean, high BC amounts especially correlated with deciduous oak pollen (Figure 5).

Relation to Prehistoric Human Activity
During the Holocene and with sedentary lifestyle, humans began to significantly affect the input of fire residues into limnic and terrestrial archives globally.However, the beginning of anthropogenic influence on burning varied around the world (Marlon et al., 2013).At Lake Prespa, we observed increased BC accumulation from about 15 ka onwards (Figure 3), and accordingly before the onset of the Holocene and while the glacial conditions of MIS 2 still prevailed (Older Dryas).As this may be an indication of an earlier human impact on fire regime, it has to be kept in mind that only shortly after, the Bølling-Allerød Interstadial (c.13.7-12.7 ka BP; Litt et al., 2007) and the Younger Dryas (c.12.9-11.7 ka BP; Rasmussen et al., 2006) emerged.They brought changing climatic conditions, and thus could have influenced fire activity as well.In general, the higher aridity during the Younger Dryas may have favored the drying of fuel, but at the same time, winter precipitation was limited during that phase, which probably reduced additional fuel build-up.Higher resolution BC data may be needed to decipher the relationship between fire and environmental conditions and possible human intervention at these variable times.
At least since the Early Neolithic, there is clear evidence of human settling in the region.For both lakes under study, several lakeshore settlements have been discovered, for example, Kallamas at Lake Prespa, dating back to ∼7 ka BP (Lera et al., 2016), and Ohridati and Pogradec at Lake Ohrid, dating to about 7.6-7.4ka BP and about 8-7.7 ka BP, respectively (Allen & Gjipali, 2014;Hafner et al., 2021;Westphal et al., 2010; locations of the sites in Figure 1).These data correlate well with the generally elevated BC influx during this period and suggests a human impact on burning.Interestingly, the notable BC increase at 8.6 ka BP at Prespa (Figure 3) could be directly related to excavated settlement remains of several Neolithic communities in northern Greece, also dated to 8.6 ka BP (Maniatis, 2014).In general, the persistence of elevated BC levels despite the decreasing July insolation from ∼10 ka BP onwards, supports the assumption of human impact on burning, as solar radiation has been shown to control fire activity in times of human absence.
A decrease in deciduous oak pollen percentages between ∼9 and ∼2 ka BP in the Prespa catchment is synchronous with a change of BC quality toward higher combustion temperatures (Figure 3), which can be typical in wood-fueled anthropogenic fires (Wolf et al., 2013).Since the beginning of the Holocene, an ecological succession seems to have taken place with oaks gradually being replaced by various deciduous tree species such as hazel, hornbeam and beech (Panagiotopoulos et al., 2014(Panagiotopoulos et al., , 2020)).As the oak-dominated forests are located in the lower elevations of the catchment and are thus most easily accessible to free up land for agriculture, livestock or direct use as fuel or building material, an acceleration of the succession by human activities seems likely.Indeed, from 6.5 ka BP onwards, large quantities of wooden pillars (mainly from oaks) were found at adjacent Lake Ohrid (Hafner et al., 2021).

Summary and Conclusions
This study presents the first BC record of the Dessarate Lake region and contributes to a comprehensive picture of the environmental history in this region.Fire residue production during the Last Glacial Period at Lake Prespa was driven predominantly by seasonality, vegetation density and composition, and, in part, insolation.The comparison with (a) nearby Lake Ohrid showed that the fire record of Lake Prespa reflects a regional rather than a local signal, and (b) Central European and Anatolian archives has strengthened recent hypotheses that fire activity was highest during periods of warm/humid climates where forests were abundant.During MIS 4-2, the geochemical tracing of BC input deviated from microcharcoal countings, and pointed therefore, to increased portions of long-distance BC transport.This changed from MIS 2 to 1, when regional fires with less contributions from long-distance transport controlled the fire signal, or exceeded the supraregional.This could be an indication of human contribution to burning events as there is evidence of early Neolithic occupation around 8.6 ka BP and a concomitant increase in fire residue input.For a better understanding of the nature of these human impacts, BPCA analyses should be combined more frequently with other markers of fire residues in the future, such as morphological charcoal analyses (e.g., Enache & Cumming, 2006), polycyclic aromatic hydrocarbon patterns (Jiang et al., 1998), and levoglucosan, a specific marker for burned altered cellulose (Schreuder et al., 2019).

Figure 1 .
Figure 1.(a) Overview map of southeastern Europe with locations of Lake Prespa (Republics of North Macedonia, Albania, and Greece) and Lake Ohrid (Republics of North Macedonia and Albania).(b) Detailed map of the two lakes with marking of the core sites (red arrows) and archeological sites (white dot: Ohridati; gray dot: Pogradec; green dot: Kallamas; Hafner et al., 2021).

Figure 4 .
Figure 4. Comparison of black carbon contents (g C kg −1 ) and B5CA/B6CA ratio values in sediment cores from Lake Prespa and Lake Ohrid plotted against age in ka cal BP.

Figure 5 .
Figure 5.Comparison of fire residue contents (black line) and forest development in sediment cores from Central Europe, the Mediterranean and East Asia.(i) Dehner Maar and Jungferweiher (Germany): black carbon (BC) input after Kappenberg et al. (2021); pollen data and forested phases after Sirocko et al. (2016); (ii) Lake Prespa (the Republics of North Macedonia, Albania, and Greece): pollen data from Panagiotopoulos et al. (2014); (iii) Lake Van (Turkey): BC input and pollen data afterKappenberg, Lehndorff, et al. (2019).All data is plotted against age in ka BP and marine isotope stages.