Diss Mere (52°22^′ N, 1°6^′ E; Fig. 1) is a small (0.034 km²), 6 m deep, presently eutrophic lake with alkaline waters and a chalk bedrock (Martin-Puertas et al., 2021). The site occupies a region dominated by maritime climate. During lake stratification, pH is high in the epilimnion (values of ∼9) and lower in the hypolimnion (∼7) (Boyall et al., 2023). The lake contains no inflow or outflow, experiences minimal groundwater input (Boyall et al., 2023) and has an estimated catchment to lake area ratio (CA:LA) of 44. The DISS-16 record exhibits 15 m of sedimentation covering the last 10 thousand years (Martin-Puertas et al., 2021) with varves preserved between 10 and 2 ka BP. The varve couplets are calcite-organic (Fig. 1; including diatoms in the organic component) with some interannual variability (Martin-Puertas et al., 2021). The calcite layer is representative of summer sedimentation with organic constituents deposited during autumn to spring following resuspension of organic remains and wind-induced mixing of the water column (Martin-Puertas et al., 2021; Boyall et al., 2023).

Nautajärvi (61°48^′ N 24°24^′ E; Fig. 1) is a small (0.17 km²), ∼20 m deep mesotrophic lake, exhibiting slightly acidic lake waters (modern pH of 5.8–6; Korkonen et al., 2017). The site occupies a region of sub-Arctic climate. The site sits in a tiered lake system within the Äväntäjärvi drainage basin and is fed by streams and Ristijärvi to the north and is drained by Lake Pitkävesi to the south (Ojala and Alenius, 2005) with a CA:LA of 71. The NAU-23 stratigraphy (Lincoln et al., 2025) contains 7.3 m of sedimentation, with 9829 varves covering the upper 6.74 m (Ojala and Saarinen, 2002; Lincoln et al., 2025). The varve couplets are clastic-organic (Fig. 1), with the clastic layer forming in spring as a product of winter precipitation and snowmelt run off once the lake becomes ice-free, with the organic layer forming in the summer to winter months (Ojala and Alenius, 2005). Inverse stratification is observed during the winter months with anoxic bottom waters at this time (Ojala and Alenius, 2005; Ojala et al., 2013).

Meerfelder Maar (50°06^′ N, 6°45^′ E; Fig. 1) is a relatively small (0.25 km²), 18 m deep, eutrophic maar lake which occupies a predominantly maritime climate region (albeit with some continental influences) within mainland Europe. The lake has a CA:LA of 6. The 11.71 m long MFM-09 sequence contains 7.45 m of sediment across the Holocene (Martin-Puertas et al., 2012). Varves are preserved from 11.6 to ca. 1.5 ka BP (Martin-Puertas et al., 2012, 2017). The Holocene varve structures preserved at Meerfelder Maar are diatomaceous (Fig. 1), with a spring and summer layer composed of a monospecific diatom bloom and the autumn and winter layer composed of organic and minerogenic detritus (Brauer et al., 2000; Martin-Puertas et al., 2012). Hypolimnetic anoxia has previously been proposed at Meerfelder Maar, with treatments performed to increase bottom water oxygen conditions (Nürnberg, 2007). This process disturbed stratigraphy of the upper 500 years of the record.

Published chronologies are derived from annual layer counting and tied to either a radiocarbon and/or tephra framework (Martin-Puertas et al., 2012, 2021; Meerfelder Maar, Diss Mere respectively) or from the cross-correlation of marker layers between an annual layer-counted sequence and newly derived sediments (e.g., Ojala and Saarinen, 2002; Ojala and Alenius, 2005; Lincoln et al., 2025; Nautajärvi). For each sequence the chronological uncertainties are decadal in scale (see Martin-Puertas et al., 2012, 2021; Lincoln et al., 2025). Following Abrook et al. (2025), we present reconstructions as ka BP (as cal. ka BP at Diss Mere and Meerfelder Maar and ka varve years BP for Nautajärvi).

TEX₈₆ is defined following Schouten et al. (2002). 1TEX86=(GDGT-2+GDGT-3+crenarchaeol')(GDGT-1+GDGT-2+GDGT-3+crenarchaeol') TEX₈₆ is calibrated to Lake Surface Temperature (LST) following Powers et al. (2010). See Supplement. 2LST=55.2⋅TEX86-14 We use the MBT5ME′ index following De Jonge et al. (2014). 3MBT5ME′=(Ia+Ib+Ic)(Ia+Ib+Ic+IIa+IIb+IIc+IIIa) Where alphanumeric values relate to different brGDGT structures and the number of cyclopentane moieties and/or methyl branches. MBT5ME′ values are converted to mean temperature of months above freezing (MAF; Martínez-Sosa et al., 2021) using a Bayesian calibration (BayMBT₀) developed for lakes (e.g., Martínez-Sosa et al., 2021). Present day mean annual temperatures are used as prior means at each site: 10.5 °C (Diss Mere); 5.3 °C (Nautajärvi) and 9.9 °C (Meerfelder Maar). Prior standard deviation is set to 10 °C.

We also use the multivariate calibration approach of Raberg et al. (2021). 4MAF=92.9±15.98+63.84±15.58⋅fIbmeth2-130.51±30.73⋅fIbmeth-28.77±5.44⋅fIIameth2-72.28±17.38⋅fIIbmeth2-5.88±1.36⋅fIIcmeth2+20.89±7.69⋅fIIIameth2-40.54±5.89⋅fIIIameth-80.47±19.19⋅fIIIbmeth

The cyclisation of branched tetraethers (CBT) index (Weijers et al., 2007) that includes both 5- and 6-methyl isomers (De Jonge et al., 2014) is used to infer lake pH: 5CBT′=10log⁡Ic+IIa′+IIb′+IIc′+IIIa′Ia+IIa+IIIa We subsequently convert CBT' values to pH using the calibration from Russell et al. (2018). 6pH=8.95+2.65⋅CBT′ The isomer ratio captures the relative abundance of 5-methyl vs. 6-methyl brGDGTs (De Jonge et al., 2015) and is defined as: 7IR6ME=(IIa′+IIb′+IIc′+IIIa′+IIIb′+IIIc′)(IIa+IIa′+IIb+IIb′+IIc+IIc′+IIIa+IIIa′+IIIb+IIIb′+IIIc+IIIc′) Deviations from the expected temperature – MBT_5ME relationship are observed in lake sediments when the IR6ME>0.5 (Bauersachs et al., 2024).

We use the methane index (MI) to evaluate anaerobic methanotrophy within the lake water column and/or sediments, whereby sediments with high values (>0.3–0.5) suggest intense anaerobic methane oxidation (Zhang et al., 2011). 8MI=(GDGT-1+GDGT-2+GDGT-3)(GDGT-1+GDGT-2+GDGT-3+crenarchaeol+crenarchaeol') Crenarchaeol and its isomer are solely produced by Nitrososphaeria (Zhang et al., 2011; Schouten et al., 2013), whereas GDGT-0 can also be produced by methanogenic archaea (Euryarchaeota) (Schouten et al., 2001). The relative abundance of GDGT-0 vs. crenarchaeol (i.e. % GDGT-0 as Sinninghe Damsté et al., 2012; related to GDGT-0/cren) is thus used to evaluate methanogenesis across each of the different lakes. 9%GDGT-0=GDGT-0(GDGT-0+crenarchaeol)⋅100 Where values >67 % are considered to have a substantial methanogen component that could render TEX₈₆ reconstructions unreliable (e.g., Blaga et al., 2009; Inglis et al., 2015).

The branched vs. isoprenoid tetraether (BIT) index (Hopmans et al., 2004; Baxter et al., 2021) evaluates the relative proportion of terrestrial vs. marine organic matter (OM) within individual samples and is defined as: 10BIT=(Ia+IIa+IIa′+IIIa+IIIa′)(Ia+IIa+IIa′+IIIa+IIIa′+crenarchaeol) Where high values (e.g., 1) traditionally reflect high terrestrial input and low values (e.g., 0) indicate high marine/aquatic input.

The ratio of hexamethylated and pentamethylated compounds (Xiao et al., 2016) is used to assess input of terrestrial vs. marine-derived brGDGTs, where ΣIIIa / ΣIIa values <0.59 suggest terrestrial provenance and high values >0.92 represents a marine/aquatic provenance (Xiao et al., 2016).

Finally, we use the branched and isoprenoid GDGT machine learning classification algorithm (BIGMaC) to constrain GDGT source environments (Martínez-Sosa et al., 2023). The algorithm assesses the distribution of all isoGDGTs and brGDGTs and establishes (dis)similarity to GDGT distributions from a global marine, soil, lake, and peat calibration dataset.

The lakes in this study are all thermally stratified in summer months. However, in terms of mixing regime, Diss Mere and Meerfelder Maar were meromictic prior to 2 ka BP. Nautajärvi was dimictic in the early and late Holocene with episodic shifts towards a more meromictic regime coinciding with episodes of reduced and strengthened overturning in the mid-Holocene (Table 1; Lincoln et al., 2025). These mixing regimes influence the distribution of oxygen through the water column, with each lake demonstrating hypoxic/anoxic bottom waters contributing to varve formation and preservation. In permanently stratified (meromictic) lakes, pentamethylated and hexamethylated brGDGTs increase in abundance with depth and are associated with anoxic conditions (Weber et al., 2018; Yao et al., 2020; van Bree et al., 2020; Baxter et al., 2024). As the lakes in this study exhibit anoxic/hypoxic hypolimneitic conditions in the Holocene, as a result of their mixing regime (above; and Table 1) high fractional abundances of pentamethylated (brGDGT-IIa, IIa^′) and hexamethylated brGDGTs (brGDGT-IIIa, IIIa^′) suggests a potential redox influence on brGDGT distributions.

In addition, we identify relationships between brGDGTs and redox sensitive elements across each site location (i.e. sulphur (S) and iron (Fe)). In lakes insoluble sulphides can form through conversion of sulphate by sulphate-reducing bacteria under anoxic conditions (Luo, 2018), whilst precipitation of Fe-hydroxide is associated with oxic conditions (Davison, 1993). At Nautajärvi, we identify a significant positive relationship between S_clr and brGDGT-Ia (r=0.45, p=0.012) and brGDGT-IIa (r=0.54; p=0.002), and a significant negative relationship between Fe_clr and brGDGT-IIIa (r=-0.52; p=0.003) (S7). Given that we observe opposing relationships between these redox-sensitive elements (Fig. 4) alongside temperature-sensitive brGDGTs (i.e., lower relative brGDGT-IIIa with high Fe, derived from mineral precipitates linked to phases of relatively greater hypolimnetic oxidation during strengthened mixing phases; see Lincoln et al., 2025), we suggest that hypoxia and shifts to relatively more oxygenated conditions may influence brGDGT production at Nautajärvi (e.g., Raberg et al., 2022, 2025).

In contrast, at Diss Mere, we observe the opposite relationship between S_clr and brGDGT-IIIa (r=-0.46; p=0.003) (S3). This is also reflected in the redundancy analysis with S_clr being positively loaded with Fe_clr at Diss Mere (Fig. 4). There is also a significant correlation between ln(Fe/Ti), which at Diss Mere captures the anoxic conditions of bottom waters (Boyall et al., 2024), and both brGDGT-IIa (r=-0.4; p=0.01) and brGDGT-IIa^′ (r=0.46; p=0.004) (S3). Increased ln(Fe/Ti) indicates increased bottom water anoxia (Boyall et al., 2024). As such, the negative relationship between brGDGT-IIa and ln(Fe/Ti) suggests that anoxic conditions in the lake, due to incomplete mixing in a meromictic setting, result in lower relative brGDGT-IIa abundance. Whilst these patterns are observable at Diss Mere and Nautajärvi, we only observe a significant relationship between S_clr and brGDGT-IIa^′ at Meerfelder Maar (r=0.36; p=0.045) in terms of redox sensitive elements (S5). Therefore, at Meerfelder Maar, like Diss Mere, anoxic conditions are considered less important than at Nautajärvi for brGDGTs, which is a pattern that has emerged across dimictic and meromictic lake settings globally (Raberg et al., 2025).

To further evaluate whether inferred redox sensitivity reflect changes in water column stratification, we compare our results with GDGT-based stratification indicators (e.g., Baxter et al. 2021). Baxter et al. (2021) use f(CREN^′) (ratio of crenarchaeol / crenarchaeol isomer) at Lake Chala to define periods of upper water column stratification, where higher values indicate greater stratification. As changes in mixing regime have been identified at Nautajärvi by Lincoln et al. (2025), we test GDGT approaches as stratification indicators here. Based on Baxter et al. (2021) increased f(CREN^′) at Nautajärvi during the mid-Holocene suggests periods of enhanced lake water stratification (S9). Physical varve evidence suggests that Nautajärvi was sensitive during the mid-Holocene with periods of increased lake water stratification (greater meromixis) interspersed with periods of periods of strengthened mixing (observed through Fe precipitation and colloid formation; Lincoln et al., 2025). Indeed, rapid fluctuations between higher and lower f(CREN^′) in the mid-Holocene may align with these observations (S9). Whilst we acknowledge that interpretations from very different, albeit meromictic systems (in the mid-Holocene at Nautajärvi), may not be wholly transferable, the data appears to agree with physical varve evidence at this location reinforcing observations from Lincoln et al. (2025).

Therefore, the strength of redox influences on brGDGT distributions vary between sites. Nautajärvi appears to exhibit the greatest sensitivity to redox conditions and changes in mixing regime whereas this is weaker at Diss Mere and Meerfelder Maar. These results indicate that although anoxia and oxycline position can influence the distribution of certain 5-methyl brGDGTs, redox effects are site specific. However, the scatter observed in the correlations between brGDGTs and redox sensitive elements, together with their contrasting behaviour across lakes, suggests that other controls (e.g., temperature and environmental factors) remain important drivers of brGDGT production in these systems.

BrGDGT distributions, cyclisation of branched tetraethers (CBT^′)-derived pH reconstructions, and the isomer ratio (IR_6ME) differ across each site (Figs. 5, 6, 7). These differences are likely influenced by differences in water chemistry owing to the global relationship between CBT^′ and IR_6ME and pH (De Jonge et al., 2014; Naafs et al., 2017) and therefore may be associated with varve formation.

At Diss Mere, calcite precipitation occurs predominantly within the summer leading to the formation of calcite laminae. This is a product of alkaline lake waters being supersaturated with CaCO₃ throughout the year, with a greater degree of precipitation occurring during the summer months. Although our sampling strategy encompasses multiple varve laminations (i.e., both organic and calcite sub-layers in sub millimetre varve structures), brGDGT-derived pH values are relatively alkaline throughout the Holocene (∼8.5–9.5). We do not observe a significant correlation between CBT^′ and Ca_clr or CBT^′ and ln(Ca/Ti) at Diss Mere (S4), likely due to this sampling approach. Nonetheless, our pH reconstructions are aligned to pH readings (∼9) from epilimnetic waters (Boyall et al., 2023). At Nautajärvi the winter season is represented by clastic lamina reflective of winter precipitation and snowmelt run off whereas the summer layer is related to lake productivity and allochthonous carbon yield from the catchment via surface runoff (Lincoln et al., 2025). This is consistent with more acidic conditions within the lake, as suggested by less alkaline pH reconstructions, which is a product of the influx of acidic snowmelt during spring (e.g., Korkonen et al., 2017). However, we also observe a slight reduction in alkalinity from the early to late Holocene at Nautajärvi, which aligns with observations from other Scandinavian locations (Rosén and Hammarlund, 2007), and may be attributed to a gradual increase in input of acidic soil or peat, as is consistent with a switch to the machine-learning-defined “peat” source for late Holocene brGDGTs. We observe positive relationships between ln(Ca/Ti) (r=0.55; p=0.001), ln(Ca/K) (r=0.54; p=0.002) and CBT′/pH at Meerfelder Maar (S6), suggesting that variability in Ti_clr and K_clr impacts the 6-methyl isomerisation response (e.g., Peaple et al., 2022). Authigenic calcite only precipitates during a short period of the early Holocene in Meerfelder Maar (Martin-Puertas et al., 2017). The strong Ca_clr RDA loadings (Fig. 4) are driven by early Holocene samples only and are not, therefore, considered a driver of the Holocene GDGT response.

The relative proportion of 6-methyl brGDGTs is variable between sites. This is reflected in differences in average IR_6ME with Diss Mere and Meerfelder Maar exhibiting greater values than Nautajärvi (Figs. 5, 6, 7). We observe a significant relationship between IR_6ME and ln(Ca/Ti) at both Diss Mere and Meerfelder Maar (S4, S6), which at the latter site appears predominantly driven by the samples from within the early Holocene. This suggests that brGDGT isomerisation responds to annual CaCO₃ saturation within lake waters (Diss Mere) and changes to winter mineral flux (Meerfelder Maar) and therefore reflect site dependent autochthonous and allochthonous processes. Given the low IR_6ME from Nautajärvi we do not discuss this further here.

Additionally, changes in organic carbon have been shown to drive brGDGT responses (Hällberg et al., 2023). Whilst no organic carbon data is available for comparison, we argue that changes in organic carbon content has a limited influence on temporal trends during the Holocene due to continued, rhythmic varve sedimentation. Such conditions suggest stable depositional and productivity environments without large shifts in organic carbon (e.g., Kril et al., 2025). Where carbon content likely increases (i.e. in the non-varved sediments <2.0 ka BP at Diss Mere) organic carbon may influence brGDGTs. However, disentangling this from processes controlling eutrophication remains difficult.

We show that differences in the distribution of brGDGTs and metrics can be related to processes that operate in seasonally stratified lakes. The data suggests that anoxia, limnological and catchment processes can influence brGDGTs in these environments but that different processes impact brGDGTs at each site. However, given the moderate correlations other factors (e.g., climate) likely remain a key component (Sect. 4.4).

To ascertain whether brGDGT temperature reconstructions can be used to infer Holocene climate variability, we compare modern (<300 year) brGDGT-based temperature reconstructions to instrumental data from Diss Mere and Nautajärvi (Fig. 8). This same analysis is not possible from Meerfelder Maar owing to disrupted core top sediments.

BrGDGT-based mean temperature of months above freezing (MAF) reconstructions from Diss Mere are similar to long-term instrumental mean annual air temperature (MAAT) data, but are ∼1 to 2 °C warmer than instrumental data between 1882 and 1950 AD but converge more closely after 1950 AD. As Diss Mere is not seasonally frozen, brGDGT-derived MAF estimates approximate MAAT albeit, on average, appear slightly warmer in modern settings. At Nautajärvi, brGDGT temperature reconstructions overestimate instrumental MAAT by >5 °C when applying a global lacustrine MAF temperature calibration. The BIGMaC machine-learning algorithm suggests that brGDGTs from samples covering the last 300 years at this site could be derived from “peat” rather than within the lake itself (see Sect. 4.2) and that a lacustrine calibration in these modern samples may be inappropriate. To explore this further, we recalculate temperatures at Nautajärvi using a combined global peat and soil dataset (Dearing Crampton-Flood et al., 2020) calibrated to either MAAT or MAF. However, MAAT reconstructions yield temperatures that are consistently lower (3–5 °C) than instrumental MAAT. Instead, we resampled instrumental MAAT to MAF to only include months where the lake is sub-aerially exposed and find a stronger overlap between the brGDGT MAF temperatures and instrumental data. This suggests that brGDGTs at Nautajärvi are recording warm season temperatures and that a lake calibration is appropriate.

Whilst Nautajärvi has a very good overlap with instrumental data, Diss Mere only approximates instrumental data with occasional divergence between reconstructed and instrumental data. However, we do note that instrumental data are within reconstruction uncertainty of the GDGT calibration. Moreover, Diss Mere and Nautajärvi are different systems. Nautajärvi presently accumulates varves and is mesotrophic (Lincoln et al., 2025), whilst the upper sediments from Diss Mere are massive and the system is currently eutrophic (despite similar processes operating as during the Holocene; Boyall et al., 2023). Eutrophication at Diss Mere is a result of increased detrital input via intensified human catchment activity over the last 1 kyr (Boyall et al., 2024). This shift in trophic state is realized through changes in diatom and blue-green algae communities (Boyall et al., 2024). We suggest that coeval shifts in the bacterial community, in a eutrophic system, may help explain why the overlap between reconstructed and instrumental data is weaker than at Nautajärvi. Shifting IR_6ME and ΣIIIa / ΣIIa ratios also suggest bacterial community change. Similar patterns have been identified elsewhere (e.g. Miller et al., 2018; Zhao et al., 2021) associated with changing bacterial communities driven by eutrophication (Miller et al., 2018) and/or degradation and sub-surface GDGT production (De Jonge et al., 2020). Given Diss Mere is presently eutrophic, this explanation seems plausible here.

This analysis shows that our data approximate modern instrumental data from these two locations yet consideration needs to be given to MAAT or MAF depending on location and present trophic conditions in the lake environment.

Previous work has suggested that application of the methylation index of branched tetraethers (MBT5ME′) in lakes with abundant 6-methyl brGDGTs can yield lower- or higher-than-expected temperature estimates (Dang et al., 2018; Russell et al., 2018; Bauersachs et al., 2024). Although the causes of this discrepancy remain unclear, it is considered to reflect a non-thermal response (Novak et al., 2025). There is no significant correlation between IR_6ME and MBT5ME′ at either Meerfelder Maar or Nautajärvi, respectively (r=0.01; p=0.97, r=-0.36; p=0.053). At Diss Mere, there is a significant correlation between IR_6ME and MBT5ME′ (r=0.65; p=<0.001) across the Holocene (S4), suggesting that input of 6-methyl brGDGTs may bias temperature reconstructions (Dang et al., 2018; Martínez-Sosa et al., 2020). However, this data incorporates sparsely sampled datapoints that track low-frequency climate variability (see Sect. 5). If the data is subset to remove the influence of low-frequency variability and includes only highly sampled mid-Holocene data, the IR_6ME and MBT5ME′ relationship is much weaker (r=0.27; p=0.141).

The observed similarities with regional reconstructions (Sect. 5) and instrumental data (Sect. 4.4.1), suggest that whilst there may be a minor influence, it is unlikely that the brGDGT distribution at Diss Mere is controlled by non-thermal factors associated with IR_6ME. Our results thus emphasize that previously identified 0.5 and 0.4 IR_6ME thresholds may vary on a site-by-site basis (e.g. Bauersachs et al., 2024; Novak et al., 2025).

Diss Mere and Meerfelder Maar exhibit ∼2–3 °C warming between 10 and 8/7 ka BP, respectively. From the available data, peak warmth occurs at Diss Mere between 8 and 4 ka BP, with the warmest temperature at 8 ka BP reconstructed from a single data point in a sparsely sampled section. Peak warmth at Meerfelder Maar appears between 5.6 and 3 ka BP with the observation at 3 ka again from a single datapoint. At Diss Mere temperatures generally follow a decreasing pattern from 8 ka BP (depending on calibration). In contrast, Nautajärvi reveals only a small temperature increase from the early to mid-Holocene from 9.8 ka BP, mid-Holocene peak warmth between 6 and 4.8 ka BP and rising temperatures across the late-Holocene (Fig. 9). Whilst caution is required when interpreting sparsely sampled datapoints, our reconstructions are broadly comparable with climate and physical proxy data obtained from these records (with the exception of warm temperatures during the late Holocene from Nautajärvi and Meerfelder Maar; S10, S11, S12), and comparative reconstructions from across Europe (Fig. 9) adding greater credence to our interpretations.

The general increase in temperature from the early Holocene as observed at Diss Mere and Meerfelder Maar, and decline in the late Holocene at Diss Mere is comparable to global temperature trends across the Holocene (e.g., Fig. 9; Kaufman et al., 2020; Salonen et al., 2024) and follows patterns of precessionally driven summer insolation (Loutre et al., 2004). However, increasing temperatures at Nautajärvi in the late Holocene appears at odds with both this pattern and Arctic neoglacial cooling (e.g., McKay et al., 2018). Orme et al. (2018) identify warming sea surface temperatures from 2 ka BP from the Greenland and Norwegian Seas and the Iceland Basin, which adds to evidence of warming from the south east Norwegian Sea, warm climate anomalies in Scandinavia and a greater abundance of warm diatom taxa in the Baltic Sea (Seppä et al., 2009; Sejrup et al., 2016; Van Wirdum et al., 2019). Despite this, it is challenging to reconcile opposing patterns in Europe, especially given that Nautajärvi is trending towards warmer climates than in the mid-Holocene. Therefore, at Nautajärvi ecological changes to bacterial communities, via changes in GDGT source in the mid–late Holocene, and changes in recorded seasonality of temperature cannot be discounted.

Despite these proxy observations, the long-term evolution of temperature between regions (e.g., in the timing of peak warmth) reveal spatial differences with sites in mainland Europe showing different trends vs. the UK. We suggest that the more maritime location of Diss Mere may be more sensitive to northern North Atlantic conditions, which, alongside certain sea surface temperature records, depending on location and proxy, also reveal earlier peak warmth (e.g., Andersson et al., 2010; Berner et al., 2011; Cartapanis et al., 2022). However, why peak warmth is delayed at Meerfelder Maar, is not yet clear. At Nautajärvi mid-Holocene peak warmth is not surprising given similarities to a multi-ensemble pollen reconstruction (Salonen et al., 2024) and a brGDGT reconstruction (Otiniano et al., 2024) from northern Finland (Fig. 9). In Finland delayed peak warmth may be related to the influence of the final wastage of the Fennoscandian ice sheet (Patton et al., 2017) and associated disruption of ocean circulation patterns (Cartapanis et al., 2022). Globally, at latitudes of >60° N, continental sites display peak warmth either prior to 10 ka BP or between 8 and 4 ka BP (Cartapanis et al., 2022). At 10 ka BP, Nautajärvi was a component of Lake Ancylus and connected to the Baltic Sea basin (Ojala et al., 2005) and therefore was not a competent lacustrine archive at this time. Therefore, regional temporal differences in temperature evolution appears variable between more maritime locations and mainland Europe.

Spatial differences in reconstructed temperatures also appear in the mid Holocene. Previous west-European GDGT-based reconstructions reveal a general increase in temperatures across the mid-Holocene (e.g., Van den Bos et al., 2018; d'Oliveira et al., 2023; Fig. 9). Whilst reconstructions from a site in eastern Europe demonstrates cooling trends (e.g., Ramos-Román et al., 2022). Further reconstruction complexity is added with a western European site displaying warming in the mid-Holocene (Martin et al., 2020). However, this has been suggested to relate to differences in reconstructed temperature seasonality alongside localised site differences, for example in environmental contexts and altitude (d'Oliveira et al., 2023). Nonetheless, the trends in temperature reconstructions from both Diss Mere and Meerfelder Maar are comparable to European Holocene brGDGT-reconstructions (Fig. 9), albeit with local and/or spatial influences (e.g., site elevation, continental position) seemingly being important in the direction of temperature change during the mid-Holocene.

By exploiting varved lake sediments, we show that during the mid-Holocene, Diss Mere and Meerfelder Maar median reconstructions exhibit ∼2–3 °C of temperature variability at multi-decadal scales. Whilst this ∼2–3 °C temperature variability may be a product of calibration uncertainty, comparative climate reconstructions reveal similar patterns. Pollen and brGDGT-based reconstructions from the European mid-latitudes reveal a similar range of variability (e.g., Seppä et al., 2009; Martin et al., 2020); albeit with some records displaying increased variability (∼4–5 °C) at multi-decadal (Ramos-Román et al., 2022) and centennial scales (d'Oliveira et al., 2023) (Fig. 9). At Nautajärvi, temperatures are stable during the mid-Holocene (Fig. 9; median variability of 1 °C). This is comparable to the brGDGT reconstruction from northern Finland between 7 and 4 ka BP (Otiniano et al., 2024). However, this is incongruent with non-brGDGT based temperature observations from Scandinavia (e.g., Shala et al., 2017; Salonen et al., 2024) where greater temperature variability (2–3 °C) is observed. Our reconstructions reveal a mostly comparable range of variability to other brGDGT reconstructions and suggest that higher latitudinal locations exhibit less variance than lower latitude locations. However, as reconstruction amplitude can be different between GDGT and traditional proxy reconstructions (pollen and/or chironomid) it is likely that some of the controls discussed above may be important in these lake settings (Sect. 4.3), or the seasonality of temperature recorded varies by proxy and by latitudinal setting.