Core GeoB23302-2 was recovered from the Celtic Margin, off the English Channel (47∘26.61′ N, 8∘28.67′ W; 2167 m water depth) (Fig. ), with the help of a gravity corer during cruise MSM 79 of the research vessel Maria S. Merian. The core location is in close proximity to the site where core MD95 2002, which has been studied in previous publications e.g., was retrieved (47∘27′ N, 8∘32′ W) (see Fig. ). The chronology of our 700 cm core was established based on seven radiocarbon accelerator mass spectrometry (14C-AMS) measurements of planktic foraminifera (G. bulloides and N. pachyderma) picked at specific depths (Table ). The preparation and measurement of these samples followed well-established protocols routinely run at the MICADAS 14C laboratory of the Alfred Wegener Institute (AWI) . The 14C ages were uploaded to the OxCal software version 4.4.2 (), and assuming that the deposition is a Poisson process, the P Sequence model was employed for the construction of an age–depth model for core GeoB23302-2 (). This deposition model (Fig. S2 in the Supplement) uses the global marine calibration curve Marine20 () and a local marine reservoir correction ΔR of 94 ± 45 14C yr . It is important to note that while Marine20 incorporates larger marine reservoir age estimates for the last glacial period than Marine13, these estimates are considered more realistic due to methodological improvements in the former (). A general outlier analysis was employed to account for possible outliers within the chronological model (). The code is available in the Supplement accompanying this paper.

The XRF characterization of core GeoB23302-2 was performed using the XRF Core Scanner II (AVAATECH serial no. 2) at the Center for Marine Environmental Sciences (MARUM), University of Bremen, Germany. Measurements were performed at 1 cm intervals for the upper 3.5 m of the core and at every 2 cm for the remaining section. The scan resolution was set to 1 cm with two running rounds, during which the elements were detected with 10 and 30 kV of tube voltage. In order to account for the closed sum effects of water content, grain size and OM amount (e.g. ), we report the elemental ratios Zr / Rb and Fe / Ca.

Sediment samples taken at 10 cm intervals from core GeoB23302-2 were freeze-dried and homogenized. For each depth, approximately 3 g of sediment was subsampled and underwent ultrasonic extraction with a mixture of dichloromethane : methanol 9:1 (v:v). This step was repeated three times, and the total lipid extracts obtained were then saponified with 0.1 M potassium hydroxide (KOH) in methanol : water 9:1 at 80 ∘C for 2 h. This procedure resulted in the separation of the neutral lipids and n-alkanoic acids fractions, which were subsequently extracted using n-hexane and dichloromethane (at pH 1), respectively. Next, silica gel chromatography was employed to further split the neutral lipids via elution with n-hexane and dichloromethane : methanol 1:1 (v:v), yielding the n-alkane and GDGT subfractions, respectively. The n-alkane concentrations were measured via gas chromatography (GC) using a 7890A GC (Agilent Technologies) equipped with a flame ionization detector (FID) and DB-5MS fused silica capillary columns (60 m, ID 250 µm, 0.25 µm film coupled to a 5 m, ID 530 µm deactivated fused silica precolumn). Retention times and the comparison with an n-alkane standard were used for the identification of different compounds, whereas quantifications were achieved through the use of an internal standard (squalane) added to the sample prior to extraction. We calculated n-alkane-derived indices, namely the CPIalk e.g., 1CPIalk=12⋅C25+C27+C29+C31+C33C24+C26+C28+C30+C32+C25+C27+C29+C31+C33C26+C28+C30+C32+C34, and the Paq , 2Paq=C23+C25C23+C25+C29+C31. Hopanes were analysed via GC coupled with time-of-flight mass spectrometry (GC-TOF-MS), and such a system consisted of a LECO Pegasus III (LECO Corp., St. Joseph, MI) interfaced to an Agilent 6890 GC, which was equipped with a temperature-programmable cooled injection system (CIS4, Gerstel). The measurements were performed using the instrumental set-up described in , and identification was achieved through the relative retention times and mass spectra. The sum of m/z 191 and 205 was used for the quantification of homohopane isomers (C31), namely the 17β,21β (H), 22R homohopane; the 17β,21α (H), 22R + 17β,21α (H), 22S homohopanes; the 17α,21β (H), 22R homohopane; and the 17α,21β (H), 22S homohopane. Next, the fββ was calculated : 3fββ=C31ββRC31ββR+C31αβS+C31αβR+C31βαS+C31βαR. The analyses of branched and isoprenoid GDGTs by high-performance liquid chromatography (HPLC) were performed on an Agilent 1200 series HPLC system coupled to an Agilent 6120 single-quadrupole MS via an atmospheric pressure chemical ionization interface (APCI), broadly following the method described in . The chromatographic separation of individual GDGTs was achieved via the use of two ultra-performance liquid chromatography (UPLC) silica columns in series (Waters Acquity BEH HILIC, 2.1 mm × 150 mm, 1.7 µm and a 2.1 mm × 5 mm pre-column of the same material) maintained at 30 ∘C. Positive-ion APCI-MS and selective ion monitoring (SIM) of (M + H)+ ions () or ion-source fragmentation products of OH-GDGTs () allowed the identification of GDGTs. Quantification was performed with the use of an internal standard (C46-GDGT) added prior to extraction. For this research, we calculated the BIT index : 4BIT=I+II+IIII+II+III+cren, where the roman numerals refer to specific GDGTs characteristic of terrestrial bacteria, and cren stands for crenarchaeol, which is derived from marine planktonic Thaumarchaeota.

Soxhlet extraction was employed for the compound-specific 14C dating of high-molecular-weight n-alkanoic acids. For that purpose, approximately 100 g of freeze-dried and homogenized sediment taken from selected depths in core GeoB23302-2 was extracted for 48 h using a mixture of dichloromethane : methanol 9:1 (v:v). Total lipid extracts were saponified with 0.1 M KOH in methanol : water 9:1 at 80 ∘C for 2 h, and the n-alkanoic acids were recovered from the saponified solution using n-hexane at pH 1. Next, n-alkanoic acids were methylated at 80 ∘C overnight in a nitrogen atmosphere with HCl and methanol of known 14C signature to yield the fatty acid methyl esters (FAMEs) that were later extracted with n-hexane. Silica gel chromatography was employed to separate FAMEs from polar compounds. The n-C26:0, n-C28:0 and n-C30:0 alkanoic acids underwent purification via preparative capillary GC (PC-GC; ) on an Agilent HP6890N GC with a Gerstel cooled injection system (CIS) connected to a Gerstel preparative fraction collector . A Restek Rxi-1ms fused silica capillary column (30 m, 0.53 mm diameter, 1.5 µm film thickness) equipped the GC. Injection was performed stepwise with 5 µL per injection, and, at the end of the process, the purity of the FAMEs was checked by analysing aliquots of the samples via GC-FID. The purified FAMEs were transferred to tin capsules (25 µL volume, ELEMENTAR) using dichloromethane, dried on a hot plate at 40 ∘C and packed. An Elementar vario ISOTOPE EA (elemental analyser) was used for the combustion of the samples, generating CO2 with carbon isotopic ratios directly determined by the connected MICADAS system. Reference standards (oxalic acid II, SRM 4990C) and 14C-free CO2 gas had their 14C content measured together with the samples. The BATS software () was used for blank corrections and standard normalization, and the final results are reported as fraction modern carbon (Fm).

The preparation procedures for CSRA introduce exogenous C, i.e. contaminants, to samples. The degree of contamination varies according to the methods employed, and, in our case, processes such as column bleed and carryover during prep-GC compound isolation may contribute to this. For this reason, assessing the Fm and the size of the blank (Fmblank and mblank, respectively) is essential for accurate results. Here, in-house reference samples of 14C-free Messel shale (Fm=0) and modern apple peel (Fm=1.029±0.001) underwent the same pre-treatment as samples of unknown age, and their results were used for blank correction following the method outlined in . Isotopic mass balance was employed in order to make a correction for the methyl group added during the derivatization of the samples. Uncertainties were fully propagated.

The Δ14C values of the n-alkanoic acids analysed here were corrected for radioactive decay between 1950 and 2021, which is the year of measurement. These values were then used to calculate the Δ14C values at the time of deposition: 5Δ14Cinitial=Δ14C1000+1⋅eλt-1⋅1000, where λ is a decay constant (1/8267 yr-1), and t is the time of deposition. The Δ14C values of the atmosphere contemporaneous with the compounds (Δ14Catm) were obtained from comparison with the IntCal20 dataset () using the age ranges given by the deposition model for the respective sediment layers. Finally, pre-depositional 14C ages for the n-alkanoic acids were given by 6A=-8033⋅ln⁡1+Δ14Cinitial/10001+Δ14Catm/1000. These calculations follow the method outlined in and later in , where more details can be found.

Carbon stable isotope (δ13C) analyses were carried out on acidified samples (Ag capsules, HCl, 1.5 M) in order to remove the inorganic C (). Analyses were performed using a Thermo Scientific DELTA Q isotope ratio mass spectrometer coupled to a Thermo Scientific FLASH 2000 CHNS/O analyser via Conflo III at the Stable Isotope Laboratory of ISP-CNR. δ13C data are expressed in the conventional delta notation (‰). Isotopic data were calibrated using the IAEA reference material IAEA-CH7 polyethylene (-32.15 ‰ vs. Vienna Pee Dee Belemnite (VPDB)). Throughout the runs, we used other standards with a sediment matrix routinely used in the laboratory to check the reproducibility of measurements. The standard deviation for δ13C measurements was lower than ±0.1‰ based on replicates of sediment standards.

Our results corroborate the previous findings of , showing a notable increase in the influx of terrigenous OM in the Bay of Biscay during the last deglacial period. Between ca. 20.6 and 15 cal kyr BP, the values of the Zr / Rb ratio reflect the deposition of coarse-grained sediments at the core location, which may be associated with enhanced fluvial activity . A period of relatively high Fe / Ca values (ca. 21–16.4 cal kyr BP) is indicative of a greater influx of sediment from land (Fig. c). This is consistent with a period of elevated terrestrial contribution to the bulk OM present in the sediment, from approximately 20.5 to 16.5 cal kyr BP, except for a sudden decrease at approximately 19 cal kyr BP, as revealed by the BIT index see e.g. (Fig. f). These Fe / Ca and BIT patterns are also recorded in core MD95 2002 see Figs. S3 and S4; and a similar pattern of marked Fe / Ca peaks, sometimes associated with peaks in OM content, during Heinrich events and much lower values throughout the Holocene has been observed at other sites .

Beyond identifying the presence of terrestrial OM transported to the Bay of Biscay via the Channel River during the LGM–Holocene transition, the comprehensive analysis of elemental, geochemical, and isotopic proxies presented here provides insights into the potential sources of this terrigenous OM. Values for the Paq proxy point to a major contribution of OM from aquatic plants between approximately 20.2 and 17.2 cal kyr BP, suggesting the presence of OM sourced from wetland vegetation (Fig. d). Our CPIalk record reflects the degree of degradation the sedimentary OM has undergone in its previous terrestrial reservoir or during transportation cf.. During the peak of terrigenous deposition, the signal of more mature OM fluvially transported to the continental slope is detected in our CPIalk and fββ records, which reach relatively low values when compared to the Holocene (Fig. e). The presence of petrogenic, i.e. thermally mature, material is another factor to consider when interpreting these records. Nonetheless, throughout the archive, the CPIalk values remain above the diagnostic value of petrogenic material (ca. 1) reported by . Additionally, in contrast with the results of for the Bering Sea, in the present study the fββ proxy is not indicative of petrogenic material in the sediment (see the Supplement). This is because the values do not decrease in response to diagenetic transformations but rather due to enhanced inputs of the C31αβR hopane, which is abundant in peat .

The compound-specific 14C results disclose that the analysed terrigenous biomarkers are ancient and pre-aged, indicating that they are not contemporaneous with sediment deposition but rather older (Fig. f). Despite the large variability in OM residence time in permafrost soils, the pre-depositional ages of some of the compounds present in core GeoB23302-2 (up to 25 000 14C yr) are considerably greater than those previously attributed to permafrost-derived OM at other sites and at different timescales e.g. up to ca. 10 000 14C yr;. Although petrogenic contributions are commonly thought to be devoid of n-alkanoic acids, this assumption does not always hold , and the erosion of organic-rich sedimentary rocks can supply fossil OM to the ocean, thereby depleting the 14C compound-specific signal in the sediment e.g.and references therein. However, the OM present in core GeoB23302-2 presents δ13C values in the range from -25.9 ‰ to -22.3 ‰, heavier than the average value of δ13C = -29.4 ‰ displayed by organic-rich rocks in the region e.g.. These values are rather comparable to those observed in peat, corroborating the results of the CPIalk and fββ proxies and pointing to a pre-aged immature source such as ancient peat. The pre-depositional ages observed during the peak of OM deposition likely result from a 14C reservoir effect, which is caused by radioactive decay and a lack of exchange with the atmosphere . This effect occurs as the OM is stored in its source reservoir before eventual erosion and transportation to the core location. Mechanisms such as deposition–resuspension loops on the continental slope could be invoked to explain pre-aged n-alkanoic acids during the most recent part of our record and references therein. To summarize, our results strongly support previous findings describing a massive remobilization of terrestrial C from the European continent to the North Atlantic Ocean during the last deglaciation, with notable peaks at ca. 19.5 and between approximately 19 and 17 cal kyr BP . In addition, the set of proxies applied in the present study allows us to go further and suggest peat-derived material as a major source of OM to the Bay of Biscay during the last deglaciation.

Wetlands are dynamic ecosystems that fix CO2 from the atmosphere, store C and contribute to the C cycle through various processes, including the decomposition of OM that releases CO2 . Therefore, the establishment of wetlands in the study region towards the end of the LGM and during the last deglaciation (Fig. d), combined with permafrost distribution data that imply gradual permafrost decomposition e.g. and records of atmospheric CO2 concentration , suggests the need to investigate thawing permafrost as a possible source of OM to the deposition site. Although some of the peatlands formed during the Eemian period were buried by glaciers and mineral sediments, Weichselian ice sheets were not as extensive as the ones from the Saalian, and some deposits remained uncovered e.g. in northern Germany, on the cliff of the Elbe river;. Peat deposits formed during the last interglacial occur widely across the studied region , from Belgium e.g. and the Netherlands e.g. to Poland e.g., through Germany e.g. and Denmark e.g.. Although the environmental conditions of the last glaciation were unfavourable for the development of peatlands, factors such as the formation of permafrost in Europe resulted in the long-term preservation of OM from older periods (e.g. frozen peat OM) (). However, during the last deglaciation, the thawing of permafrost and the presence of meltwater streams may have contributed to the erosion of these peats. We propose that Eemian peatlands represent the primary source of fossil biomarkers transported to the Bay of Biscay, with processes such as thermal and physical erosion of these deposits see e.g. leading to pre-aged material reaching the final burial site. At this point, it is crucial to emphasize that, in the context of this study, permafrost plays a more significant role as a storage mechanism for peatland OM rather than serving as a unique source of OM by itself.

In north-western and central Europe, continuous permafrost prevailed at approximately 27–17 cal kyr BP , with the deposits likely degrading and shrinking due to the warming observed at the end of the LGM. For instance, in the Netherlands, there has been evidence of widespread permafrost degradation between 22 and 21 cal kyr BP , with continuous permafrost in the Dutch coversand region completely disappearing after 20 cal kyr BP . This episode is an example of permafrost thawing that may have contributed to wetland development between ca. 21 and 16 cal kyr BP as recorded in our Paq record (Fig. d). Between 17 and 15 cal kyr BP, permafrost zones in the study region were restricted to areas near the retreating ice sheets and references therein. Afterwards, apart from a short later period of discontinuous permafrost (ca. 10.9–10.5 cal kyr BP) that has also been reported for north-western and central Europe , there is evidence of the presence of permafrost in this region during the Younger Dryas e.g..

The erosion of European permafrost and peatland deposits by glacial meltwater is a mechanism likely to have exported OM to the ocean. The decay of the European ice sheets and glaciers at the onset of the last deglaciation e.g. contributed to strengthening the Channel River discharge into the Bay of Biscay , and it has been proposed that this was the main mechanism controlling the river activity during this time period . Deglacial pulses of meltwater emanating from the BIIS (at 22 and 18.6 cal kyr BP) and the FIS (starting around 19 cal kyr BP) were routed to the North Atlantic via the Channel River. In this way, the activity of the river responded to changes in the European ice masses, being particularly influenced by the dynamics of the FIS .

Although subglacial meltwater can flow through permeable sediments as groundwater, the presence of frozen ground with reduced hydraulic transmissivity, i.e. permafrost, hinders this process . This leads to trapped pressurized water accumulating underneath ice sheets and eventually draining during catastrophic events. As the climate warmed, and the ice sheets retreated, and/or permafrost decayed, bursts of subglacial meltwater were released, carving glacial features known as tunnel valleys in the ground and discharging large amounts of eroded material into rivers (; and references therein). Subglacial channels from major Pleistocene glaciations are still present today in Europe and serve as evidence of this phenomenon e.g.. Meltwater streams from the FIS discharged through the Elbe river and provoked several flooding events of the Channel River , with remarkable episodes (R2–R5) recorded as peaks in the ratios Ti / Ca and Fe / Ca of both the GeoB23302-2 and the MD95 2002 archives (Fig. S3). Floods R2–R5 were most likely associated with enhanced processes of erosion and sediment export in the catchment see e.g., and the intensified freshwater influx, resulting from riverine discharge due to a mixture of precipitation and glacial meltwaters, is reflected in the terrestrially sourced OM signal shown by our BIT index record, which corroborates that reported for core MD95 2002 (Fig. S4). Furthermore, the process of post-glacial sea-level rise may have played a role in the erosion of coastal permafrost deposits, potentially serving as an additional pathway for the transport of OM to the ocean e.g..

The remobilization of OM from land to ocean is largely mediated by rivers, with factors such as precipitation and temperature being major regulators of fluvial C fluxes and references therein. Between 21 and 17 cal kyr BP, a temperature rise (Fig. a) observed in various Atlantic environmental records e.g. marked a transition from cold and dry to warm and wet conditions in continental Europe. For example, a gradual increase in the north-eastern Atlantic sea surface temperature (SST) starting at about 25 cal kyr BP and preceding the peak of terrigenous deposition is recorded in core SU8118 (37∘46′ N, 10∘11′ W) . This same warming trend can be observed on land, reflected in the δ13C signature of a speleothem record from western Europe (45∘30′ N, 0∘50′ E) and starting roughly at the same time (Fig. ). The speleothem time series is not high-resolution, and long-term trends may be in fact punctuated with short-term oscillations. In any case, it is important to acknowledge that, although speleothem δ13C values can potentially serve as a proxy for temperature see e.g.and references therein, the correlation must be interpreted with caution due to several other factors influencing C isotopic ratios in these archives . Enhanced precipitation in response to warming led to increases in fluvial runoff and, due to widespread permafrost, increased erosion and transport of sediment from land to the Channel River outlet . After approximately 18 cal kyr BP, as the climate warmed, the area occupied by peatlands in Europe increased . This is in agreement with our Paq index record, which indicates the re-establishment of formerly frozen wetlands, potentially including peat-rich environments (Fig. d).

Considering that the OM buried in marine sediment is only a relatively small part of the total OM entering rivers, which is predominantly returned to the atmosphere as CO2 e.g., the OM export to the Bay of Biscay via the Channel River is likely to have been accompanied by the transfer of CO2 and CH4 to the atmosphere e.g.. It follows that our comprehensive analysis, encompassing biomarkers, elemental proxies and radiocarbon dating, consistently corroborates the hypothesis of permafrost thawing in the Northern Hemisphere contributing to the observed perturbations in the atmospheric C reservoir . This essentially means that north-western and central Europe too, similar to other permafrost sites , may have contributed to the deglacial rise in atmospheric CO2 . However, it is likely that the deglacial loss of European permafrost was offset by the subsequent accumulation of significant amounts of C in permafrost-free soils and peatlands (). Our elevated pre-depositional ages at ca. 17.5 cal kyr BP (up to ca. 15 14C kyr) may partly explain the steep drop in the 14C signature of atmospheric CO2 during the period known as the Mystery Interval (17.5–14.5 kyr BP) see e.g.. In other words, this result implies that thawing European permafrost combined with the deep-ocean reservoir contributed 14C-depleted CO2 to the atmosphere during this period. However, our results show that the remobilization of C from this terrestrial pool started as early as ca. 20.2 cal kyr BP, which considerably precedes estimates of large permafrost contributions to atmospheric CO2 between 17.5 and 15 kyr BP . This suggests the need to investigate leads and lags in the permafrost carbon feedback.

After approximately 18 cal kyr BP, a rerouting of the Elbe–Weser system meant that FIS meltwater carrying ancient C was being delivered to the Norwegian Channel and references therein. After 17 cal kyr BP, sea-level rise caused a shift in the shoreline, with the Bay of Biscay no longer being suitable to record terrestrial runoff during the Holocene . This is reflected in the sudden drop observed in the BIT index record (Fig. f). Notably, although our data support what has been previously inferred for the study region , the distinct timing for the discharge peak observed in this study compared to other sites may imply different mechanisms of C remobilization, and these need to be further investigated. Indeed, factors such as the local hydrology and vegetation have been shown to play a role in the accumulation and degradation pathways of permafrost-influenced peatlands and references therein.

Peat-forming wetlands remain an important source of terrestrial OM to the ocean and of CH4 and CO2 to the atmosphere, with flux rates likely increasing due to current warming . In high-northern-latitude wetlands, it has been shown that permafrost degradation leads to wetland shrinkage . In the tropics the situation is also critical, with anthropogenic and natural factors contributing to the remobilization of pre-aged C from peatlands. Therefore, the release of large amounts of peat-derived OM described here for deglacial Europe has analogues in the present day and may be useful to inform future projections of permafrost peatland loss e.g., with our results advocating for the importance of better constraining the C cycle in wetlands.