Fine-Tuning of Sub-Annual Resolution Spectral Index Time Series from Eifel Maar Sediments, Western Germany, to the NGRIP δ¹⁸O Chronology, 26–60 ka

Abstract

Recent technological advancements in spectral imaging core-scanning techniques have proved to be a promising tool to study lake sediments at extremely high resolution. We used this novel analytical approach to scan core AU3 of the Pleistocene Auel maar, Western Germany. The resulting ultra-high-resolution RABD₆₇₀ spectral index, a proxy for the lake’s primary production, shows an almost complete succession of Greenland Interstadials of the NGRIP ice core chronology back to around 60,000 years. Using the ELSA-20 chronology and its anchor points to the NGRIP record as a stratigraphic basis, we were able to compare and fine-tune prominent climate signals occurring in both regions. This in-depth correlation yields strong evidence that the climates of Greenland and Central Europe were not only strongly coupled on timescales of stadials and interstadials but even on multidecadal scales, showing prominent climate cycles between 20 and 125 years. As climate changes in these regions were ultimately driven by variations in the North Atlantic meridional heat transport, their strong coupling becomes most apparent during cold and arid intervals. In contrast, longer-lasting warmer and more humid phases caused the activation of various regional feedback mechanisms (e.g., soil formation, forest growth), resulting in more complex patterns in the proxy records.

1. Introduction

The open maar lakes and infilled maar structures of the West Eifel Volcanic Field, Western Germany, present an exceptional archive to study Holocene and Pleistocene climate variations due to their unique geological setting and their location in a homogeneous climate zone stretching across continental Central Europe [1]. The Eifel repeatedly experienced periods of increased volcanic activity, leading to the formation of numerous maar lakes in the region during the last 500,000 years (e.g., [2,3,4,5]). A typical feature of these lacustrine basins of volcanic origin are annually laminated varves (e.g., [6,7]), which form as a result of calm sedimentation environments and long water residence times with anoxic bottom waters. These favorable depositional and preservation conditions allow for highly detailed insight into past environmental and ecological variability (e.g., [8,9,10,11]). The ELSA (Eifel Laminated Sediment Archive) project at the University of Mainz, Germany has drilled and studied many of the 68 open and infilled maar structures of the West Eifel Volcanic Field (e.g., [12,13,14,15,16]). Most recently, the ELSA-20 C_org(chlorins) climate proxy record [8] revealed a strong forcing mechanism by changes in the Atlantic Meridional Overturning Circulation (AMOC) and North Atlantic sea-ice coverage (e.g., [17,18,19]) on the frequency of multidecadal climate oscillations in Central Europe, which lasted throughout large parts of the last glacial cycle.

Organic carbon content in lake sediments can serve as a first-order proxy for primary production and biomass accumulation at a lake’s bottom, which is highly influenced by the prevailing climate conditions (e.g., [20,21,22]). Reflectance spectroscopy methods utilize the typical absorption bands of sedimentary pigments to detect changes in the primary production in both marine and lacustrine sediment deposits (e.g., [23,24,25]). Unlike conventional core sampling methods (e.g., geochemical analyses, pollen sampling), spectroscopy techniques not only allow for sampling of lake sediment records without using or destroying the core material itself, but they also significantly increase the spatial and thus temporal sampling resolution down to 40–300 μm, making it possible to obtain data with sub-annual resolution. However, recent technological advancements in high-resolution spectroscopic imaging techniques have opened up a whole new range of analytical possibilities [26]. In contrast to other point-based reflectance spectroscopy techniques, hyperspectral imaging (HSI) methods, which use optical cameras to scan the core sample at wavelengths between 400 and 1000 nm (visible to near infrared) and between 1000 and 2500 nm (short wave infrared), can provide important three-dimensional information (two dimensions plus intensity) about a sample’s properties, including mineral composition, organic matter content, sedimentary pigments, and sedimentary structures (e.g., [27,28,29,30,31]).

Here, we present an ultra-high-resolution RABD₆₇₀ (relative absorption band depth at 670 nm) spectral index record of the Pleistocene Auel infilled maar spanning from 26,000 to 60,000 yr b2k (years before the year 2000). This record utilizes the ELSA-20 stratigraphy [8], which is based on multiple dating methods, including tie-points to the Greenland Interstadial/Stadial (GI/GS) succession of the NGRIP oxygen isotope ice core chronology [32,33,34]. We then compared and fine-tuned the RABD₆₇₀ data to the NGRIP record within interstadial and stadial periods in order to demonstrate the analytical potential of this novel spectroscopy method.

2. Materials and Methods

2.1. Coring Site

The Eifel region, with its many maar lakes and infilled maar basins, is among the few locations in Europe that allow for the study of lacustrine ecosystems beyond the Holocene [35,36]. The Pleistocene Auel infilled maar lake, which is one of the largest maar structures in the region (diameter of 1325 m), is located in the northwestern part of the West Eifel Volcanic Field, Western Germany (Figure 1A,B). Due to the large catchment area of the Auel maar (12,187 km²) and resulting abundant fluvial input (Figure 1C), it has an average sedimentation rate of around 2.3 mm/year. Drill core AU3, which was used for this study, was recovered with a ‘Seilkern’ coring device and has a length of 102 m. Several prominent tephra layers are visible in the sediment deposits of the Auel infilled maar (Figure 2), i.e., the Laacher See Tephra (LST—13,056 yr b2k) [37], the Eltville Tephra (EVT—24,720 yr b2k), the Wartgesberg Tephra (WBT—28,100 yr b2k), the Unknown Tephra 1 (UT1—30,300 yr b2k), the Dreiser Weiher Tephra (DWT—40,370 yr b2k), and the Meerfelder Maar Tephra (MMT—47,340 yr b2k). The Auel Maar Tephra (AUT—59,130 yr b2k), the maar’s own eruption sequence, is stratigraphically positioned shortly before the base of the AU3 sediments. These volcanic marker layers serve as anchor points to correlate core AU3 to various other Holocene and Pleistocene maar lakes in the region [3,16,38]. The AU3 sediment record contains a complete GI/GS succession [32,33,34] back to GI17 (core base at 59,110 yr b2k) and was included in the near-annual ELSA-20 climate proxy time series [8].

2.2. Hyperspectral Imaging (HSI) and Pigment Calibration

Hyperspectral images of the AU3 core were acquired at the University of Bern with a Specim PFD-CL-65-V10E line scan camera using the method developed by Butz et al. [25]. Segments from 30 to 102 m depth were scanned. The core faces were cleaned and flattened using a blade immediately prior to scanning. The scanning system uses a push-broom method to acquire images in which a split core in a motorized tray passes under an illumination source and hyperspectral camera. Images are created by stacking single lines of pixels to create a complete image of the core. Light reflectance in the range of 400–1000 nm is measured by the camera sensor at a spectral resolution of 2.8 nm, which is then sampled at 0.78 nm and binned to 1.6 nm resolution for analysis. The raw reflectance data were normalized to a 0 to 1 scale, where 0 represents a dark reference (camera shutter closed), and 1 is a white reference (barium sulfate plate). The following settings were used for all scans: field of view = 109.33 mm, resolution (pixel size) = 0.0833 mm, exposure = 150 ms, tray speed = 0.5 mm/s, and frame rate = 6 Hz. Cracked and disturbed areas of the core were removed from the images prior to spectral index calculations. First, a threshold based on reflectance values was applied, whereby pixels with less than 0.05 reflectance in the range of 550–700 nm were removed. Second, cracks and disturbed areas that were not captured by this automated process were removed by manually masking unwanted areas. RABD₆₇₀ was calculated using the same trough endpoints and minimum as the ISRS formula of Sirocko et al. [8]. Each data point in the spectral index record represents the row-average of pixels within a 2 mm wide subset of the resulting RABD₆₇₀ raster.

Pigments were extracted from 29 samples; each sample was taken with a cube-shaped sampling device that enables sampling of 1 cm³ of sediment from a 1 cm² square on the core face. Aliquots weighing ca. 0.5 g of homogenized freeze-dried sediments were extracted using 100% acetone following a modified version of the method of Amann et al. [39]. Pigment extracts were analyzed using a spectrophotometer (Shimadzu UV-1800, Shimadzu, Kyoto, Japan). No bacteriopheopytin was detected in the pigment extracts. Total chloropigments-a (TChl-a) were determined using the extinction coefficient of Jeffrey and Humphrey [40]. To establish a calibration formula to convert RABD₆₇₀ values to TChl-a concentrations, a linear regression was calculated between measured TChl-a and the average RABD₆₇₀ value over the sampled area (Figure 3).

3. Results

The RABD₆₇₀ index is significantly correlated with concentrations of TChl-a measured in the 29 samples (R² = 0.8, p-value < 0.001; Figure 3), allowing for an interpretation of the index as a proxy of lake primary production. The linear regression calibration model indicates that TChl-a concentrations reach 0 at approximately RABD₆₇₀ = 1.01, and pigments could not be quantified with the UV–Vis spectrophotometer in samples with RABD₆₇₀ < 1.01 (Figure 3 and Figure 4). Therefore, intervals with values below 1.01 should be interpreted with caution. However, all visible interstadials are characterized by RABD₆₇₀ indices well above this quantification threshold, which allows for an in-depth comparison of the AU3 RABD₆₇₀ and NGRIP δ¹⁸O records. We chose five interstadials (Figure 5, Figure 6 and Figure 7) and four stadials (Figure 8 and Figure 9) based on data completeness within selected intervals and their representation of the different climate stages during the last 60,000 years. Due to the different sampling resolutions of the presented records, varying sedimentation rates within the records, and the different lengths of the intervals shown in Figure 4, Figure 5, Figure 6, Figure 7, Figure 8 and Figure 9, NGRIP oxygen isotope data and AU3 RABD₆₇₀ and C_org(chlorins) records were smoothed using different running averages in the various figures. This was done visually for each figure and proxy record to reduce noise as much as possible but still keep as much detail visible as possible and thus make the datasets comparable.

The ELSA-20 stratigraphy [8] and its tie-points to the GI succession (onsets and ends) of the NGRIP ice core chronology [32,33,34] were used as a basis for the constructed age model (Figure 4). The overall structure of the RABD₆₇₀ spectral index matches very well with the ELSA-20 stadial/interstadial cycles, as both records reflect the primary production inside the Auel maar lake. Thus, all interstadial periods, except for GI7, are clearly visible in the RABD₆₇₀ spectral index back to the eruption of the Auel infilled maar at around 59,130 yr b2k (Figure 4B). GI7 is also the least prominent interstadial in the ELSA-20 record. Accordingly, interstadial onsets and ends serve as anchor points for a further, more detailed fine-tuning of prominent climate signals within interstadial and stadial periods present in both the Eifel and Greenland data.

Out of the selected interstadial periods, the most striking similarities between the two regions (Greenland and the Eifel) can be observed during GI3, GI4, and GI6 (Figure 5 and Figure 6; Table S1). Especially during GI6, the two climate proxy records almost perfectly match, showing a clear multidecadal periodicity (between 27 and 48 years), and even minor climate trends recorded in the Greenland ice cores are observable in the RABD₆₇₀ data during most of the interstadial (Figure 6A). GI3 and GI4 are characterized by very similar overall structures in both records. Prominent trends in the Greenland δ¹⁸O are also present in the AU3 data, e.g., the tripartite structure of GI3 (Figure 5A) and the plateau-like shape of GI4 with the four pronounced maxima in the second half of the interstadial (Figure 5B). Within GI3 and GI4, prominent climate cycles between 55 and 71 years and between 20 and 40 years, respectively, are visible in the RABD₆₇₀ data. However, the two records do not match as perfectly as during GI6, with some single peaks or minor trends within GI3 and GI4 only being present in one of the records.

Such prominent similarities between the two regions, however, cannot be observed within every interstadial of the last glacial cycle. The onset of GI8, for example, is characterized by a very sharp and rapid increase in the ice core data, reaching the interstadial’s maximum δ¹⁸O values in the decades after its onset. The AU3 RABD₆₇₀ record, on the other hand, shows a much more moderate increase, with a periodicity between 40 and 80 years and maximum values towards the middle of GI8 (Figure 7A). Although a few prominent climate signals can be correlated during GI8 (Table S2), the general shapes of the two records within the interstadial differ significantly. During GI15.2, the overall structures of the NGRIP and AU3 data are both characterized by a plateau-like shape, and some prominent climate signals can be observed and linked in both records (Figure 7B, Table S2). However, the δ¹⁸O and RABD₆₇₀ do not match as perfectly as during GI3, GI4, or GI6, with some maxima and minima being much more pronounced in the AU3 spectral index (periodicity between 24 and 44 years) or not present in the oxygen isotope data.

Although most of the stadial phases are characterized by spectral indices below the quantification limit, stadials within the early Marine Isotope Stage (MIS) 3 warm phase [7,41,42] show values well above the threshold of 1.01 (Figure 4 and Figure 8). This allows for an in-depth analysis of the two regions’ climate proxy records not only within interstadials, but also within stadial periods. During GS16.1, the Greenland and Eifel data follow a quite similar periodic trend (between 65 and 125 years), even though most of the maxima in the two records differ significantly in duration or amplitude (Figure 8A). Furthermore, a few of the even minor structural characteristics of the δ¹⁸O and RABD₆₇₀ records almost match perfectly, e.g., the small trough within a double peak at around 56,025 yr b2k. During GS17.1, the NGRIP ice core data are again characterized by prominent multidecadal fluctuations (Figure 8B). The AU3 spectral index shows, after an initial minimum that lasted around 45 years, a much more continuous increase towards the middle of GS17.1. However, some of the minor peaks in the Eifel data could potentially be correlated to corresponding signals in the Greenland record. During the second half of the stadial, both regions show a distinct periodicity of around 40 years (Figure 8B, Table S3).

Even though the RABD₆₇₀ values lie below the quantification limit during stadial periods subsequent to the early MIS 3 warm phase (Figure 4), some major climate trends occurring in Greenland are still visible in the AU3 RABD₆₇₀ data (Figure 9, Table S3). During GS6, for example, the spectral index shows values well below the threshold of 1.01 throughout the entire stadial, but even here, structural similarities can be observed in the two records, e.g., the double peak at 32,680 yr b2k or the tripartite plateau between 33,150 and 33,200 yr b2k (Figure 9A). GS13 is also characterized by indices well below 1.01. However, the four prominent δ¹⁸O maxima during the first half of the stadial were also recorded in the Auel sediments as four distinctive peaks, even exceeding the quantification limit (Figure 9B).

An in-depth comparison between the two climate proxy records shows that the AU3 RABD₆₇₀ spectral index not only matches the NGRIP oxygen isotope data very well but exceeds it in its temporal resolution. The NGRIP δ¹⁸O record, which was sampled with a resolution of 5 cm [32], has a temporal resolution of around 1 to 3 years. The temporal resolution of the AU3 RABD₆₇₀ record lies in the order of around 10⁻¹ to 10⁻² years in sections where varves are preserved (sampling resolution of 0.083 mm), revealing well-defined structural trends on a sub-annual level (Figure 6B–D). This makes it possible to study individual blooming events of phytoplankton and gain highly detailed insight into seasonal ecological changes inside the Auel maar. For example, the pronounced variations in the RABD₆₇₀ record during early GI6 represent massive blooming events, which resulted in the formation and preservation of prominent varves due to a lack of bioturbation (Figure 6C,D).

4. Discussion

4.1. Advantages and Disadvantages of the Hyperspectral Imaging (HSI) Method for Lake Productivity Reconstruction

The HSI method provides information on sediment geochemical properties at extremely high resolution (0.083 mm; sub-annual), enabling interpretation of lake productivity and environmental change at unprecedented resolution. This high resolution is critical for this study, in which we aim to compare lake productivity in Eifel sediments with δ¹⁸O of Greenland ice cores. The HSI method offers additional advantages of being rapid and non-destructive, allowing for subsequent analyses on the same core. Furthermore, the spatial nature of the data allows for recognition of sedimentary structures such as varves, diagenetic features, or sediment disturbances, and it is possible to observe the relationship between these features and biogeochemical properties of the sediment, such as pigment abundance. A disadvantage of the method is that spectral indices can provide only relative estimates of pigment abundance, and spectral features can be non-unique, meaning it is possible for absorption troughs to be caused by sedimentary components other than chloropigments. The combination of the high-resolution HSI method with measurement of pigment extracts on discrete samples is powerful, because the spectrophotometer measurements of chloropigments enable calibration of the RABD₆₇₀ index to semi-quantitative pigment concentrations. However, neither of the pigment methods we used can distinguish between the various chloropigment compounds such as chlorophyll-a,b, pheophytin-a, pheophorbide-a, and others. Therefore, we are not able to interpret changes in the aquatic communities or changes in degradation of chloropigments. Furthermore, the method struggles to quantify pigments at low concentrations (<~5 μg/g); thus, data from sediments with scarce pigments will be qualitative only.

4.2. Comparison of Paleoenvironmental Condition in the Eifel, Germany, with the NGRIP Oxygen Isotope Record

During the past 60,000 years, the environmental conditions in the Eifel region gradually deteriorated [7,42] as a result of the continuous trend towards colder and more arid climate conditions (e.g., [15,41]). The vegetation first changed from a lush, warm temperate spruce and hornbeam forest with thermophilous tree taxa, which dominated the Eifel throughout early MIS 3 (49,000 to 60,000 yr b2k), into a cold temperate forest near the end of GI13. During subsequent stadials, the landscape further transformed into a forest-steppe scattered with birch and pine trees subsequent to GI8, into a forest-tundra at around 28,500 yr b2k, and finally into the polar desert of the Last Glacial Maximum [7].

Throughout this gradual deterioration of the prevailing climate and environmental conditions, warmer and more humid phases (interstadials) caused the vegetation cover in the Eifel to at least partly recover, and forests regrew [42]. The favorable environmental condition during these warmer episodes allowed microorganisms to thrive, resulting in a significant increase in primary production inside the Auel maar. This is reflected both in the ELSA-20 C_org(chlorins) content [8] and the RABD₆₇₀ spectral index of core AU3 presented in this study (Figure 4).

The climates of Greenland and continental Central Europe were not only deeply connected on timescales of stadials and interstadials. Especially during colder, more arid intervals, i.e., stadial periods or later interstadials, the strong connection between the two regions, even in minor trends, can be clearly seen in the presented data, as the RABD₆₇₀ spectral index shows striking similarities to the multidecadal climate cycles of the NGRIP δ¹⁸O [32,33,34] during parts of the last glacial cycle (Figure 5, Figure 6 and Figure 8). This is most apparent during GI6, when a prominent cycle between 27 and 48 years persisted in both regions (Figure 6A).

However, the strong coupling mechanism between the two regions is not quite as obvious throughout the entire last glacial cycle (Figure 7). Although the overall trends of the Greenland oxygen isotope ratio and AU3 spectral index are quite similar, for example, during GI15.2, with some prominent climate signals occurring in both regions, some maxima and minima are significantly more pronounced in the Eifel record or not present in the Greenland data (Figure 7B), or the climate proxy records of the two regions differ significantly in their general shape, as can be observed during GI8 (Figure 7A). This is because the RABD₆₇₀ spectral index represents the primary production inside the maar lake. Although mainly driven by temperature variations, this also depends on various other boundary conditions (e.g., nutrient availability, wind causing surface water turbulences, solar radiation for photosynthesis, water turbidity, lake stratification, and pigment preservation). Additionally, as microorganism growth, and thus aquatic primary production, are limited to the warmer, ice-free seasons, the RABD₆₇₀ record only represents warm growing seasons [41], whereas the NGRIP ice accumulated perennially. In addition to temperature changes, the typical sawtooth structure of the NGRIP δ¹⁸O record is also heavily influenced by sea-ice cover in the Nordic Seas and variations in Greenland precipitation [43,44]. The sharp increase in the oxygen isotope data at the onset of GI8 suggests a rapid temperature rise in Greenland within a few decades following a slow and steady cooling trend towards the middle of the interstadial. This fits well with palynological data from the Eifel region, which also show an immediate increase in tree pollen at the initial onset of GI8 [11]. The AU3 RABD₆₇₀ spectral index and ELSA-20 C_org(chlorins) [8] records, however, follow a much slower and longer-lasting rising trend, reaching peak values around 720 years after the onset (Figure 7A). This reflects the gradual development of favorable growing conditions for the lake’s microorganisms over time as Central Europe experienced warm and humid climate conditions (e.g., [11]). During the middle of the interstadial, as the moisture source for Greenland precipitation changed from relatively high to mid latitudes due to an increasing pressure gradient between high and low latitudes, Europe’s climate became more arid [45]. The vegetation of the Eifel region began to transition from a boreal forest into a steppe landscape [11], and the Auel maar experienced a sharp drop in aquatic primary production (Figure 7A). Regardless of the various feedback mechanisms impacting the proxy records, prominent climate signals during mid-GI8 can be observed in both the Greenland ice and the Eifel maar sediments, suggesting a strong coupling between the two regions during the interstadial.

The divergent structures between the NGRIP oxygen isotope and AU3 RABD₆₇₀ spectral index during long-lasting warmer intervals, e.g., the early MIS3 warm phase [7,42] or GI8 [11], may be due to favorable conditions for flora and fauna persisting long enough for complex environmental feedback mechanisms to be activated in- and outside the Auel maar, with many contributing factors (e.g., nutrient cycling, groundwater chemistry, soils and vegetation in the catchment influencing the sedimentation process). This resulted in a more complex climate signal in the Eifel sediments than in the Greenland ice (Figure 7). In contrast, during colder and more arid stadial phases, e.g., GS16.1 or GS17.1 (Figure 8), or during later interstadials, when the Eifel already transitioned into a steppe- or tundra-like landscape, e.g., GI3, GI4, or GI6 (Figure 5 and Figure 6), fewer environmental factors influenced the primary production inside the Auel maar. As the duration of the ice-free (growing) season significantly controls lake productivity [41], the coupling between the regions’ climates is much more evident during colder periods with shorter growing seasons.

As Greenland and Central Europe are around 3000 km apart, the question of what could ultimately drive climate changes over such spatial distances becomes apparent. Although the final answer remains under debate, as there are many contributing factors in the feedback mechanisms between bio-, hydro-, cryo-, geo-, and atmosphere (e.g., [38,43,46]), an increasing amount of evidence has been found in the proxy records of several marine drill sites (e.g., [17,18,19,45,47]). Variations in sea ice cover in the Nordic seas and related changes in the strength of the AMOC, which plays a crucial role in the ocean’s CO₂ storage and the meridional heat transport, ultimately drove the climates in Greenland and the European continent during the last glacial cycle, reaching as far as Northern Turkey [48]. Even Chinese speleothem records have been linked to the Greenland ice core chronology [49], highlighting the significant influence of the AMOC on the global climate.

Due to their unique location and favorable depositional and preservation conditions, the Eifel maar basins are exceptional archives that can provide important information about the impact of past changes of the AMOC on Central European climate variations. Because the various maar lakes were filled with water during different geological times, it is potentially possible to study the entire environmental history of the region, and thus Central Europe, back to the Eifel’s first volcanic eruption phase around 500,000 years ago [2,3,4]. Recently, the ELSA project extended the high-resolution ELSA-20 organic carbon (chlorins) record [8] and the ELSA pollen stack [7] back to around 130,000 yr b2k [42,50], increasing our knowledge about the region’s climate and environmental past. Another recently published study deals with the sediments of the Dottinger infilled maar, which cover the time span between around 385,000 and 457,000 years and contain a clear succession of the Holstein interglacial and the Elsterian glacial periods, MIS 11 and MIS 12, respectively [5]. With the latest technological advancements in hyperspectral core-scanning techniques, however, these windows into the past can now potentially be investigated in ultra-high resolution.

5. Conclusions

The strong coupling between Greenland and Central European (Eifel) climate changes, which persisted throughout the last glacial cycle, is not only evident in the major climate shifts between warm/humid (interstadials) and colder/more arid (stadials) conditions (Figure 4), but also even in decadal-centennial-scale variations within these periods. The coupling between aquatic primary production in the Auel maar and Greenland climate is most apparent during cool and arid periods, i.e., stadial episodes and later interstadials, showing prominent multidecadal climate cycles (between 20 and 125 years; Figure 5, Figure 6 and Figure 8). In contrast, during long-lasting warm and humid intervals, favorable environmental conditions had enough time to establish complex systems of feedback mechanisms, which impacted the lake’s microorganisms and resulted in more complex patterns in the proxy record (Figure 7).

Although the climates in Greenland and Central Europe experienced some regional differences during the past 60,000 years due to the complexity of the global climate system, they both appear to be ultimately driven by changes in the strength of the AMOC and associated variations in the heat transport from lower to higher latitudes.