Winter-spring signal in the Iberian range speleothems

A previous study on the origin of rainfall based on the isotopic composition of rainfall events during three years in the Iberian range demonstrated that precipitation comes both from Atlantic and Mediterranean sources (Moreno et al., Reference Moreno, Sancho, Bartolomé, Oliva-Urcia, Delgado-Huertas, Estrela, Corell, López-Moreno and Cacho2014). At the cave locations, leeward of the Atlantic influence, spring is the rainiest season (Fig. 2A), whereas winter is relatively dry and winter precipitation is not related to the North Atlantic Oscillation (NAO) pattern (Martin-Vide and Lopez-Bustins, Reference Martin-Vide and Lopez-Bustins2006; Lopez-Bustins et al., Reference Lopez-Bustins, Martin-Vide and Sanchez-Lorenzo2008). On the contrary, a moderate negative correlation, statistically significant between δ¹⁸O rainfall values and western Mediterranean Oscillation index (WeMOi), is described previously for this area, with a Spearman’s rank annual correlation coefficient of -0.25 and a P value of 0.046 in spring (Moreno et al., Reference Moreno, Sancho, Bartolomé, Oliva-Urcia, Delgado-Huertas, Estrela, Corell, López-Moreno and Cacho2014). The WeMOi has its greatest influence in the eastern fringe of the Iberian Peninsula, being especially well correlated to rainfall amount in winter-spring. This correlation suggests that regional atmospheric teleconnections, linked to the western Mediterranean pressure system’s patterns of variability, control the incidence of storms associated with northeastern cyclogenesis (Lopez-Bustins, J.-A., Reference Lopez-Bustins, Martin-Vide and Sanchez-Lorenzo2008). Importantly, this synoptic pattern (“back-door fronts” in Supplementary material, Supplementary Fig.1) is dominant during the end of the winter through the beginning of spring, indicating the influence of a Mediterranean origin in that part of the year when the amount of precipitation is higher (Fig. 2A). Coherently, calcite precipitates on glass slides occur during seasons of positive water balance, autumn through spring, suggesting a seasonal bias in the obtained speleothem records (see Supplementary Fig. 1). Thus, elevated deposition rates may bias the isotopic signature of stalagmites fed by seasonally responsive drips toward the rainy season values in both caves.

Absolute values of δ¹³C are slightly different in HOR, MO-1, and MO-7, varying between -7.6‰ and -10.12‰ in Ejulve cave and between -6.06‰ and -11.24‰ in Molinos cave (Fig. 4). This different amplitude in the range of variation may be explained by the host rock composition, considering that both caves are below similar rock thickness (less than 20 m) and similar soil and vegetation cover development (today and very likely during the Holocene). Thus, Ejulve cave originated on Cenomanian dolostones, while the host rock of Molinos cave is formed by crystalline marine limestones of Cenomanien-Turonian age. As dolostones are less prone to dissolution than limestones, a lower range of variation should be expected in the δ¹³C profile from HOR sample compared to Molinos stalagmites, indicating less influence of the interaction with host rock in Ejulve cave. In addition, the HOR sample is characterized by higher growth rates than Molinos samples, pointing to faster dripwater flow and thus, less time for rock-water exchange. In coherence with these growth rate differences, Mg/Ca values are lower and less variable in the HOR stalagmite from Ejulve cave than in the Molinos MO-1 sample (see different y-axis in Fig. 4). We hypothesize that the higher growth rates in HOR compared to Molinos samples point to faster carbonate precipitation and, consequently, less time for Ca²⁺ to precipitate.

Figure 4 Carbon (δ¹³C) isotope values vs. time from HOR (green line), MO-7 (blue line), and MO-1 (red line), compared with Mg/ Ca from the same stalagmites. Three phases are indicated. (For interpretations of the references to color in this figure legend, the reader is referred to the web version of this article.)

In addition, the dominance of a columnar fabric (see Supplementary material) helps to interpret that these speleothems precipitated close to isotopic equilibrium (Belli et al., Reference Belli, Frisia, Borsato, Drysdale, Hellstrom, Zhao and Spötl2013). This assumption is corroborated by the large diameter of the stalagmite, which is considered typical for stalagmites that reflect the original composition of drip water (Dreybrodt and Scholz, Reference Dreybrodt and Scholz2011). Consequently, δ¹³C values should reflect processes external to the two studied caves, such as soil dissolved inorganic carbon (DIC). Thus, in spite of the different range of variation, δ¹³C profiles from the three stalagmites are similar in their general trends (Fig. 4, plotted at the same scale), this correlation suggesting a common mechanism controlling the δ¹³C records despite the notable differences in the host rock composition and the different features of the stalagmites (e.g., growth rates).

The trends in the δ¹³C variation over the whole Holocene can be described in terms of three intervals (Phases in Fig. 4). First, the lightest values (-12‰ to -8‰) are observed in the early Holocene (11.7–9.4 ka) represented in MO-1. Second, a tendency towards higher values is observed after a hiatus (9.4–8.5 ka) and later maintained during the middle Holocene (-8‰ to -6‰), recorded by MO-1, MO-7 and HOR speleothems from 8.5 to 4.8 ka. Finally, after 4.8 ka, lower values (-9‰ to -10‰) are observed again. Although these three phases are not so evident in Mg/Ca profiles from HOR and MO-1, still the correlation between Mg/Ca and δ¹³C is high, with Spearman’s correlation coefficients of 0.52 (P value<0.0001) for HOR and 0.28 (P value<0.0001) for MO-1. In contrast, the MO-7 sample, with lower resolution, displays a Mg/Ca record not similar in trends or values to the other two speleothems (Fig. 4). Therefore, the Mg/Ca profile from MO-7 is not used for further interpretations.

Interpretation of δ¹³C variation must consider that carbon in speleothem calcite has two main sources: (1) soil CO₂, which is controlled by atmospheric CO₂, plant respiration, and aged organic matter degradation; and (2) bedrock carbonate (CaCO₃) that is dissolved during seepage. In this way, in a dry period, soil activity and vegetation above the cave decrease and cause higher δ¹³C values of soil CO₂ and, later on, enriched isotope values of speleothem carbonate (Genty et al., Reference Genty, Blamart, Ghaleb, Plagnes, Causse, Bakalowicz and Zouari2006; Moreno et al., Reference Moreno, Stoll, Jiménez-Sánchez, Cacho, Valero-Garcés, Ito and Edwards2010). In this cave setting, the influence of age of respired soil gas CO₂ due to ancient organic matter degradation (important in tropical climates, Noronha et al., Reference Noronha, Johnson, Hu, Ruan, Southon and Ferguson2014) is considered negligible due to the dryness of this Mediterranean area, with expected low soil development and scarce organic matter. It is important to note that, for the studied time interval, there was no human transformation of the landscape in this region until the Iron Age (ca. 2800–2500 yr), when there was an expansion of human occupation and an increase in settlements (Ibáñez and Burillo, Reference Ibáñez and Burillo1995). Possibly, as a consequence of this increase in population, soil degradation may have happened (Ibáñez and Burillo, Reference Ibáñez and Burillo1995 and references therein). From that time on, only limited calcite precipitation was observed in both studied caves, suggesting a change in the cave hydrology.

Regarding the influence of vegetation cover in the δ¹³C records, we need to take into account that the hydrological balance in this region is negative, making moisture availability the more critical factor for forest development. In fact, recent studies have demonstrated that the variability in the amount of rain during spring season is the main cause for regional forest expansion or decline (Camarero et al., Reference Camarero, Gazol, Tardif and Conciatori2015).

An additional factor, proven to be of influence in δ¹³C records from other caves in northern Iberia (Moreno et al., Reference Moreno, Stoll, Jiménez-Sánchez, Cacho, Valero-Garcés, Ito and Edwards2010) and southern France (Genty et al., Reference Genty, Blamart, Ghaleb, Plagnes, Causse, Bakalowicz and Zouari2006), is the extent of CO₂ degassing prior to calcite precipitation on top of the stalagmite. Higher degrees of degassing accompany the slower drip rates and percolation through unsaturated epikarst conduits during periods of lower rainfall and result in differential ¹²C release and more positive δ¹³C values in precipitated calcite. However, in these speleothems, the lack of correlation among δ¹³C and δ¹⁸O and the dominant fabrics (compact columnar, see Supplementary material) indicate precipitation close to equilibrium (Frisia et al., Reference Frisia, Borsato, Fairchild and McDermott2000). Therefore, in this case, δ¹³C is reflecting the soil-carbon pool, which is mostly derived from the extent of soil microbial respiration (Fairchild and Baker, Reference Fairchild and Baker2012), which in this regional climate is limited by the amount and timing of soil moisture. Under these premises, the δ¹³C values depend on the combination of soil- and limestone-derived carbon, from which the DIC in groundwater derives (e.g., Rudzka et al., Reference Rudzka, McDermott, Baldini, Fleitmann, Moreno and Stoll2011).

Thus, interpreting the δ¹³C profiles in these terms, the three-part Holocene can be described in terms of water availability during the rainy season (Fig. 4), with a wetter period from the Holocene onset to ca. 9.4 ka (Phase I, characterized by lower δ¹³C values); drier from 8.5 to 4.8 ka (Phase II, with higher δ¹³C values); and wet again afterwards (Phase III, with generally lower δ¹³C values again). The hiatus from 9.4 to 8.5 ka may be associated to the onset of Phase II (dry) but the lack of records prevents a strong interpretation here. Differences between Phase II and Phase III are not so visually evident in HOR sample, but the average values differ by 1.3‰, with lower δ¹³C values in Phase III.

Despite some of the potential complexities of using speleothem Mg as a paleohydrological proxy, many studies have demonstrated a convincing relationship between dripwater and/or speleothem Mg and recorded rainfall (e.g., McMillan et al., Reference McMillan, Fairchild, Frisia, Borsato and McDermott2005; Fairchild and Treble, Reference Fairchild and Treble2009). Significant correlation observed in MO-1 and HOR speleothems among δ¹³C and Mg/Ca suggests either that periods of reduced soil microbial activity (heavier soil CO₂ δ¹³C) were also drier (high Mg/Ca), and/or that a small portion of the δ¹³C variation arises from degassing and prior calcite precipitation effects caused by humidity variations (Moreno et al., Reference Moreno, Stoll, Jiménez-Sánchez, Cacho, Valero-Garcés, Ito and Edwards2010). Although the second case is less acceptable here as explained above, in both cases, dry periods would most plausibly lead to higher δ¹³C and Mg/Ca ratios. Beyond the observed correlation, the abruptness of the Mg/Ca variation is markedly different from that of the isotopes, suggesting different thresholds of sensitivity (e.g., growth rate, different host rock).

Seasonality as the key for reconciling records on hydrological variability during the Holocene

The three stages defined in terms of hydrological variability appear synchronous with the signal observed in lake level records north of ca. 40°N in the central Mediterranean (see location in Fig. 1: Accesa, Cerin, and Ledro lakes; e.g., Magny et al., Reference Magny, Combourieu Nebout, de Beaulieu, Bout-Roumazeilles, Colombaroli, Desprat and Francke2013), that is, two humid periods (before ca. 9 ka and after 4.5 ka) separated by a drier phase (see Accesa lake level in Fig. 5c). Interestingly, in spite of the different timing of transitions, the same three-part pattern is reproduced in Bunker cave stalagmites (Germany; see location in Fig. 1), where a winter signal is demonstrated as the result of enhanced evapotranspiration during summer months leading to reduced infiltration into the karst aquifer (Fig. 5d; Fohlmeister et al., Reference Fohlmeister, Schröder-Ritzrau, Scholz, Spötl, Riechelmann, Mudelsee, Wackerbarth, Gerdes, Riechelmann, Immenhauser, Richter and Mangini2012). One could argue that this three-part pattern represents the more Mediterranean characteristics recorded in our studied caves compared to other Iberian records regulated by Atlantic conditions, such as Pyrenean lake records (González-Sampériz et al., Reference González-Sampériz, Valero-Garcés, Moreno, Jalut, García-Ruiz, Martí-Bono, Delgado-Huertas, Navas, Otto and Dedoubat2006; Gil-Romera et al., Reference Gil-Romera, González-Sampériz, Lasheras-Álvarez, Sevilla-Callejo, Moreno, Valero-Garcés and López-Merino2014), northwest Iberian lakes (Santos et al., Reference Santos, Vidal Romani and Jalut2000; López-Merino et al., Reference López-Merino, Silva Sánchez, Kaal, López-Sáez and Martínez Cortizas2012), or speleothem records from northern Iberian coast (e.g., Stoll et al., Reference Stoll, Moreno, Mendez-Vicente, Gonzalez-Lemos, Jimenez-Sanchez, Dominguez-Cuesta, Edwards, Cheng and Wang2013; see location of all cited records in Fig. 1). Thus, the comparison among speleothem δ¹³C profiles and other records of water availability is attempted (Fig. 5).

Figure 5 Comparison of Holocene paleo humidity records. (A) Percentage of terrigenous (aeolian flux) from ODP Site 658 (northwest African margin) indicating (arrow) the end of the African Humid Period (deMenocal et al., Reference de Menocal, Ortiz, Guilderson, Adkins, Sarnthein, Baker and Yarunsiky2000). (B) K/Al ratio from MD99-2343 (Minorca island, northwest Mediterranean) marking the decrease in fluvial input throughout the Holocene (two steps indicated by arrows; Frigola et al., Reference Frigola, Moreno, Cacho, Canals, Sierro, Flores, Grimalt, Hodell and Curtis2007). (C) Lake level oscillations in Accesa Lake (North Italy; Magny et al., Reference Magny, Bégeot, Guiot and Peyron2003). (D) Bunker cave (W Germany; Fohlmeister et al., Reference Fohlmeister, Schröder-Ritzrau, Scholz, Spötl, Riechelmann, Mudelsee, Wackerbarth, Gerdes, Riechelmann, Immenhauser, Richter and Mangini2012). (E) δ¹³C values from HOR (green line), MO-7 (blue line), and MO-1 (red line). Note that the axis-y orientation is inverted to mark drier conditions towards the top of the page. Gray bars indicate the two transitional periods: first, from the early to middle Holocene and, second, from the middle to late Holocene. (For interpretations of the references to color in this figure legend, the reader is referred to the web version of this article.)

Regarding the hydrological signal recorded from sequences under typical Mediterranean climate, fluvial supply to marine sediments from offshore Minorca island (marine core MD99-2343; Frigola et al., Reference Frigola, Moreno, Cacho, Canals, Sierro, Flores, Grimalt, Hodell and Curtis2007) are compared to our speleothem records (Fig. 5). The K/Al ratio is considered a good indicator for clay inputs (mainly illite) from river runoff. Thus, its variation (Fig. 5b) is interpreted to indicate higher humidity during the early Holocene that gradually decreases in concert with the end of the African Humid Period, that is, associated with the decrease in orbitally-driven summer insolation and thus marked by an increase in aeolian-transported particles (Fig. 5a; Harrison and Digerfeldt, Reference Harrison and Digerfeldt1993; deMenocal et al., Reference de Menocal, Ortiz, Guilderson, Adkins, Sarnthein, Baker and Yarunsiky2000). The decrease in humidity in Minorca core is in two phases, one ca. 9.5 ka and the second one, more intense, at ca. 5 ka (arrows in Fig. 5b). A recent study on one stalagmite from southern Iberia (36°N) based on calcite density variations was interpreted to reflect a decrease of humidity after 5.3 ka (Walczak et al., Reference Walczak, Baldini, Baldini, McDermott, Marsden, Standish, Richards, Andreo and Slater2015) once the speleothem becomes more porous. Other records under Mediterranean climate are similar to this pattern (e.g., El Maillo record from Iberian Central Range; Morales-Molino et al., Reference Morales-Molino, García-Antón, Postigo-Mijarra and Morla2013), although the early Holocene is not always a period with high water availability but, in contrast, characterized by long dry summers that would be probably determinant in the hydric balance recorded by lake sediments.

Other examples of the Iberian “typical Mediterranean” pattern (dry-wet-dry) are Villarquemado paleolake (Aranbarri et al., Reference Aranbarri, González-Sampériz, Valero-Garcés, Moreno, Gil-Romera, Sevilla-Callejo and García-Prieto2014) and Ojos del Tremedal sequences (Stevenson, Reference Stevenson2000) both located in the Iberian Range, or Estanya lake in the Pre-Pyrenees (Morellón et al., Reference Morellón, Valero-Garcés, Moreno, González-Sampériz, Mata, Romero, Maestro and Navas2008), where the early Holocene is still cold and relatively dry and the most humid interval covers from 9 to 5 ka, generally related to the “Holocene Climate Optimum” concept in terms of forest expansion (González-Sampériz et al., Reference González-Sampériz, Aranbarri, Pérez-Sanz, Gil-Romera, Moreno, Leunda and Sevilla-Callejo2017). In the same way, Sancho et al., (Reference Sancho, Arenas, Vázquez-Urbez, Pardo, Lozano, Peña-Monné and Hellstrom2015) identified a maximum of tufa deposition at 7 ka in northeastern Iberia under wet conditions. This pattern is also coherent with other records located at higher altitude but still under Mediterranean influence, for example Basa de la Mora lake record (Pérez-Sanz et al., Reference Pérez-Sanz, González-Sampériz, Moreno, Valero-Garcés, Gil-Romera, Rieradevall and Tarrats2013) in the Pyrenees, Siles lake in the Segura mountains of southern Spain (Carrión, Reference Carrión2002), or Borreguiles de la Virgen in the Sierra Nevada (Fig. 1; García-Alix et al., Reference García-Alix, Jiménez-Moreno, Anderson, Jiménez Espejo and Delgado Huertas2012).

Comparing Molinos and Ejulve records with sequences under the direct influence of Atlantic climate is the second step. In general, Holocene climate reconstructions from sites located in northern Iberia indicate the control of orbitally-driven insolation, with an inflection at ca. 6–5 ka (e.g., Muñoz Sobrino et al., Reference Muñoz Sobrino, Ramil-Rego, GóMez-Orellana and Díaz Varela2005). Most pollen-based climate reconstructions from Atlantic Iberia show an early Holocene rapid transition to vegetation dominated by temperate deciduous woodland, a maximum forest development up to about 8 ka, and a transition during the late Holocene to the current disturbance-adapted vegetation (see a review in Carrión et al., Reference Carrión, Fernández, González-Sampériz, Gil-Romera, Badal, Carrión-Marco, López-Merino, López-Sáez, Fierro and Burjachs2010). Similarly, the compilation of growth rates from up to 15 stalagmites in northern Spain indicates a change from wet early Holocene to drier conditions after 6 ka and the cessation of most growth by 4.1 ka (Stoll et al., Reference Stoll, Moreno, Mendez-Vicente, Gonzalez-Lemos, Jimenez-Sanchez, Dominguez-Cuesta, Edwards, Cheng and Wang2013).

Clearly, the opposite pattern (wet-dry-wet) found in our Iberian Range speleothems compared to most Mediterranean sequences (dry-wet-dry) indicates that interpretation of the δ¹³C record as changes in Mediterranean-related humidity requires an additional explanation (e.g., different season of humidity). Additionally, the lack of correlation with most Atlantic sequences leads us to confirm that our δ¹³C profiles are not directly representing the humidity variability derived from the entrance of westerly-driven Atlantic fronts. We propose that Molinos and Ejulve records are shaped mainly by Mediterranean-sourced precipitation but during the end-of-winter/beginning-of-spring, a season not generally represented by other sequences.

Additionally, information about fire frequency reconstructed from charcoal records in western Mediterranean lacustrine sites reflects a different variation in records from northern and southern Mediterranean Iberia. Thus, higher values of fire frequency were observed during the “Holocene Climate Optimum” around 8–5 ka in records from northeastern Iberia (Las Pardillas, El Carrizal, Ojos del Tremedal, and Hoya del Castillo) while records from the southern Mediterranean Iberia (including Siles Lake, Cañada de la Cruz, Villaverde, and Navarrés) indicate a decrease associated with wetter-than-present summers (Vanniere et al., Reference Vanniere, Power, Roberts, Tinner, Carrion, Magny and Bartlein2011). This pattern is very similar to our record in northeastern Iberia with the drier period during 8–5 ka. Particularly coherent is the fact that, after 5 ka, records from northern Mediterranean Iberia indicate an absence of fire activity, whereas a significant increase in fire activity was recorded in the south. This coherence likely highlights that Mediterranean precipitation in winter-spring season has been neglected in previous studies of hydrological variability along the Holocene.

Variations in both the synoptic-scale climate patterns and large-scale atmospheric circulation indexes (e.g., NAO and WeMOi) are invoked to explain the observed variability and, particularly, the lack of correlation among north and south Mediterranean records already proposed by Magny et al. (Reference Magny, Combourieu Nebout, de Beaulieu, Bout-Roumazeilles, Colombaroli, Desprat and Francke2013). In Molinos cave, precipitation of carbonate in artificial supports occurs from late fall to early spring (Moreno et al., Reference Moreno, Sancho, Bartolomé, Oliva-Urcia, Delgado-Huertas, Estrela, Corell, López-Moreno and Cacho2014), the season with higher present-day correlation between amount of rainfall and WeMOi in this area (October to March in Martin-Vide and Lopez-Bustins, Reference Martin-Vide and Lopez-Bustins2006). This index explains the pluviometric variability in the eastern fringe of the Iberian Peninsula, particularly in the Gulf of Valencia, since it allows the detection of the variability relevant to the cyclogenesis next to the western Mediterranean (Martin-Vide and Lopez-Bustins, Reference Martin-Vide and Lopez-Bustins2006). Negative values of the WeMOi are correlated to higher precipitation in relation to the entrance of northeasterly flows over eastern Iberia (Lopez-Bustins et al., Reference Lopez-Bustins, Martin-Vide and Sanchez-Lorenzo2008). Therefore, at present-day, years with negative value of the WeMOi result in wetter winter-spring seasons in the Iberian Range (leeward of the Atlantic influence) and, coherently, would produce more negative δ¹³C values in the precipitated calcite (e.g., early Holocene). This mechanism is possibly not true for southern Mediterranean Iberian records (e.g., El Refugio cave; Walczak et al., Reference Walczak, Baldini, Baldini, McDermott, Marsden, Standish, Richards, Andreo and Slater2015), where winters are wet but related to the southern position of the North Atlantic Subtropical High and the associated westerlies; this area is under the influence of the NAO mechanism today (Trigo et al., Reference Trigo, Pozo-Vázquez, Osborn, Castro-Díez, Gámiz-Fortis and Esteban-Parra2004) and also for the late Holocene (Ramos-Román et al., Reference Ramos-Román, Jiménez-Moreno, Anderson, García-Alix, Toney, Jiménez-Espejo and Carrión2016). Therefore, the operation of a mechanism similar to the present-day WeMOi with different intensity during the Holocene is considered the main force shaping Molinos and Ejulve speleothem records.