1200 years of decadal-scale variability of Mediterranean vegetation and climate at Pantelleria Island,Italy

Coring and dating

Two parallel cores (VEA and VEB) were extracted from Lago di Venere in summer 2008 with a modified Livingstone corer 4.8 cm in diameter (Lang, 1994). The cores were recovered from the deepest lake basin in 11 m water depth. The cores were visually correlated using prominent lithological layers to produce a continuous sedimentary sequence. The master core contains mostly sections from core VEB and was in total 10.29 m. The sediments of core VEB between 516 and 451 cm show various degrees of folding and therefore have been interpreted as a slump.

Terrestrial plant macrofossils (leaf, wood and charcoal) from core VEB were used for Accelerator Mass Spectrometry (AMS) analyses to infer radiocarbon dates (Table 1). The dates were converted to calibrated ages (cal. yr ad) using Calib version 6.0 with the radiocarbon age calibration set Intcal09 (Reimer et al., 2009). The two ¹⁴C dates were linearly interpolated and extrapolated to provide sample ages (Figure 3) revealing a basal age in the master core of 785 cal. yr ad. The depth–age model dates the slump around 1423 cal. yr ad (Figure 3). As this slump deposit does not contain paleoclimatic and palaeovegetational information, this section was not used for the discussion of the climate and vegetation history.

OPEN IN VIEWER

Table 1.

Radiocarbon dates. The mid-point of the calibrated 1σ age range was used for building the age model.

Laboratory code	Core	Depth (cm)	Material	14C age (yr BP)	Cal. yr ad
Poz-30194	VEB	588–590	Charcoal, leaf, wood	685 ± 30	1329 ± 52
Poz-30195	VEB	989–991	Wood	1185 ± 52	833 ± 52

OPEN IN VIEWER

Figure 3.

Radiocarbon dating of the sediments of Lago di Venere. Error bars show the 1σ error range of the calibrated ¹⁴C dates. Black dots show the midpoint of the 1σ error range, and the sampling depth of the dated material. The depth–age model was produce by linear interpolation of the dated levels.

Pollen and microscopic charcoal

Subsamples of 1 cm³ of sediment were taken every 8 cm for pollen and microscopic charcoal analyses. Standard procedures for glycerin samples (Moore et al., 1991) were followed in order to prepare a total of 95 sediment samples. Lycopodium tablets were added to estimate pollen concentrations (grains/cm³) and influx (grains/cm² per yr; Stockmarr, 1971). Pollen and spore identification followed keys (e.g. Beug, 2004; Moore et al., 1991; Punt, 1976; Van Geel, 1986), pollen atlases (e.g. Reille, 1992) and the reference collection of the Institute of Plant Sciences of the University of Bern. Pollen preservation is good for the entire record. The minimum number of pollen grains counted per slide was c. 250, excluding pollen of water plants and spores. The pollen diagram was subdivided into local pollen assemblage zones (LPAZ) using the zonation method of optimal partitioning (Birks and Gordon, 1985). The number of statistically significant zones was inferred with the program BSTICK following the broken-stick approach (Bennett, 1996). Microscopic charcoal particles longer than 10 μm were counted in pollen slides and used to calculate charcoal influx (particles/cm³ per yr; Finsinger and Tinner, 2007; Stockmarr, 1971; Tinner and Hu, 2003). The trend of the most important pollen types has been determined by smoothing with a five-point moving average.

Sedimentology and XRF core scanning

The master core was sedimentologically described following Schnurrenberger et al. (2003). An Avaatech^TM x-ray fluorescence (XRF) core scanner was applied to the master sections to infer changes in the elemental composition of the sediment (For details see Richter and van der Gaast, 2006). The measurements were conducted continuously using standard settings (voltage: 10 kV; current: 200 μA, count time: 20 s) with stepwise sampling of 1 cm. Powder x-ray diffraction (XRD) measurements with a Bruker AXS D8 Advance allowed the determination of the major minerals.

Pollen stratigraphy

The pollen and charcoal record (Figures 4, 5) is subdivided into three statistically significant local pollen assemblage zones (VEB1-3). Zone VEB1 (800–1000 cal. yr ad) is characterized by high percentages of arboreal taxa. The most important taxa are Olea, Quercus ilex-type and Pinus (from 20% to 40%). Quercus pubescens-type and Quercus suber-type are present, but with lower percentages (Figure 3). The shrub component is mostly formed by Erica arborea-type, Juniperus and Pistacia (up to 20%). The herb sum (dominated by Artemisia, Poaceae and Rumex-type) remains low through the whole zone (Figure 4).

OPEN IN VIEWER

Figure 4.

Arboreal pollen percentage, charcoal influx, and lithology diagram of Lago di Venere. Selected pollen types only. Empty curves show 10× exaggerations.

OPEN IN VIEWER

Figure 5.

Pollen and spore percentage diagram of herbs, ferns, wetland plants, aquatic plants, charcoal influx, and lithology of Lago di Venere. Empty curves show 10× exaggerations.

Zone VEB2 spans from ad 1000 to 1900 cal. yr. The main change is a prominent increase of Pinus percentages (up to 80%) and the decrease of other arboreal taxa, such as Olea and Q. ilex-type (Figure 4). The presence of Q. pubescens-type pollen is discontinuous. This taxon, after a first contraction, reappears from 1400 cal. yr ad and lasts until 1750 cal. yr ad. Shrub pollen increases in comparison to the preceding zone. E. arborea-type reveals a shift from 10% to 40% and Cistus percentages also increase. However, Pistacia and Juniperus decline. The sum of the herbs is low, although Rumex-type shows some peaks (Figure 4; 1090, 1180, and 1750 cal. yr ad).

VEB3 is a short zone that covers the last century (1890–2008 cal. yr ad). An important decrease of Pinus (from 80% to 20%) and E. arborea-type (from 40% to 2%) characterizes the whole zone. Q. ilex is the only taxon that shows an increase (up to 35%; Figures 3, 4). Herbs remain very low, although two peaks (1900 and 1990 cal. yr ad) occur that are mainly formed by Rumex-type, Poaceae, Urtica dioica-type, Cerealia-type and Artemisia.

Charcoal record

Our charcoal data (Figure 4) show that from 800 to 1000 cal. yr ad (VEB1) no prominent microscopic charcoal peak occurs. Within the pollen zone VEB2 a first microscopic charcoal peak is registered at 1300 cal. yr ad (151,350 particles/cm² per yr). Subsequently two minor peaks are recorded, at 1350 cal. yr ad (50,150 particles/cm² per yr) and at 1370 cal. yr ad (70,200 particles/cm² per yr). The highest charcoal influx values occur at c. 1500 cal. yr ad (194,600 particles/cm² per yr). Less pronounced charcoal peaks occur at 1580 cal. yr ad (115,400 particles/cm² per yr) and at 1650 cal. yr ad (131,191 particles/cm² per yr). Subsequently, abundance of microscopic charcoal generally decreases and reaches low values during the last 50 years.

Lithology and sediment composition

The core consists of carbonatic gyttja, which is in certain intervals clearly laminated, as well as sections dominated by clastic sediments of various grain sizes and increased organic content. The litho-stratigraphic profile can be subdivided into five main units (Figure 4). Unit 1 (800–1000 cal. yr ad) consists of light carbonate laminations with a maximum thickness of 3 mm intercalated with darker, organic-rich lamina partly consisting of microbial threats. This is overlain by Unit 2 (1000–1300 cal. yr ad) characterized by clastic sediment of clay, silt and sand size fraction, partly graded and sporadically containing volcanic pumice. Organic content is high. Unit 3 (1300–1400 cal. yr ad) is an interval of carbonate mud interrupted by dark sections rich in detritus. Unit 4 (1400–1650 cal. yr ad) is, like Unit 2, dominated by clastic sediment with organic rests. The stratigraphically youngest Unit 5 consists of carbonate mud with increasing carbonate content towards the top. Part of Unit 5 is laminated, whereas some parts are homogenous with increased organic content.

XRD measurements indicate aragonite is the dominant carbonate mineral for the entire sedimentary record. Therefore, changes in Calcium (Ca) abundance in the sediment correspond to variations in the amount of aragonite. We measured changes in Ca abundance using XRF (Figure 6). Although the Ca curve cannot be converted into absolute carbonate concentrations, the relative changes are a semi-quantitative indication of the varying carbonate content. Sedimentary content is high during two intervals, at the start of the record (c. 800–1000 cal. yr ad) and from c. 1850–1900 cal. yr ad to the end of the record. Ca is generally low from 1000 to 1900 cal. yr ad, with somewhat higher values from 1300 to 1500 cal. yr ad (Figure 6).

OPEN IN VIEWER

Figure 6.

Smoothed curves of sediment calcium content, important pollen types, and lithology. Dotted lines show significant pollen zones. High calcium values correspond to drier intervals, low values to moister intervals.

Sedimentology and climate

The Lago di Venere sediment varies between light carbonate lamina and darker intervals dominated by detrital influx and rather higher organic content. These variations are well expressed in the Calcium (Ca) curve from the XRF measurements (Figure 6). The water analysis of several studies (Aiuppa et al., 2007; Cangemi et al., 2010; Duchi et al., 1994) documented a supersaturation of aragonite in the water column. Analysis of short sediment cores confirmed that aragonite is also very abundant in the surface sediments (Aiuppa et al., 2007). Therefore, it is likely that during dry intervals, a reduced lake volume increases the ionic concentration of the lake water favouring the precipitation of aragonite. In contrast, during wetter climatic phases more detrital input from the volcanic catchment is washed in, diluting the sedimentary carbonate content. Thus we interpret high Ca values as indicating drier intervals, and low Ca values as indicating wetter intervals (Figure 6). On this basis, our results record a dry interval from 800 to 1000 cal. yr ad and increased moisture availability from 1000 to 1900 cal. yr ad, followed by a shift to drier conditions. During recent years the lake underwent seasonal lake-level fluctuations of several decimeters (Aiuppa et al., 2007), but also a general lowering of the lake level is observed since the 1960s (Duchi et al., 1994). Although the reason for this modern lake level lowering remains unclear, a drying trend would increase the evaporation/precipitation ratio, which would lead to increased formation of aragonite, and explain the high Ca values in the upper sediments (Figure 6).

Vegetation history and climate

Broadleaved evergreen trees (Q. ilex, Olea) dominated the vegetation of the island forming forests from 800 to 1000 cal. yr ad (Figure 4). Shrubby communities mainly consisting of thermophilous taxa such as Pistacia and Juniperus, and heliophilous herbs such as Artemisia and Cichorioideae were probably also widespread. In the present vegetation, Juniperus turbinata mainly occurs in the infra-Mediterranean semi-arid coastal vegetation, whereas Olea and Pistacia are frequent in the thermo-Mediterranean belt of the island and Q. ilex in the thermo-Mediterranean and meso-Mediterranean belts. The vegetation was probably influenced by particularly dry summer conditions between 800 and 1000 cal. yr ad. Pollen-independent evidence for dry conditions comes from sedimentology. During the period 800–1000 cal. yr ad, an interval rich in carbonate layers occurred when drought-adapted taxa such as Q. ilex, Olea and Pistacia were abundant (Figure 6). This period roughly coincides with the ‘Medieval Warm Period’ (MWP) that in the Northern Hemisphere was characterized by higher temperatures, compared with those of the centuries that immediately preceded and followed it (Bradley et al., 2003). The MWP lasted between about 900 and 1300 ad, however the timing has been demonstrated to vary geographically (Bradley and Jones, 1993; Bradley et al., 2003; Hughes and Diaz, 1994).

An important, statistically significant, change of vegetation (boundary LPAZ 1/2) occurred about 1000 cal. yr ad. Conifers and especially Pinus expanded, while broadleaved evergreen trees declined (Figure 4). Today, drought-adapted Pinus halepensis and more mesophilous Pinus pinaster both grow on the island. However, P. pinaster stands are far more abundant (5.1% of total vegetation) than those of P. halepensis (0.3%; Gianguzzi, 1999). Thus, Pinus pollen should predominantly (93%) reflect an increase in P. pinaster. Similarly, relatively moisture-demanding E. arborea communities expanded, while the drought-tolerant maquis formed by Pistacia and Juniperus declined.

A link between Pinus expansion at 1000 cal. yr ad and changing climatic conditions is again supported by sedimentology. Calcium declined between 900 and 1000 cal. yr ad, and carbonate deposition remained low from 1000 to 1850 cal. yr ad (Figure 6). Thus, the vegetational change was probably driven by a climatic shift towards moister conditions before and during the ‘Little Ice Age’ (LIA). The age of the LIA is not very well constrained but the earliest evidence dates back to c. ad 1100–1200 in the European Alps (Holzhauser et al., 2005). We assume that enhanced rainfall and/or cooler conditions during the LIA increased moisture availability, thus promoting relatively moisture-demanding species (e.g. P. pinaster, E. arborea, Arbutus unedo, Rumex, Selaginella denticulata) at the expense of more xeric vegetation (e.g. Q. ilex, Olea europaea, Pistacia, Juniperus turbinata and Phillyrea; Figure 4). The P. pinaster expansion may also mean that in contrast to today’s regime of relatively moist winters and very dry spring and summer seasons (Figure 2), warm-season moisture was more abundant during the LIA. P. pinaster shows better growth with an annual precipitation of at least 850 mm/yr, with at least 75 mm from May to September (Richardson and Rundel, 1998; Vidakovic and Soljan, 1991). Summer water stress is the main limiting factor for P. pinaster growth and because of its high stomatal sensitivity to soil water availability, this species is considered a drought-avoider (Bogino and Bravo, 2008; Chambel et al., 2007; Granier and Loustau, 1994; Picon et al., 1996). At present P. pinaster is restricted in a small area above 800 m a.s.l. on Pantelleria. However, the widespread presence of Pinus on the island from 1000 to 1900 cal. yr ad, may indicate that warm-season precipitation was more abundant during the LIA than at present.

Although lake sediment climate records from elsewhere in Italy lack the stratigraphic resolution needed to provide a detailed climate history from the last millennium (Luterbacher et al., 2012), our pollen and sedimentological data showing a shift from a dry MWP to a moist LIA is consistent with other evidence from the central and western Mediterranean. For example, surface temperatures near Southern Italy in Gulf of Taranto declined after 1000 cal. yr ad and remained low until 1800 cal. yr ad (Taricco et al., 2009). Lake records from Spain also show a dry MWP followed by a shift to moister conditions. However, dry conditions in Pantelleria lasted until 1000 cal. yr ad, whereas the dry MWP in Spain occurred between about 700 and 1400 cal. yr ad (Martin-Puertas et al., 2008; Morellon et al., 2009; Moreno et al., 2012).

The last 100 years from 1900 to 2008 cal. yr ad are characterized by a re-expansion of broadleaved evergreen taxa. Local Olea populations seem to respond faster than Q. ilex. However by the end of the sequence, Q. ilex is the most important tree species. After 1900 cal. yr ad, communities formed by E. arborea declined while Pistacia re-expanded. The vegetation change fits well with the carbonate content that increased during the past 130 years, suggesting increased evaporation/precipitation ratios after the end of the LIA. This climatic trend is well documented also by the weather station in Pantelleria that reports an increase of 1°C of the annual mean temperature from 1951 to 1999 ad (Pirazzoli and Tomasin, 2003). Cooler conditions during the LIA on the island are also recorded in written sources that report the presence of stone wall structures built to gather snow on the top of Montagna Grande (Calcara, 1853). A shift to warmer and drier conditions at the end of the LIA is evident elsewhere in Southern Italy. For example, a historical precipitation series from Sicily reports an abrupt rainfall decrease after ad 1922 (Diodato, 2006). Similarly, the recurrence of drought events in Southern Italy increased during the 20th century relative to the previous 500 years, which may relate to a reduction in spring precipitation and increasing annual temperatures (Brunetti et al., 2006; Diodato, 2007; Diodato and Bellocchi, 2011). Drying at Pantelleria is also consistent with evidence for a general shift from the wettest period of the LIA during the 18th and 19th centuries in the western Mediterranean, to a drier 20th century (Nicault et al., 2008).

Vegetation history and human impact

Vegetational changes at Pantelleria were co-determined by agriculture and land use. Owing to its position in the middle of the Strait of Sicily, the island has been for millennia an important trade centre in the Mediterranean area. Despite chronological uncertainties, it has been hypothesized that the island was already frequented by 5900 bc, mainly to exploit obsidian (Whittle, 1996). Other archaeological sources indicate that Pantelleria was inhabited since the Bronze Age (i.e. 2000–1300 bc) and the most fertile areas were cultivated (Francaviglia and Piperino, 1987). Terraces and cisterns, introduced by Punic people (Baldassari, 2007), protected crops against the winds, provided new space for cultivation, reduced slope inclination and collected rainwater to supply the island’s needs. Our new data cover the historical period that began with the Arabic domination of the island. Arabs conquered Pantelleria from the eastern Roman Empire in ad 835, colonizing the island and intensifying agriculture. The Arab period continued during the following centuries, but came to an abrupt end in ad 1123, when the Normans conquered the island unifying it to Sicily and Southern Italy (D’Aietti, 1978). Afterwards, under Bourbon domination (ad 1713–1860), the intense production of capers, wine and charcoal was established, that has never been abandoned (D’Aietti, 1978).

Various indicators of human land use are recorded in the Lago di Venere sedimentary sequence. Grazing indicators including Anthemis-type, Cichorioideae, Aster-type, Brassicaceae, Artemisia-type, Caryophyllaceae, Carduus, as well as dung spores such as Cercophora, Podospora and Sporormiella, are distributed along the sequence and signal the continuous presence of humans on the island. Plantago lanceolata, Cerealia-type, and Mercurialis annua, suggest high levels of arable farming during the 20th century (c. 1900–1970, Figure 5). Of special interest is the strong increase of herbs (>40%) during the period between c. 950 and 1100 cal. yr ad (Figure 5). This is probably the most long-lasting period of intense land-use activity. A link to climate change (e.g. drier and thus more open conditions) seems unlikely, since the expansion of Pinus and E. arborea point to increasing moisture availability. Instead, the abundance of Cerealia-type and Sporormiella point to intensification of farming at the end of the Arabic period and the onset of the Norman domination. From c. ad 1180 to 1280, the occurrence of Asteroideae, Apiaceae and Cerealia-type is synchronous to the most prominent peak of Sporormiella and to an increase of dung spores, such as Cercophora and Podospora (Figure 5). These data agree very well with historical sources that highlight the importance of agriculture and pasture in ad 1123 when Normans conquered Pantelleria (D’Aietti, 1978). Another significant episode of intensification of human activities occurred during the period ad 1450–1550, when frequent dung indicators are well linked to Cerealia-type, Plantago lanceolata, Mercurialis annua, Carduus, Cichorioideae and Anthemis-type. Pollen data (e.g. Cerealia-type) suggest high levels of arable farming during the 20th century (c. 1900–1970; Figure 5). During the first decades of the 20th century, viticulture was the main trade of the island and it is still important today (Di Lorenzo and Sottile, 1990), which is reflected in the pollen record (sparse Vitis finds).

Island fire history

Changes in charcoal frequency suggest variable fire activity. In general fire activity was higher during the LIA (peaks from 1350 to 1650 cal. yr ad) than before and during the past century (Figure 5). This may suggest a fuel-limited ecosystem (Pausas and Fernández-Muñoz, 2012), since the LIA was cooler and/or wetter than the periods before and afterwards. More likely, human control of fires and especially fire ignition may have determined the fire regime. In the Mediterranean region almost all fires are induced by humans and fire suppression or ignition is usually a key factor of fire regimes (e.g. Moreno et al., 1998). Nevertheless, climate significantly influences the occurrence of fire, for favourable fire-weather conditions affect flammability and influence the extent and intensity of fires once they are ignited (Vannière et al., 2011). Fuel-limited regimes have been reconstructed from Eski Acıgöl and Nar crater lakes in Turkey (Turner et al., 2008), whereas in the western and central Mediterranean, usually under moister conditions, fires were favoured by drier conditions (Tinner et al., 2009; Vannière et al., 2011).

After repeated fire disturbance Mediterranean ecosystems can undergo regressive succession towards a Cistus-dominated garrigue with Pteridium aquilinum (Chiappini, 1985; Rühl, 2003). For example, after burning, Juniperus communities are often replaced by Cistus creticus, C. salvifolius and C. monspeliensis (Gianguzzi, 1999). These shrubs and ferns are well recorded, especially Cistus which replaced Juniperus at around 1100 cal. yr ad (onset of VEBA2 and date of Norman conquest). Together with the increase of open spaces (pollen of herbs) and human indicators (e.g. Cerealia-type, Sporormiella), this shift suggests increased anthropogenic fire activity (Figure 5). Fire impact on the environment is also suggested by the sparse presence of Gelasinospora, the spore of a fungus growing on charcoal (Figure 5). Thus, the overall increase of Cistus spp. between 1000 and 1900 cal. yr ad may relate to increasing fire disturbance, not the climatic effects of the LIA.

Wood was constantly cut on Pantelleria to obtain timber, charcoal and firewood from the Norman period onward (11th century ad; D’Aietti, 1978). A consequence of extensive logging and burning in Pantelleria was the extinction of deciduous Quercus (probably Q. pubescens) c. 1750 cal. yr ad. Although this species is not currently present on the island, it was broadly used for shipbuilding purposes (Furia, 1863). Charcoal influx shows a decrease in the last pollen zone, when forests partly recovered. Ecological estimations point to a strong decline of cultivated land during the past century (from 81.6% to 28.1%; Rühl, 2003), however, our pollen record suggests that open, unforested land during the past century was not significantly reduced, if compared with the past 800 years of vegetation history.