The role of human-induced fire and sweet chestnut ( Castanea sativa Mill.) cultivation on the long-term landscape dynamics of the southern Swiss Alps

Study area

The Insubrian region is located on the southern slopes of the Alps across Switzerland and Italy and is characterized by warm-temperate and rainy climate, with a mean annual precipitation around 1700 mm and an average annual temperature of about 12°C. This climate features dry and mild winters with some days (40 days a year on average) having strong gusts of a katabatic (descending) dry wind from the North, which causes drops in the relative humidity to values as low as 20%. In summer, long periods without rain or even periods of drought may alternate with thunderstorms and short, heavy spells of precipitation, making the total amount of summer precipitation (about 800 mm) significantly higher than in other Mediterranean areas (Spinedi and Isotta, 2004). The region is prone to fire since the onset, around 9200 cal. yr BP, of such Insubrian climate characterized by high amounts of rainfall combined with episodic dry or drought periods (Tinner et al., 1999).

The soils are generally classified as haplic podzol (cryptopodzol) on cristalline bedrock (Blaser et al., 2005). The suitability of such environmental conditions for the cultivation of the sweet chestnut was soon recognized by the Romans that introduced the tree species cultivation in the Insubrian region (Conedera et al., 2004; Krebs et al., 2004; Tinner et al., 1999). Since the first plantation, sweet chestnut has been intensively cultivated as a monoculture, and native forest vegetation was consequently suppressed in favour of chestnut coppices and orchards for the production of wood and fruits for multiple purposes (Conedera et al., 2004). During the Middle Ages, the sweet chestnut tree even became a fundamental source of food for the local human population (Conedera and Krebs, 2008). First signs of a decreasing importance of chestnuts as a staple food were associated with the progressive improvement of agricultural techniques, the introduction of alternative crops from the Americas (e.g. maize, potatoes) and the onset of industrialization (with the subsequent exodus of the people from the countryside to the cities). However, sweet chestnut cultivation remained the standard forest use until the socio-economic expansion starting in the late 1950s (Conedera and Krebs, 2008). Traditional sweet chestnut economy consisted of intensely managed orchards (exploited for fruit harvesting, litter collection, disperse pasturing) and short-rotation coppices. After the 1950s, the traditional sweet chestnut cultivation definitively dropped as a consequence of the general abandonment of the primary sector and the contemporaneous spread of the newly introduced chestnut blight (Cryphonectria parasitica). In the absence of management, chestnut coppices and especially chestnut orchards started to be invaded by other species (e.g. Betula pendula, Acer pseudoplatanus, Tilia cordata, Fraxinus excelsior, Fagus sylvatica) and to evolve towards mixed deciduous forests (Conedera et al., 2000, 2001; Pividori et al., 2005).

Such a dramatic change in land use resulted in a collateral consequence, namely the significant increase in fire frequency starting in the mid-1950s, when the average yearly number of fires in Canton Ticino triplicated passing from 30 to 90 events over 150,000 ha of forests. Fire frequency and burnt area dropped again in the last three decades thanks to fire prevention measures and the improvement in fire fighting techniques and organization (Pezzatti et al., 2013). At present, most of these wildfires are of anthropogenic origin and occur mainly during winter and spring, from December to April, although a trend towards higher lightning fire occurrence in summer has been detected (Conedera et al., 2006a; Reineking et al., 2010).

Study site

Guèr (46°16′44″N, 9°01′18″E) is a small mire (around 0.42 ha) on a glacial terrace at 832 m.a.s.l. situated in the municipality of Claro (Canton Ticino), above the main village (300 m.a.s.l.; Figure 1). Glaciers have largely modelled the landscape of the main valley during the Pleistocene; moraines and other glacial deposits are frequent (Scapozza, 2013). The catchment covers around 0.27 km² from 832 until 1510 m.a.s.l., and is mainly covered by several centuries-old sweet chestnut orchard trees, birch (Betula pendula), alder (Alnus glutinosa), beech (Fagus sylvatica), linden (Tilia spp.) and other tree species in the lower part (until ca. 1000 m.a.s.l.), with mostly Norway spruce (Picea abies) in the upper and distal part.

OPEN IN VIEWER

Figure 1.

Top panels: location of the study site, Guér mire, in Canton Ticino (Switzerland). Lower left: climate diagram in the area (Acquarossa/Comprovasco meteorological station, modified from MeteoSwiss). Lower right: distribution of chestnut forests in the area (yellow green) with present (green) and past (dark green) chestnut orchards around Guèr; main archaeological sites are also shown. 1: Cadossola (Neolithic/Bronze Age), 2: Necropolis of La Monda (Bronze/Iron Ages), 3: Valaréne (Iron Age), 4: North of the Station (Iron Age), 5: Residence Ai Cedri (Iron Age), and 6: Duno (Iron Age).

Prehistoric settlements in the area

The area has a long history of human settlement, as indicated by the early evidence of human presence in the surroundings (Figure 1), from the Neolithic to the Middle Ages, with a likely abandonment during the Roman Times. A serpentine axe from Neolithic times discovered at the site called Cadossola is the oldest archaeological remain in the area. Archaeological findings are rare from the Bronze Age period (e.g. a bronze axe also from Cadossola) but get more abundant from the Iron Age, between 600 and 200 BC (2550–2150 cal. yr BP). In particular two cemeteries (La Monda and north of the Claro railway station), a single grave (in the site called Duno), a bronze vessels depot (at Valaréne, known in the literature as ‘Pontone’), as well as other isolated findings, probably associated with a settlement and discovered during the construction of the residence Ai Cedri (Bullo, 2011; Carlevaro, 2013; de Marinis, 2000b). Most of these sites are located on the valley bottom (Figure 1). Duno, Valaréne and residence Cedri are dated between 500 and 400 BC (2450–2350 cal. yr BP), while the necropolis north of the Claro railway station is more recent with materials dated between 400 and 200 BC (2350–2150 cal. yr BP; Carlevaro, 2013).

The necropolis of La Monda is the only archaeological site in the Claro community located at relatively high elevation (ca. 500 m above the valley bottom), on the same main glacial terrace as Guèr (Figure 1). The close proximity of the necropolis (distant only ca. 400 m) was originally the main motivation to core the Guèr site. The archaeological remains of La Monda date back to the late Bronze Age and the Iron Age. At least in the context of Canton Ticino, La Monda is a very special case of Iron Age necropolis because of its altitude (the highest in southern Switzerland) and location on the side of the valley far from the main village (Carlevaro, 2013). The existence of this cemetery could be associated with the presence of a permanent settlement, which has not been found yet. A total of 26 graves were uncovered, a cremation tomb from the Bronze Age (de Marinis, 2000a) and inhumation burials dating to the Iron Age. The objects collected from the tombs are stored in the Swiss National Museum (Zürich, Switzerland). They date from 1300 BC (3250 cal. yr BP, late Bronze Age) to 300 BC (2250 cal. yr BP, Second Iron Age), but most of them belong to the early Iron Age between 600 and 400 BC (2550–2350 cal. yr BP). The findings of the necropolis of La Monda also include amber necklaces imported from the Baltic Sea area and bronze vessels (‘situlae’, the typical container of wine), suggesting that local Celtic inhabitants played an active role in controlling the transalpine trade routes. Activities such as the direct control and maintenance of these routes and the provisions necessary for people and animals to cross the Alps gave access to the basically agriculture-oriented local Celtic population, the Lepontii, to additional source of income allowing them to possess prestige goods (Della Casa, 2007).

Coring, sampling and chronology

Two parallel cores (GUA and GUB, 396 and 375 cm long, respectively) were retrieved in June 2007 from the centre of the Guèr mire using a Streif modified Livingstone piston corer (Merkt and Streif, 1970). The uppermost 10 cm of the sedimentary sequence was collected separately as a surface block. The surface block and the two parallel cores were accurately correlated on the basis of their lithostratigraphy. The final master core consists of surface block (0–10 cm), GUA (10–72 cm) and GUB (72–375 cm).

We obtained eight AMS-radiocarbon dates from terrestrial macrofossils (see Table 1), which were converted to calendar years using the CALIB 6.0.1 program (INTCAL09 calibration curve; Reimer et al., 2009). For the final age–depth model we used a linear interpolation between adjacent radiocarbon dates (median of the calibrated ages). We selected the simplest model after considering different alternatives (GAM, locally weighted regressions), which were excluded because the age estimates were outside the 2σ ranges of calibrated ages (95.4% of the probability distribution) of several of the available radiocarbon dates (Figure S1, available online).

OPEN IN VIEWER

Table 1.

AMS radiocarbon dates from the Guèr sequence, measured on terrestrial plant macrofossils at the Poznan Radiocarbon Laboratory (Poland).

Sample depth (cm)	Laboratory number	Material	Radiocarbon age (14C yr BP)	Calibrated age ranges, p = 0.954 (cal. yr BP) a	Age in diagrams (cal. yr BP) b
47–50	Poz-28009	Wood	870 ± 30	698–905	780
101–104	Poz-31973	Wood	1730 ± 40	1548–1711	1640
146–149	Poz-31974	Bark	2110 ± 40	1991–2292	2080
249–251	Poz-31975	Betula catkin scales and seeds, bud scales	3110 ± 40	3218–3439	3340
297–298	Poz-28001	Wood	4490 ± 40	4979–5295	5170
320–322	Poz-40965	Wood	6430 ± 50	7269–7427	7360
350–351	Poz-28012	Charcoal	7750 ± 50	8420–8603	8520
363–367	Poz-40966	Periderm, charcoal	9190 ± 50	10,242–10,495	10,350

a

Radiocarbon dates have been converted into calendar years using the CALIB 6.0.1 program (Stuiver and Reimer, 1993) and the INTCAL09 calibration curve (Reimer et al., 2009).

b

Median of the probability distribution of calibrated ages rounded to the nearest decade.

Macroscopic charcoal analysis

The core was continuously sampled every centimetre for macroscopic charcoal analysis. A total of 327 sediment samples of 1 cm³ were obtained from slices 1 cm thick. We took sub-samples until the depth of 327 cm because of the presence of a thick sandy layer, probably associated with erosion, below this depth between 347 and 328 cm.

For the analysis, sediment samples were gently washed through a 100-µm mesh sieve. Charcoal particles were then counted under a stereomicroscope at a magnification of 40× (analyst: C Morales-Molino). Macroscopic charcoal concentrations (particles cm⁻³) were converted to charcoal influx or accumulation rates (here denoted CHAR, particles cm⁻² yr⁻¹) using the sediment deposition rates (cm yr⁻¹) derived from the age–depth model (Figure S2, available online). Afterwards, CHAR values were resampled to the median sample resolution (16 ± 22 years) to achieve a constant time interval. For the identification of CHAR peaks linked to local fire episodes (within 0.5 km distance of the study site; Higuera et al., 2007) a three-step decomposition of the CHAR curve was followed (Higuera et al., 2010; Long et al., 1998). The slowly varying component of the CHAR series, called background CHAR which is associated with regional fire activity, changes in fuel load availability and/or to taphonomic processes such as reworking (Long et al., 1998; Whitlock and Larsen, 2001), was modelled by fitting a locally weighted regression (lowess) robust to outliers with a 1000-year smoothing window. We tested the sensitivity of our macroscopic charcoal data to the use of different age–depth models by comparing the resultant CHAR curves (Figure S3, available online), and to different smoothing functions (i.e. standard lowess and moving average) and windows to obtain the background CHAR curve (Figure S4, available online). Differences were minimal, which allowed us to conclude that the reconstructed fire frequency is robust.

Residuals (peak series) were then obtained by subtracting the background CHAR component from the raw CHAR record. Finally, a threshold was used to separate the noise component (due to random variability, sediment mixing, distant fires and redeposition; Clark and Royall, 1996; Long et al., 1998; Whitlock and Larsen, 2001) from the peaks that probably represent local fire episodes. The noise distribution of the CHAR peak series was modelled using a Gaussian mixture model (Gavin et al., 2006), and then a global threshold was defined at its 99th percentile. After this, a minimum count screening (Gavin et al., 2006) was performed to avoid the identification of peaks that are not statistically significant. To account for the variability in charcoal production that could be associated with changes in vegetation composition and fuel loads, we also used a locally defined threshold to determine which peaks could be associated with local fire episodes (Higuera et al., 2009), allowing comparison with the results obtained with the global threshold. As the introduction of Castanea sativa led to a vegetation transition, we ran again the analyses considering two different zones: one included the period prior to the local establishment of Castanea sativa (estimated from the palynological ‘rational’ limit, that is, when the percentages of Castanea pollen reached the 5% threshold, at around 1600 cal. yr BP), and the other after Castanea sativa expansion (i.e. after 1600 cal. yr BP). Once the fire-related peaks were identified, inferred fire frequencies and fire return intervals were smoothed over 1000-year intervals. A signal-to-noise index (SNI; Higuera et al., 2009; Kelly et al., 2011) was calculated to assess how the signal of local fires is separated from the noise component. All the macroscopic charcoal analyses were performed using the CharAnalysis software (Higuera et al., 2009).

Pollen and microscopic charcoal analyses

A total of 38 sediment samples of 1 cm³ were taken at approximately every 10 cm for pollen analyses. The sediments were chemically (HCl, KOH, HF and acetolysis) and physically processed (0.5 mm sieving and decanting) following standard procedures (see, for example, Moore et al., 1991). Lycopodium tablets (Stockmarr, 1971) were added to estimate pollen concentration (grains cm⁻³) and microscopic charcoal influx or accumulation rates (particles cm⁻² yr⁻¹). For the identification of pollen types (analyst: E Vescovi), the reference collection at the Institute of Plant Sciences of the University of Bern and identification keys (e.g. Beug, 2004; Faegri and Iversen, 1989; Moore et al., 1991; Punt et al., 1976–2009) were used. The samples were analysed at 400–1000× magnification; the minimum pollen sum was 444 pollen grains. Pollen of water and wetland plants as well as spores were excluded from the pollen sum. In the following text and figures, we present a selection of most relevant tree and shrub taxa (Pinus sylvestris-type, Alnus glutinosa-type, Betula, Corylus, Quercus pubescens-type) and anthropogenic indicators (Castanea sativa, Juglans, Cerealia-type, Plantago lanceolata) to highlight vegetation composition before and after the introduction of Castanea sativa and the human impact at this transition. Microscopic charcoal particles >10 µm (longest axis) were counted in pollen slides under a transmitted light microscope following Tinner and Hu (2003) and Finsinger and Tinner (2005). Microscopic charcoal is mostly linked to regional fire activity within a distance of 20–50 km of the coring location and does thus not require background and peak analysis (Conedera et al., 2009; Tinner et al., 1998).

Sediment description and chronology

The sediment consists mostly of peaty gyttja (328–78 cm) with gyttja in the intervals 208–164 cm and 117–86 cm, and peat between 226 and 208 cm (Figure 2), which reflects a pattern of alternating shallow lake and mire phases. From 78 to 0 cm the sediment is mainly peat and fibrous peat (Figure 2), following the final conversion of the shallow lake into mire. According to the chronology, a lake was present at the beginning of the Holocene (ca. 11,600 cal. yr BP; Figure 2). The sedimentary environment turned into a shallow lake around 9800 cal. yr BP and finally became a mire ca. 1200 cal. yr BP. These changes in the sedimentary basin were probably due to natural infilling over the millennia and by water exploitation for cultivation purposes (Figure S5, available online).

OPEN IN VIEWER

Figure 2.

Raw macroscopic charcoal concentration, age–depth model, lithology and raw charcoal accumulation rates (CHAR) from the Guèr record.

Macroscopic and microscopic charcoal

Raw CHAR values usually range between 0 and 5 particles cm⁻² yr⁻¹ between ca. 7600 and 3300 cal. yr BP, followed by increasing values from ca. 3300 cal. yr BP (Figure 2). Maximum values over 30 particles cm⁻² yr⁻¹ are present around ca. 2000, 1500 and 1200 cal. yr BP. Finally, CHAR values drop again during the last millennium (0–5 particles cm⁻² yr⁻¹). The BCHAR curve shows low values at the beginning of the sequence (0.5–2.5 particles cm⁻² yr⁻¹ between 7600 and 3300 cal. yr BP) followed by a rising trend from ca. 3300 cal. yr BP onwards, and reaching a maximum between ca. 2200 and 2000 cal. yr BP (approximately 5 particles cm⁻² yr⁻¹). BCHAR values start to decrease afterwards, with very low values during the last millennium (0.3–1.8 particles cm⁻² yr⁻¹; Figure 3a). Generally, macroscopic charcoal concentrations and influx (or CHAR) show similar trends, although with different amplitudes, until ca. 3300 cal. yr BP (Figure 2).

OPEN IN VIEWER

Figure 3.

(a) CHAR (interpolated to the median sample resolution) and background component of the CHAR series (BCHAR). (b) Positive residuals of the Guèr macroscopic CHAR after subtracting the BCHAR from the interpolated CHAR series. Fire episodes identified using a global threshold are also shown, together with the inferred fire frequency (IFF) per millennium.

Using the 99th percentile global threshold, 29 peaks were identified as probably associated with local fire episodes (Figure 3b). As the median sample resolution is 16 years, detected peaks likely reflect clusters of individual fire events (Whitlock and Larsen, 2001). The global SNI is 5.65, indicating that peaks are well separated from the noise (Kelly et al., 2011). The inferred fire frequency (IFF; Figure 3b) remains low (<1–3 episodes/1000 years) from the beginning of the record until ca. 3600 cal yr BP, with a slightly increasing trend (from 1.1 to 2.4 episodes/1000 years) in the period between ca. 5500 and 4200 cal. yr BP. A sharp increase in IFF starts at ca. 3600 cal. yr BP (ca. 2 fire episodes/1000 years), peaking around 2600 cal. yr BP with the overall maximum value (9 fire episodes/1000 years; Figure 3b). Afterwards, the IFF remains high although following a slightly decreasing trend. The interval between 2600 and 1700 cal. yr BP is an interesting period given the vegetation dynamics (described below). This can be subdivided into two phases: first, between 2600 and 2250 cal. yr BP, fires are more frequent, and second, between 2250 and 1700 cal. yr BP, fire episodes became less frequent (at 2250, 2050 and 1700 cal. yr BP) but with higher peak values. This suggests that more biomass burnt during each fire episode (Figure 3b). The IFF decreases then slightly to reach values of 5–6 fire episodes/1000 years during the last millennium. In summary, the mean fire return interval (mFRI) throughout the record is 262 years (95% confidence interval: 159–380 years). Shifts in the positive residuals show important changes in the amount of biomass burnt along the record, with high peaks (15–28 particles cm⁻² yr⁻¹) concentrated mainly between ca. 2000 and 1000 cal. yr BP, while positive residuals are low during the first millennia of the sequence (up to 5 particles cm⁻² yr⁻¹) and the last millennium (maximum of 7 particles cm⁻² yr⁻¹; Figure 3b). An alternative fire reconstruction based on a locally defined threshold (Figure S6, available online) suggests the occurrence of 28 fire episodes, which corresponds to 271 years mFRI (95% confidence interval: 200–354 years). The rise in fire activity during the last millennium as detected with the locally defined threshold does neither match the marked peaks in the charcoal series nor the available historical archives; therefore, the reconstruction based on the globally defined threshold is preferred.

Microscopic CHAR values (Figure 4) are mostly low until 3200 cal. yr BP, but increase between 3200 and 2000 cal. yr BP with the two highest maxima in the sequence at ca. 2600 and 2000 cal. yr BP. A less marked increase occurred between ca. 5000 and 4000 cal. yr BP, resembling quite closely the background trends in the macroscopic charcoal record (Figure 4 vs Figure 3a). As microscopic charcoal is a good proxy for regional fire activity, this may suggest that the macroscopic BCHAR mainly derives from regional fires and not, as alternatively possible, from reworked (older) material.

OPEN IN VIEWER

Figure 4.

Summary pollen and charcoal diagram with most relevant pollen types, microscopic and macroscopic CHAR and the main parameters of local fire regime reconstruction (fire episodes, IFF).

Pollen and anthropogenic indicators

Considering the small size of our study site, with a catchment of less than 1 ha, we expect pollen assemblages mostly to reflect vegetation communities in the Guèr area (Conedera et al., 2006b; Jacobson and Bradshaw, 1981). As shown by calibration studies, most pollen of trees in the southern Alps, including insect- and wind-pollinated Castanea sativa, is dispersed within a few hundred metres (Conedera et al., 2006b), thus overlapping with the provenience of microscopic and macroscopic charcoal as a proxy for biomass burning (Conedera et al., 2009).

The bottom part of the record (roughly until ca. 3800 cal. yr BP; Figure 4) shows high tree pollen percentages (75–90%) suggesting the presence of a closed deciduous forest with Alnus glutinosa (26–40%), Betula (13–22%), deciduous Quercus (8–26%) and Pinus sylvestris (10–17% and a peak of 33%). The shrub Corylus is also present (4–5%), and was probably growing in the understory and the forest gaps. In the study area, Alnus glutinosa is not confined to lakes, peatlands or river shores; instead, it can also co-dominate or even dominate forests on upland slopes because of the high summer moisture availability. This largely forested landscape dominated for a relatively long period of low fire activity and human impact. Notwithstanding this general trend, a certain forest opening is apparent at 5000–4000 cal. yr BP (tree pollen decline from ca. 85% to 75%) linked to higher local (and partly also regional; see Figure 4) fire activity. This episode is followed by a rapid and remarkable spread of Pinus sylvestris (reaching its maximum of 33%) and a minor increase in Corylus (4–9%).

A first important decline in tree pollen is detected from ca. 3800 to 3200 cal. yr BP (ca. 80–60%), mostly driven by the drop in Pinus sylvestris (30–5%) and Alnus glutinosa (25–10%). The simultaneous increase of deciduous Quercus (11–27%) and Corylus (8–17%) suggests a progressive recovery of deciduous forest at the expense of pine stands. From ca. 3200 to 3000 cal. yr BP, a rise in tree pollen occurs (from 60% to 80%) mainly due to noticeable increases of Alnus glutinosa (10–30%) and Betula (12–22%). A marked decline in deciduous Quercus (27–11%) is coupled with the increase of these post-disturbance pioneer trees (i.e. Alnus glutinosa, Betula). Both processes are probably linked to enhanced local and regional fire activity starting from this period (Figures 3 and 4). Around 3000 cal. yr BP a more marked decline of tree pollen is observed (i.e. decreasing from 78% to 34%), involving mainly Alnus glutinosa (30–10%), Betula (22–7%), deciduous Quercus (11–5%) and Pinus sylvestris (10–2%). This major episode of forest clearance dated at ca. 3000–2600 cal. yr BP is synchronous with the period of highest fire activity, both at regional and local scales (Figures 3 and 4). The almost continuous record of Cerealia-type is in good match with the occurrence of Plantago lanceolata and suggests the presence of arable and pastoral farming close to the coring site between ca. 3000 and 2100 cal. yr BP. Additionally, the close link between land use indicators and fire activity suggests that the high fire activity recorded approximately between 3000 and 2100 cal. yr BP was caused by humans for agricultural purposes.

Tree pollen increased again (reaching 86%) after 2300 cal. yr BP suggesting a gradual forest recovery. Pioneer trees such as Betula (up to 30%) and Alnus glutinosa (maximum of 55%) dominated forest stands with a significant presence of deciduous Quercus (13–17%). These taxa established in a period of declining but still rather frequent fire occurrence, as indicated by macroscopic and microscopic charcoal records. Pollen of important tree crops such as Juglans regia (walnut) and Castanea sativa (sweet chestnut) is first recorded during this reforestation stage (at ca. 2000 and 1800 cal. yr BP, respectively).

Pollen data suggest that between 1700 and 1400 cal. yr BP, a marked change in tree species composition coincided with a moderate opening of the forest (tree pollen fell from 86% to 48%). Castanea sativa shows an abrupt rise in pollen percentages up to 28% and Corylus increases from 4% to 14%, significantly expanding at the expense of deciduous Quercus (decreasing from 17% to 4%), Alnus glutinosa (from 54% to 9%) and Betula (from 13% to 1%). During this stage of Castanea sativa establishment local fires were still frequent. Other anthropogenic indicators such as Cerealia-type, Juglans regia and Plantago lanceolata increase at ca. 1600 cal. yr BP (Figure 4) suggesting more intense agricultural activities.

Castanea sativa expanded significantly after 1200 cal. yr BP, retaining its dominance during the last millennium. The period with dominance of Castanea sativa was characterized by an open landscape (decrease of tree pollen other than Castanea) and intensified land use (increase of anthropogenic indicators).

The final establishment of Castanea sativa stands after 1600 cal. yr BP brought a considerable reduction in mFRI, passing from 324 years (95% confidence interval: 223–429 years) in the pre-Castanea period to 164 years (95% confidence interval: 94–249 years) afterwards. As suggested by the magnitude of positive residuals, even fire impact reached its maximum precisely during the Castanea sativa expansion and then decreased after the establishment of Castanea sativa orchards around Guèr (Figure 5). The most recent pollen spectra show increases in the representation of deciduous broadleaved trees, mostly Betula (23% at present) and Alnus glutinosa (16%), and a trend towards denser forest stands (68% of tree pollen).

OPEN IN VIEWER

Figure 5.

CHAR positive residuals for the periods before the first occurrence of Castanea sativa pollen (at ca. 2000 cal. yr BP), during the introduction and expansion (between ca. 2000 and 1600 cal. yr BP; ‘empirical limit’ at ca. 1800 cal. yr BP), and after the establishment (‘rational limit’ ca. 1600 cal. yr BP) of Castanea sativa stands.

Fire and human impact on vegetation prior to sweet chestnut introduction (7600–1800 cal. yr BP = 5650 BC–AD 150)

At the beginning of our record, during the mid-Holocene, between 7600 and 3200 cal. yr BP (5650–1250 BC), fire frequency was moderate (IFF = 1–3 fire episodes/1000 yr) and did not cause substantial forest opening but probably co-determined vegetation composition. The dominant trees Alnus glutinosa, Betula and deciduous Quercus are all known to be favoured by moderate fire occurrence (Tinner et al., 2000; Figure 4). This is also true for Pinus sylvestris (Hille and den Ouden, 2004), which persisted at least until ca. 3000 cal. yr BP, when land use significantly increased. At around 5000–4000 cal. yr BP (ca. 3050–2050 BC), during the late Neolithic and early Bronze Age, forests became more open as a consequence of increased fire activity and land use. Thus, human impact was the main driver of increased burning already before the onset of the Iron Age. Anthropogenic fire activity during the Neolithic and in the early Bronze Age increased also at Lago di Origlio and Lago di Muzzano (distant ca. 25 and 30 km from our study site, respectively; Tinner et al., 1999). The impact of fire on vegetation at Guèr was generally less severe than at lower altitude sites such as Origlio and Muzzano, where shrublands became more widespread (Tinner et al., 1999). Higher summer temperatures and enhanced seasonality during the mid-Holocene (Heiri et al., 2003; Renssen et al., 2009) likely favoured fire ignition and spread by humans. A comparison with palaeoclimatic proxies from the Alps suggests that anthropogenic Neolithic fires were more abundant during dry and warm climatic phases (Tinner et al., 1999).

The frequency of local fires during the mid-Holocene around Guèr (1–3 fire episodes/1000 years) is comparable to that at high-elevation sites in the Alps (e.g. Engadine, Valais; Colombaroli et al., 2010; Stähli et al., 2006) but lower than at mid-altitude sites in the Pyrenees (Rius et al., 2011). We explain the moderate fire activity at the study site by the relatively low human activity compared with the lowlands (e.g. Origlio, Muzzano), which were better suited for settlements and agriculture, because of warmer climatic conditions and the proximity to the major trade routes (Tinner et al., 2003). This interpretation is supported by archaeological evidence which, in the surroundings of Guèr as in other regions, is much more frequent in the valley bottom than on the slope terraces at mid-elevations (Carlevaro, 2013).

Fire activity reached a minimum during the Bronze Age between 3800 and 3300–3200 cal. yr BP (1850–1350/1250 BC). The modest forest opening at 3800–3200 cal. yr BP (tree pollen decreased from 80% to 60%) leading to a replacement of Pinus sylvestris with deciduous Quercus–Corylus stands was thus not primarily caused by fire. Instead, clusters of cultural indicators (e.g. Plantago lanceolata) suggest that this minor opening was related to pastoral activities. In addition, the shift in forest composition can be interpreted as the result of secondary succession following late Neolithic/early Bronze Age fires. A major shift in regional and local fire activity (microscopic charcoal record and BCHAR, IFF) started during the late Bronze Age at ca. 3200 cal. yr BP (1250 BC). Burning gradually increased to reach a maximum at 2600–2500 cal. yr BP (650–550 BC), in the middle of the Iron Age (see Figure 3b). At the beginning of this period, between 3200 and 3000 cal. yr BP (1250–1050 BC), there was an overall increase in tree cover under moderate fires, mainly due to rises in the pollen curves of Alnus glutinosa and Betula coupled with a moderate decline of deciduous Quercus. These responses of the main tree species to fire are in agreement with previous studies on fire sensitivity classifying Alnus and Betula as ‘fire-enhanced’ and deciduous Quercus as ‘fire-indifferent’ under moderate fire activity (Tinner et al., 2000). A major deforestation episode (tree pollen drop from 78% to 34%) occurred around the Bronze Age–early Iron Age transition, at 3000–2600 cal. yr BP (1050–650 BC), affecting also rather resistant tree species such as Alnus glutinosa and Betula which had been favoured by more moderate fire activity some centuries earlier. A mechanistic explanation is that increasing fire frequency (reflected by the charcoal record) crossed the post-fire forest recovery threshold. This may have caused several fire-resistant tree species, such as Alnus glutinosa, Betula and deciduous Quercus to become ‘fire-sensitive’ with increasing fire frequency as described in earlier studies (Gil-Romera et al., 2014; Tinner et al., 2000). When crop cultivation and weed abundance (Cerealia-type, Plantago lanceolata; Figure 4) increased, grasslands expanded greatly indicating that livestock farming was also important for local human communities. The rise of local and regional fires was one of the most prominent effects of increased human presence in the surroundings of Guèr during the late Bronze Age and especially the Iron Age, attested by several archaeological findings (see Figure 1). In addition to deforestation, and as a consequence, soil stability was affected leading to more erosion, that is, increased sedimentation rates (Figure 2; Figure S1 (available online)).

The highest regional fire activity was recorded at 2600–2000 cal. yr BP (650–50 BC; Figures 3a and 4), while burning in the vicinity of Guèr continued at high levels during this phase of the Iron Age and also the Roman Period up to 1700 cal. yr BP (AD 250). Frequent fires at 2600–2250 cal. yr BP (650–300 BC) promoted the persistence of a landscape dominated by pasturelands with reduced forest cover. Afterwards, between 2250 and 1700 cal. yr BP (300 BC–AD 250), fire episodes became less frequent and this change probably allowed the recovery of forest stands dominated mainly by Alnus glutinosa and Betula (Figure 4), i.e. trees favoured by moderately frequent fires (Tinner et al., 2000). It is likely that these trees, which can also behave as phreatophytes, (i.e. plants closely associated with aquifers for water uptake) could have additionally benefited from hydrological changes in the wetland, at that time probably at an intermediate stage between shallow lake and mire (Figure 2). The presence of several archaeological sites close to Guèr dating back the late Bronze Age and the Iron Age (especially La Monda; Figure 1) supports our interpretation of human controlled fire activity during this period.

The rise in fire activity during the Bronze and Iron Ages was a common feature on the southern slopes of the Alps as recorded in the microscopic charcoal records from previously studied lowland sites (Gobet et al., 2000; Hofstetter et al., 2006; Tinner et al., 1999). In the same way as at Guèr, fires were mainly set by humans to favour agricultural and pastoral activities. Enhanced fire frequency and human activities around Guèr during the late Bronze and Iron Ages were almost coeval with two phases of forest clearance and intensified farming in the southern Swiss Alps and the Swiss Plateau, dated at 3300 and 2600–2350 cal. yr BP (1350 and 650–400 BC; Tinner et al., 2003), when fire activity rised noticeably (Tinner et al., 2005). This pattern can be followed as south as Sicily (Bisculm et al., 2012; Tinner et al., 2009) with a prominent land use and fire peak at ca. 2700 cal. yr BP (750 BC). Archaeological findings support an increase of human activities such as livestock transhumance on the southern slopes of the Alps and the use of transalpine routes from the Bronze Age onwards (Della Casa, 2007; Marzatico, 2009). The same occurs with the reestablishment of woodlands around Guèr by the end of the Iron Age, which may be attributable to land abandonment following successive harvest failures, due to more humid climate during this period (Tinner et al., 2003). Similarly, Bronze and Iron Ages were also characterized by an intensification of farming activities and fire activity elsewhere in the Mediterranean (e.g. Carrión et al., 2001; Morales-Molino et al., 2011, 2013; Vannière et al., 2008, 2011).

The role of fire and humans for the establishment and management of sweet chestnut forests (1800 cal. yr BP–Present = AD 150–Present)

The introduction and cultivation of Castanea sativa is probably the most striking and important vegetation and land use change recorded during the past 7600 years at Guèr. On the one hand, the continuous pollen record of Castanea sativa (‘empirical limit’), occurring at ca. 1800 cal. yr BP (AD 150), probably indicates the introduction of the first sweet chestnut trees on the valley floor next to the main villages. On the other hand, the rapid increase in Castanea pollen percentages (‘rational limit’, >5% of the pollen spectra) that took place at ca. 1600 cal. yr BP (AD 350; see Figure 4) may reflect the first diffusion of chestnut stands on the slopes around the study site at mid-elevation (Figure 4). Finally, the first maximum in the Castanea sativa curve (at 1440 cal. yr BP, that is, AD 510) indicates that chestnut stands were certainly well-established by that time at least on the main glacial terraces. Juglans regia cultivation was introduced almost synchronously (ca. 2000 cal. yr BP/50 BC; Figure 4), supporting the idea of a radical change in the surviving-oriented economic strategy over these low- and mid-altitude areas during the Roman Period. The chronology of sweet chestnut cultivation from Guèr is in accordance with previous well-dated palynological records from the southern slopes of the Alps which showed that the introduction and expansion of Castanea sativa at a regional scale started at around 1900–1600 cal. yr BP (AD 50–350; Gobet et al., 2000; Hofstetter et al., 2006; Tinner et al., 1999).

The high-resolution charcoal record from Guèr suggests important changes in fire regime before, during and after the establishment of well-developed Castanea sativa stands (Figures 4 and 5). Overall, fire frequency increased almost twofold (i.e. mFRI dropped from 324 to 164 years) at the introduction time of Castanea sativa (Figure S7, available online). The first significant expansion of Castanea sativa took place at the expense of mixed deciduous forests dominated by Alnus glutinosa, Quercus and Betula, which experienced extensive clearance during the late Roman and the Migration Periods at 1700–1400 cal. yr BP (AD 250–550). Frequent and severe fires (Figures 3 and 5) occurred during this period, therefore showing that the introduction and establishment of Castanea sativa stands implied the use of fire for forest clearance and a partial conversion to more open landscapes. Increases in the representation of anthropogenic pollen indicators (Cerealia-type, Juglans and Plantago lanceolata; see Figure 4) suggest intensified farming activities in the surroundings of Guèr coeval with sweet chestnut cultivation. These land transformations have also been reported for other lowland sites in the Insubrian region (Gobet et al., 2000; Hofstetter et al., 2006; Tinner et al., 1999). Hazel (Corylus avellana) also experienced an important expansion at the start of sweet chestnut cultivation (6–31%), until ca. 1100 cal. yr BP, probably colonizing forest clearings.

Despite a clear decline in regional fire activity (BCHAR, microscopic charcoal; Figures 3a and 4) after the establishment of Castanea sativa at ca. 1800–1600 cal. yr BP (AD 150–350), which is in agreement with previous studies (e.g. Tinner et al., 1999), local fire impact remained high for several centuries, until 1100–1000 cal. yr BP (AD 850–950; Figures 3 and 5). This phase of high local burning partially overlapped the most abrupt increase of Castanea sativa in the whole record, starting already in the early Middle Ages at ca. 1200 cal. yr BP (AD 750; Figure 4). Considering the profile of the valley with large glacial terraces separated by steep slopes, sweet chestnut trees were probably planted first on the valley floor next to the main villages, and second, on gentle slopes around the secondary mountain settlements with their surrounding meadows and cultivated fields. The diffusion of the Castanea sativa stands to the steeper stretches of the hillsides outside the glacial terraces was a huge task that was carried out over a rather long time. It probably required the use of fire or controlled burning for the clearance of forest and to manage the cultivated land. This may explain the persistence of local moderate to high fire impact for several centuries after the first introduction and the establishment of sweet chestnut trees in our marginal high-altitudinal site.

The second maximum of Castanea occurred around 800–500 cal. yr BP (AD 1150–1450), coinciding with the flourishing use of chestnut for staple food production during the late Middle Ages (ca. 950–450 cal. yr BP = AD 1000–1500; Conedera et al., 2004). The apex development of sweet chestnut orchards coincided with planting in marginal areas (far from human settlements, on steep slopes, at high altitudes) and was probably caused by increasing demographic pressure. Our data show a marked drop in fire activity by 1000 cal. yr BP that persisted for several centuries (Figures 3 and 5), with fire frequency decreasing to 5–6 fire episodes per millennium. Low fire incidence occurred together with the largest diffusion of sweet chestnut orchards, which brought the need for its protection (fire exclusion) while forest clearance became a minor priority. In agreement with the growing dominance of sweet chestnut stands over extensive areas and the importance of chestnut production for the local human population, setting fire was regulated by a number of medieval bylaws (Conedera et al., 2007). Fire reconstruction from Guèr and other sites in the Ticino (e.g. Tinner et al., 1999) reflects this change in the local land use, showing a very important decrease in fire frequency until the last century.

Implications of sweet chestnut cultivation decline for nature management in the southern Alps

At present, the Insubrian vegetation is largely the result of several millennia of land use. The establishment of Castanea sativa stands was actually promoted by Roman and medieval people through intensive land management involving the use of fire in combination with farming activities. Thus, recurrent and probably intense fires set by local inhabitants (Figures 4 and 5) prepared, accompanied and finally favoured the introduction and establishment of sweet chestnut over other tree species (Conedera et al., 2004; Krebs et al., 2012). As other disturbance-adapted tree species such as Betula and Alnus glutinosa would have also benefited from that moderate fire activity, local inhabitants might however have actively eliminated their stands to facilitate the establishment of Castanea sativa and the expansion of pasturelands and crop fields (Figure 4). Following this phase of cultural landscape ‘design’, intensive land use including grazing, crop cultivation and fire management allowed the persistence of the sweet chestnut landscape for the last millennium. Consequently, present coppices and orchards of Castanea sativa represent a very important bio-cultural heritage, including several centuries-old giant sweet chestnut trees (Krebs et al., 2012; Figure 1). Nevertheless, due to the prolonged state of abandonment and the absence of cultural inputs during the last decades, sweet chestnut orchards are being invaded by other tree species, with a general increase in canopy cover and a decrease in highly diverse open areas, what leads to a more homogeneous forest mosaic and a dominance of mixed stands (Conedera et al., 2001). In the short term, this causes negative effects on forest functions in terms of landscape architecture and ecological value (Muster et al., 2007; Schiess and Schiess-Bühler, 1997).

The accumulation of biomass in abandoned sweet chestnut stands increases the risk of forest fires, as postulated by Arnaud et al. (1997) and Maltoni et al. (1997) and as recorded in sedimentary high-resolution records from the Insubrian region by Tinner et al. (1998). This trend towards more frequent and severe wildfires may favour species tolerant to fire, stimulating the post-fire resprouting of sweet chestnut stools (Delarze et al., 1992). Previous palaeoecological research in the southern Swiss Alps showed that Castanea sativa is favoured by moderate fire activity, but it remains unknown how this species could respond to high fire activity in the long term (Tinner et al., 2000). As a consequence, it could be outcompeted by other plant species even more adapted to fire disturbance if fire regime intensifies in the near future (including alien invasive species; Maringer et al., 2012). The ability to predict both the post-cultural development of abandoned sweet chestnut groves and the potential role of enhanced fire regimes is crucial for future decision-making policy concerning the management and protection of forests and the environment. Castanea sativa stands contain, in fact, more historical, cultural and ecological elements than most other forest stands in the Alps, so that the development of new concepts for their sustainable management is particularly worthwhile (Conedera et al., 1997). From a long-term perspective, it seems that the persistence of this valuable cultural landscape is associated with intensive human management involving moderate fire activity, livestock husbandry and probably selective cutting of trees different from Castanea sativa. Changes in this disturbance regime including fire fighting, reductions in forest cuttings and/or grazing suppression will sooner or later lead to the disappearance of this cultural element.

In summary, ecological restoration efforts largely depend on the target. Direct interventions aimed at recovering the ecosystem prior to the onset of severe human disturbances may apply for protected areas (e.g. National Parks; Valsecchi et al., 2010), where naturalness is the goal, but are anachronistic for managed landscapes, as important environmental changes have occurred during the last millennia (Jackson and Hobbs, 2009). In this latter sense, maintaining and restoring cultural landscapes such as sweet chestnut orchards interspersed with highly diverse grasslands (also a legacy of millennia of land use; Colombaroli et al., 2013; Colombaroli and Tinner, 2013) would contribute to the survival of a unique culture and a highly diverse system which is by far more desirable (e.g. Obrist et al., 2011; Jackson and Hobbs, 2009). This would also be a way of taking into account in ecosystem management the deep historical, cultural and long-lasting ecological imprints that millennia of human activity has left on many European areas.