Plant Ecology

, 201:723

Effects of fire on the vegetation of a lowland heathland in North-western Italy


    • Department of Biological SciencesUniversity of Illinois at Chicago

DOI: 10.1007/s11258-008-9459-1

Cite this article as:
Borghesio, L. Plant Ecol (2009) 201: 723. doi:10.1007/s11258-008-9459-1


This study focuses on the effect of fire on lowland heathlands at the extreme southern edge of their European distribution (Vauda Nature Reserve, NW Italy). Forty-nine plots (50 m radius) were surveyed between 1999 and 2006. Each year, fire occurrences were recorded and per cent cover of four vegetation types (grassland, heath, low shrubland, and tall shrubland) was estimated in each plot. Vascular plant species richness was also recorded in 255, 1 m2 quadrats. After a fire, grassland vegetation expanded, but then declined rapidly as heath and shrubland recovered: 7 years after a fire, tall shrubland encroached on to more than 40% of the plots, and grassland declined from 50% to 20% cover. Between 1999 and 2006, Betula pendula shrubland greatly expanded, while grassland decreased over most of the Reserve, even where fire frequency was high. Tall shrubland had low plant diversity and was dominated by widespread species of lower conservation value. By contrast, early successional vegetation (grassland and low shrubland) had higher richness and more narrowly distributed species, indication that the development of tall shrubland causes significant species loss in the heathland. Italian lowland heathlands are characterized by high rates of shrubland encroachment that threatens both habitat and species diversity. Burning frequencies of once in 3–6 years seem appropriate in this habitat, but burning alone might not suffice without actions to increase herbivore grazing.


Betula pendulaBirchCalluna vulgarisConservationFireHeatherShrubland


Lowland dry heathlands are mosaics of dwarf ericaceous shrubs (among which Calluna vulgaris (L.) Hull is the commonest species), intermixed with patches of grassland and pioneer trees. Although they occur on acidic, nutrient-poor soils that check the growth of woody vegetation, lowland heathlands are almost never a true climax, but rather an anthropogenic formation maintained by man through cutting and burning the original woods (Ellenberg 1988). Thus, heathland maintenance depends on a mix of human activities, among which grazing by domestic herbivores, burning, and mowing are the most commonly practiced (Webb 1986, 1998). These activities prevent the development of woody vegetation by physically damaging plants and restricting the accumulation of nutrients in the soil. In this way, the heathland remains in an early successional stage dominated by dwarf shrubs and grasses.

Heathlands characterized the landscape of central and western Europe at least since the late Neolithic, or c. 5,000 years ago (Webb 1986; Ellenberg 1988). However, the abandonment of traditional management practices and developments in agricultural techniques that allowed the expansion of crops on infertile soils caused a spectacular decrease of heathlands across the continent. Conservative estimates of loss in western Europe range between 60% and 95%, and the remaining heathlands are severely fragmented and threatened by a variety of factors (Rebane and Wynde 1997). The conservation of lowland dry heathlands is one of the key targets of the environmental policy of the European Union, as stated in the European Habitat Directive (92/43/CEE).

One of the conservation issues in European heathlands is woodland encroachment, which causes the disappearance of the majority of the species of conservation concern (Gimingham 1992; Price 2002). When burning and grazing are discontinued, the expansion of woody species is an inevitable natural process that, in Italy, is initiated by pioneer trees such as Betula pendula Roth. and Populus tremula L. In time, pioneers facilitate the invasion of late successional species such as Quercus spp, and Robinia pseudoacacia L., an invasive tree introduced from North America (Sindaco et al. 2003).

Heathland encroachment was studied intensively in northern Europe (e.g. Gimingham 1992; Bullock and Webb 1995; Mitchell et al. 1997; Snow and Marrs 1997; Marrs et al. 2000; Pakeman et al. 2002; Manning et al. 2004, 2005), but little research was done in southern Europe, where heathlands reach the edge of their distribution. Lack of research on southern European heathlands is particularly problematic, since they are floristically different from similar habitats in northern Europe, and contain rich floras as well as unique habitat types (Devillers et al. 1991; Webb 1998). Available information suggests that encroachment is faster in southern Europe than in the northern part of the continent (Bartolome et al. 2005). Thus, the last heathland remnants in southern Europe are particularly threatened and in need of management, which, however, currently cannot be based on a sound base of scientific information as is possible in northern Europe.

This paper has two aims. First, data collected between 1999 and 2006 are used to illustrate patterns of fire recurrence and their effects on vegetation in a poorly known habitat, the lowland heathland of North-western Italy. Second, plant species richness is compared across habitat types to assess the impact of periodic burning (or lack of burning).

Material and methods

Study area

The Vauda Nature Reserve (45°15′N 7°39′E) was created in 1993 to preserve a fragment of the “Po basin heathland”, a habitat unique to northern Italy, located at the extreme southern edge of the range of European heathlands (CORINE code 31.229, (Devillers et al. 1991)). The total heathland area in the Reserve is 910 ha, divided into four sectors (Fig. 1; sector 1: 234 ha; 2: 116 ha; 3: 276 ha; 4: 284 ha).
Fig. 1

Map of the study area, showing the borders of the Vauda Nature Reserve (hatched line), the extent of the heathland habitat within the Reserve (bold lines), and the areas burnt between 1996 and 2006. Darker shades correspond to higher burning frequency. The small circles mark the 49 plots; numbers (1–4) refer to sectors described in the text

The Reserve is located on a fluvio-glacial terrace datable to the Mindel glaciation (c. 400,000 years ago). Soils are ancient and leached, with fairly low pH (4.8 ± 0.1, n = 14, L. Borghesio, unpublished data) and high clay content. Elevation ranges from 270 to 440 m. The climate is Prealpine, with marked yearly temperature excursions (>20°C difference between the hottest and the coldest month), and a relatively well-spread rainfall, with no arid months (Sestini 1957). Rainfall averages 1100 mm/year, with maxima in May and October and minima in January and July. Average yearly temperature is 12°C, with minima (1.3°C) in January, and maxima (22°C) in July (Biancotti and Bovo 1998).

The Vauda is a fragment of a much larger heathland that occupied the area until the beginning of the 20th century. Since then, most of the heathland has been converted to agriculture and settlements. The Nature Reserve was spared because it has been used as military training area since the early 19th century. Military activities continue to this day.

The study area has a rich flora, with 750 species of vascular plants reported in the early 20th century (Ferrari 1913). However, many species have not been observed for many years, and the flora of the Reserve is impoverished compared with one century ago (L. Borghesio, unpublished). Evidence for widespread biodiversity erosion was also reported for birds, butterflies, and dung beetles (Borghesio 2004) and was linked to a variety of causes, including conversion of the heathland to agriculture, military activities, change in traditional management practices, and encroachment of woody plants and allochthonous herbs.

The Vauda has been used for centuries as a pasture for cattle and sheep, and grazing is still practiced. However, grazing intensity has decreased in recent decades. In particular, sector 2 of the Vauda has not been grazed since 2003. Three herds of cattle (totaling c. 200 animals) graze sectors 1, 3, and 4 from late March to early July. One herd of c. 500 sheep also grazes sectors 3 and 4 in May. Fires occur usually in winter or early spring (December to April), and are probably lit by the shepherds who graze their animals in the Vauda, but no clear information is available on this, since burning natural vegetation is illegal, and those who lit the fires do so anonymously.

Field methods

Burnt areas were mapped each year between 1999 and 2006. Data were collected in late May, at a time of the year when no fire occurs and burnt vegetation is easily identifiable. Between 1999 and 2001, burnt areas were mapped manually on 1:10,000 maps, with an estimated resolution of c. 50 m. Beginning in 2002, a GPS was used to draw the boundaries of burnt areas with 10 m horizontal resolution. Surfaces were calculated with ArcInfo 8.3 with the Spatial Analyst extension (©ESRI 2003).

Vegetation development was studied in a network of 49 circular plots (50 m radius) located on randomly selected nodes of a 250 m-sized grid superposed to the study area. Positioning error due to GPS selective availability (Adrados et al. 2002) resulted in some slight irregularities in the placement of the plots (Fig. 1). The plots were visited each year between 15 May and 15 June and percent cover of the following four vegetation types was visually estimated:
  • Grassland: areas with prevailingly grassy vegetation. Common species include Molinia arundinacea Schrank, Brachypodium pinnatum (L.) Beauv, Danthonia decumbens (L.) D.C., and Festuca tenuifolia Sibth. (Poaceae); Genista tinctoria L. and Genista germanica L. (Fabaceae); Carex panicea L. and C. pilulifera L. (Cyperaceae); and Potentilla erecta (L.) Rauschel (Rosaceae). These species are also found in the following vegetation types, which differ from open grassland due to the high cover of particular species

  • Heath: areas with 40% or more cover of Calluna vulgaris

  • Low shrubland: with at least 40% cover of shrubs and low trees (<1.5 m high), such as Betula pendula, Populus tremula, and Frangula alnus Miller

  • Tall shrubland: same species as low shrubland, but average height >1.5 m and per cent cover of woody plants 70% or more

Accuracy of visual estimates was checked with aerial (1999 and 2000) and satellite photos (2004) of the study area. When necessary, plots were revisited to improve the accuracy of the estimates.

Vascular plant richness within each of the four vegetation types was estimated by surveying 1 m2 quadrats between 21 June and 9 July 2002. Two quadrats were located 10 and 50 m from the center of each of the 49 main plots along three compass directions (north, southeast, and southwest). Thus, six quadrats were located in each of the 50-m radius plots; 39 quadrats were discarded as they fell on ponds, roads, mown meadows, or Robinia pseudacacia woods growing on former cultivated land. The remaining 255 quadrats were unequivocally assigned to one of the four vegetation types and categorized as burnt or not burnt during the current year. There were no quadrats in the burnt-heath category, because fire strongly reduces Calluna, thus changing the vegetation to the burnt-grassland type. All the species of vascular plants whose aerial parts fell inside the quadrat were recorded to estimate species richness at the quadrat level. In total, 122 species of native vascular plants were observed; eight allochthonous species (73 records, 2.2% of total) were excluded from the analyses. The range size of each species was estimated by tallying their presence in the 55 administrative provinces of the Alps according to Aeschimann et al. (2004). Use of this reference dataset is justified as the Vauda lies at the foot of the Alps, within the area covered by Aeschimann et al. (2004). As range size and population abundance are positively correlated (Gaston et al. 2000), species with smaller ranges are assumed to have a higher conservation value.

Statistical analyses

The data allowed the fire history of the 49 circular plots to be followed between 1999 and 2006. Additional information allowed the detection of fires that occurred in 1997 and 1998 in 12 of the 49 plots. The resulting database contained 353 yearly estimates of vegetation cover in plots whose fire history was known. Plots were grouped according to the number of years elapsed from the last fire (i.e., group 0 contains plots that burned on the current year, group 1 those that burned in the previous year, etc). Plots that had not burned for seven or more years were lumped in a single group. Due to repeated burning, some plots went through the same development stage multiple times (i.e. a plot that burned three times on alternate years went through stages 0 and 1 three times). To avoid overrepresentation of some plots in particular development stages, multiple estimates of the same plot in the same stage were averaged, resulting in a final dataset comprising 238 estimates. Sample size varied between 17 (age classes 5 and 6) and 44 plots (age class 2).

General linear models (GLM) were used to study changes of per cent vegetation cover across the time elapsed since the last fire. Data were arcsin square-root transformed to achieve normality. Spatial autocorrelation was controlled by regressing the vegetation cover estimates on Easting coordinate and using the residuals of this regression in the analyses. Moran’s I values showed that residuals of this regression did not have any significant residual spatial autocorrelation. Simple (YEAR) and quadratic terms (YEAR2) were entered as continuous covariates in the GLM to account for non-linear effects of time. Plot identity was modeled as a random factor to account for the non-independence of repeated estimates done in the same plot on multiple years.

To better describe changes from one year to the following, per cent variation of each vegetation type in each plot between pairs of subsequent years was estimated, and Spearman’s rank order correlations were calculated between all pairs of vegetation types. This analysis highlights spatial relationships between vegetation types, i.e., which vegetation type expands when another decreases due to fire or succession.

To show how vegetation changed in the four sectors of the study area over the entire period, two-way repeated measures ANOVA (sector X year) was used to compare per cent cover (arcsin square-root transformed) of the four vegetation types in the 49 plots from 1999 to 2006. When two-way ANOVA was significant, each sector and vegetation type was analyzed separately with one-way repeated measures ANOVA to highlight significant variations more precisely.

One-way ANOVA was used to compare average species richness and range sizes per 1 m2 quadrat across the four vegetation types and between burnt and unburnt quadrats. Preliminary analyses showed a spatial trend of species richness in the study area: quadrats in the westernmost sector are more species-rich than those in the eastern sectors. Slopes of regression lines did not differ among the four vegetation types (ANCOVA, interaction species richness X coordinate, F7,247 = 0.2, P = 0.86). Thus, log(Easting coordinate) was entered as a covariate in the analyses to control this spatial trend. Untransformed variables conformed to normality and homogeneity of variances.

Throughout the paper, data are presented as mean ± SE and the threshold for statistical significance was set at P ≤ 0.05. Statistical calculations were done with Statistica 6.1 (©Statsoft 2003). Taxonomic nomenclature follows Pignatti (1982).


Patterns of fire recurrence

During the period considered, 25 fires occurred in the study area. The average extent of a fire was 47 ± 12 ha. The four largest ones (>100 ha each) made up 57% of the total area burnt between 1999 and 2006. Fires whose date were known occurred between 15 December and 16 April (N = 7). Fires usually started on days with strong eastward wind, and burned the heathland but stopped at the edges of dense Betula or Quercus woodland.

During the eight-year period, 53% of the study area burned one or more times. However, fires were unequally distributed over the four sectors (Figs. 1 and 2). More than 80% of sector 4 burned one to six times, while 71% of sector 1 never burned.
Fig. 2

Percentage of area burnt in the four sectors (totals of 1999–2006)

Effects of fire on vegetation development

All vegetation types were significantly affected by fire. Heath cover increased after a fire (GLM, F50,187 = 15.1, P < 0.0001). Both linear and quadratic trends were highly significant (YEAR, F1,48 = 68.9, P < 0.0001; YEAR2, F1,48 = 68.9, P < 0.0001), suggesting that the expansion of Calluna slows about 3–5 years after a fire (Fig. 3a). Grassland declined sharply and linearly after fires (GLM, F50,187 = 21.8, P < 0.0001; YEAR, F1,48 = 46.9, P < 0.0001; YEAR2, F1,48 = 0.1, P = 0.85; Fig. 3b). Low shrubland initially increased and then decreased, as shown by the significant effect of YEAR2 (GLM, F50,187 = 9.3, P < 0.0001; YEAR, F1,48 = 16.2, P < 0.0001; YEAR2, F1,48 = 26.3, P < 0.0001; Fig. 3c). Tall shrubland increased over time (GLM, F50,187 = 23.7, P < 0.0001). Only quadratic terms in the model were highly significant (YEAR, F1,48 = 2.5, P < 0.11; YEAR2, F1,48 = 58.6, P < 0.0001), suggesting an accelerating spread of tall shrubland in late succession years (Fig. 3d).
Fig. 3

Changes in per cent vegetation cover with increasing time after a fire. a: heath; b: grassland; c: low shrubland; d: tall shrubland

Considering change in percent cover between pairs of consecutive years (Fig. 4), grassland had a large, negative correlation with heath and low shrubland, suggesting that the expansion of these two vegetation types occurs at the expense of grassland. Low shrubland was negatively correlated with tall shrubland. Heath had a marginally significant positive correlation with low shrubland, suggesting that these two vegetation types follow parallel trends of development after a fire, and that shrubland has little ability to encroach Calluna-dominated areas.
Fig. 4

Spearman rank correlation coefficients between differences of per cent cover of the four vegetation types in plots assessed on two consecutive years (N = 249 estimates). *** P < 0.01; * P = 0.1

Between 1999 and 2006 there were marked changes in the study area, but these differed in amplitude and direction among the four sectors. Heath cover changed significantly, but the four sectors followed contrasting trajectories (Two-way repeated-measures ANOVA, Year, F1,45 = 19.0, P < 0.0001; Sector, F3,45 = 13.0, P < 0.0001; interaction F3,45 = 16.8, P < 0.0001). Heath increased in sectors 2 and 3, which burned in 1999, decreased in sector 1, which burned rarely (Fig. 2), and remained on low levels in sector 4, which burned frequently (Table 1).
Table 1

Per cent vegetation cover in the four sectors in 1999 and 2006.

Vegetation type










19.8 ± 5.3

14.8 ± 3.6





5.1 ± 2.4

33.4 ± 4.7





8.4 ± 2.6

13.6 ± 3.4





1.3 ± 0.2

1.5 ± 0.5






59.6 ± 5.4

39.9 ± 5.3





76.0 ± 3.9

20.4 ± 4.8





45.8 ± 5.4

16.8 ± 4.5





37.2 ± 3.0

30.6 ± 4.8




Low shrubland


8.2 ± 2.0

17.2 ± 4.2





14.9 ± 3.4

15.3 ± 1.9





31.9 ± 4.4

19.9 ± 3.4





51.2 ± 4.4

44.4 ± 4.6




Tall shrubland


12.3 ± 4.4

28.1 ± 6.3





3.9 ± 1.5

30.9 ± 3.8





13.9 ± 3.9

49.7 ± 5.5





10.4 ± 5.4

23.6 ± 6.9




Table gives means ± SE and the results of one-way repeated-measures ANOVA

Grassland cover decreased significantly in sectors 1, 2, and 3 between years (Table 1), although the magnitude of change differed (Year, F1,45 = 99.5, P < 0.0001; Sector, F3,45 = 5.0, P = 0.004; interaction F3,45 = 10.5, P < 0.0001).

At the scale of the entire study area, low shrubland cover did not change across years, but this concealed contrasting trends in the sectors (Year, F1,45 = 0.3, P = 0.57; Sector, F3,45 = 18.7, P < 0.0001; interaction F3,45 = 4.0, P = 0.01). Low shrub increased in sector 1, had a nearly significant negative trend in sector 3, and did not change in sectors 2 and 4 (Table 1).

Tall shrubland expanded significantly in the entire area (Year, F1,45 = 36.8, P < 0.0001; Sector, F3,45 = 1.5, P = 0.23; interaction F3,45 = 1.0, P = 0.40), although the increment was only marginally significant in sector 2 (Table 1).

Effects of fire and vegetation type on vascular plant richness

Burnt and unburnt quadrats did not differ either in species richness (ANOVA, F1,207 = 0.1, P = 0.71), or in the average range size of the plants (F1,207 = 0.04, P = 0.83).

Species richness differed between vegetation types (ANOVA, F3,250 = 9.3, P < 0.0001), which were ranked as low shrubland > grassland > tall shrubland > heath (13.1 ± 0.4, 11.0 ± 0.3, 9.8 ± 0.4, 9.7 ± 0.4 species/quadrat, respectively). Multiple comparisons showed significant differences between low shrubland and the other vegetation types (Tukey test, all P < 0.001) and a marginally significant difference between grassland and tall shrubland (Tukey test, P = 0.1). Range size differed between vegetation types (ANOVA, F3,250 = 10.7, P > 0.0001): species observed in grassland (49.9 ± 0.2 provinces/species) and low shrubland (49.7 ± 0.3) had smaller ranges than those recorded in tall shrubland (51.6 ± 0.3) (Tukey test, all P < 0.001).


This work is one of the first analyses of the effects of fire in an Italian lowland heathland. In this study, vegetation followed a predictable succession after a fire. Grassland increased in recently burnt areas while heath decreased, but, as time progressed, Calluna-dominated heath grew and reinvaded the grassland. Low shrubland initially expanded in areas occupied by grassland immediately after a fire (Fig. 4), but, after 3 years, it started to decrease, as woody plants grew above 1.5 m and entered the tall shrubland stage. This pattern has been found in other studies (e.g., Clement and Touffet 1981; Mallik and Gimingham 1983; Hobbs and Gimingham 1984; Forgeard 1990), but a major difference with more northern latitudes is that, in the Vauda, shrubland expands more rapidly after a fire. Bullock and Webb (1995) reported an increase of Betula pendula shrubland from about 8% to 15% 8 years after a severe fire in southern England, while in this study tall shrubland expanded to more than 40% cover in seven years (Fig. 3). In the UK, rates of shrubland encroachment similar to those found in Italy are usually observed only after 35–40 years (Marrs et al. 1986; Rose et al. 2000). Plant diversity was low in tall shrubland, and vascular plants observed in this vegetation were more widespread in the Alps than those found in grassland. This suggests that rapid expansion of woody species reduces both habitat and species diversity, and is probably the most important direct threat to the heathlands of Southern Europe.

Variation of Calluna vulgaris cover between consecutive years was strongly and inversely correlated with that of grassland, but much less so with tall and low shrubland, suggesting that the increase of heather after a fire occurs mainly at the expense of grassland. The leveling out of Calluna cover after the third year suggests a competitive equilibrium between shrubland and Calluna. Mature Calluna strongly suppresses other plants, as is confirmed by the low species richness of Calluna-dominated quadrats. In northern Europe, Calluna suppresses the establishment of competitors for 20 or more years (Watt 1955; Gimingham 1972). As this study lasted only 8 years, it is not clear if that would be the case also in the Italian heathlands, or if the encroachment of Calluna stands by shrubland will occur in shorter times. In any case, this study showed that in the Italian heathlands the dominance of Calluna is less marked than in northern Europe, where this species often blankets large expanses of land. On the contrary, Italian heathlands are mosaics where grassland and shrubland occupy 70% or more of the area (Table 1). This makes Italian heathlands constitutionally more prone to encroachment by woody species such as Betula pendula or Populus tremula.

Between 1999 and 2006 there were marked changes in the amount and distribution of vegetation types. The most consistent change was the increase of tall shrubland in all sectors, in spite of strong differences in fire frequency. If the increase of shrubland in unburnt areas is easily explained, that observed in frequently burnt sectors might be due to two factors. First, fire stimulates seed germination and seedling establishment by providing patches of bare soil where competition is low. As a result, Betula pendula and other woody species often increase in burnt areas (Atkinson 1992; Bullock and Webb 1995). Second, due to reduced fuel abundance, fire intensity decreases in frequently burnt areas, and low-intensity fires spare many woody plants, especially when they have developed a protective layer of bark around their foot (Mallik and Gimingham 1983). The rapid growth of woody plants observed in this study enhances their chances of reaching a stage where they are safe from low-intensity fires. Thus, in the Vauda, both the absence of fires and frequent, low-intensity fires appear to stimulate shrubland expansion.

How could Italian heathlands persist over time? Ancient maps show that the Vauda has been a heathland since the beginning of the 19th century, and, in all likelihood, heathland existed there since late Middle Age. This is remarkable, because Italian heathlands rest on a fragile equilibrium between fires and the rapid encroachment by shrubland and woodland. Most European heathlands are created and preserved by human action, but the rapid vegetation succession observed in this study suggests that Italian heathlands will require a particularly intense and frequent management. Frequent fires have probably been important to maintain Italian heathlands over the centuries, but this study also showed that shrubland expands even in frequently burnt areas. Indeed, other research in the area confirms that the short-term effects of fire in controlling woody plants are unclear (Ascoli et al. 2006). This suggests the importance of other factors, especially of herbivore grazing, to complement the action of fire in checking the growth of woody plants.

Conclusion: the conservation of lowland heathland in Italy

This study suggests that fire frequency is a critical variable in the management of Italian heathlands, since both too high and too low frequencies will stimulate shrubland encroachment. Burning once in 3–6 years appears to be the most appropriate frequency, as this will maximize the amount of heath, and at the same maintain species-rich grassland, while controlling the expansion of tall shrubland dominated by Betula pendula. This frequency is markedly shorter than the 15–20 years usually recommended in the British Isles (Price 2002). Thus, attempts should be taken to reduce fire frequency in the western part of the Vauda, while the opposite target should be pursued in the eastern sectors. At the same time, this study suggests that burning alone might not be sufficient to stop shrubland expansion. Increased and more prolonged grazing might also be an essential management action.

Considering the large size of the Vauda, fire management will likely pose a huge logistic and financial challenge. Intervals of 3–6 years between fires will require burning a total of 150–300 ha in each year, which, considering the low amount of resources available for environmental management in Italy, is probably well beyond the current possibilities of the Reserve administrators. The likely outcome of this lack of resources will be a foreseeable large decrease, if not the total disappearance, of the last remnants of lowland heathland in Italy.


I thank Paola Laiolo and Roberto Sindaco for help in the field and commenting on the first draft of this paper, and Ente di Gestione delle Aree protette del Canavese for supporting this survey. Partial funding was provided by Regione Piemonte and the European Union through the project Interreg IIIA ALCOTRA, “Conservation and management of flora and habitats of the south-western Alps”, which was coordinated by IPLA (Istituto per le Piante da Legno e l’Ambiente) and Ente di Gestione delle Aree Protette del Canavese.

Copyright information

© Springer Science+Business Media B.V. 2008