Silver fir (Abies alba Mill.) is able to thrive and prosper under meso-Mediterranean conditions

The potential ecological envelope of silver fir ( Abies alba Mill.) based on its present distribution suggests a high suitability for moist rather than warm and dry environments. This contrasts with paleoecological evidence reporting its former presence at low elevations under meso -Mediterranean conditions. In this study, we evaluated the growth performance of silver fir at low elevation (20 – 60 m a


Introduction
European climate is getting warmer and drier (IPCC 2013) and the frequency and intensity of extreme climatic events and related disturbance regimes are changing (Bowman et al. 2014;Seidl et al. 2014;Seidl et al. 2017).This has a potentially large impact on forest dynamics, tree species distributions and the survival probability of trees in particular (Allen et al. 2010; van Mantgem et al. 2009), eventually affecting ecosystem services (Scheffers et al. 2016).
In a climate change context, adaptive management strategies are thus needed to increase forest resistance and resilience and to mitigate possible negative impacts (Thurm et al. 2018;Vilà-Cabrera et al. 2018).Different strategies to adapt to climate change exist (Hagerman and Pelai 2018;Madrigal-González et al. 2017), such as increasing the resistance and/or resilience of current forests (i.e.shortening rotation periods and reducing mortality risk or decreasing forest stand density; Rais et al. 2014;Seidl et al. 2011), regulating tree species mixtures by favoring the most suitable ones (Lebourgeois et al. 2013;Pretzsch et al. 2013), or fostering and introducing more drought-tolerant species (Buras and Menzel 2018;Dyderski et al. 2018;Scherrer et al. 2017).Developing adaptive strategies based on the selection of suitable tree species requires an in-depth understanding of their behavior under the expected climatic and other environmental conditions (Keenan 2015).
The debate on the potential role of silver fir (Abies alba Mill.) under a future European climate is very emblematic in this respect (Vitasse et al. 2019).The current distribution of silver fir ranges from the Pyrenees to the Carpathians and from southern Italy to Poland, mainly following mountain ranges (Mauri et al. 2016; Fig. 1).In such ecosystems, it represents an important keystone species significantly contributing to biodiversity conservation, stand resistance to disturbances such as wind and insects in mixtures with Norway spruce (Ott et al. 1997;Rüegg 2015), as well as to forest ecosystem services, such as timber production and protection from gravitative natural hazards (Tinner et al. 2013;Wolf 2003).
Yet, there is contrasting evidence regarding its suitability to cope with future climate change (Vitasse et al. 2019).Based on its current distribution, silver fir is traditionally considered well adapted to moist conditions (Bucher 2014) and rather poorly adapted to warm and summer drought conditions (Macias et al. 2006;Rigling and Stähli 2020).As a consequence, the modelled potential ecological envelope based on its present distribution suggests a future contraction of its range due to climate change, describing the species as widely unsuited in a climate-adapted European forestry context (Maiorano et al. 2013;Zimmermann et al. 2016).On the contrary, studies based on paleoecological evidence report its former presence at low elevations from southern Switzerland to southern Italy, reaching the Mediterranean coast of central Italy, featuring summer and annual mean temperatures up to 7 • C higher than in its actual distribution range (Tinner et al. 2013).According to this view, the current absence of silver fir under such warm conditions is mainly due to the long-lasting human pressure since 5000-7000 years ago, which caused the decline of this latesuccessional and strongly fire-and browsing-sensitive species (Carcaillet and Muller 2005;Di Pasquale et al. 2014;Tinner et al. 1999).Combining this paleoecological evidence with process-based modelling, some authors postulated a greater-than-expected climatic tolerance of silver fir, especially in terms of higher temperatures (Henne et al. 2013;Henne et al. 2015;Ruosch et al. 2016).
Dendroecological studies highlighting an increase in the average annual ring width of silver fir during the warming period from 1990 to 2010 confirmed the adaptation of the species to higher temperatures (Büntgen et al. 2014;Boše ľa et al. 2018), suggesting a higher than expected drought tolerance (e.g., Vitali et al. 2017).Other studies, in contrast, suggested sensitivity to Mediterranean summer droughts (Gazol et al. 2015).Unfortunately, only few and partially outdated studies described growth and regeneration dynamics of silver fir stands growing under low-elevation Mediterranean conditions (Cesarini 1940;Cortini Pedrotti 1967).
In this paper, we aim to fill this gap by conducting a dendroecological study in a meso-Mediterranean low-elevation silver fir forest that is characterized by a steep geomorphological gradient.We particularly address the following research questions: -Does silver fir develop and regenerate under meso-Mediterranean climatic conditions?-What are the main climatic factors influencing silver fir radial growth?-How may the species respond to a climate getting warmer and drier?

Study area and sampling design
The study area features ca. 12 ha of silver fir-dominated stands located at an elevation from 20 to 60 m a.s.l.along the lower part of three small valleys extending in the north-south direction in Varramista (Tuscany, Central Italy; 43 • 39 ′ 45 ′′ N, 10 • 42 ′ 55 ′′ E; Fig. 1) that were probably reclaimed at the end of the 18th century (Baroni 1810;Giorgi 2009).The forests originate from a plantation dating back to the beginning of the 19th century or even earlier (Capponi 1882;Cesarini 1940;Cortini Pedrotti 1967).Although the provenance of the planted stock is unknown, the historical and botanical descriptions allow us to assume that the trees are Abies alba very likely originating from the Apennines (Susmel 1957;Cortini Pedrotti 1967).No originally planted trees survived to date and current stands originate from natural regeneration (Cesarini 1940) At present, in these stands silver fir grows in mixture with typical broadleaved and coniferous Mediterranean species, such as manna ash (Fraxinus ornus L.), holm oak (Quercus ilex L.), and the strawberry tree (Arbutus unedo L.).Towards the top of the hills, silver fir is increasingly substituted by Turkey oak (Quercus cerris L.), sessile oak (Q.petraea (Matt.)Lieb.), holm oak, Italian stone pine (Pinus pinea L.), maritime pine (P.pinaster Ait.), and other minor species.
The shrubs and herbaceous layers are heterogeneous and characterized by taxa typical of the Abies alba and evergreen and thermophilous oak vegetation types, respectively.The list of phanerogamic flora in Varramista includes 140 taxa, 65% of them belonging to the geoclimatic boreal element, while the Mediterranean species make up only 26%  (Cortini Pedrotti 1967).A few eastern species (Illyria, Pontus and Eurasia) are present, summing up to about 5%.The biological spectrum of Raunkiaer life form types (Raunkiaer 1934) shows a prevalence of hemicryptophytes (H), phanerophytes (P) and geophytes (G) and a low percentage of therophytes (Th) due to the unsuited temperature and humidity conditions of the understory, while chamaephytes (Ch) are in general not frequent in forests (P 30%, Ch 6%, H 38% ,G 19%, Th 7%;Cortini Pedrotti 1967).
According to the Tuscan Regional Hydrologic Service (SIR 2019a, 2019b), annual mean temperature in the study area is 15.3 • C, with a mean temperature of 6.9 • C in the coldest month (January) and 24.5 • C and 24.4 • C in the warmest months (July and August, respectively).Annual precipitation amounts to 867 mm, mainly occurring from late autumn to early winter.The driest month is July (31.5 mm in average) while the wettest is November (114.4 mm).Temperatures are rarely below 0 • C and it hardly ever snows.These conditions were classified by Fratianni and Acquaotta (2017) as hot temperate (Csa after the Köppen-Geiger classification), pointing out the importance of summer arid periods, which usually occur between June and August.
The geology is characterized by sand layers of Pliocenic origin and younger continental deposits, with sparsely emerging clay formations (Cortini Pedrotti 1967;SGR 2006).Soils are acidic Luvisols and have different properties within the investigated area, ranging from the presence of an albic horizon in the slopes and stagnic horizon in the depressions, indicating water flow from the hilltops towards the mesic valley bottoms (Fig. S1; Corongiu et al. 2016;Regione Toscana 2018).
The study design followed the existing geomorphological gradient by selecting three macroplots, going from the depression to the lower and the upper part of the slopes, respectively (Fig. 1).The valley is quite open and all macroplots, including the one in the depression, are rich in woody and herbaceous vegetation of meso-Mediterranean character.This allows us to exclude any significant micro-climatic effect due to the terrain morphology.The east-facing lower slope macroplot is steeper than the west-facing upper slope, with an average slope of 25-30 • and 10-15 • , respectively (Table 1).Within each macroplot, at least 34 adult, dominant and visibly healthy fir trees were selected for coring.In addition, three circular plots of 12 m in radius (corresponding to 452 m 2 in area) were placed in each macroplot (i.e., 9 circular plots in all) for assessing stand structure, including tree regeneration.

Field work
From each selected dominant silver fir tree, two opposite cores at 1.3 m height were taken perpendicular to the slope direction using an increment borer (3-threaded, 40 cm long with a diameter of 5 mm, Haglöf Co.).Each cored tree was georeferenced (x, y coordinate) and characterized by measuring diameter at breast height (DBH), height, height of the crown base, and crown width (average of the four cardinal directions).
In each circular plot, species, DBH, social position, height, and height of the crown base of all trees with DBH > 5 cm were measured.In this study we considered as established regeneration (saplings) individuals with height > 1.3 m and DBH < 5 cm.Their location, species, stem diameter at the ground, and height were recorded.In each plot, the number of seedlings (H < 1.3 m) was assessed for each species using a square frame (1 × 1 m 2 ) placed at distances of 2, 6 and 10 m from the plot center and along the eight cardinal directions, for a total assessed area of 24 m 2 per circular plot.Species diversity was calculated using the Shannon diversity index (Shannon 1948).A selected number (i.e., 49 individuals for the whole study area) of seedlings and established regeneration of silver fir were cut or cored in order to determine the germination year.Finally, light transmittance (ratio of photosynthetically active radiation -PARreaching the forest floor) through the canopy was estimated using a radiation sensor.The measurements were carried out in each macroplot, along 8 cardinal directions, at 12:00 a.m. during a sunny summer day.Field work took place between October 2018 and January 2019, except for PAR measurements which were carried out between July and August 2019.

Dendrochronological analyses
The tree cores were glued on wood supports, dried and sanded with a belt sander using progressively finer grit sizes (150, 320, 600 and 800) in order to increase the visibility of the tree rings.Tree-ring widths (TRWs) were measured at the WSL research campus in Cadenazzo (Switzerland) with a precision of 0.005 mm on a measuring table (DENDROTAB 2003) connected to the software T-Tools Win Pro, both produced by WALESCH Electronic GmbH.
The measured tree-ring sequences were imported in the TSAP Win Professional software v4.5 (RINNTECH, Heidelberg).The matching of the two sequences belonging to the same tree was first visually assessed to identify possible measurement errors.The pairs were then averaged to get a single TRW series for each tree.The TRW series were cross-dated within each macroplot.The quality of this process was evaluated with the average inter-series intercorrelation (Cerrato et al. 2018).The TRW series were then detrended with a spline function and standardized by dividing their values by the corresponding estimates of the spline, creating unitless ring-width index (RWI) series that were pre-whitened using standard procedures to remove autocorrelation (Cook and Kairiukstis 1990).
Next, residual chronologies were calculated for every macroplot by averaging the RWI series belonging to the same macroplot.To enhance the robustness of the signal, sampling depth for the chronologies was set to a minimum of five RWI series.The similarity between chronologies was quantified by calculating their Gleichläufigkeit (Buras and Wilmking 2015;Eckstein and Bauch 1969).

Climatic data
Daily precipitation and daily maximum and minimum temperature data were retrieved from the San Miniato Cimitero meteorological station (102 m a.s.l., 43.68403 • N, 10.83152 • E) of the Tuscan Regional Hydrologic Service (SIR 2019a(SIR , 2019b)).This station was chosen as reference because it is the closest weather station to Varramista (9.8 km away) with long data series for both temperature (since 1951) and precipitation (since 1958).The suitability of San Miniato Cimitero as Missing precipitation data between 1997 and 1999 and missing temperature data in 1955, 1956, 1969, between 1996 and 1999 as well as single monthly gaps in precipitation (2000, 2004, and 2010) and temperature (2000, 2004, and 2013) were filled by estimates from a linear regression with precipitation data from Fattoria Scaletta (22 m a. s.l., 43.70739 • N, 10.83063 • E) and temperature data from Orto Botanico Lucca (16 m a.s.l., 43.84144 • N, 10.51126 • E).Moreover, precipitation data were extended back to 1951 (i.e., by 7 years) using data from Pontedera (14 m a.s.l., 43.66522 • N, 10.63120 • E).These weather stations were chosen considering their distance from Varramista and the availability of data during the measurement gaps for San Miniato Cimitero meteorological station.
The three-month Standardised Precipitation-Evapotranspiration Index (SPEI; Vicente-Serrano et al. 2010) was used as a proxy for drought conditions because it considers the role of evapotranspiration and temperature and is sensitive to short-term events (Abramopoulos et al. 1988;WMO and GWP 2016).Longer periods than three months have been shown not to be suitable in the analysis of drought effects on silver fir (Toromani et al. 2011), whereas shorter periods would not take into account the influence of precipitation and temperature of previous months.Estimates of potential evapotranspiration (PET) were calculated with a modified version of the Hargreaves equation (Beguería et al. 2014;Droogers and Allen 2002;Hargreaves and Samani 1985).

Climate-growth relationships 2.5.1. Correlation and response functions analysis
For the correlation and response function analysis of the RWI, we first detrended the climate series by means of a local regression (Baker 1990;Cleveland et al. 1992).Monthly precipitation, maximum temperature and SPEI data were then related to annual tree growth data with a correlation and response function analysis.Maximum temperature was preferred over minimum temperature because of its higher correlation with tree growth in other Italian silver fir populations (Gentilesca and Todaro 2008;Santini and Martinelli 1991) and the ecological assumption that maximum temperatures should be more important at this site than the mellow (typically above-zero) minimum temperatures.
Since the growth of the current year may be affected by the previous year's climatic conditions and related carbon and nutrients reserves, the time span for the analysis was set from March of the previous year to October of the current growing season.The growing season was assumed to start in March and end in October.This resulted in a total of 60 potential predictor variables (i.e., 20 independent parameters for three climatic variables).Pearson correlation and response function coefficients between variables and chronologies were calculated as described by Fritts et al. (1971) and Fritts (1976), differentiating from the correlation analysis by taking into account the possible multicollinearity of climatic variables.This analysis firstly runs a Principal Components Analysis and then estimates response coefficients using multiple regression (Rozas 2005).
Significance was tested at the level of p < 0.05 for both correlation and response functions, applying a bootstrap technique with 1000 resampling runs.

Superposed epoch analysis (SEA)
The reaction of tree growth to summer droughts was assessed using a SEA.Summer drought was quantified by averaging SPEI values between June and September, thus expressing summer drought intensity (Tan et al. 2015).Years with a value < − 1, indicating moderate to extreme drought conditions (McKee et al. 1993), were selected as drought years, and the behavior of growth during these drought years and up to four years after the drought year was analyzed.Significance was tested using the four previous years as well.The growth reaction was quantified from scaled values: firstly, the mean chronology value was subtracted from the RWIs, and secondly RWIs were divided by their standard deviation.Therefore, growth differences are expressed as standard deviation difference.Significance at p < 0.05 of the growth reactions was tested with a bootstrap technique with 5000 simulations over the whole site chronologies.

Pointer year analysis
For the analysis of the relation between extreme growth and climate, years with growth extremes (pointer years) were selected from the raw tree-ring series.Pointer years were defined at the macroplot level as years where at least 75% of the cored trees showed a relative growth increase or decrease of at least 20% compared to the average of the previous four years.Since similarities in the selected years were high, we decided to unify them all in a single dataset.Climatic conditions during positive and negative pointer years were analyzed qualitatively and compared to mean climatic conditions.

Stand structure
The average length of the TRW series ranges from 54.3 (±9.7, upper slope macroplot) to 61.6 (±13.1) years (lower slope macroplot), with a maximum value of 89 years (Table 1).The average height of dominant trees in the macroplots confirms the existence of a gradient in growing conditions (Figure S3).
The three macroplots feature distinctly different basal area, with the lower slope macroplot having the lowest value of 28.6 (±7.3) m 2 ha − 1 , the depression 35.1 (±2.7) m 2 ha − 1 , and the upper slope the highest with 36.7 (±5.0) m 2 ha − 1 .Silver fir accounts for 99.7% of the basal area in the depression, 82.0% in the lower slope and 71.4% on the upper slope.Deciduous species contribute with a relative basal area of 10.9% in the lower slope, despite their absence from DBH classes >30 cm (Fig. 2), and 18.7% in the upper slope (58.0% of which were >30 cm DBH), respectively.
In every macroplot, the highest tree density is found in the lowest DBH class, i.e. below 5 cm, where shrubs and small trees are present.Silver fir regularly regenerates in all macroplots (Fig. S4).Species diversity is lower in the depression macroplot and in high DBH classes (Table 2), while the relative amount of silver fir increases with DBH, becoming the most frequent species starting from a DBH of 10 cm.
The macroplot in the depression has an overall lower tree density (Fig. S5), is characterized by a one-layered structure of silver fir only and no trees between 10 and 20 cm DBH, and displays with 8.9% the highest fraction of photosynthetically active radiation reaching the forest floor, which is more than twice as much as in the upper (3.2%) and lower slope (3.0%) macroplot, respectively (Table 1).At the established regeneration class (individuals below 5 cm DBH), the Shannon index is highest in the depression (Table 2), where deciduous species such as manna ash (Fraxinus ornus L.) prevail over evergreens.The two slope macroplots show a very similar species structure, with the relative amount of silver fir increasing monotonically with DBH, and Mediterranean evergreens and English holly (Ilex aquifolium L.) in particular dominating the established regeneration stage.This pattern changes to a higher amount of deciduous species, such as sweet chestnut (Castanea sativa Mill.) and Turkey oak (Quercus cerris L.) for larger DBH values (Fig. S4).
In terms of seedlings (i.e., trees < 1.3 m in height) in the depression macroplot, the density of silver fir is twice as high compared to the slope macroplots (Fig. 3).Along the macroplot gradient, seedlings of deciduous species are highly significantly more frequent in the two slope macroplots, whereas Mediterranean evergreens are significantly denser in the upper slope macroplot only, compared to the depression.

Recent climate evolution in the study area
The 1951-1990 time series of annual precipitation and temperature shows no clear increasing or decreasing trends in inter-and intra-annual variability (Fig. 4).Starting in 1990, temperatures tend to increase, in the 2010s reaching a mean decadal value more than 1 • C higher compared to the 1970s, whereas precipitation tends to be lower.As a consequence, SPEI has been decreasing since ca.1990.

Tree-ring width series
The overall average annual ring width is 2.8 (±1.5) mm, ranging from 2.5 (±1.0) mm in the upper slope macroplot, across 2.7 (±1.2) mm on the lower slope macroplot, to 3.3 (±1.5) mm in the depression macroplot (Table 1).At the macroplot level, the tree-ring series show very similar patterns (Fig. 5), with a first phase until about 1980 that is characterized by a slightly decreasing growth trend.Between 1980 and 2000, TRWs experienced an increase in all three macroplots, before decreasing again.
Average series intercorrelations range from 0.62 to 0.64, whereas Gleichläufigkeit values are 79.4% between upper slope and lower slope chronologies, 81.0% between upper slope and depression, and 81.4% between lower slope and depression, respectively.Similarly, the site chronologies show synchronous peaks, although at different magnitudes (Figure S7).

Climatic response functions
Significant correlation coefficients are mostly found in the period between May and September of the current growing season (Fig. 6), even though the previous year's conditions yield significant correlations as well, with differences between the climate variables, as follows.
Temperature features significant negative correlations with TRW for at least three months during the growing season of the current year in all macroplots and for November of the previous year.Previous summer precipitation has a somewhat negative influence on TRW, contrary to current summer, whose influence is distinctly positive for at least two months across all macroplots.This is also highlighted by the highest response function coefficient in the upper slope macroplot for current June.Established regeneration = DBH < 5.0 cm and height > 1.3 m.Lower main layer = DBH > 5.0 cm and < 15.0 cm.Upper main layer = DBH > 15.0 cm.Please note that the Shannon index of the lower main layer in the depression has no standard deviation due to the presence of only one tree.
Three-month SPEI shows positive and very high correlation coefficients during three to five months of the current summer.In contrast, previous summer SPEI is negatively correlated with growth, although a moving response analysis reveals that this influence was higher between 1970 and 2000 than at present (Fig. 7).The SPEI correlation and response coefficients reflect the patterns of the precipitation values, while they tend to be opposite to the coefficients calculated for temperature.
As a general trend, the upper slope macroplot shows a higher number of large coefficients and significant values than the other macroplots, whereas the lower slope is least sensitive to the climate of the current summer season (Fig. S8).
The years selected as drought years (i.e., with summer SPEI values < − 1) are 1957, 1958, 2003, 2006, 2009, 2011, and 2017.The Superposed Epoch Analysis (SEA) indicates a significant growth decrease (p < 0.01) during the drought years, but no significant consequences in the following years, irrespective of the macroplot (Fig. 7).The average ringwidth index (RWI) in these drought years is 0.82 in the depression, 0.84 along the lower slope and 0.86 on the upper slope, respectively (Fig. S7).
Selected negative pointer years based on growth deviations are 1946, 1948, 1955, 1958, 2009, 2011, 2015, and 2017, while positive pointer years are 1953, 1956, 1963, 1984, 1991, 1996, and 2014.In general, the negative pointer years have below-average yearly mean SPEI values (-0.62).In particular, monthly values between May and November are lower than 0.75 in negative pointer years, reaching a minimum of − 1.32 and − 1.31 in June and July, respectively.In positive pointer years, the average SPEI is 0.62, with a maximum in February (1.13) and values from 0.66 to 0.79 between June and November.
In negative pointer years, climate is characterized by a mean temperature of 15.9 • C and a mean precipitation sum of 683 mm.Pointeryear temperature is thus comparable with the overall mean temperature in Varramista, but annual precipitation is more than 20% lower than the average.In negative pointer years, monthly precipitation exceeded the threshold of 100 mm in December only, i.e. outside the vegetation period, and reached the lowest value in July, with less than 20 mm.February and March were wetter than average, while summer and autumn were drier.In these years, the dry period starts almost two months earlier with respect to a standard season (when it usually spans from late June to August) and extends from May to August, which is more than twice as long as the average duration of a usual dry season.
Positive pointer years featured a mean annual precipitation of 1041 mm (i.e., 20% higher than the average) and a number of very moist months with rainfall exceeding 100 mm (i.e., January, September, October, and November).In these years, spring was wet and the dry period was only half as long with respect to the overall mean (e.g., dry July only).In addition, mean temperature amounted to 14.8 • C, i.e. 1.1 • C lower than during negative pointer years.

Stand structure and regeneration dynamics
In Varramista, silver fir grows and regenerates together with Mediterranean species such as holm oak (Q.ilex L.) as well as thermophilous  S6. deciduous species such as Turkey oak (Q.cerris L.).This confirms the suitability of the species for such environmental conditions as reported by several palynological studies and models (Bellini et al. 2009;Di Pasquale et al. 2014;Tinner et al. 2013).
Nevertheless, regeneration capacity and growth performance of silver fir clearly follow a geomorphological gradient.In comparison to the slope macroplots, in the depression the stand structure is nearly single-layered and features a higher silver fir density in the seedling class (Fig. 3) and at DBH values above 40 cm (Fig. 2).Although no detailed information exists on the effective impact of the 2015 windstorm, which was concentrated on the stands of the valley bottom, we can assume that the event highly influenced the stand structure of the depression macroplot, which was open enough to allow for the development of rather light-demanding deciduous species, such as manna ash (Fraxinus ornus

L.).
For the highest DBH classes, silver fir densities tend to decrease (Fig. 2) and species diversity increases (Table 2) from the depression towards the slope macroplots, in particular the macroplot on the upper slope, which has a multi-layered structure and includes deciduous tree species that have a DBH larger than 30 cm, such as sweet chestnut and Turkey oak.The differences between the upper and lower slope macroplots may be due not only to their position on the slope but also to their aspect.In fact, west-facing slopes have been shown to usually host more dry-adapted vegetation communities than east-facing ones (Bennie et al. 2008).In the established regeneration stage, the dominance of Mediterranean evergreens in the slope macroplots mostly refers to the high densities of the shade-tolerant English holly (Ilex aquifolium L.), which was favored by the lower light transmittance of these stands.
The observed differences in stand structure and tree growth as a function of geomorphology are in line with previous studies on the susceptibility of the species when growing at low elevations under Mediterranean climate conditions (Henne et al. 2013;Henne et al. 2015;Tinner et al. 2013).In our case, however, Turkey oak played a much more prominent role than holm oak, which may be related to the broad use of Q. cerris for coppice management in this climatic zone (Fabbio and Cutini 2017).

Trends in tree-ring growth
The long-term mean tree-ring widths in Varramista are comparable for similar time spans with the growth rates of silver fir in sub-Mediterranean and Mediterranean thermophilous mixed oak forests (e. g., Battipaglia et al. 2009;Lebourgeois et al. 2010;Mazza et al. 2014) as well as in the range of European beech and beech-fir forests of western, central and southeastern Europe, where annual precipitation rates are almost twice as high as in Varramista (i.e., 1500 mm/year;Fig. 8 and Fig. S9;Bijak 2010;Lebourgeois et al. 2010;Toromani et al. 2011;van der Maaten-Theunissen et al. 2013).In Varramista, mean TRWs in the depression macroplot and in the upper slope macroplot, which can be assumed to be most exposed to drought, do not substantially differ from the growth performance of silver fir in the Black Forest (van der Maaten-Theunissen et al. 2013), which is considered to represent a well-suited mesic environment for the species.
Furthermore, the observed multi-annual (e.g., decadal) TRW trends in Varramista are in line with corresponding data at the continental level (Gazol et al. 2015;Vitasse et al. 2019), although with different magnitudes along the geomorphological gradient.In particular and rather unexpectedly, silver fir growing on the upper slope responds less sensitively to dry years.
The Varramista stands feature a decrease in TRWs from the early 1950s until the middle of the 1980s.Similar trends are reported from Central Europe (Germany and Poland in particular;Boše ľa et al. 2014;Büntgen et al. 2014;Elling et al. 2009;Łuszczyńska et al. 2018;Uhl et al. 2013) to the Apennines in the Italian Peninsula (Clauser 1981;Clauser and Gellini 1986) and are usually ascribed to air pollution-induced stress due to high atmospheric SO 2 and NO x concentrations.
Starting from the 1980s, the TRW trends in Varramista followed again the general European pattern, i.e. the reduction in SO 2 and nitrogen oxides (NO x ) emissions (Boše ľa et al. 2014;Łuszczyńska et al. 2018) combined with climate warming (Büntgen et al. 2014) allowed silver fir to recover and to considerably increase its growth rate until the end of the 1990s.A partial exception to this general growth release after 1980 is represented by silver fir stands in southern Italy and in the Pyrenees, where drier local climates combined with frequent drought events counterbalanced the positive effects of increased air quality (Büntgen et al. 2014).
A generalized increase in summer drought frequency and intensity as revealed by the SPEI values of the last 20 years in Varramista caused a TRW reduction starting at the beginning of the present century.This is in line with what is reported at the European scale (Boše ľa et al. 2018;EEA 2017;Gazol et al. 2016).In this context, Lebourgeois et al. (2010) emphasized the importance of soil water holding capacity for buffering the negative consequences of increasing summer drought intensities and frequencies for silver fir stands at low elevation.

Climate-growth relationships
The correlation and response functions analysis confirmed the high sensitivity in terms of annual ring width of lowland silver fir to water shortage and summer drought in particular.Therefore, longer-thanusual season aridity starting in May usually results in negative pointer years.A severe and early-starting water shortage period may in fact significantly restrict the available time for sustained growth, which in silver fir mostly occurs in early summer (Aussenac 2002;Lebourgeois 2007;Santini and Martinelli 1991).For the same reason, wet spring years allow silver fir to continuously foster the growth rates, which eventually results in positive pointer years (Cailleret et al. 2013).
The key limiting factor for tree-ring growth in silver fir at low elevation is thus summer drought, whereas with increasing elevation the main growth-regulating factors are winter and spring temperatures (Lebourgeois et al. 2013;Rolland et al. 1999;van der Maaten-Theunissen et al. 2013;Vitali et al. 2018).
Surprisingly, no negative impact or even a slight positive impact of previous-year drought conditions on TRWs has been observed in Varramista, especially during the growth release phase in the 1980s (Vitali et al. 2017).This lack of negative influence of dry summers on the growth in the next season contrasts with the immediate and mid-term negative post-drought growth reactions reported by other authors (Battipaglia et al. 2009;Macias et al. 2006;van der Maaten-Theunissen and Bouriaud 2012).According to Guicherd (1994), such mid-term negative reactions are due to the trait of silver fir to suspend droughtinduced stomatal closure even during short and weak rainfall events, which possibly leads to water stress and lack of reserves for the next growing period (Toromani et al. 2011).In some cases, severe droughts and related decreases in water availability are even reported to have caused silver fir dieback, e.g. in the Pyrenees (Camarero et al. 2011;Hernández et al. 2019), in southern France (Cailleret et al. 2013) and in Switzerland (Rigling and Stähli 2020).There is no information on past dieback events in Varramista, however.Nevertheless, large diebacks following drought periods cannot be excluded, even if today the stands do not seem to have experienced any.Furthermore, the analyzed silver fir individuals are relatively young (ca.60 years) and their response to climatic stress may feature an age-dependent component, as found for other coniferous tree species (e.g., Carrer and Urbinati 2004;Carrer 2011, Wu et al. 2018).
Clearly, the provenance of silver fir influences its response to drought (Matias et al. 2016;Mihai et al. 2021), as well as local adaptations of trees growing under frequent drought stress, which may to some extent buffer the negative consequences of water shortage (Csillery et al. 2020a;Csillery et al. 2020b;Nourtier et al. 2012).The study case of Varramista provides however the opportunity to further analyze this issue.
Lastly, the question remains open whether the high winter temperatures in Varramista may allow silver fir to restore part of the summer drought-induced missing reserves through a reactivation of the photosynthetic activity and the accumulation of carbon before the start of the next growing season (Carrer et al. 2010;Guehl et al. 1985;Kozlowski et al. 1991), as recently highlighted in different evergreen species of Mediterranean and temperate environments (Saarinen et al. 2011;Zhang et al. 2017).To elucidate this, ecophysiological measurements for several years would need to be made.

Conclusions
Our dendroecological investigation on meso-Mediterranean silver fir stands in Varramista confirms the potential of the species to thrive at the warm edge of its distribution, reaching similar growth performance as in Central Europe, in spite of substantially higher temperatures and much lower summer precipitation sums.The silver fir stands in Varramista follow the general long-term (decadal) growth trends of most European silver fir populations, including a continuous reduction of tree-ring width over the last ca.25 years as a consequence of increasing water stress.
The most limiting factor for silver fir growth under these conditions is summer water shortage, but not temperature.We propose that in Varramista soils have a sufficient water holding capacity, allowing the local population of silver fir to not only overcome the typical and strong Mediterranean summer drought lasting at least two months, but also to regularly regenerate naturally and to successfully compete with meso-Mediterranean species such as Quercus ilex and Q. cerris.
The potential climatic envelope of silver fir should be extended to warm Mediterranean conditions, provided a sufficient annual precipitation (i.e., 850 mm) and the absence of extreme (i.e., three and more months lasting) summer droughts.In the face of ongoing climate change, future research should focus on the genetics and the provenances of existing crypto-Mediterranean silver fir occurrences, such as the Varramista stands, and on their physiological adaptation mechanisms to summer drought in particular (e.g., growth performance, cambial activity and reserve use capacity under drought-stress during the growing season).Our study suggests that silver fir may have a bright future also under warmer and partially drier conditions in central Europe.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
, presently representing the lowest fir occurrence in Tuscany (Municipality of Montopoli in Val d'Arno 2008).In the last few decades, silvicultural interventions were limited to the phytosanitary clearing of fallen and damaged trees after a storm in March 2015, which mainly affected the valley bottoms (A.Casarosa 2018, personal communication).

Fig. 2 .
Fig. 2. Tree density distribution per macroplot, species group and DBH class.The white column refers to the established regeneration (height > 1.3 m and DBH < 5 cm).Note that the y-axis has a logarithmic scale.

Fig. 5 .
Fig. 5. Tree-ring-width (TRW) series of the cored trees in each macroplot.(a) upper slope, (b) lower slope, (c) depression.The black lines represent the TRW average series of the macroplot, whereas the dashed grey lines report the sampling depth.

Fig. 6 .
Fig. 6.Correlation and response functions coefficients for temperature, precipitation, and threemonths SPEI monthly data for each macroplot.Bars show correlation coefficients and dots represent response coefficients.Dark grey bars refer to significant values, light grey bars to non-significant ones.Months of the current year are in capital letters, months of the previous year in lowercase.The analysis covers the period 1953-2018 for the upper and lower slope macroplots and 1955-2018 for the depression macroplot.

Fig. 8 .
Fig. 8. Mean tree-ring widths of Silver fir for the period after 1900 AD and under different climate conditions (mean annual precipitation and temperature) and in different forest types.Point size represents the mean TRW of the series, the grey-scale color refers to the vegetation type following Lang (1994).The crossed point represents the depression macroplot of Varramista.

Table 1
Characteristics of macroplots, forest stands, sampled trees and related tree ringwidth series and chronologies.stationwasadditionally confirmed by the high mean Pearson correlation coefficient (r = 0.88) with the data of a rain gauge in Varramista, which is operated since 2012 (see Fig.S2for details).
SD = Standard Deviation.D. Walder et al. reference

Table 2
Shannon index values (±SD)in the different macroplots and stand layers.