Drought-induced growth decline of Aleppo and maritime pine forests in south-eastern Spain

Climate warming may enhance the negative effects of droughts on radial growth in areas with severe water deficit, such as Mediterranean mountains under semi-arid conditions. The impacts of drought on growth decline of Mediterranean pines have not been evaluated considering species with different vulnerability and areas with contrasting climates. Dendrochronological methods were used to assess the responses of basal area increment to drought in Pinus pinaster and P. halepensis plantations. We compared growth trends of trees with different defoliation degree in two sites in south-eastern Spain (Sierra de los Filabres and Sierra de Baza) with contrasting climatic conditions. In the more xeric area (Filabres) both pine species showed a sharp growth reduction in response to extreme droughts such as those observed in 1994-1995, 1999 and 2005. The radial growth of both species was enhanced by May and June precipitation of the year of tree-ring formation. P. pinaster showed higher defoliation in the xeric area (Filabres) than in the more mesic area (Baza) but needle loss was not linked to an abrupt growth reduction. Contrastingly, divergent radial growth patterns between trees showing high and low defoliation degrees were found for P. halepensis in the more xeric area, where a negative relationship between recent basal area increment and defoliation was found. Pine plantations in Mediterranean mountains under semi-arid conditions are highly vulnerable to warming-induced droughts. Such marginal stands constitute valuable monitoring systems to assess the negative impacts of drought on tree growth, and to test if management strategies as thinning can mitigate the negative impacts of climate warming on similar drought-stressed forests.


Introduction
Forests are responding to climate warming through changes in growth and vigor thus acting as monitors of the effects of climate change on terrestrial ecosystems (Bonan, 2008).An increase in temperature and evapotranspiration and a greater frequency of severe droughts have been predicted for the Mediterranean Basin in the near future (IPCC, 2007).Forest dieback and growth decline are usually linked to severe droughts occurrence in areas with pronounced water deficit such as the semi-arid USA (Allen and Breshears, 1998).Thus, the negative effects of warming-induced drought might also affect negatively the performance of Mediterranean mountain forests, which are mostly drought-prone ecosystems.
Droughts can modify the hydraulic conductivity of trees reducing their vigor and causing a decline in radial growth (McDowell et al., 2008).Indeed, droughtinduced dieback, declining radial growth and increased mortality rates have been described for drought-prone forests throughout the world (Allen et al., 2010).These dieback episodes usually show a great spatial variability across geographical gradients (Van Mantgem and Stephenson, 2007;Peñuelas et al., 2008).However, the responses of forests to warming-induced drought, including growth decline, can also vary as a function of species-specific resistance to drought and local climatic conditions (Suárez et al., 2004).
In Europe, several Mediterranean pine species reach their southernmost limit of distribution in the mountains from the southern Iberian Peninsula.Then, we can expect that these southernmost populations of pines growing in xeric sites may be more vulnerable to warming-induced drought stress than similar populations from growing in mesic sites (Jump et al., 2006;Macías et al., 2006;Linares et al., 2009).For instance, populations in xeric sites of southern Iberia may be more sensitive to drought-linked xylem embolism and can show greater growth decline than northern populations from mesic sites (Martínez-Vilalta et al., 2008).However, southern pine populations may also show adaptive features to withstand the negative effects of drought on growth and hydraulic conductivity (McDowell et al., 2008).These adaptations should depend not only on the species' resistance or vulnerability to drought but also on local conditions such as soil water holding capacity which may modulate the effects of climatic stressors (Camarero et al., 2004;Macias et al., 2006;Linares and Tíscar, 2010).
Recent episodes of growth decline associated to drought events have been reported for Pinus sylvestris L. in central Europe (Rebetez and Dobbertin, 2004;Bigler et al., 2006) and NE Spain (Martínez-Vilalta and Piñol, 2002).Moreover, mortality and growth decline of several conifers throughout the Iberian Peninsula have been the subject of considerable study and debate (Camarero et al., 2004).In the Iberian Peninsula, relevant episodes of forest decline were detected in response to the 1980s and 1990s droughts, mostly affecting conifer forests in xeric sites from Mediterranean mountains (Lloret and Siscart, 1995;Peñuelas et al., 2001;Camarero et al., 2004;Linares and Tíscar, 2010).These studies focused on natural forests despite recent episodes causing a sharp growth decline and leading to massive defoliation events have been also described in pine afforestations (Navarro-Cerrillo et al., 2007).Extensive pine afforestations are highly relevant in the Mediterranean Basin from both ecological and socioeconomic points of view.For instance, ca.3.5 million ha were reforested with conifers since the 1940s in Spain (Allen et al., 2010).Nevertheless, the effects of droughts on tree dieback and growth decline have been rarely evaluated in pine reforestations despite its ecological and economic importance.
Climatic trends in eastern Andalusia, southern Spain, during the second half of the 20 th century were characterized by a high drop of spring rain (De Luis et al., 2008).Such increase in spring aridity was particularly noticeable in mountains from SE Spain (Fernández-Cancio et al., 2010).Concurrently, a dieback process was detected in 2002 in Sierra de los Filabres (Andalusia, SE Spain) affecting at least 10,000 ha of Pinus afforestations that showed massive defoliation, spreading to the neighboring range in Sierra de Baza (Navarro-Cerrillo et al., 2007;Sánchez-Salguero et al., 2009).The lack of visual symptoms of forest pathogens and pests, and its coincidence with previous extreme droughts in the mid 1990s, suggested that the decline might be linked to drought stress (Fernández-Cancio et al., 2010).
In this study, we used dendrochronological methods to evaluate the relationships between radial growth and drought severity in P. pinaster and P. halepensis plantations located in two mountain areas (Filabres, Baza) with contrasting climatic conditions in southern Spain.We employed dendrochronology as the best tool to quantify the severity of growth decline of trees in the context of the last 2-3 decades and its relationship with drought stress and defoliation events (Camarero et al., 2004).We aimed to (i) quantify the changes in recent radial growth in response to severe droughts, (ii) determine if there are different responses to drought in the growth of both species in two study areas, and (iii) evaluate the correspondence between the recent defoliation degree and the growth response to drought severity, as inferred through the assessment of climategrowth relationships.

Study area
The study area includes afforestations of P. pinaster and P. halepensis located in the Sierra de Baza (37°1 3' N, 2°32'W, elevation range 845-2,269 m) and Sierra de los Filabres (37°22' N, 2°50' W, elevation range 300-2,186 m) (Fig. 1).The mean annual rainfall (1950-2009 period) ranged between 320 mm in Sierra de Filabres (from now on «Filabres») to 400 mm in Sierra de Baza (from now on «Baza») and the estimated mean annual temperature was 13.4°C at 1,000 m.These values correspond to a Mediterranean semi-arid climate.The soils had a greater water holding capacity in Baza than in Filabres, being respectively cambisols on limestone substrate and regosols on limestone substrate.

Field sampling
The trees were classif ied into two vigor classes according to their current defoliation degree: «healthy» trees with low to moderate defoliation (hereafter abbreviated as H trees) lower than 25% of the crown, and defoliated trees (hereafter abbreviated as D trees) with crown defoliation higher than 25% (Ferretti, 1994).Stratified sampling was done for each species, study site and vigor class, with the sampled trees randomly selected from each canopy stratum.Selected trees represent every vigor class and they were located at a distance greater than 100 m from the closest stand edge.The diameter at breast height (dbh, cm), total height (m) and vigor class (H or D) were registered for each tree (Table 1).The selected trees were felled and a transversal disk of the bole at 1.3 m height was obtained.

Dendrochronological methods
The disks were dried and polished with sandpapers of successively fine grain until its growth rings were clearly visible.Two opposite radii per section perpendicular to the maximum slope were selected to avoid reaction wood.The ring series along these radii were visually cross-dated based on characteristic rings (Yamaguchi, 1991).The polished surfaces of the disks were subsequently scanned to a resolution of 300 ppp.
The annual growth rings in the synchronized tree-ring series were measured with a resolution of 0.01 mm using the semi-automatic measurement system WinDendro TM (Regents Co., Canada).Cross-dating of the tree-ring series was evaluated using the COFECHA program (Holmes, 1983).
To analyze changes in growth patterns, we calculated annual basal area increment (BAI) for both pine spe-cies and study areas, considering both health classes separately, using the following formula: where R is the radius of the tree and t is the year of tree-ring formation.
We also obtained the residual chronologies of ringwidth indices by eliminating the long-term growth trends related to increasing tree size and age and reducing the first-order autocorrelation.Detrending of the tree-ring series was done by fitting a negative exponential curve to each series.Dendrochronological statistics were used to compare growth of less or more defoliated trees.We calculated the mean sensitivity (MSx), a measure of the relative difference of growth between consecutive rings, and the first-order autocorrelation (AR1), a measure of similarity in width between consecutive rings (Fritts, 1976).Dimensionless Residual indices of tree growth were produced by dividing the raw ring width values by the values of the fitted curve and performing autoregressive modeling of the results with the program ARSTAN (Cook, 1985).

Climate data
We used two sources of climate data for different purposes.First, to assess climate-growth relationships we used data from thirty local meteorological stations with long and continuous records located near both study areas (source AEMET).Second, to quantify changes in climatic trends of both study areas during the 20 th century we obtained data interpolated data from the period 1900-2006 for the of 0.5°-grid (coordinates 37°00'-37°30' N and 2°30'-3°00' W) from the database CRU TS 2.1 (Mitchell et al., 2001).Missing data of the local series were reconstructed by linear regressions based on local interpolation considering the nearest and most complete stations (Fernández Cancio and Manrique Menéndez, 1997).
A regional series of monthly climate variables (mean temperature, total rainfall) for each study area (Filabres, Baza) was obtained using the subroutine MET of the Dendrochronology Program Library package (Holmes, 2001).The homogeneity of the climatic series was assessed using the subroutine HOM of the same package.Finally, we calculated an index of annual water deficit or drought index (DRI, in mm) since this may be a better indicator of the effects of water availatibility on tree growth than temperature or precipitation itself.The DRI was calculated since 1950 using mean values of the two regional climate series derived from local data.The DRI is defined as the difference between the accumulated precipitation (P) and the potential evapotranspiration (PET) since August of the previous year up to July of the year of tree-ring formation, i.e.DRI = P -PET (Thornthwaite, 1948, Bigler et al., 2006).Thus, lower (higher) DRI values indicate higher (lower) water deficit.

Growth response to climate
The relationships between radial growth and climate were evaluated for each species, study area and vigor class using the residual chronologies and the regional climate series of mean monthly temperature and monthly precipitation.Growth-climate relationships were quantified using Pearson correlation coefficients.Growth indices and monthly climatic series were compared from August of the previous year to September of the year of tree-ring growth because this is the most influential period for radial growth of both studied pines (Richter and Eckstein, 1991;De Luis et al., 2007;Camarero et al., 2010).Correlation analyses were performed using the program DENDROCLIM 2002 (Biondi and Waikul, 2004).

Results
Mean maximum temperatures in the study area have signif icantly increased since 1950, whereas spring rainfall has decreased in the same period (Fig. 2).The lowest DRI values were observed in 1994, 1995, 1999 and 2005, and corresponded to years characterized by severe droughts.
BAIs of both species showed similar trends and sharp reductions in both study areas corresponding to years with low DRI values, i.e. 1995, 1999 and 2005 (Fig. 3).We also noted a greater divergence in BAI patterns between P. halepensis with different defoliation degree in Filabres and, to a lower degree, in Baza.Contrastingly, no BAI divergence was noted for P. pinaster according to the trees' defoliation, despite this species showed the highest defoliation degree in the xeric area (Filabres).
Mean tree-ring widths in the period 1980-2006 significantly (P < 0.05) differed between vigour classes for P. halepensis in both study areas, being higher for trees with low defoliation degree (Table 2).In the xeric Filabres area, more defoliated P. halepensis trees had a higher mean sensitivity and first-order autocorrelation than less defoliated trees (Table 2).
We found significant positive (negative) relationships between growth and precipitations (temperatures) of May and June for both pine species (Fig. 4).In Filabres, the growth of both pines was also related positively to the rainfall of January and March, whereas precipitation in the previous September also favored growth.
Overall, in the xeric area (Filabres), the more defoliated trees of both species showed a stronger negative (positive) association with monthly maximum temperatures in summer (precipitation in winter and spring) than less defoliated individuals, In this area, the temperatures of June and July showed negative relationships with growth of P. halepensis trees with low or high defoliation levels, respectively (Fig. 4).Despite the stronger negative growth-temperature associations of defoliated trees in the xeric area, the mean relationship with annual maximum temperature was more negative for less defoliated than for more defoliated P. halepensis trees (Table 3).Finally, the basal area increment of P. halepensis after 1995 was consistently and negatively related to current defoliation in the xeric area (Filabres) but not in the mesic area (Baza) [This was analyzed in SPSS v.15 software (SPSS Inc., Chicago IL)] (Fig. 5).1950 1955 1960 1965 1970 1975 1980 1985 1990 1995   The tree-ring width was calculated for the period 1980-2006. 2Variables' abbreviations: MSx: mean sensitivity of residual chronologies; AR1, first-order autocorrelation of standard chronologies.The annual MSx measures the relative difference of indexed tree-ring width from one year to the next, and it is calculated by dividing the absolute value of the differences between each pair of ring-width indices by the mean of the paired index (Fritts, 1976).

Discussion
On the whole, the climate-growth associations found for P. pinaster and P. halepensis were similar to those found in other Iberian forests under Mediterranean climatic conditions (Bogino and Bravo, 2008;De Luis et al. 2009;Vieira et al., 2009).Nevertheless, pine plantations of both Mediterranean pine species (P.pinaster, P. halepensis) in south-eastern Spanish mountains are undergoing acute processes of growth decline and forest dieback, despite they are theoretically better adapted to withstand drought than boreal species as P. sylvestris (Richardson, 1998).This apparently drought-induced decline is characterized by a sharp reduction in basal area increment and high levels of needle loss.The high sensitivity of growth of defoliated trees to precipitation and maximum temperatures suggest that their growth decline and subsequent needle loss were caused by warming-incited drought stress.Specifically, warming-induced water deficit in spring seems to be the main climatic trigger for similar droughtinduced growth declines in other Iberian forests (Martínez-Vilalta and Piñol, 2002;Camarero et al., 2004).The fact that growth of P. halepensis trees with low and high defoliation levels responded negatively to June and July temperatures, respectively, suggests that individuals with different vulnerability to drought stress could have different radial-growth phenology.Further studies should evaluate xylogenesis in cooccurring individuals of the same species with different defoliation degree to test this hypothesis.
P. pinaster showed higher defoliation degree in the xeric than in the mesic area, which reflects its greater  1994-1995and 1999 (low DRI values) (low DRI values).Bars correspond to the standard error.(Tardif et al., 2003;Bogino and Bravo, 2008).Indeed, the negative effects of drought stress on tree growth are greatly modulated by local factors as topography (Linares and Tíscar, 2010).On the other hand, P. halepensis seems to be better adapted to the semiarid conditions of this area, as the moderate defoliation levels and the sustained growth patterns of healthy trees indicated.The responses of growth to climate in P. pinaster and P. halepensis are age-and size-dependent (De Luis et al., 2009;Vieira et al., 2009), but these factors were not relevant in our case since all trees were of similar age and size.Therefore, neither age nor size seems to be among the main causes of the differential response of growth and defoliation to drought stress in our study case.The divergence of radial growth after extreme droughts was observed in P. halepensis, being more evident in the xeric than in the mesic site, but this was not noted in P. pinaster.Although a high degree of defoliation and mortality were observed since 2001 in Filabres (Navarro-Cerrillo et al., 2007), growth divergence and a negative association between basal-area increment and the degree of defoliation was evident in P. halepensis since the severe droughts of 1994-1995.Such findings suggest a lagged response of needle loss to drought stress and growth decline, which in our study may have lasted from 1995 (first sharp reduction in basal-area increment) to 2002 (defoliation), i.e. 7 years.The lagged responses to climatic stress of growth decline and defoliation greatly complicate the disentangling of the cause-effect relationships in episodes of forest dieback (Pedersen, 1999;Dobbertin, 2005).
The increased climate variability associated with the current climate change (Manrique and Fernández Cancio, 2000) could lead to growth decline in Iberian pine species (Andreu et al., 2007).Such decline may be linked to the described dieback episodes in marginal plantations under severe drought stress and with a high year-to-year variability in precipitation.Rainfall variability was responsible of the sharp growth declines in 1994-1995 and 1999, but also caused high basal-area increment in wet years (e.g.1992, 1997) when radial growth of P. pinaster was similar in both study sites.Furthermore, more research is required to unravel the relative contributions of long-(e.g., temperature rise) and short-term (e.g., droughts) climatic stressors on growth decline.
Precipitation in April-June influenced positively radial growth, whereas mean temperature in March-July had a negative effect on wood formation.In the xeric area (Filabres), water availability in late spring and early summer seems to be a major driver for tree growth since maximum radial growth rates of both pines occur in this time (De Luis et al., 2007;Bogino and Bravo, 2008;Camarero et al., 2010).Specifically, P. halepensis trees with low defoliation increased growth after the 1999 drought meanwhile P. pinaster maintained low growth rates after this drought regard-466 R. Sánchez-Salguero et al. / Forest Systems (2010) 19(3), 458-469 Year 199019911992199319941995199619971998199920002001200220032004  less of their defoliation degree.This could be explained by a lower adaptation to drought resistance of P. pinaster, in particular in the xeric area.A difference between both study areas is the greater positive influence of February temperatures on P. halepensis growth in Baza than in Filabres, which suggests a greater thermal limitation to growth in Baza than in Filabres.
The observed defoliation and decline in radial growth of P. halepensis and P. pinaster plantations was strongly linked to drought stress in spring.Even if P. pinaster is less tolerant to drought stress than the co-occurring P. halepensis, we suggest that the specific responses of growth were conditioned by contrasting climatic conditions and local variability in soil water holding capacity.Therefore, xeric areas and sites with low water holding capacity, such as south-facing slopes on limestone substrate, may predispose pines to drought-induced decline, as was the case of declining P. pinaster stands.
The severe droughts of 1994-1995 and 1999 induced a decline in radial growth and the selective defoliation of Aleppo and maritime pine plantations through a reduction in spring water availability.We found a divergence of radial growth between trees currently showing different defoliation levels for P. halepensis in the more xeric area (Filabres) where basal-area increment and recent defoliation were negatively related.Contrastingly, P. pinaster showed a growth reduction irrespective of the trees' defoliation level which suggests that this species is less adapted to the increasingly arid conditions of the study areas than P. halepensis.

Figure 1 .
Figure 1.Study area and sampling sites (different symbols or lines correspond to the two study species and areas, respectively).The upper inset maps show the distribution of both species and the location of the study area in SE Spain (Andalusia).The distribution maps are based on information provided by Alia and Martin (2002) and Fady et al. (2003).

Figure 2 .
Figure 2. Regional trends of mean annual maximum temperature, spring precipitation and annual drought index (DRI) during the 1950-2006 period.Lower (higher) DRI values correspond to higher (lower) water deficit.The linear regression shows the significant warming trend in the study area.Negative (positive) values of the drought index indicate higher (lower) water deficit.The regional climatic data are based on a regional mean calculated using data from 30 local stations.

Figure 3 .
Figure3.Recent trends in basal area increment (BAI) of the two pine species studied in the Filabres and Baza study areas according to the crown defoliation of trees.Trees were classified as healthy with low defoliation (H, white symbols, defoliation < 25%) or declining with high defoliation (D, black symbols, defoliation > 25%).The grey line corresponds to the annual drought index (DRI) with negative (positive) values indicating highest (lower) water deficit.Note the severe droughts in1994-1995  and 1999 (low DRI values) (low DRI values).Bars correspond to the standard error.

Figure 4 .
Figure 4. Relationships between radial growth (ring-width indices) and monthly climatic variables (T, mean temperature -black bars; P, total precipitation -white bars; Tmx, mean maximum temperature -grey dashed lines) for the two studied pine species in the Filabres and Baza study areas considering separately trees with low (H, healthy trees; defoliation < 25%) or high defoliation (D, declining trees; defoliation > 25%).Growth is related with climate data from the previous (months abbreviated by lowercase letters) and current (months abbreviated by uppercase letters) years, being the current year that of tree-ring formation.The significance levels (P < 0.05) are indicated by dashed horizontal lines.

Figure 5 .
Figure5.Relationships between annual basal area increment (BAI) values and current defoliation degree for P. halepensis in each study area.Filled symbols are significant correlation coefficients (P < 0.05).The dashed vertical lines indicate sharp changes in the BAI-defoliation associations after the severe droughts in1994-1995 and 1999.

Table 1 .
Description of the trees sampled in the Baza (B) and Filabres (F) study areas bValues are means ± SE.Different letters indicate significant (P < 0.05) differences between study areas (Mann-Whitney U test).