Twenty-six sampling sites (8 P. halepensis sites, 8 P. pinaster sites and 10 P. sylvestris sites) were selected in their distribution area in Spain (Figure 1; Table 1). Pinus halepensis sampling sites consist of an upper storey of P. halepensis and an understorey formed by broadleaved Mediterranean species (Quercus ilex L., Q. coccifera L. and Q. faginea Lamk.). Silviculture in the sampling area of P. pinaster is traditionally based on natural regeneration following a seed tree system and focused on multifunctional uses (recreation, timber and resin). Pinus sylvestris sampling sites are at its southern and western distribution threshold. These dry areas of distribution of this species that usually grows in humid environments are the logical places to investigate the effects of increased aridity (Martínez Vilalta & Piñol, 2002). Besides, in assessing the impact of global warming on ecosystems, any changes in tree growth are likely to occur first in those tree stands placed at the ecological boundary of the species (Tessier et al., 1997). At each sampling site, 15 dominant trees were randomly selected. Two cores were extracted at 1.30 m above ground from each selected tree. The increment cores were air dried, mounted on wooden supports and dated according to standard dendrochronological techniques (Stokes & Smiley, 1968).

Absolute dating is essential for any dendroclimatological study, and it is impossible to compare climatic variables in one specific year with tree-ring growth if the individual tree-ring series are not dated correctly (Fritts, 2001). To assess measurement and dating accuracy, the v6.06P COFECHA program (Holmes, 2001; Grissino-Mayer, 2001; available at www.ltrr.arizona.edu) was applied. This program calculates the Pearson correlation indices between the indexed tree-ring series and a master reference chronology in a series of consecutive, partially overlapped segments of a length specified by the user. According to standard dendrochronological methods, tree-ring series exhibiting correlation values with the master chronology below 0.4 were excluded.

Standardization removes geometrical and ecological trends while preserving inter-annual high-frequency variations that are presumably related to climate. To eliminate biological trends in tree-ring series and to minimize growth variations that are not shared by most trees, the v6.05P ARSTAN program (Cook & Holmes, 1984; Holmes, 2001; available at www.ltrr.arizona.edu) was used. The long-term trend was removed from each time series of ring width measurements by fitting and calculating an index defined as actual ring-width for each year divided by the curve-fit value. The standardized series were averaged in order to obtain a master chronology at each study site.

The accurately dated cores were visually examined for IADF using a stereomicroscope (magnification up to 25x). In contrast to the annual rings, IADFs show a non-sharp transition boundary between earlywood and latewood cells (Fritts, 2001). IADFs were only counted when present in both cores in the same tree ring, and they were identified by considering the position of the density fluctuation within the ring. Only IADF type E (latewood-like cells within the earlywood) were considered for our study since IADF type L (earlywood-like cells within the latewood) were rarely present in our sample (Figure 2). As the number of samples changed over time, the relative frequency was calculated with the following formula [1]:

Precipitation is the main factor influencing tree growth of pine species in semiarid Mediterranean conditions (Raventós et al., 2001). In our study sites, the growth of P. halepensis, P. pinaster and P. sylvestris is mainly controlled by precipitation at different times of the year. Despite the lack of biological significance of some low correlation values and the site variability, higher correlation values between seasonal climatic conditions and species reflect differences in the influence of climatic conditions between species. Winter and spring precipitation is related positively with tree-ring growth in P. halepensis and P. pinaster, while the growth of P. sylvestris is mostly influenced by summer precipitation. These results are consistent with those of previous studies in P. halepensis in Greece and Spain (Papadopoulos et al. 2008; Olivar et al. 2012), P. pinaster in Portugal (Vieira et al., 2010; Campelo et al., 2013) and central Spain (Bogino & Bravo, 2008), and P. sylvestris at its southern and western distribution limits (Bogino et al., 2009).

However, the influence of these climatic variables on the growth of these species changed over the studied period. The positive influence of winter and spring precipitation on P. halepensis growth increased beginning in the 1990s and the positive influence of spring precipitation on P. pinaster growth increased beginning in the 1970s, while the positive influence of summer precipitation on P. sylvestris growth remained stable. These results agree with previous reports on pine species in the Mediterranean area, which suffered a change in growth response to climatic conditions in the second half of the 20th century (Andreu et al., 2007; Bogino & Bravo, 2008; Vieira et al., 2010; Campelo et al., 2013).

Global studies around the Mediterranean basin indicate that winter and spring precipitation increased and summer precipitation decreased during the second-half of the 20th century (Bradley et al., 1987; Maheras, 1988; Díaz et al., 1989). Mediterranean pines evolved during the Pliocene under tropical-like climate, before the onset of the Mediterranean climate, as a component of the pre-Mediterranean Arcto-Tertiary flora (Verdú et al., 2003; Petit et al., 2005). This species survived to a gradual increase of aridity during the transition to Mediterranean conditions, which may have led to its characteristic growth plasticity (Chambel et al., 2007). Mediterranean Pinus species are considered well adapted to withstand drought by reducing growth as water availability decreases and increasing growth as conditions become favourable (Pasho et al., 2012). This increase of winter and spring precipitation combined with the increasingly harsh climatic conditions during summer may have enhanced the importance of precipitation at the beginning of the growing season on the growth of species subject to higher drought stress conditions during summer, such as P. halepensis and P. pinaster. On the other hand, P. sylvestris, growing in mountainous environments with higher water availability during the whole year, didn’t suffer that severity under the climatic conditions.

The occurrence of IADFs is mainly influenced by precipitation in these species. IADFs may appear at different positions within a tree-ring depending on the time of the year when the triggering factor occurred (Campelo et al., 2007; Edmondson, 2010; de Micco et al., 2012). IADF type E is triggered by dry periods during spring and early summer. In contrast, IADF type L is triggered by precipitation during late summer and (or) early autumn (Wimmer et al., 2000). Previous studies in the Mediterranean area showed a high frequency of IADFs in latewood (de Luis et al., 2007; Vieira et al., 2010; Rozas et al., 2011; Campelo et al., 2013; Novak et al., 2013). However, the low frequency of IADF type L and the high frequency of IADF type E observed in our samples indicate a higher occurrence of water stress episodes inhibiting cell division and enlargement during the first part of the growing season. The ability of species to produce different types and forms of cells in different periods may also be interpreted as an important adaptation of trees for maintaining the balance among the capacity to conduct water, resistance to cavitation and mechanical stability (Novak et al., 2013).

The formation of IADFs is triggered by above-average precipitation in the previous winter and spring in P. halepensis and in spring and early summer in P. pinaster and negatively influenced by precipitation in July. These climatic conditions (precipitation at the beginning of the growing season and summer droughts) have been increasingly favoured over the second half of the 20^th century, explaining the increasing occurrence of IADFs our study area. This result agrees with previous studies that found an increase in IADF frequencies after 1980 in P. pinaster in Spain (Bogino & Bravo, 2009) and Portugal (Vieira et al., 2010; Campelo et al., 2013). Despite being at its southern distribution threshold, where a species that usually grows in humid environments could suffer from the effects of increased aridity (Martínez Vilalta & Piñol, 2002), P. sylvestris showed the lowest IADF frequency of the three species on our sample. As pointed out by Battipaglia et al. (2010), the frequency and the triggering climatic factors promoting different anatomical characteristics may vary among populations, depending on different environmental conditions.