Strength of parenchyma chronologies
The classical estimators for chronology quality were weak for ray parenchyma time-series, with values close to zero or even negative (Table 1). This situation is common in chronologies based on anatomical features (Eckstein & Frisse, 1982; Yasue et al., 2000; Fonti & García-González, 2004; Olano et al., 2012) and has been attributed to either a reduced interannual variation in anatomical parameters owing to functional constraints, to a great interindividual variability, or to the reduced area examined in anatomical studies, as they are commonly based on microscopic preparations (García-González & Fonti, 2008). However, it has been shown that merging individual ring-width chronologies into a mean chronology may produce a robust climatic signal, even if individual trees have lower climatic signals or no significant responses to these limiting factors (Carrer, 2011; Rozas & Olano, 2013). The concomitant poor chronology statistics and robust climatic signals in our data and other chronologies based on anatomy (Yasue et al., 2000; Campelo et al., 2010) might respond to the same phenomenon of common climatic signal enhancement. A preliminary analysis of our data shows that the climatic signal strongly increases with larger sample sizes for those climatic factors that significantly influence wood anatomy, although remains unchanged for nonsignificant factors (see Fig. S2).
Climatic factors affecting ray parenchyma formation
Our results suggest that ray parenchyma formation in J. thurifera depends on conditions before and during its formation, and this concurs with previously observed effects of climate on cambial activity, xylogenesis and tracheid traits (Camarero et al., 2010; Olano et al., 2012). Parameters linked to parenchyma total quantity (TOTPAR, PERPAR and TOTRAY) shared climatic signals at critical stages of the xylogenetic process, mainly at the end of previous year xylogenesis (October), radial growth resumption (May) and latewood initiation after summer arrest (August). Interestingly, parameters related to ray formation (NEWRAY) or to the continuity of existing rays (CONRAY) responded to winter rainfall conditions, probably through their effect on winter levels of carbohydrates stored in the stem.
Higher temperatures at the end of the previous growing season exerted a negative effect on PERPAR and TOTRAY (Figs 4, 5). This is the time of latewood tracheids’ wall thickening (Camarero et al., 2010), a process that is usually positively related to temperature (Yasue et al., 2000; Wang et al., 2002). Thus, high October temperatures may induce carbohydrate investment in lignin deposition in latewood cell walls (Gindl et al., 2000), reducing carbohydrate levels and inducing a subsequent decline of parenchyma production the following growth season. [Correction added after online publication 14 January 2013: ‘low’ has been replaced with ‘high’ relating to the October temperatures.]
The detrimental influence of winter rainfall on TOTRAY may be a side-effect of its detrimental effect on ring width, as is suggested by the absence of a winter rainfall signal in the parenchyma chronologies after removing the RW effect (Fig. 4). Winter rainfall exerts a strong negative effect on Juniperus thurifera secondary growth in our study area (Rozas et al., 2009), and in general throughout the northern sector of its Spanish distribution range (DeSoto et al., 2012). Rainy conditions and cloudiness may notably reduce solar radiation, photosynthetic rates and the amount of glucose assimilated by evergreen conifers (Medlyn et al., 2002). A decrease of winter recharge of stem carbohydrates under rainy/cloudy conditions may lead to lower reserve levels at the beginning of the growing season (DeSoto, 2010; Gimeno et al., 2012), a factor that usually induces reduced rates of xylem increment the next year (Hoch et al., 2003; Daudet et al., 2005). However, the negative effect of winter rainfall on the appearance of new rays is slightly different because it occurs within an earlier window, that is, December–February, and remains strong even after the RW effect has been removed (Fig. 4). This finding indicates that the formation of new rays in response to dry winters is independent from ring width. A tentative hypothesis may be that the formation of new rays is promoted by high carbohydrate levels at the beginning of the growing season. Moreover, the positive effect of January rainfall on CONRAY and the negative correlation between CONRAY and NEWRAY, after removing the RW effect, also indicates that the formation of new rays is partly related to the end of previously existing rays.
Summer drought is a major factor constraining secondary growth in Mediterranean environments (Cherubini et al., 2003). In our study area, water deficit occurs in June and July in 50% and 70% of the years analysed, respectively. Consequently, June and July rainfall signals are commonly reflected in juniper tree-ring increment (Fig. 3; Notes S1; Table S1; Rozas et al., 2009; DeSoto et al., 2012; Gimeno et al., 2012; Rozas & Olano, 2013) and tracheid size (Olano et al., 2012). By contrast, May conditions are generally optimal for photosynthesis and growth, with water deficits occurring in only 5% of years, leading to a weak signal of May rainfall in the RW chronology. An explanation other than water limitation should be sought to interpret the strong positive effect of May rainfall on PERPAR and TOTRAY. Ray parenchyma formation derived from high concentrations of nonstructural carbohydrates produced under optimal photosynthetic conditions (Gartner et al., 2000) is a possible explanation that should, however, be discarded, because active secondary growth in May (Camarero et al., 2010) depletes storage carbohydrate levels in J. thurifera (DeSoto, 2010).
Recent studies with Arabidopsis mutants show that interfascicular cambium, which contains ray initials, is activated through the tension generated from dividing fascicular cambium and an associated hormonal triggering by jasmonate (Sehr et al., 2010; Agusti et al., 2011). Under this hypothesis, the interfascicular cambium divides as a response to cambial pressure generated by the dividing cells of the fascicular cambium, and this process is regulated by hormonal signals related to a mechanical disturbance. Ethylene, a hormone related to mechanical disturbance, has been linked to a higher production of parenchyma rays after wounds or external tensions (Andersson-Gunneras et al., 2003; Chehab et al., 2009): it is known that ethylene plays a clear role in both cambial activity (Love et al., 2009) and the control of interfascicular cambium activity (Lev-Yadun & Aloni, 1995), and elevated ethylene levels have been described in conifer stems during the period of maximal cambial activity (Klintborg et al., 2002). If the same mechanism is applied to J. thurifera, it would offer a plausible explanation, as abundant May rainfall may increase stem internal water pressure, increasing jasmonate and ethylene levels, and activating fascicular cambial divisions, with this process leading to the activation of ray initials. The observed positive effect of May rainfall on PERPAR and TOTRAY fits with this hypothesis, with the timing in the climatic response occurring during the period of cambial activity initiation in J. thurifera (Camarero et al., 2010).
Late summer weather conditions in the current growth year did not play a significant role for J. thurifera growth, probably because of the reduced contribution of latewood to the final ring width (Fig. 1; Olano et al., 2012). However, our results suggest that the conditions during August, that is, at the onset of latewood formation after the arrest of mid-summer growth (Camarero et al., 2010), are relevant for parenchyma formation. The development of a greater number of new rays is associated with a relatively humid August, whereas a consistently warm August has a strong detrimental effect on PERPAR, probably through the additional effect of ray finalization. Interestingly, no effects from the conditions in or after the current September were observed, which is consistent with the end of latewood forming cambial divisions in late August (Camarero et al., 2010).
Potential use of ray-parenchyma chronologies
In comparison with other anatomical parameters, ray parenchyma is relatively quick to measure, especially in conifers with uniseriate rays. This fact is extremely relevant as the need for time-consuming procedures precludes the widespread use of other anatomical proxies based on conductive elements (Fonti et al., 2010). The potential to analyse large sample sizes is critical to effectively evaluate the climatic sensitivity observed in the chronologies, as common climatic signal is generally enhanced by increasing sample size (Fig. S2; Carrer, 2011). In addition, the interest of parenchyma chronologies may go well beyond its value as a climatic proxy. Medium- and low-frequency signals in the interannual variation of ray parenchyma may provide keys that enable the interpretation of plant responses to diverse abiotic factors (such as disturbances, fires, avalanches and extreme climatic events) and biotic factors (such as competition, herbivory and defoliator outbreaks). As such, they may contribute to a more thorough understanding of the response of trees to global climatic change through retrospective analysis of both growth dynamics and reserve levels. Increasing the potential of this technique requires a better understanding of the physiological process controlling ray parenchyma formation and its relationships with the vegetative phenology and environmental conditions.