This study included two deciduous (Quercus faginea Lam. and Q. pyrenaica Willd.) and two evergreen Quercus species, one of which has a mean leaf longevity of about one year (Quercus suber L.) and the other one (Q. ilex L. subsp ballota (Desf.) Samp.) has a mean leaf longevity of slightly more than two years (Mediavilla & Escudero, 2003 a). These species are typically Mediterranean species and they occupy large areas of open woodlands in the interior of the Iberian Peninsula.

The species were studied over four years (2007-2010) at two sites (“Garcirrey” and “Valdelosa”) located in the province of Salamanca (central-western Spain) at latitudes 40º 55´ N and 41º 08´ N and longitudes 6º 07´ W and 5º 47´ W respectively. All species were present at each of the selected sites, although Q. suber was monitored only during three years (2007-2009). The sites have sparse populations of isolated mature trees (about 50 specimens ha^-1) over 100 years old, with open pasture areas among them. The soils are Cambisols, poor in organic matter and nutrient contents in both cases. However, both sites show differences in both temperature and water-holding capacity (Dorronsoro, 1992), allowing us to study the effects of these factors on possible differences in the timing of abscission of the same species between sites.

Temperature data for each site were obtained by means of data loggers (Hobo Pendant temperature/light datalogger, Part UA-002-08, Onset Computer Corporation, Pocasset, MA, USA) placed in the shade under the canopy of one tree next to the specimens selected for phenological measurements. The sensors, programmed to obtain a measurement every 10 minutes, were monitored at each site throughout the study period. Annual rainfall data from the station nearest to each site were provided by the National Meteorological Institute of Spain (AEMET). Two sensors (HOBO soil moisture smart sensor ECH2O-10 Dielectric Aquameter) buried at different depths (between 50-80 cm the first and between 1-1.5 m the second) allowed us to study the seasonal changes in soil moisture. The depths were selected according to those suggested in the literature to obtain estimates of soil water availability for tree species (Lloret et al., 2004; Bader & Körner 2010). We are aware that the maximum depth reached by the sensors is probably insufficient to record the total absolute water amount that is available to the roots of the surveyed trees. However, it can be assumed that most of their root mass is concentrated in the selected range of depth, so that the data supplied by the sensors can be used to record the changes in water availability over time and the differences between sites, allowing us to identify possible relationships between leaf mortality and decreases in water availability.

The data provided by these sensors, programmed to obtain a measurement every 30 minutes, were then used to estimate the soil water reserve. First, we estimated the monthly soil water balance for each site, applying the following input-output equation (Landsberg, 1986):

Δθs = P – AET – q,

where Δθs = monthly variation in the soil water reserve, P = precipitation, AET = actual evapotranspiration and q = surplus (surface runoff plus deep percolation). All variables were measured in mm. We assumed soil water reserve was zero at the end of summer. From this time we calculated the cumulative water balance until the time when the electric signal from the soil moisture sensors reached saturation (usually at the end of autumn). The water balance accumulated up to this point was taken as an estimate of the maximum amount of water retained by the soil. Soil water reserve at each time interval was then calculated as a function of the electric signal from the sensors as compared with the signal at saturation. These data were used only for comparative purposes, and were not meant to represent true soil water reserve available for plants because of the difficulty of knowing the total volume of soil occupied by roots. Potential evapotranspiration was calculated applying the equation of Jensen & Haise (1963). Solar radiation data from the station nearest to each site were provided by the National Meteorological Institute of Spain (AEMET).

Five specimens of each species were selected in each of the two study sites. Selected individuals were mature trees (over 100 years old) of similar size in both cases. Trunk diameter at 1.3 m height ranged from 20 to 60 cm and mean heights were 6-10 m.

To study leaf phenology, we monitored the number of leaves of the different age classes found in the crown. On each specimen two branches situated at mid-height in the sun-lit portion of the crown were selected. The different leaf age classes were labelled with wires of different colors at the beginning of the study. The number of surviving leaves of each age class, counted at intervals of 15 days (one week during the time of the year with more intense growth or abscission rates) was used to elaborate life-tables, according to standard methods (Begon & Mortimer, 1986). Relative mortality rates were calculated for the leaves of each age class as the difference in the number of leaves between two consecutive dates with respect to the number of leaves present at the beginning of that interval, all divided by the number of days elapsed between both dates. The mortality rates were used to identify the typical abscission period of each species and to detect possible differences at the level of a single species that could be related to changes in environmental variables between sites and years.

On all sampling dates, predawn water potentials (Ψ_pd) were measured as an estimate of plant water availability, using a pressure chamber (PMS Instruments Co., Mod. 1002, Corvallis, OR, USA) in twigs sampled from the same specimens and from the same position in the crown as those used for the leaf phenology measurements. The data were analyzed separately for each of the two periods in which the growth season can normally be divided as a function of water availability (the most favorable period was termed “spring” and that with the highest stress level “summer”). The seasonal variation in the predawn water potential was estimated as the difference between the mean maximum and mean minimum predawn water potentials recorded for each study year.

To test whether the mortality of old leaves occurs after these leaves have acted as nutrient sources for the production of new tissues, we analyzed the seasonal changes in the nitrogen resorption efficiency values for the two evergreen species. At each site we placed five 0.25 m² leaf-litter collection traps below the crowns of five specimens of the two evergreen species; the contents of these traps were collected at 15-day intervals. On each sampling date, green leaves of the different leaf age classes present on the crown were also collected from the trees used for litter sampling. All samples collected were taken to the laboratory, where the total projected leaf area was determined by image analysis (Delta-T Devices Ltd, Cambridge, UK). The samples were then oven-dried at 70º C to constant mass and the total dry mass was determined. From the data thus obtained, we calculated the leaf mass per unit area (LMA, g m^-2). Leaf N concentrations were determined with a CE-Instruments NA-2100 autoanalyser (ThermoQuest, Milan, Italy). The N content was expressed per unit leaf area (N_area, g m^-2), obtained as the nutrient concentration (mg g^-1) multiplied by the LMA. For each individual tree we obtained a value of the resorption efficiency for each date in which the leaf litter was collected as [100* (N_area in green leaves – N_area in shed leaves)/N_area in green leaves]. For the N content in living leaves, we took the maximum N_area registered for the green leaves collected throughout the year in each tree, which was compared with that of the shed leaves collected from the same tree in each sampling date. Resorption efficiency of each species was obtained as an average of the five specimens studied at each site.

One-way Analysis of Variance was used to establish significant differences among site means (temperature, precipitation, predawn water potentials and soil water reserve) for P ≤ 0.05 after applying the Levene test to check for homogeneity of variances.

To analyze the effect of environmental factors on abscission, we used the values of the mortality rates obtained at each site and year for each species (and the age class of the leaves on the evergreen species). Having identified the dates with significant mortality values, we calculated the average mean daily temperature, mean maximum and minimum daily temperatures, absolute maximum and minimum temperatures, precipitation and soil moisture data for different time intervals prior to each mortality peak. Then, for the leaves of each species and, in the case of evergreen species, the leaves of each age class, we used a linear regression (after transforming the data to a logarithmic scale) to analyze the relationships between mortality rate and environmental factors, repeating the analysis with the values of the environmental variables obtained for each of the different time intervals. The most significant relationships were found for a period of 15 days before each mortality peak, similarly to previous studies (Wielgolaski et al., 2011). We took this interval as the optimum one to identify the effect of the environmental factors on leaf abscission. In the case of leaf water potential we used data from the dates closest to those of the mortality peaks. Multiple regression (after co-linearity analysis, where necessary) was used to separate the effect of two independent variables, as applied in previous studies (Doi & Katano, 2008; Gordo & Sanz, 2010). All statistical analyses were carried out using the SPSS program (SPSS Inc., Chicago, IL, USA).

Table 1 shows the seasonal temperatures at each site for the four years of study. Garcirrey always showed higher temperatures than Valdelosa, with significant differences between sites especially during the summer and in the maximum temperatures (Table 1). Rainfall displayed similar levels and seasonal distribution between the stands. In all cases, rainfall was particularly low in the summer period, when less than 10% of the annual rainfall was recorded (Table 1). However, although there were no differences in the levels of precipitation, differences between sites in soil water content were apparent, significantly higher in Valdelosa for almost the entire study period (Table 1), due to higher water-holding capacity of the soil in this site.

For all species, predawn water potentials underwent a decline along the growth season and this was always more marked at Garcirrey, for which significantly lower Ψ_pd summer values and higher seasonal Ψ_pd range than at Valdelosa were found. However, differences between sites were already present in the favorable period, with mean Ψ_pd values higher at Valdelosa, although in this case the differences were not significant (Table 2).

In the two deciduous species, leaf emergence occurred between the second fortnight in April and the first fortnight in May, the leaves appearing earlier at Garcirrey than at Valdelosa (Fig. 1). Both species retained around 80% of their leaves until November when senescence began, although with some differences between species and sites. In Q. pyrenaica, leaf abscission started during early November at Valdelosa, and two weeks later at Garcirrey (Fig. 1). Similar trends were seen for Q. faginea, in which the onset of senescence occurred around one month later in Garcirrey with respect to Valdelosa, and, in both sites, around two weeks later compared to Q. pyrenaica (Fig. 1). Similar results were observed with little differences between years of study.

Leaf mortality rates in the deciduous species increased as mean, maximum and minimum temperatures declined and the number of days with frosts increased. The highest percentage of the variation in leaf mortality during a specific period was explained by the maximum temperatures in the two weeks preceding that period (data not shown). In multiple regressions, the effects of mean temperatures, minimum temperatures and days with frost on leaf mortality tended to be non-significant, unlike the effect of maximum temperatures (in particular absolute maximum temperatures during the two previous weeks, Table 3).

Quercus suber had a later leaf emergence than the other three species (between the end of May and the beginning of June), with no appreciable differences between sites (Fig. 2). Most of the leaves survived their first year, with losses of around 15-20% at the end of their first year of life. During the second year, the surviving leaves underwent massive mortality that coincided with the appearance of the new cohort (Fig. 2). The same trends, with few variations, were observed at both sites. There were no relationships between the different environmental factors (temperature and water availability) and the mortality rates observed for the old leaves in Q. suber (data not shown). However, although abscission of the older leaves extended along several weeks, the maximum mortality rates were observed during the dates closest to the emergence of the new leaves (Fig. 3). The maximum N resorption value recorded at each site and year was always obtained on the date closest to that of maximum mortality rate of the old leaves (Fig. 4).

The other evergreen species, Q. ilex, unfurled its leaves during the last days of April or beginning of May, one month earlier than Q. suber (Fig. 5). Leaf mortality was almost negligible during their first year (less than 10%), and became significant coinciding with the emergence of a new cohort the following year, thereafter extending along the summer. Leaf mortality rates during the second year of life were higher at Garcirrey, where only about 55% of the leaves survived as compared to the 60-65% observed at Valdelosa (Fig. 5). For the leaves lost during the second year after leaf development, the maximum mortality was recorded in the pre-summer period in Valdelosa (around 60%, as opposed to 14% during the summer), whereas in Garcirrey mortality was more gradual (around 40% of the leaves that died in their second year was lost between May and June and 25% in the summer months). Similar results were obtained for the leaves during their 3^rd year: the maximum mortality rates were obtained in the pre-summer period in Valdelosa, whereas in Garcirrey mortality was more gradual (Fig. 5)

The regression analysis failed to reveal any significant relationship between the mortality rates of leaves and the different environmental parameters in Q. ilex (temperature and water availability) (data not shown). Analysis of the temporal evolution of the N resorption indicated that the highest resorption values were always recorded in the dates closest to the emergence of a new cohort of leaves (Fig. 4). The only exception was obtained in 2010 at Valdelosa, where the highest N resorption values were recorded in late July, well after the emergence of the new cohort.