The study was carried out in a natural Quercus ilex oak forest in the Prades mountains in southern Catalonia (north-east Spain) (41°13′ N, 0°55′ E) on a south-facing slope (25%). The soil is a stony Dystric Xerochrept (Soil Survey Staff 1999) lying on a bedrock of metamorphic sandstone. Its depth ranges between 35 and 100 cm, with the depth of Horizon A ranging between 25 and 30 cm. The average annual temperature is 12 °C and an average rainfall of 658 mm, with a period between September to November experiencing the maximum of rainfall. Summer drought is pronounced and usually lasts for 3 months. The vegetation consists of a dense forest with a canopy height average of 8–10 m dominated by Quercus ilex L. (20·8 m2 ha−1 of trunk basal area at 50 cm of height) accompanied by abundant Phillyrea latifolia (7·7 m2 ha−1 of trunk basal area at 50 cm of height and Arbutus unedo L. A number of other evergreen species well-adapted to drought conditions such as Erica arborea L., Junniperus oxycedrus L., Cistus albidus L., and occasional individuals of deciduous species such as Sorbus torminalis L. Crantz and Acer monspessulanum L. are also present. In winter 1999, the above-ground biomass (AB) of Quercus ilex represented 77·1% of the total biomass, while Phillyrea latifolia represented 12·6% and Arbutus unedo 7·8%; the sum of the aerial biomass of these three species thus represented 97·6% of the whole ecosystem AB. In the winter of 2005, the biomass for the same three species were 75·6%, 13·3% and 8·7%, respectively, representing in total a 97·6% of the total AB.
Eight 15 × 10 m plots were established at the same altitude (930 m above sea level) on a slope. Four of the plots received the drought treatment and four plots left as controls. All the plots were established in an area with the same aspect and altitudinal level, with a minimum distance between plots of 15 m. The treatments were randomly assigned to different plots. The drought treatment consisted of partial rainfall exclusion by suspending transparent PVC strips at a height of 0·5–0·8 m above soil level and covered approximately 30% of the total soil surface. Four plastic strips 14 m long and 1 m wide were placed along the drought treatment plots and a 0·8–1 m deep ditch was dug along the entire top edge of the upper part of the treatment plots to intercept runoff water. The water intercepted by the strips and ditches was channelled to the bottom edge of the plots. The drought treatment began in March 1999 (Ogaya et al. 2003). Soil moisture content was measured every 2 weeks throughout the experiment period by time domain reflectometry (Tektronix 1502 C, Beaverton, OR, USA; Zegelin, White & Jenkins 1989). Three stainless steel cylindrical rods, 25-cm long, were driven into the soil at four randomly selected places in each drought plot. The time domain reflectometer was connected to the ends of the rods to determine the soil moisture content.
biomass and litter determination
Just before the treatment was begun, all living stems of the three dominant species with a diameter of over 2 cm at 0·5 m height above the ground were tagged and their circumferences measured at 50 cm height with a metric tape. In January 2005, the circumferences of the stems were measured again to calculate the annual stem diameter increment.
Allometric relationships between above-ground tree biomass and the diameter at 50 cm (D50) were calculated for Quecus ilex and Phillyrea latifolia in the studied area (outside the plots). Total AB, leaf biomass (LB) and stem biomass were measured by weighing plant material after it had reached a constant weight in an oven at 70 °C. The allometric relationships in Quercus ilex (ln AB = 4·9 + 2·277 ln D50, r2 = 0·98, n = 12) and in Phillyrea latifolia (ln AB = 4·251 + 2·463 ln D50, r2 = 0·97, n = 13) were used thereafter to estimate the above-ground standing biomass of these two species in the studied area (see Ogaya et al. 2003). To estimate Arbutus unedo biomass, we used the allometric relationship (ln AB = 3·830 + 2·563 ln D50, r2 = 0·99, n = 10) previously calculated in the same area by Lledó (1990). LB was calculated by the following allometric relationships for Quercus ilex: ln LB = 3·48 + 1·70 ln D50, r2 = 0·97, n = 12, for Phillyrea latifolia: ln LB = 1·43 + 2·43 ln D50, r2 = 0·94, and for Arbutus unedo: ln LB = 1·887 + 2·157 ln D50, r2 = 0·95. Stem biomass was calculated by the difference between total AB and total LB.
Litterfall was collected in 20 circular baskets (27 cm diameter with 1·5 mm mesh diameter) randomly distributed on the ground of each of the eight plots. The fallen litter was collected every 15 days during 1999 and every 2 months during 2004. Total litterfall was estimated by the proportion of the surface area of the plots covered by the collecting baskets.
biomass and soil sampling process
Prior to the start of the experiment (March 1999), eight soil samples (four control plots and four drought plots) were analysed to test the spatial variability of the soils. There were no significant differences in P and K availability and contents between control and drought plots. In January 2005, just 6 years after the experiment was initiated, all the soil and biomass samples were collected, in order to evaluate the total contents in soil and in stand biomass, at the same time. Eight samples of leaf and stems from the three dominant species (Quercus ilex, Phillyrea latifolia and Arbutus unedo) were randomly sampled in each plot (four samples in the sun and four samples in the shade). The leaves were sampled from between 1·5 and 6 m where most foliar biomass was located. Sample collection was standardized in order to avoid bias due to differences in the age of tissues and their position with respect to sunlight. The leaves sampled were those from current year leaves of 1998 and 2004 and represented the majority of the leaves of the plants of these three species. Stems were collected separately and stems of 0·3–2 cm and more than 2 cm in diameter were differentiated. We collected four samples of each stem diameter class per plot. We only sampled the trees and shrubs of the diameter class between 2 and 12 cm of BD (at 5 cm), that represents most of the community biomass (Ogaya et al. 2003; Ogaya & Peñuelas 2007). The concentrations in the stems of diameter 0·3–2 cm did not differ from those with a diameter greater than 2 cm and thus only one stem concentration was calculated for all stem diameters, which was taken as the wood concentration. All leaves and stems were collected from different plants in each plot. In 1999 and in 2005, five leaf litter samples from each species (Quercus ilex, Phillyrea latifolia and Arbutus unedo) from each plot, were analysed separately. The leaf litter that represented 87% of the total litter mass, was analysed in the same way as the biomass.
We conducted the soil sampling in January 2005, i.e. following 6 years of drought treatment. We randomly sampled five cores from the first 30 cm of soil profile (Horizon A) in control plots and 10 in each drought plot. In the drought plots, we distinguished two levels of drought: that of the soil between the strips (D) (runoff exclusion) and that under the strips (DD) (runoff exclusion plus rainfall exclusion). We analysed these two soil fractions separately because we had previously observed that soil moisture decreased more under plastic strips than between plastic strips, being the soil moisture 27% lower under plastic strips than between plastic strips in winter. Five soil cores were taken between strips and five from under the strips, at a minimum distance of 1 m from the nearest tree or shrub; in each control plot only five soil cores were randomly sampled. We collected and analysed separately the 0–15-cm deep soil and the 15–30-cm deep soil in each soil core, as horizon A had an A1 subhorizon (first 15 cm) rich in organic matter (7·25% W/W) and an A2 subhorizon (15–30 cm) with only moderate amounts of organic matter (1·3% W/W). However, as extractable soil P and K can have great variations through the year, we conducted a seasonal study with additional samplings in spring, summer and autumn in order to investigate possible fluctuations in these variables.
Additionally five soil holes per plot under Quercus ilex trees were dug and roots of this species were sampled (φ > 5 mm) in order to study the effects of drought on root P and K concentration.
All the samples were taken to the laboratory and stored at 4 °C until the analyses were carried out. In order to analyse P and K in foliar tissues, leaves were washed with distilled water as in Porter (1986).
For the analyses of total P and K, biomass and soil samples were washed and dried in an oven at 60 °C until constant weight was obtained. Then, they were ground up in a CYCLOTEC 1093 (Foss Tecator, Höganäs, Sweden) – in the case of the biomass samples – or in a FRITSCH Pulverisette (Rudolstadt, Germany) – in the case of the soils and bedrock samples.
P and K concentrations in all biomass and soil samples were measured using ICP-AES (Atomic Emission Spectroscopy with Inductively Coupled Plasma) in a JOBIN IBON JY 38 (Longjumeau, HORIBA Jobin Ibon S.A.S., France). Before the biomass ICP-AES analyses, an acid digestion of the samples was carried out with an acid mixture of HNO3 (60%) and HClO4 (60%) (2 : 1) in a microwave oven (SAMSUNG, TDS, Seoul, South Korea). Two millilitres of the mixed acid solution were added to 100 mg of dry biomass for each sample. The digested solutions were brought to 10 mL of final volume (HClO4 3%). During the acid digestion process, two blank solutions (2 mL of acid mixture without any sample biomass) were also analysed. In order to assess the accuracy of digestion and analytical procedures of biomasses, a standard certified biomass (DC73351) was used.
For the determination of total P and K soil samples, digestion was carried out in a microwave oven at 120 °C for 8 h with 0·25 g of ground sample in 9 mL of HNO3 (65%) and 4 mL HF (40%) (Bargagli, Brown & Nelli 1995). The digested solutions were adjusted to 50 mL final volume of HNO3 (3%), filtered with a Millex 0·45 µm filter and stored at 4 °C until required. The analytical precision for soil and bedrock analyses, as verified by parallel analyses of international (GSR-6) standards, was better than 5% for all trace elements analysed.
To determine the available P in soil, soluble Pinorganic (Pi), an extraction with 0·5 m NaHCO3 was conducted. We analysed the soil soluble Pi in 0·5 m NaHCO3 soil extracts by the Olsen Method (Olsen et al. 1954). We analysed Olsen-P soil fractions in the soil samples of January 2005 and in order to detect possible annual variations in Olsen-P fractions an additional soil sampling and Olsen-P determination was conducted in April 2005.
To determine total soil extractable P (Olsen-P), an aliquot of the NaHCO3 extracts (3 mL) was digested and analysed by the Olsen method (Olsen et al. 1954). Soil extractable organic P (Olsen-Porganic) was determined as the difference between total Olsen-P and Olsen-Pinorganic.
The extractable K fractions were analysed for each soil sample. The K extracts were obtained by shaking 2 g of soil (or pulverized bedrock) with 12 mL of solvent (0·01 m NaNO3) as according to Yin et al. (2002) and van Elteren & Budic (2004). The soil and the 0·01 m NaNO3 solvent were mixed in 50-mL plastic centrifuge tubes and were used according to the method given by Blaser et al. (2000). Two soil suspensions were prepared for each sample. The soil mixtures were equilibrated by shaking in a reciprocal shaker (100 strokes min−1) for 5 h, a technique based on batch extraction studies by Gupta & Mackay (1966). After equilibrium, soil solids were separated from the solution by centrifugation and then by filtration through a 0·45 µm pore-size membrane filter. The concentrations of K in the filtrates were determined as described above for biomass and soil digests. In order to detect possible annual variations in K concentrations in soil extracts an additional soil sampling and K concentrations in soil extracts was conducted in April 2005.
The effects of drought treatment on plant P and K concentrations and contents were investigated by t-test using plot mean values of each variable. In the case of soil P and K contents, we differentiated between soils under plastic strips (runoff plus partial rainfall exclusion) and soils between plastic strips (runoff exclusion). Thus, for soil analyses we used an anova post-hoc test (Bonferroni/Dunn) to compare the three levels of water availability (Control, D and DD). These analyses were conducted with the statview 5·01 program (Abacus Concepts, SAS Institute Inc., Berkeley, CA, USA).