Physiology–phenology interactions in a productive semi-arid pine forest


  • Kadmiel S. Maseyk,

    1. Department of Environmental Sciences and Energy Research, Weizmann Institute of Science, Rehovot 76100, Israel;
    Search for more papers by this author
  • Tongbao Lin,

    1. Department of Environmental Sciences and Energy Research, Weizmann Institute of Science, Rehovot 76100, Israel;
    Search for more papers by this author
  • Eyal Rotenberg,

    1. Department of Environmental Sciences and Energy Research, Weizmann Institute of Science, Rehovot 76100, Israel;
    Search for more papers by this author
  • José M. Grünzweig,

    1. Robert H. Smith Institute of Plant Sciences and Genetics in Agriculture, Faculty of Agricultural, Food and Environmental Quality Sciences, the Hebrew University of Jerusalem, Rehovot, Israel
    Search for more papers by this author
  • Amnon Schwartz,

    1. Robert H. Smith Institute of Plant Sciences and Genetics in Agriculture, Faculty of Agricultural, Food and Environmental Quality Sciences, the Hebrew University of Jerusalem, Rehovot, Israel
    Search for more papers by this author
  • Dan Yakir

    1. Department of Environmental Sciences and Energy Research, Weizmann Institute of Science, Rehovot 76100, Israel;
    Search for more papers by this author

Author for correspondence:
Dan Yakir
Tel: +972 89342549
Fax: +972 89344124


  • • This study explored possible advantages conferred by the phase shift between leaf phenology and photosynthesis seasonality in a semi-arid Pinus halepensis forest system, not seen in temperate sites.
  • • Leaf-scale measurements of gas exchange, nitrogen and phenology were used on daily, seasonal and annual time-scales.
  • • Peak photosynthesis was in late winter, when high soil moisture, mild temperatures and low leaf vapour pressure deficit (DL) allowed high rates associated with high water- and nitrogen-use efficiencies. Self-sustained new needle growth through the dry and hot summer maximized photosynthesis in the following wet season, without straining carbon storage. Low rates of water loss were associated with increasing sensitivity of stomatal conductance (gs) to soil moisture below a relative extractable water (REW) of 0.4, and decreased gs sensitivity to DL below REW of approx. 0.2. This response was captured by the modified Ball–Berry (Leuning) model.
  • • While most physiological parameters and responses measured were typical of temperate pines, the photosynthesis–phenological phasing contributed to high productivity under warm-dry conditions. This contrasts with reported effects of short-term periodical droughts and could lead to different predictions of the effect of warming and drying climate on pine forest productivity.


The impact of water stress on plants and ecosystems is an important aspect of biosphere–climate interactions when considering the potential effects of a warming climate. Climate projections indicate drying trends, owing to decreasing precipitation and increasing evaporative demand, in regions of most continents (Christensen et al., 2007). Water shortage is the main factor limiting plant production worldwide (Boyer, 1982; Flexas et al., 2004) and both episodic drought events and a long-term increase in aridity have profound effects on crop, forest and natural ecosystem productivity and distribution (Angert et al., 2005; Breshears et al., 2005; Ciais et al., 2005; Mueller et al., 2005).

Information on the underlying plant physiological responses is necessary for robust model parameterization and reliable predictions of forest productivity under warm and dry conditions. Recently, long-term measurements of ecosystem productivity at the driest forest site in the global Fluxnet network (the Yatir Forest, in southern Israel) have revealed that annual forest carbon (C) sequestration under warm and dry summer conditions can be equivalent to that of more mesic sites (Grünzweig et al., 2003). Estimates of net ecosystem productivity (NEP) in this forest may reach 3.5 t C ha−1 yr−1 in some years (6-yr mean = 2.1 t C ha−1; E. Rotenberg, unpublished; see also These values are similar to the European mean of 2.9 t C ha−1 yr−1 (Janssens et al., 2001).

In addition to the effects on plant physiology, the warming climate has been observed to influence phenology (Menzel, 2000; Peñuelas et al., 2002, Menzel et al., 2006, Parmesan, 2007). Often the key for the initiation of leaf development is temperature and photoperiod (Rathcke & Lacey, 1985; Dougherty et al., 1994), but may involve water availability in Mediterranean and desert plants and in tropical dry forests, where low temperatures are not a critical issue (Kemp, 1983; Borchert, 1994; Peñuelas et al., 2004). However, Mediterranean pines, including Pinus halepensis, have their origins in a pre-Mediterranean climate (Klaus & Ehrendorfer, 1989) and while some adaptation to Mediterranean climate conditions has occurred in a few species (Liphschitz & Lev-Yadun, 1986), most species also share many characteristics similar to temperate pine species. Needle phenology in pines, including P. halepensis, involves spring flushing and summer growth (Dougherty et al., 1994), although the high level of genetic divergence between populations (Bucci et al., 1998) and individual plasticity in phenological traits (Pardos et al., 2003) may account for the up to 4-wk difference observed in the timing of phenological events across different ecotypes of P. halepensis (Weinstein (1989). As a result, needle growth in P. halepensis and other Mediterranean pines occurs through the potentially most stressful period of the year.

The persistence of summer growth suggests that leaf phenology is an evolutionarily constrained feature (Herrera, 1984; Orshan et al., 1989) or that dry-summer growth may confer some advantage to the species. Dry season leaf phenology may become a feature of trees in drying regions in the future and, through affecting the amount of photosynthesizing tissue, the timing and extent of annual leaf growth will influence ecosystem productivity. Therefore, information on the interactions between phenology and physiology in dryland pine forests are of relevance to currently temperate regions, and will be required for model predictions of ecosystem responses under drier conditions.

Aleppo pine has long been recognized for its drought tolerance (Oppenheimer, 1947; Schiller, 2000), and its widespread use in reforestation and land reclamation, and ongoing natural establishment in the Mediterranean Basin makes it one of the most ecologically and economically important species in the region (Quézel, 2000; Maestre & Cortina, 2004). The species is a significant component of Mediterranean forests and will therefore play a defining role in the C balance of the region under the warm and dry conditions expected for the Mediterranean Basin. Dendrochronology studies have shown that P. halepensis productivity has shown little response to recent climatic changes, but sensitivity to increasing CO2 concentrations and consequently higher water-use efficiency may have a positive effect on productivity in the future (Rathgeber et al., 2000). While a large amount of work on various aspects of P. halepensis ecology exists (Ne’eman & Trabaud, 2000), the focus of physiological investigations has largely been limited to seedlings in controlled environments (Melzack et al., 1985; Wellburn et al., 1996; Inclan et al., 2005). Field-based studies under natural conditions have shown that transpiration occurs predominantly in the winter season (Schiller & Cohen, 1998).

In this study, the relationship between dry summer leaf phenology and photosynthetic activity is explored to provide insights into the way such interactions can contribute to the high ecosystem productivity observed in a dryland pine forest.

Materials and Methods

Site description and meteorological measurements

The study was conducted in a 2800 ha mature (35–40 yr old) Aleppo pine (P. halepensis Mill.) forest (31°21′N, 35°03′E, 650 m above sea level (a.s.l.)) located in the southern Hebron Mountains, Israel, between October 2000 and April 2005. The selected stand had a density of approx. 300 trees ha−1, with mean tree height of 10 m, mean diameter at breast height (DBH) of 17 cm, and leaf area index (LAI) of 1.5 (Sprintsin et al., 2007). The average annual rainfall over the lifetime of the forest is 280 mm, having fluctuated between 147 mm and 496 mm, and was 325 mm over the study period. Typically, rainfall is between November and March, resulting in a dry season of 6–7 months. The seasonal cycle is defined by the hydrological regime, and the year is denoted to start on October 1st of each calendar year (day of season, DOS = 1). The average annual air temperature is 18.2°C, with the mean monthly daytime air temperatures ranging from 10°C in January to 25°C in July. The aridity index (ratio of precipitation to potential evapotranspiration) is 0.18, which is typical of arid regions. The soil is shallow and poor (0.2–1 m deep rendzina soil above chalk and limestone) with a deep (approx. 300 m) ground water table. Environmental conditions, including photosynthetically active radiation (PAR), air temperature, vapour pressure deficit (D) and precipitation, were monitored continuously 5 m above the canopy on an eddy covariance tower at the site as described in Grünzweig et al. (2003), and are shown in Fig. 1.

Figure 1.

Site meteorological variables over the study period. (a) Monthly precipitation; (b) daily soil water content (SWC); (c) daily maximum vapour pressure deficit (Dmax); (d) average daily air temperature; (e) daily maximal photosynthetically active radiation (PAR).

Volumetric soil water content (SWC) to 30 cm depth was determined from gravimetric measurements every 1–3 wk during winter and every 2–6 wk during summer for the years 2000–2003, and converted to volumetric units using soil bulk density. From 2003–2005, volumetric SWC to 30 cm depth was continuously measured with three Time-Domain Reflectrometry sensors (CS616; Campbell Scientific, Logan, UT, USA). Daily SWC values for 2000–2003 were estimated using an empirical model developed from the period of continuous measurements, constrained by the gravimetric data. The increase in SWC at each rain event was determined as a function of pre-event SWC and the event rain amount, using separate functions for SWC above and below 15%. Soil drying between events was determined from a first-order exponential decay function fitted to gravimetric measurements where there were three or more measurements uninterrupted by rain events. Where gravimetric measurements were insufficient, the rate of drying was determined as a function of initial SWC. Volumetric SWC was scaled according to the minimum soil water content (5%) and field capacity (38%) and expressed in terms of plant relative extractable water (REW) according to Granier et al. (1999).

Phenology measurements

Current year (y0) needle growth was measured periodically using digital callipers (accuracy of 0.1 mm) through the growth period (March–October) on three apical shoots from 12 trees. Litterfall was collected in 25 litter traps of 0.5 m2 each. Litter was removed from the traps every 1–2 months and sorted into needle, reproductive, woody and residual fractions.

Leaf nitrogen content

Leaf nitrogen (N) concentration was measured on needles collected through 2001 and 2002. Needles were oven dried (48 h at 80°C) and ground to pass through 250 µm mesh. The N content was determined in an elemental analyser (EA 1108; Carlo-Erba, Milan, Italy) and is expressed on a % dry weight (DW) basis.

Gas exchange measurements

Net photosynthetic rates (A), leaf transpiration (E), stomatal conductance to water vapour (gs) and intercellular CO2 concentration (Ci) were measured in situ with an LI-6400 photosynthesis system equipped with an LI-6400-01 CO2 injector (Li-Cor Inc., Lincoln, NE, USA), calibrated regularly with standard gasses and an LI-610 Dew Point Generator (for H2O; Li-Cor Inc.). All gas exchange parameters were expressed on a projected needle area basis (determined using an LI-3000 leaf area meter; Li-Cor Inc.). Intrinsic and instantaneous water-use efficiency (WUE) was calculated as A/gs and A/E, respectively.

The first measurements on the current year needles were made in May or June, when needles were c. 50% of their final length. Needles of this age class were designated y0 needles, and the older age class (previous year needles) were designated y1 needles. Needles remained in their age class until the first measurements on the new needles the following year. Occasional measurements were made on 2-yr-old needles (y2 age class, needles persist on the tree for usually between 2 and 3 yr).

Instantaneous gas exchange measurements were made on both current year and 1-yr-old needles throughout the study period under light saturating conditions. Measurements were made on 12 trees at mid-morning between 09 : 00–11 : 00 h in winter and 08 : 00–10 : 00 h in summer (local standard time) on relatively clear days, initially at c. 2-wk intervals, and at 3–4 wk intervals over the last 18 months. Ambient temperature, relative humidity, PAR and CO2 concentration were maintained in the leaf cuvette, with the exception that the Li-6400-02B red-blue LED light source (Li-Cor Inc.) was used (set to 1200 µmol m−2 s−1) when ambient irradiance was low and intermittent cloudiness prevented stable readings. Diurnal patterns of leaf gas exchange under ambient conditions were measured on 10 occasions over the course of the study period at different times of the year. In the diurnal measurements, gas exchange measurements were made five or six times over the course of the photoperiod.

The photosynthetic responses to PAR (A/Q) at ambient CO2 concentration was measured under 8–12 levels of decreasing PAR from 1400–0 µmol m−2 s−1, supplied by the LI-6400-02B red-blue LED light source. The photosynthetic response to Ci (A/Ci) at saturating PAR (1200 µmol m−2 s−1) was measured under 9–11 different CO2 concentrations, controlled by the LI-6400-01 CO2 injector and adjusted in approx. 50 p.p.m. steps. The A/Ci procedure involved starting at ambient CO2 concentration, decreasing to 50 p.p.m., returning to ambient, and then increasing to 1200–1500 p.p.m. Measurements for both A/Q and A/Ci response curves were taken after maintaining the needles for 5 min at each CO2 or light level. Over summer, lack of a stomatal response made it difficult to obtain meaningful response curves, so data analysis was confined to those curves determined between November–June and November–May for the A/Q and A/Ci response curves, respectively.

Data analysis and parameter estimation

The light response data were fitted with an asymptotic exponential model by nonlinear regression (Iqbal et al., 1996):

image(Eqn 1)

(Am is maximum (gross) photosynthetic rate at high light; α is the quantum efficiency of CO2 fixation; Q is the PAR level (data at PAR = 0 were excluded); and Rd is dark respiration rate). The light intensity required to saturate A (PARsat) was determined as the light intensity at 95% of Am: PARsat = ln(0.05)Am/−α.

The CO2 response data were fitted to the well-described model of photosynthesis (Farquhar et al., 1982) following studies on other Pinus species (Leuning, 1995; Walcroft et al., 1997), using nonlinear regression to estimate the maximum rates of carboxylation (Vcmax) and electron transport (Jmax). Data with Ci less than 200 µmol mol−1 were used to determine Vcmax from:

image(Eqn 2)

(Γ* is the CO2 compensation concentration in the absence of day respiration; Kc and Ko are the Michaelis–Menton constants for carboxylation and oxygenation, respectively; Oi is the intercellular partial pressure of oxygen; Γ*, Kc and Ko are temperature dependent parameters, and were calculated according to the leaf temperature during measurement) (Leuning, 1995; Walcroft et al., 1997).

Using the calculated values of Γ* and the values of Rd determined from Eqn 2, the value of J, the electron transport rate, was determined from:

image(Eqn 3)

A nonrectangular hyperbola function of J was used to determine Jmax:

image(Eqn 4)

(φ is the quantum yield of electron transport and θ describes the convexity of the curve and were taken as 0.2 and 0.9, respectively) (Walcroft et al., 1997).

The temperature dependence of Vcmax and Jmax was modelled by fitting the values determined for Vcmax and Jmax to the peaked equation:

image(Eqn 5)

(Vcmax0 is the value of Vcmax (substituted with Jmax0 and Jmax, respectively) at a reference temperature T0 (298 K); Ha and Hd are the energies of activation and deactivation, respectively; R is the universal gas constant (8.31 J mol−1 K−1); Sv is an entropy term).

The stomatal response to leaf vapour pressure deficit (DL) was analysed using a Lohammer-type hyperbolic function (Lohammer et al., 1980):

image(Eqn 6)

(gsm is the estimated maximum gs at a DL of 0; D0 describes the sensitivity of gs to DL). To separate the effects of the co-occurring seasonal decreases in soil water content and increases in DL on gs, Eqn 6 was fitted to data obtained when REW > 0.5, where soil water content was assumed to not be limiting gs (Granier et al., 2007), providing a reference gs value (gsref) for each DL. The measured gs data were then expressed relative to the gsref value estimated for the measurement DL.

Two commonly used empirical models describing the coordinated changes between assimilation and stomatal conductance, the Ball–Berry model (Ball et al., 1987) and the Leuning modification (Leuning, 1995) were evaluated using the stomatal conductance data from the seasonal and diurnal measurements. The Ball–Berry model describes gs by:

image(Eqn 7)

and the Leuning model by:

image(Eqn 8)

(RH is relative humidity; Ca is the CO2 concentration at the leaf surface; Γ is the CO2 compensation concentration). The Leuning model incorporates the Lohammer function of gs sensitivity to DL and the values of g0 and a are the intercept and slope determined by linear least-squares regression.


Needle phenology

Needle growth occurred between March and September (i.e. through the dry season), and both the period of growth (approx. 200 d) and the relative growth rate through the season was highly conservative between years (Fig. 2a). The average date of budburst of the 3 yr was Julian day 82 (SD = 4), or March 22. The Spring-Warming Model described in Chuine & Cour (1999) was used to estimate the sum of degree-days (forcing units, F*) from a fixed onset date of November 1 until bud-burst. The threshold base temperature giving the lowest variance in F* between years was 7°C, with an average F* of 796 (SD = 8.3). Final needle lengths were 68.5 ± 1.8 mm, 58.8 ± 1.6 mm and 57.0 ± 2.3 mm in 2002, 2003 and 2004, respectively (mean ± SE).

Figure 2.

Needle phenological characteristics of Pinus halepensis measured at the Yatir Forest site. (a) Phenograms of needle growth (open symbols) and litterfall (closed symbols) relative to the seasonal maximum (i.e. final needle length and total litterfall). Data are from 3 yr normalized to the seasonal maximum, and curves are fits to the data of all years combined. A Gompertz growth function was fitted to the normalized needle growth data: Nr = 107 · exp[−exp(−0.194(x – 234))], r2 = 0.99, where Nr is needle length relative to the final length and x is day of season (DOS). The seasonal litterfall was separated into two periods (before and after needle budburst). Separate Boltzmann sigmoidal curves were fitted to the normalized litter retention data: Lr = (A1 + A2)/(1 + exp(x – x0)/dx) + A2, where Lr is the relative litter retention (proportion of total seasonal litterfall still on the trees) and x is DOS. A1, A2, x0 and dx are empirically fitted parameters and are 1.0, 0.806, 53.0, 22.5, respectively, for DOS 1 – 165 (r2 = 0.99), and 0.80, 0.0, 282.5, dx = 17.7 (r2 = 0.99), respectively, for DOS 166–365. (b) Seasonal leaf area index (LAI) dynamics based on the phenology patterns in (a), with the left axis showing a normalized scale of LAI relative to the seasonal maximum and the right axis showing the absolute LAI, assuming a constant LAI of 1.5 between years. (c) Seasonal dynamics of needle N content in y0 (open symbols), y1 (black symbols) and y2 (grey symbols) needles. Different symbol shapes represent data from different years, as indicated for the y0 needles in the legend, and the same symbol shapes refer to the same years across the needle age classes. The dashed vertical line represents the point of emergence of the new needles for that season and the change in age class notation for needle cohorts (e.g. y0 of 2000–2001, i.e. open squares, become y1 of 2000–2001, i.e. black squares). Tick marks indicate the start of each month.

Litterfall occurred in two phases. The majority (80%) fell concurrently with needle growth, predominantly between June and August (Fig. 2a), and the remaining 20% fell over the winter period (October–March). Similarly to needle growth, the relative rate of litterfall through the season was very similar between the years. Needle litterfall totals were 135 ±6 g m−2, 132 ± 10 g m−2 and 148.0 ± 8 g m−2 in 2001–2002, 2002–2003 and 2003–2004, respectively (mean ± SE, in g dry mass).

Seasonal leaf area index

Using the general phenology equations (Fig. 2a), representative intraseasonal LAI dynamics for the case of a constant LAI between seasons (i.e. litterfall = growth) were reconstructed by scaling total growth and litterfall data according to the normalized phenology equations and are shown in Fig. 2b. The representative pattern shown uses an initial LAI of 1.5 and a value of 140 g m−2 (= 0.615 m2 m−2) for the litterfall data, converted to area using specific leaf area data from the site (4.39 10−3 m2 g−1, SD = 0.86 10−3, n = 88, representing samples from two canopy heights, three age classes and three plots). There was a c. 10% variation both above and below the initial seasonal value through the year as a result of the growth and senescence patterns. Maximum LAI occurred in the middle of June, when the new season's growth was approx. 55% complete, adding to the previous year and older needles. By the start of the wet season, the canopy consisted of a full cohort of newly developed (y0) needles, a cohort of previous years (y1) needles and approx. 45% of a cohort of y2 needles. The minimum LAI occurred in the middle of February, following the winter litterfall (Fig. 2), which consisted of about half of the remaining y2 needles.

Leaf nitrogen

There was a seasonal fluctuation in leaf N that was qualitatively similar for all age classes and between years (Fig. 2c). The N content of mature needles was highest between January and May (0.90–1.19% DW, excluding the y0 value soon after budburst, i.e. y0 in April 2002), and lowest over summer and autumn (0.69–0.95% DW). The high initial N content of the new needles (1.65% DW) reflected N loading of new leaves at the start of growth (Millard, 1994), which constituted c. 30% of the total N in the new needles by the end of growth.

Seasonal and diurnal leaf gas exchange

Seasonal patterns of Asat, E and gs were similar (Fig. 3a–c), and the seasonal and diurnal variation in Asat was closely coupled with gs, described by an asymptotic exponential relationship common between age classes and years (Fig. 4, r2 = 0.94). Photosynthesis rates were maximal in the wet season (Asat of 12–18 µmol m−2 s−1, with gs of 0.1–0.25 mol m−2 s−1), typically between January and May, and low (0–4 µmol m−2 s−1, with gs of 0–0.05 mol m−2 s−1) in the five dry months between July and November. However, in the drier year of 2003–2004, the high rates were only observed between February and April (Fig. 3a). The reductions in gs more than compensated for the seasonal increase in atmospheric vapour pressure deficit (from approx. 1 kPa to approx. 4 kPa), as E declined from a peak of c. 4 mmol m−2 s−1 in March and April to < 1 mmol m−2 s−1 in the summer months (Fig. 3b,e). Midday reductions in stomatal conductance in the warmer and drier months (May–August, Fig. 3f) resulted in midday decreases in net CO2 assimilation rates, which were near zero or negative at midday in mid-summer (i.e. June and August, Fig. 3d). However, even though assimilation was highly suppressed for much of the day, some photosynthetic activity continued through the entire seasonal drought period. Analysis of all the morning peak activity data (Fig. 3a) showed that the measured values were significantly greater than zero (P > 0.05) on all but two occasions (September 11, 2002, and November 4, 2003), during occasional heat waves that characterize this system and result in high vapour pressure deficits all day.

Figure 3.

Seasonal (a,b,c) and diurnal (d,e,f) patterns of gas exchange parameters (CO2 assimilation rate, A; transpiration, E; and stomatal conductance, gs) in Pinus halepensis measured on mature trees under field conditions. Seasonal measurements were made under light-saturated conditions during daily peak activity, and values of A are referred to as Asat in (a). The different symbols represent the same year in (a), (b) and (c), as indicated in the legend in (a). Diurnal measurements were made at different times of the season, as indicated in the legend in (d), which refers also to (e) and (f). Data shown is for fully elongated needles from the October in the year of their growth until the following October. For clarity, error bars (standard errors of the means) are shown on the diurnal data only, but are representative of the errors in the seasonal data. The average standard error from across the dataset of each parameter in the seasonal data is shown in the upper left corner of (a–c).

Figure 4.

The relationship between stomatal conductance (gs) and light saturated CO2 assimilation rate (Asat) in y0 (open symbols) and y1 (closed symbols) Pinus halepensis needles from all seasonal mid-morning and diurnal (for photosynthetically active radiation (PAR) > 1000 µmol m−2 s−1) measurements. The curve is an asymptotic exponential fit to all data (Asat = 18.61 – 19.80 · 0.00009 gs, r2 = 0.94).

The first gas exchange measurements of the new season's needles (y0) were made in late May, when they were c. 40–50% of their final length and had photosynthetic rates of c. 30% of the y1 needles (not shown). By the end of June (c. 65% final length) area based photosynthetic rates of the y0 needles were equivalent, or greater than, the y1 needles. The y0 photosynthetic rates over the latter part of the growth period (July–September) were weighted according to the growth curve in Fig. 2a for an estimate of their contribution to shoot photosynthesis relative to the y1 needles over this period. The shoot-based ratio of y0 : y1 photosynthesis (assuming equal numbers of y0 and y1 needles in a shoot) was 1.01 ± 0.23 (n = 5), 1.04 ± 0.24 (n = 4) and 1.10 ± 0.28 (n = 6) in the months of July, August and September, respectively (mean ± SD of measurements from the same month across the years pooled). Occasional measurements on y2 needles indicated photosynthetic rates of < 80% of the co-occurring y1 needles (not shown).

Much of the variation in Vcmax and Jmax determined from the A-Ci response curves over the wet season was related to temperature (r2 = 0.856 and 0.721 for Vcmax and Jmax, respectively, when fitted with Eqn 5; not shown). The parameters from the temperature dependence of Vcmax and Jmax (Eqn 5) are shown in Table 1. There was a strong linear correlation between Vcmax and Jmax, (Fig. 5, r2 = 0.82; P < 0.0001) but the Jmax : Vcmax ratio decreased from 2.3 to 1.0 as temperature increased (regression equation: ratio = − 0.029Tleaf + 2.269; r2 = 0.19; P = 0.002; not shown).

Table 1.  Parameters values describing the temperature dependence of Vcmax and Jmax (Eqn 5)
  1. As maximum temperatures were not reached for either Vcmax or Jmax, a constant value of 200 kJ mol−1 was used for Hd for both Vcmax and Jmax (Medlyn et al., 2002). The Sv term was determined from the Vcmax response and the same value was used for fitting the Jmax response.

Sv0.642J mol−1 K−1
Vcmax0113.1µmol CO2 m−2 s−1
Jmax0175.5µmol e m−2 s−1
Ha (Vcmax)67.39kJ mol−1
Ha (Jmax)57.55kJ mol−1
Figure 5.

The relationship between the maximum rates of carboxylation (Vcmax) and electron transport (Jmax) in Pinus halepensis needles from y0 (open symbols) and y1 (closed symbols) age classes. The regression is fitted to data form both age classes: Jmax = 1.30Vcmax + 26.3 (r2 = 0.82; P < 0.0001).

Influence of soil water content on gas exchange parameters

The ratio of gs to gsref is shown in relation to REW in Fig. 6a, where gsref was the gs estimated for the measurement DL under nonlimiting soil water content (REW > 0.5) using Eqn 6. When REW was above c. 0.4 (equivalent to 20% vol. SWC), gs was variable but without a clear dependence on soil water content. Below an REW of 0.4 gs decreased linearly with REW. Therefore the point at which soil water deficits start to limit transpiration and photosynthesis, defined as the critical REW (REWc, Granier et al., 1999), lies at about a REW of 0.4 in this system, based on soil water content measurements of the upper 30 cm.

Figure 6.

The relationship between relative extractable water (REW) and gas exchange parameters in Pinus halepensis: (a) the ratio of gs to a reference gs (gsref, gs estimated from the relationship with DL at high soil water content); (b) the ratio of A/E; (c) the ratio of A : gs; (d) photosynthetic nitrogen (N)-use efficiency (PNUE = Asat/N); (e) quantum efficiency (α) from A/Q response measurements. Data in (a–c) are from fully expanded y0 needles only, data in (d) and (e) are from both y0 and y1 needles. Curve fits in (a) and (e) are separate linear regressions to data below and above REW of 0.4: (a) y = 2.04x + 0.029, r2 = 0.66, for REW < 0.4 and y = 0.088x + 0.924, r2 = 0.003, for REW > 0.4 and (e) y = 0.142x + 0.007, r2 = 0.53, for REW < 0.4 and y = 0.005x + 0.053, r2 = 0.01, for REW > 0.4; in (b) is an asymptotic exponential fit: y = 5.43 – 5.02 · (6 × 10−7)x, r2 = 0.13; in (c) is a log-normal fit: y = −39.1 + 718/(sqrt(2π) · 1.79x) · exp(−(ln(x/4.87))2/(6.41)), r2 = 0.38; and in (d) is an asymptotic exponential fit: y = 6.91 – 6.45 · 0.032x, r2 = 0.86.

Over much of the seasonal variation in soil moisture, including periods below REWc, the instantaneous WUE of photosynthesis (or transpiration ratio, A/E) was between 3 and 6 mmol CO2 mol−1 H2O (Fig. 6b), with an average value of 5.1 mmol CO2 mol−1 H2O across all measurements. The intrinsic WUE (A/gs), on the other hand, showed a general increase as soil water content decreased, from values at saturation of c. 70 µmol CO2 mol−1 H2O to reach maximal values of c. 150 µmol CO2 mol−1 H2O at low REW (Fig. 6c). At very low soil water contents (REW < 0.1) and gs values (Fig. 6a), both A/E and A/gs approached zero. These periods were also typically associated with high temperatures and DL > 4 kPa (see Fig. 7).

Figure 7.

Relationship between leaf vapour pressure deficit (DL) and (a) gs and (b) A/E from the light-saturated gas exchange measurements. The different symbol shadings represent data from different soil water content at the time of measurement and are the same for both (a) and (b): open symbols for a relative extractable water (REW) < 0.2; grey symbols for REW of 0.2–0.4 and black symbols for REW > 0.4. The curves in (a) are the fits of Eqn 6 to data from each soil water class: solid line for REW 0.0–0.2 (gsm = 0.065, D0 = 1.89, r2 = 0.37); dashed line for REW 0.2–0.4 (gsm = 0.303, D0 = 1.08, r2 = 0.56); dotted line for REW 0.4–1.0 (gsm = 0.372, D0 = 1.39, r2 = 0.41). The fit to the A/E data in (b) is an exponential function: A/E = 14.1 · exp(−DL/2.02) + 0.227 (r2 = 0.79).

The seasonal variation in Asat and in leaf N were correlated when Asat was above approx. 4 µmol m−2 s−1 (r2 = 0.76 for the linear regression between Asat and leaf N, P < 0.0001, n = 13, not shown). The ratio, Asat : N (photosynthetic N-use efficiency, PNUE) was between 5.3 and 7.5 µmol CO2 g−1 N s−1 when soil REW was above c. 0.3 (Fig. 6d). The proportionally larger decrease in Asat relative to changes in leaf N at low SWC resulted in a decrease in PNUE of 50–80% to values between 0.9–3.2 µmol CO2 g−1 N s−1.

There was no difference in the average values of PARsat and α between the age classes. The average values of α and PARsat were 0.046 ± 0.003 mol mol−1 and 866 ± 35 µmol m−2 s−1 (mean ± SE), respectively, for the data of both age classes combined across all measurements (n = 54). Both PARsat and α were highest in March and April, when soil water and photosynthetic rates were high, and decreased sharply as REW decreased below 0.4, shown in Fig. 6e for α.

Vapour pressure deficit as a driver of water use efficiency

Stomatal response was analysed in response to changes in both soil water and DL, and three distinct soil moisture regimes could be defined. First, at REW > 0.4, when gs was insensitive to soil moisture (see Fig. 6a), there was a typical hyperbolic decline in gs in response to increasing DL (r2 of fit with Eqn 6 of 0.41; Fig. 7a). Second, when REW was intermediate, between 0.2 and 0.4, there was an apparent decrease in maximal gs and increase in the sensitivity of gs to DL compared with the higher soil moisture regime (i.e. gsm decreased from 0.372 to 0.303 and the D0 parameter decreased from 1.39 to 1.08; Fig. 7a). However, the influence of co-occurring decreases in REW with increasing DL in this transition period can confound the gs response to DL. Third, when REW was less than c. 0.2, there was a clear separation in the response of gs to DL compared with the other regimes, with reductions in both the maximal gs (to 0.065 mol m−2 s−1) and the sensitivity of gs to DL (D0 parameter of 1.87).

There was an exponential decrease in A/E with increasing DL that was independent of the soil water class (r2 = 0.79; Fig. 7b); hence, DL explained much of the variation in A/E observed across the range of soil moistures (Fig. 6b). Low values of A/E (A/E < 2 mmol mol−1) were observed at low soil moisture when associated with concurrently high DL. A/E values were high even at low soil moisture, when DL was low (DL < 2 kPa), such as in autumn before the rain arrived (A/E 4–12 mmol mol−1; Fig. 7b).

Stomatal model parameters

The slope, a, and intercept, g0, derived from the regressions of gs with the relevant model index of both the Ball–Berry and Leuning modified Ball–Berry model were determined for all data combined, and on the basis of separation into the three soil water content divisions (Fig. 8, Table 2). In all cases the correlation between gs and the relevant index was linear and significant (P < 0.0001), with the Ball–Berry model performing better at the mid and low soil moisture, and the Leuning model giving better fits at the high SWC and with all data pooled.

Figure 8.

The relationship between stomatal conductance (gs) and (a) the Ball–Berry model index (A · RH/Ca) and (b) the Leuning model index (A/[(Ca – Γ)(1 + DL/D0)]). The different symbol shadings represent data from different soil water content at the time of measurement and are the same for both (a) and (b): open symbols for a relative extractable water (REW) < 0.2; grey symbols for REW of 0.2–0.4 and black symbols for REW > 0.4. The thick solid line is the linear fit to all data combined and the thin lines are regressions of the separate soil water classes; solid line, REW < 0.2; dashed line, REW 0.2–0.4; dotted line, REW > 0.4. Values of the slope and intercept for all regressions are given in Table 2.

Table 2.  Values of the intercept (g0) and slope (a) derived from the relationship between gs and the model indices of the Ball–Berry and Leuning models of stomatal conductance (Fig. 8), determined for the separate soil water (relative extractable water, REW) classes and across all data combined
REW rangeBall–Berry modelLeuning model
< 0.20.0125.040.653< 0.00010.01637.50.528< 0.0001
0.2–0.40.0405.790.702< 0.00010.04715.90.650< 0.0001
> 0.40.0717.040.510< 0.00010.06415.10.655< 0.0001
0–1.00.0119.660.806< 0.00010.02819.10.879< 0.0001


Needle and photosynthesis phenology are off phase

The early spring peak-photosynthesis in our system is characteristic for the Mediterranean climate (Harley et al., 1987; Epron et al., 1992), but is 3–5 months earlier than the summer maximum of conifers in temperate regions (Teskey et al., 1994; Rundel & Yoder, 1998). Stomatal regulation and tight coupling of assimilation with gs (Fig. 4) is a key element of plant performance in dry environments (Harley et al., 1987; Damesin & Rambal, 1995; Flexas et al., 2001). In our system, the imposition of stomatal limitations to gas exchange occurred here when REW value decreased below c. 0.4 (Fig. 6a), a value characteristic of many temperate forests (Granier et al., 2007). Irrespective of soil moisture, stomatal conductance was sensitive to DL, and even at high soil water contents gs could be reduced to below 0.1 mol m−2 s−1 (and assimilation rates to approx. 50% of maximal observed rates) when DL increased above approx. 2 kPa (cf. Figs 4, 7a). However, when REW was below a critical point of approx. 0.2, both maximal gs (approx. 0.05 mol m−2 s−1) and sensitivity to DL were greatly decreased, resulting in greatly reduced water loss over summer and autumn (Schiller & Cohen, 1998). The strong coupling between stomatal conductance and photosynthesis and the regulation of gs to limit water loss shows that the seasonality of photosynthesis seems, therefore, to be primarily driven by that of soil water content and atmospheric DL.

By contrast, needle phenology at Yatir was fundamentally that of temperate pines, with summer growth and senescence (Fig. 2; Dougherty et al., 1994). The time of bud-burst observed at our site (mid-March) is earlier than other P. halepensis sites in the region (Weinstein, 1989; Borghetti et al., 1998; Mediavilla & Escudero, 2003), but consistent with variation often observed across geographical gradients whereby phenological events occur earlier in the season at warmer sites (Weinstein, 1989; Kramer, 1995; Ahas et al., 2002). It seems, therefore, that while temperature and photoperiod maintain a constrained phenology of leaf elongation, leaf physiology showed the necessary plasticity for a large change in the seasonality of photosynthesis under the dry-summer conditions. Consequently there was a separation between the main period of C gain and needle growth not typically observed in pines.

The separation of photosynthesis and growth phenology may, however, confer an advantage in terms of maximizing whole-canopy photosynthesis. The cohort of new needles had fully expanded by the start of the following wet season (for an analogous pattern in dry tropical forest see Singh & Singh, 1992). Photosynthetic capacity is typically highest in younger needles (Warren, 2006), and was similarly high between the two age classes at our site that made up the majority of the canopy at the start of the season. Hence there was no lag in the development of maximal leaf area of new foliage once conditions for photosynthesis were favourable at the onset of the short wet season. The advantage of the observed leaf phenology is further enhanced by the results that support the evidence that, despite the low rates, photosynthesis during the growth phenophase, rather than winter period storage, contributes significantly to needle growth (Klein et al., 2005). Indeed, estimates of ecosystem C gain integrated over the growth period made from flux measurements at the site (106 ± 45 g C m−2 between March and September, mean ± SD of 6 yr, not shown), were consistent with the biomass in the new foliage growth (84 ± 18 g C m−2; K. Maseyk, unpublished). While it is possible that C assimilated during the period of higher photosynthesis early in the growth phase (March–April) is transiently stored and used to sustain growth later in summer, it appears that the long growth phase enables needle growth to be sustained under the limiting summer conditions without relying on seasonal storage.

High nitrogen-use efficiency underlie high photosynthetic rates during the active season

Leaf N concentrations over winter and spring (1.0–1.2%) were similar to those reported elsewhere for unfertilized P. halepensis (0.9–1.25%, (Mediavilla & Escudero, 2003; Sardans et al., 2005). The maximal photosynthetic rates (15–18 µmol m−2 s−1, similar between the two youngest age classes of needles) were relatively high compared with other pines (Teskey et al., 1994; Rundel & Yoder, 1998; Panek & Goldstein, 2001), indicating a relatively high allocation of N to photosynthesis. The maximal photosynthetic N-use efficiency values (PNUE, A/N), observed in winter, were 6–7 µmol CO2 g−1 N s−1 (Fig. 6d, equivalent to 80–100 µmol CO2 mol−1 N s−1), compared with, for example, values of c. 4 µmol g−1 s−1 reported for Pinus pinaster and Pinus ponderosa (Bond et al., 1999; Warren, 2006).

The relatively high PNUE values are consistent with the observations of photosynthesis parameters. While the range of Vcmax and Jmax values and the nature of their response to temperature across the seasonal scale were consistent with other pines, values of Vcmax and Jmax at 25°C (113.1 and 175.5 µmol m−2 s−1, respectively; Table 1) were at the upper range reported for other Pinus species (Walcroft et al., 1997; Medlyn et al., 2002; Warren et al., 2003). The relationship between Jmax and Vcmax (slope of 1.3 in the Jmax–Vcmax relationship, Fig. 5, and a value of 1.55 for the Jmax : Vcmax ratio at 25°C, Table 1) was low relative to other C3 plants and pines (e.g. slopes of 1.92 (Leuning, 2002) and 1.67 (Medlyn et al. 2002)) indicating a relatively higher N allocation to carboxylation compared with electron transport processes.

That high leaf N and allocation to photosynthesis occurred during the period favourable for photosynthesis is a further consequence of the decoupling of photosynthesis and leaf growth. The decrease in leaf N content in the mature needles over summer was consistent with vigorous remobilization to support the current season's new growth (Nambiar & Fife, 1991), but these growth demands were not acting as a N sink alternative to photosynthetic components during peak activity. The decrease in leaf N from winter maximums of 1.1 to 0.8% in summer (Fig. 2c) represents resorption of c. 30% of leaf N. These resorption estimates are within the range of values reported for other pines (e.g. 20–40%; Nambiar & Fife, 1991), although rates of resorption from senescing needles (55% of leaf N; Grünzweig et al., 2007) were relatively high compared with other studies (e.g. range of 15–50%; Aerts, 1996; Cherbuy et al., 2001; Sardans et al., 2005). Soil N uptake by the trees is likely to be low over summer, as soil N availability is low (Gelfand & Yakir, 2008) and drought impairs root nitrate reductase activity in P. halepensis (Wellburn et al., 1996). Therefore, foliage development relies on internal N reallocation, but this can be independent of soil N availability even under wet conditions (Nambiar & Fife, 1991; Proe et al., 2000). Maintaining a degree of independence between leaf growth and soil N availability may therefore represent an adaptation that confers an advantage in terms of buffering phenology against drought stress in some evergreen species.

High photosynthetic capacities maintain high water-use efficiencies

An important aspect of dryland productivity is the amount of C assimilated for the available water. The average instantaneous WUE (A/E) across all measurements was 5.1 mmol CO2 mol−1 H2O, compared with typical values in pines of 2–4.4 mmol mol−1 (Teskey et al., 1994), and were high relative other Mediterranean species (approx. 3 mmol mol−1 at 2.5 kPa; Moriana et al., 2002; Llorens et al., 2003). The A/E was strongly dependent on DL in a nonlinear manner and largely independent of soil water content, so for a given DL a similar WUE was observed under both high and low SWC (cf. Figs 6b and 7b). The winter–spring seasonality of photosynthesis, with the highest rates of assimilation when transpiration demand was relatively low (DL was typically < 2.5 kPa when soil water was not limiting; see Figs 1 and 7) results in particularly high WUE in the active season.

Similar relationships between A/E and DL, independent of SWC, have been observed at leaf and canopy scales and are attractive for their use in estimating carbon fluxes if DL and E can be estimated (Verhoef et al., 1996; Moriana et al., 2002). However, unlike the results discussed above, the relationships between DL and A/E have also been reported to be linear (Ponton et al., 2006) and dependent on soil water content. It is therefore important to further explore to what extent changes in factors such as photosynthetic capacity and plant hydraulic conductance may be affecting the seasonal A/E relationship with DL (Law et al., 2002). Note also that the unique nature of our research site enabled observations under a large range of DL (< 1–5 kPa) at high and low soil contents, providing a more extensive view of the response to DL under natural settings than is typically observed in other forest systems.

Significant reductions in PNUE and quantum efficiency were observed at soil water contents below REWc (Fig. 6d,e), suggesting a reduction in photosynthetic capacity associated with the seasonal reductions in leaf N (Fig. 2). However, intrinsic WUE (A/gs) generally displayed a typical increase with decreasing soil water (Fig. 6c), and was only really depressed at very low REW when air temperatures and DL were also high. Dry season maximum values of A/gs were comparable to other Mediterranean trees and shrubs (80–120 µmol mol−1; Flexas et al., 2001; Gulias et al., 2002). The high levels of intrinsic WUE from the gas exchange measurements are consistent with measurements of organic matter Δ13C, which also indicate a high intrinsic WUE in P. halepensis (Ferrio et al., 2003; Klein et al., 2005). Therefore, it can be concluded that the phenology related reductions in N were not limiting photosynthesis given the low stomatal conductance in summer, but the high leaf N and photosynthetic rates over winter contributed to an efficient use of the available water when evaporative demand was low.

Stomatal model parameters

The wide range of field conditions under which our measurements were made, including high and low DL across the range of soil water contents, provided a good opportunity to test two empirical stomatal models under drought conditions. Such models are commonly used in coupled photosynthesis–conductance models for canopy and ecosystem scale simulations of carbon and water fluxes. Therefore, it is important to evaluate their performance under conditions relevant for simulating responses to warmer and drier conditions. Both the Ball–Berry model and Leuning (1995) formulation that replaced RH with a function of DL performed well, although when treating all data as one set, the Leuning form gave a better overall fit (Fig. 8). The value of the slope (a) in the Ball–Berry model from all data combined (9.19) was similar to that observed in other species and environments (9.31; Ball et al. 1987, 8.88; (Xu & Baldocchi, 2003), but different slopes were observed when data were separated on the basis of the three REW phases. Although some authors have found that a is unaffected by water stress (Xu & Baldocchi, 2003), a has been shown to decrease with water stress in Mediterranean systems (Sala & Tenhunen, 1996) and we observed that a was lower when REW was below REWc. Using the Leuning model resulted in a more constrained relationship between the high and mid soil water content classes, where gs was sensitive to DL and the data covered a range of assimilation rates and conductances.


The close association in the timing of leaf phenology and seasonal photosynthesis typical of pines was separated in this system and resulted in a combination of factors that helped maximize annual scale photosynthesis. High allocation of N to photosynthesis, maximal LAI of high photosynthetic capacity needles, and a high WUE of photosynthesis contributed to high rates of C gain in the wet season. Intrinsic photosynthetic capacity was not compromised over the long summer drought, enabling opportunistic use of favourable periods for C gain (diurnally and seasonally).

We have observed a number of responses and traits that help explain the relatively high productivity observed in this semi-arid pine forest (c. 2.3 t C ha−1, similar to many mesic sites). Furthermore, at the basis of our observations were climate responses of photosynthesis and leaf phenology that are typical of pines and conifers. The broader application of such observations from climatic transition zones to understand forest response to long-term (regular) seasonal drought is apparent when considering the persistent predictions of warming and drying for Southern Europe, the entire Mediterranean basin and other regions (Christensen et al., 2007). These responses, adjustments and high productivity are clearly different from those observed following short-term periodical drought (e.g. the transition to net CO2 release in the 2003 drought in Europe; Ciais et al., 2005) and would necessarily lead to different predictions.


We thank Y. Moshe and the JNF for cooperation and logistics at the field site. Na’ama Raz-Yaseef was responsible for supplying the continuous soil water data. Emanuela Negreanu, Ruth Ben-Meir, Hagai Sagi and Avraham Pelner provided much-appreciated technical assistance for various aspects of this study. Helpful comments on a previous version of the manuscript were provided by Heikki Hänninen and two anonymous reviewers. This multi-year project was funded by grants from ISF (695/99), BSF (2000293), EU (Carboeuroflux, EVK2-CT-1999-00032), GLOWA-JR (01-02-01752), IALC (04-DG-1132403-019) and the Minerva–Avron Photosynthesis Center. K.M. gratefully acknowledges the financial support of a Feinberg Graduate School PhD fellowship.