Change in hydraulic properties and leaf traits in a tall rainforest tree species subjected to long-term throughfall exclusion in the perhumid tropics

Abstract. A large-scale replicated throughfall exclusion experiment was conducted in a pre-montane perhumid rainforest in Sulawesi (Indonesia) exposing the trees for two years to pronounced soil desiccation. The lack of regularly occurring dry periods and shallow rooting patterns distinguish this experiment from similar experiments conducted in the Amazonian rainforest. We tested the hypotheses that a tree's sun canopy is more affected by soil drought than its shade crown, making tall trees particularly vulnerable even under a perhumid climate, and that extended drought periods stimulate an acclimation in the hydraulic system of the sun canopy. In the abundant and tall tree species Castanopsis acuminatissima (Fagaceae), we compared 31 morphological, anatomical, hydraulic and chemical variables of leaves, branches and the stem together with stem diameter growth between drought and control plots. There was no evidence of canopy dieback. However, the drought treatment led to a 30 % reduction in sapwood-specific hydraulic conductivity of sun canopy branches, possibly caused by the formation of smaller vessels and/or vessel filling by tyloses. Drought caused an increase in leaf size, but a decrease in leaf number, and a reduction in foliar calcium content. The δ13C and δ18O signatures of sun canopy leaves gave no indication of a permanent down-regulation of stomatal conductance during the drought, indicating that pre-senescent leaf shedding may have improved the water status of the remaining leaves. Annual stem diameter growth decreased during the drought, while the density of wood in the recently produced xylem increased in both the stem and sun canopy branches (marginally significant). The sun canopy showed a more pronounced drought response than the shade crown indicating that tall trees with a large sun canopy are more vulnerable to drought stress. We conclude that the extended drought prompted a number of medium- to long-term responses in the leaves, branches and the trunk, which may have reduced drought susceptibility. However, unlike a natural drought, our drought simulation experiment was carried out under conditions of high humidity, which may have dampened drought induced damages.

perhumid climate where the trees should be less experienced in coping with drought because rainless periods occur only irregularly (Aldrian and Susanto, 2003;Aldrian et al., 2004;Erasmi et al., 2009). Experiments on the drought response of perhumid tropical forests with continuously high soil moistures and air humidity do not yet exist. To fill this gap, a replicated throughfall displacement experiment (Sulawesi Throughfall 15 Displacement Experiment, STDE) was carried out in a premontane perhumid rainforest in Central Sulawesi, Indonesia, to investigate the response of the trees and soil biological activity to a 24-months drought period. The study region is characterized by high amounts of rainfall throughout the year and air humidity at the canopy height that rarely drops below 80%. 20 Both observational studies on natural drought events and the Amazonian throughfall displacement experiments showed that under prolonged drought especially large and tall canopy trees (and species) as well as lianas experienced higher mortalities than trees of smaller size (Slik et al., 2004;Van Nieuwstadt and Sheil, 2005;Nepstad et al., 2007;da Costa et al., 2010;Phillips et al., 2010). Drought may harm trees through 25 two pathways, exposure to increase of xylem embolism and reduced assimilate supply due to stomatal closure. Besides cell dehydration and a consequently reduced leaf expansion growth, carbon starvation could be one consequence of severe drought (Farooq et al., 2009), but this hypothesis has been questioned (Sala, 2009). explain drought damage to trees. We used tree climbing equipment in each of the seven mature C. acuminatissima in the roof and control plots in order to study the response of sun-lit upper canopy leaves and branches. Because leaf exposure and canopy position is known to exert a great influence on leaf morphology and physiology in trees, we investigated leaves and branches of both the sun and shade canopy and 10 compared their response to the two-year desiccation. We further hypothesized that sun canopy leaves and branches are more susceptible to desiccation than are shade canopy organs. 15 The Sulawesi Throughfall Displacement Experiment (STDE) was established in 2006 in a premontane rainforest in the Pono Valley on the western boundary of Lore Lindu National Park in Central Sulawesi, Indonesia (01 • 29.6 S 120 • 03.4 E elevation 1050 m).

Site description
The climate of the study area is perhumid with a mean annual precipitation of 2901 mm, a mean annual temperature of 20.6 • C and a mean relative air humidity of 88.7% (data 20 derived from measurements in 2008). The heavily weathered soils of this old-growth forest developed on metamorphic rocks. The clayey-loamy soil texture with dominant kaolinite and hematite has been classified as Acrisol (World Reference Base for Soil Resources, Leitner, 2010

Experimental design
The STDE consisted of six floristically and structurally similar plots of 0.16 ha (40 m × 40 m) that were spread in a stratified random design over an area of approximately five ha. While three plots served as control, the remaining three plots were covered by sub-canopy roofs to displace a large fraction of the rainfall. The roofs were 5 constructed with a large number of removable transparent plastic-lined bamboo-frames placed on a wooden gutter construction to collect the throughfall water. The desiccation period started in May 2007. At the beginning, approximately 70% of the plot area was covered by the bamboo frames. In early 2008, the roof closure was further increased to approximately 90% by building custom-sized panels to close gaps around the tree 10 stems and odd-sized openings. To avoid lateral soil water movement or infiltration of surface runoff into the plots and to disable trees to take up water from the surroundings of the study plots, all plots were trenched along the perimeter to 0.4 m soil depth and lined with plastic foil. Since 74.3% of the fine root and 91.1% of the coarse root biomass are located in the upper 20 cm of the soil profile (Hertel et al., 2009), we assumed this 15 trenching depth to be sufficient to effectively prevent root water uptake from beyond the plot edges. The litter, which had accumulated on top of the roof construction or in the runoff channels, was transferred back to the soil surface.

Microclimatic and hydrologic measurements
Above-canopy global radiation was measured with a pyranometer (CS 300, bell Scientific, UK). Air temperature and relative humidity were recorded with a combined temperature and humidity probe (CS 215,Campbell Scientific,UK). Rainfall was measured to the nearest 0.1 mm with a tipping bucket rain gauge (ARG100, Campbell Scientific, UK). All sensors were mounted on a 16 m tall tower located in a natural forest gap approximately 50 m away from the first study plot. Data were collected every Introduction (Campbell Scientific, UK). Atmospheric vapor pressure deficit was calculated from air temperature and relative humidity according to Goff and Gratch (1946). Volumetric soil water content was continuously measured in one main and two additional soil pits per plot. In the main soil pit, time domain reflectometry (TDR) probes (CS616, Campbell Scientific Inc., Logan, UT, USA) were installed at 10, 20, 40, 75, 150 5 and 250 cm soil depth. The two additional soil pits were equipped with TDR probes in 10, 40, and 75 cm soil depth. The probes were inserted horizontally in the undisturbed soil at the end of a 30 cm long hole dug into the soil pit wall. In total, 36 TDR probes were used per treatment and the data was logged hourly (CR1000, Campbell Scientific Inc., Logan, UT, USA). The TDR probes were calibrated for four soil depths following 10 the procedure described by Veldkamp and O'Brian (2000).
The hydrological and physiological measurements began on 27 March 2007 in the roof plots and on 31 May 2007 in the control plots. For the delayed onset of the measurements in the control, a lightning strike in March 2007 was responsible that damaged both dataloggers and TDR probes.

Relative extractable water and soil water potential
The soil moisture measurements were used to calculate the relative extractable water (REW) in the soil using the following equation (Breda et al., 2006;Granier et al., 2007): where θ t is the fractional volumetric soil water content on the respective day, θ min 20 the soil water content at which all plant-available water is extracted (corresponding to the water content at wilting point in a given soil depth), and θ max the maximum water content measured during the study in a given soil depth. The volumetric soil water content at the permanent wilting point was calculated from laboratory derived soil water retention curves (van Genuchten, 1980)  Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | measured soil water content) and 0 (permanent wilting point). According to Granier et al. (2007) temperate trees typically experience drought stress when REW drops below a threshold of 0.4. The soil water retention curves were further used to calculate soil matric potentials from the soil moisture measurements for all soil depths in the two treatments.

Tree selection and plant material
In June 2009, after 24 months of experimental drought, 14 tree individuals (7 in the roof, and 7 in the control plots) of the most abundant and tallest upper-canopy tree species of the studied forest, C. acuminatissima, were chosen to collect branch, twig and leaf material from the upper sun-lit canopy and from the lowermost deeply shaded 10 part of the crown using tree climbing equipment. These samples were used to obtain data on branch wood specific gravity (wood density), twig hydraulic properties, branch wood anatomy, leaf morphology and foliar nutrient status. In addition, wood cores were taken from the trunk of every tree at 1.5 m height to determine the wood density and wood anatomy of the trunk.

Wood density and saturated water content determination
Wood density (ρ) was determined for trunk wood cores with a diameter of 5 mm and a mean core length of 69.5 ± 7.2 mm using an increment corer (Haglöf, Långsele, Sweden) and for branch wood samples from the upper and lower crown with a mean segment diameter of 33.1 ± 7.8 mm and a mean segment length of 120.0 ± 15.3 mm. In 20 total, 109 branch segments were harvested in the 14 C. acuminatissima trees. The fresh volume of the wood cores was calculated from the diameter of the increment corer and the length of the core sample after removing bark and phloem; the fresh volume of the branch samples was determined gravimetrically by water replacement. After volume determination, all samples were oven-dried at 105 • C for at least four Introduction at a precision of 10 mg. The dry mass of each sample was then divided by its volume to obtain ρ. The trunk wood cores were further used to determine the saturated water content of the wood (SWC). The cores were submerged in deionized water and allowed to equilibrate overnight. Afterwards, the cores were lightly blotted with tissue paper and weighed at a precision of 0.1 mg.

5
For comparison, we also used a non-destructive Pilodyn wood tester (Pilodyn 6J, Proceq, Switzerland) prior to trunk wood core extraction in the same trunk area to obtain a second independent measure of wood density. A circle-shaped area with a diameter of 5 cm had to be removed to apply the Pilodyn tester (Hansen, 2009). 10 Annual stem diameter increment was measured with increment measurement tapes (UMS, Munich, Germany) that were installed in December 2006 on 16 tree individuals of C. acuminatissima (7 trees in the control plots: DBH 23-150 cm and 9 in the roof plots: DBH 22-91 cm). Stem diameter increment was documented monthly until the end of the desiccation period in May 2009. The relative annual stem increment (incre-15 ment as a fraction of basal area) for the trees was calculated separately for the first and the second year of the desiccation experiment.

Experimental determination of axial hydraulic conductivity
The technique introduced by Sperry et al. (1988) was applied to measure axial hydraulic conductivity in twig segments. In total, 116 measurements were analyzed 20 (control n = 56, roof n = 60). For each tree individual, eight twig segments of 139.7 ± 34.7 mm length and 10.6 ± 1.5 mm in diameter were harvested, four from the upper canopy and four from the lower crown. These segments were immediately transferred to polyethylene tubes filled with water containing a sodium-silver-chloride complex (16 µg l −1 Ag, 8 mg l −1 NaCl, Micropur katadyn, Wallisellen, Switzerland) to pre- laboratory. Additionally, all leaves distal to the twig segment were harvested. In the laboratory, each twig segment was recut under water with a razor blade and mounted on the tubing system. We used distilled water containing the same sodium-silver-chloride complex as described above for the conductivity measurements. Before entering the twig segment, the solution was forced through a 0.2 µm membrane filter (Maxi Capsule, Pall, USA). The segments were flushed at a pressure of 0.12 MPa to achieve maximum axial hydraulic conductivity (k h ). Subsequently, length and mean diameter of the segments were determined and the samples stored in 70% ethanol for further anatomical analyses. Hydraulic conductivity (k h , kg m MPa −1 s −1 ) was calculated as: where J v is the flow rate through the branch segment (kg s −1 ) and ∆P/∆X is the pressure gradient across the segment (MPa m −1 ). k h was used to calculate vessel lumen area-specific (k s ) and leaf area-specific conductivity (LSC, kg m −1 MPa −1 s −1 ) by dividing the maximum conductivity by the microscopically determined lumen area of the vessels (m 2 ) or the supported leaf area (m 2 ) of the twig segments.

Xylem anatomy, vessel size distribution and theoretical hydraulic conductivity
Anatomical measurements were conducted in all harvested twig segments and trunk cores. We used a stereo-microscope (SteREO V20, Carl Zeiss MicroImaging GmbH, Germany) to obtain high quality top-view images of the cross-sectional cuts of the twigs 20 and trunk cores. Before analysis, the twig segments and trunk cores were dyed with safranin and treated with chalk. The base of every twig segment was photographed to calculate the size of the xylem (sapwood area, A Xylem , m 2 ). In the trunk core samples, only the outer-most centimetre of the core was analyzed. On average, we analyzed by this procedure an area of 45.3 ± 4.8 mm 2 . The images were analyzed with the soft- 25 ware ImageJ (v1.42q, http://rsb.info.nih.gov/ij) using the particle analysis-function to 8562 Introduction

Tables Figures
Back Close

Full Screen / Esc
Printer-friendly Version

Interactive Discussion
Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | estimate idealized radii (r) from lumen area (A = πr 2 ), vessel density (VD, n mm −2 ) and cumulative vessel lumen area (m 2 ). Additionally, we calculated the hydraulicallyweighted mean vessel diameter, subsequently referred to as hydraulic mean diameter (d h ) using the expression Σd 5 i /Σd 4 i after Sperry et al. (1994). By applying this transformation, every vessel is weighted according to its contribution to total hydraulic conduc- . 15 All leaves distal to the analysed twig segments were stripped off and analyzed for their leaf area (WinFOLIA 2005b, Regent Instruments Inc.). On average, 105.7 ± 59.6 leaves per twig segment (control n = 56, roof n = 60) were scanned to obtain the total leaf area per twig and the mean leaf size (A L , cm 2 ). Afterwards, the whole leaf material was oven-dried for at least 72 h at 70 • C and subsequently weighed at a precision of 20 10 mg to relate dry mass to the total leaf area to obtain the specific leaf area (SLA, cm 2 g −1 ). The Huber value (HV) was calculated as the ratio of sapwood cross-sectional area to the dependent leaf area or number of leaves distal to the measured twig segment. The specific leaf number stands for the number of leaves supported per twig which was calculated by dividing the total number of leaves distal to the twig segment by the twig cross-sectional area (n spec L , n mm −2 ). In the leaf dry mass, the concentrations of C, N, P and Ca, K, Mg, Fe and Mn 5 were analyzed and expressed on a mass and leaf area basis (control n = 56, roof n = 63). The foliar signatures of δ 13 C and δ 15 N were determined with a Delta plus isotope mass spectrometer (Finnigan MAT, Bremen, Germany), a Conflo III interface (Thermo Electron Coorperation, Bremen, Germany) and an NA2500 element analyzer (CE-Instruments, Rodano, Milano, Italy) using standard δ notion:

Leaf morphology and nutrients
The concentrations of P, Ca, Fe, K, Mg and Mn were determined with an ICP spectrometer Optima 5300 DV (PerkinElmer Inc., USA).

Statistical analyses
All data sets were tested for Gaussian distribution with a Shapiro-Wilk test. Com- 15 parisons of normally-distributed parameters were made with three-way general linear models (GLM). In case of non-Gaussian distribution, the datasets were tested with the non-parametric Mann-Whitney U test for pair-wise comparison of means. Significance was assumed at p ≤ 0.05 in all cases. These calculations were conducted with the SAS System for Windows 9.1 (SAS Institute, Cary, NC, USA). Linear regressions were 20 calculated with the program Xact 8.03 (SciLab, Hamburg, Germany). When comparing upper and lower canopy of the trees in a given treatment, the analyses are labeled with "canopy position", when comparing either upper canopies or lower canopies, between the roof and control plots, the label "treatment" is used.  (Table 1). 10 Due to installation delays for the TDR probes caused by a lightning strike, no pretreatment comparison between control and roof plots could be established. However, soil moisture content (volumetric soil water content, θ) in the roof plots at 0.1 m depth before roof closure was similar to later measured values from the control plots, indicating no differences between the treatments. 15 The drying of the soil proceeded in two steps that reflected the roof closure by 70% (  plots, θ was on average by 30% lower in the upper soil and by 15% lower in the lower profile in the period June 2008-May 2009. According to the soil water retention curves established in the laboratory, the calculated soil matric potential (Ψ soil ) decreased in the topsoil (0.1 m) up to −3 MPa during the driest phase from March 2009 until roof opening. As an average for the three inves-5 tigated upper soil layers (0.1, 0.2 and 0.4 m), Ψ soil dropped to −1.5 MPa at the end of the desiccation (Fig. 1). In contrast, no significant differences in Ψ soil were detected in the lower soil layers (0.5-3.0 m) between roof and control plots, even though θ differed by about 15%.

Soil moisture status during the desiccation
The calculated relative extractable water for the upper soil layers (REW top ) dropped below the threshold value of 0.4 in the roof plots immediately after the beginning of the second phase of the desiccation in early 2008, enhanced by rather low rainfall in this period (Fig. 1). Interrupted by a short recovery due to strong rainfall in March/April 2008, REW top decreased further in the roof plots, leading to a 90% smaller amount of available water in the roof plots compared to the control. On the other hand, the relative 15 extractable water of the lower soil layers (REW low ) only dropped below the threshold of 0.4 in the driest phase of the experiment from February 2009 onwards. Nevertheless, REW low was by 50% smaller in the roof than in the control plots from June 2008 until May 2009. 20 Branches of C. acuminatissima harvested after the two-year desiccation period had significantly lower axial hydraulic conductivities in the xylem than samples from the control trees (Fig. 4). This was found for leaf-specific hydraulic conductivity (LSC) and for axial conductivity normalized by vessel lumen area (k s ) and was valid for both branches of the sun and shade canopies. While the ratio branch sapwood area: dependent leaf 25 area (Huber value) did not alter, we recorded a significantly reduced number of leaves of about 30% that depend on a unit branch sapwood area (leaf number-based Huber value, Fig. 2, Table 4). The droughted trees also showed a higher wood density in 8566 Introduction

Tables Figures
Back Close

Full Screen / Esc
Printer-friendly Version

Interactive Discussion
Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | the branches of the sun canopy, but not of the shade canopy, which is consistent with the reduced hydraulic conductivity. Probably linked to the reduced leaf number per branch sapwood, we found a significant increase in mean leaf size after two years of desiccation in the roof plots both in the sun and shade canopy branches. Analysis of isotope composition in leaf dry mass indicated no drought effects on δ 13 C 5 and δ 15 N (Fig. 3). The δ 13 C-N A relationship was not different between roof and control plots (Fig. 5). Leaves harvested at the end of the experiment contained in the roof plots significantly less Ca, Mg and Fe (sun and shade canopy) and N (shade canopy) per dry mass than the control (Tables 3 and 4).

Stem increment, wood anatomy and wood density as affected by
10 the desiccation In the roof plots, the annual stem diameter increment was by 10% lower in the first year of the experiment, and by 26% in the second year than in the control plots in the C. acuminatissima trees. However, the differences were not significant (p = 0.12 for the second year). Drought-induced alterations in the outermost xylem of the trunk are 15 further documented by a significant decrease in mean vessel diameter (p = 0.03), an increase in the wood density of the peripheral xylem sections (by 5%, p = 0.09) and an associated significant reduction in saturated water content (SWC, p = 0.05) of the xylem by 10% when comparing roof and control plots (Table 2).
3.5 The factor "canopy position": differences between sun and shade canopy 20 The large majority of leaf morphological, chemical and branch hydraulic traits differed significantly between sun-lit upper canopy and lower shade canopy of C. acuminatissima. Sun leaves were smaller with a lower SLA, had a less negative δ 13 C and a more positive δ 15 Tables 3 and 4). Shade leaves with lower N per leaf area discriminated stronger against δ 13 C (more negative δ 13 C) than sun leaves; the δ 13 C-N A relationship was not different between drought-exposed and control trees (Fig. 5). Sun canopy branches had a much higher sapwood area : leaf area ratio (Huber value), and thus leaf-specific conductivities (LSC) than shade canopy branches. Vessel density and wood density in 5 the branches differed between sun and shade canopy only in the drought-exposed roof plots, while lumen area-specific conductivity was the same (Table 4).

Desiccation effects on leaf traits and twig hydraulic properties
After 24 months of throughfall reduction, the topsoil layers of the Pono forest were 10 strongly desiccated, exceeding conventional thresholds of critical soil water availability for plant growth (Ψ soil < −1.5 MPa, REW < 0.4). In accordance with other root system studies in perhumid environments (Schenk and Jackson, 2002;Hertel et al., 2003;Jimenez et al., 2009) the trees of the Pono forest most likely did not develop deepreaching roots that could tap water reserves in deeper soil layers. In support of this as- 15 sumption, Hertel et al. (2009) observed that 74.3% of the fine root biomass (Ø ≤ 2 mm) in the soil of the study plots was located within the top 20 cm and only 4.4% reached 40-60 cm soil depth. The coarse roots (Ø > 2 mm) showed a similar depth distribution (91.1% in the top 20 cm, 1.2% in 40-60 cm soil depth) with an exponential biomass depth distribution decrease with depth and only extremely small fine root densities 20 (0.12 g L −1 ) at 100-300 cm depth. Our data on the depth distribution of tree roots strongly indicate that deep-reaching roots are far less important in this tropical perhumid forest than in the Amazonian forest with short dry periods studied by Nepstad et al. (2002). More than three months of exposure to water availabilities in the topsoil below the conventional "wilting point" of crop plants must have represented drought Anatomical investigations of the conducting system and hydraulic measurements in a rather large number of sun-canopy and shade branches of C. acuminatissima showed that the terminal twigs, which must have been grown during the two-year experimental period, had a significantly reduced axial conductivity in their xylem when expressed per vessel lumen cross-sectional area (k s ; 25% reduction) or leaf area distal to the 5 measuring point (LSC; 10-33% reduction). The reason may be a smaller mean vessel diameter or decreased vessel densities in the twig xylem, and hence a higher wood density in the branches of the desiccation treatment. Several authors have reported that trees adjust the shape of their vessels when exposed to drought (e.g. Sass and Eckstein, 1995;Eilmann et al., 2006), reflecting plant water status at the time of cell 10 differentiation (García- Gonzáles and Eckstein, 2003).
Remarkably, we found an impaired hydraulic performance of the terminal twigs. The comparative investigation of about 60 twigs each in the roof and control plots produced evidence that drought may also have affected processes of leaf bud initiation and leaf expansion because our data show a significant reduction in the number of leaves per 15 twig sapwood area (lower Huber-value normalized to leaf number) in the sun crown, and an increase in mean leaf size by 30-40% in twigs in the sun and shade crown of the desiccation plots at the end of the treatment. Since we found no scars of abscised leaves on the investigated twigs, we assume that twigs grown during the desiccation treatment have formed a smaller number of new leaf buds, thereby reducing the leaf 20 area to be supplied with water, thus improving the water status of the remaining leaf buds and allowing them to enfold larger leaves. A similar effect has been observed in saplings of silver birch (Betula pendula) that produced fewer but larger leaves under drought (Aspelmeier and Leuschner, 2006). The same was found along a precipitation gradient in Central Germany for beech (Fagus sylvatica, Meier and Leuschner, 2008). 25 This reasoning could also explain why we did not find a decrease in δ 13 C signatures in the drought-exposed leaves, as would be expected when leaf conductance and leaf expansion growth were reduced during periods of water shortage (Lambers et al., 1998;Lösch, 2001). However, an alternative strategy is to reduce the number of leaves in BGD Introduction

Tables Figures
Back Close

Full Screen / Esc
Printer-friendly Version

Interactive Discussion
Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | order to maintain, or even improve, the water status of the remaining leaves, which may have happened in C. acuminatissima in our experiment. Foliar nutrient analyses did not provoke evidence for the hypothesis that soil desiccation considerably influences the metabolism of trees through nutrient, mainly N, shortage (e.g. Gessler et al., 2004;Beier et al., 2008;Fotelli et al., 2009;Kreuzwieser 5 and Gessler, 2010). Neither foliar N nor P were significantly altered after 24 months of desiccation treatment. However, significant smaller leaf Ca contents per dry mass and also per leaf area may indicate either reduced transpiration rates or smaller Ca concentrations in the soil solution of the roof plots, because the element mostly is transported passively with the mass-flow of water in soil and xylem (Gollan et al., 1992;McDonald and Davis, 1996). We speculate that the droughted trees extracted water from deeper soil layers where the concentrations of Ca and other nutrients were lower.
Reduced stomatal conductance in periods of high atmospheric saturation deficits and low soil moisture often have been found to result in less discrimination against δ 13 C in the course of CO 2 assimilation (e.g. Saurer et al., 1997;Handley et al., 1999;15 Jäggi et al., 2003;Sala and Hoch, 2009;Fichtler et al., 2010), while the δ 15 N signature of leaves typically shows no strong drought signal (Peuke et al., 2006;Hartman and Danin, 2010). Thus, the lack of differences in δ 15 N between roof and control trees fits to the expectation, while the absence of δ 13 C differences comes as a surprise. It appears that leaf conductance was not significantly reduced in response to the 24-month 20 desiccation. Further, the foliar nitrogen-leaf area relationship was similar in roof and control trees (see Fig. 5) which suggests that stomatal and biochemical photosynthesis were not larger in the drought-exposed trees than in the control. The δ 13 C signal is viewed as support of the assumption, that soil desiccation led to a reduction in leaf numbers, while the water status of the remaining leaves was not deteriorated.

Desiccation effects on the hydraulic system of the stem
In concert with the alterations observed in the xylem of sun canopy and also shade canopy twigs, we detected reductions in mean vessel diameter and axial conductivity 8570 Introduction in the outermost xylem of the trunks of the roof plots which also showed up in a higher wood density and reduced saturated water content of this recently developed section of the xylem. These anatomical responses may also partly explain the 26% reduction in stem diameter growth observed in the drought-exposed C. acuminatissima trees during the second year of the experiment, even though stem shrinking most likely has 5 also contributed to the relatively small diameter increase in these trees. It has to be mentioned that the difference in stem increment between roof and control plots despite its absolute size was not significant (p = 0.12 for the second year) which was mainly caused by the relatively small number of large C. acuminatissima trees (9 or 7) that grew inside the roof and control plots.
In contrast to our expectations of a sensible drought-response of the tree of this perhumid forest, we found no signs of major damage in the adult trees after 24-months desiccation, while most of the tree saplings and herb layer plants had already died. Rather, our data indicate adaptive responses in the hydraulic system and canopy leaf area of the tall C. acuminatissima trees that are suited to lower the risk of cavitation 15 and reduce canopy transpiration. One might conclude that C. acuminatissima is not as drought-sensitive as we assumed due to its frequent occurrence in this perhumid forest with only exceptionally occurring droughts. However, several facts make it likely that this conclusion is premature. First, other desiccation experiments in forests showed that severe damage to the trees may occur only after two or more years of soil desic- Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | simulated in our experiment where relative humidity did not drop below 88% in the experimental period (see Table 1) despite the long-lasting and marked soil desiccation. There is also the possibility that the trees were profiting from the high rainfall in the area by foliar uptake of water, thereby mitigating the effects of soil water shortage. For example, some studies obtained evidence for water absorption through the leaf cuticle of 5 tropical trees (Yates and Hutley, 1995;Diaz and Granadillo, 2005;Oliveira et al., 2005). Thus, it is likely that soil desiccation in a natural dry spell will have a much stronger effect on the studied tree individuals of C. acuminatissima than it was simulated in our experiment.

4.3
Is the sun canopy more drought sensitive than the shade canopy? 10 Our second hypothesis postulated that the sun canopy of tall trees is more susceptible to drought than lower crown parts in the deep shade where air humidity is higher. Except for wood density in the twigs, all hydraulic and leaf parameters that showed responses to the desiccation treatment for sun canopy twigs, reacted in a similar way in shade twigs as well. Indeed, the decrease in twig axial hydraulic conductivity, in the 15 number of leaves per sapwood area (modified Huber value) and the increase in leaf size upon drought were observed in the shade canopy in a similar manner as in the tree top. Moreover, the reduction in LSC was even greater than in the sun canopy. Thus, a stress-mitigating effect of the humid forest interior did not occur; the physiological consequences of soil desiccation seem to develop in tall C. acuminatissima trees rather 20 independently of height in the tree and the specific microclimate. This is astonishing given the large differences in leaf and hydraulic traits between the sun and shade canopy. Shade leaves of C. acuminatissima were on average 60-70% larger and had 20-30% higher SLA than sun leaves, while leaf-specific conductivity in the twig xylem and the Huber value were about 40-60% smaller in the shade canopy 25 due to lower evaporative demand. On the other hand, lumen area-specific conductivity was about 10% higher in the xylem of the shade branches than in the sun canopy (differences not significant) which is associated with a smaller wood density in the twig 8572 Introduction xylem. Thus, despite a greater exposure to atmospheric drought, sun leaves and twigs did not differ from shade leaves and twigs in their response to soil desiccation.

Conclusions
The Sulawesi Throughfall Displacement Experiment is the first experimental study about the effects of an extended soil desiccation period on the trees of a perhumid 5 tropical rainforest where natural droughts occur only exceptionally. The very shallow depth distribution patterns of fine and coarse roots are interpreted as resulting from the continuously high rainfall and permanently low atmosphere saturation deficit; these hydrologic characteristics allows to contrast the Sulawesi experiment with the two throughfall displacement experiments in Eastern Amazonia where regular dry peri-10 ods occur and certain trees may have deep-reaching roots (e.g. Markewitz et al., 2010 and references therein). While no signs of canopy dieback or other critical damage were observed in the tall Castanopsis acuminatissima trees or the other trees in the stand, the long and severe desiccation of the upper soil caused marked reductions in the hydraulic conductivity of the xylem of the trunk and of the terminal twigs, a reduction in leaf number per conducting sapwood in the twigs (but no reduction in leaf size), and a tendency for reduced stem diameter growth. We conclude that the tall C. acuminatissima trees in this perhumid forest were -in contrast to our second hypothesisnot more drought-susceptible in the upper sun canopy than in the shade crown. Neither the C. acuminatissima trees nor other smaller tree species showed signs of critical 20 damage which reflects our first hypothesis. We assume that the constantly high air humidity in this environment, which was not reduced by the throughfall displacement, plays an important role for the vigor of these trees and may have buffered against critical drought-induced damages as they were expected from the soil water status data.

Tables Figures
Back Close

Full Screen / Esc
Printer-friendly Version

Interactive Discussion
Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | of partial throughfall exclusion on canopy processes, aboveground production, and biogeochemistry of an Amazon forest, J. Geophys. Res.-Atmos., 107, 8085, 2002. Nepstad, D. C., Tohver, I. M., Ray, D., Moutinho, P., and Cardinot, G   Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Table 3. Foliar contents of C, N, P and cations per mass and per leaf area in leaves that were harvested distal to the twig segments used for hydraulic measurements in the control and roof plots for upper (sun) and lower (shadow) crown. The given unit is g kg −1 or g m −2 . Lower-case letters indicate significant differences between the two crown positions, and upper-case letters between the two treatments. All values are means ± 1 SE. The number of replicates for the control are n = 56, for the roof n = 63. Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Table 4. P values for the comparison between means of control and roof plots or sun and shade crown of 27 parameters measured in C. acuminatissima trees. The first column gives the ratio of the means (for canopy position: shade over sun canopy; for the treatment: roof over control treatment), the second column indicates the significance of the difference (parametric or non-parametric traits). Level of significance are presented as p ≥ 0.05 = * , p > 0.01 = * * and p > 0.001 = * * * . Not significant relations = n.s.   ) of C. acuminatissima trees in the control and roof plots. Lower-case letters indicate significant differences between the crown positions of a given treatment, and upper-case letters stand for significant differences between the two treatments. Values are means ± SE. Number of replicates for LSC, k s and HV were: control n = 56, roof n = 60; for ρ branch : control n = 52, roof n = 57.  Table 3.