This study was conducted in Lambir Hills National Park in Sarawak, Malaysia (4°20′N, 113°50′E). The study area has a humid tropical climate with minimal seasonal changes in rainfall or temperature (Kato et al., 1995). The mean annual precipitation and air temperature are 2,600 mm and 25.8°C, respectively (Kume et al., 2011). Although this location lacks a regular dry season (Kume et al., 2011), it sometimes experiences reduced precipitation in association with supra-annual climatic events, such as ENSO (Nakagawa et al., 2000; Sakai et al., 2006; Gomyo and Kuraji, 2009).

We selected six dipterocarp species for this study; Dipterocarpus pachyphyllus Meijer (DP), Dryobalanops aromatica Gaertn.f (DrA), Shorea beccariana Burck (SB), Shorea curtisii Dyer ex King (SC), Shorea parvifolia Dyer (SP), and Shorea smithiana Symington (SS). All of these species are evergreen and dominate the canopy and emergent layers at the study site. DrA, SB, SC, and SS are among the most important tree species in Lambir Hills National Park in terms of appearance frequency or basal area (Davies et al., 2005), and DP and SP are locally abundant, especially on the lower slopes of the forest (Momose et al., 1997; Ashton, 2004). DrA, SB, and SC are mainly distributed in upper slope with sandy soil, while DP, SP, and SS are distributed in lower slope with clay soil (Itoh et al., 1995; Hirai et al., 1997; Lee et al., 2002; Ashton, 2004; Hiromi et al., 2012; Kenzo et al., 2016). However, the distribution of these species is unrelated to soil pH (Lee et al., 2002). All six species are used for enrichment planting in tropical rainforests in Southeast Asia (e.g., Appanah and Weinland, 1993; Vincent and Davies, 2003; Hattori et al., 2013; Tuck et al., 2016; Perumal et al., 2017; Widiyatno et al., 2020).

Seeds were collected from at least three mother trees per species in Lambir Hills National Park during the landscape-scale masting event in January 2014 (Iku et al., 2017). More than 300 seeds (50 per species) were collected, each of which was planted in a plastic pot (8 cm × 20 cm) filled with a 50:50 mixture of clayey soil (collected from the nearby Lambir Forest) and sand. The seedlings were grown in a nursery under two layers of 70% shade cloth, which reduced direct sunlight to about 5%, for 7 months. This is because dipterocarp species consume stored nutrients in their seeds for about 1 month after germination and then become autotrophic (Ichie et al., 2001).

From August to October 2014, watering was stopped after all pots were moved to a concrete-paved area under the eaves, where light conditions were comparable to those inside the nursery (Supplementary Appendix 1, 2). Each pot was randomly moved throughout the experiment so that all individuals experienced similar light conditions over time. Volumetric soil moisture was measured to assess soil drying using a soil moisture sensor (SM-200; Delta-T Devices, Cambridge, UK). We measured soil water content (SWC) five times within 1 day per pot, and the mean was taken as the value for the day. This procedure was implemented periodically for all pots. Soil water potential was measured simultaneously with SWC using a soil moisture sensor (Watermark 200SS-15; IRROMETER, Riverside, CA, USA) equipped with a temperature sensor (Watermark 200TS; IRROMETER) to determine the relationship between the SWC and soil water potential. Soil drying reached a minimum of −2.5 MPa (Supplementary Appendix 2A). The decrease in SWC after the experiment showed similar patterns among the six dipterocarp species studied (Supplementary Appendix 2B).

We measured the leaf gas exchange rate and water potential before irrigation was stopped, at 20, 15, 13, and 10% SWC, and finally every 2 days after the SWC fell below 10%. The leaf light-saturated photosynthetic rate (A_max), transpiration rate (E_max), and stomatal conductance (gs_max) at light saturation, and the dark respiration rate (R_d) in fully expanded and apparently non-senescent leaves were measured with a portable photosynthesis apparatus (LI-6400, Li-Cor, Lincoln, NE, USA). We randomly selected five to eight individuals per species and continued the measurements all day until the SWC of each reached the desired value. To avoid midday photosynthesis depression, we measured A_max, E_max, and gs_max from 08:00 to 11:30, and R_d from sunset to predawn (Kenzo et al., 2015). The light intensity was set to 1,500 μmol photon m^–2 s^–1 for the measurement of A_max, E_max, and gs_max and at 0 μmol photon m^–2 s^–1 for the measurement of R_d using an internal light-emitting diode (Li-640B; Li-Cor) as a light source. The CO₂ concentration in the chamber was maintained at 400 ppm, and the air temperature and relative humidity were approximately 30°C and 60%, respectively. The water-use efficiency (WUE), i.e., the ratio of A_max to gs_max, was also calculated (Kitahashi et al., 2008; Ichie et al., 2016).

We measured the maximum quantum yield of photosystem II (F_v/F_m) of each individual using a fluorescence meter (MINI-PAM; WALZ, Effeltrich, Germany) at predawn (around 05:00). To quantify F_v/F_m, dark-adapted leaves were exposed to a saturating pulse of 6,000 μmol m^–2 s^–1 with a wavelength of 650 nm, for 0.8 s under predawn conditions. In the leaves that maintained photosynthetic functions, the values of F_v/F_m show around 0.8, though less than 0.2 is the threshold corresponding to drought-induced mortality (Woo et al., 2008; García-Fernández et al., 2013). Three leaves were selected from each individual and the average value was taken.

Predawn leaf water potential (Ψ_pd), water potential at the turgor loss point (Ψ_tlp), and osmotic potential at full turgor (Ψ₁₀₀) were determined using a pressure chamber (Model 1002; PMS instruments, Corvallis, OR, USA). Ψ_pd was measured from 05:00 to 06:00 in five individuals per species selected randomly on the day when the SWC reached the specified values. Ψ_tlp and Ψ₁₀₀ were determined from pressure–volume curves (Tyree and Hammel, 1972). The base of each twig was cut off under water after the Ψ_pd measurement and rehydrated with tap water for at least 12 h under dark conditions (Tyree et al., 1974). The roughly 15-cm-long tip of each shoot bearing 5–10 leaves was cut from the twig for the measurement, and a pressure chamber was used to generate pressure–volume curves. We were unable to measure Ψ_pd, Ψ_tlp, and Ψ₁₀₀ at 25 and 20% SWC in SP and at 20% in SC due to equipment trouble.

Four or five individuals per species were selected randomly before the irrigation was stopped and the tree height, ground diameter, and leaf area of each were measured. During the experiment, the seedlings used for leaf water potential measurement were dug out of the pots, washed to carefully remove soil, and divided into root, stem, leaf, and bud organs. After the fresh weight of leaves was measured, the leaf area was determined using a flatbed scanner and image analysis software (Lia32 freeware; http://www.agr.nagoya-u.ac.jp/~shinkan/LIA32/index.html). We measured the dry weight of each organ sample after drying at 60°C for 48 h. The leaf water content during the experiment was calculated as follows: (Fresh weight–Dry weight)/Fresh weight × 100 (%).

Using the organ samples from the dry weight measurement, we determined the starch and total sugar contents of the leaves, stems, and roots in each species before the start of the experiment, and at 10 and 7% SWC. The total sugar content was determined following the phenol–sulfuric acid method of Dubois et al. (1956), as modified by Ashwell (1966). The dried samples were ground into a fine powder using a Hi-speed Vibration Sample Mill (TI-100; CMT, Fukushima, Japan), and about 100 mg of each sample was extracted at 90°C overnight using 80% ethanol. The fractional transmittance was read using an ultraviolet–visible light spectrophotometer (UV-1400; Shimadzu, Kyoto, Japan) at 490 nm. The starch content was determined following modified the methods of Lawrence et al. (1990) and Newell et al. (2002). The analytical procedure followed Ichie et al. (2005). The TNC content was calculated as the sum of the starch and total sugar contents.

To evaluate the response of each physiological trait to soil drought, we assessed the relationships between physiological parameters, including A_max, gs_max, E_max, F_v/F_m, and leaf water content, with SWC using a logistic regression model that estimated three parameters: the upper limit of each physiological trait under moist conditions before the experiment, the slope coefficient showing the rate of decline or rise, and the SWC at 50% of the upper limit. We also used a four-parameter logistic regression model to examine the relationships of R_d, WUE, and Ψ_pd with SWC. In addition to the parameters in the three-parameter model, this model estimated the lower limit of each physiological trait under dry conditions. Before conducting the analyses, A_max, gs_max, E_max, R_d, and WUE values were converted into values relative to those under well-irrigated conditions just before stopping irrigation to eliminate intraspecific error. Both analyses were carried out using the drm function in the drc package of R statistical software (version 4.2.1; R Development Core Team, Vienna, Austria) (Ritz and Streibig, 2005). A similar analysis was also performed at the individual level. Finally, we determined the SWC at which each physiological trait reached 90 and 50% of the upper limit, and the average was taken for interspecific comparison.

Linear regression was used to examine the relationships of Ψ₁₀₀ and Ψ_tlp with SWC. Analysis of variance was used to assess the interspecific differences of all physiological and morphological characteristics before the experiment, changes in sugar, starch, and TNC concentrations in each organ throughout the experiment, and differences in SWC among the six species in the SWC at which each physiological trait reached 90 and 50% of the upper limit. When the result was significant at P < 0.05, Tukey’s HSD test was used to compare the differences and changes. Principal component analysis (PCA) was conducted to estimate the physiological responses of the six species to soil drought using the prcomp function. R software (version 4.2.1) was used for all analyses (R Core Team, 2022).

The species studied differed widely in size immediately before the experiment (Supplementary Appendix 3). SB had much higher values for all parameters than the other species, except that SS was equivalent to SB in height and total leaf area; there were few differences among DP, SC, and SP.

The dry weight of each organ differed significantly among the species (Supplementary Appendix 3). The root dry weight was highest in SB, i.e., more than twice that of the other species. The species also differed considerably in their TR (shoot-to-root) ratio, with SS having a value significantly higher than the other five species.

Gas exchange rate parameters immediately before the experiment showed interspecific differences (Supplementary Appendix 4). A_max, E_max, and gs_max were significantly lower in DrA and SC than in the other four species. R_d was significantly lower in DP than SP. WUE was significantly higher in DrA than in the other five species.

During the experiment, interspecific differences were observed in A_max, E_max, and gs_max with decreasing SWC (Figures 1A–C and Table 1). SP showed a decrease in gs_max immediately after the irrigation was stopped, and A_max and E_max decreased accordingly. Meanwhile, in SB, gs_max was maintained for a short period after irrigation was stopped, and gs_max, A_max, and E_max decreased rapidly thereafter. In contrast, DP, DrA, SC, and SS responded to the decrease in SWC by maintaining gs_max values for a long period, but their stomata closed rapidly below 12% SWC; at 7–10% SWC, the gs_max, A_max and E_max values of these four species were half the pre-drought-values (Figures 1A–C and Table 1).

After irrigation was stopped, the relative R_d began to increase under relatively high SWC in SP and SC, but only began to increase at low SWC for the other four species (Figure 1D and Table 1). Among the six species, DrA showed the lowest R_d with decreasing SWC. There was no interspecific difference in SWC at 50% of the difference between the initial and maximum R_d values (Table 1).

All species showed a sharp decline in WUE at 7–11% SWC (Figure 1E and Table 1). DP and SS reached 50% of the initial value at a significantly lower SWC compared to SB and SC (Table 1).

In all six species, F_v/F_m was about 0.8 before irrigation was stopped, and was maintained at approximately the same value after initially stopping irrigation (Figure 2 and Table 1). However, F_v/F_m decreased sharply when SWC fell below 8%. A significant interspecific difference was observed in the SWC at which F_v/F_m reached 90% of the initial value; however, there was no interspecific difference in the SWC at which F_v/F_m reached 0.2, i.e., the threshold corresponding to drought-induced death.

Ψ_pd began to decrease at 12–16% SWC in all six species. Ψ_pd decreased more rapidly in SC than the other species, and showed the lowest values throughout the soil-drying period (Figure 1F). The Ψ_pd of DP and SS began to decrease at 15% SWC, and decreased slowly as the SWC dropped. Meanwhile, the Ψ_pd of SB and SP declined rapidly beginning around 12% SWC, but remained higher compared to the other four species. Finally, the Ψ_pd of DrA decreased slowly beginning around 12% SWC (Figure 1F). Overall, the values of Ψ₁₀₀ and Ψ_tlp decreased with decreasing SWC in DrA, DP, and SP, but remained relatively constant, or even increased, in SB, SC, and SS (Figure 3).

The leaf water content immediately before the experiment was significantly differed among the six species; DP and SP showed the highest values (Supplementary Appendix 3). The leaf water content was positively related to A_max (P < 0.05, r = 0.84), E_max (P < 0.05, r = 0.83), and gs_max (P < 0.05, r = 0.87), while other leaf traits, such as LMA and stomatal size, were not significantly related to gas exchange traits (data not shown).

After stopping irrigation, the leaf water content of all species remained stable for a while, but varied with increasing drought intensity according to species (Figure 1G). In DrA and SC, the leaf water content started to decrease at a relatively fast rate at 15% SWC, but decreased slowly at lower SWC. SP and SB showed a rapid decline in leaf water content beginning around 10% SWC and reached below 40% the fastest. Finally, DP and SS maintained steady leaf water content until 7% SWC, below which they decreased rapidly.

TNC contents in leaves, stems, and roots tended to increase in all six species after stopping irrigation (Figure 4). Total sugar and starch contents did not change, or tended to increase in all the species; they did not decrease for any species.

The PCA revealed different response patterns to soil drought among the six dipterocarp species studied. The first principal component (PC1) explained 48.0% of the variance; this dimension was related to the gas exchange response to drought (e.g., SWC at 50% of the initial gs_max, E_max, and A_max). The second principal component (PC2) explained 34.2% of the variance; this dimension was related to drought tolerance and death (e.g., SWC at 50% of the initial R_d and at a F_v/F_m of 0.2), and to variations in Ψ_tlp when SWC changed from 25 to 10%, (Figure 5 and Supplementary Appendix 5).

SP and SB exhibited an isohydric response, as illustrated by a rapid decline in gs_max and E_max, whereas DP, DrA, SS, and SC exhibited an anisohydric response, as evidenced by the late decline in PC1. Interestingly, DP, DrA, and SP showed increased drought resistance during the experiment, whereas SB, SC, and SS showed lower drought resistance in PC2. No significant correlations were found between PC1 or PC2 and the initial morphological and physiological characteristics of the six species (Supplementary Appendix 6).