Water-use characteristics of Syzygium antisepticum and Adinandra integerrima in a secondary forest of Khao Yai National Park in Thailand with implications for environmental management

Study site and measurements of the environmental variables

The study was conducted in Khao Yai National Park, Thailand (14°26′31″N, 101°22′55″E). Khao Yai National Park covers an area of about 200 km² in Nakhon Ratchasima, Saraburi, Prachinburi and Nakhon Nayok Provinces in Thailand. This region is dominated by monsoon climate, where the dry season usually lasts from November to April and from May to October for the wet season (Brockelman, Nathalang & Maxwell, 2017). Based on recorded data between 1994–2018, the overall mean annual temperature was 22.4 °C. The mean annual rainfall was 2,100 mm. Khao Yai National Park is characterized by different stages of forest succession comprising primary forests and various stages of secondary forests. In this study, we performed the study in a secondary forest representing a young forest in Nakhon Nayok Province. The study site has an area of 2 ha and an age of approximately 10 years (Chanthorn, Hartig & Brockelman, 2017). Its mean canopy height is 15 m and its tree density of 1,226 trees ha⁻¹. The soil is gray-brown ultisol which was degraded by shifting agriculture by burning before regeneration (Chanthorn et al., 2016; Chanthorn, Hartig & Brockelman, 2017). The bulk density was 1.24 g cm⁻³ and soil texture was sandy clay-loam with the sand contents of 64.4% and 56.4% as measured in September 2020 and February 2021, respectively (Rodtassana et al., 2021). In 2020, a 20 m tall tower was constructed for installing weather sensors above the forest canopy in the plot. Environmental conditions that influence T including atmospheric humidity, solar radiation, and soil moisture have been continuously monitored since then. Air temperature (T, °C), relative humidity (RH, %), and photosynthetically active radiation (PAR, µmol m⁻² s⁻¹) were measured by a temperature and relative humidity probe (EE181-PT; Campbell Scientific, Logan, UT, USA) and a quantum sensor (LI190R-PT, Campbell Scientific), respectively. Soil moisture sensors (Water content reflectometer, CS616-PT-U; Campbell Scientific) were installed to monitor volumetric soil moisture at 5, 10, 15, and 30 cm depth because tree roots may access water from multiple depths in the soil (Wang et al., 2019). We randomized the points to install soil moisture sensors around the tower. Two soil moisture sensors were installed at each depth of 5, 10, and 15 cm. However, soil moisture at 30 cm depth was monitored by one soil moisture sensor because soil moisture in subsoil was less sensitive to changing environmental conditions than topsoil (Rong et al., 2017). Rainfall (mm) was measured by tipping rain gauge bucket (TE525MM-PT; Campbell Scientific). All sensors were connected to a datalogger (CR1000 series; Campbell Scientific, Logan, UT, USA) which recorded data every 30 min. Air temperature and relative humidity are used to calculate vapor pressure deficit (VPD, kPa), which is the difference between actual vapor pressure and saturated vapor pressure (SVP), from the following equations (Monteith & Unsworth, 1990).

(1)${\displaystyle \mathrm{SV\; P}=610.7\times 10^{\frac{7.5\mathrm{T}}{237.5+\mathrm{T}}}}$ (2)${\displaystyle \mathrm{V\; PD}=\left(1-\frac{\mathrm{RH}}{100}\right)\times \mathrm{SV\; P}.}$

Because we did not have any information regarding rooting depth, which determines the depth of soil moisture data to be used in the analysis, we used the average of soil moisture data from all soil water probes, covering soil depth up to 30 cm, as the soil moisture data (θ, m³m⁻³) for further analysis. Based on previous studies in the central Amazon which reported the most fine root distribution within 20 cm soil depth (Noguchi et al., 2014), we assumed that the average soil moisture across 30 cm depth represents the soil water that largely influences tree water use. To facilitate the cross-site comparison with other or future studies, relative extractable water (REW) was used in the analysis and was calculated according to Granier, Loustau & Bréda (2000) (3)${\displaystyle \mathrm{REW}=\frac{\theta -\theta m}{\theta \mathrm{FC}-\theta m}}$ where θ is the average soil moisture of all sensors across 30-cm soil depth, θ_m is minimum volumetric soil moisture and θ_FC is the soil water at field capacity. In the plot where soil water at field capacity has not been measured, maximum volumetric soil moisture during the study period can be used as θ_FC for the REW calculation (Tor-ngern et al., 2018). Accordingly, we used the maximum and minimum θ_average that were determined from our data during the study period to represent θ_FC and θ_m, respectively.

Species selection and tree sampling

The tree species were chosen based on the relative abundance of basal area in this plot, which was calculated from the basal area of one species relative to total basal area of all species within the site. To examine the difference in tree water use, two dominant tree species with similar leaf phenology were selected for this study. As a result, Syzygium antisepticum and Adinandra integerrima, hereafter Sa and Ai, respectively, which have evergreen leaf habit, were chosen to measure water flow rate. We attempted to select trees to cover the range of size distribution within the plot, based on the inventory data from the site (W Chanthorn, pers. comm., 2018), by partitioning the tree size classes into 10-cm intervals and sampled three trees from each size class. However, due to the requirement of trees being within 25 m radius from the data logger, 4 trees of Sa and 5 trees of Ai were selected for the measurement (Table 1).

Sap flux measurement and scaling up from the point measurement to whole-tree water use

Sap flux density (J_s), which represents water mass flowing through a unit area per time in trees, was measured using self-constructed thermal dissipation probes (TDPs) (Granier, 1985). Each TDP set contains one non-heated and one heated probe being supplied with a constant ∼0.2 W electrical power. Before inserting TDPs into the stems, debarking around the drilling point was done before drilling the holes for TDP installation. Two holes were drilled with approximately 10–15 cm spacing between two probes. Based on previous studies in pine trees, the patterns of radial variation in J_s along the sapwood depth were observed with higher J_s in the outer sapwood layers than in the inner sapwood layers (Ford et al., 2004; Oishi, Oren & Stoy, 2008). Therefore, ignoring the radial variation of J_s may produce an error when scaling up from J_s to T. However, previous studies in tropical forests that use similar sap flow sensors only measured J_s at the outer sapwood because of the unknown pattern of sapwood area in tropical tree species (Horna et al., 2011; Raquel Salas-Acosta et al., 2022). In addition, most tropical trees have diffuse-porous wood without distinct annual rings and tend to have a sap flow rate that is similar along the radial sapwood depth (Lu, Urban & Ping, 2004). Therefore, we assumed that J_s was uniform along the sapwood depth of the selected trees when scaling from single-point measurements to the whole-tree level, and only measured J_s at the outer 2-cm sapwood at breast height (∼1.3 m above ground). In addition, azimuthal variation of J_s may produce variation when scaling up from J_s to T (Lu, Müller & Chacko, 2000; James et al., 2002; Tateishi et al., 2008). This variation depends on the effect of forest canopy shading by neighboring trees. In other words, trees may be obstructed from sunlight by canopy shading from surrounding trees leading to varying J_s along the circumference. In this study, the surrounding trees were equally distributed around the measured trees. Nevertheless, we installed two sensors in the north and the south directions in some trees which may be influenced by canopy shading at certain times during the day. Data from TDPs were recorded as 30-minute means of voltage difference between the probes (ΔV, mV) by the same data logger (CR1000; Campbell Scientific, Logan, UT, USA) that recorded environmental data. For the analysis, the voltage difference was converted to J_s (g m⁻² s⁻¹) using an empirical equation (Granier, 1987): (4)${\displaystyle {\mathrm{J}}_{\mathrm{s}}=118.99\times 10^{\text{−6}}\times {\left(\frac{\Delta V_m-\Delta V}{\Delta V}\right)}^{1.231}}$

where ΔV_m is the maximum voltage difference under no flow conditions which usually occurs at night and when VPD is low. The Baseliner program version 4.0 was used to select ΔV_m to calculate J_s (Oishi, Hawthorne & Oren, 2016). The program automatically determines the maximum daily ΔV to represent ΔV_m. Maximum voltage difference may occur every night if air humidity is very high, or VPD reaches 0 kPa, resulting in potentially zero water flow. However, this assumption is not valid for many ecosystems due to nighttime transpiration (e.g., Caird, Richards & Donovan, 2007; Forster, 2014; Dayer et al., 2020) or recharge of stem water (Phillips & Oren, 1998). For these reasons, no universal rule exists for identifying ΔV_m. The Baseliner software takes an approach to ΔV_m estimation by first identifying points in time where flow is likely zero and allowing the user to visually inspect and modify those points.

To scale up from J_s to T, we employed the following approach. Daily sum J_s (g m⁻² day⁻¹) was considered in the analysis to avoid issues related to the nighttime recharge of stem water that may increase as soil moisture becomes more depleted (Phillips & Oren, 1998). When nighttime recharge increases with decreasing soil moisture, the proportions of sap flux at night relative to sap flux during the day become larger. This problem can be avoided when calculating T as a daily sum (Phillips & Oren, 1998). For trees with sensors in the north and the south direction, daily sum J_s from both sensors were averaged to a mean daily J_s (J_mean) for each of them (Kunert et al., 2012). The following equation was used to estimate T: (5)${\displaystyle T=1800\times 10^{\text{−7}}\times {\mathrm{J}}_{\mathrm{mean}}\times {\mathrm{A}}_{\mathrm{S}}}$

where T is daily tree water use (L d⁻¹), J_mean is mean daily sum J_s (g m⁻² day ⁻¹) and A_S is sapwood area (cm²). In both species, A_S was estimated based on an allometric equation which was derived from 13 dominant species in an old growth and a secondary forest (the same plot as this study site) at Khao Yai National Park as follows (Yaemphum, Unawong & Tor-Ngern, 2022): (6)${\displaystyle y=0.728{\times }^{1.998}}$

where y is sapwood area (cm²), x is diameter at breast height (cm).

Data analysis

For the analysis, we used the environmental data and T between 18 September 2020 to 26 November 2022. The data covered two years which represents a wide range of environmental conditions. To avoid the potential effects of wet canopy conditions that may inhibit T when the leaf surface is covered with water droplets (Aparecido et al., 2016), we selected the days under rain-free conditions to perform the analysis.

To evaluate the responses of T to environmental factors including VPD and REW, we performed a boundary line analysis (Schäfer, Oren & Tenhunen, 2000) to obtain the response of T to environmental factors under non-limiting conditions. Trees were categorized based on the size distribution of each species as presented in Table 1 into large trees (DBH ≥ 10 cm for Ai and DBH ≥ 20 cm for Sa) and small trees (DBH <10 cm for Ai and DBH <20 cm for Sa). This results in 2 trees for both species in the small class, and three Ai trees and two Sa trees in the large class. After that, T from all trees in the same category was averaged to mean T (T_mean) for each day. Tree water use varies with VPD, REW and PAR (Phillips & Oren, 2001). Based on our data during the study period, VPD and PAR were highly correlated (r = 0.79, p ≤ 0.001), therefore we focused on VPD and REW as environmental driving variables. We performed boundary line analysis after partitioning data into three REW classes based on the REW distribution including low soil moisture (REW < 0.1), intermediate soil moisture (REW 0.1−0.4), and high soil moisture (REW > 0.4). With two classes of tree size (large and small), we had six subsets of data in both species. Each subset was subjected to the boundary line, designed to select data representing the maximum T_mean for each tree size in each REW class along the range of VPD. The upper boundary line was derived by (1) partitioning T_mean data of each REW class into at least five VPD intervals for appropriate number of data points in regression analysis (at least five data points per analysis), (2) calculating the mean and standard deviation of T_mean in each interval, (3) removing outliers using Dixon’s test, (4) selecting the data falling above the mean plus one standard deviation and (5) averaging the selected data for each VPD interval. For each tree size and REW class, the mean T_mean values of all VPD intervals obtained in step (5) were analyzed by regression analysis. All regression analyses were performed in SigmaPlot version 12.0 (Systat Software, Inc., San Jose, CA USA). Data management and analysis were performed with Rstudio, version 1.3.1073 (RStudio Team, 2020).

Environmental conditions in the study site

During the study period, there were 52% rainy and 48% rain-free days. The average daily VPD and PAR inversely corresponded with rainfall, being low when rainfall occurred and vice versa. The maximum and minimum values of PAR during the study period were 575 and 57.3 µmol m⁻² s⁻¹, respectively, with an average of 345.76 ±103.47 µmol m⁻² s⁻¹. The average daily VPD was 0.34 ±0.23 kPa. Volumetric soil moisture of all depths was averaged into θ_average. The maximum and minimum θ_average during the study period were 0.2 and 0.04 m³m⁻³, respectively. The θ_average was then used to calculate REW with an average value of 0.44 ± 0.25. Figure 1 summarizes the environmental conditions during the study period.

Tree water use of Syzygium antisepticum and Adinandra integerrima

Tree water uses of both species during the study period are shown in Fig. 2. The average T values with one standard deviation of Sa and Ai were 21.48 ± 7.73 and 10.01 ± 4.04 L d⁻¹, respectively. Comparing T between both species, we found that the T of Sa was significantly higher than that of Ai under high soil moisture and high light conditions (p < 0.0001).

Responses of tree water use to environmental factors in different tree size classes

Figure 3 summarizes the results of the responses of T to VPD under various REW ranges, with the regression statistics in Table 2. At low soil moisture (REW < 0.1, black circles), T of Sa increased with increasing VPD and gradually saturated at high VPD while that of Ai did not respond to the changing VPD, regardless of tree size. Under intermediate soil moisture conditions (REW 0.1− 0.4, gray squares), the T of both species in both sizes did not respond to VPD. Under high soil moisture (REW > 0.4, red triangles), the T of both species in both sizes followed the saturating exponential pattern as previously described in the case of low soil moisture. However, the sensitivity of increasing T at low VPD was different between the species and size class. In both species, T of large trees was less sensitive to rising VPD than small ones (Table 2).

Implications for environmental management

The results from this study imply that Sa may provide ecosystem disservice in dry areas due to its high water consumption which results in low water supply for the downstream community, but it may slow down runoff in the region that experiences heavy precipitation. In contrast, Ai may provide ecosystem benefits by conservatively using water, even under drought conditions, but may increase runoff when storms come with high rainfall. Another implication is that Ai may be suitable for reforestation in the area where droughts frequently occur in downstream ecosystems through its conservative water-use behavior, thus maintaining runoff from the forests during drought. Moreover, because Ai showed relatively constant water use regardless of tree size, the species would still provide such benefits to the ecosystems even when it grows larger in the future. In contrast, Sa may be appropriate for reforestation in the area with frequent floods because it has high water consumption during high water availability which may decelerate runoff from forests into downstream ecosystems. This would benefit the downstream ecosystems when storms occur. Regardless, mixed planting species seem to be suitable for reforestation in the areas where extreme events do not frequently occur because both species can maintain their water use at moderate soil moisture regardless of tree size, therefore preventing the depletion of soil water availability. In addition, mixed planting species could reduce the competition for limited water resources because the differences in root structures of different tree species lead to less competition for water (Schwendenmann et al., 2015). Nevertheless, reforestation projects should emphasize the use of native species to avoid competition with other native trees on the site (Hooper, Condit & Legendre, 2002).