Tree water use is a critical component of the forest water cycle and is influenced by global climate changes, such as shifts in precipitation patterns. These changes may disproportionately affect forest runoff depending on how sensitive tree water use is to environmental conditions. Therefore, understanding the water-use strategies of different tree species is essential for predicting how forests will respond to environmental changes. This study investigated how daily sap flux density (J_s), representing water flow per a unit of sapwood area, varies with environmental factors in eight common tree species in successional tropical forests in Thailand. Using thermal dissipation probes, we measured J_s in an old-growth (OF) and a young forest (YF). When averaging across all species, trees in OF were highly sensitive to rising vapor pressure deficit (VPD) under low soil moisture, responding to high atmospheric demand, while those in YF maintained their water use rate regardless of changing VPD. Additionally, species-specific response patterns were observed across different soil moisture conditions at both sites. In OF, Syzygium syzygoides and Cinnamomum subavenium exhibited conservative water use under low soil moisture, potentially preventing them from negative effects from droughts. In YF, J_s of Adinandra integerrima saturated earlier than other species in dry soils, possibly indicating greater drought tolerance compared to others. These findings provide valuable insights into species-specific water-use patterns across different successional forests which may benefit forest management, particularly the selective planting of suitable species in reforested areas.

The tree species were selected based on their relative basal area abundance and number of trees within each plot, using inventory data from the sites (W. Chanthorn, pers. comm, 2018). This was determined by calculating each species' basal area as a proportion of the total basal area across all species in each site. We focused on common tree species, i.e., those in the top 20 tree species with high relative basal area ranking and more than 100 trees. However, our choice of species selection was limited to the requirement that the sampled trees for sap flux measurement had to be within 25 m radius from the data logger. This requirement resulted in 21 trees of eight evergreen species in total, with one species existing in both sites (Syzygium antisepticum). Sap flux density (g cm^-2 s^-1) was measured using self-constructed thermal dissipation probes (Granier, 1985). Details about sap flux density measurement and its estimation are described in Ampornpitak et al. (2023).

Our monitoring period covered January 2021 to April 2022 in OF, and September 2020 to June 2023 in YF. The data covered at least one year in both sites, representing a wide range of environmental conditions. Wet canopy conditions may inhibit tree water use when the leaf surface is covered with water droplets (Aparecido et al., 2016); hence, we selected the days under rain-free conditions to perform the analysis. Unfortunately, rainfall data in OF were unavailable from 20 February 2021 to 6 October 2021 due to technical issues with the rain gauge. Nevertheless, we found a strong correlation between the rainfall data from OF and YF (R = 0.60, p < 0.0001), allowing us to assume similar rainfall patterns. Therefore, we referred to the rainfall from YF to filter out rainy days for further analysis in OF during the period of missing data.

To evaluate how sap flux density varies with environmental factors in common tree species in successional tropical forests, we performed a boundary line analysis (Ewers et al., 2001; Schäfer et al., 2000) to obtain the response of daily sum sap flux density (J_s; g cm^-2 day^-1) to environmental factors on the daily scale under non limiting conditions. This technique allows us to obtain responses with a single factor without the effects from other confounding factors. For each site, the species average J_s was used to assess the response of J_s to environmental factors with the boundary line analysis. Generally, VPD, REW, and PAR are the environmental variables that affect J_s (Phillips & Oren 2001; Wright et al., 2023; Zhao et al., 2021). Based on the data from the monitoring period, we found a high correlation between VPD and PAR in both sites (R = 0.71, p ≤ 0.001 in OF and R = 0.73, p ≤ 0.001 in YF). Therefore, we mainly focused on VPD and REW as the environmental driving factors in this study. Before performing the analysis, we filtered the data by selecting those under no rainfall conditions. Next, we partitioned soil moisture data into two classes based on REW distribution, including low soil moisture (REW < 0.1 for YF and < 0.2 for OF) and high soil moisture (REW > 0.1 for YF and > 0.2 for OF). These filtering and partitioning resulted in two subsets of data in each forest plot and each common species. To perform the boundary line analysis, we then (1) partitioned J_s data of each REW class into at least five VPD intervals, (2) calculated J_s mean and its standard deviation of each VPD interval, (3) removed outliers using Z score (95% confidence interval), (4) chose the data above mean plus one standard deviation, and (5) averaged the chosen data for each VPD interval. The data set from step (5), which represents the maximum of mean J_s, were then analyzed using regression analysis with VPD in different soil moisture conditions (i.e., REW classes). Data management (e.g., partitioning, filtering) and data analysis were performed in R software (version 1.3.1073), and regression analyses were done in SigmaPlot version 15.0 (Systat Software, Inc., San Jose, CA USA).