Seasonal Dynamics, Environmental Drivers, and Hysteresis of Sap Flow in Forests of China’s Subtropical Transitional Zone

Abstract

The subtropical transitional zone of China exhibits highly complex climatic conditions and diverse forest ecosystems, making it a critical region for understanding vegetation–water interactions. This study employed the Thermal Dissipation Probe (TDP) method to monitor sap flow in three typical forest types—evergreen broad-leaved forest, bamboo forest (Dendrocalamus latiflorus), and Chinese fir (Cunninghamia lanceolata)—in a subtropical transitional watershed in southern China. The aims were to quantify seasonal and annual variations in sap flow, to examine the effects of environmental drivers, and to analyze the hysteretic responses between sap flow and the drivers. The main findings were as follows: (1) bamboo forests exhibited significantly higher sap flow density than evergreen broad-leaved and fir forests at both annual and seasonal scales, though the overall transpiration of bamboo forests was lower than the others due to its limited sapwood area; (2) sap flow was positively correlated with potential evapotranspiration, solar radiation (Ra), vapor pressure deficit (VPD), air temperature, and soil temperature, while it was negatively correlated with relative humidity, atmospheric pressure, soil moisture, and precipitation; (3) Ra and VPD were identified as the dominant drivers of sap flow variations, with nonlinear increases that leveled off once thresholds were reached; (4) clear hysteresis patterns were observed, with sap flow peaks consistently lagging behind Ra but occurring earlier than VPD. These results advance our understanding of forest water-use strategies in the subtropical transitional zone and provide a scientific basis for improving water resource management and ecosystem sustainability in this region.

1. Introduction

Evapotranspiration is one of the most active hydrological components of the global water cycle and introduces considerable uncertainties to water resource security [1]. As a critical process in the hydrological system, transpiration (Tr) accounts for approximately 61.54% (±0.44%) of the precipitation that is returned to the atmosphere [2]. Numerous studies have demonstrated that the variation in Tr across different forest types is driven by regional environmental factors and species-specific physiological characteristics [3,4]. Therefore, a clearer understanding of these mechanisms is vital for advancing ecohydrological research and guiding the sustainable management of forest water resources under changing environmental conditions. At present, a variety of instrument-based techniques and ecological models are used to quantify vegetation Tr. At the local scale, the lysimeter method [5], the pan evaporation method [6], the gas exchange method [7], and the sap flow method [8] are widely applied in small-scale experiments to accurately estimate Tr. Among these methods, the Thermal Dissipation Probe (TDP) method, proposed by Granier in 1987, is an important approach for accurately, continuously, and rapidly measuring tree Tr [9]. In the TDP method, Tr is calculated as a function of measured sap flow and the sapwood area of monitored trees. So far, the TDP method has played a critical role in various fields, including investigating vegetation eco-physiological processes [7,8], assessing forest water resource [10], detecting drought [11], and managing agricultural production [12].

Numerous studies have demonstrated that sap flow varies across plant species due to differences in physiological characteristics, and that the driving effects of the same environmental factors on plant Tr are not always consistent among species [8,11,13]. For example, in the Chinese Loess Plateau, the sap flow of Salix matsudana was significantly higher than that of Populus simonii from July to September, whereas the opposite pattern was observed during other months [14]. A study on fruit trees in the Loess Plateau revealed that the primary driving factor of Tr in peach (Prunus persica) and apple (Malus pumila) trees was solar radiation (Ra), whereas vapor pressure deficit (VPD) was considered the predominant factor in regulating the Tr of walnuts (Juglans regia L.) [15]. Additionally, it was found that the stomata responses of walnuts to variations in Tr were more sensitive than those of peach and apple trees [15]. In Canadian boreal forests, the sap flow of balsam fir (Abies balsamea (L.) Mill.) and black spruce (Picea mariana (Mill) BSP) was mainly influenced by VPD and Ra, with stomatal conductance decreasing as VPD increased, resulting in a diminishing rate of increase in Tr once VPD reached a certain threshold [13]. Excessively high VPD can subsequently reduce plant photosynthesis and respiration rates, ultimately increasing the risk of carbon starvation and hydraulic failure [16]. Collectively, these findings highlight the complex interactions between plant physiological traits and environmental factors in regulating sap flow. In addition, most existing studies on variations in Tr and their controlling factors have focused on tropical, temperate, and arid regions, while the characteristics of vegetation transpiration and its response to environmental drivers in the subtropical zone of China are not yet well-grasped.

The subtropical transitional zone of China is characterized by diverse forest types and complex topography, which contribute to the spatial and temporal heterogeneity of Tr in this region. Chen et al. (2019) investigated seasonal variations in the Tr of Schima superba, Eucalyptus citriodora, and Acacia auriculaeformis under different environmental conditions [17]. Their results revealed that summer transpiration was considerably higher than in other seasons, primarily driven by increased photosynthetically active radiation. Among the species studied, Schima superba exhibited markedly greater transpiration compared with the other two species [17]. Besides, several studies have reported that a hysteresis phenomenon exists between environmental drivers and sap flow in various tree species across subtropical areas of China [18,19]. For example, the peak sap flow of Populus euphratica occurred approximately one hour later than the peak Ra [20]. In addition to native forests, Dendrocalamus latiflorus Munro (Poaceae) bamboo forests and Cunninghamia lanceolata (Lamb.) Hook. (Cupressaceae) Chinese fir forests are two other dominant planted forest types in the subtropical zone of China. The sap flow characteristics of these planted forests in the subtropical transitional zone and their relationship with environmental factors need to be further explored [21,22]. Additionally, it remains uncertain whether a hysteresis effect exists between variations in environmental factors and sap flow in the subtropical region of China [18].

To address the above gaps, this study investigated differences in Tr and the primary environmental drivers across three dominant forest types in the subtropical transitional zone of southern China. Natural evergreen broad-leaved, bamboo, and Chinese fir forests are widely distributed in this region and collectively form a distinctive subtropical forest ecosystem. Given that the rapid growth of bamboo forests may impact regional water consumption [23], it is essential to compare the ecohydrological characteristics and sap flow controls of these forest types. Specifically, based on one-year high temporal resolution (10 min) measurements of sap flow and associated environmental variables in evergreen broad-leaved, bamboo, and Chinese fir forests, the objectives of this study are (1) to quantify differences in transpiration among the three forest types, (2) to identify the primary environmental drivers to trigger Tr variation, and (3) to determine whether a time gap exists between the peak of sap flow and that of environmental factors in the three forest types. The findings of this research will improve our understanding of water-use strategies among these three forest types and provide a scientific basis for sustainable forest water resource conservation and management.

2. Methods

2.1. Study Area

As a representative mountain watershed (51 km²) in southern China, the Niupanshi Watershed (NPSW), defined as the drainage area above the Niupanshi Hydrological Station, and located in northern Guangdong Province, was selected as the study area (Figure 1). The average annual precipitation and evapotranspiration were 1332 mm and 1062 mm, respectively, based on field observations from 2021 to 2024. The mean annual temperature was about 20.1 °C, with the lowest mean temperature of −0.2 °C recorded in January 2023. The NPSW has various land cover types, and the subtropical evergreen broadleaf forests (EBFs) are the most widely distributed vegetation (around 80.42%). Cropland and deciduous broadleaf forests (DBFs) are distributed in the south of the NPSW, and subtropical shrubland and evergreen needleleaf forests (ENFs) are mainly distributed in the northern part of the watershed. The main plant species in the three EBF sites (Du_tree, LAO1, and LAO2) are Diospyros morrisiana Hance (Ebenaceae), Heptapleurum heptaphyllum (L.) Y. F. Deng (Araliaceae), and Myrsine seguinii H. Lév. (Primulaceae).

2.2. Measurement and Calculation of Sap Flow

To measure the water transport processes of different forest types, a total of 30 pairs of thermal dissipation probes [9] were installed in five sampling plots. Each tree or bamboo was equipped with a pair of TDP probes. The upper probe consists of a copper–constantan thermocouple for signal transmission and a constantan heating element powered by a constant 3.0 V DC supply. Continuous power was maintained using a rechargeable battery coupled with a solar panel, and the equipment’s integrity was inspected monthly. The lower probe contains only a copper–constantan thermocouple, serving as a temperature reference. Before installing the probes, the quality of each probe was checked: the resistance of the constantan heating element was around 21–22 Ω, and that of the copper–constantan thermocouple was 4–5 Ω. Before installing the TDP probes, the outer bark at a height of approximately 1.3 m was removed using a knife. Two vertically aligned holes (2.5 mm in diameter) were drilled, and the probe assembly was positioned into the two holes of the trees or bamboos, with the upper and lower probes spaced approximately 10 cm apart [9,24,25]. The probes were installed on the southern side of each tree or bamboo to capture more pronounced sap flow patterns. After placement, the contact points between the probes and stem were sealed with modeling clay to secure them. The probes were wrapped in plastic for mechanical protection, and a sun-blocking film was applied to shield them from direct sunlight and rain. Due to the distinct structure of bamboo, the probes used for bamboo culms had a tube length of 8.0 mm, whereas those used for other tree species had a tube length of 22.0 mm. Given that the TDP probe can introduce uncertainties due to probe stability, temperature sensitivity, radial symmetry of trees, and individual variability [26], we took several measures to improve the reliability of our results. We conducted regular monthly checks and replacements to avoid damage to the TDP probes. To minimize the impact of temperature fluctuations on our results, we used shading films to shield the probes, reducing interference from high temperatures or rainfall. Moreover, considering the potential variability in radial symmetry and the individual growth conditions of trees, we selected trees that were in average growth condition and exhibited relatively uniform growth in all directions.

The locations of the sampling plots and the diameter at breast height (DBH) of the sampling trees and bamboo culms are detailed in

Table 1. Basic information of sap flow sampling site.

Site	Lon (°)	Lat (°)	DBH (cm)	DEM (m)	Slope (°)	Aspect	Forest Type	Dominant Species (Amount)	Relative Dominance (%)
Dutree	113.3508	24.4685	13.38 ± 5.64	292	11.87	Southeast	EBF	Heptapleurum heptaphyllum (3); Litsea elongata (1); others (2)	42.6; 27.4; 30
LAO1	113.3530	24.4740	9.53 ± 1.27	392	8.21	Northeast	EBF	Myrsine seguinii (2); Diospyros morrisiana (3); Daphniphyllum oldhamii (1)	26.3; 47.8; 31.9
LAO2	113.3529	24.4744	12.48 ± 6.90	391	10.20	North	EBF	Myrsine seguinii (1); Litsea elongata (1); others (4)	12.7; 19.7; 67.6
Dubam	113.3508	24.4686	12.26 ± 5.87	294	11.45	East	Bamboo	Dendrocalamus latiflorus (6)	100
Midfir	113.3539	24.4632	14.43 ± 4.63	372	9.16	East	ENF	Cunninghamia lanceolata (5); Vernicia fordii (1)	86.1; 13.9

Note: The uncertainties of the DBH are within the 95% confidence interval. DEM: digital elevation model.

. The sampled individuals were relatively mature, with no visible dry branches or leaf discoloration. The vegetation is evergreen broad-leaved forest at the Du_tree, LAO1, and LAO2 sampling plots, bamboo forest at the Du_bam plot, and needle-leaved forest at the Midfir plot. The measured sap flow was recorded by a CR1000X data logger (Campbell Scientific, Inc., Logan, UT, USA) with a time interval of 10 min.

The variations in sap flow can cause temperature changes between the two probes, and the temperature difference can be related to sap flow flux as a function of the maximum temperature differential between the two probes [9]. The relative difference between the maximum temperature difference ($∆T_m$) and an instantaneous temperature difference ($∆T$) could be expressed as Equation (1).

$$\mathrm{K}=\left(∆T_m-∆T\right)/∆T,$$

(1)

where K refers to the relative difference between $∆T_m$ and $∆T$, which is a unitless value. Then, the sap flow can be calculated with an empirical formula [9] as Equation (2).

$$J_{s1}=119\times K^{1.231},$$

(2)

where $J_{s1}$ refers to the instantaneous sap flow density of the measured tree ($\mathrm{g}·{\mathrm{m}}^{-2}·{\mathrm{s}}^{-1}$). The two constants in Equation (2) are empirical parameters, which could be adjusted to fit the sap flow of the different tree species. Unlike trees, bamboo is a kind of perennial grass (Poaceae) whose stem structure distinctly differs from that of trees. Its thin-walled culms have limited water-conducting capacity, with vascular bundles clearly visible but unevenly distributed and varying in size. Therefore, we used the adjusted Equation (3) to calculate the sap flow of bamboo [27] as follows:

$$J_{s2}=360.6\times K^{1.7459},$$

(3)

where $J_{s2}$ refers to the instantaneous sap flow density of bamboo ($\mathrm{g}·{\mathrm{m}}^{-2}·{\mathrm{s}}^{-1}$).

The Baseliner 4.0 program was used to remove anomalous data, select $∆T_m$ anchor points, and obtain the final K value [28]. The daily average sap flow density can be calculated by Equation (4) [24], which is expressed as

$$J_{d,n}={\displaystyle \frac{{\sum }_{i=1}^{144}{J}_{s,i,n}}{144}},$$

(4)

where $J_{d,n}$ refers to the daily average sap flow of each tree ($\mathrm{g}·{\mathrm{m}}^{-2}·{\mathrm{s}}^{-1}$), and $J_{s,i,n}$ is the average sap flow of each tree during the ${\mathrm{i}}_{th}$ 10 min of each day ($\mathrm{g}·{\mathrm{m}}^{-2}·{\mathrm{s}}^{-1}$). The average sap flow for each forest type ($J_d$, $\mathrm{g}·{\mathrm{m}}^{-2}·{\mathrm{s}}^{-1}$) can be calculated by Equation (5) [24], which is expressed as

$$J_d={\sum }_{i=1}^n{J}_{d,n}/n,$$

(5)

where $n$ is the total number of dominant trees (or bamboos) in the sampling sites.

2.3. Calculation of Sapwood Area and Transpiration

It is well known that the sapwood of trees plays a vital role in conducting sap flow, storing nutrients, and providing mechanical support because its vessels are highly efficient in transporting water. The sapwood is not only an important organ for transporting water, but also influences plant Tr. Therefore, the sapwood area (AS, m²) is a key factor in calculating Tr. We used an increment borer (Haglöf, Bromma, Sweden) to obtain the thickness of the sapwood of measured trees at the sampling sites, and then established a regression relationship with their DBHs, as in Equation (6).

$${AS}_1=0.6177\times {DBH}^{2.0738}\left(R^2=0.996, n=10\right),$$

(6)

where ${AS}_1$ refers to the sapwood area of the trees.

Since the physiological structure of bamboo differs from that of trees, we directly measured the thickness of bamboo at approximately 1.3 m height near the sampling sites and established a regression relationship with their DBHs, as shown in Equation (7).

$${AS}_2=0.339\times {DBH}^{1.8926}\left(R^2=0.945, n=38\right),$$

(7)

where ${AS}_2$ refers to the sapwood area of bamboo.

The whole-tree transpiration (${Tr}_i$) could be obtained by multiplying the sapwood area with the sap flow density, as in Equation (8).

$${Tr}_i=AS\times J_{s,i},$$

(8)

Then, the daily mean whole-tree transpiration (Tr) can be expressed as Equation (9).

$$Tr={\sum }_{i=1}^{144}{Tr}_i/144,$$

(9)

2.4. Observation of Meteorological Variables

We continuously measured the meteorological factors at 10 min resolution using a Dyanmet6 weather station installed near the Du_tree sampling site. The measured meteorological data spanned September 2023 to August 2024, including precipitation (PPT, mm), total solar radiation (Ra, W/m²), wind speed (WS, m/s), air temperature (Ta, °C), air relative humidity (RH, %), atmospheric pressure (Pr, hPa), and potential evapotranspiration (PET, mm) logged using a CR1000 data logger (Campbell Scientific, Inc., Logan, UT, USA) (Figure S1). Soil moisture (SM, m³/m³) and temperature (ST, °C) were also recorded simultaneously at sap flow monitoring sites using soil probes installed at a depth of 5 cm (ECH2O-5TE, Meter Group, Inc., Pullman, WA, USA) connected to CR1000X data loggers. The atmospheric vapor pressure deficit (VPD, kPa), defined as the difference between saturated (e_s, kPa) and actual water vapor pressure, serves as an indicator of atmospheric dryness, a key climatic factor regulating plant transpiration. The saturated actual water vapor pressure can be calculated with air temperature by Equation (10) [29] and is expressed as

$${\mathrm{e}}_s=0.61078\times {\mathrm{e}}^{{\displaystyle \frac{17.27\times \mathrm{T}\mathrm{a}}{\mathrm{T}\mathrm{a}+237.3}}},$$

(10)

Using ${\mathrm{e}}_s$ and RH, VPD can be calculated by Equation (11) [29] as follows:

$$\mathrm{VPD} = {\mathrm{e}}_s-{\mathrm{e}}_s\times \mathrm{R}\mathrm{H},$$

(11)

2.5. Statistical Analysis

Before analysis, we first preprocessed both raw sap flow data and meteorological data using Excel (version 2020, Microsoft, Redmond, WA, USA) and then used Baseliner 4.0 to remove anomalous sap flow data [28]. The relationship between AS and DBH was established through exponential curve regression to estimate the AS of different trees. Due to weather and the failure of data loggers or probes, missing data was inevitable. For missing data of less than 2 h, a linear regression method was used for interpolation. To study the nocturnal sap flow, nighttime was defined as the period when solar radiation was below 2 W/m² [30]. The period from April to September was designated as the wet season, and the other months were defined as the dry season [25]. Pearson’s correlation method was used to analyze the relationships between environmental factors and sap flow. Principal component analysis (PCA) was employed to quantify the influence of individual environmental factors on sap flow and to obtain the main contributing factors. In PCA, the eigen value refers to the variance accounted for by each principal component. We selected the components whose eigen value was greater than 1.0 and the cumulative variance percentage was greater than 80% as the main influencing component. The collinearity analysis and contribution analysis were conducted to remove environmental factors with high collinearity and contributions below 5%. The Lindeman, Merenda, and Gold (LMG) method was used to calculate the contributions of major environmental factors to sap flow variation in both dry and wet seasons. Multiple regression analysis was conducted using the Least Absolute Shrinkage and Selection Operator (LASSO) regression model to simulate the relationships between sap flow and environmental factors. All statistical analyses and figures were generated using MATLAB (version R2023a, MathWorks, Natick, MA, USA).

3. Results

3.1. Sap Flow Differences Among Evergreen Broad-Leaved, Bamboo, and Fir Forests

The average sap flow demonstrated similar intra-annual variation patterns in EBFs as well as bamboo and fir forests (Figure 2). In general, sap flow declined from September to February of the following year and then increased until the end of summer. For example, the sap flow in EBFs decreased at a rate of 0.03 $\mathrm{g}·{\mathrm{m}}^{-2}·{\mathrm{s}}^{-1}·{\mathrm{d}}^{-1}$ from September to March of the next year and then increased at a rate of 0.05 $\mathrm{g}·{\mathrm{m}}^{-2}·{\mathrm{s}}^{-1}·{\mathrm{d}}^{-1}$ till the end of August. In addition, there are apparent seasonal differences in sap flow, with the highest mean sap flow occurring in summer and lowest in winter across all three forest types. Specifically, the mean summer sap flow was 8.47 $\mathrm{g}·{\mathrm{m}}^{-2}·{\mathrm{s}}^{-1}$ in EBFs, 15.62 $\mathrm{g}·{\mathrm{m}}^{-2}·{\mathrm{s}}^{-1}$ in bamboo forests, and 4.40 $\mathrm{g}·{\mathrm{m}}^{-2}·{\mathrm{s}}^{-1}$ in fir forests, which was greater than the mean winter sap flow of 2.74 $\mathrm{g}·{\mathrm{m}}^{-2}·{\mathrm{s}}^{-1}$ in EBFs, 6.04 $\mathrm{g}·{\mathrm{m}}^{-2}·{\mathrm{s}}^{-1}$ in bamboo forests, and 1.42 $\mathrm{g}·{\mathrm{m}}^{-2}·{\mathrm{s}}^{-1}$ in fir forests, respectively.

Nevertheless, significant differences in sap flow were observed among the three forests. First, the decreasing rate in sap flow from September to March of the next year was highest in bamboo forests, at a rate of 0.08 $\mathrm{g}·{\mathrm{m}}^{-2}·{\mathrm{s}}^{-1}·{\mathrm{d}}^{-1}$, which was considerably greater than that in EBFs (0.03 $\mathrm{g}·{\mathrm{m}}^{-2}·{\mathrm{s}}^{-1}·{\mathrm{d}}^{-1}$) and fir forests (0.02 $\mathrm{g}·{\mathrm{m}}^{-2}·{\mathrm{s}}^{-1}·{\mathrm{d}}^{-1}$). Second, sap flow was more variable in bamboo forests compared to the other two forest types, as illustrated by the boxplots (Figure 2b,d,f). For example, sap flow in bamboo forests ranged from 3.35 to 25.16 $\mathrm{g}·{\mathrm{m}}^{-2}·{\mathrm{s}}^{-1}$ during summer, whereas the range was narrower in EBFs (0.51 to 19.91 $\mathrm{g}·{\mathrm{m}}^{-2}·{\mathrm{s}}^{-1}$) and fir forests (0.35 to 8.19 $\mathrm{g}·{\mathrm{m}}^{-2}·{\mathrm{s}}^{-1}$). Third, the magnitude of sap flow was substantially higher in bamboo forests. The annual mean sap flow was 10.98 $\mathrm{g}·{\mathrm{m}}^{-2}·{\mathrm{s}}^{-1}$ in bamboo forests, which was much greater than 5.19 $\mathrm{g}·{\mathrm{m}}^{-2}·{\mathrm{s}}^{-1}$ in EBFs and 2.84 $\mathrm{g}·{\mathrm{m}}^{-2}·{\mathrm{s}}^{-1}$ in fir forests. Overall, the similarity of intra-annual variability of sap flow between EBFs and fir forests was higher than that between bamboo and fir forests. For example, the correlation coefficient between EBFs and fir forests was 0.77, which was higher than the value of 0.62 between bamboo and fir forests.

On a daily scale, vegetation Tr showed a distinct unimodal pattern, with relatively low values during the nighttime and elevated levels during the daytime (Figure 3). During the daytime, Tr in EBFs was comparatively and significantly higher than that in both bamboo and fir forests (p < 0.001). Seasonally, daytime Tr in EBFs and fir forests was remarkably higher in summer compared to other seasons, whereas in bamboo forests, daytime Tr in summer was roughly comparable to that in autumn. Among the three forests, the daily mean summer Tr in EBFs exceeded that of bamboo and fir forests by approximately 0.0638 g/s and 0.0170 g/s, respectively. In spring, the daily mean Tr in EBFs was about 0.0411 g/s and 0.0349 g/s higher than in bamboo and fir forests, respectively. The Tr differences among the three forests were minimal during winter. For example, the daily mean winter sap flow in EBFs was only 0.0058 g/s higher than that in fir forests. Interestingly, a notable difference was observed in the timing of peak Tr. In EBFs and fir forests, peak Tr generally occurred around 2:00 p.m., whereas in bamboo forests, it occurred earlier, around 12:00 p.m.

3.2. Effects of Meteorological Factors on Transpiration

The sap flow of vegetation was significantly and positively correlated with PET, Ra, VPD, Ta, and ST, and negatively correlated with RH, Pr, SM, and PPT at a 10 min temporal scale across all three forests (Figure S2). Among the positive correlations, PET and Ra exhibited stronger associations with Tr than the other factors. Specifically, the correlation coefficients were 0.74 between Tr and PET and 0.72 between Tr and Ra, which is greater than 0.49 between Tr and VPD, as well as 0.48 between Tr and Ta. Regarding negative correlations, RH played the strongest inhibitory effect on vegetation Tr, and this pattern was consistent across all seasons. Compared to RH and Pr, the negative effects of PPT on vegetation Tr are negligible. Overall, the relationships between Tr and individual environmental factors were largely consistent across the three forest types.

A further principal component analysis was conducted to examine the key environmental factors influencing Tr, and the results are presented in Figure 4. The results indicated that VPD, Ra, PET, and Tr showed similar contributions to the first two principal components (PC1 and PC2) across the three forests. PC1 explained the highest proportion of variance, accounting for 35.30%, 37.43%, and 34.53% in evergreen broadleaved, bamboo, and fir forests, respectively. After extracting the eigen values, we found that the cumulative variance contribution of the top four components was 81.66%, 81.10%, and 81.11% for the three forests (

Table 2. Eigen values and percentage of variance of environmental factors at 10 min timescales in NPSW based on PCA.

Vegetation Types	Indicates	PC1	PC2	PC3	PC4
EBF	Eigen value	3.88	2.66	1.41	1.03
	Percentage of variance (%)	35.30	24.18	12.81	9.37
	Cumulative variance (%)	35.30	59.48	72.29	81.66
Bamboo	Eigen value	4.12	2.33	1.41	1.06
	Percentage of variance (%)	37.43	21.23	12.78	9.66
	Cumulative variance (%)	37.43	58.66	71.44	81.10
Fir	Eigen value	3.80	2.71	1.39	1.03
	Percentage of variance (%)	34.53	24.64	12.60	9.34
	Cumulative variance (%)	34.53	59.17	71.77	81.11

), indicating that these components effectively captured the environmental drivers of vegetation Tr change. In EBFs and fir forests, SM exhibited a stronger correlation with RH, while in bamboo forests, it was more closely related to Pr and WS. The contributions of VPD, Ra, PET, and Tr to PC1 and PC2 were notably similar, indicating a stronger relationship between each of these environmental factors and Tr. Furthermore, Ta and ST showed significant contributions to PC1, implying that temperature may also play an important role in regulating forest Tr. Detailed loading contributions for environmental factors on the principal components are listed in Table S1.

Multiple regression analysis was used to develop regression models (

Table 3. Multiple regression and contribution analysis between sap flow and main environmental factors in dry and wet seasons.

Forest Type	Time Scales	Regression Equation	Main Contribution (%)
EBF	Year	Js = −0.0374 + 0.0004·Ra + 0.0027·Ta + 0.0601·VPD	Ra: 79.4; Ta: 10.2; VPD: 10.4
	Dry	Js = 0.0078 + 0.0002·Ra + 0.0540·VPD	Ra: 75.7; VPD: 24.3
	Wet	Js = −0.1345 + 0.0004·Ra + 0.0065·Ta + 0.0494·VPD	Ra: 86.6; Ta: 8.2; VPD: 5.2
Bamboo	Year	Js = −0.0038 + 0.0001·Ra + 0.0006·Ta	Ra: 89.3; Ta: 10.7
	Dry	Js = −0.0042 + 0.0001·Ra + 0.0008·Ta	Ra: 83.4; Ta: 16.6
	Wet	Js = −0.0642 + 0.0001·Ra + 0.0028·Ta	Ra: 63.9; Ta: 36.1
Fir	Year	Js = −0.0403 + 0.0003·Ra + 0.0026·Ta + 0.0861·VPD	Ra: 59.3; Ta:13.2; VPD: 27.6
	Dry	Js = −0.0021 + 0.0002·Ra + 0.0879·VPD	Ra: 53.6; VPD: 46.4
	Wet	Js = −0.1236 + 0.0003·Ra + 0.0059·Ta + 0.0781·VPD	Ra: 68.7; Ta:11.1; VPD: 20.2

Note: Wet season: April–September; Dry season: October–March of the next year.

), and the contributions of the main environmental factors were quantified with the LMG method. The results indicated that Ra, Ta, and VPD were positively correlated with vegetation sap flow at both the annual and seasonal scales. Among these factors, Ra emerged as the most important environmental factor, contributing more to Tr than to Ta and VPD in all regression models. On an annual scale, Ra accounted for 89.3% of the Tr variation in bamboo forests, which was higher than its contribution in EBFs (79.4%) and fir forests (59.3%). Ta also played a non-negligible role, contributing 10.2% and 13.2% of Tr variation across all three forests. In contrast, VPD made substantial contributions to vegetation Tr variation in EBFs (10.4%) and fir forests (27.3%). Seasonal analysis showed that Ra contributed more to vegetation Tr during the wet season compared to the dry season in both EBFs and fir forests, whereas VPD had the opposite effects. Notably, the contribution of Ta to Tr was consistently higher during the wet season across the three forest types.

3.3. Variation of Sap Flow Under Shifting Environmental Conditions

To further investigate how sap flow responds to changing environmental conditions, additional analyses were conducted, and the results are shown in Figure 5 and Figure 6. Overall, sap flow increased with rising Ra and VPD, but the increasing rates of sap flow were different across the three forest types. For Ra, the increasing rate of sap flow with an increase in Ra was 0.04 $\mathrm{g}·{\mathrm{s}}^{-1}·{\mathrm{W}}^{-1}$ in EBFs, 0.08 $\mathrm{g}·{\mathrm{s}}^{-1}·{\mathrm{W}}^{-1}$ in bamboo forests, and 0.02 $\mathrm{g}·{\mathrm{s}}^{-1}·{\mathrm{W}}^{-1}$ in fir forests. However, the increase in sap flow as Ra increases differed slightly among the three forests. A clear linear relationship between sap flow and Ra was observed in these three forests (Figure 5a,d,g). However, in bamboo forests, the rate of increase in sap flow began to level off once Ra exceeded 300 W/m². These patterns were consistent across both the dry and wet seasons (Figure 5). During the dry season, the increasing rate in sap flow with Ra was 0.03 $\mathrm{g}·{\mathrm{s}}^{-1}·{\mathrm{W}}^{-1}$ in EBFs, 0.06 $\mathrm{g}·{\mathrm{s}}^{-1}·{\mathrm{W}}^{-1}$ in bamboo forests, and 0.02 $\mathrm{g}·{\mathrm{s}}^{-1}·{\mathrm{W}}^{-1}$ in fir forests. Likewise, sap flow exhibited a linear increasing trend across all forest types during the wet period, with the relationship leveling off again in bamboo forests when Ra surpassed 300 W/m² (Figure 5f). In addition, the increasing rate of sap flow was greater in the wet period than in the dry period. For example, the increasing rate of sap flow in the dry period was approximately 0.02 $\mathrm{g}·{\mathrm{s}}^{-1}·{\mathrm{W}}^{-1}$, 0.02 $\mathrm{g}·{\mathrm{s}}^{-1}·{\mathrm{W}}^{-1}$, and 0.01 $\mathrm{g}·{\mathrm{s}}^{-1}·{\mathrm{W}}^{-1}$ lower in both EBFs and bamboo and fir forests compared to the wet season.

As for the relationship between vegetation sap flow and VPD, sap flow also tended to increase as VPD increased across all three forest types. The increasing rate was 9.99 $\mathrm{g}·{\mathrm{m}}^{-2}·{\mathrm{s}}^{-1}·{\mathrm{k}\mathrm{P}\mathrm{a}}^{-1}$ in EBFs, 12.64 $\mathrm{g}·{\mathrm{m}}^{-2}·{\mathrm{s}}^{-1}·{\mathrm{k}\mathrm{P}\mathrm{a}}^{-1}$ in bamboo forests, and 6.76 $\mathrm{g}·{\mathrm{m}}^{-2}·{\mathrm{s}}^{-1}·{\mathrm{k}\mathrm{P}\mathrm{a}}^{-1}$ in fir forests. However, the increased rate of sap flow as the VPD increases varied among the three forests. Specifically, there appeared a continuously increasing trend in sap flow in EBFs as VPD increased (Figure 6a), while such an increasing trend was flattened as VPD reached above 1.2 kPa in both bamboo and fir forests (Figure 6d,f). This trend of increasing sap flow with rising VPD was observed in both the dry and wet periods (Figure 6). For example, the sap flow increasing trend during the dry period with VPD was 5.55 $\mathrm{g}·{\mathrm{m}}^{-2}·{\mathrm{s}}^{-1}·{\mathrm{k}\mathrm{P}\mathrm{a}}^{-1}$ in EBFs, 9.91 $\mathrm{g}·{\mathrm{m}}^{-2}·{\mathrm{s}}^{-1}·{\mathrm{k}\mathrm{P}\mathrm{a}}^{-1}$ in bamboo forests, and 4.69 $\mathrm{g}·{\mathrm{m}}^{-2}·{\mathrm{s}}^{-1}·{\mathrm{k}\mathrm{P}\mathrm{a}}^{-1}$ in fir forests. Moreover, when VPD approached 1.5 kPa, the upward trend in sap flow tended to saturate or even show a declining slope in the three forest types (Figure 6b,e,h), which was consistent across both the dry and wet periods (Figure 6c,f,i). Finally, the increasing rate of sap flow was generally greater in the wet period (about 11.15 $\mathrm{g}·{\mathrm{m}}^{-2}·{\mathrm{s}}^{-1}·{\mathrm{k}\mathrm{P}\mathrm{a}}^{-1}$ in EBFs and 6.57 $\mathrm{g}·{\mathrm{m}}^{-2}·{\mathrm{s}}^{-1}·{\mathrm{k}\mathrm{P}\mathrm{a}}^{-1}$ in fir forests, respectively) compared to the dry period.

3.4. The Difference in the Variability of Sap Flow and Major Environment Factors

A comparison of the diurnal patterns of sap flow, Ra, and VPD revealed that their respective unimodal curves differed in timing, as illustrated in Figure 7. Specifically, the peak sap flow consistently lagged slightly behind the peak Ra across all three forest types. On average, the time lag between peak Ra and peak sap flow was approximately 29.4 min in EBFs, 1.7 min in bamboo forests, and 68.0 min in fir forests (

Table 4. Correlation coefficients and time lags (minutes) between Ra vs. sap flow and VPD vs. sap flow in EBFs as well as bamboo and fir forests (STD: standard deviation).

Indicator	Time Scale	EBFs		Bamboo Forests		Fir Forests
		R	Time Lag (STD)	R	Time Lag (STD)	R	Time Lag (STD)
Ra	Year	0.68	29.4 (16.4)	0.74	1.7 (3.7)	0.74	68.0 (4.0)
	Dry	0.67	53.8 (24.2)	0.72	10.0 (0)	0.83	94.0 (4.9)
	Wet	0.76	21.3 (12.7)	0.81	0 (0)	0.76	48.0 (7.5)
VPD	Year	0.58	−146.3 (24.7)	0.66	−171.7 (6.9)	0.67	−102.0 (7.5)
	Dry	0.59	−127.5 (25.9)	0.66	−170.0 (11.0)	0.75	−90.0 (9.0)
	Wet	0.65	−153.7 (28.5)	0.72	−178.3 (13.4)	0.71	−112.0 (7.5)

). Such phenomena applied to both the dry (October–March of the next year) and wet (April–September) periods (Figure 7b vs. Figure 7c, Figure 7e vs. Figure 7f, Figure 7h vs. Figure 7i). Among the three forest types, fir forests exhibited the longest time lag, while bamboo forests showed the shortest. These results suggest that the sensitivity of sap flow to solar radiation varies across forest types, potentially due to differences in canopy structure, leaf physiology, and stomatal behavior.

On the contrary, the peak sap flow occurred significantly earlier than the peak VPD across all forest types, with time lags of 146.3 min in EBFs, 127.5 min in bamboo forests, and 102.0 min in fir forests. This earlier occurrence of peak sap flow relative to VPD was consistently observed during both the dry and wet periods. Among the three forest types, the time lead of sap flow over VPD was greatest in bamboo forests and smallest in fir forests.

4. Discussion

4.1. Comparison of Sap Flow Among Different Vegetations

Sap flow in EBFs, bamboo forests, and fir forests in this study all decreased from September to February and then increased until the end of the next summer. These intra-annual variations in sap flows were consistent with previous findings for forests in southern China [24,31]. The seasonal changing trends in mean daily sap flow were similar across the three forest types, with the peak sap flow occurring in the middle of the wet season (July–August) [31]. However, the magnitudes of sap flow differed among species [21]. In this study, the mean annual sap flow of EBFs was 5.19 $\mathrm{g}·{\mathrm{m}}^{-2}·{\mathrm{s}}^{-1}$, slightly lower than the approximately 6.55 $\mathrm{g}·{\mathrm{m}}^{-2}·{\mathrm{s}}^{-1}$ observed for similar forest types in South China Botanical Garden [32,33] (Table S2). This discrepancy may be attributed to differences in tree age, height, AS, and DBH, as well as the generally lower air temperature in our study area compared with the South China Botanical Garden, which likely constrained transpiration in the NPSW. In general, the sap flow measured in this study fell within a reasonable range, and it was higher than Zelkova schneideriana (2.26 $\mathrm{g}·{\mathrm{m}}^{-2}·{\mathrm{s}}^{-1}$) and Euonymus bungeanu (2.51 $\mathrm{g}·{\mathrm{m}}^{-2}·{\mathrm{s}}^{-1}$) in northern cold climate, but much lower than Hevea brasiliensis (23.00 $\mathrm{g}·{\mathrm{m}}^{-2}·{\mathrm{s}}^{-1}$) in tropical climate [34,35]. Greater soil moisture and higher Ra in tropical regions promote photosynthesis, which in turn leads to greater sap flow in tropical vegetation.

In this study, the average sap flow of D. latiflorus bamboo forests was 10.98 $\mathrm{g}·{\mathrm{m}}^{-2}·{\mathrm{s}}^{-1}$, which was 2.35 times higher than that of B. blumeana bamboo forests (4.67 $\mathrm{g}·{\mathrm{m}}^{-2}·{\mathrm{s}}^{-1}$) [36]. Despite the higher sap flow observed in bamboo forests, the lower AS of bamboo forests restricted its Tr. For example, the relatively small AS in bamboo forests limits its overall transpiration. Specifically, the average AS in bamboo forests was 24.27 cm², substantially smaller than that of EBFs (112.66 cm²) and fir forests (175.15 cm²), which explains why the Tr in bamboo forests was much lower than in EBFs and fir forests. Moreover, sap flow in fir forests was slightly lower than that in EBFs, likely due to differences in growth form and stand structure. In particular, in order to grow and compete for light, fir forests (Cunninghamia lanceolata) gradually shed lower branches and foliage, reducing photosynthetic capacity and leading to lower sap flow [37,38].

Previous studies have demonstrated that nocturnal water transport processes cannot be ignored when estimating nocturnal Tr, as water continues to move within plants during the night [39,40]. In this study, nocturnal sap flow accounted for 5.21%–18.60% of the total sap flow across forest types (Table S2), confirming the presence of nocturnal sap flow in the NPSW. Specifically, the mean nocturnal sap flow proportion in bamboo forests (9.71%) was lower than that in fir forests (12.88%) and EBFs (13.16%). Previous research has shown that nocturnal sap flow can account for approximately 5%–25% of daytime Tr [41], with the proportion varying across climates and species. For example, the nocturnal sap flow proportion may reach up to 60% of daytime sap flow in arid regions [42]. The proportion of nocturnal sap flow also varies across forest types. A review showed that nocturnal sap flow accounted for 25.26% of sap flow in tropical rainforests, 19.15% in tropical sparse forests, 13.34% in subtropical EBFs, 6.14% in subtropical sclerophyllous forests, and 6.47% in temperate deciduous forests [41]. Our observed mean nocturnal sap flow accounted for 13.16% of daytime flow in EBFs, very close to the value (13.2%) reported for subtropical EBFs comprising 19 species [41]. From a physiological perspective, excessive daytime transpiration can deplete water stored in plant tissues, necessitating stem water refilling at night [43]. The stem water refilling process plays a critical role in sustaining photosynthesis and transpiration the next day.

4.2. Influence of Environmental Factors on Sap Flow

The results showed that Ra and VPD are the two dominant environmental drivers of sap flow across the three forest types. This result was highly consistent with numerous previous studies demonstrating that Ra and VPD made non-negligible contributions to vegetation Tr [18,44]. However, the relative effects of Ra and VPD on sap flow varied among forest types and across different regions. For instance, a study conducted in the Haihe River Basin, China, reported that the contribution of Ra to variations in sap flow ranged from 81.8% to 96.4% [18], which was higher than the contributions we observed for EBFs (79.4%), bamboo forests (89.3%), and fir forests (59.3%). Compared to the Haihe River Basin, southern China has more rainfall, and soil moisture is generally higher, which may decrease the relative importance of Ra in Tr. Similarly, Ouyang et al. (2022) found that the sap flow of EBFs was positively correlated with Ra and Ta in southern China, which agrees with our findings [25]. Moreover, our results suggest that Ta played a more important role in driving sap flow during the wet period than the dry period, as shown in Table 3. Increased radiation typically promotes stomatal opening, thereby enhancing transpiration and elevating sap flow. In addition, high radiation raises leaf temperature, which intensifies the vapor pressure gradient and further accelerates water loss [45]. However, under conditions of excessive radiation and limited soil water availability, trees may partially close their stomata to safeguard hydraulic function and prevent excessive water depletion [46]. Therefore, Ra regulates sap flow not only through its direct influence on leaf temperature but also indirectly via stomatal regulation.

As for the relationship between vegetation sap flow and VPD, we found that sap flow was more strongly influenced by VPD during the dry period than during the wet period. This pattern was highly consistent with the results of Ma et al. (2008) [44], who observed a shorter hysteresis (~40 min) in the dry period than in the wet period (~60 min). VPD regulates the opening and closure of the stomata of leaves and thus canopy conductance, thereby influencing plant Tr [47]. Previous research has also shown that sap flow tends to increase with rising VPD when VPD is below approximately 1.0 kPa [19]. Our study indicated that sap flow still increased as VPD rose beyond 1.0–1.5 kPa, but at a diminished rate. A higher VPD generally increases the atmospheric demand for water, thereby stimulating transpiration and sap flow. At moderate levels, this promotes the efficiency of water transport and gas exchange. However, when VPD becomes excessively high, trees may partially or fully close their stomata to reduce water loss and prevent the risk of xylem embolism [19]. Consequently, the effect of VPD on sap flow reflects a balance between driving transpiration and the physiological regulation of stomata aimed at maintaining hydraulic safety. What is more, a previous study has indicated that Ra exerted a stronger influence on sap flow during the daytime, while VPD becomes a more dominant driver at night.

Our results suggested that time intervals and other environmental factors play an important role in regulating sap flow, as reported in previous studies [25,48,49]. Some studies have indicated that sap flow is more strongly correlated with Ra and VPD at shorter, hourly timescales [18], while ST and SM exert greater influence at longer time scales [40]. According to Figures S2–S4, sap flow was negatively correlated with RH, PPT, and SM across different time intervals. This is partly because frequent rainfall can decrease solar radiation and thus restrain the plant transpiration process. For example, Ouyang et al. (2022) found that the Tr of Schima superba was positively correlated with Ra and Ta, but negatively correlated with SM and PPT during the dry period [25]. Under limited soil water supply, Schima superba may improve water-use efficiency as a drought adaptation strategy, allowing it to maintain stable transpiration even during dry periods [25]. Nevertheless, the overall impact of precipitation on sap flow remains unclear. Some studies have indicated that peak sap flow often occurs during wet periods because increased precipitation improves soil moisture conditions and supports transpiration [50]. However, another study found that precipitation was not necessarily the primary factor driving Tr [51], and both excessive precipitation and prolonged drought may reduce sap flow [44].

The leaf area index (LAI) is also an important environmental factor influencing canopy transpiration and thus plays a critical role in shaping plant water-use strategies. When the LAI is below a certain threshold, there is a significant positive correlation between sap flow and LAI. However, once the threshold is surpassed, sap flow is primarily driven by environmental factors [52]. Previous studies have shown that the growth state and the root depth of vegetation are influenced by climatic conditions, leading to variations in water-use efficiency and hydraulic traits, which in turn affect sap flow [53,54]. During the growing season, root-mediated water redistribution is generally low. However, under extreme drought conditions, vegetation roots can absorb large amounts of water. In such cases, roots may contribute up to 64.6% of the total daily sap flow, demonstrating their ability to self-regulate in response to changing hydraulic conditions [54].

4.3. The Time Lag of Sap Flow in Different Species

The hysteresis phenomenon between sap flow and both Ra and VPD observed in this study has been reported previously [20,55]. A study involving 52 tree species showed that the peak Ra preceded peak sap flow by approximately 80.4 (±37.2) minutes [55], which is substantially longer than the lag times observed in both EBFs (29.4 min) and bamboo forests (1.7 min) in our study, but comparable to that of fir forests (68.0 min). These results indicate that sap flow in EBFs and bamboo forests was more sensitive to changes in Ra than in fir forests. Zhao et al. (2024) reported that peak sap flow in Populus euphratica lagged peak Ra by approximately one hour [11]. The process of water release from leaves was regulated by sunrise and sunset, and it took time for water to be transported from the trunk to branches and leaves, which might explain the hysteresis phenomenon between sap flow and Ra [56]. Although all three forest types showed similar sap flow cessation at sunset, the increase at sunrise was notably slower in fir forests than in EBFs and bamboo forests (Figure 8). This suggested that fir leaves were less sensitive to Ra than the other two forest types. For example, in P. tabulaeformis, a needle-leaved species, sap flow usually lagged Ra by about 0–3 h, and it was even longer than 3.5 h during dry periods [57]. Wan et al. (2023) proposed that the hysteresis phenomenon in needle-leaved forests would intensify with increasing sapwood area [55]. Additionally, soil moisture influences hysteresis by prolonging the time lag between peak Ra and peak sap flow [55].

The pronounced hysteresis between sap flow and VPD was also evident across all three forest types. The time at which peak sap flow preceded peak VPD was about 146.3 min in EBFs, 127.5 min in bamboo forests, and 102.0 min in fir forests. These findings were consistent with those of Wang et al. (2008), who observed that the peak VPD generally lagged the peak sap flow around 1–3 h during both the dry and wet periods [58]. Similarly, Zhao et al. (2014) found that the peak sap flow in Populus euphratica would occur around 2 h earlier than the peak VPD [20]. Another study indicated that the peak sap flow in Ulmus macrocarpa could be 20–40 min earlier than the peak VPD under high soil moisture conditions, and 40–60 min earlier than at lower soil moisture [19]. The hysteresis between peak sap flow and VPD was primarily attributed to differences in soil hydraulic conductivity and stomatal closure. Variations in stomatal conductance sensitivity to VPD across species represent a key driver of this lag [16]. In addition, tree hydraulic conductivity and trunk water recharge capacity have also been implicated in producing hysteresis between sap flow and VPD [58]. Furthermore, the lag of VPD behind Ra that we observed in this study has also been documented in Eucalyptus and Callitris glaucophylla-dominated forests in Australia [59], as well as in a tropical forest comprising 10 species at a wet tropical research station [60].

4.4. Practical Implications

This study advances our understanding of ecohydrological processes in subtropical transitional zones by quantifying sap flow dynamics and examining their responsive mechanisms across different forest types. The results revealed that bamboo forests exhibited lower transpiration than EBFs and fir forests, suggesting that large-scale bamboo plantations will not lead to excessive water consumption. However, it should be noted that extensive bamboo cultivation may decrease regional evapotranspiration, which has the potential to exacerbate flood vulnerability in precipitation-prone areas by weakening the water-regulating capacity of the natural ecosystem. These findings provide valuable insights for forest management and land-use policy, highlighting the importance of a balance between bamboo cultivation and the conservation of natural forests to maintain ecosystem stability and hydrological function. Furthermore, an accurate estimate of forest water demand based on measured transpiration rates and canopy coverage will offer practical value not only for precise irrigation scheduling in managed plantations but also for afforestation and reforestation planning in subtropical regions. Such information can guide the selection of tree species and management practices that optimize water-use efficiency while sustaining ecosystem services.

This study has several limitations. First, the NPSW represents only a small fraction of the subtropical transitional zone, the monitoring period was relatively short, and the number of sampled trees was limited. These constraints may affect the generalizability of our conclusions. It is necessary to include other forest types and sites across broader spatial and temporal scales in future studies. Second, the variation in nocturnal sap flow and its environmental controls were not considered in this study. Installing additional sensors at different trunk heights would provide a better understanding of vertical heterogeneity in nocturnal stem refilling. Finally, regional water demand was not quantified in this study. The development of water demand models based on transpiration data will optimize water allocation and improve ecosystem productivity, thereby strengthening forest management and policy decisions.

5. Conclusions

The results showed that sap flow in bamboo forests was higher than that in EBFs and fir forests across seasons. Nevertheless, the lower sapwood area of bamboos resulted in lower transpiration than EBFs and fir forests. Among the environmental drivers examined, Ra and VPD were identified as the most important factors regulating sap flow across the three forest types, although their relative contributions differed substantially. Specifically, Ra exerted a stronger influence on sap flow in bamboo forests than in EBFs and fir forests, while VPD had a comparatively greater effect on EBFs and fir forests. Furthermore, bamboo sap flow was more closely coupled with Ra, but exhibited the longest lag time in responding to VPD peaks compared to EBFs and fir forests. Overall, these findings provide an integrated understanding of plant water-use strategies in the subtropical transitional zone of China and offer valuable insights for enhancing forest water conservation and management in the future.