Field measurements were conducted for one entire year (2019) in the Zhangjiang Estuary Mangrove National Nature Reserve (23.9240°N, 117.4147°E) in the Fujian Province of China. The annual weather conditions in the estuary followed the patterns typical of a subtropical monsoon climate. Daily mean relative humidity (RH) ranged from 24.2% to 100% with a mean of 83.2%. Annual temperature extremes ranged from 9.35°C to 37.9°C with an annual mean air temperature of 23.7°C. The annual precipitation was 1714 mm with the peak in spring and summer. The estuary experiences a regular semidiurnal tide with an annual mean tidal range of 2.32 m. The sediment surface in mangrove forests is normally flooded with a maximum inundation depth of around 1.0 m. Salinity fluctuates seasonally, with measured minimum values of approximately 13-18 practical salinity units (PSU) during November-December and maximum values of 3-13 PSU during June-September. The mangrove forest naturally consists of three co-occurring species, Kandelia obovate (dominant), Avicennia marina, and Aegiceras corniculatum. Four trees of each species in close proximity to one another were randomly selected for sap flow monitoring and related measurements. The morphological features of the trees were measured before sap flow monitoring ( Table 1 ).

The canopy projection area (m²) was calculated by an elliptic equation after measuring the canopy length (C _l ) and canopy width (C _w ) of each tree (Wang et al., 2018):

To investigate plant physiological responses to the microclimate, a weather station was installed in the study area. A sensor of HOBO MX2301 (Onset Computer Corp., Bourne, MA, USA) was installed on an Eddy tower above the canopy to monitor air temperature (T_a, °C) and relative humidity (RH, %). Vapor pressure deficit (VPD, kPa) was calculated from T_a and RH. Photosynthetically active radiation (PAR) was measured from the data collected on an Eddy tower above the canopy using a PQS1 PAR Quantum sensor (Kipp & Zonen). Wind speed were measured using a 010C Wind speed Sensor (Met One Instruments, Inc) made above the mangrove canopy. Meteorological data were recorded every 30 s and stored at 10 min average in a data logger (CR1000 datalogger, Campbell Scientific, Inc., Logan, UT, USA). Tidal surface water salinity was also continuously measured and estimated from conductivity and temperature measurements using HOBO U24-002-C Conductivity Logger (Onset) deployed just above sediment surface. Raw 10-minute tidal data were consistently converted to half-hourly time series for data analysis.

Sap flow was measured using Granier-type thermal dissipation probes as per Zhao et al. (2018) ( Figure 1 ). In order not to damage the sample trees, we selected 4 trees of different diameter classes in each of the three species to establish the relationship between diameter at breast height and sapwood depth, and the operation method referred to the sapwood measurement method mentioned by Leng & Cao, 2020. Two different size probes (TDP10 with a probe length of 10 mm, and TDP30 with a length of 30 mm) were used to fit different sapwood widths. Each probe was installed at 1.3 m above the sediment surface and on the north-facing side of the stem to avoid the effects of sunlight. The probes were covered with reflective aluminum sheets and wrapped with a plastic tarp to protect them from solar irradiation and rain damage. Each probe consisted of two needles with thermocouples and a heater. The upper needle was continuously heated with a constant power of 0.2 W, while the lower needle stayed at the temperature of the unheated sapwood. The temperature difference between the two probes was measured at 30 s intervals, and the averaged values for each 10 min were recorded with a data logger (CR1000, Campbell Scientific Inc., Logan, UT, USA). The monitoring period was from January 1 to December 31, 2019. The sap flux density (SFD, g cm⁻² s⁻¹) was calculated based on the differential temperature (ΔT, °C) between the heated and reference probes following an empirical equation (Granier, 1987):

where ΔT is the measured differential temperature, and ΔT _m represents the ΔT when Sap flow density is zero.

The xylem wood of the three species in this study is diffuse porous (Lin et al., 1998). According to previous studies, the radial sap flow density is not significantly different in diffuse-porous wood (Phillips et al., 1996; Clearwater et al., 1999). Therefore, the daily sap flow (Q, kg d⁻¹) for each sampled tree was calculated as follows:

where A_s is the cross-sectional sapwood area of an individual tree, and t is 10 min because the data are being stored as 10 min averages.

To quantify the contribution of the two components (E _n and R _n) to the Q _n, the sap flow of three sample trees of each species was monitored at breast height and the base of the live crown ( Figure 1 ) (Daley & Phillips, 2006). In this analysis, the 24-h sum of sap flow at the base of the live crown was assumed to be equal to the 24-h sum of sap flow at breast height (Goldstein et al., 1998; Daley & Phillips, 2006). We normalized the 24-h sum of sap flow at the base of the live crown in each sample tree so that it equaled the 24-h sum of the sap flow at breast height. Specifically, 10-min sap flow at the base of the live crown multiply the ratio of the 24-h sum of sap flow at breast height to 24-h sum of sap flow at the base of the live crown. This is similar to the approach described by Goldstein et al. (1998). Previous studies showed lower water storage in the crown than in the stem (Waring et al., 1979), so we assumed the normalized sap flow at the base of the live crown could represent the total transpiration (Daley & Phillips, 2006). We calculated the difference between the sap flow rates at the stem breast height and mean transpiration at 10-min intervals. The positive values represented stem water refilling, whereas the negative values represented stem water storage was used by transpiration (Goldstein et al., 1998; Daley & Phillips, 2006; Zeppel et al., 2010).

Since their first appearance in 1970, point or ring dendrometers have been increasingly used for the measurement of stem diameter variation. These first observations suggest a close link between diurnal variation in plant water status and diurnal variation in stem thickness. Radial movement of water back and forth through the storage tissue in the stem leads to measurable changes in stem diameter (Barraclough et al., 2020). The stem diameter variations were measured with precision dendrometers ( Figure 1 ). To explore the radial movement of water in the stem, electronic point dendrometers (RRDIM10, Rainroot, China) were installed at the breast height of the main stem on two of the four sampled trees of each species. The resolution of the electronic point dendrometers was 10 μm. Data were obtained every 15 s and instruments were logged at 5 min intervals with a data logger (RR-1004, Rainroot, China). To get the largest diurnal variations of stem diameter during the same measuring period, stem diameter was measured during seven typical clear days in summer (from August 22 to August 28, 2019). We took the stem diameter value at 0:00 on a day after the equipment was stabilized as the stable point of reference.

Leaf transpiration (T_r ) and stomatal conductance (g_s ) from three trees of each species were measured on a typical sunny day in summer (August 26, 2019). Three to five mature, healthy, and fully expanded leaves on a south-facing branch were randomly selected from lower and upper crown locations for leaf gas exchange measurement using a portable photosynthesis system (Li-6400xt, Li-cor, USA) ( Figure 1 ). Measurements were taken at 2-hr intervals for 24 hours from 10:00 on the first day to 10:00 on the following morning. During the measurements, PAR, T_a, RH, and CO₂ concentrations in the leaf chamber were set to match ambient environmental conditions. The measurement chamber was not illuminated at night.

Night was defined as the period from 18:00 to 6:00 on the following day with PAR lower than 20 μmol m⁻² s⁻¹ (Daley & Phillips, 2006; Wang et al., 2011; Siddiq & Cao, 2018). We subdivided night into a first part (18:00 to 24:00) and second part (0:00 to 6:00). To calculate the amount of stem expansion and shrinkage during the night, we averaged stem diameter data every 10 minutes (D_i , where 1 ≤ i ≤ 72). When D_i+1 - D_i > 0, the stem was expanding and when D_i+1 - D_{i <} 0, the stem was shrinking. Thus, the amount of stem expansion and shrinkage during the nighttime was calculated as follows:

where D_i is the mean stem diameter at the i^th 10 minutes, and D_i+1 is the mean stem diameter value at the (i+1)^th 10 minutes. D _e is the cumulative amount of stem expansion and D _s is the cumulative amount of stem shrinkage.

The vapor pressure deficit (VPD, kPa) was determined using measured atmospheric temperature (T_a) and relative humidity (RH), and the calculation formula is as follows (Buck, 1981):

Since the sap flow data for the entire year did not have a normal distribution, we used the nonparametric Kruskal-Wallis test to analyze the difference between Q, Q _n, and the contribution of Q _n to Q (Q _n/Q) among the three mangrove species. An independent sample t-test was conducted to test whether there were significant differences between the mean cumulative amount of stem expansion and the mean cumulative amount of stem shrinkage for each species. The general linear regression and curvilinear model regression were fitted to identify the relationships between (1) salinity and sap flow (daily sap flow and nocturnal sap flow), (2) sap flow density (SFD) (daytime SFD, nocturnal SFD, SFD in the first part of the night and SFD in the second part of the night) and VPD, (3) wind speed sap flow (daily sap flow and nocturnal sap flow), (4) Q _n and Q. At the same time, the analysis of Covariance (ANCOVA) was used to test whether there were significant differences in the slopes of fitted linear relationships for different species within the same salinity environment. The Gompertz models (three parameters) was used to test the relationship between the Q _n/Q of different tree species and latitude. The Q _n/Q data (n=56) for different tree species were based on data provided by Forster (2014). The significance level was set at P< 0.05. All data were analyzed using SPSS 25.0 (SPSS, Chicago, Illinois, USA).

For the species Kandelia obovata, Avicennia marina, and Aegiceras corniculatum, the annual mean Q per sapwood area ranged from 162.0 Kg m⁻² d⁻¹ to 348.1 Kg m⁻² d⁻¹ ( Figure 2A ). The magnitude of the annual mean Q _n per sapwood area varied from 16.9 ± 0.7 Kg m⁻² d⁻¹ to 35.2 ± 1.1 Kg m⁻² d⁻¹ ( Figure 2B ). The annual mean Q _n/Q was 24.0 ± 0.6%, 5.5 ± 0.2%, and 12.1 ± 0.4% for Kandelia obovata, Avicennia marina, and Aegiceras corniculatum, respectively. These values were up to 45.6%, 17.9% and 32.0%, respectively, higher on some days ( Figure 2C ). For Q, Q _n, and Q _n/Q, significant differences were observed among the three species. The Q of Avicennia marina was significantly higher than the Q of the other two species, while the Q _n of Avicennia marina was significant lower than the Q _n of the other two species ( Figures 2A–C ).

A substantial amount of SFD and leaf T_r was detected during the night. For all three species, SFD, leaf T_r and g_s decreased sharply after 16:00 and then remained steady until sunrise on the following day ( Figures 3A, C, E ). During the night, Kandelia obovata and Aegiceras corniculatum maintained low g_s , less than 0.02 mol H₂O m⁻² s⁻¹. These values decreased to about 9% and 6%, respectively, of day values. T_r was also low, below 0.30 mmol H₂O m⁻² s⁻¹ ( Figures 3C, E ). There was no significant difference in g_s and T_r between Kandelia obovata and Aegiceras corniculatum ( Figures 3D, F ). Although the nocturnal SFD of the Avicennia marina was significantly lower than that of the other two species, the g_s and T_r were significantly higher ( Figures 3B, D, F ). The mean nocturnal g_s of Avicennia marina was 0.05 mol H₂O m⁻² s⁻¹, and the value decreased to 16% of day value. The mean nocturnal T_r of Avicennia marina was 1.09 mmol H₂O m⁻² s⁻¹ ( Figures 3C, E ).

Early in the morning (about 6:00-10:00), sap flow at breast height was lower than sap flow at the base of the live crown, indicating that the three mangrove species utilized water stored in the stem for transpiration at sunrise ( Figures 4A–C ). Stem water refilling and stem water storage withdrawal during the night occurred in all three species, but stem water refilling occurred most of the night for Kandelia obovata and Aegiceras corniculatum, while Avicennia marina experienced mainly stem water storage withdrawal ( Figures 4D–F ). The contribution of E _n and R _n to Q _n in Kandelia obovata was 43.7% and 56.3%, respectively. Similar to Kandelia obovata, the contribution of E _n and R _n to Q _n in Aegiceras corniculatum was 32.7% and 67.3%, respectively. In contrast, the Q _n of Avicennia marina consisted mainly of E _n and the E _n exceeded the Q _n (E _n accounted for 103% of Q _n). In Avicennia marina, stem water refilling occurred during the day and stem water storage was partially depleted at night during the nocturnal transpiration process ( Figure 4E ).

For Kandelia obovata and Aegiceras corniculatum, stem diameter decreased in the morning and started to increase again in the afternoon. Although the increasing trend of stem diameter in Kandelia obovata and Aegiceras corniculatum at night was weaker than the increasing trend before the sunset, the stem diameter of these two species remained in an expanded state at night ( Figures 4G, I ). However, the stem diameter changes of Avicennia marina showed different diurnal patterns from the former two species, with an increase during the morning and a decrease in the afternoon, which follow the variation of the SFD ( Figure 4H ). Unlike Kandelia obovata and Aegiceras corniculatum, little the cumulative amount of stem expansion can be detected during certain nighttime periods in Avicennia marina. By the independent sample t-test analysis, the cumulative amount of stem expansion was significantly higher than the cumulative amount of stem shrinkage in Kandelia obovata and Aegiceras corniculatum. However, the cumulative amount of stem expansion was significantly lower than the cumulative amount of stem shrinkage in Avicennia marina ( Figure 5 ).

There was a significant negative linear relationship between the salinity and the daily sap flow of the three species ( Figure 6 ). The slopes of fitted linear relationships between salinity and daily sap flow of Kandelia obovata was significantly lower than that of Aegiceras corniculatum and Avicennia marina. When the ambient salinity increased from 5 PSU to 15 PSU, the daily sap flow of Kandelia obovata, Avicennia marina and Aegiceras corniculatum decreased to 26.3%, 48.3% and 50.6%, respectively. During the night, different species showed different relationship between salinity and sap flow. There was a positive linear relationship between salinity and sap flow of Kandelia obovata while a negative linear relationship between salinity and sap flow of Avicennia marina ( Figures 6A, B ). For Aegiceras corniculatum, there was a non-linear relationship between salinity and sap flow ( Figure 6C ).

As can be seen from Figure S1 , wind speed had no significant effect on the daily sap flow and nocturnal sap flow. The relationships between SFD and VPD differed between day and night. During the day, there was an exponential relationship between SFD and VPD (SFD = i + a · (1 - e^-bVPD)), and the VPD explained 57.0%–60.4% SFD variations in three mangrove species. During the night, a weak generally linear relationship (SFD = i + a · VPD) existed ( Figure 7 ). Furthermore, the correlation coefficient (R²) differed among the three mangrove species ( Figure 7 ). For Avicennia marina, the VPD explained 27.3% of the nocturnal SFD variation, whereas for Kandelia obovata and Aegiceras corniculatum, it explained only 9.0% and 12.6%, respectively ( Figures 7A–C ). SFD presented a sharp decreasing trend after sunset in Kandelia obovata and Avicennia marina, which was similar to the change of VPD. The SFD of Kandelia obovata showed a slow declining trend after 20:00 and maintained a high SFD until sunrise of the next day. After a sharp decline, the SFD of Avicennia marina stabilized and remained low in the second part of the night (N _s). In contrast, the SFD of Aegiceras corniculatum showed an increasing trend after a brief decline at night, and then maintained a steady-state with high SFD in N _s. The variation in nocturnal SFD of Aegiceras corniculatum was inconsistent with the variation of VPD ( Figure S2 ). For Kandelia obovata and Avicennia marina, the responses of the nocturnal SFD to VPD occurred in the first part of the night (N _f). VPD explained 21.1% and 50.3% of the nocturnal SFD variation, respectively. Interestingly, there was a significant negative correlation between the nocturnal SFD and VPD of Aegiceras corniculatum ( Table 2 ).

The general linear regression analysis showed a significant and positive relationship between Q _n and Q in Kandelia obovata (y = 4.718 + 0.215x, R^2 = 0.539, P< 0.0001) and Aegiceras corniculatum (y = 3.543 + 0.096x, R^2 = 0.573, P< 0.0001). These results indicate that the Q _n increased with the increasing water lost from transpiration during the day. However, this linear relationship was not found between Q _n and Q of Avicennia marina (R^2 = 0.062 P = 0.056) ( Figure 8 ), and the Q _n of Avicennia marina was almost not affected by the water lost from transpiration during the day.

The curvilinear model regression analysis showed a significant and positive relationship between the Q _n/Q of different tree species and latitudes (y = 6.203 · 39.518 ^ e^{(-0.0634 x )}, R^2 = 0.693, P< 0.0001) ( Figure 9 ).

Previous studies confirmed that Q _n occurs in species across all vegetation types and biomes, making a significant contribution to Q (Daley & Phillips, 2006; Zeppel et al., 2010; Forster, 2014). By analyzing 98 tree species from different biomes, Forster (2014) reported that Q _n/Q was 7.6%, 16.5%, and 29.1% for forests in warm temperate, tropical, and equatorial, respectively. Our results demonstrated that Q _n occurred in each of the three mangrove species during most of the night. Q _n/Q ranged from 5.5% to 24.0% of these subtropical mangrove species ( Figure 2C ). In the dataset of Forster (2014), a significant negative relationship (y = 6.230 · 39.518 ^ e^{(−0.0634 x )}) was observed between Q _n/Q and latitude ( Figure 9 ). The highest Q _n/Q occurred in equatorial and tropical biomes, under conditions of high atmospheric evaporative demand, suggesting that tree species growing at lower latitudes would have higher Q _n/Q ( Figure 9 ). We predict that mangrove species growing in equatorial and tropical regions have even greater Q _n, which has significant implications for the water balance between soil and plant and hydrology.

This study is one of the few to quantify inter-species differences of Q _n among the co-occurring species. Previous studies showed that Q _n rates varied among plant species because of different adaptive strategies (Phillips et al., 2010; Zeppel et al., 2010). In this study, significant differences were observed in the Q _n/Q between species ( Figure 2C ), and this result supports the conclusion that different adaptive strategies lead to Q _n/Q diversity. The relatively high average Q _n/Q of Kandelia obovata and Aegiceras corniculatum in this study, indicating that the two species were highly dependent on Q _n and Q _n was important for maintaining water balance in these species. In contrast, Avicennia marina had little dependence on Q _n. The reason for the difference in Q _n/Q between species may be that stem water refilling of Kandelia obovata and Aegiceras corniculatum occurred mainly after sunset while the stem water refilling of Avicennia marina occurred during the daytime ( Figures 4D, E, F ). Due to the special habitat of mangrove species, we had to consider the effect of salinity changes on sap flow. For the Q of the three mangrove species, high salt conditions contribute to more conservative water use in mangroves ( Figure 6 ). During the nighttime, high salt conditions drove high Q _n of Kandelia obovata and Aegiceras corniculatum while the high salt conditions inhibited the Q _n of Avicennia marina. The high salt conditions drove high Q _n may be related to mangrove species overcoming the high osmotic potential of saline environments through enhance both osmotic and hydrostatic potential in their xylem sap during the nighttime. The various responses of different species to salinity at night were also a reason for their different Q _n/Q.

Q _n was associated with two physiological processes, E _n and R _n. Direct evidence for E _n comes from leaf gas exchange measurements (Daley & Phillips, 2006). A previous study showed that weak stomatal conductance at night might result from slight E _n (Rosado et al., 2012). However, some reported not low levels of Q _n with incomplete stomatal closure during the night in many plant species (Snyder et al., 2003; Dawson et al., 2007). For Kandelia obovata and Aegiceras corniculatum, the SFD at night remained at a relatively high level with a low g_s and T_r , indicating that both species occurred weak E _n and had strict control over their stomata to reduce the water loss at night. In contrast, the g_s and T_r of Avicennia marina at night were significantly higher with lower SFD than the other two species ( Figures 3B, D, F ), indicating that a large part of the Q _n of Avicennia marina was dissipated into the atmosphere through the stomata. These different controls of leaf stoma in co-occurring species of the same environment were also reported in a mixed deciduous forest (Daley & Phillips, 2006).

The water-use pattern in which trees use stem stored water during the day to buffer water deficits and refill the stem water during the night time when evaporative demand is low, has been observed in different tree species (Goldstein et al., 1998; Phillips et al., 2003). Here, we found that in Kandelia obovata and Aegiceras corniculatum, R _n was a major component (over 60%) of Q _n. The high ratio of R _n to Q _n indicates a strong reliance of these plants on stem storage water, which helps them to maintain a balance between water supply and water demand during periods with high transpiration demands (Scholz et al., 2011; McCulloh et al., 2014). High salinity conditions promote more conserved daytime water use in Kandelia obovata and Aegiceras corniculatum and drive higher levels of Q _n, which maintains water balance in mangroves in highly negative water potential environments in the form of R _n. Therefore, we deemed Q _n to be an effective water-use strategy to avoid physiological dehydration and adapt to high salt environments in some mangrove species. Unlike Kandelia obovata and Aegiceras corniculatum, the Q _n of Avicennia marina was mainly used for E _n, and the stored water was released from the stem to supply the water needed for E _n, indicating that the water consumption during transpiration at night was higher than the water absorbed by the roots from the soil. This leads to the question of how Avicennia marina maintained water balance at night after experiencing high water consumption by transpiration during the day. Different from Kandelia obovata and Aegiceras corniculatum, Avicennia marina is able to maintain water refilling while transpiring and uptake water from the saline soil more efficiently during daytime ( Figure 4E ). This process enables the water storage tissue to retain sufficient water after the water loss during the day and beneficial for maintaining stem water balance at night. Furthermore, previous studies showed that Avicennia marina had unique wood anatomy in which successive or multiple cambia resulted in consecutive bands of xylem interspersed with internal secondary phloem strands connected by parenchyma tissue (Schmitz et al., 2008; Robert et al., 2011). This anatomy is conducive to internal water storage and security and is deemed to be an adaptive trait to deal with physiological drought (Robert et al., 2011). At the same time, our results showed that Avicennia marina would limit Q _n under high salt conditions to reduce water loss at night, which was also a strategy to overcome physiological dehydration. In addition, Coopman et al. (2021) found that Avicennia marina can absorb atmospheric moisture through deliquescence of salt on leaf surfaces, driving top-down sap flow and replenishes the water needs of the stem. This may be one of the mechanisms for maintaining the water balance of Avicennia marina at night. Due to technical limitations, however, the thermal dissipation probe could not detect this part of the reverse flow. The neglect of reverse flow leads to an underestimation of its nocturnal sap flow and this will be a limitation of the experiment.

The changing patterns of stem diameter indirectly support our findings. From hourly to daily scale, the fluctuations in stem diameter reflected changes in the elastic tissue water content (Zweifel et al., 2001; Pfautsch et al., 2015). The cumulative amount of stem expansion was significantly higher than the cumulative amount of stem shrinkage in Kandelia obovata and Aegiceras corniculatum ( Figure 5 ), which indicated that R _n occurred more frequently than stem water storage withdrawal at night. In contrast, the cumulative amount of stem expansion was significantly lower than the cumulative amount of stem shrinkage in Avicennia marina ( Figure 5 ), indicating that relatively large of the stored water in the stems was released into the xylem flow to participate in the E _n. The unusual stem variation pattern of Avicennia marina has also been reported by Barraclough et al. (2020).

The partitioning method used in this study had its limitations because the sap flow in terminal branches could not be measured, the same as the previous study (Daley & Phillips, 2006). For Kandelia obovata and Aegiceras corniculatum, the nocturnal g_s and T_r were low ( Figures 3D, F ), and SFD was low sensitive to VPD at night ( Figure 7 , Table 2 ), leading to low E _n. Therefore, the contribution of E _n to Q _n may have been overestimated in these two species. For Avicennia marina, the sap flow at the base of the live crown exceeded which at breast height ( Figures 4B, E ), indicating that part of the stem storage water was used for E _n, while there was no R _n in the terminal branches. Thus, we can reliably estimate the contribution of E _n to Q _n in Avicennia marina.

In the water stress environment, mangrove species developed water-saving mechanisms to reduce water loss via transportation during the day (Ball, 1988). Our results indicated that a significant water compensation mechanism used by Kandelia obovata and Aegiceras corniculatum operated at night. Thus the more water was lost owing to diurnal transpiration, the higher the Q _n was at night to make up for the lost water. This newly found stratagem in mangrove species provides evidence of a nocturnal water recharging mechanism in addition to their diurnal water-saving strategies. We suggest that mangrove species with nocturnal water recharging mechanisms and tight control of stomata will be better adapted to future environmental conditions of enhanced atmospheric evaporation demand than species without these capabilities. As for the Avicennia marina, its unique daytime hydration and timely restriction of water loss under high salt conditions provides its ability to survive under conditions of limited water.

Q _n is driven by environmental factors and biotic factors, separately or synergistically. Previous studies suggested that VPD was the main environmental factor driving E _n because VPD between the leaf surface and the atmosphere provided conditions for stomata to open or partially open at night (Caird et al., 2007; Dawson et al., 2007; Zeppel et al., 2010). Wind speed affected the moisture movement around plant canopy (Fricke, 2019) and thus influenced on nocturnal water loss (Chen et al., 2020), which also could not be ignored. In this study, we found that wind speed was not a driving force in sap flow, whether during the daytime or nighttime, which is consistent with the result that wind speed had little effects on the nocturnal water loss by Wang et al., 2018 and Wu et al., 2020. Rosado et al. (2012) found strong positive relationships between VPD and Q _n in woody species in rain forest and indicated that the E _n was driven by atmospheric demand for water. Wang et al. (2018) concluded that VPD had little effect on Q _n because the correlation coefficient was low (R^2 = 0.100, P< 0.0001). Similarly, our findings also showed that VPD had only a slight effect on the Q _n of Kandelia obovata and Aegiceras corniculatum, indicating that VPD was not the main driver of Q _n. In addition, there was a significant positive correlation between Q and Q _n in Kandelia obovata and Aegiceras corniculatum. Based on the principle of water balance, Q _n of these two species was thus mainly driven by the demands of stem water refilling after diurnal water depletion. In addition, high salinity conditions were also a driver of Q _n in Kandelia obovata and Aegiceras corniculatum, and the high salinity environment prompts both of the species to refilling more water at night. However, the Q _n of Avicennia marina occurred mainly in the early night and was driven by VPD (R^2 = 50.3%) ( Figure S2 , Table 2 ). Our results suggest that Q _n is driven by VPD, high salinity conditions or the demand for stem rehydration, separately or synergistically. The main driving force differs among species.