The SAPFLUXNET database provides harmonized sap flow data from forest stands across 202 sites across the globe, covering the period 1995 to 2018. The sites are primarily located in Europe and the USA, with additional sites in Australia and South Africa, and a few in the tropics and high latitude regions. Each site offers half hourly or hourly time series, with an average duration of three years (Fig. ).

The database includes sap flow data expressed at different levels – plant, per sapwood area, or per leaf area – which are derived from heat dissipation, heat pulse and heat balance methods . Sensors are typically inserted into the sapwood at the trunk, providing a sap flux density, i.e., the rate of flow per unit sapwood area. Plant-level data can then be calculated by multiplying this density by the total sapwood area of the tree. To obtain leaf-level data, the plant-level sap flow is divided by the total leaf area, yielding an estimate of transpiration at the leaf scale. Not all site specific datasets include measurements for all three levels simultaneously. In this study, we selected only sites with plant-level data because it represents the largest scale of sap flow measurement, making it particularly useful for understanding whole-plant water dynamics .

In addition to sap flow (SF), the database includes observations of hydrometeorological variables such as photosynthetically active portion of incoming radiation (PAR, quantified as PPFD), 2 m surface air temperature (TAir), precipitation (PRECIP), volumetric top soil moisture (0–30 cm, TSM) or vapour pressure deficit (VPD). With these ancillary variables, the SAPFLUXNET database allows the study of rapid tree responses to environmental conditions and helps to bridge the gap between ecosystem flux networks and remote sensing.

For the broader analysis of sub-daily sap flow dynamics and response to climatic drivers, we identified 15 unique sites where plant-level SF, as well as VPD, TAir and TSM are available. Specifically, we selected sites where days with these data overlapped for at least 80 % of the growing season (defined based on day length >12 h and TAir >5 °C) across two or more years. These 15 sites represent less than 15 % of all sites in the SAPFLUXNET database. Figure  shows how the mean temperature and precipitation varies across the 15 study sites. Furthermore, site-level information on mean Diameter at Breast Height, Leaf Area Index, soil texture, tree species, biome, plant group, treatment and site-specific publications are provided in Tables  and .

Sap flow (SF) represents the rate at which a volume of water moves through the tree per second and can be considered a proxy for transpiration and, thus, often serves as an indicator of carbon assimilation . SF has a clear diurnal cycle, driven by variations in radiation and atmospheric water demand, and regulated by plant functioning. Here, we analyse the relationship between sub-daily SF and VPD, the partial pressure deficit relative to saturated conditions, an indicator for atmospheric dryness and water demand (Fig. ). We then analyse how this relationship varies across the growing season and over multiple years, and how it relates to other environmental variables, including temperature, soil moisture and radiation, as described below.

For each of the 15 sites, we aggregated half-hourly VPD and average SF data across all measured trees into hourly values per site. Since some sites only provided hourly data, this step ensured consistency across sites. We then analyse the diurnal cycle of paired SF–VPD values. As discussed in the introduction, we hypothesise that the sub-daily dynamics can be used to understand plant stress, and derive two metrics at a daily basis to characterize the diurnal SF–VPD relationship: the magnitude of the morning slope of the SF–VPD curve, referred to as SLOPE, and the area of the hysteresis loop , referred to as AREA. Other metrics were tested in preliminary analysis, such as the magnitude of the afternoon SF–VPD slope, or the temporal lags between peak SF and peak VPD (analogous to e.g., ), but we found the information in these metrics to be strongly correlated, and thus redundant, with the two presented here.

The SLOPE is calculated as the coefficient from a linear regression of VPD and SF between the times of minimum VPD and maximum SF. AREA corresponds to the area enclosed by the hysteresis curve of SF and VPD during the diurnal cycle, measured as the area of a polygon formed by the observation points. The variable SLOPE shows outliers (upper 5 %) and sometimes negative values, particularly when VPD gradients throughout the day are too small, e.g., in winter. These values were excluded from the analysis, resulting in a filtered SLOPE. In order to make the values comparable across sites, we standardize the filtered SLOPE values with the median and scale the AREA to 0–1 per site. These quantities are indicated as sSLOPE and nAREA.

We analyse the seasonal variability of the two metrics for the selected sites along with the seasonal changes in PPFD used as a measure of solar radiation, TAir and TSM for the three focal sites. For this, we first evaluate the seasonal cycle of each variable per site and in a second step correlate the seasonal cylces of SLOPE and AREA to each of the hydrometeorological drivers. To reduce redundancy, when a site was represented by multiple plots, we averaged the correlation coefficients and p-values across plots to obtain a single site-level estimate.

Since we want to evaluate plant functioning and water fluxes under environmental stress, we analyse combined extremes of TAir and TSM (upper and lower 20 % of all days within the growing season at each site). We obtain the following four extreme conditions: cold and wet, hot&wet, hot&dry and cold&dry. Within each of the remaining clusters, we calculate the values of sSLOPE and nAREA and the hysteresis of the mean diurnal cycle. In order to get a more general picture of diurnal hysteresis during extreme conditions, we evaluated the metrics at all 15 sites (Fig. ). In order to make the sites comparable, we evaluated sSLOPE and nAREA at anomalies of TAir and TSM (upper and lower 20 %, after subtraction of the mean).

Finally, to evaluate the feasibility of large-scale monitoring of plant stress, for example, through remote sensing of VWC (see Fig. ), we investigate how these two sub-daily metrics depend on the temporal resolution of the sub-daily observations. For this, we tested four scenarios of sub-daily sampling: 3-hourly, starting at mid-night (8 times per day, 8TPD), 6-hourly, starting at mid-night (4TPD) and 6-hourly with mid-night or noon left out (3TPDday, 3TPDnight). We then derived the curves of SF–VPD at the coarser temporal resolution and calculated the corresponding sub-daily metrics. To evaluate whether the metrics could still be reliably estimated at coarser temporal resolution, we determined the corresponding coefficients of determination (Pearson R2) between the metrics derived from the resampled time series and the metrics derived from the hourly data for the complete data record at each site.

Next, we compare diurnal SF-VPD dynamics and the associated AREA and SLOPE values under varying hydrometeorological conditions, classified according to combinations of upper and lower 20 % percentiles of TAir and TSM (Fig. , and see Section ). This procedure yields four clusters of moderate to extreme cold and wet, hot&wet, hot&dry and cold&dry conditions. The clusters and their corresponding mean diurnal cycles of absolute SF-VPD are illustrated for the three focal sites in Fig. . Given that we selected the three sites based on their different climate zones, the values of TAir and TSM differ strongly in their absolute ranges for the different clusters as does absolute SF, given the differences in evaporative demand, but also differences in DBH and leaf area index (Table ). Nevertheless, this clustering approach allows us to compare relative responses of SF-VPD dynamics to different to hydrometeorological conditions and evaluate how the SLOPE and AREA metrics reflect these responses.

Across the three focal sites, and despite their different background climate, sub-daily dynamics exhibits similar responses to the different hydrometeorological clusters (Fig. ). Hot days, generally reaching higher VPD values, are characterized by higher hysteresis (and thus higher AREA values) than cold days at all three sites. However, hot and wet days tend to be associated with steeper morning and afternoon slopes compared to hot and dry days, with the latter associated with a lower morning SLOPE, which is indicative of suppression of SF when soil moisture is most likely to be a limiting factor, even under very high VPD.

Despite these similarities, we also note differences between the three sites (Fig. ). FRA_PUE shows the clearest separation between hydrometeorological regimes (panel a and d): cold–wet days are associated with low values of AREA with high SLOPE values, hot–wet days show high AREA values with reduced SLOPE and the highest peak SF, and hot–dry days reveal negative SLOPE anomalies with lower AREA than hot–wet days but still higher than cold–wet days. At this site, nighttime VPD minima do not return to zero in hot days, although SF can still cease. GUF_GUY_GUY (panel b and e) also shows contrasting dynamics for high- and low-TSM regimes, although it is worth noting that dry days are still considerably wetter than dry days in FRA_PUE and RUS_POG_VAR. SLOPE is higher for the wet regimes than dry ones, while AREA is higher in hot days than cold days, although the difference of hot versus cold is only 5 °C at this site. At hot-wet days, peak SF exceeds all other regimes by more than a factor of two. In RUS_POG_VAR (panel c and f), hot conditions are associated with higher AREA and dry days are associated with lower SLOPE, although the variability observed in SF–VPD hysteresis between relative hydrometeorological extremes at this site is less marked than at FRA_PUE and GUF_GUY_GUY.

Based on these results, the AREA and SLOPE descriptions appear to reflect characteristic hydrometeorological regimes across the three focal sites, allowing to distinguish distinct types of responses to environmental stressors. To evaluate whether these differences can be generalized across sites, we analyse the distributions of sSLOPE and nAREA at all 15 sites during the four types of hydrometeorological regimes (Fig. ).

For the standardized SLOPE (sSLOPE, panel a), distributions show significant differences between wet and dry hydrometeorological regimes: for both cold and hot days with high TSM, the distributions of sSLOPE show a large spread with positive median values (median =0.12), while for days with low TSM, the distributions are strongly skewed towards negative sSLOPE values, with median values of -0.16 and -0.05 for hot and dry and cold and dry days, respectively.

In the case of the normalized values of AREA (nAREA, panel b), we find stronger differences in the between hot and cold hydrometeorological regimes: cold days show a predominance of very small values, and very similar distributions for cold days whether the soil is wet or dry, with comparable median values (0.03) and a similar tail distribution. By contrast, hot days show very long tails and median values shifted towards higher values (0.11 and 0.35 respectively), which is particularly pronounced for the regime corresponding to hot conditions and high TSM.

While in situ sensors can provide continuous sap flow data, we consider the potential to estimate these descriptors of the diurnal cycle using temporally sparse data (see Sects.  and ). In particular, as sap flow is connected to changes in water storage, which can be estimated using microwave remote sensing, we examine the degree to which the SLOPE and AREA can be estimated for several acquisition strategies with different numbers of observations and acquisition times.

Figure b) shows the distribution of the coefficients of determination (Pearson R2) obtained by linear correlation of the daily values of the SLOPE and AREA metrics derived from the hourly data, and those derived from sparser sub-daily sampling frequencies and the values for each site. We tested frequencies ranging from 8 (3-hourly), 4 (6-hourly), and 3 (6-hourly with either mid-night or mid-day left out) times per day (TPD) starting at 6am, as illustrated in Fig. a for FRA_PUE.

The AREA descriptor calculated with coarser temporal sampling agrees well with the AREA estimated using hourly data, though the agreement obviously declines when fewer data are used in the estimation. With 8TPD, the AREA values are well preserved, with a median annual Pearson R2 of 0.95 and an interquartile range of 0.9–0.97. For 4TPD, the relationships between AREA values derived from sparse sampling and the higher temporal resolution become weaker and less well constrained, as given by the median correlation of 0.76 and interquartile range of 0.67–0.83. There is very little further information loss in AREA estimation when sampling at 3TPD when there is an observation at noon (3TPDday, median R2: 0.74; IQR: 0.66–0.83). However, there is significant degradation in the relationship for 3TPDnight, where the observation at noon is missing (median R2: 0.30; IQR: 0.22–0.45)

Compared to AREA, the estimates of SLOPE are more sensitive to sampling frequency and exhibit greater variability across the sites, especially at 8TPD. At 8TPD, the correlation with the hourly sampling shows a median value of 0.5 with interquartile range of 0.26–0.82. Similarly to the performance of AREA, the agreement of SLOPE calculated with lower sampling rates deteriorates with decreasing frequency: with median correlations of 0.16, 0.01 and 0.14 for 4TPD, 3TPDnight and 3TPDday, respectively. Furthermore, the correlations of SLOPE for 4TPD and 3TPDday show large spread, with interquartile ranges of 0.03–0.55 and 0.02–0.37, respectively. In general, AREA is underestimated at lower sampling frequencies.

The diurnal cycles of SF–VPD under sparse sampling, shown in Fig. a), illustrate the poor performance of 3TPDnight in estimating SLOPE at least for FRA_PUE. 3TPDnight cannot accurately capture maximum SF (see Fig. a), even though it does effectively capture minimum VPD. In contrast, 4TPD and 3TPDday slightly overestimate minimum VPD but still yield better SLOPE estimates, with 3TPDday performing best among the reduced sample rates.

With the sample times considered, maximum VPD tends to be underestimated, reducing accuracy in metrics relying on its diurnal extremes at this site. Nevertheless the sample rates of 4TPD and 3TPDday still capture the metrics similarly at all sites, with better results of AREA at 4TPD and better results for SLOPE at 3TPDday.

Sap flow (SF) alone is not a reliable indicator of vegetation response to environmental conditions, because absolute rates depend on tree size, species, and stand conditions . This is seen by the large differences in absolute SF across the three selected sites (Fig. ). Instead, previous studies have shown that diurnal SF–VPD dynamics provide clearer insight into regulation processes . Therefore, the diurnal cycle of SF-VPD provides diagnostic indicators that consistently discriminate the drivers of extreme stress across all fifteen forested sites. Here we propose that two such indicators can be used in combination to identify characteristic responses to changing environmental conditions, and particularly hydrometeorological extremes. Our results show that the morning SLOPE and the AREA of the diurnal cycle of SF-VPD vary across seasons and systematically change from cold–wet to hot–dry conditions across most sites. Our results indicate that the combination of higher climatic demand and lower soil water availability reduce the coupling of SF to VPD.

We have shown, that cold and wet conditions correspond to high SLOPE and low AREA of the diurnal cycle (dashed lines in Fig. , blue curves in Fig. ), as SF aligns tightly with VPD and hysteresis is minimal. This dynamics indicates a strong coupling of SF and VPD, i.e., under conditions of limited demand and high supply, trees exhibit weak stomatal regulation, prioritizing carbon uptake over preventing water losses. By contrast, in hot and dry conditions (dotted lines in Fig. , red curves in Fig. ), AREA increases and the SLOPE declines, reflecting the onset of regulation of SF before peak VPD and stagnation of SF during periods of high atmospheric demand. These patterns may result from increased stomatal regulation or limitations to plant hydraulic water supply.

Despite these general patterns, the variations of SLOPE and AREA and their relationships to hydrometeorological drivers differ among climate zones (Fig. ). AREA shows consistently positive correlations with TAir and PPFD across sites, indicating a robust response to atmospheric energy availability that is largely independent of climate regime or plant group. In contrast, correlations between AREA and soil moisture are weaker and more variable, suggesting that AREA integrates canopy responses that are only indirectly constrained by short-term water availability.

The relationships between SLOPE and hydrometeorological drivers are markedly more site-dependent. Sites with pronounced dry seasons show positive correlations between SLOPE and soil moisture and negative correlations with TAir and PPFD, consistent with water-limited canopy dynamics. Conversely, cold and temperate sites, as well as semi-arid sites with irrigation, exhibit negative or no correlations between SLOPE and soil moisture and positive or neutral relationships with energy-related drivers, suggesting that irrigation at otherwise water-limited sites may dampen or mask typical stress-related responses. Overall, AREA behaves as a broadly comparable metric across ecosystems, primarily reflecting energy limitation, whereas SLOPE is strongly shaped by local hydrometeorological regimes and management. Together, both metrics capture differences in the strength of biosphere–atmosphere coupling.

The full exploration of the complexity of the correlations between metrics and climate drivers is beyond the scope of this analysis. Furthermore, different responses of the metrics were found across sites, but still they help to reveal whether down-regulation occurs under increasing atmospheric demand. The underlying cause of down-regulation, be it stress, adaptive behavior, or species-specific traits—must be interpreted in context. Next to site-specific differences, the still short SAPFLUXNET time series limit the calibration of the metrics against long-term means. Without such references, it remains uncertain whether regulation corresponds to stress or adaptive behavior. Importantly, a lack of regulation does not necessarily indicate optimal conditions; it may instead signal failure to respond to stress or overshooting, risking excessive water loss and hydraulic damage .

The metrics derived from the SF–VPD relationship thus can serve as early indicators of regulation patterns and water use strategies, some of which can be attributed to stress . While SLOPE measures the strength of the coupling between SF and atmospheric demand, the AREA is an indicator of regulation timing in relationship to the atmospheric demand. Combining SLOPE and AREA helps to reduce ambiguity, as in the diagnostic case of low SLOPE with high AREA, which can be used as an indicator for strong regulation. Still, limitations remain: coarse temporal resolution may miss the identification of downregulation from breakpoint patterns in morning SF, and lack of regulation over consecutive days may reflect either risky strategies or optimal conditions. While interpreting these metrics requires integration with local water balance, species traits, and climatic drivers, they can serve as a useful proxy to detect changes in vegetation functioning in response to environmental conditions.

More detailed assessment of the hydraulic hysteresis reveals novel insights into the variation in physiological dynamics across sites, with implications for understanding their adaptation to existing climate and potential vulnerability to future climate extremes.

At the Mediterranean site (FRA_PUE), stomatal regulation is triggered when atmospheric and soil moisture drought coincide (Fig. ), reflecting interactions between water and temperature, through VPD. The sensitivity of AREA and SLOPE to both atmospheric demand and TSM is shown by the sign and strength of the correlation with all hydrometeorological drivers (Fig. ). Furthermore, the negative correlation of AREA with TSM indicates that stomatal closure is a primary response when topsoil water becomes a limiting factor. Increasing AREA signals moderate regulation at high temperature, while declining SLOPE throughout the day represents strong conservative water use under severe diurnal drought and heat stress, triggering stomatal closure (Figs. , , ). A strong ability to close stomata is also indicated, when SF is absent at night at non-zero VPD. This site also has the widest range in TAir and TSM, which are as well negatively correlated, suggesting that plants frequently operate near or at their physiological limits. While the adaptation strategies have traditionally allowed these species to withstand seasonal drought, the additional compounding stress from increasing heat and VPD leaves them vulnerable to conditions that exceed historical variability . Thus, the site’s long-term water conservation strategies, although beneficial in historical contexts, now signal a heightened exposure to future extreme events. Empirical evidence of recent mortality supports the conclusion that these adaptive strategies are being overcome under more severe and compound climatic stresses .

In contrast, stomatal regulation at the tropical site is less sensitive to heat. Here, the SF–VPD dynamics is mostly controlled by radiation and soil moisture (Fig. ), where seasonal drought results in the decreasing of the morning SLOPE. AREA primarily reflects afternoon or nighttime stagnation of SF, which can also be a sign for high TSM enabling the trees to maintain SF during the night (Fig. ). Previous studies have shown the strong importance of VPD for SF dynamics in these regions , here we show that radiation still contributes to modulate the sensitivity of SF to VPD, as seen by the strong correlations between AREA and SLOPE with PPFD and TSM (Fig. ). At this site, tree height and hydraulic structure likely contribute to the high absolute SF rates, supported by deep rooting and access to water beyond topsoil layers . The results highlight an ecosystem optimized for heat under conditions of high water supply. Trees at this site are capable of transpiring large volumes of water to maintain levels of high stomatal conductance and photosynthesis.

The Russian site displays an inverted response pattern compared to the warm sites: SLOPE and AREA increase and decrease together at the beginning of the season until they are aligned (Fig. ), which is supported by the strong correlation of AREA-PPFD and AREA-TAir (Fig. ). The cluster analysis (Fig. ) indicates that the trees at this site are predominantly energy-limited, with an optimal temperature threshold, which is likely to be occur earlier within a season with rising temperatures . Contrary to the other two sites, SLOPE decreases with increasing TSM, thus contradicting the expectation that additional water would sustain SF. The variability of AREA throughout summer months might be linked to cloud cover. This suggests unusual or site-specific regulation dynamics, potentially influenced by snow melting early in the season, growth and senescence of the larch canopy, or the fact that the seasonal drought is often outside of the growing season (see Fig. ). The overall absence of strong stomatal downregulation at this site suggests that trees follow a non-conservative water-use strategy, possibly at the risk of depleting water for more carbon uptake. However, this may mask delayed stress onset or increased vulnerability to prolonged warming .

Projected climate change will likely intensify the stressors already identified here. In Mediterranean regions, increased frequency and duration of hot–dry extremes may amplify regulation and risk hydraulic failure, pointing to high vulnerability. In tropical forests, warming and drying may shift regulation triggers, potentially exposing hidden hydraulic limitations despite deep water access. At cold continental sites, continued warming may lift energy constraints, increasing SF without corresponding regulation–yet such non-conservative strategies may conceal longer-term risks of water imbalance.

Current research is actively exploring how satellite-based measurements of VWC, e.g., through vegetation optical depth (VOD) or other microwave observables, and derived indicators of plant water status can contribute to a more in-depth understanding of plant health and ecosystem dynamics . The need for diurnal measurements of VWC to better quantify vegetation responses to drought has been highlighted by . As discussed in Sect.  and Fig. , sub-daily satellite observations of vegetation water storage from microwave remote sensing can in principle be used to derive sub-daily variability in SF dynamics, a variable more directly related with plant functioning.

Here, we analysed the effect of temporal resolution on the two metrics proposed. This allows us to evaluate, in cases where 30 min or hourly resolution is not feasible, e.g., for remote-sensing applications, what the minimum temporal resolution would be required in order to adequately capture the SF–VPD dynamics. Generally, no sample rate can capture all details of the diurnal cycle as constrained by the hourly data (Fig. ).

The 8TPD represents an ideal case assuming unlimited observations or coverage, consistent with characteristic frequencies from geostationary satellites for example. While time series from measurements at 8TPD exhibit very good correlation with the reference from hourly observations of SF and VPD for AREA, the results for SLOPE already show a degradation of the signal, with a broad spread across sites. For lower sample rates, results for both metrics are less accurate than for 8TPD, with the 75 % percentile for SLOPE already falling below 0.6 at 4TPD. Compared to 4TPD, the correlations of 3TPDday for SLOPE demonstrate only slight differences, since only nighttime is excluded, where sap usually does not flow significantly. Therefore, measurements at 3TPDnight, which include only nighttime flux, would not be a suitable choice for both metrics.

The mixed correlation results of SLOPE from resampled SF and VPD are likely due to the low signal-to-noise ratio of the hourly metric itself. These mixed results originate in dynamics resulting in negative SLOPE or high positive and negative outliers, such as rainy days or dew, which are hard to interpret. It is possible that SLOPE calculated from SF and VPD at lower sample rates could better infer the signal needed for early warning of tree stress hotspots by neglecting the values of SF and VPD that produce outliers. Another potential reason for the low accuracy of SLOPE from resampled observations is the possibility of both under- and overestimation, depending on the hysteretic behavior of the diurnal cycle, which shifts the maximum value of VPD behind the maximum of SF. Furthermore, the peak times of SF and VPD at some sites (e.g. the russian site) did not match solar noon. While the dynamics at the proposed times and sample rates would be well captured at FRA_PUE and GUF_GUY_GUY, both metrics of the diurnal cycle could be underestimated at sites with dynamics similar to RUS_POG_VAR.

These results raise the question of whether the choice of metrics is adequate to capture the diurnal cycle of SF and VPD from satellites and how they can be improved. AREA is generally underestimated, but relative trends in AREA tend to be consistently retrieved with coarser temporal resolution. SLOPE has lower Pearson R2 than AREA, however, the relative trends are still sufficient to indicate water limitation at most sites. Nevertheless, alternative approaches should be evaluated. For example, instead of the absolute slope, the ratio of the actual to a potential morning regression slope could be considered. This metric could identify collapse days, when the ratio is high, and exclude outliers when it exceeds 1. The challenge here mainly lies in how a potential slope could be defined.

The reduced variance of the metrics under high environmental stress conditions indicates a stabilization of underlying processes and enhances the predictability of stress patterns during these periods. Therefore, SLOPE and AREA can detect early warning signs of SF regulation, which indicates tree health, from the sub-daily response of SF to VPD, which could then be further investigated in detail through in situ observations to enhance our knowledge of specific stress responses outside of laboratory settings and mitigate subsequent tree mortality events. For this purpose, a sample rate of 3TPD centered at day-time, would suffice.

One limitation of the current study is the reliance on sap flow data (a flux), while the the quantity more likely to be provided by satellite remote sensing is VOD (or derived VWC), a measure of vegetation water storage. Recall from Fig. 1 that the expected shapes of the hysteresis curves are different for SF and VWC. It is therefore possible that a SLOPE metric may be more reliably estimated when monitoring VWC rather than SF under a 3TPDday scenario. The application of conservation of mass to translate between the two is complicated by a mismatch in terms of the volume being considered. Sap flow measurements quantify the movement of water through the sapwood of individual tress. Although sensing depth of VOD and its sensitivity to water content in different compartments of the vegetation are known to depend on frequency, VOD (and derived VWC or biomass) are commonly assumed to represent storage of the vegetation as a whole, and spatially integrated across some footprint. Therefore, the dynamics observed in sap flow data, and the derived daily slope and area metrics should not be considered a trivial proxy for VOD or a change in VOD. That said, sap flow data reveal when transpiration is occurring which allows us to identify when, and the degree to which storage is likely to be changing at sub-daily scales. While it seems logical that steeper slopes and larger areas in sap flow would also result in similar changes in VOD, this should be confirmed with sub-daily VOD estimates from tower-based radars and/or radiometers and emerging GNSS-related techniques .