The Niwot Ridge Subalpine Forest AmeriFlux site (US-NR1; ) is located on a remnant glacial moraine in the Colorado Rocky Mountains about 8 km east of the Continental Divide. The forest near the 26 m US-NR1 main tower is primarily composed of subalpine fir (Abies lasiocarpa var. bifolia), lodgepole pine (Pinus contorta), and Englemann spruce (Picea engelmannii) with a tree density of around 4000 trees ha⁻¹ (0.4 trees m⁻²), a leaf area index (LAI) of 3.8–4.2 m² m⁻², and tree heights of 13–15 m . The trees near the tower (as identified by lidar) are shown in Fig. . The mean annual temperature at US-NR1 is around 2 °C. According to the Köppen–Geiger climate classification system the site is type Dfc, which corresponds to a cold, snowy/moist continental climate with precipitation spread fairly evenly throughout the year; the long-term mean annual precipitation at the site is around 800 mm and snow typically covers the ground from mid-November until late May . During the warm season, precipitation primarily occurs during afternoon convective storms generated by upslope mountain plain wind patterns (e.g., ). More information about the US-NR1 site can be found in and references therein or online at http://ameriflux.lbl.gov (last access: 10 October 2025).

Frozen tree boles have been shown to limit sap flow within the US-NR1 trees and control seasonal transpiration onset and dormancy as well as changing the bole flexural properties, which modifies the tree sway frequency (e.g., ). The freezing of internal tree water also changes the water permittivity, which affects VOD (e.g., ). To avoid the complications of frozen boles and align with previous work done during the warm season at the US-NR1 site (e.g., ), we restrict our analysis to the “warm” season, which we define as starting when the boles thaw in spring and ending when they freeze in the fall.

The measurements used in our study are summarized in Table . Most of these measurements have been described in previous publications (e.g., . In the Appendices we provide additional information on the measurements of ET (Appendix ), turbulence (Appendix ), net radiation (Appendix ), surface wetness (Appendix ), dielectric permittivity εa and tree bole water content (Appendix ), and airborne lidar (Appendix ). Precipitation measurements were collected at the nearby NOAA US Climate Reference Network (USCRN) site (see Table ). Below, we describe the VOD and tree sway measurements and our use of the above-canopy and subcanopy ET measurements.

Because the magnitudes of VOD and tree sway frequency are related to water content within the canopy, we decided it was worthwhile to compare these observations with canopy water content determined by a land surface model. Previous studies of canopy evaporation at US-NR1 by used the big-leaf land surface model CLM4.5, making these results readily available for inclusion in our current study.

When liquid water is present on the canopy, CLM4.5 uses the leaf boundary layer resistance rb scaled by the potential evaporation from wet foliage to evaporate the canopy water. The expression for rb is 3rb=1CvUavdleaf-0.5, where Cv is the turbulent transfer coefficient between the canopy surface and canopy air (Cv = 0.01 m s^−0.5), Uav is an estimate of the wind speed within the subcanopy, and dleaf is the characteristic dimension of the leaves in the direction of wind flow (with a default value of 0.04 m). The big leaf of CLM4.5 is divided into areas where evaporation (wet areas) and transpiration (dry areas) occur and scaled accordingly. The water and energy budgets within CLM4.5 must be balanced and a 30-step iterative process is used to simultaneously solve the system of equations. For full details on these iterative steps see Chap. 5 in .

The CLM4.5 parameter maximum leaf wetted fraction fwetmax⁡ limits the area of the leaf surface that is wet. A larger fwetmax⁡ value decreases the dry portion of the canopy undergoing transpiration and increases the leaf area for evaporation of canopy water. For fwetmax⁡ = 1, the entire leaf is covered in water, which is the CLM4.5 default value . In , 14 different CLM4.5 configurations were examined; they concluded that fwetmax⁡ = 0.02 was more appropriate for a needleleaf forest than fwetmax⁡ = 1. One of the CLM4.5 variables (“H2OCAN”) represents the amount of intercepted water on the canopy surfaces. Because changes in H2OCAN should be similar to changes in VOD, we use four of the CLM4.5 cases from to examine “CLM Canopy H2O”. The cases we chose to include are A1 and F2 (fwetmax⁡ = 1) and B0 and G1 (fwetmax⁡ = 0.02). Other differences between the A1, F2, B0, and G1 configurations are related to the subcanopy turbulent transfer coefficient, stability criteria, and use of friction velocity; anyone curious for additional details can refer to Table 1 in . The mechanisms that control the evaporation of warm-season intercepted water in CLM4.5 are the same as those in CLM5 .

To study the impact of rainfall interception on the evaporative fluxes and VOD, we followed the methodology of and label days when the daily rainfall exceeded 3 mm as “wet” days. The choice to use 3 mm as the wet-day criterion was a balance between effectively capturing the effect of precipitation and providing enough wet periods to improve the wet-day statistics. To examine the effect of precipitation on VOD, ET, and the other variables we use the following strategy: first, we designate the precipitation state of the day of interest as either “Wet” or “Dry”, and we then indicate the preceding-day precipitation state with a lowercase letter (“w” for wet or “d” for dry) and create composite diel cycles for dDry (dry day followed by a dry day), dWet (dry day followed by a wet day), wWet (wet day followed by a wet day), and wDry (wet day followed by a dry day) days. Only 30 min periods when all variables of interest were available were used to create the diel cycle composites.

It is important to realize that eddy covariance fluxes are sensitive to rain accumulating on sonic anemometer transducers and gas analyzer windows, so therefore on dWet and wWet days either gap-filling of ET is needed or periods with heavy rain are excluded from the analysis. Therefore, our results focus on wDry days where there is zero to little precipitation (daily total < 3 mm) and the flux instruments have a chance to dry out. In order to improve the statistics for certain variables, we use additional years of data. For example, for the ET diel cycle analysis, we used ET data from the warm season between the years 2004 and 2022. Whenever this occurs, it will be described in the figure caption.

Unless otherwise noted, ET and all other statistics are calculated over 30 min periods. The sampling rates of the primary variables are listed in Table . To ease comparison with other studies (e.g., ), we have expressed ET and the rate of precipitation in units of millimeters of H₂O per hour (mm h⁻¹).

Figure  demonstrates how VOD, tree sway frequency fsway, and the dielectric permittivity εa (our proxy for tree water content) were modified by precipitation in fall 2022. The “wet” days are highlighted by thin vertical black lines and the blue squares in Fig. c. Similar time series for the 2023 spring/summer period are in the Supplement (Fig. S7). As mentioned in Sect. , our analysis focused on the period after the boles had thawed in spring and before they froze in fall, and the precipitation was rain.

As one would expect, bole temperature varied between the daytime and nighttime air temperatures (Fig. a). We ended our 2022 analysis period when the bole temperature went below 0 °C on 4 October 2022 (DOY 277). The freezing of tree sap water and dependence of water permittivity on temperature have been shown to have a large effect on VOD (e.g., ) as well as changes to the tree flexural stiffness, which affects tree sway (e.g., ). This change in behavior in VOD, fsway, and εa occurred just before DOY 300 in Fig.  when the air temperature dipped below -5 °C and the boles became frozen.

Variations in VOD were primarily due to changes in the subcanopy GNSS (gnssB signal strength anomaly), whereas the gnssA anomaly at the top of the tower was stable over the entire period (Fig. b), including in colder conditions. This indicates that the method used to calculate VOD worked as expected, and it was gnssB in the forest that was affected by the canopy-intercepted precipitation. The bias in the signal strength anomalies between the two GNSS receivers is also indicative of the effect of the canopy.

During precipitation tree sway frequency dropped on the order of 0.05 Hz (Fig. b), which is consistent with previous results (e.g., ). In addition to the frequency drops due to precipitation, there was a more gradual low-frequency change over time, which we have removed as described in Sect. . The VOD time series also has a very weak low-frequency trend that we did not remove (because its removal would not affect our results). The dielectric permittivity of the sap water in the tree boles increased in magnitude following precipitation events and decreased sharply as the boles froze (Fig. d).

To summarize how precipitation affected our observations, net radiation Rnet, ET, VOD, detrended tree sway frequency fd, surface wetness, and precipitation from our analysis period (fall 2022 and spring/summer 2023) are composited in Fig.  based on the dDry, dWet, wWet, and wDry precipitation states described in Sect. . During wet periods Rnet (Fig. a) and ET (Fig. b) were both reduced below dDry conditions; however, on a wDry day there was an increase in ET (relative to dDry conditions) of around 0.1 mm h⁻¹ at midday, which is due to evaporation of recently fallen rainwater on the vegetation and ground .

On a dDry day, VOD was around 0.37 and fairly constant over the diel cycle (Fig. d). As precipitation occurred (i.e., dWet and wWet days), VOD increased to between 0.55 and 0.61. The maximum VOD value of 0.61 occurred on the afternoon/evening of a wWet day, which matches the timing of the highest precipitation amounts (Fig. e). On a wDry day, VOD was initially high (at 00:00 MST) but by the afternoon had recovered to near the dDry value.

The impact of precipitation on the tree sway frequency (Fig. c) is almost the mirror image of the VOD pattern (Fig. d), and the linear relationship between VOD and sway frequency is shown in Fig. . Here, it is clear that the low-frequency detrending of the tree sway data (see Sect. ) did not significantly modify the results (i.e., the slope of VOD versus fsway was -4.26 compared to -4.34 for VOD versus fd). It is highly encouraging that these two completely independent methods (VOD and tree sway) show consistent changes as the canopy becomes wet and dries.

The surface wetness data closely matched the VOD trends with an important exception: the wetness sensor rapidly dried in the early morning (at around 08:00 MST, i.e., when Rnet increased sharply) for the dWet, wWet, and wDry days. We describe some of the limitations with the wetness sensor in Appendix  and summarize our thoughts here: (i) the wetness sensor was located near the top of the canopy and therefore had increased exposure to radiation and wind, (ii) the wetness sensor is a small flat plate that does not properly represent the water-holding capacity of an evergreen branch/needles, and (iii) the mountain plain wind pattern at the site leads to clear mornings (which encourage evaporation), while clouds/precipitation typically occurred in the afternoon or evening (Fig. e). Therefore, wetness data should be considered an estimate of how long it takes the water to evaporate from a flat, hard, open surface near the canopy top. At best, it could represent leaf wetness in the upper part of the forest canopy, but it is not indicative of the entire forest drying (and therefore reasonable to see the wetness sensor drying faster compared to what can be inferred from VOD or tree sway motion).

The time series of dielectric permittivity (Fig. d) suggests that εa would provide valuable information about how the internal sap water of the trees was modified by precipitation. However, the dielectric permittivity included many large jumps and shifts that we could not explain and were corrected as described in Appendix . Over the diel cycle, it is expected that morning increases in transpiration will reduce the trunk water content and trunk diameter (e.g., ). As shown in Fig. S8c, on dDry, dWet, and wDry days, εa reached a maximum just after sunrise (at around 07:00 MST) followed by an afternoon minimum (after the midday transpiration maximum). On wWet days, when transpiration was reduced, the diel cycle of εa was fairly constant. In general, εa on wet days was slightly elevated compared to dry days, but the dDry and wDry days were not different enough to extract insightful information, as we had for VOD and tree sway (Fig. c, d).

There is an important implication from the dDry periods (when most of the ET is transpiration, not evaporation). Though εa had a dDry diel cycle with a clear early-morning maximum and late-afternoon minimum, VOD had relatively small variation without any apparent pattern (Fig. ). This suggests that, at our site, VOD changes were largely controlled by water on the canopy surfaces, not the internal water content of the trees. In contrast, tree sway frequency had a diel pattern on dDry days (Fig. c) that was similar to that of εa (Fig. e), suggesting that tree sway frequency was more affected by internal tree water content changes than VOD. also show that tree sway motion followed changes in internal tree water content.

The lack of a clear diel pattern in VOD during dDry conditions was a surprising result because previous studies (e.g., ) have shown a VOD diel pattern that they related to the changes in internal water content of the forest/trees. As we looked closer at this, we realized that US-NR1 VOD in dDry conditions was much lower than VOD from the other sites (both the mean value and the diel range in dry conditions). For example, the US-NR1 dDry VOD diel range is 0.36 to 0.38 (Fig. d), whereas the VOD diel range in the study by was on the order of 0.85 to 1.1 (their Fig. 4) and that of had a range of 0.62 to 0.65 (their Fig. 2). Both of these studies are from deciduous forests in the eastern USA. Though we cannot definitively explain the reason for low VOD at the US-NR1 site, we offer a few possible explanations: (i) the US-NR1 forest has a lower tree density, (ii) the internal water content of the coniferous US-NR1 trees is lower and stored differently within the tree bole than broadleaf trees in the more humid locations , or (iii) there is too much noise in the VOD measurements to properly capture the true diel cycle in internal tree water content during dDry conditions (which is when the VOD signal is weakest).

We now take a closer look at the wDry diel cycle in Fig. , which is re-plotted in Fig.  with a few important changes. First, for ET we used the warm seasons between 2004 and 2022 to create the wDry ET composite and include the subcanopy (2.5 m) ET fluxes (Fig. a; the other precipitation states for 2004–2022 are shown in Fig. S9). Second, we use the ET diel cycle difference between dDry and wDry conditions to estimate the 2.5 m ground evaporation and 21.5 m canopy and ground evaporation (Fig. b). Third, we add a panel of CLM4.5 wDry Canopy H2O results for comparison to the observations (Fig. e). We also note that wDry days have much less actual rain than wWet and dWet days (Fig. e), so ET on a wDry day has fewer missing or gap-filled data caused by instrument issues due to heavy rain (Sect. ).

When we examine the VOD and tree sway data over the 24 h wDry diel cycle, a fairly obvious change in slope versus time occurred at around 14:00 MST (Fig. c, d). We highlighted this by a linear fit through each period, as shown by the red lines in Fig. c and d. For tree sway (Fig. c), the slope changed from 0.052 to 0.008 Hz (d)⁻¹ (a slope decrease by a factor of 6.5). For VOD (Fig. d), the slope changed from -0.193 to -0.034 (d)⁻¹ (a slope decrease by a factor of 5.6). For both tree sway and VOD the smaller slopes after 14:00 MST were indicative of the intercepted water being removed from the canopy by evaporation. The day-to-day variability in VOD and sway frequency was also smaller in the afternoon/evening of a wDry day compared to the early-morning and midday values (Fig. S10c, d). As discussed in Sect. , the wetness sensor rapidly dried between 06:00 and 08:00 MST (Fig. d) due to the sensor characteristics and location.

Above-canopy and subcanopy ET on dDry days both peaked around an hour before noon MST at around 0.25 and 0.06 mm h⁻¹, respectively (Fig. a). In contrast, the corresponding ET values on wDry days had a similar pattern, but with peak midday values of 0.33 and 0.08 mm h⁻¹. The higher-magnitude ET on wDry days (relative to dDry) was due to the presence of liquid water on the trees and ground that enhanced evaporation; we show the enhancement amount as ΔET in Fig. b. Between 00:00 and 06:00 MST on a wDry day, 21.5 m ΔET was fairly constant, suggesting a canopy evaporation rate of around 0.02 mm h⁻¹. It is important to note that 2.5 m ET during the nocturnal hours was very small (< 0.01 mm h⁻¹), which suggests that nocturnal evaporation was primarily from the canopy, not the ground. By the end of a wDry day (between 18:00 and 24:00 MST), there was no more liquid water to evaporate, and 21.5 m ΔET became small (< 0.01 mm h⁻¹) (Fig. b). This ET result is consistent with the VOD and tree sway observations, which also suggest that the intercepted canopy water was fully evaporated by the afternoon or early evening on a wDry day.

If we only consider the nocturnal data on a wDry day, then the VOD, tree sway, and ET results are in good agreement (Fig. ). VOD and tree sway data suggest that the evaporation rate of canopy-intercepted rain is fairly constant (Fig. c, d), while the ET data tell us that nocturnal ground evaporation was small/negligible and the evaporation of canopy-intercepted rain was fairly constant at around 0.02 mm h⁻¹ (Fig. b).

For the wDry daytime data, the comparison between VOD and fd with ET is less clear. The VOD and fd results imply that the canopy evaporation of intercepted water was nearly linear with time from the early-morning (nocturnal) hours up until the time that the intercepted water was completely (or mostly) evaporated, at around 14:00 MST. This would mean that the morning sunrise (i.e., increases in Rnet, transpiration, and air temperature) had a minimal effect on canopy evaporation. In contrast, between around 06:00–10:00 MST, 21.5 m ΔET increased from a value of around 0.02 to around 0.08 mm h⁻¹ (Fig. b). A portion of this increase can be explained by increased ground evaporation starting just after 08:00 MST. However, the increase in 21.5 m ΔET between 06:00 and 08:00 MST is inconsistent with the VOD/tree sway measurements, which suggest the canopy evaporation was fairly constant during that period. One possible explanation for the increase in ΔET between 06:00 and 08:00 MST is that transpiration between dDry and wDry days is not the same. However, previous work at the US-NR1 site has suggested that transpiration is similar on dDry and wDry days and tree-to-tree variability from the 2007 sap flow data of was small (Fig. S11b–d).

Overall, our ET-based canopy evaporation estimates of 0.02 to 0.08 mm h⁻¹ are on the lower side compared to other studies' canopy evaporation rates, which typically range between 0.03 and 0.45 mm h⁻¹ (Table ). The ratio of midday canopy evaporation to total ET on a wDry day seems reasonable (i.e., E ET⁻¹ ≈ 0.2). Furthermore, previous studies have suggested that canopy evaporation measurements tend to overestimate modeling results (e.g., ).

The lack of a change in VOD and sway frequency during the morning transition is consistent with previous canopy evaporation estimates in evergreen forests (e.g., ) but counter to the general consensus that net radiation plays an important role in evaporation of canopy water (e.g., ). Our results suggest that turbulent mixing of dry air into the canopy airspace has a primary control on canopy evaporation, whereas the early-morning increases in net radiation and air temperature are of secondary importance. The US-NR1 site is in sloping terrain (≈ 6 % slope) that is subject to nocturnal drainage flows, which provides a mechanism/energy for turbulent mixing (and horizontal advection) to occur. If the turbulence within the canopy is controlling canopy water evaporation, one would expect it to be approximately constant from early morning to past sunrise. Previous studies have shown that above-canopy friction velocity u∗ typically increases from a nocturnal value of around 0.4 to over 0.6 m s⁻¹ at midday . We found that subcanopy local (u∗)z at 5.7 m on a wDry day was fairly constant through the morning transition period with a slight increase at midday (Fig. ). In addition, between 00:00 and 08:00 MST, 5.7 m (u∗)z on a wDry day is elevated over that of a dDry day (Fig. c). This topic goes beyond the goals of the current study but suggests that turbulence within the canopy could be playing a primary role in the evaporation of canopy-intercepted rainwater.

As one would expect, CLM Canopy H2O increases for the dWet and wWet precipitation states depending on the value of the maximum leaf wetted fraction fwetmax⁡ (Fig. c). Furthermore, the B0 and G1 cases (fwetmax⁡ = 0.02) retain more water within the canopy than the A1 and F2 cases (fwetmax⁡ = 1; see Sect.  for details about each case), which is reflected in the differences in CLM4.5 ET between cases (Fig. b). Also note that on a wWet day, there is a sharp decrease in CLM Canopy H2O following sunrise on a wWet day, which is similar to what we observed with the tower wetness sensor (Figs. d and S9c).

If we focus on the wDry conditions, canopy H₂O shows a distinct decrease in the intercepted canopy water content between 06:00 and 11:00 MST (Fig. e), where the A1 and F2 cases (fwetmax⁡ = 1) had rapid drying between 06:00 and 08:00 MST and the B0 and G1 cases (fwetmax⁡ = 0.02) between 06:00 and 11:00 MST. It appears that in the B0 and G1 cases, a small amount of water is retained by the canopy in the wDry afternoon and evening (the reason for this behavior is unknown). Relevant to our study, the important result is that all four cases show that CLM4.5 canopy evaporation increased at sunrise, which is inconsistent with the VOD and tree sway frequency results. In fact, the CLM Canopy H2O behavior at sunrise is strikingly similar to that of the wetness sensor (Fig. d). This suggests that both the CLM4.5 model and the wetness sensor do not properly represent the water-holding capacity of an evergreen forest (as discussed in Sect. ). If the nearly linear change in intercepted water content found with the VOD and tree sway observations is also found in other ecosystem types, this could be an area to improve CLM.

Up to this point, we have focused on the composite mean results from our measurements (i.e., Fig. ). The variability of tree sway frequency and VOD is also an important consideration, and Table  lists the precipitation amounts and timing for each of the 17 wDry days. Figure  shows each of the 17 wDry day time series of VOD and sway frequency. Here, we observe that (i) VOD has more rapid temporal variations than sway frequency, (ii) case-by-case variability in VOD and sway frequency during the drier conditions is smaller than during wet conditions (this is shown quantitatively in Fig. S10c, d), (iii) rainfall occurred on wDry days but is less than our 3 mm (d)⁻¹ criterion, and (iv) rainfall occurred at all times of day during wWet and dWet days (Table ). By highlighting the wet days with the highest and lowest precipitation with bold lines in Fig. , it is clear that the largest wet-day precipitation amount (red bold line) induces a significant change to VOD and fd, whereas the lowest precipitation amount (olive dashed line) results in a rather small change to VOD and fd.

We considered how the timing of the rainfall affected our results by shifting the precipitation periods so they align at the end of the wWet or dWet day (Fig. ). The maximum shift to the time series was 9.5 h, but most were less than 4 h (Table ). For wet days with multiple precipitation events, the final precipitation event of the day was used to align the time series. The choice of aligning the time series to the end of the wet days was arbitrary but useful for comparison purposes. In Fig. a and b, the mean composite value of fd and VOD is shown as a light blue line (with filled circles) in the “Post-precip” period, which can be compared with the black line that is the composite of VOD and tree sway from Fig. . While the line from the shifted data is slightly smoother, the trends that we observed in the change of slope in VOD and sway frequency at around 14:00 MST are similar.

As an attempt to determine the key variables driving the changes in VOD and fd, we created scatter plots for VOD (Fig. ) and fd (Fig. , where we have multiplied fd by -1 to make the pattern similar to that of VOD). First, we confirmed that the fit of VOD versus fd shown in Fig.  was robust over all 30 min data (Fig. a). Because of the close relationship between VOD and fd, the scatter plots in Figs.  and  show similar patterns, and descriptions with either VOD or fd are interchangeable. We note the following: (i) though there is a lot of scatter, the largest rainfall amounts occurred during low-wind periods (i.e., 21.5 m WS < 3 m s⁻¹, see Fig. a), (ii) as one would expect, there was a tendency for higher VOD values (smaller fd) with larger precipitation amounts (Figs. b, b), and (iii) drier air (i.e., larger VPD) led to smaller VOD (larger fd) values (Figs. c, c). The relationship between VOD and horizontal wind speed and friction velocity (u∗)z was less clear (Fig. d, e, f), but there was a slight tendency for lower (u∗)z to result in higher values of VOD; this relationship is confounded by the tendency for higher precipitation to occur at lower wind speeds (Fig. a). The lack of a relationship between fd and wind speed is especially apparent at moderate to higher WS values in Fig. d (e.g., 21.5 m WS > 2.5 m s⁻¹) and consistent with the mechanical theory discussed in Sect.  and shown in Eq. (). This supports our interpretation that variations in the fd time series are primarily influenced by canopy interception and subsequent evaporation. Finally, it should be noted that some of the short-term variability in VOD is inherent to the uneven and constantly varying distribution of satellites and orbits through the canopy. This noise is only partially addressed by the data-processing algorithms.

We found that the timing of evaporation of canopy-intercepted rainwater in a subalpine forest was consistently measured by two very different techniques, GNSS-based VOD and accelerometer-based tree sway frequency. Neither technique directly measures evaporation, but both use proxy variables that are affected by the amount of liquid water within the canopy. It is encouraging to see good agreement between these techniques, especially considering that tree sway measurements were from a single tree, whereas the VOD footprint area was around 2000 to 3000 m². In contrast, eddy covariance ET measures the amount of vertically transported water vapor, but it is subject to theoretical and practical challenges, especially in the subcanopy (e.g., ). A few of these challenges are (i) separating the total water vapor (ET) flux into the component fluxes of transpiration and canopy/ground evaporation, (ii) decoupling between the true surface flux and the flux measured by the eddy covariance sensor (e.g., ), (iii) taking into account non-steady flow and horizontal water vapor transport (which could be taking place below our 2.5 m subcanopy sensor), (iv) eddy covariance instruments not working properly during active rainfall, and (v) footprint issues that are difficult to evaluate. Most of these concerns do not affect the VOD and tree sway measurements, which make them appealing to combine with eddy covariance ET.

We end this section with a few limitations to our study and recommendations for future studies:

The VOD results are based on only 17 wDry days, while the ET results are based on 176 wDry days. So, there are significant statistical differences (and time periods covered) between these two data sets. If we had a multiyear record of VOD data that matched ET, then we could better examine other effects on the canopy evaporation, such as the dryness of the air or the rainfall amounts and intensity.