Water flux components and soil water-atmospheric controls in a temperate pine forest growing in a well-drained sandy soil



[1] The influences of soil water supply and atmospheric demand on transpiration were studied to gain insight into the physical mechanisms limiting forest water use within the broader context of total canopy water loss to the atmosphere. Evaporation from forests (E) can be partitioned in to four main source components: canopy transpiration (Ec), understorey transpiration (Eu), evaporation from the soil (Es), and the evaporation of intercepted water (EI). Ec and EI usually make up most of E. Ec estimated from sap flow measurements and modeled EI estimates were compared with eddy covariance measured values of E to quantify the components of the above canopy water flux to the atmosphere, in a temperate pine forest ecosystem established on a well-drained sandy plain at Turkey Point in southern Ontario, Canada. Daily values of E averaged 2.4 mm d−1 and reached maximums of 4 mm d−1 while daily values of Ec averaged 1.2 mm d−1 over the growing season. The evaporation of intercepted water (EI) was generally between 2 and 3 mm per event. EI accounted for 34% and Ec accounted for 47% (31 to 67% range on a monthly basis), together accounting for 81% of E during the growing season. Ec increased linearly with vapor pressure deficit (VPD) until a transition point was reached, after which mid-day Ec rates remained more or less constant. For analysis purposes, data were segregated by early morning VPD (or VPDin) in an attempt to characterize the atmosphere at the beginning of the daily transpiration cycle. This technique revealed that shifts in the timing and magnitude of Ec rates masked the response of Ec to changes in soil water content. Analysis also suggested that while increasing VPDs may limit maximum transpiration rates, daily total transpiration is a conservative quantity. This study improves the understanding of the physical mechanisms limiting water loss in forested ecosystems growing on water-stressed soils by investigating the effects of VPD and soil water content on Ec.

1. Introduction

[2] The impact of global climate change on precipitation and temperature regimes in temperate regions threatens to significantly alter forest water budgets [Wullschleger and Hanson, 2006]. Over many landscapes plants exert a strong control on evapotranspiration processes because of their ability to access, transport, and evaporate water that would otherwise be decoupled from terrestrial water cycles [Calder, 1998; Nosetto et al., 2005]. Total evaporation from a forest (E) can be partitioned into four source components: canopy transpiration (Ec), understorey transpiration (Eu), evaporation from the soil (Es), and the evaporation of intercepted water (EI). With E given as:

equation image

[3] Despite a large body of literature there is still considerable uncertainty about partitioning E into its source components and their key controls in forest ecosystems. Adequate characterization of E components in mature planted conifer forests, where the natural succession toward a mixed-wood forest has begun can be difficult due to the spatial dynamics of these forests. Previous studies have recognized canopy wetness as the main determining factor of the dominant component sources of above canopy water flux (i.e. when the canopy is wet EI dominates and when the canopy is dry Ec dominates) [Barbour et al., 2005]. The degree to which the two main components (Ec and EI) dominate the partitioning of forest water flux over daily or annual timescales is largely determined by the openness of the canopy and character (amount, frequency etc.) of the rainfall. Closed-canopy forests often have limited contributions from the other sources of forest water flux (understorey transpiration, Eu and evaporation from the forest floor, Es) because of the small amounts of advected and radiant energy available below the canopy [Wullschleger et al., 1998]. In contrast, open-canopy forests may see much larger contributions from these components [Barbour et al., 2005; Unsworth et al., 2004]. The emergence of gaps within the generally closed canopies of planted coniferous forests therefore poses a unique challenge when trying to characterize the components of E.

[4] Ec is generally the largest component of the total water vapor flux to the atmosphere in forested ecosystems [Schafer et al., 2002]. Xylem sap flux can be used to estimate tree transpiration by scaling point measurements of sap flow velocity to represent the whole stem [Granier, 1987]. These tree transpiration estimates can then be scaled to stand level to calculate Ec. Extensive work has been conducted investigating the environmental controls of Ec by comparing measurements of sap flow velocity with soil and environmental variables [Bovard et al., 2005; Kurpius et al., 2003; Wullschleger et al., 1998; Hogg and Hurdle, 1997]. Studies have shown that much of the variation in Ec can be explained by variation in vapor pressure deficit (VPD) [Kurpius et al., 2003; Hogg et al., 1997; Hogg and Hurdle, 1997]. They suggested a strong linear relationship until a transitional VPD threshold is reached, after which Ec tends to remain relatively constant. This relationship can be sensitive to soil water supply [Cinnirella et al., 2002; Wullschleger et al., 1998]. Barbour et al. [2005] reported a strong linear regulation of Ec by VPD until a threshold VPD of 0.6 kPa in a temperate coniferous rainforest. Similarly, linear dependencies of Ec to VPD were found to exist until a threshold VPD of 1.0 kPa in both a temperate hardwood forest in Michigan, USA [Bovard et al., 2005] and a boreal trembling aspen (Popolus tremuloides) stand in Saskatchewan, Canada [Hogg and Hurdle, 1997]. In each of these cases the lack of a transpiration response to VPD above a threshold value was concluded to be a stomatal response to high VPD.

[5] Other studies have shown that when soil water supply is limited Ec varies with VPD over a much narrower range of VPD and the flux of Ec is suppressed or dampened at all but the lowest VPD values [Wilson et al., 2001; Wullschleger et al., 1998]. However, the patterns of the limitation response of Ec to high VPD and low soil water deficit appear similar and are often concurrent, sometimes making interpretation difficult [Kurpius et al., 2003; Oren and Pataki, 2001]. It is unclear how Ec responds to the combined effects of VPD and soil water content (θ) [Kurpius et al., 2003]. The sensitivity of the relationship between Ec and VPD to variation in soil water supply is poorly understood [Bovard et al., 2005].

[6] Relatively little is known about how θ and VPD interact with each other to influence Ec [Bovard et al., 2005]. Previous studies have suggested that a strong negative feedback between VPD and canopy conductance and a general insensitivity to typical θ variations may result in similar growing season Ec rates between forests growing in comparable climates [Humphreys et al., 2003; Oren and Pataki, 2001; Roberts, 1983]. Some studies have shown that stomatal conductance is generally unaffected by θ until a deficit occurs at which point trees close stomata in an effort to conserve water limiting transpiration [Cinnirella et al., 2002; Phillips and Oren, 2001; Irvine et al., 1998; Wullschleger et al., 1998]. The deficit required to induce change in canopy transpiration differs from site to site, most likely due to differences in vegetation type and soil texture, as different plants respond differently to given soil water states [Roberts, 2000].

[7] In this study, water fluxes were measured at the whole tree and whole ecosystem scales in a temperate forest ecosystem, growing on well-drained sandy soils, to gain new knowledge of the physical mechanisms limiting forest water use, with the specific objectives: (1) to determine the magnitudes of the source components of canopy water flux, and (2) to characterize the sensitivity of the relationship between Ec and VPD to variation in soil water supply.

2. Methods

2.1. Site Description

[8] The forest site is located near Turkey Point, Ontario, Canada (42°71′N, 80°35′W). This landscape is part of the Norfolk sand plain and lies upon a gently undulating wind-blown sand dune system (Corporation of Norfolk County, Long Point Region Conservation Authority, and Haldimand-Norfolk Health Unit, unpublished data, 2003). The stand was planted with regularly spaced rows of eastern white pine (Pinus strobus L.) in 1939, about 3 km north of Lake Erie. Overall, the surface of the site can be characterized as flat (0.5–3.0° slopes) [Restrepo and Arain, 2005]. The spacing of the planted white pine varies from 3 × 3 m to 5 × 6 m and has lead to a nearly homogeneous canopy height and structure. The current species composition of the stand is white pine (82%), balsam fir (Abies balsamea L.) (11%) and a population of native Carolinian species mixed through the stand in the intermediate and understorey layers emerging from the gaps in the canopy left by fallen white pines. Some of the Carolinian species include: oak (Quercus velutina L., Quercus alba L.) (4%); red maple (Acer rubrum L.) (2%); and wild black cherry (Prunus serotina Ehrh.) (2%) [Peichl and Arain, 2006]. The understorey is patchy and largely dominated by the deciduous tree species, young white pines and a variety of forest floor plants. A large patch of oak trees (approximately 100 m × 160 m) is located 200 m east of the tower. Important stand characteristics are presented in Table 1.

Table 1. Stand Characteristicsa
  • a

    Means include standard deviations in parentheses where applicable.

  • b

    From LAI measurements made at this site by [Chen et al., 2006].

Max. Leaf Area Index, LAIb8.0
Diameter at breast height, DBH (cm)34.6 (±5.9)
Base diameter (cm)39.4 (±7.9)
Tree height (m)20.2 (±2.1)
Stand basal area (m2)37.3
Stem density (stems ha−1)429 (±166)

[9] The soil is a brunisolic grey brown luvisol, with a very fine sand to fine sandy loam texture, making it predominantly well drained with low water holding capacities. Other soil characteristics are given in Arain and Restrepo-Coupe [2005]. Fine root (<2 mm) biomass, important for the uptake of water and nutrients, decreases rapidly with depth from 1.2 mg cm−3 in the 0–15 cm depth class to 0.49 mg cm−3 in the 15–35 cm depth class, and finally to 0.35 mg cm−3 in the 35–55 cm depth class [Peichl, 2005]. Approximately two thirds of total root biomass is found in the 0–15 cm depth range. The water table in the area is quite deep and generally lays 5–9 m below the surface (Corporation of Norfolk County et al., unpublished data, 2003). Based on standard literature values relating soil water content to soil water potentials for a sandy soil [Oke, 1987], the soil water content at the field capacity is estimated to be 0.16 cm3 cm−3 and the water content at the wilting point is estimated to be between 0.04 and 0.08 cm3 cm−3.

[10] The region has a mean annual temperature of 7.8 (±1.3) °C, an annual precipitation of 1010 mm, with 878 mm (87%) precipitating as rain and 133 mm (13%) as snow, and an average of 2021 hours of bright sunshine per year (1971–2000 (30 years) climate normals established at Delhi located approximately 20 km north of the site (Environment Canada, unpublished data, 2005). Daily precipitation accumulations of ≥2, ≥5, ≥10, and ≥25 mm have an approximate exceedance interval of 13, 20, 23, and 51 days, respectively, as shown by 30-year climate normals at Delhi (Environment Canada, unpublished data, 2004).

[11] The phenology of the broadleaf deciduous species was monitored during the growing season of 2006 with first leaf emergence occurring during the 4th week of April, and full leaf development (approximately 95%) occurring during the 3rd week of May. Senescence began during the 2nd week of October and completed in the 3rd week of November. Therefore the growing season for this study was defined as 1 May to 31 September.

2.2. Climatic Measurements

[12] Continuous micrometeorological measurements were averaged half-hourly at 28m (unless stated otherwise) on top of a scaffold tower. Measured variables included: air temperature and relative humidity (at 2, 14 and 28 m heights), wind speed and direction, incoming and outgoing components of short-wave and long-wave radiation (used to calculate net radiation) and incident and reflected photosynthetic photon flux density (PPFD). Most other measurements were conducted in a 20 × 20 m study plot centered 10 m north of the tower, including sapflow and throughfall. Soil temperature was measured at locations in the study plot at depths of 2, 5, 10, 20, 50 and 100 cm. Meteorological measurement details are given in Arain and Restrepo-Coupe [2005]. Leaf wetness was measured in the canopy at a height of 18 m and near the forest floor at 1 m located in the study plot using electrical impedance grid leaf wetness sensors (model 237; Campbell Scientific Inc., Utah, USA). Each day of the growing season was separated into classes based on rainfall amount and leaf wetness duration in order to compare water fluxes under differing canopy wetness states. We adopted a classification scheme similar to the one used by Barbour et al. [2005]. Wet days were defined as days where more than 2 mm of rain fell, the below canopy leaf wetness sensor was wet for more than 30% of the day and the within-canopy leaf wetness sensor was wet for any length of time. Days were considered dry if less than 2 mm of rain fell, both leaf wetness sensors were completely dry for the entire day and no more than a total of 2 mm of rain had fallen over the previous 3 days. All remaining unclassified days were deemed as intermediate.

[13] Half hour variations in the average volumetric soil water content (cm3 cm−3) in the 100 cm soil profile were estimated using 10 water content reflectometers (CS615; Campbell Scientific Inc., USA) at two locations in the study plot at depths of 5, 10, 20, 50 and 100 cm. Each estimate was corrected for temperature using the manufacturer's temperature correction based on concurrent soil temperature measurements, but variation in soil water content remained unaltered. The spatial variability of volumetric soil water content of the 0–20 cm soil layer was measured 12 separate times, at 12 points along a 50 m transect extending from the study plot toward the north east during the growing season of 2006 using a hand held water content reflectometer (CS620, Campbell Scientific Inc., USA). Average daily coefficient of variation of the mean soil water content was 20% over the course of the growing season, ranging from 12% to 39%. Due to the moderate amount of spatial variability and inherently large size of an Eddy flux footprint it should be understood that the soil water content measurements are more representative of the 20 × 20 m study plot than of the entire stand, and that this is one of the limitations of the study. The root zone volumetric soil water content (θ0−25cm) was calculated using weighted-average soil water content from the six sensors located in the 0 to 25 cm soil profile. A weighted average was used because of the unequal spacing of the three sensor depths in the root zone. Each sensor was assumed to represent a 10 cm column of soil (5 cm above and 5 cm below the wave-guides) and thus the domains of the sensors located at 5 and 10 cm over lapped at the 5–10 cm depth. Their weightings were reduced accordingly to compensate. Comparison of field and lab measurements of soil water content, using water content reflectometers, indicated that any additional uncertainty caused by the temperature sensitivity of the associated above ground electronics as observed in previous studies [Unsworth et al., 2004] was not introduced. Thus the influence of varying soil properties on moisture content could be accepted and did not have to be corrected in our analysis.

[14] Gross precipitation (PG) was measured using a heated tipping bucket rain gauge (52202; R.M. Young Company, USA) mounted on a boom extending from the tower at a height just below the tree tops (20 m) to avoid under-catch during high wind conditions. These measurements were cross checked with data from three additional sources, including Environment Canada's Delhi weather station, a similar tipping bucket rain gauge located approximately 20 km west of the tower site, and three standard bucket rain gauges deployed near the tipping bucket gauge at the top of the tower and one in a small nearby clearing. All gauges showed consistent precipitation data.

[15] Net precipitation, otherwise known as throughfall (Pn), was measured using an 8.0 × 0.1 m v-shaped aluminium trough that funnelled collected water into a single tipping bucket rain gauge (CS700; Campbell Scientific Inc., USA), with a bucket resolution of 8 ml per tip, installed below the canopy in the study plot. Trough location and orientation within the study plot was chosen based on typical canopy cover conditions within the planted forest. Frequent clogging of the throughfall apparatus and tipping bucket overflow during intense storms proved problematic and most of the data from the larger storms was unusable. Parameters for the Gash rainfall interception model [Gash, 1979] were derived from the relationship between rainfall and throughfall for 15 lower-intensity storms free from apparatus clogging and preceded by at least 8 hours of daylight. Derived stand structure parameters were used to estimate the canopy storage capacity (S) and free throughfall coefficient (p) using the mean method outlined by Klaassen et al. [1998]. The free throughfall coefficient (p) was estimated to be 0.34. The canopy storage capacity (S) was estimated to be 2.4 mm. These data were then used in conjunction with micrometeorological measurements to calculate the daily evaporation of intercepted water (EI) using a modified version of the Gash analytical model of rainfall interception by forests [Gash, 1979; Klaassen et al., 1998; Link et al., 2004]. It was assumed that stem flow made a negligible contribution to the water balance and was therefore omitted [Link et al., 2004]. The Gash model describes EI based on an analysis of individual storm events. The evaporation from a saturated canopy (Ew) is estimated from the Penman–Monteith equation with the canopy resistance fixed at zero [Monteith, 1965].

[16] The physical process of interception evaporation is very well understood and has been predicted to acceptable accuracy for many years using less sophisticated methods than the model used here. Estimates used for comparison in water flux studies are often calculated based simply on interception fractions found in similar forests [Hogg et al., 1997] or taken as the remainder of a mass balance equation [Barbour et al., 2005]. However, it has been shown that assuming static values for p and S in interception modeling can introduce significant errors [Link et al., 2004]. Therefore the model was only used to give an estimate of interception loss for comparison with total evapotranspiration and canopy transpiration. The interception model (Gash model) used in this study was modified slightly to enable it to calculate total intercepted evaporation on an event basis rather than summing over longer periods of time (e.g. a month or entire season, as described by Gash [1979]). This was done by calculating separately the daily interception total for days when the recorded precipitation was enough to saturate the canopy, and for days when the recorded precipitation was not large enough to saturate the canopy. Details of the modified interception model are given in Appendix A.

2.3. Sap Flow and Canopy Transpiration

[17] Continuous measurements of white pine xylem sap flow velocity (Js) (m s−1) were averaged over half-hourly intervals using 30 mm long, 1.3 mm diameter, continuously heated thermal dissipation probes (TDP-30; Dynamax, Houston, Texas, USA [Granier, 1987]). Probes were installed at breast height (1.3 m above the base on the northern side) in six mature White pine trees, located within the 20 × 20 m study plot. The sample trees were chosen to be within ±2 standard deviations of the mean stand DBH, in order to represent approximately 95% of the size distribution within the stand. The biometric characteristics of all six trees are given in Table 2. All probes were shielded from the effect of direct solar heating using a reflective, foil-coated, bubble wrap, to minimize non-sap flow related temperature fluctuations. The bark at the probe insertion point was removed to a depth just above the boundary with the phloem. Sapwood depth near the insertion point was determined using a wood core to estimate the amount of non-conducting tissue in contact with the probes. The sapwood depth of each tree was quite uniform and between 5–7 mm of non-conducting tissue was in contact with each probe. Corrections were made for the proportion of the probes in contact with non-conducting tissue following Clearwater et al. [1999]. The Granier style probes integrate the sap flow over the entire radial thickness of the sapwood they are in contact with, and in each case our probes were in contact with 100% of the radial sapwood at their insertion points. The xylem sap flow velocity of each sample tree was calculated following Granier [1987].

Table 2. Sapflow Sample Tree Characteristicsa
TreeDBH (m)Aw (m2)As (m2)
  • a

    DBH is diameter at breast height (1.3 m above the ground), As is tree sapwood area and Aw is total tree wood area at DBH.


[18] Two of the main issues in scaling-up sap flow velocity to stand level canopy transpiration are heterogeneity in sapwood conductance and heterogeneity in stand properties [Kurpius et al., 2003]. However, the even spacing of trees within planted forests tends to minimize differences in xylem sapwood properties between trees within the stand [Kostner et al., 1998]. Thus, this heterogeneity should be reduced in our planted forest dominated by even-aged evenly spaced white pine trees located on a relatively flat texturally homogeneous and well drained soil.

[19] Sap flow velocity was scaled to stand level canopy transpiration (Ec) by calculating the average transpiration per unit ground area for each of the sample trees. Cross-sectional sapwood area of each sample tree was calculated using a site specific allometric equation. This equation and the stand basal area were developed from a harvesting and forest inventory experiment conducted at the site in 2004 (M. Peichl and M. A. Arain, Sapwood and canopy biomass allocation in an age-sequence of eastern white pine forests, submitted to Canadian Journal of Forest Research, 2008, hereinafter referred to as Peichl and Arain, submitted manuscript, 2008). The equation relates sapwood area (As) (m2) to DBH (m) (Peichl and Arain, submitted manuscript, 2008):

equation image

Transpiration (kg m−2 s−1) for each tree was calculated as:

equation image

where As: Aw is the ratio of tree sapwood area to total tree wood area (Aw) at DBH and ρw is the density of liquid water (kg m−3). Stand level canopy transpiration per unit ground area, Ec (kg m−2 s−1) was then calculated by averaging the ET (kg m−2 s−1) from the sample trees and multiplying by the basal area of the stand, BA (m2 m−2):

equation image

where n is the number of trees sampled (n = 6). Values were subsequently converted to half-hourly hydrologic units (mm hh−1) of water flux.

2.4. Ecosystem Water Flux

[20] Whole-ecosystem water flux (E) was measured using a closed-path eddy covariance (EC) system. The EC system consisted of a sonic anemometer (CSAT3; Campbell Scientific Inc., USA), an infrared gas analyzer (IRGA) (LI-7000; LI-COR, USA), and a 12.5 μm diameter (fine-wire) thermocouple. A detailed description of the EC system, calibrations, flux corrections, and gap-filling procedures applied are provided in Arain and Restrepo-Coupe [2005]. The EC system measures water fluxes over a large variable “footprint” source area, but due to practical and economic restrictions its output must be compared with hydrological and soil based measurements made at a much smaller scale. The average distance at which the cumulative flux density is equal to 80% of the total is 402 m (±232) from the tower during the growing season. The character of the EC flux footprint for this study was similar to the sapflow plot in terms of mean tree height and structure. Differences in stem density (429 ± 166 stems ha−1 [Peichl and Arain, 2006]) across the footprint were also small. However, due to the emergence of deciduous trees within canopy gaps left by fallen mature white pine, there are differences in species composition that were not possible to capture within the sapflow plot.

2.5. Data Processing and Statistical Analysis

[21] All data were post-processed using the Matlab 7.1 software package (Math Works Inc., USA). Statistical tests were conducted using the SAS 9.1 statistical software package (SAS Institute Inc., USA). A statistical analysis of the relationships and interactions between Ec, VPD and θ0−25cm was conducted using a simple multivariate approach [Otomo and Liaw, 2003]. The association between VPD and Ec was determined by a simple linear regression. A multivariate approach was used to investigate if the relationship between Ec and VPD varied for three different VPDin classes. VPDin classes were defined as follows: low, VPDin < 0.005 kPa; moderate, 0.05 kPa ≥ VPDin ≤ 0.005 kPa; and high, VPDin > 0.05 kPa. Mean values are presented with associated standard deviations given in parentheses (i.e. μσ)), when applicable. Turbulent flux outliers were removed following Restrepo and Arain [2005], and water vapor flux data were gap-filled following Amiro et al. [2006]. The aerodynamic resistance (ra) (s m−1) was calculated by dividing horizontal wind speed (u) (m s−1) by friction velocity (u*) (m s−1) i.e. [Arain and Restrepo-Coupe, 2005].

3. Results

3.1. Microclimate

[22] Variations of daily total photosynthetic photon flux density (PPFD), mean mid-day saturation vapor pressure deficit (VPD), mean daily root zone soil water content (θ0−25cm), and gross precipitation (PG) for the 2006 growing season (1 May to 30 September) are shown in Figure 1. The 2006 growing season would be best characterized as warm and wet, with temperatures regularly exceeding 20°C and reaching a maximum of 32.5°C on 16 July. Total growing season PG (697 mm) was 34% greater than the 30-year normal PG (520 mm) for the region. PPFD fluctuations indicated frequent cloudy conditions, with highs above 30 mol m−2 d−1 regularly occurring on clear days. Maximum incoming PPFD occurred in mid June and mid July. These maximums corresponded with observed maximums in VPD. There was a strong coupling between the canopy and the atmosphere, as indicated by the low daytime aerodynamic resistance, ra at the site, which generally ranged from approximately 1.0 × 10−5 s m−1 to 3.5 × 10−5 s m−1 (data not shown).

Figure 1.

Daily meteorological and water flux measurements of (a) daily cumulative photosynthetic photon flux density (PPFD), (b) mean daily mid-day saturation vapour pressure deficit (VPD), and (c) mean daily root zone soil water content (θ0−25cm) and gross precipitation (PG), and (d) whole ecosystem water flux (E), and canopy transpiration (Ec), for the growing season of 2006. X axis tick marks indicate the beginnings and mid points of months.

[23] May 2006 was cool and wet until a sharp increase in temperature late in the month. The precipitation total in May (107 mm) was 29% greater than the 30-year normal of 83 mm. Consequently θ0−25cm remained high for most of the month (see Figure 1). June was warm (0.6°C above the normal daily mean maximum) and dry, with only 62 mm of rain falling, approximately 25% less rain than normal. During an early summer dry period θ0−25cm remained quite low, at values below 0.10 for one month (10 June to 10 July 2006) despite numerous rain events totaling 50 mm. θ0−25cm reached a minimum value of 0.06 on 2 July and was taken to indicate the presence of drought conditions. The rest of the summer was uncharacteristically wet. Repeated storms brought 528 mm of rain over the months of July, August, and September. For the same period the 30-year normal precipitation is 270 mm (49% less than observed). During one short dry period in early August θ0−25cm returned to levels below 0.10, but otherwise θ0−25cm steadily rose as temperatures declined and the unseasonably frequent and large precipitation events continued to occur.

[24] Overall, θ0−25cm values at this site ranged between 0.06 and 0.29. Half-hourly observations showed that θ0−25cm decreased rapidly after precipitation due to the moderate to high infiltration rates within these sandy soils. However, after reaching the θ associated with the estimated field capacity, observed θ0−25cm would decrease much less rapidly in a “stair-case” pattern due to transpiration and evaporation losses during the daylight hours followed by a plateau during the night, when θ0−25cm would remain largely unchanged (see Figure 2). During several dry periods through the growing season when θ0−25cm would approach its seasonal minimum the 0–25 cm soil depth would gain water at night without any apparent atmospheric input. Figure 2 presents a comparison of the observed change in θ0−25cm during one of these periods (5 to 13 August 2006) and the predicted change in θ0−25cm over the same period without the observed nocturnal recharge. The predicted θ0−25cm was estimated by simply replacing the nocturnal recharge periods with constant water content values to mimic the typical “staircase” pattern mentioned above. Root zone soil water storage (in mm of water) was calculated by summing the average water content of the 6 sensors located in the 0–25 cm column (i.e. 250 mm × θ0−25cm = θstored). During this period the nightly increases in stored water, of up to 0.7 mm, prolonged the reduction of soil water content from around 0.08 to 0.07 by an estimated 3 days, preventing the additional reduction of approximately 3 mm of stored water from the root zone predicted in the absence of the recharge. Therefore we conclude that the nocturnal recharge helped to reduce the severity of drought effects during extremely dry periods in this forest.

Figure 2.

Comparison of time course of measured root zone soil water content (θ0−25cm), and root zone soil water content, predicted in the absence of hydraulic redistribution 4–12 August 2006.

3.2. Water Flux Patterns

[25] Daily estimates of E based on eddy-covariance measurements and Ec estimates based on scaled-up sap-flow measurements are presented in Figure 1d. E varied from 0.3 to 5.8 mm d−1, tending to be higher in the late summer after the dry period. The seasonal daily mean of E was 2.4 (±1.1) mm d−1. The highest E fluxes occurred during late July and early August. The lowest E fluxes were recorded in May. Ec varied from 0.1 to 2.2 mm d−1 with a daily mean of 1.2 (±0.5) mm d−1. The highest sustained Ec rates occurred in early August, but the period from May to late July contained frequent high daily Ec rates. E and Ec show roughly the same degree of variability, with coefficients of variation of 46% and 41% respectively. The qualitative response of the two variables is similar, but their magnitudes differ greatly depending on the time of the season. Daily mean water flux components and monthly totals of water balance components are summarized in Table 3. Ec was 47% of E over the season, but varied between 31–67% on a monthly basis. During the early growing season (May) Ec accounted for a much larger proportion (67%) of E than at any other time. A divergence from this initially high proportion began in late May. The difference between the two variables increased until the early-summer dry period, when observed E began to fall slightly relative to Ec. E increased over June and July from a mean daily flux of 2.6 (±0.9) mm d−1 to 3.0 (±1.0) mm d−1, while Ec remained unchanged from 1.2 mm d−1. Despite the month-long dry spell through June and July, E and Ec remained relatively high, at levels just below spring values and the difference between the estimates increased. After the dry period ended in mid-July, the difference between E and Ec increased again and both variables reached their highest seasonal means in July and August, respectively: 3.0 (±1.0) mm d−1 for E in July and 1.5 (±0.5) mm d−1 for Ec in August. Both variables began to decrease soon thereafter until the end of the growing season.

Table 3. Seasonal Water Flux Components, Precipitation, and Canopy Wetnessa
MonthPG (mm)E (mm)Mean E (mm d−1)Ec (mm)Mean Ec (mm d−1)EI (mm)% of PG(EI)% of E(Ec)(EI)DaysDryInter.Wet
  • a

    PG, E, Ec and EI represent gross precipitation, ecosystem water flux, canopy transpiration and loss of intercepted precipitation, respectively.

May10749.11.6 (0.8±)33.01.1 (0.5±)2119.967.343.43113126
June6279.02.6 (0.9±)37.01.2 (0.3±)2133.046.826.03011145
July15393.53.0 (1.0±)36.21.2 (0.5±)2918.738.730.5317168
August15486.72.8 (1.3±)46.81.5 (0.5±)2516.053.928.43110138
September22158.72.0 (0.8±)19.51.0 (0.5±)2812.733.247.9305178
Season697367.02.4 (1.1±)172.51.2 (0.5±)12317.747.033.6153467235

[26] Modeled net precipitation, or throughfall (Pn), and interception (EI) components of daily sums of precipitation (PG) are shown in Figure 1c. Modeled EI was in good agreement (r2 = 0.85, p < 0.001) with measured interception from the 15 selected low-intensity storms. Modeled EI increased linearly with the size of the rainfall event, averaging 18% of PG. Expectedly, the proportion of rainfall that was intercepted decreased as the size of the rain event increased [Gash and Morton, 1978; Grelle et al., 1997; Link et al., 2004]. EI was generally between 2 and 3 mm per event with a maximum value of 4 mm during a single 55 mm event. The proportion of EI to PG ranged from 7% to 65%. Modeled EI usually exceeded measured total E on wet days.

[27] Monthly EI totals varied conservatively from 21 mm to 29 mm throughout the growing season, while PG varied significantly (from 62 mm to 220 mm), although each month had roughly the same number of days with rain (varying from 9 to 11 per month). Consequently, the proportion of EI to PG and E varied greatly through the growing season. Table 3 presents the monthly EI and the proportion of EI to PG and E. The maximum amount of rain was intercepted in June, the month with the least amount of rainfall, while the least amount of rain was intercepted in September, the month with the highest amount of rainfall. This indicated that a greater proportion is evaporated on days with lesser amounts of precipitation. These extremes were likewise matched by monthly EI proportions of PG: 33% in June and approximately 12% in September. The proportion of EI to E was the highest in September (approximately 47%) and lowest in June (approximately 26%).

3.3. Controls and Limitations of Canopy Transpiration

[28] A period with an overall decline in soil water contents to near drought conditions was chosen to characterize the relationships and interactions ± between the parameters of soil water supply and atmospheric demand. This was done to investigate the relative individual and combined influences of soil water supply and atmospheric demand on the control and limitation of transpiration. Half hourly diurnal water fluxes and key meteorological variables of this 15-day period in June 2006 are presented in Figure 3. Both E and Ec tended to peak in the early afternoons, circa 13:00 to 15:00 daily (12:00 is indicated by the x axis ticks in Figures 3a3c). The consistent lag of Ec behind E was likely due to differences in the timing of the water uptake measured by thermal dissipation probes embedded in the stem and actual water loss at the crown, as is commonly observed in transpiration studies [Unsworth et al., 2004]. VPD also consistently lagged behind PPFD. Many diurnal meteorological and water flux patterns tend to be more or less symmetrical around the mid-day. E tended to increase rapidly in the morning and decrease rapidly in the afternoon, almost mimicking the response of PPFD. VPD tended to peak much later in the day than the other variables, usually in the late afternoon or early evening (see Figure 3b).

Figure 3.

Diurnal meteorological and water flux measurements for a selected analysis period, 4 to 18 June 2006. (a) Half hourly ecosystem water flux (E) and canopy transpiration (Ec), (b) half hourly photosynthetic photon flux density (PPFD), and saturation vapour pressure deficit (VPD), (c) half hourly root zone soil water content (θ0−25cm) and precipitation (PG), and (d) half hourly diurnal relationship between Ec and VPD.

[29] Figure 3d illustrates the trend in the diurnal time courses of the EcVPD relationship spanning the 15 selected days and includes a “drying” period (11 to 18 June, where θ0−25cm fell from 0.10 to 0.07). As time progressed through this drying period the daily maximum Ec rate decreased concurrently with an increase in daily maximum VPD and an increase in early morning VPD. Figure 4 offers a close-up of three of the diurnal time courses of the EcVPD relationship during this drying period. On the first selected day (11 June) the slope of the relationship between Ec and VPD was maintained until reaching an apex at 15:30. On the second selected day (14 June) the relationship began much the same as the first day, but the slope was reduced in the morning (at around 9:00) due to a reduction in the change of Ec rate relative to the change in VPD. The maximum Ec rate of the second day (approximately 0.06 mm hh−1) was reduced by roughly 25% from the first selected day (approximately 0.08 mm hh−1). The second day also saw roughly a doubling of early morning VPD relative to the first day. Surprisingly, the total Ec flux was exactly the same for the two days (1.8 mm d−1), despite the reduction in maximum half hourly Ec rate. The results from a similar study suggested that a general compensatory water use strategy in pine trees may exist in which Ec reaches the maximum daily rate earlier in the day, followed by moderate consistent Ec rates during dry periods, due to stomatal closure [Kurpius et al., 2003]. This implied that during warm dry days the early arrival of moderate, but steady, Ec rates were minimizing the differences in total daily Ec between days. This was due to the ecosystem maintaining moderate, but constant, Ec rates for longer portions of the day rather than experiencing large fluctuations in Ec. The final selected day (18 June) also began much the same as the other days and again the incline of the relationship decreased at 9:00, but this time much more dramatically. The Ec rate of the final day reached a plateau and was maintained at an almost constant level (of approximately 0.04 mm hh−1) as the VPD continued to climb throughout the rest of the day. The maximum Ec for the final day (approximately 0.04 mm hh−1) was reduced by roughly 25% relative to 14 June (approximately 0.06 mm hh−1), and 50% relative to 11 June (approximately 0.08 mm hh−1). Despite this increased limitation in maximum rate, the total Ec for the final day was 1.2 mm, a 33% reduction (0.6 mm) from the daily totals of 11 and 14 June. Interestingly, the early morning VPD of the final two days was almost 3 times that of the first day. The relationships between Ec and VPD for all days have very similar slopes in the mornings, before their respective reductions in slope.

Figure 4.

Diurnal relationship between half-hourly canopy transpiration (Ec) and vapour pressure deficit (VPD) for three selected 24 hour periods: 11 June (closed circle), 14 June (open circle), and 18 June (closed triangle). The closed star symbols mark 1200 hours.

[30] In an attempt to characterize the control of Ec by atmospheric demand, the relationship between daytime half hourly (6:00 to 18:00 hours) Ec and VPD for a cloudless eight day period near the beginning of the growing season (1 May to 9 May), when soil moisture and light levels were not likely to severely limit Ec, is shown in Figure 5. As anticipated, a positive linear relationship existed between Ec and VPD (r2 = 0.70, p < 0.001). During this period, maximum half hourly Ec rates were limited to 0.07 mm hh−1, resulting in a plateau beyond 1.00 kPa, when midday Ec remained constant despite increasing VPD. This implies that Ec is being limited by VPD when neither light level or soil water supply are limiting factors.

Figure 5.

Relationship between daytime half-hourly canopy transpiration (Ec) and vapour pressure deficit (VPD) for 1 May–9 May. The dashed line represents the divide between the relationship when VPD < 1.00 kPa, and the relationship when VPD > 1.00 kPa. The solid line represents least squares regression (y = 0.077x − 0.022, r2 = 0.7, p < 0.001) between Ec and VPD when VPD < 1.00 kPa.

[31] In the early mornings, an inherent atmospheric stability and low radiation input occur at the site. Therefore we investigated if early morning VPD could be used as a potential integrator of recent weather trends that would characterize the condition of the atmosphere at the beginning of the diurnal transpiration cycle. This characterization of the atmosphere was then used to investigate typical diurnal EcVPD relationships and Ec magnitudes under differing soil water contents, but with similar initial conditions. Through this analysis we were able to determine to what extent increases in daily initial VPD (or the daily mean VPD between 6:00 and 8:00 hours (VPDin)) masked the influence of soil water content on Ec during dry periods.

[32] To illustrate the effect of VPDin on the relationship between Ec and VPD for the entire growing season we present diurnal ensemble averages of the relationship segregated into VPDin classes in Figure 6. Using half-hourly data from dry and intermediate days over the entire growing season (n = 5531), VPDin classes were defined as follows: low, VPDin < 0.005 kPa; moderate, 0.05 kPa ≥ VPDin ≤ 0.005 kPa; and high, VPDin > 0.05 kPa. A similar pattern in the relationship between Ec and VPD was observed in the ensemble averaged data, when compared with the set of selected days (11 to 18 June, Figure 4). The low VPDin class (n = 1251) tended to maintain the slope (0.6 mm hh−1 kPa−1) of the EcVPD relationship through the morning, but had higher mean afternoon Ec rates, with a maximum mean rate of 0.08 mm hh−1 at 15:00. The middle class (n = 3799) had a slightly reduced slope beyond 10:00 (0.11 mm hh−1 kPa−1), later reaching a maximum mean Ec rate of 0.07 mm hh−1 at 14:00 hours, but also had higher morning Ec rates compared to the low class. The high VPDin class (n = 481) tended to have a slightly steeper EcVPD slope (0.22 mm hh−1 kPa−1, compared to 0.11 mm hh−1 kPa−1) in the morning (with the “morning slope” in this case defined between 8:00 and 9:00 hours) and higher morning Ec rates relative to days with lower VPDin. The high VPDin slope then steadily reduced reaching an Ec plateau of approximately 0.05 mm hh−1 at 12:00.

Figure 6.

Ensemble averaged hourly diurnal relationship between canopy transpiration, Ec, and vapor pressure deficit (VPD), separated by daily initial VPD (or the daily mean VPD between 600 and 800 hours (VPDin)) class: low (VPDin < 0.02 kPa, open circles), moderate (0.05 kPa ≤ VPDin ≥ 0.02 kPa, closed circles), and high (VPDin > 0.05 kPa, open triangle). The closed star symbols mark 1200 hours. Arrows mark EcVPD slopes between 1100 and 1200 hours.

[33] EcVPD slopes between 11:00 and 12:00 (marked with arrows in Figure 6) progressively decreased with increasing VPDin class from 0.11 mm hh−1 kPa−1 (low VPDin) to 0.05 mm hh−1 kPa−1 and 0.03 mm hh−1 kPa−1 (low, mid and high VPDin, respectively). This correspondence of reductions in slope with increases in VPDin, on a day-to-day basis, suggested the following three conclusions: (a) that increases in early morning VPD tended to reduce the sensitivity of Ec to changes in mid-day VPD; (b) that increased early morning VPD may be causing reductions in the magnitude of maximum daily Ec rates, on a day-to-day basis; and (c) that, as early morning VPD increased, maximum daily Ec rates arrived earlier in the day. Again, as observed on selected days, the average Ec rates, when VPDin values were high, exhibited an increase in the early morning hours (between 6:00 and 8:00) with concurrent decreases in average VPD. However, note that while the maximum daily Ec rates decreased with increasing VPDin, mean daily Ec rates did not. This implied that the changes in the diurnal pattern of Ec, with rising VPD, may be compensatory and do not significantly affect daily tree water use. The mean daily Ec rate on dry days remained the same at 1.2 mm d−1, when progressing from the low VPDin class to the moderate class, and increased slightly to 1.4 mm d−1 for the high class.

[34] To determine the statistical significance of the VPDin classes on the EcVPD relationship we compared the coefficients of determination (i.e. R2) of the simple EcVPD model with the EcVPD model in which all three VPDin classes were accounted for (Table 4). Accounting for VPDin classes improved the model fit by 11% (i.e. R2 went from 0.44 up to 0.55). Furthermore, all slopes and intercepts between the three VPDin classes were found to be statistically different at p < 0.0001. From this we can conclude that the EcVPD relationship at the site varied depending on VPDin characterization. The slope of the EcVPD relationship increased from high to low VPDin class (Table 4), supporting our findings described above where we showed that Ec was most sensitive to incremental changes in VPD for low VPDin periods and least sensitive during high VPDin periods.

Table 4. Comparison of a Regression Analysis of the EcVPD Relationship to a Linear Multivariate Analysis of the EcVPD Relationship (Both Using Half-hourly Data From All Dry and Intermediate Day Over the Growing Season) to Test the Significance of Segregation by VPDin Classa
Ec versus VPD ModelEc versus VPD Model Which Accounted for VPDin Classes
VariableEstimated Coefficientt-ValueVariableEstimated Coefficientt-Value
  • a

    All estimated coefficients were significant at p < 0.001 (n = 5531).

Intercept0.00511.3Intercept (VPDin High)−0.010−4.7
Slope (VPD)0.03663.2Intercept (VPDin Mid)0.000
   Intercept (VPDin Low)0.01111.9
   Slope (VPDin High)0.031−6.9
   Slope (VPDin Mid)0.04367.7
   Slope (VPDin Low)0.0515.3
 R20.44 R20.55

[35] From our analysis (see Figures 3 and 6, and Table 4), it appears that increases in early morning VPD tend to reduce the sensitivity of Ec to changes in VPD, cause reductions in maximum daily Ec rates, and cause maximum daily Ec rates to arrive earlier in the day. The onset of the reduction in this slope tended to occur earlier, as VPDin increased. This reduction in slope often resulted in an Ec rate plateau and was likely an expression of the limitation of transpiration in response to environmental changes in water availability, or increases in temperature.

[36] One of the most interesting observations of the study was that the limitation of transpiration was not only sensitive to the atmospheric demand at the onset of the daily course of transpiration, but also to the soil moisture deficit as well. The relationship between VPD and Ec for different soil water contents was investigated and initially no significant differences were found using a simple linear regression (p > 0.05). We recognized however, that VPD and soil water content are not necessarily independent of one another and that low soil water contents can cause diurnal time shifts in maximum transpiration rates relative to maximum VPD [Kurpius et al., 2003]. We also recognized that a response to soil water deficits might be obscured by changes in Ec rates, caused by a change in the daily pattern of VPD over drying soils. It was only when VPDin classes were accounted for that a change in the response of Ec was observed under differing θ0−25cm values. Figure 7 illustrates the relationship between Ec and VPD for all growing season days, where the VPDin value was below 0.02 kPa (the low VPDin class), separated between periods when the θ0–25cm was above 0.07 and below 0.07. This demarcation of soil water content was found by observing the relationship between Ec and VPD under iteratively decreasing soil water content states. This point was identified by a Ec plateau at low soil water contents, similar to the clear leveling-off response to high VPD. Interestingly, the chosen soil water content demarcation roughly corresponded with the above-mentioned plateau value observed in θ0–25cm during dry periods. This plateau was also corresponded with the observed onset of the apparent nocturnal increases in θ0–25cm during periods with hydraulic redistribution. The relationship between Ec and VPD, in both form and control, was similar to the one shown earlier in Figure 5. Again, a positive linear relationship existed between Ec and VPD (r2 = 0.64, p < 0.001 and r2 = 0.97, p < 0.001, when θ0–25cm was above 0.07 or below 0.07, respectively) until a threshold VPD value was reached, after which a plateau with strictly limited Ec rates formed. The VPD transition point and the Ec rate plateau were both considerably lower, when θ0–25cm was below 0.07. This relationship was upheld until a VPD of approximately 0.60 kPa for θ0–25cm above 0.07 and until a VPD of approximately 0.40 kPa for θ0–25cm below 0.07. While the slopes and intercepts of the two regression equations were similar, the plateau of observed Ec rates for days of low soil water content (i.e. θ0–25cm > 0.07) was limited to approximately 0.06 mm hh−1, compared to approximately 0.10 mm hh−1 when the soil water content was higher in the well-drained sandy soils at our site (see Figure 7). A similar response was observed for all three VPDin classes, but was most clearly expressed for the low VPDin class.

Figure 7.

Relationship between canopy transpiration (Ec) and vapour pressure deficit (VPD) for all growing season days, where the VPDin value was below 0.02 kPa (low VPDin case), separated between periods when the root zone soil moisture, θ0−25cm, was above 0.07 (open circles) and below 0.07 (closed circles). Dashed line represents least squares regression (y = 0.12x − 0.0013, r2 = 0.64, p < 0.001), when the θ0–25cm was above 0.07, while solid line represents least squares regression (y = 0.1x − 0.0017, r2 = 0.97, p < 0.001), when the θ0–25cm was below 0.07.

4. Discussion

4.1. Water Fluxes

[37] The observed daily E and Ec values declined in response to cloudy periods, and also showed a gradual decline over the course of the growing season in response to shorter day-length and reduced incoming solar radiation. On a daily basis, Ec consistently lagged behind E. This lag continued until the end of the day when Ec often exceeded E, indicating that the tree tissues were replenished after transpiration ceased. This has been readily observed in forest transpiration studies and implies that the water storage capacity of living tissues is also important in this forest [Unsworth et al., 2004]. The highest fluxes (both E and Ec) were recorded on hot summer days immediately following precipitation events, and the lowest rates were recorded on cool, cloudy, wet days in the spring. The maximum daily E rate of 5.8 mm d−1 at the Turkey Point white pine forest site was somewhat larger than many other mature temperate conifer forests located in wet environments (4.4 mm d−1 [Barbour et al., 2005]; 4.0 mm d−1 [Grelle et al., 1997]; 3.7 mm d−1 [Humphreys et al., 2003] and 3.6 mm d−1 [Unsworth et al., 2004]). Two extremely high daily water losses were recorded during the study. Without these two values, the maximum observed E rates would have been approximately 4.0 mm d−1. Both extremes (5.7 mm on 15 July and 5.8 mm on 4 August) corresponded with high mean daily VPDs (greater than 1.3 kPa) and occurred less than two days after rain storms. It is likely that these extreme estimates were the result of a large evaporation contribution from the wet canopy, forest floor, and soil.

[38] The estimated S for the forest was similar to other high-LAI conifer forests [Link et al., 2004]. The S of a Douglas-fir stand with an LAI of 8.6 was 2.7 mm [Link et al., 2004] and was 2.4 mm for another Douglas-fir stand with an LAI of 9–13 [Klaassen et al., 1998]; both similar to the S of 2.4 mm and LAI of 8.0 found at this site. Growing season EI accounted for 18% of PG, well within the range of observed net interception losses in temperate conifer forests [Link et al., 2004]. Hormann et al. [1996] published values of interception loss from a number of interception studies, showing temperate conifer forests having interception losses ranging from 9–48% of gross precipitation, depending on the size of the precipitation event. Despite large fluctuations in monthly PG (62 mm to 221 mm), the relatively large contribution of monthly EI to E and narrow range of observed monthly EI (21 mm to 28 mm) illustrated the influence of canopy storage capacity on forest water loss and interception under varying storm size. As storm size increases, proportional interception loss decreases [Gash and Morton, 1978; Link et al., 2004], resulting in EI accounting for 33% of PG in June, the month with the least amount of rain, and EI accounting 13% of PG in September, the month with the greatest amount of rain (see Table 3).

[39] The water status of a plant is related to both soil water supply and atmospheric demand. Interception is an important control on the amount of available soil water. Interestingly, during the drought of late June and early July, 50 mm of rain fell. However, due to the low intensity of the rainfall events (never exceeding 7 mm hh−1), a greater proportion of rain was intercepted. It is likely that the relatively high water holding capacity of the canopy (2.4 mm) is exacerbating periodic droughts in the forest growing on well drained sandy soils. The high LAI and S of this forest make the water balance at the site particularly sensitive to reductions in precipitation characteristics, such as frequency, duration or intensity, because the resulting reduction in net precipitation would be amplified by the canopy storage capacity.

[40] On a daily basis, the remainder of evaporation (after subtracting Ec from E) cannot always be accounted for by modeled EI, due to the different temporal frameworks used to make the estimates. Our estimates of EI on the event scale must be compared to fluxes on a daily scale. The reconciliation of these two scales is not possible because rain events and the evaporation of water afterward are not always confined to a single calendar day. For this reason EI usually exceeds E on wet days, but is under-estimated on the days immediately following rainfall events, when E usually greatly exceeds Ec. This occurs because modeled estimates of EI are given on a per rain event basis rather than on a daily basis, with reported EI representing the sum of the evaporation of water intercepted during and after the event, but can only be reported as the sum for the day the rain event began. It should be noted that for any significant rain event, a large proportion of the reported EI would have likely evaporated within hours following rainfall.

[41] The range of Ec rates in our forest was similar to those reported from other studies conducted in mature wet temperate conifer forests. Unsworth et al. [2004] reported mean summer Ec rates of 1.5 mm d−1 and 1.4 mm d−1 (June and July). Barbour et al. [2005] reported a range of Ec rates of between 0 and 1.8 mm d−1 in their study on a temperate conifer rainforest and Irvine et al. [1998] report the same range for a temperate Scots pine (Pinus sylvestris L.). Apart from the responses to storm events, there was a distinct lack of seasonality and variability in Ec rates at our forest site compared to the variability exhibited in other studies. This was also evident in the daily ensemble Ec averages for each month of the study, which showed little change until the beginning of the dormant season. However, small but noticeable declines of Ec rates in late June and early July did show a broad response to reductions in soil water content. This relative stability was an indication of the importance of the physiological and climatic limitations on canopy transpiration. It also lends support to the idea that for a single species the range of variation in transpiration is likely to be dampened by an overall coupling with climate and a strong negative feedback between stomatal conductance and VPD [Roberts, 1983].

[42] The proportion of Ec to E was quite low (47%) in this forest when compared to other conifer forests. Oren et al. [1998] reported that Ec accounted for 69% of E in a temperate Loblolly pine (Pinus taeda) plantation, Unsworth et al. [2004] reported 65% in a temperate Douglas fir–Western hemlock (Pseudotsuga menziesiiTsuga heterophylla) old growth forest, and Grelle et al. [1997] reported 75% in a boreal mixed conifer (Picea abiesPinus sylvestris L.) forest. Two studies however, did find similar partitioning in conifer forests: one in a Ponderosa pine (Pinus ponderosa) plantation in California, USA, found that 53% of E came from Ec [Kurpius et al., 2003]; the other in a temperate old growth coniferous (Dacrydium cupressinum) rainforest, in New Zealand, reported that 39% of E was derived from Ec over the growing season and 51% on dry days [Barbour et al., 2005]. Both of these forests had much smaller LAIs than our forest, 2.2 and 2.9 respectively.

[43] The low Ec proportion and high LAI of our forest imply that we under-estimated stand Ec, because the unmeasured components of soil evaporation (Es) and understorey transpiration (Eu) are unlikely to entirely account for the 49% of forest water lost on dry days from this site. The likely reason for the possible underestimation is the difference between the spatial scales involved in scaling-up of sapflow measurements from the 20 × 20 m study plot to the stand level for estimating canopy transpiration. The eddy covariance technique measures water fluxes over a large variable “footprint” source area, but due to practical and economic restrictions sapflow had to be measured at a much smaller fixed local scale and thus can only estimate the Ec of the dominant species. Our initial assumption that the sapflow from the dominant white pine would represent the response of all species in the forest adequately, and therefore suffice to represent total canopy transpiration, was inadequate. The confounding factor was the unmeasured contribution from the young emergent deciduous trees (or gap species, which make up at least 8% of the species composition) and patches of heavy understorey growth distributed unevenly throughout the forest. Though we have no way of estimating their individual contributions, all three sources (gap species transpiration, understorey transpiration and soil evaporation) do contribute and can plausibly account for the remaining 49% of forest water lost on dry days from this site. Further support for this reasoning may be found in the seasonality of the divergence between E and Ec estimates and its relationship to the approximated deciduous phenology. The difference between the two estimates was relatively small during the earliest part of the growing season (Ec was 73% of E from 1 May to 26 May). This difference grew in late May, corresponding with the timing of gap deciduous species full leaf date, and then shrank again in October, corresponding with the onset of senescence. Hogg et al. [1997] similarly reported that eddy covariance measurements showed a rapid increase in water flux from a deciduous aspen canopy in late May, due to rapid leaf expansion. This reasoning is also supported by the good agreement of our Ec rates with the typical transpiration rates from wet temperate conifer forests dominated by a single species reported by previous studies [Barbour et al., 2005; Unsworth et al., 2004; Irvine et al., 1998]. It is likely that the unmeasured gap species at our site contributed significantly to the total water flux measured at the tower from June to September, though presently we have no way of estimating their contribution due, in part, to their uneven spatial distribution. We assumed that errors associated with the eddy covariance evaporative flux measurements did not contribute significantly to the underestimation. Another possible reason for the seemingly low transpiration proportion of this forest maybe shoot morphology. Chen et al. [2006] observed that both within-shoot and beyond-shoot clumping was particularly high in this forest. The reported needle-to-shoot area ratio of 1.91 is considerably larger than the needle-to-shoot area ratio values reported for several conifer forests across Canada [Chen et al., 2006]. Beyond-shoot clumping value (0.98) for this site was also high when compared with the other sites investigated by Chen et al. [2006]. Therefore, despite high LAI (8.0), the understorey and soil of this forest may have been exposed to higher levels of radiation compared to other forest ecosystems with high LAI, resulting in relatively higher evaporative losses. However, this hypothesis needs to be confirmed and should be considered in future studies at this site.

[44] These findings highlight the difficulties in adequately measuring canopy transpiration in a mature planted conifer forest, where the natural succession toward a mixed-wood forest has begun. The results also provide strong support for the simultaneous use of sapflow and eddy covariance measurements to monitor forest water fluxes. Since the errors of the two techniques are different and generally independent, their simultaneous measurements helped to identify possible problems and methodological weaknesses [Hogg et al., 1997]. Despite these limitations, we were still able to account for the majority (81%) of the total growing season evapotranspiration from the forest by estimating the two largest components (transpiration from the dominant white pine and interception evaporation).

[45] During the driest periods in which θ0−25cm would approach the seasonal minimum and the water table depth was presumably at its deepest, the shallow soil layers began to gain more noticeable amounts of water at night without any apparent atmospheric input. A possible reason is the presence of a strong water potential gradient between non-adjacent soil layers during these periods, leading to the passive nocturnal redistribution of soil from deeper depths to the root-zone, sometimes referred to as Hydraulic Redistribution [Brooks et al., 2002]. This recharge may explain why soil water contents were maintained above those associated with the approximate wilting point through dry periods at the site (see Figure 2). We hypothesize that night-time increases during extremely dry conditions, which were large enough to change net water use (calculated as the difference between the maximum soil water storage (mm) of a single day and the maximum soil water storage of the following day), were likely due to hydraulic redistribution. Hydraulic redistribution has been shown to occur in a number of different soil types, climates and species [Brooks et al., 2002; Burgess et al., 1998; Caldwell et al., 1998; Dawson, 1993; Oliveira et al., 2005; Unsworth et al., 2004]. Numerous possible benefits have been proposed as a consequence of hydraulic redistribution, including; providing water to shallow rooted seedlings and understorey plants, enhancing mineral nutrient availability, enhancing microbial processes and heightening the acquisition of nutrients by roots, by keeping the fine root zone hydrated [Caldwell et al., 1998].

4.2. Controls and Limitations on Canopy Transpiration

[46] Shifts in the timing and magnitude of Ec rates, caused by increased VPDin, masked an important relationship between Ec rate and θ0−25cm. Several studies have indicated that the pattern of the limitation responses of Ec to high VPD and to soil water deficit appeared similar and were often concurrent, sometimes making interpretations difficult [Kurpius et al., 2003; Oren and Pataki, 2001]. Our results suggest that VPDin could be used as an analytical tool to investigate the combined effects of soil water content and atmospheric demand on transpiration.

[47] It has been postulated that Ec rates are linked to VPD through stomata, which work to maintain leaf-needle water potential above a critical minimum value, thereby, limiting maximum Ec rates [Hogg and Hurdle, 1997]. Our results seem to support this hypothesis, though we did not test it using a direct measurement of leaf water potential. Since our Ec rates reflect water uptake into the stem and not actual transpiration rates at the needles, it is likely that leaf water potentials are reaching critical minimums earlier in the day, and at lower water uptake rates, when early morning VPD is high. This is because when VPD is high at the beginning of the diurnal transpiration cycle, the initial rates of water uptake are inherently lower regardless of VPD and can be exceeded by lower rates of leaf water loss than usual. Thus the critical minimum leaf water potentials that signal a stomatal response (and hence the daily maximum Ec rates) are met earlier in the day. There is a growing body of evidence supporting the notion of isohydric regulation (the homeostasis of minimum leaf water potential) of plant water status [Bucci et al., 2005; Buckley, 2005; Brodribb and Holbrook, 2006; O'Grady et al., 2008]. A recent study by O'Grady et al. [2008] found that as pre-dawn leaf water potentials in a Tasmanian Eucalyptus plantation declined, in response to soil desiccation and increasing VPD, the difference between pre-dawn leaf water potentials and midday leaf water potentials declined, in turn causing declines in both transpiration and canopy conductance. Their observations lead to a conclusion that there was a strong case for the existence of a critical minimum leaf water potential as soil water deficits increase. Similarly, Bucci et al. found minimum leaf water potentials in a Brazilian savanna (Cerrado) were isohydric, but that pre-dawn leaf water potentials decreased as soils dried.

[48] Though the maximum Ec rates observed were shown to decrease and occur earlier in the day as VPDin increased, mean daily Ec rates remained largely unchanged, indicating that the trees were still able to maintain relatively high levels of daily transpiration, even when VPD's were quite high. The shifting of the diurnal center of Ec into the morning hours has been linked to a specific water use strategy in pine trees. In pines, transpiration rates are maximized in the morning on warm dry days, rather than throughout the day thus maximizing water use efficiency during optimal conditions for stomatal opening (i.e. when light levels (PPFD) are close to maximum, but atmospheric demand (VPD) has not yet peaked) [Kurpius et al., 2003]. Similarly, our findings showed that the timing of peak Ec shifted toward the morning on days when the air was warm and dry. Furthermore, this shifted peak was followed by a steady, but reduced Ec, relative to days when the early morning atmospheric demand was not as high. We found a general correspondence between high morning VPDs and early peaks in Ec, during the driest period of the study (June). Our results suggest that, during dry periods, the dominant white pines at our site moderate the timing of stomatal behavior to maintain a constant rate of Ec with changes in VPD. This interpretation of generalized stomatal function supports the hypothesis that stomatal behavior is shaped by selection, such that the underlying control mechanisms approach a quantifiable goal [Buckley, 2005], where by transpiration is being constrained to a critical rate by the need to both maximize water use efficiency and avoid cavitation at the same time.

[49] As a result of these findings, we hypothesize that a feedback exists between VPD and Ec at our site, whereby initial daily increases in VPD cause increases in Ec. In turn, increases in Ec cause decreases in leaf water potential. If VPD continues to increase, the decreasing leaf water potentials will eventually meet a critical minimum, prompting a stomatal response, lowering the maximum obtainable Ec rate, which ultimately results in Ec rates leveling-off with subsequent increases in VPD. Further to this point of isohydric regulation, we also hypothesize that either an increase in the magnitude of this negative feedback on Ec with increasing mean early morning VPD, or increasing soil water deficit, would similarly prompt stomata to close progressively sooner and sooner. Likewise, Buckley [2005] made the case that stomata respond similarly to any perturbation in the hydraulic continuum of a tree (i.e. including either soil water supply or atmospheric demand) because the effect on leaf water status is the same. It is thought that plants respond to the cumulative effects of daily weather over extended periods [Schwartz et al., 2006], and it may also be the case that trees can respond to the cumulative effects of daily weather on short time scales in this manner.

[50] The fact that diurnal Ec rates were significantly affected by atmospheric demand at the onset of the daily course of transpiration is an important finding of our study. The shifts in the timing and magnitude of daily Ec rates, caused by increased early morning VPD, were masking the relationship between transpiration and root zone soil water content. The reason behind the masking was likely that high mean early-morning VPDs and low soil water contents have much the same effect on the relationship between water loss at the leaf surface and water uptake at the root surface. Both can cause a reduction in leaf water potential, signaling a similar stomatal response [Buckley, 2005]. This suggests that the actual degree of VPD is not as critical to stomatal operation as is the balance between the rate of water loss at the crown and the rate of water uptake into the stem. It is likely that critical minimum leaf water potentials can be met at any VPD and could explain the relative insensitivity of Ec to changes in soil water content over short time scales, reported in several past studies (mentioned above). This being said, the results outlined in detailed literature reviews suggest that there doesn't seem to be a single response mechanism that can explain all features of stomatal behavior and it is likely that stomatal operation is controlled by a combination of multiple mechanisms, including root-shoot signaling (via abscisic acid or ABA), leaf water status, atmospheric demand and soil water supply [Jones, 1998].

5. Conclusions

[51] As expected, transpiration from the dominant White pine was the largest component of forest water flux to the atmosphere (47%) when the canopy was dry, while the evaporation of intercepted water dominated the flux when the canopy was wet. Together, these two fluxes accounted for the majority (81%) of the water flux from the forest over the growing season. Transpiration from the dominant white pines was controlled by VPD, until a variable transition value was reached, after which mid-day transpiration rates remained relatively constant. Transpiration rates that were independent of VPD were limited to approximately 0.10 mm hh−1 through the study period and this limitation was sensitive to early morning atmospheric demand and soil water deficit. The relationship between transpiration and atmospheric demand is broadly sensitive to soil water supply, but didn't appear to be sensitive to fine variations in soil water content. Increases in early morning atmospheric demand caused maximum transpiration rates to arrive earlier in the day and to be reduced in magnitude, but did not affect mean daily transpiration rates significantly. Transpiration rates were sustained at relatively high levels through periods of high atmospheric demand. This was likely because stomata were operating to maintain leaf-needle water potentials above a critical minimum value and limiting Ec rates in response to increasing VPD. On days when the early morning VPD was relatively high, these critical minimums were met earlier in the day. Atmospheric demand and soil water supply often co-vary during warm dry periods and likely interact with one another via their mutual influence over the stomatal response to reductions in leaf water potentials, which directly limits Ec. A better understanding of Ec limitation could help refine transpiration models and link Ec to a feedback that includes a mechanistic connection between declining soil water contents and increasing VPD. The evidence that Ec rates were responding to changes in early morning VPD, on a day-to-day basis, indicated that trees can respond to the cumulative effects of weather over short time periods through the same feedback between VPD and Ec that causes Ec to level-off with increasing VPD over longer timescales. This finding warrants further investigation. An improved understanding of the response of Ec to cumulative weather would allow better assessments of the effects of future climate changes on water budgets, at varying temporal and spatial scales. This study advances our understanding of the physical mechanisms limiting forest water use, particularly in planted forests and in forests situated in moist environments overlying well drained soils, and could also help refine or parameterize transpiration models.

Appendix A

[52] The Gash model describes EI based on an analysis of individual storm events. The evaporation from a saturated canopy (Ew) is estimated from the Penman–Monteith equation [Monteith, 1965]. Ew (kg m−2 s−1) is calculated for half hours, when the PG is greater than 2.25 mm using:

equation image

where Δ is the slope of the saturation vapor pressure curve at air temperature (kPa °C−1); Rn is net radiation (W m−2); ρ is the density of air (kg m−3); Cp is the specific heat of air at constant pressure (J kg−1 °C−1); γ is the psychrometric constant (kPa °C−1); and λ is the latent heat of vaporization of water (J kg−1). In the model, evaporation after rainfall, evaporation from storms too small to fully saturate the canopy, and the wetting-up of the canopy are all treated as separate terms [Gash, 1979]. Working under the assumption that those half hours with PG ≥ 2.25 mm represented conditions where the canopy was fully saturated, these values were then averaged to estimate the mean evaporation rate from a saturated canopy during rainfall:

equation image

where m represents the number of half hour periods (which are denoted as the ith period), when PG ≥ 2.25 mm. The mean rainfall rate (equation image) during those same half hour periods is given as:

equation image

The rainfall amount at the point at which the canopy reaches saturation (PG) is then given by:

equation image

Assuming that these mean rates apply, and that stem-flow contributed a negligible amount to evaporation during storm events and need not be included, it follows that the evaporation generated by a storm event large enough to saturate the canopy, EIj (denoted as the jth event) is then given by:

equation image

and EIk, the evaporation generated by a storm event that was not large enough to saturate the canopy (denoted as the kth event) is given by:

equation image


[53] Funding for this study was provided by the Natural Sciences and Engineering Research Council (NSERC) of Canada Discovery and Strategic Project Grants. Support from the Canadian Foundation of Innovation (CFI), the Ontario Innovation Trust (OIT), and McMaster University is also acknowledged. In-kind support from the Canadian Carbon Program (CCP), Ontario Ministry of Natural Resources (OMNR), the Long Point Recreation and Conservation Authority (LPRCA), and the Canadian Forest Service (CFS) of Natural Resources Canada is appreciated. Thanks to Miles McLaren, Gabe Thompson, Wojtek Stepien, Natalia Restrepo-Coupé, Mahmoud Pejam, David Spittlehouse, Alan Cameron and Steve Williams for their advice and support for Turkey Point Flux Station data collection.