Stand-level gas-exchange responses to seasonal drought in very young versus old Douglas-fir forests of the Pacific Northwest, USA

This study examines how stand age affects ecosystem mass and energy exchange response to seasonal drought in three adjacent Douglas-fir (Pseudotsuga menziesii (Mirb.) Franco) forests. The sites include two early seral (ES) stands (0-15 years old) and an old-growth (OG) (~450-500 years old) forest in the Wind River Experimental Forest, Washington, USA. We use eddy covariance flux measurements of carbon dioxide (FNEE), latent energy (^E) and sensible heat (H) to derive evapotranspiration rate (ET), Bowen ratio (B), water use efficiency (WUE), canopy conductance (Gc) , the Priestley-Taylor coefficient (x) and a canopy decoupling factor ( ). The canopy and bulk parameters are examined to find out how ecophysiological responses to water stress, including changes in relative soil water contentrs.) and vapour pressure deficit (de), differ among the two forest successional stages. Despite different rainfall patterns in 2006 and 2007; we observed site-specific diurnal patterns of ET, x, Gc, tie and 0r during both years. The largest stand differences were (1) at the OG forest high morning Gc (> 10 mm S) coincided with high net CO2 uptake (FNEE = -9 to -6 umol m·-2 S -1 ), but a strong negative response in OG Gc to moderate tie was observed later in the afternoons and subsequently reduced daily ET and (2) at the ES stands total ET was higher (+ 72 mm) because midday Gc did not decrease until very low water availability levels (0 r


Introduction
The landscape of the Pacific Northwest, USA is dominated by tall, long-lived evergreen conifer species that are well adapted to a distinct seasonal climate. This climate regime includes a cool and wet season, and a warm and dry season that are determined by the locations of the Aleutian low during the wet winter months and the Pacific high during the dry summer. Climate change scenarios currently predict that while the Pacific Northwest region will likely receive more precipitation during the winter than it does now, the summers are expected to be warmer and drier creating ecophysiological stress on forest communities and species (Mote et al 2005). Some ecologists predict that a more intense summer drought will generally have a greater impact on Pacific Northwest tree species, including the dominant low-elevation species Douglas-fir (Pseudotsuga menziesii var menziesii (Mirb.) Franco) and western hemlock (Tsuga heterophylla (Raf.) Sarg.), than the expected annual temperature increases (Mote et al 2003). Enhanced water stress is a particular concern for the regeneration of Douglas-fir stands in the Western Cascade Mountains. This is an area where the youngest age class makes up 40% of the total forest coverage on harvested lands (Cohen et al 1996). Mature trees in intact, closed stands will likely be partially buffered from increased drought stress (Waring and Franklin 1979), but early seral (ES) conifer species (< 15 years old) may have trouble surviving prolonged, extremely dry summers.
In Douglas-fir/western hemlock forests, stomatal conductance and net photosynthesis are strongly dependent on the canopy's microclimate and decline throughout the course of the drought season as vapour pressure deficit (de) increases and soil water content (0v) decreases (e.g., Waring and Franklin 1979, Winner et al 2004, Falk et al. 2008, although stand age certainly affects this response. Stomatal control is an important driver of variation in the surface energy budget of forests (Stewart 1988) because stomatal aperture controls how available energy is partitioned between sensible heat and latent heat (or transpiration). Partial stomatal closure restricts water vapour exchange between the leaf and atmosphere and consequently increases leaf temperature and sensible heat transfer. Douglas-fir and western hemlock foliage have the ability to induce stomatal closure and conserve water in the short term by limiting transpiration but this occurs at the expense of reducing the photosynthetic rate, and over the long term, it also reduces tree growth and wood production (Bower et al. 2005).
In developing forest stands, productivity is closely dependent on available soil water because (1) soil moisture determines the growth of foliage biomass and (2) soil moisture stress affects the photosynthetic efficiency of that foliage (Jarvis and Mullins 1987). Prior studies give us an idea of how evergreen needleleaf trees in southern Washington survive the annual, 5-month drought event. Site water availability and tree water use and demand measurements have been taken in a 20-year-old Douglas-fir and the 450-year-old Douglas-fir/western hemlock forest in the Wind River Experimental Forest, Washington, USA. Warren et al. (2005) showed that soil moisture in the shallow 0.15-0.60 m soil layer declined by 40-45% during the summer drought period, while deep soil (2 m) released only 5% of its original volume. The importance of hydraulic redistribution (HR) at these forest sites, whereby roots are able to lift water from deeper, moist soil horizons and release it into shallower, drier soil portions is reinforced in the work by Domec et al. (2004).Their data showed that HR at the highest rates replenished 60% of the previous day's water use and was critical for maintaining shallow root function and preventing total stomatal closure during drought conditions in the Douglas-fir stands. Brooks et al. (2002) further found that the effects of summer drought in the 20-year-old Douglas-fir forest were partially muted by HR, which accounted for an additional 16 days of stored water to remain in the upper soil horizons after a 60-day drought.
Although experiments have shown that mature stands have access to deep water reserves during drought periods, foliage at the top of very tall trees often exist near critical values for cavitation, a condition largely set by the distance between the water table and the hydraulic capacity of the xylem Yoder 1997, Ryan et al. 2006). High hydraulic path-length resistance may decrease productivity in taller, older trees because stomata in tall trees are more often closed than in the younger, shorter trees and consequently carbon gain is reduced during the midday hours (see 'hydraulic limitation hypothesis', Ryan and Yoder 1997). Several branch-level, gas-exchange studies have shown that stomatal conductance decreases with increasing tree age and height (Yoder et al. 1994, Ryan et al. 2000, while other studies have shown that stomatal conductance measurements fail to conclusively support the hydraulic limitation to gas-exchange hypothesis (Bauerle et al. 1999, Phillips et al. 2002, McDowell et al. 2005. Identifying any universal stand age-effect responses from these types of studies needs to be done with caution because 'branchlevel' data are scaled up to the 'stand level' based on short measurement periods and small sampling sizes (e.g., Raulier et al. 2000).
Our study uses the eddy covariance (EC) technique that makes measurements of microclimate, and mass and energy fluxes between the whole forest ecosystem and the atmosphere. We collected stand-level data over two growing seasons (May to October 2006 and2007) at the Wind River old-growth (OG) forest and at Early Seral North (ESN) in 2006 and Early Seral South (ESS) in 2007. All our study sites were either dominated (the ES stands) or co-dominated (the OG stand) by Douglas-fir, an extremely long-lived (maximum age~1000years old) pioneering species and the three stands are representative of the youngest and oldest Douglas-fir successional stages. Our objectives were to (1) assess how summer-time reductions in soil moisture availability and increases in vapour pressure deficit influence ecosystem-atmospheric carbon dioxide (FNEE)' evapotranspiration rate (ET) and energy exchange in two , distinct forest age classes and (2) compare site ecophysiological responses (e.g., canopy conductance, the Priestley-Taylor coefficient and canopy decoupling factor) to microclimate conditions during the seasonal drought period.

Site description OC forest and canopy crane
The 87-m-tall canopy crane is in the Thorton T. Munger Research Natural Area (RNA) (45°49'13.76" N and 121°57'06.88" W; 371 m a.s.l.), a preserved 478 ha section of OG forest in the Gifford Pinchot National Forest, Washington, USA. The RNA is classified as part of the Western Cascades Lowlands and Valleys ecoregion with a wet, mild climate and forests that are dominated by Douglas-fir and western hemlock (Thorson et al. 2003). Topography within the RNA is gentle and elevations range from 335 to 610 m ). The stand is co-dominated by Douglas-fir (mean height = 52 m and maximum height = 65 m), the colonizing tree species, and western hemlock (mean height = 19 m and maximum height = 55 m) (Ishii et al. 2000), a shade tolerant, climax species, which will eventually dominate the stand. Other tree species found in the T.T. Munger RNA include the conifers: western red cedar(Thuja plicata Donn.), noble fir (Abies procera Rehd.), grand fir (Abies grandis (Dougl.) Forbes), western white pine (Pinus monticola Dougl.), Pacific silver fir (Abies amabilis (Dougl.) Forbes) and Pacific yew (Taxus brevifolia Nutt.); and the angiosperms: cascara buckthorn (Rhamnus purshiana (DC.) Cooper), Pacific dogwood (Comus nuttallii (Aud.) Torr.) and red alder (Alnus rubra Bong.). Ground species include salal (Gaultheria shallon Pursh), vine maple (Acer circinatum Pursh), Oregon-grape (Berberis nervosa Pursh), vanillaleaf (Achlys triphylla (Smith) DC.), bracken fern (Pteridium aquilinum (L.) Kuhn), sword fern (Polystichum munitum (Kaulf.) C. Presl) and deer fern (Blechnum spicant (L.) Roth) (Franklin 1972, DeBell and Franklin 1987. The vertical canopy is structurally complex due to a mixture of tree ages (0 to ~ 500 years old) among shade-tolerant and shade-intolerant species. Strong vertical temperature and humidity gradients are periodically present , Pyles et al. 2004, Falk et al. 2005 and are indicative of low aerodynamic mixing conditions throughout the canopy and high boundary-layer resistance around the foliage. Overhead canopy gap fraction is 0.52 (Parker et al. 2002) and leaf area index (LAI) measurements range from 8.2 to 9.2 m 2 m-2 Winner 2000, Parker et al. 2004) and are divided unevenly among the middle and upper canopy (4.8 m 2 m -2), lower canopy (2.i m 2 m-2 ) and understory (1.7 m 2 m-2 ). Western hemlock and western red cedar represent more than half (55%) of the stand LAI but Douglas-fir foliage dominate the upper canopy (Thomas and Winner 2000), the driest and brightest microenvironment, and have disproportionate control over ecosystem mass and energy exchange, including transpiration (Lewis et al. 2000, Winner et al. 2004. Deciduous vegetation are insignificant components of stand biomass and during the summer months are no more than 15% of the canopy LAI (Thomas and Winner 2000).

ES canopies
The ESN (45°49'37.2 N and 121°57'39.6"W; 361 m a.s.l.) is a re-established evergreen needleleaf forest on a 7 ha clear-cut patch, 1.25 km northwest of the canopy crane. This stand represents a third generation Douglas-fir ecosystem: the original OG forest was logged in 1920 and a clear-cut harvest was done in 1994 on the 80-year-old Douglas-fir trees. In 1997, the stand was seeded with Douglas-fir saplings at 741 trees ha-1 . The second most common tree was the deciduous species red alder, which was found predominately in the southern portion of the stand. Western hemlock and western white pine seedlings were also present but in insignificant amounts. Douglas-fir height and diameter at breast height (d.b.h.) measurements were taken in September 2005: mean height = 4.4 m, height range = 1.2-5.3 m and mean d.b.h. = 5.7 cm. Ground cover species were diverse in the summer months and included salal, Oregon-grape, bracken fern, sword fern and blackberry (Rubus ursinus Cham. & Schltdl.).
The ESS stand (45°48'47.4" N and 121°57'32.9" W; 371 m a.s.l.) is an abandoned clear-cut patch (1990), 1.1 km southwest of the canopy crane, and was naturally established with Douglas-fir from surrounding cone crops. In July 2007, the average Douglas-fir tree height was 3.58 ± 0.84 m and d.b.h. was 4.47 ± 1.6 cm (n = 95 trees). The estimated tree density was 1063 trees ha-I 961 (biomass survey included eight plots of 100 m 2 each). Tree cores showed that stand-representative Douglas-fir trees were between 9 and 12 years of age in 2007 (n = 10 trees). Other tree species included western white pine, red alder, and planted Pacific silver fir and Pacific yew seedlings, each in insignificant amounts. Bracken fern was the dominant ground species from May to September. Grasses and scotch broom bushes (Cytisus scoparius (L.) Link) were also common in the more open areas.
LAI eff is effective,single-sided LAI and was calculated using the software program, x. is woody-to-plant ratio and was set at 0.20, YE is needle-to-shoot ratio and was set at 1.61 and E is the foliage element clumping index and was set at 0.91 (parameter values were based on the measurements taken by Chen 1996, Chen et al. 2006). The hemispheric photographs were taken at a height of 10 ern with a Nikon COOLPIX E4300 digital camera adapted with a Nikon Fisheye Converter lens (Nikon Inc., Melville, NY). For logistical reasons DHP surveys were done just once at both sites. The photographs were taken just past sunset on 1 September 2006 at ESN and 30-31 August 2007 at ESS07. Fern and other ground species were cleared before the photographs were taken to ensure that only trees were included in the canopy LAI estimates. At ESN, 15images were taken along a 150 m west-to-east transect (centred on the flux tower) at 10 m intervals. At ESS, 17 images were taken along a 170 m west-to-east transect at 10 m intervals. DHP average estimates of canopy LAI were 1.5 m 2 m-2 at ESN and 0.9 m 2 m-2 at ESS. Separate LAI measurements of the herbaceous/understory species were not taken at ESN. ESS LAI of ground species (predominately bracken fern) was estimated from DHP and was between 0.4 and 0.5 m 2 m-2 .

Belowground description
Coarse roots of mature Douglas-fir extend 1-2 m deep although most root biomass in the OG stand is concentrated within the first 0.5 m of the soil profiles . At the ES stands coarse roots were observed down to 0.5 m and fine roots were primarily in the 0-0.3 m soil layer. Area soils are medial, mesic, Entic Vitrands and are 2-3 m deep, well drained and derived from volcanic material . These soils are classified as silt loams and are generally stone-free, high in organic material and nitrogen deficit in the root zone at the Wind River sites. Stand-specific soil properties are listed in LAI was indirectly measured at the ES stands using digital hemispheric photography (DHP), and estimated using HemiVicw 2.1 (Delta-T Devices Ltd., Cambridge, UK) and the following equation (Chen 1996, Chen et al. 1997 both spatially and temporally heterogeneous at the OG forest and ranges from 0.3 to 0.5 m in the winter months to 2.0-2.4 m in the dry summer months ). Water table depth was not measured at the ES stands but is assumed to be within the range found at the OG forest. Field capacity (0v-10 kPa, 0v at matrix potential = -10 kPa) for this soil type is 0.30 m 3 m-3 , permanent wilting point (0v-1500 kPa, 0v at matrix potential = -1500 kPa) is 0.14 m 3 m-3 and volumetric available water content of the soil (0v, where 0v =0v-1O kPa -0v-1500 kPa) is 0.16 m 3 m-3 (Lambers et al. 2008). For root-zone depths of 1.5 and 0.6 m, respectively, at the OG forest and ES stands, maximum available water storage was estimated to be 240 mm (OG) and 96 mm (ES) (By multiplied by root-zone depth).

Instrumentation and flux calculations
OG forest Ecosystem carbon dioxide, water vapour and energy fluxes were measured using EC methodology (see, e.g., Wofsy et al. 1993, Hollinger et al. 1994, Goulden et al. 1996,Paw U et al. 2000). The EC system consisted of a sonic anemometer (Solent HS, Gill Instruments, Lymington, UK) and a closed-path infrared gas analyser (IRGA) (LI-7000, Li-Cor Inc., Lincoln, NE), which measured the wind velocity vectors and air temperature, and concentrations (mixing ratios) of water vapour (H20) and carbon dioxide (C02), respectively, at 10 Hz. The IRGA and sonic anemometer were mounted on a horizontal boom extending off the canopy crane at a height of 67 m so that the anemometer faced west, the predominant wind direction and direction of greatest homogenous fetch (> 1 km). Carbon dioxide (Fco, , umol CO2 m-2 S -1 ), sensible heat (H, W m-2 ) and latent energy (^E, W m -2 ) fluxes were computed with FORTRAN90 code using a 30-min averaging period and a horizontal coordinate rotation. The rate of change in CO2 concentration (storage flux, Sea" umol C02m-2 S-I) within the canopy volume was estimated using 30-min changes in the mean CO2 mixing ratio measured at the top of the canopy (Falk et al. 2008). To account for any CO2 stored within the canopy and below the detection height of the instruments, Sco, was added to F co2 to estimate the net ecosystem exchange of carbon (F NEE umol CO2 m-2 S -1 ) on a half-hourly basis. Half-hour measurements of F NEE and ^E were further screened for outliers and gap-filled (16% of data in 2006 and 11% in 2007) using a running-mean and look-up table approach DROUGHT RESPONSE IN TWO DOUGLAS-FIR AGE CLASSES (Reichstein et al. 2005). We report no nighttime flux data in this study except in the daily evapotranspiration sums. For further details on the EC post-processing, refer to Paw U et al. (2004) and Falk (200S).
Meteorological instrumentation at the canopy crane included air temperature/relative humidity (sheltered HMP-35C, Vaisala Oyj, Helsinki, Finland) and incident photosynthetically active radiation sensors (PAR) (190-SB, Li-Cor Inc.) mounted at heights of 2 (below canopy measurement) and 70 (above canopy measurement) m along the crane tower, and a four-stream (up-and downwelling and short-and long-wave radiation) net radiometer (CNR 1, Kipp and Zonen, Delft, The Netherlands) was mounted at a height of 85 m. Additionally, soil temperature was measured at depths of 0.05, 0.10 and 0.15 m (CS106B, Campbell Scientific Inc., Logan, UT). Ground heat flux was measured with a HFT-3.1 soil heat flux plate (Radiation and Energy Systems, REBS, Seattle, Washington) buried 0.075 m below the surface. The meteorological measurements were collected as 30-min averages and were logged continuously from May 2006 to October 2007.

ESN and ESS
Ten hertz measurements of horizontal (u and v) and vertical (w) wind velocity and air temperature were made using a CSAT-3 sonic anemometer (Campbell Scientific Inc.), and densities of CO2 and H20 vapour were measured with an open-path fast response IRGA (LI-7500, Li-Cor Inc.). F co2 ^E and H were calculated in real time using a 30-min averaging period with the CRl000 EC program (Campbell Scientific Inc.). F co2 , and ^E were also corrected for any measurement errors associated with density fluctuations (Webb et al. 1980, referred to as WPL80 corrections). During post-processing, all scalar and energy fluxes were re-calculated after the mean cross-wind (v) and vertical wind (w) velocities were rotated to zero (following the natural wind coordinate system). The rate of change of CO2 concentration (S c02 ) within the canopy was estimated using the half-hourly changes in the CO2 mixing ratio measured at the top of the canopy and was added to F co2 to estimate FNEE. Half-hour scalar and energy fluxes were quality controlled for non-preferred wind directions, inadequate fetch (using a parameterized footprint model), low turbulence conditions (determined by a ratio of mean wind velocity to a turbulent energy velocity scale; methodology described in Wharton et al. 2009), heavy precipitation events and times of general instrument failure. Missing or excluded scalar and energy fluxes were gap-filled using a running-mean approach (Reichstein et al. 2005).
At ESN, the sonic anemometer was mounted facing westsouthwest with a fetch of 210 m over homogeneous Douglas-fir trees. Both LI-7500 and CSAT-3 were mounted at 5.5 m a.g.l., 1.1 m above the ESN canopy, on a boom extending from a 6 m tall tower. Tower-based micrometeorological data included half-hour measurements of air temperature/relative humidity (sheltered HMP-35C, 963 Vaisala Oyj), net radiation (Q7.1, REBS) and soil temperature (O.O5, 0.10 and O.15 m) (CS106B, Campbell Scientific Inc.). Ground heat storage was calculated from the soil temperature profiles. Fluxes and micrometeorological data were collected at ESN from May to October 2006.The instrument setup at ESS was identical except that the LI-7500 and CSAT-3 were mounted at 5 m a.g.l., 1.4 m above the canopy facing south (170°) and up-and down-welling PAR sensors  were added at the top of the tower at a height of 6 m a.g.l. EC data at ESS were collected May to August 2007 (EC fluxes end in August due to instrument failure) and meteorological data from May to October 2007.
A simple, parameterized footprint model (Kljun et al. 2004, http://www.footprint.kljun.net/index.php) was used to determine the extent of which measured turbulent fluxes were influenced by scalar sources outside the ES stands. The model showed that daytime (10:00-14:00) footprint estimates ranged from 75 m (east upwind direction) to 100 m (north upwind direction) at ESN and 77 m (east upwind direction) to lIS m (north upwind direction) at ESS, translating into fetch-to-Be instrument height ratios of 14:1-23:1. Available (i.e., homogeneous vegetation) fetch-to-EC instrument height ratios averaged 33:I and 34:1 at ESN and ESS, respectively, but ranged from 10:1 to 44:1 depending on the wind direction. Half-hour fluxestaken under inadequate fetch conditions were removed from the data series, as well as wind directions at ESN where red alder made up a significant portion of the flux footprint.

Water supply and demand measurements
Direct measurements of canopy water availability included precipitation and volumetric soil moisture. Precipitation (P, mm day -1 ) was measured at the Carson Fish Hatchery (CFH) National Oceanic and Atmospheric Administration (NOAA) weather station (45°31'12" N and 121°34'48" W; 345.6 m a.s.l.) using a rain and snow gauge (385 heated, Met One Instruments, Inc., Grants Pass, Oregon). The NOAA weather station is located 5 km north of the canopy crane in the Wind River Valley at a similar elevation. Soil water content (0v, m 3 m-3) at the OG forest was measured over an integrated depth of 0-0.30 m (three replicates) in 2006 with a time-domain reflectometry (TDR) system (TDR100,Campbell Scientific Inc.) and in staggered depths down to 2 m in 2007 with Sentek soil moisture probes (four replicates) (Sentek EnviroSMART, Scntek Sensor Technologies, Stepney, Australia). At ESN, 0v was measured with the TDR100 system over integrated depths of 0-0.30 and 0.30-0.60 m (two replicates). The TDR100 system was moved to ESS in 2007, where 0v was measured over integrated depths of 0-0.30, 0.30-0.60 and 0.60-0.90 m (two replicates). Our study reports soil moisture measurements in relative soil water content (0v), where 0 r = (0v -0v -1500 kPa)/(0v-10 kPa -0v-1500 kPa). 0 r is a dimensionless number that ranges from 0 to 1 and represents the fraction of available water in the soil root zone. 964 WHARTON ET AL.
Water vapour loss from the canopy was estimated using the 30-min averaged ^E to calculate the evapotranspiration rate (E r mm half hour -1 ), and summed over daily and monthly intervals. Equilibrium evapotranspiration rate (ETeq, mm half hour -1 ) was calculated based on the energy-balance technique (^E + H = Rn -S, where S = energy storage flux) and a modification of the Penman equation (Penman 1948) following Denmead and McIlroy (1970), In Eq. (2), ^is the slope of the saturation vapour pressure curve (kPa K-I ), Y is the psychrometric constant (kPa Ksure of the climatologically expected evapotranspiration rate over a moist surface based only on temperature and available energy. The Priestley-Taylor coefficient, X, a ratio of measured E T to equilibrium ET, was calculated using (Priestley and Taylor 1972) The Priestley-Taylor coefficient is site dependent and varies with surface vegetation (Denmead 1969) and microclimate conditions, including soil water availability (Slatyer and Denmead 1964, Priestley and Taylor 1972, Black 1979. Equation (3) gives a maximum X value of 1 assuming that there is no upwind advection of heat added to the system (e.g., an 'oasis effect'). Magnitudes of X approach one as E T approaches ETeq, and measured evapotranspiration is largely controlled by aerodynamic resistance (Ra = 1/G a) and the subsequent water vapour gradient between the canopy surface and atmosphere, and less by canopy resistance (R c = 1/G c ) to water vapour transfer. As X approaches zero, the measured evapotranspiration rate is less than the expected, energylimited rate (E Teq ) and ecosystem water loss is strongly controlled by canopy resistance (i.e., the degree of stomatal closure) to surface-atmospheric water exchange.

Bulk canopy and mechanistic measurements
Bulk canopy and mechanistic variables including water use efficiency (WUE), canopy conductance and a canopy decoupling factor were calculated for daytime (downwelling shortwave radiation > 10 W m-2) half-hour periods only. The WUE is defined as the total mass of dry matter produced by photosynthesis for every kilogram of water lost by vegetation through transpiration (e.g., Rosenberg et al. 1983). Here, we defined a midday WUE as the ratio of F NEE (g C m-2 half hour -1 ) to E T (kg H20 m-2 half hour -1 ), averaged between the hours of 10:00 and 15:00 when ^E was greater than zero (following Berbigier et al. 2001). Following Stewart (1988), canopy conductance (Gc) was estimated using the inverted Penman-Monteith equation (Monteith 1964): In Eq. (4), Gc is canopy conductance (m s-1 ), p is air density (kg m-3 ), cp is specific heat (J kg-1 K-1 de is vapour pressure deficit (kPa), ^E is latent energy (W m-2), B is the Bowen ratio (H/^E) and G a , is aerodynamic conductance for momentum transfer (Ga = "U 2 /U, m S-1, where u* is friction velocity, m S -1 and U is mean wind speed, m S-1). Equation (4) includes both a canopy conductance-driven component (first term, right-hand side (RHS) of Eq. (4)) and a radiation-driven component (second term, RHS of Eq. (4)) so that the proportion of E T controlled by the two drivers can be represented by a canopy decoupling factor, ^ (Jarvis and McNaughton 1986): where ^is a dimensionless number that ranges from 0 to 1 depending on whether E T is controlled strongly by Gc and de (^approaches 0) and is an aerodynamic-driven process, or whether E T is determined by the amount of available energy (Rn -S) to the canopy (^approaches 1). If a canopy is completely dry at the surface and one assumes that E T is approximately equal to the transpiration flux, then ^refers to the degree to which transpiration is uncoupled to atmospheric de. In a forest canopy where surface roughness is high, ^is mostly dependent on wind speed, and gas exchange will be strongly coupled to atmospheric saturation conditions (^< 0.2). As soil moisture decreases, ^also decreases and canopy air coupling is enhanced (Jarvis and McNaughton 1986).

Climate and stand microenvironments
Water  (Reichstein et al. 2005). We report no nighttime flux data in this study except in the daily evapotranspiration sums. For further details on the EC post-processing, refer to Paw U et al. (2004) and Falk (2005). Meteorological instrumentation at the canopy crane included air temperature/relative humidity (sheltered HMP-35C, Vaisala Oyj, Helsinki, Finland) and incident photosynthetically active radiation sensors (PAR) (190-SB, Li-Cor Inc.) mounted at heights of 2 (below canopy measurement) and 70 (above canopy measurement) m along the crane tower, and a four-stream (up-and downwelling and short-and long-wave radiation) nct radiometer (CNR 1, Kipp and Zonen, Delft, The Netherlands) was mounted at a height of 85 m. Additionally, soil temperature was measured at depths of 0.05,0.10 and 0.15 m (CS106B, Campbell Scientific Inc., Logan, UT). Ground heat flux was measured with a HFT-3.l soil heat flux plate (Radiation and Energy Systems, REBS, Seattle, Washington) buried 0.075 m below the surface. The meteorological measurements were collected as 30-min averages and were logged continuously from May 2006 to October 2007.

ESN and ESS
Ten hertz measurements of horizontal (u and v) and vertical (w) wind velocity and air temperature were made using a CSAT-3 sonic anemometer (Campbell Scientific Inc.), and densities of CO2 and H20 vapour were measured with an open-path fast response IRGA (LI-7500, Li-Cor Inc.). F c02 ^E and H were calculated in real time using a 30-min averaging period with the CR1000 EC program (Campbell Scientific Inc.). F CO2 an ^E were also corrected for any measurement errors associated with density fluctuations (Webb et al. 1980, referred to as WPL80 corrections). During post-processing, all scalar and energy fluxes were re-calculated after the mean cross-wind (v) and vertical wind (w) velocities were rotated to zero (following the natural wind coordinate system). The rate of change of CO2 concentration (S co2 ) within the canopy was estimated using the half-hourly changes in the CO2 mixing ratio measured at the top of the canopy and was added to F C02 to estimate FNEE. Half-hour scalar and energy fluxes were quality controlled for non-preferred wind directions, inadequate fetch (using a parameterized footprint model), low turbulence conditions (determined by a ratio of mean wind velocity to a turbulent energy velocity scale; methodology described in Wharton et al. 2009), heavy precipitation events and times of general instrument failure. Missing or excluded scalar and energy fluxes were gap-filled using a running-mean approach (Reichstein et al. 2005).
At ESN, the sonic anemometer was mounted facing westsouthwest with a fetch of 210 m over homogeneous Douglas-fir trees. Both LI-7500 and CSAT-3 were mounted at 5.5 m a.g.l., 1.1 m above the ESN canopy, on a boom extending from a 6 m tall tower. Tower-based micrometeorological data included half-hour measurements of air temperature/relative humidity (sheltered HMP-35C, 963 Vaisala Oyj), net radiation (Q7.l, REBS) and soil temperature (0.05,0.10 and 0.15 m) (CS106B, Campbell Scientific Inc.). Ground heat storage was calculated from the soil temperature profiles. Fluxes and micrometeorological data were collected at ESN from May to October 2006. The instrument setup at ESS was identical except that the LI-7500 and CSAT-3 were mounted at 5 m a.g.l., 1.4 m above the canopy facing south (170°) and up-and down-welling PAR sensors  were added at the top of the tower at a height of 6 m a.g.l. EC data at ESS were collected May to August 2007 (EC fluxes end in August due to instrument failure) and meteorological data from May to October 2007.
A simple, parameterized footprint model (Kljun et al. 2004, http://www.footprint.kljun.net/index.php) was used to determine the extent of which measured turbulent fluxes were influenced by scalar sources outside the ES stands. The model showed that daytime (10:00-14:00) footprint estimates ranged from 75 m (east upwind direction) to 100 m (north upwind direction) at ESN and 77 m (east upwind direction) to 115 m (north upwind direction) at ESS, translating into fetch-to-EC instrument height ratios of 14:1-23:1. Available (i.e., homogeneous vegetation) fetch-to-EC instrument height ratios averaged 33:1 and 34:1 at ESN and ESS, respectively,but ranged from 10:1 to 44:1 depending on the wind direction. Half-hour fluxes taken under inadequate fetch conditions were removed from the data series, as well as wind directions at ESN where red alder made up a significant portion of the flux footprint.

Water supply and demand measurements
Direct measurements of canopy water availability included precipitation and volumetric soil moisture. Precipitation (P, mm day -1 ) was measured at the Carson Fish Hatchery (CFH) National Oceanic and Atmospheric Administration (NOAA) weather station (45 o 31'12" N and 121 o 34'48" W; 345.6 m a.s.l.) using a rain and snow gauge (385 heated, Met One Instruments, Inc., Grants Pass, Oregon). The NOAA weather station is locatcd 5 km north of the canopy crane in the Wind River Valley at a similar elevation. Soil water content (0v m 3 m -3 ) at the OG forest was measured over an integrated depth of 0 0.30 m (three replicates) in 2006 with a time-domain reflectometry (TDR) system (TDR100, Campbell ScientificInc.) and in staggered depths down to 2 min 2007 with Sentek soil moisture probes (four replicates) (Sentek EnviroSMART, Sentek Sensor Technologies, Stepney, Australia). At ESN, 0v was measured with the TDR100 system over integrated depths of 0-0.30 and 0.30-0.60 m (two replicates). The TDR100 system was moved to ESS in 2007, where 0v was measured ovcr integrated depths of 0-0.30, 0.30-0.60 and 0.60-0.90 m (two replicates). Our study reports soil moisture measurements in relative soil water content (0r), where 0r = (0v -0v-1500 kPa)/(0c-10 kPa-0v-I500 kPa). 0r is a dimensionless number that ranges from 0 to 1 and represents the fraction of available water in the soil root zone.
Water vapour loss from the canopy was estimated using the 30-min averaged E to calculate the evapotranspiration rate (ET, mm half hour ), and summed over daily and monthly intervals. Equilibrium evapotranspiration rate (E q, mm half h -) was calculated ased on the energy-balance technique (^E + H = Rn -S, where S = energy storage f ux) and a modification of the Pemnan equation (Penman 1948) following Denmead and McIlroy (1970), In Eq. (4), Gc is canopy conductance (m S -), p is air density (kg m-3 ), cp is specific heat (J kg -), be is vapour pressure deficit (kPa), E is latent energy (W m-2), is the Bowen ratio ( ) and Ga is aerodynamic conductance for momentum transfer (Ga = u,2/U, S -1 , where u* is friction velocity, m S-1 and U is mean wind speed, m S-1). Equation (4) includes both a canopy conductance-driven component (first term, right-hand side (RHS) of Eq, (4)) and a radiation-driven component (second term, RHS of Eq. (4)) so that the proportion of E T controlled by the two drivers can be represented by a canopy decoupling factor, ^(Jarvis and McNaughton 1986): In Eq, (2), A is the slope of the saturation vapour pressure curve (kPa K -), y is the psychrometric constant (kPa ), Rn is the net radiation (W m-2 ) and S g is the ground heat storage fiux (W m-2 ). Equilibrium E T is a measure of the climatologically expected evapotranspiration rate over a moist surface based only on temperature and available energy. The Priestley-Taylor coefficient, «; a ratio of measured E T to equilibrium E T, was calculated using (priestley and Taylor 1972) where ^is a dimensionless number that ranges from 0 to 1 depending on whether E T is controlled strongly by Gc and be (^approaches 0) and is an aerodynamic-driven process, or whether E T is determined by the amount of available energy (Rn -S) to the canopy (^approaches 1). If a canopy is completely dry at the surface and one assumes that E T is approximately equal to the transpiration flux, then ^refers to the degree to which transpiration is uncoupled to atmospheric de. In a forest canopy where surface roughness is high, Q is mostly dependent on wind speed, and gas exchange will be strongly coupled to atmospheric saturation eonditions (Q < 0.2). As soil moisture decreases, ^also decreases and canopy air coupling is enhanced (Jarvis and McNaughton 1986).
The Priestley-Taylor eoeffieient is site dependent and varies with surface vegetation (Denmead 1969) and microclimate conditions, including soil water availability (Slatyer and Denmead 1964, Priestley and Taylor 1972, Black 1979. Equation (3) gives a maximum r:J. value of I assuming that there is no upwind advection of heat added to the system (e.g., an 'oasis effect'). Magnitudes of approach one as approaches ETeq, and measured evapotranspiration is largely controlled by aerodynamic resistance (Ra = ) and the subsequent water vapour gradient between the cano surface and atmosphere, and less by canopy resistance Rc = ) to water vapour transfer. As X approaches zero, the measured evapotranspiration rate is less than the expected, energylimited rate (ETeq) and ecosystem water loss is strongly controlled by canopy resistance (i.e., the degree of stomatal closure) to surface-atmospheric water exchange.

Bulk canopy and mechanistic measurements
Bulk canopy and mechanistic variables including water use efficiency (WUE), canopy eonductance and a canopy decoupling factor were calculated for daytime (downwelling shortwave radiation > l W m -2 ) half-hour periods only. The WUE is defined as the total mass of dry matter produced by photosynthesis for every kilogram of water lost by vegetation through transpiration (e.g., Rosenberg et al. 1983). Here, we defined a midday WUE as the ratio of F NEE (g C m-2 half hour -) to E T (kg H20 m-2 half h -), averaged between the hours of 10:00 and 15:00 when E was greater than zero (following Berbigier et al. 2001). Following Stewart (1988), canopy conductance (Gc) was estimated using the inverted Penman-Monteith equation (Monteith 1964):

Climate and stand microenvironments
Water Relative available soil water content also varied among stands and years although the seasonal drought pattern remained a dominant feature. In 2006, near-surface (0-0.30 m) 0r equaled one at both ESN and OG during the spring months and began declining between June and July. Relative water availability dropped to a minimum of 0.27 in August at ESN and to 0.26 in September at OG (Table 2). During the 2007 summer months, nearsurface 8r did not reach these low levels. 8r dropped to a weekly minimum value of 0.67 at ESS in September and 0.52 at OG in August. While the near-surface water availability was less at the OG stand than at the ES stand, the 0.9-2.0 m depth 0v measurements in 2007 revealed that deeper soil layers were not water deficient (0r = 1) at OG. At ESS, the deepest 0v measurements (0.6-0.9 m) showed that relative water availability was not significantly higher in this soil layer and available water content was nearly identical to the near-surface measurements from July to September.

Diurnal and monthly fluxes
Net radiation was higher at tile OG stand than at either ES stand during the months of May to August. Monthly Rn averaged 470 MJ m -2 mo -I in 2006 and 468 MJ m-2 mo -1 in 2007 at OG, and 410 MJ m-2 mo-I at ESN and 393 MJ m-2 mo -1 at ESS (Tables 3 and 4).
During the summer months at the OG stand, a greater amount of available energy was on average partitioned into sensible heat (average daily maximum = 350 W m-2) than latent heat (average daily maximum = 200 W m-2 ). The May to August Bowen ratios at OG were higher than those observed at either ES stand and ranged from 2.05 (June) to 2.61 (July) in 2006 and 1.94 (June) to 2.S8 (May) in 2007. Peak daytime latent heat fluxes were constant at the OG stand from May to August with the exception of a middaŷ E decline in July 2006. A more distinct ^E pattern was observed at the ES. stands with peak ^E occurring in June and July (Figures 1B and 2B). July E T in 2006 was nearly twice as great at ESN (103 mm mo -1 ) than at OG (53 mm mo -1 ) (Table 3), while a smaller increase over 2007 OG E T was also measured at ESS (Table 4).
During May, midday CO2 fluxes were more than twice as great at the OG stand as at either ES stand (Figures 1C and  2C). Midday net CO2 uptake at the OG stand peaked in June and declined throughout the latter summer months at OG. At the ES stands, the greatest differences in midday CO2 fluxes occurred between the months of May and June. Mean midday FNEE increased from -3.2 to -9.0 umol m-2 S-1 at ESN and from -4.6 to -8.7 umol m-2 S-1 at ESS during this period. A lag of `2 h occurred in the timing between daily peak flux exchange at ESS and OG in 2007 but not at ESN and OG in 2006. This time lag created a longer period of net CO2 uptake in May at the OG stand but a reduced period of CO2 uptake at OG during .Tulyas compared to ESS (compare boxed regions in Figure 2C). Peak ^E at ESS also occurred later in the afternoon than at OG from May to August 2007 resulting in higher total daily E T at the younger stand. 966 WHARTON ET AL.

Ecosystem response to water stress Precipitation. Ov and de effects on ET
A ratio ofprecipitation to evapotranspiration ( ) showed that all forest canopies lost more water via evapotranspiration than gained from precipitation during the months of June to August (Tables 3 and 4). Precipitation was assumed equal at all stands so any differences were due only to variations in canopy evapotranspiration.
In 2006, we observed very low (P/E T ) values of just 0.01 at the young stand and 0.02 at the older forest in July and August due to nearly zero precipitation. The ESN was likely water stressed by September 2006 as relative soil moisture fell to 20% in the root zone. Nearsurface 0r at the OG stand was also extremely low and deeper water measurements were not made that year to accurately determine water availability throughout the entire root zone.
On average, the 2007 summer was wetter and larger stand differences were observed between the OG and ES stand. From May to August, (P/ET) was 0.35 (more water limited) at ESS and 0.42 (less water {imited) at OG.
A time series of daily maximum de and daily total E T at ESS and the OG forest is shown in Figure 3. de was consistently higher at ESS during the 2007 summer, often by > 1 kPa (circled data points in Figure 3). In June, higher de coincided with increased canopy water loss (1-2kg H20 m-2 day-1) at the ES stand, while in August, daily E T was moderate and closer to E T observed at OG (e.g., compare squared data points in Figure 3) even though de remained higher at the ES site. Figure 4 shows that E T was not related to de at the OG forest (R2 = 0.0) during the month of June in both 2006 and 2007 while a stronger relationship between ET and be was observed at ESN (R2 = 0.47) and ESS (R 2 = 0.78). ET at the OG forest was more closely linked to vapour pressure deficit in August (R 2 = 0.6 in 2006 and R2 = 0.34 in 2007) than during early summer.
For equivalent daily maximum de (e.g., 2 kPa) total daily ET was smaller in August (1.5-2.5 kg H20 m-2 day-1) than in June (2.5-3.5 kg H20 m-2 day-1) at all forests. Summer x ranged from a low of 0.35 at OG (July) to a high of 0.74 at ESN (July) in 2006 and from 0.39 (July) at OG to 0.63 (June) at ESS in 2007 (Tables 3 and 4). Overall, both ES stands had higher Priestley Taylor coefficients (~0.5-0.6) than the OG forest (~0.4) during the drought seasons. The Priestley-Taylor coefficient increased logarithmically with canopy conductance in all stands (e.g., R2 = 0.34 at ESN) so that measured E T was closer to equilibriurn ET at the higher Gc values. The relationship between x and relative available soil water was less straightforward ( Figure 5). At the OG gorest and ESS a correlation between the Priestley-Taylor coefficient and 0r was not found (R2 = 0.0). At ESN x dropped from an average of 0.73-0.55 as 0r conditions decreased and approached 20%.

Bulk canopy parameter and mechanistic responses
We measured higher WUE at the OG stand than at either of the ES stands. Mean midday WUE during the summer TREE PHYSIOLOGY VOLUME 29, 2009 DROUGHT RESPONSE IN TWO DOUGLAS-FIR AGE CLASSES 967 drought was 2.5 ± l.l g C kg-1 H20 at OG and 1.6 ± l.0 g C kg-1 H 2 0 at ESN in 2006 (Table 3), and 2.2 ± l.0 g C kg-1 H 2 0 at OG and 1.5 ± 0.7 g C kg-1 H20 at ESS in 2007 (Table 4). The OG stand was slightly more water use efficient in 2006 than in 2007 and consistently more water use efficient than either of the ES stands. Figure 6 shows average canopy conductance from May to October at the OG and ES stands in 2006 and 2007. Overall, higher values of Gc were observed in 2006 than in 2007 and Gc magnitudes were higher at the beginning of the drought season than at the end. Site differences in diurnal Gc values were also observed. Canopy conductance was higher at the ES stands during the afternoon hours than at the OG forest, while morning G; was often higher at the mature forest. Canopy conductance began declining earlier in the day at OG suggesting that stomates are shutting down at lower be levels in the OG Douglas-fir/western hemlock forest than in the 4 m tall ES Douglas-fir trees. Figure 7 further details the difference between mean canopy conductance in May and July 2007 at ESS and OG during the hours of 10:00, 13:00 and 16:00. In May, G; declined at sirnilar rates (~0.2 rum S-1 per half hour) at both stands and averaged 7.4 mm s-1 (OG) and 7.6 mm S-1ESS) at 10:00, 6.4 mm S-1 (OG) and 6.5 mm S-1 (ESS) at 13:00, and 4.7 mm s-1 (OG) and 4.6 mm 8-1 (ESS) at 16:00. In contrast, August Gc declined more rapidly at the OG stand (0.3 mm S-l per half hour) than at ESS (0.1 mm S-1 per half hour) beginning at the noon hour. By early evening, Gc averaged less than 4 mm S-1 at OG but remained around 6 mm s-1 at ESS.
A close look at canopy conductance in Figure 8 reveals both site and monthly differences in leaf-atmosphere gasexchange response to evaporative demand at ESS and OG. At low de values ( < 0.5 kPa), the OG stand had higher Gc (> 2-3 mm S-1) than the young stands but beyond values of 1-l.5 kPa, Gc at OG rapidly declined with increasing Be even though available soil moisture was not low (e.g., 0r < 50%) in May and June. In general, Gc was lower in July and August than in May and June at both stands for all vapour pressure deficit levels below 2.5 kPa. Beyond 2.5 kPa, the rate of canopy conductance decline with increasing vapour pressure deficit was strongest at OG. The minimum de threshold to produce very little response in Gc (i.e.,~~~approaches 0) was 3 kPa and 4 kPa at ESS in May to June and in July to August. The OG stand had no observable de-Gc threshold response in May to June but was 3.5 kPa in July to August. Figure 9 illustrates how the de-Gc responses at OG and ESS differ drought was 2.5 ± 1.1 g C kg-1 H20 at OG and 1.6 ± 1.0 g C kg-1 H 2 0 at ESN in 2006 (Table 3), and 2.2 ± 1.0 g C kg-1 H20 at OG and 1.5 ± 0.7 g C kg-1 H20 at ESS in 2007 (Table 4). The OG stand was slightly more water use efficient in 2006 than in 2007 and consistently more water use efficient than either of the ES stands. Figure 6 shows average canopy conductance from May to October at the OG and ES stands in 2006 and 2007. Overall, higher values of Gc were observed in 2006 than in 2007 and Gc magnitudes were higher at the beginning of the drought season than at the end. Site differences in diurnal Gc values were also observed. Canopy conductance was higher at the ES stands during the afternoon hours than at the OG forest, while morning Gc was often higher at the mature forest. Canopy conductance began declining earlier in the day at OG suggesting that stomates are shutting down at lower oe levels in the OG Douglas-fir/western hemlock forest than in the 4 m tall ES Douglas-fir trees. Figure 7 further details the difference between mean canopy conductance in May and July 2007 at ESS and OG during the hours of 10:00, 13:00 and 16:00. In May, Gc declined at similar rates (~0.2 mm S-1 per half hour) at both stands and averaged 7.4 mm S-1 COG) and 7.6 mm s-1 (ESS) at 10:00, 6.4 mm S-1 (OG) and 6.5 mm S-1 (ESS)at 13:00, and 4.7 mm s-J COG)and 4.6 mm s-1 (ESS)at 16:00. In contrast, August Gc declined more rapidly at the OG stand (0.3 mm S-1 per half hour) than at ESS (0.1 mrn S-1 per half hour) beginning at the noon hour. By early evening, Gc averaged less than 4 mm S-1 at OG but remained around 6 mm S-1 at ESS.
A close look at canopy conductance in Figure 8 reveals both site and monthly differences in leaf-atmosphere gasexchange response to evaporative demand at ESS and OG. At low e values 0.5 kPa), the OG stand had higher G c (> 2-3 mm S-) than the young stands but beyond values of 1-1.5 kPa, Gcat OG rapidly declined with increasing oe even though available soil moisture was not low (e.g., < 50 n May and June. In general, Gc was lower in July and August than in May and June at both stands for all vapour pressure deficit levels below 2.5 kPa. Beyond 2.5 kPa, the rate of canopy conductance decline with increasing vapour pressure deficit was strongest at OG. The minimum e threshold to produce very little response in Gc (i.e.,~~~approaches 0) was 3 kPa and 4 kPa at ESS in May to June and in July to August. The OG stand had no observable e-Gc threshold response in May to June but was 3.5 kPa in July to August. Figure 9 illustrates how th e-Gc responses at OG and ESS differ highest Be levels (2-3.5 kPa), while ESS Gc dropped only from 5 to 3 mm S-1 (grey triangles in Figure 9). In contrast, in July and August, ESS Gc declined sharply from 4 to 2 rum s-I across the highest Be levels (3.0-4.5 kPa), while Gc remained at 2 rum S-I at the OG forest (grey circles in Figure 9).
At very high e (> 4.5 kPa), canopy conductance was low (about 2 mm S-) in all stands but we found that evapotranspiration was more directly related to available energy than to stomatal control as shown by higher decou-pIing coefficients (e.g., at ESN, = 0.27 for e > 4.5 kPa and = 0.l4 for 2.5 < e < 3.5 kPa). The decoupling coefficients at the OG stand in 2006 were on average (^= 0.18) equal to those at ESN (n = 0.18), while OĜ values in 2007 were significantly higher (^= 0.31) than in 2006 and higher than those found at ESS (^= 0.16).

Discussion and conclusions
Douglas-fir/western hemlock stands in the Pacific Northwest have adapted to seasonal moisture constraints on photosynthesis by assimilating large amounts of carbon during the wetter and cooler spring months while during the summer, stomatal closure is induced as vapour pressure gradients between the leaf and atmosphere increase (Waring and Franklin 1979). Our study showed that Gc was notably reduced after the noon hour in mid-summer at the OG forest regardless of soil moisture availability. While relative near-surface soil moisture was noticeably lower at the OG forest in 2006 (0r~25%) than in 2007 (0r~55%), the mature stand likely had access to deep soil water supplies during both drought seasons because (1) relative soil TREE PHYSIOLOGY ONLINE at http://www.treephys.oxfordjournals.org from the beginning of the drought season to conditions at the end. In May and June 2007, canopy conductance declined sharply at OG from 4 to 1 mm S-1 across the moisture never dropped below 90% at the 1 m depth when these measurements were available, (2) the root zone extends down to 2 m for the oldest trees and (3) there is prior evidence of HR in the soils at this stand (Brooks et al. 2002, Warren et al. 2005. The phenomenon of stomatal closure in OG trees regardless of soil moisture has also been noted by Zweifel et al. (2002) in a 250-year-old Norway Spruce stand, whereby they observed midday stomatal closure on most sunny days during permanently wet soil conditions. Even under moderate de levels, foliage at the tops of tall evergreen conifer trees often reach near critical values for cavitation due to a long path distance between the water table and the hydraulic capacity of the xylem, and as a result shut their stomata frequently (Ryan and Yoder 1997). Our observation of a rapid decline in Gc between 0.5 and 1.0 kPa in the older forest is consistent with the findings of Ryan and Yoder (1997) as well as with the observations of a diurnal hysteresis in sap flow measurements taken at the bottom and top of tall Douglas-fir trees at Wind River by Cermak et al. (2007). Taken together, these datasets suggest that upper canopy Douglas-fir foliage are driving our observed monthly and diurnal fluxes of mass and energy at the 0G forest. At the ES stands, we found that canopy conductance was also inhibited by vapour pressure deficit but not until later in the afternoon hours when the highest be levels were reached and later in the drought season when relative soil moisture was low. Our overall Gc observations (e.g., canopy conductance was 2-4 mm S-1 higher at the ES stands than at the OG agree with branch-level measurements taken by Yoder et al. (1994), whereby they found that stomatal conductance is lower in the Wind River OG trees than in the younger Douglas-fir stands, and by Fessenden and Ehleringer (2002) who, using d 13 C isotopes, found evidence that decreased hydraulic conductance in the 450-year-old stand led to lower stomatal conductance in the mature forest than in the younger, shorter trees. The Be was almost always lower at the top of the OG forest than over the ES stands suggesting that the upper canopy stomates are generally closing at lower de levels in the older stand.
In addition to our successional-stage research, gasexchange studies comparing a Wind River 20-year-old Douglas-fir stand and the OG forest were made in 1998 and 1999 by Chen et al. (2002Chen et al. ( , 2004. In their study, Chen et al. (2002) report higher WUE (1.7 g C kg-1 H 2 0) and Bowen ratio (2.9) at a Wind River 20-year-old stand than at the OG forest (WUE = 1.0 g C kg-1 H 20, B = 1.6), while total E T was greater at OG. In contrast, we found higher B, higher WUE and lower E T at the OG forest than the ES stands. Lower WUE values have also been measured for the 10-year-old Douglas-fir age class by Thomas and Winner (2002), whereby they found that WUE was twice as high at the OG forest than at nearby younger trees using 'branch-level' measurements. We suggest that the conflicting nature of succession-stage, gas-exchange results at Wind River is a consequence of (1) a rapidly growing Douglas-fir canopy during the first 0-20 years, (2) a rapid shift in secondary species distributions in the first 0-20 years and (3) different measurement techniques (e.g., 'stand-level' versus 'branch-level') and measurement periods were used in the studies.
It is important to note that other plant species, particularly bracken fern, were a non-negligible component of ecosystem biomass at the younger stands during the summer months. Ground species certainly played a role in determining the carbon, water and energy budgets at the ES stands although we were not able to quantify how significant that role might be. At the OG forest, the canopy is diverse (eight evergreen conifer and three angiosperm species are present) and structurally complex because of this mixture of shadetolerant and shade-intolerant species. Our measurements of gas exchange are for the forest ecosystem as a whole and as such we were not able to determine what contribution each speciesmade to ecosystem fluxes. Therefore, we are making the following conclusions based on the two successional stages investigated in this study and not specifically on tree species, tree height or tree age: (I) Total evaporation (and fraction of expected ET) is higher in ES stands than at the OG forest during the summer months as a result of higher soil temperatures, higher air temperatures and higher vapour pressure deficits in the open canopies during the mid-afternoon hours. Available soil moisture limited ET at the ES stands but only when 0r dropped below 30%. Total E T was limited at the OG stand during moderate vapour pressure deficits and moderate soil water availability.
(2) The ES stands are likely to be more susceptible to increased water stress than mature stands if the Pacific Northwest drought season becomes longer or more intense due to the young stand's open canopy and extreme microclimate, limited root system (i.e., lack of access to deeper water), and inability to induce stomatal closure and conserve water under moderate levels of vapour pressure deficit.
Our results have impacts beyond our specific sites since Pacific Northwest forest productivity during the drought season is strongly coupled with evapotranspiration through stomatal control on water vapour loss. Since forest productivity models are often used to estimate present and future carbon stocks and hydrological processes for this region (see, e.g., Thornton et al. 2002, Law et al. 2004, Turner et al. 2006, our results show the critical need for using stand-specific, ecophysiological response functions in these models, especially for properly capturing the ecosystemlevel impacts of drought. We found significant differences in the timing, magnitude and environmental controls of ecosystem exchange between the OG and the ES stands, indicating the presence of distinct successional-stage mechanisms between the microenvironments and the canopies. Considering that Pacific Northwest forests are characterized by regular silviculture harvest rotations and are predicted to have strong water availability changes in the future from regional climate change, we suggest that more successional-stage studies are needed to properly predict future CO2, water and energy fluxes in these evergreen conifer forests. cavitation due to a long path distance between the water table and the hydraulic capacity of the xylem, and as a result shut their stomata frequently (Ryan and Yoder 1997). Our observation of a rapid decline in Gc between 0.5 and 1.0 kPa in the older forest is consistent with the findings of Ryan and Yoder (1997) as well as with the observations of a diurnal hysteresis in sap flow measurements taken at the bottom and top of tall Douglas-fir trees at Wind River by Cermak et al. (2007). Taken together, these datasets suggest that upper canopy Douglas-fir foliage are driving our observed monthly and diurnal fluxes of mass and energy at the OG forest. At the ES stands, we found that canopy conductance was also inhibited by vapour pressure deficit but not until later in the afternoon hours when the highest oe levels were reached and later in the drought season when relative soil moisture was low. Our overall Gc observations (e.g., canopy conductance was 2-4 mm S-1 higher at the ES stands than at the OG) agree with branch-level measurements taken by Yoder et al. (1994), whereby they found that stomatal conductance is lower in the Wind River OG trees than in the younger Douglas-fir stands, and by Fessenden and Ehleringer (2002) who, using d 13 C isotopes, found evidence that decreased hydraulic conductance in the 450-year-old stand led to lower stomatal conductance in the mature forest than in the younger, shorter trees. The oe was almost always lower at the top of the OG forest than over the ES stands suggesting that the upper canopy stomates are generally closing at lower oe levels in the older stand.
In addition to our successional-stage research, gasexchange studies comparing a Wind River 20-year-old Douglas-fir stand and the OG forest were made in 1998 and 1999 by Chen et al. (2002Chen et al. ( ,2004. In their study, Chen et al. (2002) report higher WUE (1.7 g C kg-1 H20) and Bowen ratio (2.9) at a Wind River 20-year-old stand than at the OG forest (WUE = 1.0 g C kg-1 H20, B = 1.6), while total E T was greater at OG. In contrast, we found higher B, higher WUE and lower E T at the OG forest than the ES stands. Lower WUE values have also been measured for the l0-year-old Douglas-fir age class by Thomas and Winner (2002), whereby they found that WUE was twice as high at the OG forest than at nearby younger trees using 'branch-level' measurements. We suggest that the conflicting nature of succession-stage, gas-exchange results at Wind River is a consequence of (1) a rapidly growing Douglas-fir canopy during the first 0--20 years, (2) a rapid shift in secondary species distributions in the first 0-20 years and (3) different measurement techniques (e.g., 'stand-level' versus 'branch-level') and measurement periods were used in the studies.
It is important to note that other plant species, particularly bracken fern, were a non-negligible component of ecosystem biomass at the younger stands during the summer months. Ground species certainly played a role in determining the carbon, water and energy budgets at the ES stands although we were not able to quantify how significant that role might be. At the OG forest, the canopy is diverse (eight evergreen conifer and three angiosperm species are present) and structurally complex because of this mixture of shadetolerant and shade-intolerant species. Our measurements of gas exchange are for the forest ecosystem as a whole and as such we were not able to determine what contribution each species made to ecosystem fluxes. Therefore, we are making the following conclusions based on the two successional stages investigated in this study and not specifically on tree species, tree height or tree age: (1) Total evaporation (and fraction of expected ET) is higher in ES stands than at the OG forest during the summer months as a result of higher soil temperatures, higher air temperatures and higher vapour pressure deficits in the open canopies during the mid-afternoon hours. Available soil moisture limited E T at the ES stands but only when 0r dropped below 30%. Total E T was limited at the OG stand during moderate vapour pressure deficits and moderate soil water availability.
(2) The ES stands are likely to be more susceptible to increased water stress than mature stands if the Pacific Northwest drought season becomes longer or more intense due to the young stand's open canopy and extreme microclimate, limited root system (i.e., lack of access to deeper water), and inability to induce stomatal closure and conserve water under moderate levels of vapour pressure deficit.
Our results have impacts beyond our specific sites since Pacific Northwest forest productivity during the drought season is strongly coupled with evapotranspiration through stomatal control on water vapour loss. Since forest productivity models are often used to estimate present and future carbon stocks and hydrological processes for this region (see, e.g., Thornton et al. 2002, Law et al. 2004, Turner et al. 2006, our results show the critical need for using stand-specific, ecophysiological response functions in these models, especially for properly capturing the ecosystemlevel impacts of drought. We found significant differences in the timing, magnitude and environmental controls of ecosystem exchange between the OG and the ES stands, indicating the presence of distinct successional-stage mechanisms between the micro environments and the canopies. Considering that Pacific Northwest forests are characterized by regular silviculture harvest rotations and are predicted to have strong water availability changes in the future from regional climate change, we suggest that more successional-stage studies are needed to properly predict future CO 2 water and energy fluxes in these evergreen conifer forests.