S.S.O. Burgess, (present address) School of Plant Biology, University of Western Australia, 35 Stirling Highway Crawley WA 6009 Australia. Fax: + 61 86488 1186; e-mail: email@example.com
Fog is a defining feature of the coastal California redwood forest and fog inputs via canopy drip in summer can constitute 30% or more of the total water input each year. A great deal of occult precipitation (fog and light rain) is retained in redwood canopies, which have some of the largest leaf area indices known (Westman & Whittaker, Journal of Ecology 63, 493–520, 1975). An investigation was carried out to determine whether some fraction of intercepted fog water might be directly absorbed through leaf surfaces and if so, the importance of this to the water relations physiology of coast redwood, Sequoia sempervirens. An array of complimentary techniques were adopted to demonstrate that fog is absorbed directly by S. sempervirens foliage. Xylem sap transport reversed direction during heavy fog, with instantaneous flow rates in the direction of the soil peaking at approximately 5–7% of maximum transpiration rate. Isotopic analyses showed that up to 6% of a leaf's water content could be traced to a previous night's fog deposition, but this amount varied considerably depending on the age and water status of the leaves. Old leaves, which appear most able to absorb fog water were able to absorb distilled water when fully submersed at an average rate of 0.90 mmol m2 s−1, or about 80% of transpiration rates measured at the leaf level in the field. Sequoia sempervirens has poor stomatal control in response to a drying atmosphere, with rates of water loss on very dry nights up to 40% of midday summer values and rates above 10% being extremely common. Owing to this profligate water use behaviour of S. sempervirens, it appears that fog has a greater role in suppressing water loss from leaves, and thereby ameliorating daily water stress, than in providing supplemental water to foliar tissues per se. Although direct foliar absorption from fog inputs represents only a small fraction of the water used each day, fog's in reducing transpiration and rehydrating leaf tissues during the most active growth periods in summer may allow for greater seasonal carbon fixation and thus contribute to the very fast growth rates and great size of this species.
Fog is a significant climatic contributor to a number of terrestrial ecosystems (Weathers 1999). These include coastal mountain regions where orographic lifting of moist ocean air produces adiabatic cooling, condensation and fog formation. Sections of the Chilean and Peruvian Andes are good examples of this and cloud forests such as in Costa Rica are also included in this category. Low-lying coastal regions that are skirted by cold ocean currents experience advection fog, where warm ocean air crosses a band of cold water before reaching land. Examples include Newfoundland (Labrador current), Namibia (Benguela current), and California (California current). Orographic and advection fogs rely on wind to move air over a physical feature (mountain or band of cold water) which creates cooling; often both features are present, such as in the case of Chile. Because these fogs are wind driven, interaction with land surfaces is greatly increased (see Oberlander 1956). A third, fairly ubiquitous type of fog occurs whenever radiative heat loss cools the Earth's surface below the dew point of the surrounding air. Such ‘ground fogs’ tend to form on still nights and since a strong wind component is lacking, there is less interaction with terrestrial surfaces.
The small (1–40 µm, Prada & da Silva 2001) water droplets that comprise fog do not readily precipitate out of air unless they encounter the surface of solid objects (the vertical settling rate is too slow). As a consequence, fog rarely registers on a typical rain gauge and is therefore termed occult (hidden) precipitation. Precipitation of fog droplets mostly occurs by horizontal interception by terrestrial surfaces. Objects such as plant canopies, which are fairly permeable to the through-flow of air and have a large surface area, are ideal fog interceptors (e.g. see Kerfoot 1968).
Fog as an ecological phenomenon
There are a number of potential outcomes when fog interacts with plant canopies. First, fog largely eliminates the atmospheric vapour pressure deficit that drives evaporation and transpiration from plant surfaces. Second, as wind drives a steady stream of fog droplets through the canopy, fog water builds up and wets leaf surfaces. Considerable water can be held or ‘stored’ on canopy surfaces and if this capacity is not exceeded during fog deposition, this water simply returns to the atmosphere with subsequent evaporation (Juvik & Nullet 1995). If water accrues on plant surfaces beyond a certain storage capacity, water will drip from leaves onto soil or be funnelled to soil via stemflow (Hutley et al. 1997). At a study site in northern California, Dawson (1998) estimated that 34% of the annual hydrological input was via fog drip. Once fog drips onto soil, it can be absorbed by the root systems of plants and Dawson (1998) demonstrated that a large fraction of water transpired by S. sempervirens and understorey plants in California's fog zone originates from fog. A third possibility is that the leaves of plants absorb water vapour directly from a saturated atmosphere (Breazeale, McGeorge & Breazeale 1950; Haines 1952, 1953) or from wetted leaf surfaces (Yates & Hutley 1995; Martin & von Willert 2000). There is substantial anecdotal evidence, and some experimental data largely from laboratory work (Kerfoot 1968), to suggest that leaves can absorb fog water by both processes, with greater absorption being possible from wetted surfaces than directly from the atmosphere (Breazeale et al. 1950; Yates & Hutley 1995). For both processes, quantitative data are generally lacking, particularly for plants growing in field conditions and, in addition, the pathways for foliar absorption of water are poorly understood. In certain species (e.g. Crassula spp.), epidermal hydathodes (which are usually involved in guttation) play a role (Martin & von Willert 2000). Hydathodes represent a gap in both the leaf cuticle and epidermis and their ability to exude water presumably correlates with their ability to absorb water. In the absence of hydathodes, a cuticular pathway is usually presumed (e.g. Yates & Hutley 1995) and cuticular wettability, integrity and permeability vary considerably (Boyce, McCune & Berlyn 1991; Jagels 1991; Schreiber et al. 2001). The bark of twigs has also been shown to absorb water (at least under pressure, Katz et al. 1989) but in this article we will focus on foliar uptake only.
An ecological role for foliar uptake of water has long been recognized (Rodin 1953; see Stone 1957 for review), particularly for arid habitats where other forms of precipitation are lacking (e.g. the Namib desert, Chilean highlands). In places such as coastal California that do receive rainfall, the timing of fog events is important (Dawson 1998). California's Mediterranean-type climate is characterized by long summer droughts and it is during these periods that advection fogs are common. The role of fog in ameliorating summer water deficits in coastal vegetation is therefore potentially large. This is particularly true for the coast redwood Sequoia sempervirens (D. Don) whose native distribution today appears constrained to fog-inundated regions of the Californian and Oregon coastline. Although northern limits to the range of coast redwoods are likely constrained by other factors (e.g. competition, cold temperature; Noss 2000), the southern and eastern limits appear tightly defined by areas subject to frequent and intense summer fogs (Cannon 1901; Cooper 1917). Sequoia sempervirens is a shallow-rooted species and therefore presumably does not benefit from access to stable groundwater resources. Dawson (1998) has previously demonstrated that fog-drip onto soils is an important supplement to shallow soil water resources. Nevertheless, S. sempervirens is one of the tallest tree species and transporting limited soil water to leaves growing at heights well over 100 m involves significant resistance due to friction and gravity. In this present study, our aim was to build on this earlier work on inputs to soil water via fog drip and investigate other potential contributions of fog to the plant's water relations: conservation of plant water resources by suppression of transpiration and direct foliar absorption of fog water as an alternative pathway of plant hydration. We employed a number of complimentary methods and used both field observations and controlled-environment experiments to explore the importance of fog to the water relations of the world's tallest tree species.
MATERIALS AND METHODS
Field site description
The study site consisted of a small (11 ha) patch of old growth redwood forest growing on a flat ridge (300 m altitude) in Sonoma County, CA (38°24′ N, 122°59′ W) approximately 8 km from the ocean. Satellite photographs suggested the site was subject to frequent fog inundation (http://www.nrlmry.navy.mil/sat-bin/lowcloud.cgi). The land between the forest and the ocean is largely under viticulture and the lack of physical barriers allows fog to move unimpeded inland to the site.
Two groups of three trees were selected for study, one of the extreme west (seaward) edge of the forest stand and one approximately 200 m east within the forest stand. The groups of three trees were selected such that the trees were of a similar height (60–70 m) and girth (approximately 1.5 m diameter) and grew within a 12 m radius (owing to the logistics of equipment deployment). Preliminary data (Roden, Johnstone and Dawson, unpublished) indicate that the trees were approximately 350 years old. At each group of three trees, climbing ropes were installed to facilitate canopy access.
We opted to use the heat ratio method (HRM: Burgess et al. 2001a, b) to measure sap flow in S. sempervirens specimens because of its ability to measure the slow rates of flow expected during foggy conditions that result in low vapour pressure deficits. The HRM's ability to measure reverse flows (Burgess, Adams & Bleby 2000) was also key, as we hypothesized that foliar deposition and potential uptake of fog water could result in transport of water through the xylem in the opposite direction to that due to transpiration.
Thermocouple temperature probes were constructed by placing two insulated fine gauge (36 AWG) copper–constantan thermocouple junctions inside a stainless steel hypodermic needle (18 gauge/1.3 mm diameter and 38 mm in length). The thermocouple junctions were spaced such that their final positions within the stainless steel needle were 5 and 20 mm from the needle tip. The needle tip was sealed with solder and the needle base filled with epoxy resin to secure and waterproof the contents of the probe.
Line heater elements were formed by tightly coiling a 15-cm length of nichrome wire (36 AWG, 95 Ω m−1) and inserting the coil into a 10-µL glass capillary before placing it in a 38-mm long needle.
For each group of three trees at each site, 16 sap flow sensors, each capable of measuring sap velocity at two radial depths (15 mm apart) were installed. Eight sensors were installed at breast height in the stems of the three trees and spaced approximately 120° around the circumference of each tree (only two sensors were available for the third tree). Due to the thick bark of S. sempervirens (up to 15 cm for our trees), an 8-cm diameter hole-saw, followed by hand chisels, was used to remove bark until just before the living cambium (light pink colour). A steel drill guide was built into a frame that allowed the drill guide to be inserted and secured flush on the stem within the 8-cm hole. A drill was equipped with a long-shank chuck small enough to pass within the frame and drill the three 1.35 mm diameter, 40 mm-deep holes required for probe installation. Otherwise probe installation was the same as described by Burgess et al. (2001a).
The probes were connected by 6 m-long cables to a centrally placed AM416 multiplexer (Campbell Scientific Inc., Logan, UT, USA), with thermocouple circuits differentially wired to minimize electrical noise and with heater probe circuits connected to a distribution strip mounted in the multiplexer enclosure. A temperature thermistor was placed on the multiplexer to serve as a reference for thermocouple measurements. Two shielded four-conductor cables were used to connect the multiplexer to a datalogger (CR10X, 2MB extended memory; Campbell Scientific Inc.) mounted 25–30 m above the multiplexer in one of the trees. Multiplexer control and data transmission circuits were routed to separate four-conductor cables to avoid electrical interference.
An additional eight sensors were deployed at between 50 and 55 m height in the upper canopy of one the three trees. Two sensors were placed in the stem and the remaining six placed in three branches. For each branch, one sensor was placed on the top and one on the underside of the branch. Selected branches were generally 0.1–0.2 m in diameter and faced different aspects. The stem at this height was usually approximately 0.5 m diameter. Due to the thinner bark in stems and branches at this height within the tree, the above-mentioned tools and techniques were not required to install probes.
Once again the eight sensors were multiplexed via an AM416 and connected to the datalogger 25–30 m below using two four-conductor cables. A third, two-conductor cable was used to connect a 30-W solar panel to the datalogger's battery (40 A h sealed gel-electrolyte type). During measurements, the solar panel was disconnected from the battery using a relay so as to avoid electrical interference. A 30-m nine-pin serial cable was also attached to the datalogger so that data could be collected from ground level. The central placement of the datalogger in this two-multiplexer, 16-sensor system minimized cable lengths for data transmission and therefore the effects of electrical noise. Power supply for the heater probes was switched using a relay placed in the datalogger enclosure. Voltage drop was substantial with this wiring scheme, but heaters still provided up to 1 °C heating. Further increases to heater circuit resistance (e.g. longer cable-lengths) would warrant power supplies closer to the heaters.
At the end of the data collection period, a subset of the sap flow data was validated by drilling a series of holes to form a 5–6 cm-wide slice immediately above the site of probe installations. This was done to stop sap flow near the sites of probe installation without damaging the trees excessively. In this way, a reference velocity (zero) could be imposed to calculate the zero offset value of the probe and allow data to be corrected according to the procedures outlined by Burgess et al. (2001a). This approach permits the most accurate discrimination between water loss (indicated by xylem sap flow in the direction of leaves) and uptake (reverse flow). To prevent damage to the study trees and allow other sensors to continue operating, data from only a few representative sensors were validated and the data corrected this way. Based on the results from these examples, environmental conditions were established which allowed us to estimate when flow rates were zero in the remaining sensors. The conditions were: humid (> 94%) and still pre-dawn periods (0300–0600 h) when leaves were not wet. At the potential cost of reduced accuracy, this indirect method allowed us to interpret data from probes that were not cut. Figure 2 includes two probes that were corrected by interrupting flow to establish zero flow and a third sensor (stem at 1 m) corrected using the above criteria.
In each group of three trees, the following meteorological measurements were made at the same height (50 m) where sap flow was measured in branches and the upper stem.
1An index of within-canopy photosynthetically active radiation and total radiation, using a pyranometer and quantum sensor (LI-COR Inc., Lincoln NE, USA) placed on a horizontal branch.
2Temperature and relative humidity (Model CS500; Campbell Scientific Inc.).
3Leaf wetness, using three wetness-sensing grids (Model 237; Campbell Scientific Inc.) placed horizontally 2 m from the stem on three branches facing different aspects. Leaf wetness sensors measure resistance of a gold-plated sensing grid in kΩ. When completely dry, the sensor reads as an open circuit and a mostly dry circuit reads in the high thousands (kΩ). A very wet circuit (i.e. with visible deposition of water) reads approximately 0–10 kΩ. The general response of the sensor is logarithmic and so data have been log-transformed (see Fig. 1) and then multiplied by 10 to give a crude index of leaf wetness where zero is completely saturated and 10 or above is dry.
These measurements were recorded by a datalogger placed at 50 m within the canopy but controlled from ground level via a 60 m, nine-pin serial cable (at 1200 baud to accommodate the longer cable length).
A fog collector was constructed after the design of Juvik & Nullet (1995) and installed at 50 m height in a tree crown at the forest edge. The collector was constructed of a cylinder of louvered aluminium screen measuring 17.8 cm in diameter and 60 cm in height. This cylinder was mounted onto a standard tipping bucket rain gauge (RainWise Inc. Bar Harbor, ME, USA) that was connected to a datalogger. A circular rain shield measuring 60 cm in diameter was mounted on the top of the fog collector to prevent vertical precipitation (rain) from entering the collector. Rain driven by strong winds is probably problematic for this instrument and for this reason a standard rain collector makes a useful adjunct so that rain events can be recognized and discounted. During the course of this experiment, the fog collector underwent development due to unreliable measurements in the early part of the experiment. As a consequence, we did not obtain continuous data records from the fog collector during the study period. Instead, to construct long-term fog records for the study period we used leaf wetness sensors and rain gauges as a proxy. Whenever leaves were wet for more than 3 h in the absence of a rain event, fog was presumed as the cause. This method was compared with, and agreed well with the portions of extant data from the fog collector (see Fig. 1). Dew events cannot be separated from fog using this method, but for the purposes of this investigation this is not important since the physiological effects of dew (see Boucher, Munson & Bernier 1995) are probably similar to those of fog.
Two, 2-m-tall potted redwood saplings were placed in a clear polyvinyl chloride (PVC) ‘fog’ chamber (2 m × 1.2 m × 1.2 m) inside a glasshouse. A single HRM sensor containing two pairs of thermocouples was placed in the base of the sapling's stem (50 mm diameter). Two further, miniaturized HRM sensors (see Burgess, Dubinsky & Dawson 2001 for more details) were installed on two different 6-mm-diameter branches. Readings were made every 10 min and averaged hourly. As absolute flow rates were not required to investigate the importance of water uptake by leaves relative to transpiration rates, sap flow data are expressed as a percentage of maximum rates measured for each sensor during the experiment. Although the experiment was carried out during winter, the maximum rates of flow measured on warm, sunny winter days in the glasshouse were probably not much lower than true maximum values. Zero-offset values and appropriate corrections for all sensors were obtained by taking night values when the PVC chamber was fully closed to prevent water loss.
An ultrasonic device (Chaoneng Electronics, Nanhai, Guangdong, China) was placed in a reservoir of distilled water maintained at a given capacity by means of a float switch and used to generate fog at a rate of 500 mL h−1. This device was chosen to produce a water aerosol that resembled ‘typical’ fog as closely as possible, namely with droplets in the 1–40 µm range, as opposed to spraying techniques which have more similarity with light rain. Since the artificial fog rapidly coated leaf surfaces with a film of liquid water, droplet size was probably a less important distinction than the evenness with which an aerosol can coat surfaces. A small electric fan was used to disperse the fog and circulate it throughout the chamber. By means of an electronic timer, fog events of various durations could be applied to the tree. After each fog event, the sides and roof of the PVC chamber were removed to ventilate the plant and allow leaves to dry off. Prior to initiation of any fog event, extreme care was taken to double or triple bag the pot and root system of the sapling to prevent any fog water from reaching the soil by any means other than xylem transport (i.e. including stem flow).
As described for the above glasshouse experiment, 2-m-tall saplings were placed in the ‘fog chamber’. A ‘labelled’ water source with a hydrogen isotopic composition (δD) of +6‰, was added to the reservoir that supplied the ultrasonic fog generator. We measured the isotopic composition of the resultant fog and it did not differ from the water in the reservoir (data not shown) because ultrasound atomizes, rather than evaporates water.
The isotopic composition of tap water used to irrigate the saplings had a δD of −78‰. All isotopic values were expressed in standard delta notation (parts per thousand, or ‰) relative to the V-SMOW (Vienna Standard Mean Ocean Water) standard:
A simple proportional mixing model (as reviewed by Dawson et al. 2002) was used to determine the proportion of water in leaves that was derived from fog water versus irrigation water.
Two treatments were considered: water-stressed plants versus well-watered plants and young leaves versus old leaves. Well-watered plants were irrigated such that leaf water potentials were maintained at approximately −0.5 MPa. Irrigation was withheld from plants in the ‘water-stressed’ treatment until leaf water potentials reached −1.65 MPa.
Saplings were fogged overnight using the same methods described above and then removed from the PVC chamber for sampling. Droplets of water shaken from branches were collected to verify that deposited fog water retained the 6‰ isotopic label. For each treatment, nine leaf samples (three leaves each from three different saplings) were collected. Leaves were quickly but thoroughly dried with paper towel until no water remained on leaf surfaces. Additional leaf samples were also dried and used to measure leaf water potential. Water was extracted from sample leaves by cryogenic vacuum distillation (Ehleringer, Roden & Dawson 2000). The isotope ratio of each water sample was determined by combusting approximately 2 µL of water in the presence of hot (800 °C) Cr using the H/Device interface attached to a Finnigan MAT delta-plusXL isotope ratio mass spectrometer (Bremen, Germany).
In order to measure the rate at which a s. sempervirens leaf can absorb water when its surface is completely wet, the following experimental apparatus was devised: an electronic balance (Denver Instrument, Denver, CO, USA; resolution 0.01 mg) was attached to a computer to allow weight changes to be recorded every 30 s. A 50-mL vial was placed on the balance and filled with deionized water. For each experiment, a single leaf-bearing branchlet, representing 1 year's growth, from a potted redwood sapling was immersed in the vial and the supporting branch held firmly in place by a retort stand clamp. A layer of vegetable oil approximately 3 mm thick was added to the water in the vial to prevent evaporation. The sapling and balance were placed in a dark room and left without disturbance for approximately 3 d. At the end of each experiment, the portion of branchlet and leaves immersed in water was excised and the leaf area was measured with a leaf area meter (LI-3000: LI-COR Inc.).
As an adjunct to our absorption experiments, we also examined the surfaces of a number of S. sempervirens leaves using scanning electron microscopy (Electroscan E3 ESEM; Wilmington, MA, USA).
Figure 2 shows sap flow and meteorological conditions over a four day period at the end of the summer fog season (mid-May to late October) in early October 2001, when no significant rain had been recorded for 3 months. Data from this period are representative of similar sap flow patterns associated with heavy fog on over 20 occasions during the rainless ‘fog season’ in 2001 and in the subsequent 2002 and 2003 fog seasons. The sap flow patterns displayed in Fig. 2 agreed with patterns in all other trees measured during this period. Figure 2a shows xylem sap flow measured by a single sensor placed in the lower stem (1 m), upper stem (50 m height) and lateral branch (50 m height) of a redwood at the edge of the forest site in Sonoma County. For the purposes of comparison in Fig. 2, the differing flow rates found in the different parts of the tree measured were ignored and values were normalized with respect to maximum values for each sensor for this time of year. This permits a clear comparison of the timing and relative proportions of positive and reverse sap flow at each measurement position within the tree.
Figure 2b demonstrates a range of meteorological conditions over the 4 d period presented, including two fog events of different magnitude. When compared with Fig. 2a, a number of important aspects of redwood physiology and water relations are revealed. The first night had low relative humidity (approximately 20–40%), resulting in night-time water loss from the canopy (see also Fig. 3). Meteorological conditions on day 1 produced a typical pattern of daytime transpiration, although maximum rates under October conditions of solar insolation were approximately 70% of mid-summer rates. Night 2 indicates heavy fog conditions. Relative humidity was 100% and in addition, leaf wetness sensors indicated near-saturation. During this period the fog collector logger was damaged, so fog was inferred by the combination of high humidity, wet leaves and absence of rain as measured in an adjacent vineyard. Sap flow during night 2 was near zero or very slightly negative. On day 2, heavy fog persisted, resulting in a complete absence of transpiration. Instead, rates of reverse flow increased at all positions within the tree; maximum reverse flow rates measured were approximately 5–7% of maximum transpiration rates. By approximately 1600 h, the leaf wetness sensors showed some evidence of drying, however, this was incomplete and relative humidity remained extremely high. Relative humidity remained above 90% for the duration of night 3; sap flow was approximately zero during this time. Sap flow rates were fairly normal on day 3, with humidity levels being low enough to permit some transpiration (relative humidity = 75%, vapour pressure deficit = 0.4 kPa). Humidity was once again very high on night 4 (ranging from 90 to 97%), however, the absence of readings by the leaf wetness sensor suggests there was no fog or fog-deposition on this night. Note that by dawn, sap flow rates were zero in all parts of the tree. Day 4 was similar to day 3. Finally, on night 5, fog returned, as indicated by relative humility values of 100% and leaf wetness sensors indicating deposition of water; again reverse flow was detected.
A strong relationship between night-time (0030–0530 h) sap flow and vapour pressure deficit was evident for S. sempervirens (Fig. 3). Expressed as percentage of maximum flow rate measured in the study tree, night-time rates of water loss often exceeded 20% of peak water loss rates measured at noon under hot summer conditions. On one or two extremely dry nights, water loss rates exceeding 40% of maximum were recorded (see Fig. 3); smaller values of 10–12% such as in Fig. 2a are very common. The possibility of very large rates of night-time water loss as indicated in Fig. 3 is supported by data from the 24 other extant sets of measurements at the study site: peak rates of night-time water loss averaged 37% of summer maximum.
Figure 4a shows the number of days and nights with rainfall events for each month during the sampling period. Figure 4b shows the number of days and nights with fog events. Figure 4c shows monthly totals of heat pulse velocity (a simple proxy for sap flow; see Burgess et al. 2001a) averaged for trees at both the edge and interior sites within the forest. Evapotranspiration measured by the City of Santa Rosa approximately 40 km inland from our coastal site (beyond the typical reach of fog) is also plotted next to the rainfall and fog data for each period. Comparing these data with each other at each single time period and also across the entire year, we note that monthly sap flow totals generally tracked closely the measurements of evapotranspiration inland from our site. However, large discrepancies appear particularly during August when fog was heaviest.
Figure 5 shows xylem sap flow rates (expressed as percentage of maximum) in the stem and a single branch of a 2-m-tall redwood sapling over a 5 d period. At approximately 1100 h on day 3, a 10 h fog event was initiated. For context, we calculated that water uptake by the two 6 mm branches (averaged in Fig. 3) of the 10 h fog period totalled 18% of water lost by these branches on day 5. In the main stem, water uptake totalled 6% of water loss of day 5. This value was decreased in part by the brief period of water loss in the early stages of the fog event when fog had not built up sufficiently to cover the top of the sapling. Apparently at this point, water was still being lost by upper branches while lower branches with wet leaves were absorbing water. Although sap flow reversals in response to fogging were easily repeatable, we could not detect any repeatable increases in soil moisture in response to foliar uptake of fog water using capacitance sensors (CS615; Campbell Scientific Inc., data not shown).
Based on a simple mixing model, the proportion of leaf water sourced from fog water was 6.4% in old leaves from well-watered plants (Fig. 6). In contrast, young leaves from well-watered plants only contained 1.8% fog water. For water-stressed plants, values were 1.6% for old leaves and 1.2% for young leaves.
When placed in distilled water, leaf-bearing branchlets (1–3 years old, n = 3) absorbed water fastest during ‘usual’ daylight hours (the plant remained in continuous dark) and hardly at all during hours of usual darkness (Fig. 7). The average rate of absorption, approximately 0.89 mmol m2 s−1, was about 80% of typical leaf-level transpiration values measured (n = 58) in the field during August using a portable photosynthesis measurement system (LI-6400; LI-COR Inc.). Note that the weight changes shown in Fig. 7 represent export of water from the portion of leaf submersed in the vial into the rest of the plant: localized absorption of water into leaf tissue would increase the leaf's volume and displacement resulting in no net change in the weight of the vial.
Despite the mesic habitat of S. sempervirens, leaves had a thick cuticle, stomata were recessed and each stomatal pore was capped by an epicuticular wax plug. Non-pathogenic fungal hyphae were often observed entering stomatal pores (Fig. 8) and this appeared to be more common on mature leaves where the epicuticular wax plug was degraded.
The prolonged and heavy fog event documented in Fig. 2 yields one of the clearest examples of sap flow reversal that indicates direct foliar uptake of water by a large Sequoia sempervirens specimen under field conditions. While numerous, brief fog events were accompanied by indications of sap flow reversal in all of the trees we measured, the largest magnitude and therefore most definitive flow reversals were seen when fog events were heavy for a whole daylight period (as shown in Fig. 2). Sap flow reversals during fog events were always small in magnitude (sap velocities 1–1.5 cm h−1), often approaching the limits of accuracy of the HRM (approximately 0.5 cm h−1) and thus we are only confident to discuss sap flow reversals for the largest examples. Furthermore, the data we present in Fig. 2 were recorded just days before the xylem surrounding the probes was interrupted by cutting to establish a true zero value. The possibility for any drift in the zero line for these sets of measurements was therefore minimized. By means of these prompt validations we obtained very fundamental evidence that flow reversals did indeed increase the ratio of heat flux in the direction of the soil versus the leaves, as compared with baseline heat ratios measured when xylem was severed a few days later. The fact that flow reversals were measured simultaneously throughout all parts of the tree adds strength to our interpretation of the sap flow data and further suggests a whole-plant, leaf–soil flux was involved. Laboratory data such as those of Slayter (1956) which showed whole-plant reverse water transport through Pinus echinata seedlings in response to a vapour pressure gradient create something of a precedent for our observations. Unlike Slayter's (1956) laboratory experiment, we were not in a position to gauge the amounts of water delivered to roots and/or the soil/rhizosphere from our measurements in the stem, but for reverse flows to move past our lowest measurement position (1 m), one would expect increases in the moisture content of root tissue at the very least. As an aside, it is interesting to note that during reverse flow, the water potential gradient in the tree will be less than the hydrostatic gradient due to gravity. Gravity dictates that the minimum tension in the xylem at the top of a 60-m tree is approximately 0.6 MPa, but during reverse flow this value is clearly reduced.
The results from the glasshouse experiments agree well with the above conclusions drawn from our field-based measurements. Flow reversals were of similar magnitude to those measured in the field, although because the glasshouse experiments were conducted in winter, flow reversals were probably slightly overestimated relative to true maximum transpiration rates. The fact that we could not detect increases in soil moisture with capacitance sensors is puzzling in view of the measured reversals in sap flow, however, the sensitivity and positioning of these sensors may not have been optimal to detect such changes.
While attempting to measure sap flow reversals in the saplings, we made unconfirmed observations that suggested that the longer the water was withheld from the plants, the harder it was to induce reverse flow by fogging. Although we did not set up an experiment to test this, the results of our isotope labelling experiment support these observations and indicate that, at least for old leaves, water-stress reduces the capacity for foliar absorption of water. This agrees with the findings of Vaadia & Waisel (1963), who showed that water stress reduced rates of foliar uptake in Pinus halepensis. The labelling experiments also suggest that if leaves are well watered, old leaves are more conducive to water uptake than young leaves; this was also our (undocumented) experience with the submerged leaf experiments (see below). The various factors and conditions that may enhance or reduce foliar uptake of water are worthy candidates for further, controlled experimentation. In particular, the role of fungal hyphae in altering the properties of leaf surfaces warrants exploration. As leaf surfaces degrade with age and are invaded by fungal endophytes, hyphae may form a ‘wick’ to move water past epicuticular waxes.
Magnitude and ecological significance of foliar uptake
Based on an allometric relationship supplied by Stancioiu and O’Hara (unpublished results), the tree from which the data presented in Fig. 2a were collected had a projected leaf area of 660 m2. The laboratory-based water uptake experiments using submerged leaves yielded an average absorption rate of 0.058 L h−1 m−2 projected leaf area, or the equivalent of 38 L h−1 for the whole tree. Assuming a crude estimate of rooting area of 80 m2 for each tree at our site, then over a 12-h period (e.g. midnight to noon) of foliar absorption each tree would capture an amount of water equivalent to a 5–6 mm rain event being fully absorbed by the root-zone. Sap flow measurements did not suggest uptake anywhere near this magnitude, with a maximum rate of only about 1.5 L h−1 of water moving downward through the stem at breast height (0.225 mm rain equivalent over 12 h). We see three likely reasons for this discrepancy. First, fully submersing leaves in distilled water for a prolonged period of time probably establishes conditions for water absorption that are quite different to those experienced in the field. Second, the isotope experiments demonstrate huge variability in rates of uptake depending on leaf age and physiological condition (e.g. leaf water potential): the results from one type of leaf thus cannot be scaled to a whole canopy comprised of a range of leaf types. These two reasons probably render our laboratory measurements of foliar uptake a gross overestimation from a whole plant perspective. A third reason for the discrepancy is instead a potential cause for underestimation of foliar uptake as measured by sap flow: much of water that is absorbed through leaves probably first refills storage capacity in leaves, twigs and stem, allowing only a smaller remainder of fog water to move past sap flow sensors toward roots and soil.
Results of the submerged leaf experiments compared quite poorly with our sap flow and isotopic studies. Because the latter two studies involved deposition of fog droplets onto leaves rather full submersion of leaves into water, we are inclined to consider results of these experiments as more realistic indicators of the potential of foliar uptake of water. This position is further strengthened by the fact that the sap flow and isotope experiments agreed quite well with one another. The fact that these experiments indicate that foliar absorption of water is small in comparison to transpiration losses does not mean that fog is insignificant to the water budget of redwood trees. First, Dawson (1998) demonstrated a large role for unabsorbed fog water dripping onto soil where it can be accessed by roots. Second, fog dramatically suppresses transpiration by eliminating atmospheric vapour pressure deficit (Byers 1953). Indeed, even if the flow reversals shown on day 2 in Fig. 2a were ignored, the water savings from a lack of positive flow (transpiration loss) are large compared to water losses of adjacent days. From a water budget perspective, every fog event like this shortens the drought period by 1 d, while at the same time augmenting available water in the soil via fog drip and in the plant via foliar uptake. This may have important consequences for daily and seasonal carbon fixation for S. sempervirens, which is one of the fastest growing conifer species known (Noss 2000). The fact that, in proportional terms, monthly sap flow totals (Fig. 4c) were reduced compared to inland evapotranspiration during August, which was the heaviest fog period (Fig. 4b), clearly demonstrates that coastal fog attenuates transpiration during what would otherwise be a hot dry time of the year. In other words, the coast-hugging S. sempervirens experiences a much milder Mediterranean-type drought than it would inland.
Given the strong relationship between vapour pressure deficit and night-time water loss (Fig. 3, see also Benyon 1999) it appears that S. sempervirens has either poor stomatal control in response to changes in vapour pressure deficit, or incomplete stomatal closure (due to asymmetric guard cells resulting from polyploidy, Noss 2000). We routinely measured night-time water loss rates up to 20% of the most rapid summertime transpiration rates. Given the large proportion of time S. sempervirens spends in the dark, significant amounts of water loss are possible during a time when there is no benefit of carbon gain. Figure 4b shows that nearly one-third of the nights as opposed to only one-fifth of the days during August had heavy fog. Prevention of the rather disadvantageous loss of water during night is probably just as important as attenuating daytime transpiration rates in dictating the observed distribution of S. sempervirens along fog-inundated coastlines.
As fog is most frequent at night, there are still plenty of days during the summer growing season when solar radiation is high and leaf surfaces are dry to permit maximum gas exchange and photosynthesis. It is likely that the avoidance of desiccation, coupled with supplements to stem, leaf and soil water storage via foliar uptake and fog-drip would greatly enhance leaf water potential (Menges 1994; Yates & Hutley 1995), CO2 fixation and growth rate during these conditions.
In addition to previously demonstrated hydrologic inputs via interception and the dripping of fog water from canopies to the soil (Dawson 1998), we have demonstrated that S. sempervirens can absorb small quantities of fog-water directly through their leaves. The small amounts of foliar uptake are difficult to quantify, so we employed several complimentary methods to build what we believe is a cautious estimate of fog uptake by leaves under conditions when it is maximal. Although amounts are small, they are sufficient to initiate reverse water flux throughout large trees and improve the water status of the entire plant, presumably including the root-zone. An impact sufficient to be seen at the whole plant level suggests that foliar uptake could contribute meaningfully to recharge of water stores within plants, repair of cavitated conduits (Tognetti et al. 2001), drive cell expansion and even lead to increased carbon fixation. In addition, the potential for nutrient fluxes into leaves from particulates deposited on leaf surfaces or dissolved in the fog itself (Azevedo & Morgan 1974; Weathers 1999; Weathers et al. 2000) is raised and warrants further study. Although the wax-plugged leaf surfaces of S. sempervirens do not appear to be efficient at absorbing water (e.g. Feild et al. 1998 state that wax plugs prevent water films from forming on leaf surfaces), this may change as leaf surfaces degrade with age and are invaded by hyphae from fungal endophytes (Fig. 8; see Rollinger & Langenheim 1993). Furthermore, in common with the wax-plugged leaves of Drimys granadensis (Feild & Holbrook 2000), S. sempervirens are demonstrably inefficient at preventing water loss and it is in mitigating this that we believe fog plays the most important role in this species that today is restricted to the coastal fog-belt of California.
Funding for this work was provided by the Andrew W. Mellon foundation and Global Forest (GF-18-2000-112/113). Many thanks go to Caryl & Mickey Hart and the Sonoma Co. LandPaths organization for site access and assistance, Walt Chavoor & Theo Chavas (Kendall Jackson Winery) who provided rainfall data, the City of Santa Rosa for evapotranspiration data, Chet Moritz and Danika Gilbert for risking their necks rigging the trees, Eric Dubinsky and Jia Hu for their dedicated hard work on all aspects of this project and also Primrose Boynton, Adeline Fabré, Ken Peer, Vanessa Schmitt and Spencer Hawkins for all of their assistance. We thank John Rumbold for helpful comments on the manuscript.