Making dew in the Atacama Desert: a paradigmatic case of plant water uptake from an unsaturated atmosphere fails a test

Abstract

Background and Aims

Nolana mollis is a dominant plant species in the hyper-arid Atacama Desert. A previous hypothesis states that N. mollis owes its success to the condensation of atmospheric water from undersaturated air onto its leaf surfaces by exuded salts, and absorption of this water by its leaves, or by shallow roots following drip onto the soil surface; living roots of N. mollis were suggested to only exist near the soil surface.

Methods

We conducted a field experiment with three treatments to establish the source of water for N. mollis: control, root cutting to block uptake of all soil moisture, and plastic skirting at the soil surface to block leaf drip of atmospheric water.

Key Results

Xylem tensions monotonically increased after root cutting until the plants wilted irreversibly, diverging clearly from the skirted and control treatments showing diurnal patterns of increasing tension in the day followed by recovery overnight.

Conclusions

Hydration in N. mollis requires access to deep soil water, motivating an alternative hypothesis: imperfect salt exclusion at the root surface and salt exudation by the leaf results in less root fouling and lower xylem tensions, while during the day evaporation of the surface brine, condensed overnight, increases the water use efficiency of carbon gain.

INTRODUCTION

On the coast of the hyper-arid Atacama Desert of northern Chile, a region with an annual precipitation of less than 25 mm, the succulent C₃ shrub Nolana mollis dominates a sparse flora covering a mere 1–2 % of the area of the dry river valleys (quebradas) that extend inland below the fog zone (Rundel et al., 1991). The species ecological success has been linked to the presence of a brine, absent from co-occuring species, that coats the leaves and green stems of the plant (Mooney et al., 1980; Rundel et al., 1980). Mooney et al. (1980) hypothesized that the internal tissues of N. mollis effectively pump water from the epidermal brine against a gradient in water potential by using energy from respiration. Despite the absence of a known molecular pumping mechanism by which extraction of water molecules from the brine could occur, the hypothesis that internal leaf tissues can absorb pure water from surface brines continues to shape the interpretation of ecological and physiological experiments, leading to reports of such uptake in mangrove systems (Coopman et al., 2021), arid coastal sandflats (Al-Handawi et al., 2023) and hyper-arid deserts (Wang et al., 2016).

The physical and biological processes leading to the formation of a brine on N. mollis are themselves well understood. Alone among the species of the quebradas, N. mollis has salt glands on its leaf surfaces (Rundel et al., 1991). Salts on the leaf surfaces condense atmospheric water vapour (i.e. deliquescence), resulting in a voluminous brine forming pendant droplets on the bulbous leaves. Mooney et al. (1980) absorbed this brine onto filter paper and showed it condensed almost as much water from the night air [82 % relative humidity (RH)] as a saturated NaCl solution, which they report has an equilibrium relative humidity of 75 %. Mooney et al. (1980) further concluded, based on a daytime leaf energy balance argument, that nighttime growth of the brine should be generally sufficient to balance daytime losses due to evaporation, explaining the diurnal persistence of the brine.

One way that this brine might confer an ecological advantage on N. mollis follows directly from the observation that the brine grows at night and shrinks in the day: evaporation of surface water from the leaf during the day should in part suppress water loss through open stomata, through both cooling and local humidity effects, potentially conferring upon N. mollis an advantage in water use efficiency versus its dry-leafed competitors. The hypothesis of Mooney et al. (1980) proposes a second advantage, that the internal water balance of N. mollis is maintained, despite the inevitable loss of some water to transpiration, by uptake from the brine. The authors discuss two possible pathways: water uptake from the brine directly across the leaf surfaces, and water uptake by shallow roots from brine dripping from the leaves onto the soil. In both cases, Mooney et al. (1980) acknowledge that water would be moving against its gradient in free energy (i.e. water potential).

Just how formidable the energetic barrier to extracting pure water from the brine is can be seen by expressing relative humidity in terms of an equilibrated liquid-phase water potential. The diurnal RH range for air proximal to N. mollis has a typical nighttime maximum of 82 % RH and an afternoon minimum RH of 75 % RH, as for a saturated NaCl brine, corresponding to a range of water potentials from approximately −27 to −40 MPa, far outside the diurnal range of pressure chamber data for shoot water potential of −1.8 to −2.3 MPa (Rundel et al., 1980; this study). At a minimum then, under isothermal conditions a difference of 25 MPa would have to be overcome to move water out of the brine into the plant. For a plant to perform this work would require the existence of a mechanism of plant water uptake unknown to science. Nevertheless, Mooney et al. (1980) argued that N. mollis theoretically has the required energy available from respiration to perform such work. They further proposed that this mechanism most probably occurs in the fine roots of N. mollis in the upper few centimetres of soil, rather than in the leaves. If such a mechanism that uses metabolic energy to transport water against its gradient in free energy exists and could be incorporated into agricultural plants, it would afford substantial protection from drought stress to crops.

Yet, transport mechanisms aside, the hypothesis that N. mollis subsists on inputs of atmospheric water presents other difficulties. Foremost among these is the origin of the salts in the brine. Rundel et al. (1980) report that the soils of the quebrada are not intrinsically saline, and yet the ash accumulating beneath all of the woody shrubs found there contains high concentrations of various salts: surviving in the quebrada appears to inevitably result in the accumulation of salts in leaves and stems. Nolana mollis, despite its ability to export salts to its surfaces, is no exception: sodium and chlorine account for 18 % of its leaf dry matter, the highest concentration among the five most common woody shrubs (Rundel et al., 1980). These facts are perhaps most easily reconciled with each other by the hypothesis that all of the woody shrubs of the quebrada are dependent on a saline groundwater source. The existence of saline groundwater is known from the occurrence of saline springs (aguadas) in the lower reaches of the quebradas where saline underground aquifers reach the surface: these aquifers are re-charged by a sub-surface flow derived ultimately from runoff from the Andes into the highly saline Salars in the interior Atacama (Rundel et al., 1991). Yet, a groundwater hypothesis appears to be eliminated by the report that N. mollis root systems extend to a maximum depth of about only 50 cm, with living fine roots (those involved in water uptake) confined to the upper 2 cm of the soil profile, an architecture that supports near-surface absorption of water (Rundel et al., 1980).

In this study, starting from the premise that to maintain water balance against transpirational losses N. mollis requires access to some external source of water, we sought to test the importance of three possible paths, and two possible sources. The first source, atmospheric water condensed in the brine, might be accessed by two paths: either directly by foliar absorption, or indirectly by shallow root absorption of leaf and stem drip. The latter is the preferred hypothesis of Mooney et al. (1980). A second hypothesized source, groundwater, could be reached by deep roots, should the root system of N. mollis be found to extend below 50 cm. To distinguish between these three possibilities, we adopted a design with two treatments and a control. The first treatment was designed to exclude the drip path (but not the foliar absorption path) by isolating the crown from the soil surface with plastic sheeting (‘Skirt’). The second treatment sought to exclude both the drip and deep-soil paths by severing the root system from the shoot (‘Cut’), leaving only the foliar absorption path open. Finally, the control allowed for all three paths (‘Native’). Our strategy was to monitor the response of leaf water potential to these treatments over time until a clear difference from the control emerged in at least one of the other two treatments. We reasoned that if all three groups continued to thrive, then foliar absorption of water from the leaf surface brine was supported; if both the cut and skirt treatments declined, shallow root absorption of atmospheric water must be important. And finally, if only the cut treatment declined, then groundwater absorbed by deep roots must make the most important contribution to maintaining water balance in N. mollis.

MATERIALS AND METHODS

Experimental design

The study took place in the Pan De Azucar National Park in the Atacama Desert in Chile, 16–23 March 2018. Our experimental plants were naturally established shrubs growing in the quebrada agua chica, 26°07ʹ15″S, 70°37ʹ44″W, below the high-elevation fog zone in which the air occasionally reaches saturation (Thompson et al., 2003). Daily mean air temperature, dew point and relative humidity in Puerto Pan de Azucar over the study period, as reported by a weather station at the park headquarters and research station, were 20 °C, 15 °C and 71 % respectively, consistent with conditions observed in a previous multi-year local climate study (Thompson et al., 2003). We applied three treatments to 15 vigorous N. mollis shrubs: ‘Native’ (n = 5), plastic ‘Skirt’ (n = 5) and ‘Cut’ root (n = 5). For the Native treatment, no manipulations were performed. For the plastic Skirt treatment, we installed the skirts (black plastic sheeting) underneath the branches and around the single stem in the afternoon of 16 March 2018 (Day 0). To avoid albedo effects on soil and root temperature, the skirts were covered with loose sand and gravel from the vicinity. For the Cut root treatment, we dug out the soil around the root collar of the plant on the afternoon of 16 March but did not cut the root until the following morning. We delayed the cut to the subsequent morning so that we could test for an effect of our excavation (e.g. damage to the shoot) on plant water status. Sampling for pressure chamber and psychrometer measurements was conducted each day at 0700 h (AM), between first light and sunrise, and in the afternoon at 1500 h (PM). After taking the pre-dawn samples on 17 March (Day 1), we cut the main root of each plant ~10 cm below the soil surface and above any root branching.

Nolana mollis sampling protocol and pressure chamber measurements

The Scholander pressure chamber measures xylem tension by establishing a mechanical equilibrium between the atmosphere and a flat meniscus of water on the cut surface emerging from the chamber (Scholander et al., 1965; Boyer, 1995). The water potential of a sample may then be calculated as the sum of xylem tension (the negative of the balancing pressure reported by the chamber gauge) and xylem osmotic potential. The latter is typically assumed to be negligible (less than 0.05 MPa, or within the typical error of pressure chamber measurements; Boyer, 1995), allowing balancing pressures to be interpreted as equivalent to the magnitude of total sample water potential, but in halophytes the assumption that the xylem sap is essentially pure water may not hold (Scholander et al., 1966). We therefore report balancing pressure data as xylem tension rather than water potential.

Initial attempts at measuring Nolana shoots in a Scholander-type pressure chamber (custom built, parts from Soilmoisture Equipment Corp., Goleta, CA, USA) resulted in two types of failures: the soft fleshy and green cortex around the branch xylem either became crushed when the gasket seal became too tight, releasing water to create artifactually low balancing pressures, or its slipperiness caused the shoot to eject from the chamber; the seal, when loosened to avoid crushing, failed to hold. We found that decorticating the shoot (removing the epidermis and cortex tissue layers to expose the vasculature) so that the seal could be made against a parafilm-reinforced xylem cylinder, eliminated both problems.

However, by decorticating a second time and re-measuring balancing pressure in the pressure chamber, we found that decortication did in some instances hydrate a shoot to below its prior balancing pressure, suggesting the first decortication could also hydrate the shoot in a non-detectable way. We therefore developed a sampling protocol to avoid such effects. We selected one branch (~15 cm long) per plant per sampling time point. We then decorticated ~3 cm of the shoot at the base of where we planned to cut and wrapped the decorticated region in parafilm to stop evaporation from the exposed wood. We then waited for 10 min before harvesting the shoot to allow time for any locally released water to dissipate into the rest of the plant. We then harvested the shoots, placed them in plastic bags in a cooler and returned to the lab (a 4 × 6-m out-building at the park headquarters in Pan De Azucar provided with electricity, but no water or climate control) to make the water potential measurements. To avoid confounding any storage-time effects with treatment effects, we processed samples in an order determined by blindly removing bags from the cooler.

Psychrometer measurements of water potential and osmotic potential

At each time point, two of the five plants in each treatment were selected for additional psychrometer measurements (PST-55 peltier psychrometers; Wescor, Logan, UT, USA). Shoots destined for psychrometer measurements were washed in the field with a spray bottle to remove the surface brine and patted dry with paper towels. The washing was done while shoots were still attached to the parent plant to allow any hydration effects due to washing to dissipate. Following harvest and pressure chamber measurement, for whole leaf psychrometer water potential measurements we removed three bulbous leaves from a shoot and dusted them with a lint free tissue (Kimwipe; Kimberley-Clark, Irving, TX, USA) to remove any remaining salts prior to sealing them inside psychrometer chambers, which had a volume of 0.5 mL. To measure leaf osmotic potential we followed a procedure described by Boyer (1995): four leaves were placed in a 10-mL syringe with a filter paper disc at the base, which was then placed in a freezer (−18 °C) overnight to rupture the cell membranes, as confirmed qualitatively by a change in colour due to a change in the index of refraction of the leaves as formerly symplastic fluid spread into the intercellular airspaces. The syringes were then thawed and the leaf sap expressed onto filter paper discs, which were then sealed inside psychrometer chambers. To reduce thermal noise in and the need for temperature corrections of the psychrometer output, all calibration measurements were taken with chambers immersed in a temperature-controlled water bath held at 30 °C (Savage et al., 1981). Psychrometer measurements were made every 15 min with 5 s of cooling time at −6000 μA using a CR6 programmable datalogger (Campbell Scientific; for the CR6 program see Supplementary Data Notes S1), thereby condensing water onto the ‘wet’ junction. At the cessation of cooling, evaporation from the wet junction depresses its temperature, and the resulting differences (μV) between the wet and dry thermocouple junctions (i.e. the cooling curve) were collected for 60 s. Psychrometer output (μV) over the plateau of the cooling curve was converted to water potential by calibration with NaCl solutions (Lang, 1967).

Direct measurements of xylem sap solute potential

For two samples in each treatment, after a pressure chamber measurement the shoots were over-pressurized and the sap that came out of the decorticated xylem was collected onto filter paper discs. The first disc was discarded due to concerns of possible contamination with cellular water. A second disc was then collected and immediately placed in a psychrometer.

Rooting depth and soil water potential

We excavated the root system, following the tap root, of three N. mollis individuals to a depth of at least 1 m. For one excavation to ~1 m depth, soil samples were collected by inserting a 25-cm soil corer laterally into the soil face at 0.5, 0.75, 1 and finally at 1.25 m by inserting the corer vertically to sample 25 cm below the soil surface at the bottom of the excavation. For the two other excavations only the soil at 0.5 m was sampled. Soil from the corer tip was sealed in plastic bags and transferred into psychrometer chambers in the lab, and total water potential was measured using the psychrometers.

Measurements on a co-occurring woody shrub

We made xylem tension (pressure chamber) and whole leaf water potential (psychrometer) measurements over the course of 3 d on two individuals of Gypothanium pinifolium at the same pre-dawn and afternoon timepoints as N. mollis. Gypothanium pinifolium was chosen as it was the most abundant in number and showed the lowest amount of salt accumulation in its tissues among shrubs investigated by Rundel et al. (1980) in Pan De Azucar, and as it does not exude salts like N. mollis, it appears to follow a strategy of salt exclusion at the root surface.

Statistics

Our experimental design included three treatments, sampled at two times of day, repeated over five (non-contiguous) days of a single week. While we initially planned to analyse the experiment with a single balanced statistical model, this became problematic due to practical problems faced in the field. For example, as on the afternoon of Day 0 we could not sample as we were preparing treatments, we lack afternoon data for that day, and on Day 6 (the final day of the experiment) the cut treatment was too wilted to sample, further unbalancing the design. In addition, the occasional failure of a shoot in the pressure chamber (ejection under pressure) meant that some treatments on some days had only four replicate measurements. As our null hypothesis was simply that on any given day and sampling time there would be no difference between treatments, we constructed individual ANOVAs for each day and sampling period (AM/PM), with xylem tension as the response variable and treatment as a fixed factor.

RESULTS

Xylem tension experiment and soil water profiles

Pre-dawn samples of xylem tension on Day 0 showed no significant differences between any treatment groups (Fig. 1). On the morning after the installation of treatments, Day 1 AM estimates of xylem tension showed a non-significant (within 1 s.d.) elevation of tension in the cut treatment versus the undisturbed native and skirt treatments (mean ± s.d.: native = 1.1 ± 0.21 MPa, n = 5; skirt = 1.15 ± 0.19 MPa, n = 5; cut = 1.34 ± 0.34 MPa, n = 4). This result suggests that our excavations of the root–stem junction did not materially disturb the plants in the cut treatments, and so any effects of that treatment may be attributed to cutting of the main root, and not to diffuse damage that our excavation work might have inflicted on some branches.

Fig. 1.

Nolana mollis xylem tensions, as measured by the pressure chamber, in response to no treatment (Native, ‘N’), installation of plastic skirts to prevent the possibility of brine dripping from leaves and stems to the soil and roots (Skirt, ‘S’), or severing of the root system at a point below the root collar but above any branch roots (Cut, ‘C’) to prevent possible water uptake from both deep and shallow roots; n = 5 plants per treatment. Morning sampling (AM panels) occurred at 0700 h between first light and sunrise, and afternoon sampling (PM panels) at 1500 h. Asterisks indicate significant differences between treatments within a sampling period (*P < 0.05, ***P < 0.001). Bars are means and whiskers are standard deviations for n = 4 or 5, and 3 for ‘Day 4 PM Cut’, due to mechanical failures of some samples during measurement. Due to extreme wilting of the Cut plants on Day 6, pressure chamber data are missing (red cross). Cloud and sun symbols represent overcast or clear sky conditions at the time of sampling.

Over the 6-d course of the experiment (Fig. 1), both the skirt and native treatments showed a consistent diurnal pattern between elevated tension in the afternoon (combined mean PM tension = 1.79 ± 0.3 MPa, n = 37) and recovery to lower tensions pre-dawn (combined mean AM tension = 1.13 ± 0.19 MPa, n = 48). On Day 1, afternoon tensions for the skirt and native treatments were not as high as on the other days: this may be explained by the fact that the sky on the first day was overcast, and so the energy loading and evaporative demand experienced by the plants would have been lower than on the subsequent sunny days. A light and brief drizzle of rain in the pre-dawn period of Day 4 had no observable impact on plant hydration, consistent with the observation that lightly scuffing the soil showed, by colour change, that the moisture did not penetrate the dusty soil surface. We interpret the observed sensitivity of the skirt and native xylem tensions to the daily cycle of transpiration and rehydration as reasonable assurance that our in situ decortication strategy did not prevent us from capturing physiologically meaningful variations in the water status of N. mollis.

In contrast to the skirt and native plants, the cut treatment rapidly and irreversibly developed large xylem tensions (Fig. 1). Afternoon (PM) sampling on Day 1 already showed a significantly elevated tension in the cut treatment, 0.6 MPa above the other groups (mean ± s.d.: native = 1.6 ± 0.44 MPa, n = 5; skirt = 1.45 ± 0.31 MPa, n = 4; cut = 2.17 ± 0.34 MPa, n = 4; ANOVA F_2,10 = 4.9, P = 0.033). Tensions in the cut treatment further increased to near 2.5 MPa by the next morning and reached a value near 3 MPa later in the afternoon of Day 2, where they remained until severe wilting on Day 6 prevented further measurement of the cut treatment.

Whole leaf water potential estimates from psychrometry

Psychrometer estimates of leaf water potential and leaf sap osmotic potential were collected on Day 0, and in the morning and afternoon of Days 1 and 4 (Fig. 2). Sampling was limited logistically to two replicates per treatment, and in a few cases contamination of the psychrometer chambers (detected by anomalous output as well as visually) reduced a sample to a single replicate. Nevertheless, the results for leaf water potential on Day 4 were consistent with the results for xylem tension, with the cut treatment drier by ~1 MPa than the native and skirt treatments in both the morning and afternoon.

Fig. 2.

Psychrometric measurements of leaf water potential (ψ) and osmotic potential (π) for Nolana mollis leaves sampled during the experiment in Fig. 1. Bars are means and whiskers are standard deviations for n = 3–4 for the aggregated samples collected on Day 0 and Day 1 (prior to cut), and n = 2 or 1 thereafter. Positive turgor is indicated by a magnitude of π greater than ψ .

Inferences of leaf turgor and xylem sap osmotic concentration

The difference between whole leaf water potential and osmotic potential, both measured by psychrometry, provides an estimate of turgor pressure. Aggregating samples across treatments for the mornings of Day 1 and Day 0 (i.e. prior to cutting of the roots in the cut treatment), we estimate pre-dawn turgor as 0.5 ± 0.21 MPa (n = 10). That these values are physiologically reasonable provides some assurance that our efforts to remove surface salts from the leaves before enclosing them in the psychrometer were successful, as any such salts should have produced very negative whole leaf water potentials and obscured the expected differences between leaf and osmotic potential.

By contrast, our estimates of xylem solute potential based on xylem exudate absorbed onto filter discs did not yield reasonable values (−1.34 ± 0.34 MPa, n = 10). These values, if combined with the xylem tension measurements, imply total pre-dawn leaf water potentials of about −3 MPa, almost twice as negative as the observed psychrometer measurements (Fig. 2, Day 0 + Day 1). We are inclined to trust the psychrometer measurements as it does not seem probable that our ‘dry’ cleaning of the leaves could have biased the psychrometer estimates of leaf water potential to the wet side (i.e. to be less negative and so wetter than the true value), since residual salt missed by our cleaning would have biased the psychrometer output to the dry (more negative) side. Instead, we suspect that during collection of xylem exudate some evaporative enrichment occurred, as fully saturating the leaf disc often required overpressures lasting a minute or more, during which time gas was also escaping through the stem turning the exudation into a spray, and the filter disc receiving the mix of sprayed droplets and bulk flow was also exposed to the surrounding air. We therefore alternatively calculated xylem osmotic potentials by evaluating the difference between xylem tensions measured by the pressure chamber, and whole leaf water potentials measured by psychrometery. For the aggregated pre-dawn data from Day 0 + Day 1, we estimate an average xylem solute potential of −0.49 ± 0.27 MPa (n = 10), about 10 times more concentrated than expected for mesic plants in non-saline soils (Boyer, 1995).

Comparison of N. mollis xylem tensions to those of a co-occurring woody shrub

To provide some ecological context for our observations of N. mollis, we also collected pre-dawn and afternoon water potential data for Gypothamnium pinifolium, a woody shrub presenting no difficulties for measuring in the pressure chamber. Sampling two plants over 3 d demonstrated that mean xylem tensions in Gypothamnium in both the morning and afternoon were about a megapascal larger than those seen in N. mollis (Fig. 3: mean ± s.d., AM = 2.18 ± 0.28 MPa, n = 6, and PM = 3.03 ± 0.2 MPa, n = 6, versus for N. mollis AM tension = 1.13 ± 0.19 MPa, n = 48; PM tension = 1.79 ± 0.3 MPa, n = 37). While we were only able to make two psychrometer measurements of whole leaf water potential in Gypothamnium, calculations of xylem solute potential were consistent with a more dilute sap (−0.18 and −0.03 MPa) than N. mollis. For another common shrub at our site, Tetragonia maritima, a non-salt-secreting semi-succulent with woody stems, a single set of pre-dawn pressure chamber and whole leaf water potential psychrometer measurements were in line with Gypothamnium, with xylem tension equal to 2.17 MPa, leaf water potential −2.26 MPa and a calculated xylem osmotic potential of −0.09 MPa.

Fig. 3.

Pre-dawn and afternoon mean xylem tensions and standard deviations for the salt-secreting N. molis and non-secreting G. pinifolium. For N. mollis, the data are aggregates of the Skirt and Native treatments over the whole experiment, with n = 48 for AM over 5 d, and n = 37 for PM over 4 d. For G. pinifolium, the data for both the AM and PM periods derive from two observations each day for 3 d (n = 6).

Soil measurements and root excavations

For each of the three N. mollis plants excavated, the tap root extended deeper than 1 m. Samples from the soil surface and 0.25 m depth were too dry to give interpretable output in the psychrometers, suggesting water potentials drier than −6 MPa. Three samples taken at 0.5 m depth had an average potential of −2.95 ± 0.72 MPa. Water potential measured at 0.75, 1 and 1.25 m was −1.96 ± 0.09 MPa (n = 3), approaching the range of values for pre-dawn leaf water potential observed on Days 0 and 1 (−1.81 ± 0.22 MPa, n = 10).

DISCUSSION

The failure of the cut treatment xylem tensions to relax to typical AM values overnight on any day, despite observation of a persistent presence of brine on the leaf and stem surfaces during all site visits throughout the study period (for a daily total of ~3 h split between pre-dawn and afternoon – see Methods), clearly eliminates the possibility that direct uptake of moisture from leaf and stem surface brines was the dominant process by which uncut N. mollis maintained water balance during our experiment. From a steady-state perspective, the lack of any differences between the skirt and native treatments further indicates that drip and uptake by shallow roots did not play an important role in supplying water for either transpiration or overnight rehydration over the course of the experiment. This leaves deep soil moisture as the only possible source of water for the observed diurnal maintenance of internal water balance in N. mollis.

An objection might be raised that it remains possible that the experiment did not run long enough for the skirted plants to deplete near-surface soil moisture that might have accumulated from drip prior to the installation of the skirt treatments. In this view, the cut treatment could have wilted due to a loss of connection to the top few centimetres of soil, rather than the loss of access to deep water. Against this hypothesis, we note that our psychrometer measurements indicated the soil water potential in the top 25 cm of soil was more negative than −6 MPa, and the soil was too dry to remain in the barrel of our soil-corer, suggesting that the absolute quantity of moisture in the upper soil layers prior to the experiment was negligible. Furthermore, on a return visit to the study site in 2023 we found that the skirted plants were all healthy and had increased in diameter since the installation of the skirts (which remained in place after the experiment) in 2018. Nothing remained of the cut plants but piles of white ash, as is characteristic of decomposition by dry oxidation in this hyper-arid environment.

While this field experiment has allowed us to make strong inferences about the dominant contributions to water balance, we cannot fully exclude the possibility that N. mollis extracts a minor component of its water budget from the brine on its shoot surfaces. We remain sceptical, as we know of no pumping mechanism by which plants can move water against its water potential gradient. In the case of Nolana, we expect this gradient to be always adverse: given a relative humidity that rarely exceeds 80–85 % (Mooney et al., 1980; Thompson et al., 2003), and an expected maximum nighttime leaf temperature depression of ~1.5 °C below ambient air, the water potential for the brine could range from −10 to −19 MPa (Fig. 4), versus a minimum leaf symplast water potential (wilting point, Fig. 1) for Nolana of −3.5 MPa. We constrain the nighttime leaf temperature depression based on an energy balance for a shoot under clear skies with a characteristic width of 2 cm, wind speed of 0.1 m s⁻¹, in evaporative equilibrium (i.e. maximum brine dilution) with ambient air at a relative humidity of 80 % (model code is provided in Supplementary Data Notes S2, adapted from Rockwell et al., 2014).

Fig. 4.

Water potential of a liquid phase in equilibrium with the relative humidity of ambient air. (A) The isothermal case that leaf temperature equals air temperature as calculated from the Kelvin equation (Nobel, 2005). We define an ‘unsaturated atmosphere’ conservatively as relative humidities (RH) below 90 %, corresponding to a water potential for an equilibrated liquid phase at the same temperature of ~−15 MPa. The onset of deliquescence for NaCl and ocean salts occurs at ~75 % RH, or −40 MPa. (B) The non-isothermal case that leaf temperature is below air temperature, calculated from the Clausius–Clapeyron–Kelvin equation (Rockwell et al., 2014). Leaf surface temperature depression has large effects on the water potentials that are in equilibrium with ambient RH. For Nolana our energy balance (Supplementary Data Notes S2) estimates 1.5 °C as an upper bound on leaf surface temperature depression at evaporative equilibrium with an atmosphere at 80–85 % RH, leading to a brine water potential of −10 to −19 MPa as shown in B. Given a lower bound for symplastic water potential of about −3.5 MPa before turgor loss, the water potential gradient from surface brine to symplast is expected to always be adverse, even as the surface brine grows and dilutes overnight.

Notwithstanding the absence of a known mechanism by which it could occur, a number of studies have taken inspiration from Mooney et al. (1980) to investigate other salt-exuding halophytes or arid land species and report evidence of water uptake from highly unsaturated sources (Wang et al., 2016; Coopman et al., 2021; Al-Handawi et al., 2023). A close look at these reports finds an absence of clear evidence. Al-Handawi et al. (2023) studied deliquescence, salt composition and surface chemistry of Tamarisk, and in addition reported uptake by leaf internal tissues of an aqueous fluorescent dye applied to the leaf surfaces. This cannot be considered evidence of uptake against a water potential gradient as the dye would have simply diffused along its own concentration gradient. Wang et al. (2016) used humidifiers to aerosolize an aqueous solution containing an apoplastic fluorescent tracer, and used hygrometers to monitor humidity within the treatment chambers, which was maintained at 75–95 % by manually switching the humidifiers on and off. They reported however that the leaves treated in this manner developed a thin water film on their surface, which suggests that locally, despite unsaturated conditions measured by the hygrometers in the air above the leaves, the leaves were in a saturated condition, as could occur by deposition of microdroplets produced by the humidifiers.

Coopman et al. (2021: figs 2 and 3), in a field study of the mangrove Aviccenia marina, report negative sap flow occurring at 85 % RH, which they attribute to leaf uptake of water from surface brines condensed by salts from the unsaturated air. The sap flow rates were plotted as means ± 1 s.e., which corresponds to a 68 % confidence interval. As these low confidence error bars touch zero in the region in which the mean was negative, doubling their span to find a 95 % confidence interval (i.e. ± 1.96 s.e.) would include zero, meaning these rates were not necessarily negative. This study also reported that leaf water potential converged with, and then marginally became less negative than, soil pore water potential during deliquesence, thereby providing a driving force for reverse sap flow. We note that leaf water potentials were measured with the pressure chamber without correction for xylem osmotic potential, and so leaf water potentials would have been more negative than reported, and gradient inversion to drive reverse sap flow therefore unlikely (Coopman et al., 2021: fig. 4). Finally, we note that in the most controlled experiment of Coopman et al. (2021: fig 7), sap flow of branches exposed to deliquescent conditions remained significantly positive, even as there was a clear demonstration of negative sap flow in branches washed with fresh water and bagged with fresh water sources to maintain saturated conditions: this manipulation shows that foliar uptake of pure water can occur in mangroves, but apparently not from brines with an equilibrium humidity of 85 % and water potential of −22 MPa (Fig. 4A).

At the physiological scale, oscillatory growth and shrinkage of brine droplets emerging from salt glands in epidermal peels of Avicennia officinalis might be interpreted as demonstrating water movement through salt glands against a water potential gradient (Tan et al., 2013). Yet despite these apparent transient reverse flows, brine volumes on the leaf surfaces increased over time in accord with the overall water potential gradient, raising the question of whether the gradient in water potential across the epidermis might have reversed periodically at the local scale due to some combination of unstirred layers, periodic solute movement and cellular damage from preparation of the peel.

To our knowledge, while plants that exude salts onto their surfaces do condense water from unsaturated (<90 % RH) atmospheres, such water remains far too dry to be accessed by the plant. Nevertheless, this hypothesized phenomenon continues to be misreported as fact (Ramadan, 2001; Wang et al., 2016; Dassanayake and Larkin, 2017; Coopman et al., 2021; Al-Handawi et al., 2023). Notably, all of the above reports characterize the chemical potential of external water (vapour and brine) in terms of RH, and so do not confront how adverse the leaf tissue to brine water potential gradient is when leaves are expected to be near or above ambient air temperature (Fig. 4A). At night, we acknowledge that leaf surface and brine temperature depression relative to ambient air, due to radiative cooling under clear skies, has the potential to dramatically shift the equilibrium water potential–RH relationship (Fig. 4B) into a regime more conducive of foliar uptake from undersaturated atmospheres. For example, a nighttime 80 % RH is in equilibrium with dew on a leaf sitting 3.5 °C below air temperature, though such large temperature depressions may be rare. In the Coopman et al. (2021) study, temperature effects on dew formation were discounted due to overcast conditions, and with respect to our study on N. mollis, overnight relaxation of daytime water stress proved to be contingent on the presence of roots.

A hydrological advantage hypothesis for Nolana

Given the high energetic costs calculated by Mooney et al. (1980), for example that harvesting atmospheric water from air at 80 % RH would supply only one-sixth of transpiration at the cost of more than half the energy available from respiration, a strategy of extracting water from daytime surface brines, even if possible, might offer little selective advantage. Yet, if N. mollis does not have access to atmospheric moisture as a means of hydration, what accounts for its ecological dominance? Our hypothesis, suggested by our estimates of large xylem osmotic potentials in N. mollis relative to Gypothamnium and Tetragonia, is that N. mollis excludes salts from the root xylem less efficiently than its competitors. A leaky filter would lead to less root fouling, lower root turnover and less tension in the xylem per unit transpiration. By allowing more salts into the xylem, less salt will remain in the rhizosphere of deep roots, which should allow deep roots to persist longer before being rendered non-functional by the accumulated salt content. Creating a more negative solute potential of the xylem sap lowers the xylem tension accompanying a given plant water potential, potentially lowering construction costs. While our own data are sparse, this interpretation is consistent with the data reported by Rundel et al. (1980) showing that N. mollis has the lowest xylem tensions in the healthy shrub class, and that across all species declining vigour (crown death) above ground is associated with increasing xylem tensions, as would be expected if rhizosphere osmotic concentrations increase due to salt accumulation (i.e. fouling). The second part of our hypothesis is that only N. mollis can tolerate such a leaky filter, as only N. mollis, in addition to sequestering salts in its tissues like most of the other shrubs of the quebrada (Rundel et al., 1980), has an additional sink for salts: foliar exportation via salt glands.

The quebrada shrub showing by far the lowest amount of salt sequestration in its tissues is Gypothamnium (Rundel et al., 1980). This implies that the roots of Gypothamnium achieve a very high reflection coefficient for salts, and perhaps alone among the shrubs of the quebrada follows a salt avoidance rather than tolerance strategy. Rundel et al. (1980) further report that most of the shrubs of the quebrada accumulate seed banks beneath them, with new individuals establishing in the aftermath of intense rainfall events, associated with the El Nino phase of the El Niño-Southern Oscillation (ENSO), that occur on a return interval of ~10 years (Rundel et al, 1991). Young plants then must grow their roots downward faster than soil moisture retreats away from the surface. The demographic data for Gypothamnium support this picture. In the data of Rundel et al. (1980), Gypothamnium accounted for 44 % of all individuals, indicating a high level of success in the post-rainfall establishment phase. Yet Gypothamnium only comprised 11 % of the currently healthy shrubs, implying a relatively high degree of success during establishment followed by a deterioration of relative survivorship as the system dried out. Such a trajectory would be expected if Gypothamnium were more susceptible to root fouling due to excluded salt accumulation than its competitors. Notably, in that same study N. mollis accounted for 25 % of the individuals, but 66 % of the healthy ones: less successful in the establishment phase, but increasing in relative survivorship as the soil dries down. The 11 other salt-accumulating (but not exuding) species established less successfully, most contributing only a few per cent of individuals, and all but one maintaining more vigorous canopies than Gypothamnium.

The advantages of salt exudation versus simple exclusion in the quebradas may also apply in other systems in which salt-excluding and salt-exuding woody plants co-occur, such as mangroves. Mangrove species that can exude salt from their leaves may tolerate higher salinity environments better than ultrafiltering salt-excluding species that rely on greater suberization and lignification of root exo- and endodermises (Cheng et al., 2014; Krishnamurthy et al., 2014). Exudation allows Avicennia germinans to tolerate a range of xylem sap osmotic potentials reported as wide as −0.24 to −1.36 MPa, after exclusion of about 85 % of root zone salt. By contrast, ultrafiltering species exclude up to 99 % of soil pore water salt (Scholander et al., 1966; Paliyavuth et al., 2004; Lopez-Portillo et al., 2005; Krishnamurthy et al., 2014). Tolerance of such large salt concentrations in the xylem should, in theory, allow Avicennia to maintain water uptake even as the root zone becomes hypersaline relative to seawater due to the balance of convection of salt toward, and back diffusion away from, active roots in sediments 10–15 cm below the free flowing tide (Passioura et al., 1992). Even in this tidally flushed system, modelling suggests that the balance of salt transport at the root surface limits expected mangrove transpiration to about one-sixth of potential evaporation, and the problem worsens in more stagnant areas (Passioura et al., 1992). Ultimately, foliar salts must end up back at the surface, transported either by rainfall or leaf shedding, but where tidal flushing is efficient it prevents the transport of this salt back to the root surface, maintaining the sediment surface at ocean levels of salinity (Passioura et al., 1992).

In irrigated or rain-fed saline terrestrial systems where inputs of water limit evapotranspiration, foliar salt exclusion appears potentially futile in the long run, as water uptake transports surface-deposited salt back to the root zone. Yet, in a terrestrial system with little annual rainfall like the Atacama, salts transmitted through the xylem and exuded to the surface of the leaf (and ultimately to the surface soil) are effectively sequestered away from the root zone. Thus, salt exudation by Nolana should allow the plant to slow the accumulation of rhizosphere salt that inevitably accompanies transpiration relative to other shrubs. Resolving this question requires further exploration of soil salinity profiles underneath the plants of the quebrada, and measurements of xylem salt concentrations, in relation to shrub health and transpiration and assimilation rates.

The hydrological hypotheses discussed above are focused on below-ground processes, as our experiment implicates deep soil water as critical for hydration maintenance in N. mollis. However, above-ground processes linked to the surface brine may also contribute to the ecological dominance of Nolana in the quebrada. As the brine grows at night and shrinks during the day (Mooney et al., 1980), daytime evaporation of the brine would be expected to suppress water loss from the stomata through two channels: first, by humidifying the boundary layers of the leaves and plant canopy and so quenching evaporative demand, and second, through latent cooling of the leaf surfaces. Detailed modelling of these processes is required to test whether these expected effects are likely to be physiologically important in the environment of the coastal Atacama, and to show how brine formation provides sufficient benefits to offset the energetic costs of salt transporters (Munns et al., 2020)

CONCLUSION

Given the sparseness of the land surface covered by plants in the coastal valleys of the Atacama, the ecological dominance of N. mollis points towards some physiological advantage in survival and growth. Previous work has suggested the unique presence on N. mollis of a brine condensed from unsaturated air by salts exuded onto its surfaces as the source of its advantage, positing the existence of an unknown mechanism by which a plant could extract water from the brine against its free energy gradient. Here we tested the hypothesis that this brine provides N. mollis with a source of atmospheric water, unavailable to other species, capable of replacing the water inevitably lost due to carbon uptake through stomata. Our results do not support the hypothesis that an atmospheric source maintains water balance in N. mollis, and instead indicate a dependence on deep (below 1 m) soil moisture. We hypothesize that the ecological dominance of N. mollis derives from partial mitigation of root fouling problems by salt exudation, and an increased water use efficiency of carbon gain by daytime evaporation of the resulting nocturnally condensed brine. More broadly, our work, and our review of reports purporting to show the capacity of plants to uptake water from a significantly unsaturated (<90 % RH) atmosphere, calls into question whether such uptake is a demonstrated phenomenon.

SUPPLEMENTARY DATA

Supplementary data are available at Annals of Botany online and consist of the following.

Notes S1: Code for the Campbell CR6 datalogger psychrometer program. Notes S2: Code for the leaf energy balance model.

AUTHOR CONTRIBUTIONS

NMH and JD designed the study. JG, AM, UH and FER developed the measurement protocols. AM, JG, FER, JD, NA and PL collected the data. FER, JG, AM, UH, NMH and JD analysed the data and wrote the manuscript.

FUNDING

This project was financed by a Harvard-UAI Collaborative Research grant from the David Rockefeller Center for Latin American Studies, by the National Science Foundation (NSF-MRSEC DMR-2011754), and by the US Air Force Office of Sponsored Research (AFOSR FA9550-21-1-0283).

CONFLICT OF INTEREST

The authors have no competing interests.