Water ‘on the rocks’: a summer drink for thirsty trees?

Water ‘on the rocks’: a summer drink for thirsty trees?

Andrea Nardini

, Francesco Petruzzellis

, Daniel Marusig

, Martina Tomasella

, Sara Natale

,

Alfredo Altobelli

, Chiara Calligaris

, Gabriele Floriddia

, Franco Cucchi

, Emanuele Forte

and Luca Zini

Author for correspondence: Andrea Nardini Email: nardini@units.it Received: 12 June 2020 Accepted: 27 July 2020 New Phytologist (2021)229: 199–212 doi: 10.1111/nph.16859

Key words: available water, bedrock, dolostone, drought, karst, limestone, tree mortality, water potential.

Summary

Drought-induced tree mortality frequently occurs in patches with different spatial and tem-poral distributions, which is only partly explained by inter- and intraspecific variation in drought tolerance. We investigated whether bedrock properties, with special reference to rock water storage capacity, affects tree water status and drought response in a rock-domi-nated landscape.

We measured primary porosity and available water content of breccia (B) and dolostone (D) rocks. Saplings of Fraxinus ornus were grown in pots filled with soil or soil mixed with B and D rocks, and subjected to an experimental drought. Finally, we measured seasonal changes in water status of trees in field sites overlying B or D bedrock.

B rocks were more porous and stored more available water than D rocks. Potted saplings grown with D rocks had less biomass and suffered more severe water stress than those with B rocks. Trees in sites with B bedrock had more favourable water status than those on D bedrock which also suffered drought-induced canopy dieback.

Bedrock represents an important water source for plants under drought. Different bedrock features translate into contrasting below-ground water availability, leading to landscape-level heterogeneity of the impact of drought on tree water status and dieback.

Introduction

Over the last decade, an overwhelming body of evidence has shown that anomalous drought and heat waves, probably associ-ated with global warming, are increasing tree mortality rates and forest decline in several biomes (Allen et al., 2010; Carnicer et al., 2011; Menezes-Silva et al., 2019). Because of significant ecologi-cal (Anderegg et al., 2013) and economic impacts (Hanewinkel et al., 2013) of drought-induced disruption of forests, the possi-ble physiological drivers of tree decline have been intensively scrutinized (McDowell et al., 2013; Choat et al., 2018). The mechanistic causes of tree death under drought are debated, but there is consensus on the ubiquitous association among drought, water stress, hydraulic failure, desiccation and mortality (Nardini et al., 2013; Rowland et al., 2015; Adams et al., 2017; Hammond et al., 2019; Tomasella et al., 2019). Co-occurring factors like carbon starvation (McDowell et al., 2011) also play a role (Sevanto et al., 2014; Savi et al., 2019a) and set the functional limits for recovery of tree physiological status following drought relief (Savi et al., 2016; Klein et al., 2018; Pagliarani et al., 2019; Tomasella et al., 2020).

Drought-induced tree mortality does not occur over uniform spatial and temporal scales, but rather sequentially affects differ-ent species and/or conspecific individuals in patches of differdiffer-ent extension and distribution (MacGregor & O’Connor, 2002; de Toledo et al., 2011; Baguskas et al., 2014). This is partly

explained by species-specific differences in vulnerability to hydraulic dysfunction (Lopez et al., 2005; Nardini et al., 2013). Moreover, even different individuals of the same species in a pop-ulation might display different tolerance to water stress (Rice et al., 2004; Petrucco et al., 2017). However, experimental evi-dence suggests that intraspecific variability of vulnerability to xylem embolism is generally low (Matzner et al., 2001; Skelton et al., 2019). Inter- or intraspecific differences in rooting depth have been also related to spatial variability in tree decline (Padilla & Pugnaire, 2007; Nardini et al., 2016; Chitra-Tarak et al., 2018), calling for more studies on below-ground components of tree resistance and resilience to anomalous drought (Pierret et al., 2018; Brum et al., 2019).

Another important driver of spatial variability in drought-in-duced tree mortality is landscape heterogeneity in terms of topog-raphy, aspect and soil properties (Tai et al., 2017; Schwantes et al., 2018). As an example, the probability of tree mortality in a tropical forest was higher on steep slopes or on sandy soils in val-leys than on plateau areas with clayey soils (de Toledo et al., 2011). In a semiarid environment dominated by Pinus halepensis, mortality risk was highest for trees growing in older-aged stands on southern aspects (Dorman et al., 2015). Very local soil and subsoil differences in terms of depth and available water influ-enced the response to drought of Quercus ilex individuals (Carri´ere et al., 2020), leading to differential mortality risk under drought.

Ó 2020 The Authors

New PhytologistÓ 2020 New Phytologist Trust

New Phytologist (2021) 229: 199–212 199

A frequently overlooked determinant of ecosystem hydrology is the rock component, including bedrock features (Schwinning, 2010; Dawson et al., 2020). Several studies investigating forest responses to drought were focused on habitats with relatively deep soils that contribute to the largest share of water available to plants over the growing season (Elliott & Swank, 1994; Fensham & Fairfax, 1997; Renne et al., 2019). However, large portions of terrestrial habitats are dominated by forests on shallow soils over-laying substrates that strongly limit root growth and water stor-age. Indeed, a very large fraction of terrestrial habitats is characterized by a soil depth to bedrock < 50 cm (Shangguan et al., 2017). This is particularly apparent for karst habitats, char-acterized by soluble rocks like limestone, dolostone and gypsum (Geekiyanage et al., 2019). Karst soils are generally 5–30 cm deep (Querejeta et al., 2007; Nie et al., 2012), and hence can hold only very small water volumes. The karst bedrock is charac-terized by fractures enlarged by dissolution and filled by soil, rep-resenting a potential water source for plants (Nardini et al., 2016; Savi et al., 2018). However, the overall soil and cracks volume is a minor percentage (< 20%) of the epikarst (Estrada-Medina et al., 2013), where most of the volume is occupied by rocks. Notably, the rock matrix itself can accumulate substantial amounts of water in the primary micropores which can con-tribute a significant percentage of rock volume (Estrada-Medina et al., 2013), leading to storage capacities of potentially available water in the range 0.1–0.2 m3_m−3 _{(Jones & Graham, 1993;} Schwinning, 2010), and thus representing a significant compo-nent of the hydrologic cycle. Rempe & Dietrich (2018) have shown that up to 27% of annual rainfall is seasonally stored as rock moisture in northern Californian Mediterranean forest ecosystems. Similarly, in a deciduous forest on a limestone plateau in Yucat´an, rock water storage is much higher than that of soil, owing to the larger volume fraction occupied by rocks (Estrada-Medina et al., 2013).

Despite the potential importance of the rock matrix as a water source for vegetation, the capacity of plants to exploit this water pool has rarely been investigated. Roots of woody perennials grow through fractured rock in Karst habitats (Supporting Infor-mation Fig. S1) and access soil pockets in weathered bedrock, caves and cracks (Sternberg et al., 1996; Rose et al., 2003; Quere-jeta et al., 2006; Kukowski et al., 2013; Swaffer et al., 2013; Nar-dini et al., 2016; Savi et al., 2019b). However, the possible contribution of water held in the rock matrix to the seasonal plant water balance remains largely unknown (Dawson et al., 2020). Estrada-Medina et al. (2013) showed that several woody species have large numbers of root tips in direct contact with rocks. Still, the root–rock interface might be an inefficient water pathway for exploitation of rock moisture by plants, because of its low hydraulic conductivity (Hubbert et al., 2001), although root-associated mycorrhizal fungi, as a result of narrow hyphal diameter and extensive length, might improve the extraction of water (Bornyasz et al., 2005; Allen, 2007). All these data suggest that water in the rocks might help plants to buffer water stress in habitats characterized by low soil water storage, but the potential influence of different rock porosity and water storage on plant responses to drought remains unexplored. Differences in bedrock

type have been reported to affect vegetation productivity and responses to drought (Hahm et al., 2019; Reed & Kaye, 2020). Another recent study (Preisler et al., 2019) has reported surface rock cover and soil stoniness as the only variables explaining tree mortality differences between stands of P. halepensis, identical in terms of topography, tree age or stand density. This was inter-preted as a ‘mulching effect’ of surface rocks, protecting the soil from evaporation and maintaining higher soil water availability under drought.

We report a multilevel analysis of the possible impact of differ-ent rock porosities on plant water relations and drought responses. We took advantage of the heterogeneous nature of bedrock in Classical Karst (northeastern Italy) areas covered by similar vegetation types, coupled to the observed spatial hetero-geneity of the impact of summer droughts on woody vegetation. We aimed at: quantifying the amount of water stored in the porous space of different rock types, and potentially available to plants; checking the impact of contrasting rock porosities on plant responses to water stress, both under controlled drought on potted saplings and in field-growing trees on a seasonal scale; and verifying the impact of bedrock types on vegetation productivity and water content by remote sensing.

Materials and Methods

Study sites, vegetation and geological background

The study area is located in the Classical Karst Plateau, extending for over 750 km2between Italy and Slovenia. This is a highly karstified hydrostructure (Zini et al., 2015), characterized by very high permeability (Urbanc et al., 2012) and almost complete absence of surface flow (Calligaris et al., 2018). Soils are generally shallow (< 30 cm) and overlie a carbonate bedrock dominated by limestone and dolostone. The study sites are located in the province of Trieste, near the village of Sgonico, and are character-ized by Pinus nigra plantations (Petrucco et al., 2017) intermixed with deciduous woodlands dominated by Fraxinus ornus and Quercus pubescens (Nardini et al., 2003). The climate is sub-Mediterranean, with warm summers (average July temperature: 22.7°C), and relatively cold winters (average January tempera-ture: 3.9°C). Annual rainfall totals 1200 mm, concentrated in spring and autumn. The area has experienced anomalous heat and drought waves over the last 20 yr, leading to spatially hetero-geneous impacts on vegetation (Nardini et al., 2013; Petrucco et al., 2017). Fig. S2 reports maximum and minimum air tem-peratures, as well as daily precipitation recorded during the exper-imental period in 2019 (www.osmer.fvg.it; Sgonico weather station,< 1 km from study sites).

the carbonate platform during the Cenomanian. This phase is subjected to erosion and karstification, and defines the bottom of the Monrupino Formation overlaid by pale grey, dark grey, blackish layers, black saccharoidal, fine grain-sized dolostones. Within these horizons, pockets of reddish and yellowish dolomitic breccia can be distinguished (Colizza et al., 1989). The breccia is overlaid by dark grainy bituminous limestone and platy laminated limestone rich in fossils.

A 30 m resolution digital elevation model of the study area was processed using ‘r.slope.aspect’ in GRASSv.7.8, to generate a

slope and exposure map, which was overlaid on the geological map of the area (Jurkovˇsek et al., 2016). On this basis, we selected 10 sites with an area of 10 000 m2 each, southeast-ex-posed, with similar slope (between 10% and 30%) and altitude (300–400 m above sea level (asl)), and hosting the same wood-land dominated by F. ornus and Q. pubescens (Fig. 1). Five sites were dominated by breccia bedrock, and the other five were char-acterized by dolostone bedrock. These sites were analysed for veg-etation responses to summer drought progression via remote sensing. For sampling of rocks and soils and for field measure-ments, we selected two sites (one on breccia, B; the other on dolostone, D) from those identified, extending the experimental area to 20 000 m2(Fig. S3). All measurements and analyses were performed between May and September 2019.

Ground penetrating radar

Ground-penetrating radar (GPR) is a pulsed electromagnetic geophysical technique allowing detection of electrical discontinu-ities below the ground (Jol, 2009). It is an efficient, noninvasive and high-resolution system to image the subsurface (Jayawick-reme et al., 2014). We used GPR to characterize soil and bedrock layering in B and D sites (Sucre et al., 2011). Data were collected

on 26 August 2019 with a ProEx device (Mal˚a Geoscience, Mal˚a, Sweden), equipped with 500 MHz shielded antennas, in com-mon offset configuration. GPR was triggered by an electrome-chanical odometer every 5 cm. At each site, we acquired four 100-m-long profiles randomly positioned, to obtain statistically sound information on soil thickness and bedrock characteristics. After dataset collection, we applied a processing flow including drift removal, spectral analysis, bandpass filtering, geometrical spreading correction, exponential amplitude correction and depth conversion. The latter step was performed using a constant electromagnetic wave velocity of 12.2 and 10.0 cm ns–1for sites B and D, respectively, as derived by averaging the results obtained by dedicated diffraction hyperbola fitting analyses.

Ground-penetrating radar profiles were analysed considering not only reflection (and diffraction) amplitudes and lateral phase continuity, but also specifically calculated attributes (Zhao et al., 2018). In particular, amplitude-related attributes like reflection strength discriminate GPR signatures between shallow soil with rocks, and bedrock (Fig. 1a,b). For further details on GPR attributes, readers are referred to Zhao et al. (2016).

Measurements of rock porosity

Breccia and dolostone rock samples were collected from B and D sites, and crushed to a final volume of about 1.5 cm3(15 samples per rock type). Samples were oven-dried at 70°C for 48 h, weighed (Wi), and their volume (Vrock) was measured using the water displacement method (Hughes, 2005). This took< 20 s, reducing the possibility of water entry into rock pores. Samples were water-saturated in a vacuum chamber at−70 kPa for 48 h, to obtain final weight (Wf). The pore volume for each sample (Vpores) was calculated as the difference between Wfand Wi, cor-rected for water density, and expressed as a percentage of Vrock.

Moisture release curves for rocks and soils

Breccia and dolostones rock samples were collected as above. We also collected soil samples from B and D sites, and local red soil from a nearby vineyard as a potting mix for the glasshouse experi-ment (see later). Rocks were crushed to fragexperi-ments about 30× 20 × 5 mm (five samples per rock type), to be hosted in the sample holder of a dewpoint hygrometer (WP4; Decagon devices Inc., Pullman, WA, USA) and measured for rock water potential (Ψrock). Rock samples were vacuum-infiltrated with water (see earlier), and surface water was removed by quickly touching them with filter paper. Samples were suspended in sealed vials, with the base filled with a 0.05 m KCl solution (π = −0.23 MPa) for 24 h, to favour evaporation of residual surface water while allowing equilibration of Ψrock to a near-zero value. After preparation, sequential measurements of Ψrock and corresponding weight (FWrock) were taken during controlled dehydration. After each dehydration step, samples were sealed in the sample holder for 1 h before measuring Ψrock. Experiments were stopped at Ψrock≤ −5 MPa, and samples were oven-dried for 72 h at 70°C to obtain DW (DWrock). Water content (WCrock) was calculated as:

WCrock¼ FWð rock DWrockÞ  DW1rock

Ψrockvalues were plotted against corresponding WCrock, thus obtaining moisture release curves that were modelled as exponen-tial decay regressions (SIGMAPLOT v.13.0; SyStat Software Inc.,

San Jose, CA, USA). The fraction of available water content (AWCrock) for plant uptake was calculated as the difference between WCrock values at −0.2 and −1.5 MPa, for both B (AWCrockB) and D (AWCrockD) samples.

Finally, the volumes of rock fragments (Vrock) were measured as earlier, and rock density (drock) was calculated as:

drock¼ DWrock V1rock

Soil moisture release curves were obtained following the proce-dure described earlier on five samples per soil type, after rehydra-tion to field capacity. AWC for soils of B (AWCsoilB) and D sites (AWCsoilD), as well as for that from the vineyard (AWCsoilV), were calculated.

Glasshouse experiment

To test the hypothesis that different Vpores and AWCrock can influence plant water status under drought, we initially set up an experiment on potted plants. A total of 105 saplings of F. ornus L. (3 yr old) were provided by a public nursery (Vivai Pascul, Tarcento, Italy). This species was selected because it is abundant in the field sites (see later), and because of its water-use strategy, characterized by a substantial degree of anisohydry (Tomasella et al., 2019), leading to climate- and habitat-related variations of plant water status over temporal (Nardini et al., 2003) and spatial (Gortan et al., 2009) scales. Plants were transported on 21 March 2019 to the glasshouses of the

Botanical Garden of University of Trieste. The original sub-strate was removed under a gentle jet of water. Plants were randomly divided into three groups and transplanted in 1600 cm3pots filled with three different mixtures (Fig. S4). Pots of the group S (soil) were completely filled with red soil. Pots of the other two groups were filled with c. 1050 cm3 of red soil mixed to c. 550 cm3 of rock fragments, of either breccia (group B) or dolostones (group D). Rock fragments had vol-umes of 50–70 cm3_{each (eight to 11 fragments per pot), and} were sunk in water for 24 h before potting to favour water sat-uration of the porous spaces. Plants were placed in the glasshouse in a grid of six rows, and their position was ran-domly shifted on a weekly basis to avoid artefacts from even-tual exposure of different individuals to even slightly different environmental conditions. Plants were irrigated daily to field capacity in the early morning for 95 d, and fertilized every 20 d with a liquid fertilizer (COMPO Agro Specialities Srl, Mon-za, Italy).

The 35 plants from each group were then divided into two sets. One set, comprising 30 plants, was subjected to drought stress by suspending irrigation, while other five plants continued to be irrigated. We measured a set of physiological parameters between 12:30 and 13:30 h (solar time), on a daily basis on five plants per group. This time interval was selected because previous tests on the same species grown in the same glasshouse had shown that plants reached minimum daily water potential (see later) around this time of day. Leaf conductance to water vapour (gL) was measured using a steady-state porometer (SC-1, Decagon Devices Inc.). The same leaf measured for gLwas used to measure minimum water potential (Ψmin) with a pressure chamber (1505D, PMS Instruments Co., Albany, OR, USA). A second leaf was weighed to obtain FWleaf, and rehydrated for 3 h with the rachis immersed in water. The leaf turgid weight (TWleaf) was measured, and samples were oven-dried at 70°C for 24 h, to obtain DWleaf. For each leaf, water content (WCleaf) was calcu-lated as:

WCleaf¼ FWð leaf DWleafÞ  FW1leaf

The drought treatment continued until a significant fraction of plants had reachedΨmin≤ −5 MPa, which occurred after 11 d. During the experiment, plants reachingΨmin≤ −5 MPa were progressively counted and discarded, and each plant was used for a maximum of two rounds of measurements, to avoid artefacts as a result of defoliation. TheΨminthreshold was based on the vul-nerability curve of F. ornus stems, showing 80% loss of hydraulic conductivity at the corresponding xylem pressure (Petit et al., 2016; Kiorapostolou et al., 2019; Petruzzellis et al., 2019), a value that has been reported as lethal for woody angiosperms (Barigah et al., 2013). This response was confirmed by observed desiccation of leaves at 24 h afterΨmin≤ −5 MPa was recorded. At the end of the treatment, plants withΨmin> −5 MPa were re-irrigated to field capacity, and measurements were repeated on 10 saplings per group after 2 or 4 d of recovery.

Relative growth rates in height (RGRh) and diameter (RGRd) were calculated as:

RGR¼ final value  initial valueð Þ  initial valueð Þ1 Leaves, stems and roots were harvested from 10 individuals per group, and oven-dried at 70°C for 24 h, to obtain the respective biomasses.

Field measurements

Field measurements were performed in sites B and D in sunny days on 16 May, 10 June, 27 June, 25 July and 12 August 2019. At each site, six adult individuals of F. ornus (height 4–-6 m) were randomly selected (Fig. 1). On each date, four sun-exposed leaves were sampled from each tree between 12:30 and 13:00 h (solar time). Leaves were wrapped in plastic film, stored in a cool bag and transported to the laboratory. Two leaves were used to measure Ψmin within 1 h from sampling. The other two leaves were used to measure WCleaf following the same procedure described earlier. Also leaf turgid weight (TWleaf) was measured, by fully rehydrating leaves before oven dehydration, and relative water content (RWCleaf) was calcu-lated as:

RWCleaf¼ FWð leaf DWleafÞ  TWð leaf DWleafÞ1

On 27 June and 25 July, two leaves per tree were sampled between 04:30 and 05:30 h (solar time) to measure predawn water potential (Ψpd) as a proxy for soil water potential.

Remote sensing analysis

Multispectral images of the study sites from the satellite Sentinel-2A/-2B level 2A tile 33TUL (bottom atmosphere), were retrieved from the Copernicus Open Access Hub (https://scihub.Copernic us.eu/). A series of 17 cloud-free images were selected from early June to late August 2019. Vegetation indices were calculated using optical spectral bands of red (B04, 665 nm), near-infrared (B08, 833 nm), and short-wave-infrared (B11, 1610 nm), with the ‘r.mapcalc’ function in GRASS, setting the boundaries for the

sampling areas with the ‘g.region’ and ‘r.mask’ functions. We cal-culated normalized difference vegetation index (NDVI) as a proxy of green biomass, as follows:

NDVI¼ρNIR ρred ρNIRþ ρred

where ρNIR is the near-infrared reflectance, and ρred is the red light reflectance. We also calculated the normalized difference water index (NDWI), which has been shown to correlate to the water status of F. ornus trees in our study area (Marusig et al., 2020). We used this index as a proxy for large-scale seasonal changes in water status of vegetation growing on bedrocks with contrasting porosities. NDWI was calculated as:

NDWI¼ρNIR ρSWIR ρNIRþ ρSWIR

whereρSWIRis the short-wave-infrared reflectance (1610 nm). Statistical analysis

Statistical analyses were done with R v.3.6.3 (R Core Team, 2019). Differences between mean values of AWCrockB and AWCrockD were assessed by Student’s t-test, after checking for data normality and homoscedasticity of variance assumptions. A bootstrap procedure was applied to calculate mean values, 2.5% and 97.5% confidence intervals (CIs) of AWC for soils. Specifi-cally, we applied a resampling procedure randomly sampling a restricted number of points (15 for each soil) from the moisture release curves measured on soils. Then, we modelled the subset of points with an exponential decay function and we calculated the AWC of each soil. This procedure was repeated 999 times, and then mean values and 2.5% and 97.5% CIs of AWCsoil were obtained. Differences between soil types were considered as sig-nificant when the CIs did not overlap.

Differences in WCleaf,Ψminand gLmeasured on potted plants were tested with two-way ANOVA analyses. One test was run separately for each parameter, setting day, treatment and their interaction as the response variables. Differences in RGR and biomass between treatments were tested though one-way ANOVA analyses. In these analyses, each parameter was the response variable while treatment was set as the explanatory vari-able. Normality and homogeneity of variance assumptions were checked for each analysis. For significant tests, Tukey’s honestly significant difference post hoc analysis with Bonferroni’s adjust-ment for P-values was run.

Differences of physiological parameters measured in the field (WCleaf, Ψpd and Ψmin), and remote sensing indices between sampling sites one each date were tested through linear mixed models (LMMs), using the ‘lmer’ function in theLME4 package.

One Gaussian model was calculated for each physiological parameter and remote sensing index (dependent variables), while site (B or D), sampling/acquisition date and their interaction were the fixed effects. The random intercept was the sampling/ac-quisition date nested in site. The pseudo-R2of each model was calculated using the ‘r2beta’ function in MUMINpackage.

Results

Soil and bedrock features

more compact. However, the variation of reflection strength with depth (Fig. 2c,d), as well as the depth-related variation of GPR signal (Fig. S5) allowed us to estimate a soil layer (mixed with rock fragments) of 80–90 cm at site D and 70–80 cm at site B. At both sites, peculiar situations can be observed with locally outcropping rocks or moderate local thickening of soil.

While relative soil and bedrock thickness were similar at the two sites, breccia and dolostone rocks displayed different features

in terms of density, porosity and capacity for water storage. The drockof B and D averaged 2.4 and 2.7 g cm−3, respectively. Large differences emerged in terms of primary porosity, with Vporesof 6.1 2.1% in B rocks, but only 1.4  0.6% in D rocks. The variation in primary porosity translated into clear differences in AWCrock, and values were not negligible, albeit much smaller than that of soils (see later) (Fig. 3). Less porous dolostones had AWCrockDaveraging 5 mg g−1, but the AWCrockBof breccia was

(a) (c)

(b) (d)

Fig. 2 (a, b) Comparison between two examples of processed profiles collected at a field site dominated by dolostone bedrock (a) and one dominated by breccia bedrock (b). The average vertical trend of the reflection strength (see text for details) is plotted as a function of depth for both profiles in (c) and (d), respectively. Black boxes depict the close-up view shown in Supporting Information Fig. S5.

(a) (b)

Fig. 3 (a) Moisture release curves as measured for rock samples of breccia (B, blue circles) and dolostone (D, black circles). Values of available water content (AWCrock), calculated as the difference between water content at field capacity (dashed line) and a standard permament wilting point of

−1.5 MPa (dotted line) are also reported as means  SD. (b) Moisture release curves as measured for soil samples collected at sites dominated by breccia bedrock (blue circles), dolostone bedrock (black circles), or in a nearby vineyard (red circles). Values of available water content (AWCsoil) were calculated

more than double (12 mg g−1), albeit with larger variability among samples. Soils from B and D sites showed similar mois-ture release curves, with AWCsoil between 54 and 58 mg g−1. The red soil collected in the vineyard and used for the glasshouse experiment had higher AWCsoilV(73 mg g−1).

On the basis of GPR profiles, drock and AWC of soils and rocks, and assuming a conservative maximum rooting depth of 5 m (Kukowski et al., 2013; Nardini et al., 2016; Savi et al., 2019), we calculated that the maximum amounts of total avail-able water stored at B and D sites and accessible to root systems were 150 and 190 mm, respectively.

Growth and drought response of plants potted with soil or soil–rock mixtures

Saplings planted in red soil or soil mixed with B or D rocks were subjected to a controlled drought. Under well-watered condi-tions, as well as during drought progression, daily averages of WCleaf, Ψminand gL showed no significant differences between groups (Fig. 4), although S samples had a tendency to maintain higherΨminafter day 6. After re-irrigation, all three groups par-tially recovered their water status, but not to values recorded in controls, even 4 d after irrigation.

Despite similar trends in mean values of water status, the num-ber of plants that reached Ψmin≤ −5 MPa during drought increased at a faster daily rate in group D than in group B or S (Fig. 5). After 8 d from the start of the treatment, no S plant had reached this critical threshold, compared with 2% and 10% of B and D plants, respectively. Three days later, the fractions of plants that surpassed the threshold and underwent leaf

desiccation had increased to 32%, 34% and 50% in S, B and D groups, respectively.

Plants growing in the soil–rock mixture had lower RGRhand RGRdcompared with S plants (Fig. 6). At the end of the experi-ment all plants had a similar biomass of stem and roots (Fig. 7), but leaf biomass was significantly lower in D plants (by c. 30%) than in both S and B plants.

Seasonal changes of plant water status in breccia and dolostone sites

In mid-May, plants at B and D sites had an overall favourable water status (Fig. 8). RWCleaf was close to 0.95 (Fig. 8b) and WCleafaveraged 0.7 g g−1, with no significant difference between sites (Fig. 8a).Ψminwas> −1 MPa at both sites (Fig. 8c). During late spring and early summer, only occasional rainfall events occurred (Fig. S2), leading to progressive deterioration of plant water status. By the end of July, RWCleaf was about 0.67 in D plants and 0.77 in B plants, but the difference was not significant. Significant differences emerged for both WCleafandΨmin. In par-ticular, WCleafaveraged 0.50 g g−1in D plants and 0.56 g g−1in B plants.Ψminaveraged−2.8 MPa in B plants, but was close to −4 MPa in D plants. Different water status of B and D plants was confirmed byΨpd which in late June was still relatively high (c. −0.5 MPa) and not significantly different between sites (Fig. 8c). One month later,Ψpddropped significantly, to c.−2.3 MPa in B plants and−3.4 MPa in D plants. Following abundant rainfall in late July (Fig. S2), plant water status partially recovered. WCleaf increased to 0.54 and 0.58 g g−1 at D and B sites, respectively. Ψminincreased to−2.0 MPa, with no difference between sites.

(a)

(b)

(c) Fig. 4 Changes in leaf water content (a,

WCleaf), minimum daily water potential (b,

Ψmin) and maximum leaf conductance to

water vapour (c, gL) recorded in potted

Remotely sensed vegetation biomass and water status in breccia and dolostone sites

Maximum NDVI was reached by early June, when it averaged values of 0.9 at both B and D sites (Fig. 9). NDVI remained sta-tistically similar at B and D sites until late July. From this date on, NDVI was significantly higher at B sites than at D ones. Despite similar NDVI, NDWI was significantly higher at B than at D sites throughout the season. Maximum NDWI was recorded in late May, when it averaged 0.45 and 0.40 at B and D sites, respectively. NDWI decreased progressively during summer, reaching a minimum in late July, when B sites had an average NDWI of 0.40, compared with 0.30 for D sites. After late July rainfalls, NDWI increased but not to values recorded in May.

Discussion

Karst habitats dominated by shallow soils overlying massive bedrock are challenging for plants, because of limited under-ground water storage. Roots of woody perennials can grow deeply into cracks and fissures of the bedrock, exploring caves and soil pockets (Nardini et al., 2016; Savi et al., 2018; Carri´ere et al., 2020; Dawson et al., 2020). However, not all plant species thriv-ing in rock-dominated landscapes can rely on deep root systems (Nie et al., 2014), raising the hypothesis that the bedrock itself represents a water source for vegetation (Liu et al., 2014). Our data suggest that rocks store water available to plants. Glasshouse and field experiments confirmed that plants having access to highly porous rocks maintain a more favourable water status than those in contact with less porous ones, indicating that plants can exploit rock moisture and that rock-available water potentially limits plant water status and survival under drought. This might provide an additional explanation for spatial and temporal vari-ability of canopy dieback over narrow geographical ranges.

Moisture release curves revealed that carbonate rocks store measurable amounts of water at free energy values compatible with extraction and use by plants’ roots. Based on the conven-tional permanent wilting point of −1.5 MPa, AWCrock values were 5 and 12 mg g−1 for dolostone and breccia, respectively. Contrasting AWCrock values were consistent with differences in

Fig. 5 Percentage of saplings of Fraxinus ornus reaching or surpassing the critical water potential of−5 MPa during an experimental drought. Plants were growing in pots filled with soil (red bars), or a mixture of soil and rock fragments of breccia (blue bars) or dolostone (black bars).

Fig. 6 Median values, 25th_{and 75}th_{percentiles of relative growth rate in}

terms of plant height (RGRH) (a) or basal diameter (RGRD) (b) as recorded

in potted saplings of Fraxinus ornus during the whole duration of the experiment (106 d). Plants were growing in pots filled with soil (S), or a mixture of soil and rock fragments of breccia (B) or dolostone (D). Different letters denote significant differences between treatments.

Vporesand drock, and in agreement with previous reports on pri-mary porosity of limestones and dolostones (Pulido-Bosch et al., 2004). These findings suggest that plants growing on different rock substrates have potential access to different amounts of avail-able water. At both B and D sites, AWCsoilwas five to 15 times higher than AWCrock. Nonetheless, the relative volumes occupied by soil and bedrock (Fig. 2) potentially explored by roots (Nar-dini et al., 2016) point to an important role of bedrock as a criti-cal water source for plants. GPR scans showed that soil mixed

with rock fragments occupied only about 70–90 cm of the pro-file, while deeper portions were dominated by massive bedrock. This conclusion was consistent with anecdotal data based on pit excavation in the study sites, as well as on observation of road cuts in the study area. Despite similar AWCsoilvalues at B and D sites, plants growing at B sites had c. 23% larger water volumes available in the first 5 m of soil/bedrock depth, because of higher AWCrockB, leading to total estimated below-ground available water of 190 mm. This value is consistent with results by Wang

(a)

(c)

(b)

Fig. 8 Seasonal trends of leaf water content (a, WCleaf), relative water content (b,

RWCleaf), minimum daily leaf water

potential, (c, solid circles,Ψmin), and predawn

water potential (c, open circles,Ψpd) as

recorded in the field in Fraxinus orunus trees growing at sites dominated by breccia bedrock (blue circles) or dolostone bedrock (black circles). Values are means SD. Asterisks denote significant differences between sites. ns, not significant.

(a)

(b) Fig. 9 Seasonal trends of normalized

et al. (1995), showing that woody vegetation growing on shallow soil in southern Oregon apparently used about 200 mm of bedrock water on an annual scale. Similarly, Tokumoto et al. (2014) estimated an AWC of 49–174 mm for a rocky soil in a karst savanna.

Our data suggest that different volumes of available water are stored in the bedrock of B and D sites, but a key question is whether plants can exploit these water pools. Roots can make close contact to the rock matrix (Fig. S1) (Zwieniecki & Newton, 1995), but rock-to-root water transfer is likely to be limited by rock hydraulic conductivity. Different types of carbonate rocks have hydraulic conductivities in the range 10−9– 10−12m s−1 (Boving & Grathwohl, 2001; Pulido-Bosch et al., 2004), that is, several magnitudes lower than values reported for clayey soils (10−5–10−9 m s−1), such as those dominating karstic areas (Kubota et al., 2005; Bhardwaj et al., 2007). Mycorrhizas might improve water transfer to the roots (Nardini et al., 2000; Bornyasz et al., 2005; Lehto & Zwiazek, 2011), but rock hydraulics would probably remain limiting. It is therefore unlikely that rock moisture can sustain to any significant extent the transpiration of woody plants with open stomata. However, slow rock-to-root or rock-to-soil water transfer (Rodrı´guez-Robles et al., 2015; Tetegan et al., 2015) might be sufficient to maintain plant hydration under prolonged drought, when stom-ata close and water loss by transpiration is strongly limited (Duursma et al., 2019). Under these conditions, even minimal root water uptake from the bedrock would help to maintain plant hydration status above critical limits (Sapes et al., 2019), pro-vided that root–rock physical contact is maintained (Carminati et al., 2009). Hence, root access to bedrocks with contrasting porosities might modulate the local risk of leaf desiccation and canopy dieback, contributing to the observed patchiness of drought-induced tree mortality at landscape scales. To test these ideas, we designed experiments on potted plants grown under controlled glasshouse conditions but with different soil/rock mix-tures, and at natural sites occupied by the same vegetation type over different bedrocks.

Plants grown in pots filled with soil or soil/rock showed differ-ent RGR and biomass. Plants growing in soil/dolostone had lower height and diameter when compared to plants growing in soil or soil/breccia that were not significantly different from each other (Fig. 6). This finding is remarkable, as the relative volumes of soil and rocks were the same in B and D treatments. Moreover, irrigation and fertilization were homogeneous across the groups. Even more notable is the fact that the final leaf biomass was 30% lower in D plants than in S and B plants. The variables reason-ably explaining these differences between B and D groups are the different Vpores and AWCrock of breccia and dolostone. Leaf growth and expansion are limited by cell turgor, and more criti-cally by maximum water potential reached at night (Pantin et al., 2011). Despite similar gLandΨminacross the groups, it is possi-ble that lower AWCrockD delayed the night-time recovery of water potential, leading to lower nocturnal turgor pressure and slowing down leaf expansion. Leaf growth is also controlled by chemical root-to-shoot signalling, based on a suite of plant hor-mones overall balancing root and shoot growth to acclimate plant

architecture to variations in soil water supply vs evaporative demand (Schachtman & Goodger, 2008). Finally, different growth patterns in D plants compared with B and S groups might have been caused by subtle effects of rock presence/type on plant nutritional status. It is known that bedrock type can influ-ence vegetation productivity via effects on nutrient availability (Eimil-Fraga et al., 2014). In our experiment, plants were regu-larly fertilized, so that nutritional deficiencies were unlikely. Tox-icity as a result of an excess of Mg, or Mg/Ca imbalance caused by the presence of dolostone (Mota et al., 2017), cannot be ruled out as a factor contributing to reduced growth of D plants, but this is also unlikely based on similar biomass and productivity observed at natural sites with B vs D bedrock (see later). Interest-ingly, D plants reduced leaf but not root or stem biomass, leading to an adjustment of root-to-shoot ratio, a typical response of plants to limited soil water availability (Li et al., 2005). It is possi-ble that plants growing in soil/dolostone sensed low AWCrockD, leading to chemical signalling that modulated root/leaf biomass partitioning.

Although the drydown experiment on potted plants growing in soil or soil/rocks had obvious shortcomings as a result of the fact that drought stress developed much faster compared with a natural situation, the related results offered interesting insights into the possible role of rock water in alleviating water stress pro-gression under drought. All plant groups had similar WCleaf, gL and Ψmin, both under well-watered conditions and during pro-gression of drought. However, average values of Ψmin across groups failed to capture the different numbers of plants reaching critical water potential inducing leaf desiccation (Fig. 5). Com-plete stomatal closure occurred in all groups at 4–5 d after sus-pending irrigation, when Ψmin approached −3 MPa, corresponding to turgor loss point (Nardini et al., 2003) and to the onset of critical xylem embolism levels in F. ornus (Petruzzel-lis et al., 2019). Starting from day 7, the number of D plants experiencing Ψmin< −5 MPa and undergoing leaf desiccation started to increase, reaching 50% at the end of drought, com-pared with only c. 30% of S and B plants. This is consistent with the idea that B plants had access to a larger residual water reser-voir in the rock matrix compared with D plants. This small amount of water was probably sufficient to compensate minimal water losses by cuticular transpiration, thus allowing B plants to maintain a better water status. In D plants, the rock matrix prob-ably could not release a sufficient amount of water to maintain leaf hydration, despite lower leaf biomass. Therefore,Ψminof a large number of D plants dropped below values inducing catas-trophic hydraulic failure and desiccation.

and exposure, such differences were reasonably a result of differ-ent below-ground water availability, as suggested by differences inΨpd. Because soil depth and AWCsoilwere similar at the B and D sites, we suggest that differences in plant water status arose by the contrasting values of AWCrockof breccia and dolostone.

The NDVI values, a proxy of green biomass, were similar at the two sites in spring and early summer, suggesting that different plant water status recorded during the summer drought was not a result of different amounts of foliage, leading to different water consumption rates. The finding that B and D sites had similar biomass and productivity, despite likely differences in water avail-ability related to bedrock type, is somehow counterintuitive (but see Hahm et al., 2019). It is possible that vegetation productivity is more related to the water available in soil/stones layers that can indeed efficiently supply water to transpiring plants under active photosynthesis. On the other hand, drought survival under lim-ited transpiration might depend on slow release of water from the bedrock (Tetegan et al., 2015). Indeed, at the peak of drought, NDVI decreased more at D than at B, leading to the appearance of significant differences between the sites. This was consistent with the observed onset of partial canopy desiccation at the D site. Despite similar NDVI, values of NDWI were consistently different between sites, suggesting that plants growing over dolo-stone faced a limitation in water supply compared with plants growing on breccia (Marusig et al., 2020).

Our data show that the rock matrix can represent a significant water source in rock-dominated habitats with shallow and poorly developed soils overlying massive bedrock. Most importantly, we showed that below-ground landscape heterogeneity, in terms of bedrock primary porosity and available water content, can trans-late into significant spatial variability of plant water status and risk of hydraulic failure under drought. We thus suggest that below-ground water relations might be an important driver of inter- and intraspecific variability of response to extreme drought events, especially in rock-dominated landscapes (Dawson et al., 2020). Below-ground components of the plant water envelope, with specific reference to rock physical features, should be included in models aimed at projecting tree mortality risk under climate change scenarios, to capture the spatial variability of pro-gressive deterioration of vegetation water status during intense or prolonged drought (Mackay et al., 2019). We call for more stud-ies on rock-to-root water transfer, aimed at quantifying: the hydraulic conductivity of this critical interface; its sufficiency in providing water to plants under different evaporative demand; the possible species-specific difference in root–rock water rela-tions; and how the root–rock hydraulic continuity changes dur-ing soil–rock water depletion.

Acknowledgements

We are very grateful to the ‘Direzione centrale risorse agroali-mentari, forestali e ittiche – area foreste e territorio’ of the ‘Regione Autonoma Friuli Venezia Giulia’, and to the public nursery Vivai Pascul (Tarcento, Italy) for providing the plant material for the glasshouse experiment. This research was par-tially supported by the project ‘Individuazione e caratterizzazione

delle aree carsiche del Friuli Venezia Giulia’ (Regione Autonoma FVG and University of Trieste, DMG).

Author contributions

AN, FP, LZ and MT conceived the study and designed the exper-iments. DM, FP, MT, SN and AN performed measurements of plant water status and biomass. DM, FP and GF performed mea-surements of rock and soil porosity and water status. CC, FC and LZ performed geological surveys. EF and GF performed georadar analysis. DM and AA performed analysis of remote sensing data. FP, DM and AN analysed the data. AN and FP wrote the manuscript, with contributions from all co-authors.

ORCID

Alfredo Altobelli https://orcid.org/0000-0002-4038-7014 Chiara Calligaris https://orcid.org/0000-0003-3164-5197 Emanuele Forte https://orcid.org/0000-0002-5995-9254 Daniel Marusig https://orcid.org/0000-0002-2460-6248 Andrea Nardini https://orcid.org/0000-0002-5208-0087 Francesco Petruzzellis https://orcid.org/0000-0002-3635-8501

Martina Tomasella https://orcid.org/0000-0002-1470-1030 Luca Zini https://orcid.org/0000-0001-6564-1683

References

Adams HD, Barron-Gafford GA, Minor RL, Gardea AA, Bentley LP, Law DJ, Breshears DD, McDowell NG, Huxman TE. 2017. Temperature response surfaces for mortality risk of tree species with future drought. Environmental Research Letters 12: 115014.

Allen CD, Macalady AK, Chenchouni H, Bachelet D, McDowell N, Vennetier M, Kitzberger T, Rigling A, Breshears DD, Hogg EHet al. 2010. A global overview of drought and heat-induced tree mortality reveals emerging climate change risks for forests. Forest Ecology and Management 259: 660–684. Allen MF. 2007. Mycorrhizal fungi: highways for water and nutrients in arid

soils. Vadose Zone Journal 6: 291–297.

Anderegg WRL, Kane JM, Anderegg LDL. 2013. Consequences of widespread tree mortality triggered by drought and temperature stress. Nature Climate Change 3: 30–36.

Baguskas SA, Peterson SH, Bookhagen B, Still CJ. 2014. Evaluating spatial patterns of drought-induced tree mortality in a coastal California pine forest. Forest Ecology and Management 315: 43–53.

Barigah TS, Charrier O, Douris M, Bonhomme M, Herbette S, Am´eglio T, Fichot R, Brignolas F, Cochard H. 2013. Water stress-induced xylem hydraulic failure is a causal factor of tree mortality in beech and poplar. Annals of Botany 112: 1431–1437.

Bhardwaj AK, Goldstein D, Azenkot A, Levy GJ. 2007. Irrigation with treated wastewater under two different irrigation methods: effects on hydraulic conductivity of a clay soil. Geoderma 140: 199–206.

Bornyasz MA, Graham RC, Allen MF. 2005. Ectomycorrhizae in a soil-weathered granitic bedrock regolith: linking matrix resources to plants. Geoderma 126: 141–160.

Boving TB, Grathwohl P. 2001. Tracer diffusion coefficients in sedimentary rocks: correlation to porosity and hydraulic conductivity. Journal of Contaminant Hydrology 53: 85–100.

Calligaris C, Mezga K, Slejko FF, Urbanc J, Zini L. 2018. Groundwater characterization by means of conservative (δ18

O andδ2H) and non-conservative (87Sr/86Sr) isotopic values: the Classical Karst Region aquifer case (Italy–Slovenia). Geosciences 8: 321.

Carminati A, Vetterlein D, Weller U, Vogel HJ, Oswald SE. 2009. When roots lose contact. Vadose Zone Journal 8: 805–809.

Carnicer J, Coll M, Ninyerola M, Pons X, S´anchez G, Pe˜nuelas J. 2011. Widespread crown condition decline, food web disruption, and amplified tree mortality with increased climate change-type drought. Proceedings of the National Academy of Sciences, USA 108: 1474–1478.

Carri`ere SD, Ruffault J, Pimont F, Doussan C, Simioni G, Chalikakis K, Limousin JM, Scotti I, Courdier F, Cakpo CBet al. 2020. Impact of local soil and subsoil conditions on inter-individual variations in tree responses to drought: insights from Electrical Resistivity Tomography. Science of the Total Environment 698: 134247.

Chitra-Tarak R, Ruiz L, Dattaraja HS, Kumar MSM, Riotte J, Suresh HS, McMahon SM, Sukumar R. 2018. The roots of the drought: hydrology and water uptake strategies mediate forest-wide demographic response to precipitation. Journal of Ecology 106: 1495–1507.

Choat B, Brodribb TJ, Brodersen CR, Duursma RA, L´opez R, Medlyn BE. 2018. Triggers of tree mortality under drought. Nature 558: 531–539. Colizza E, Cucchi F, Ulcigrai F. 1989. Caratteristiche geolitologiche e strutturali

del Membro di Rupingrande della Formazione dei Calcari del Carso Triestino. Bollettino della Societ`a Adriatica di Scienze 71: 29–46.

Dawson TE, Hahm WJ, Crutchfield-Peters K. 2020. Digging deeper: what the critical zone perspective adds to the study of plant ecophysiology. New Phytologist 226: 666–671.

De Toledo JJ, Magnusson WE, Castilho CV, Nascimento HEM. 2011. How much variation in tree mortality is predicted by soil and topography in Central Amazonia? Forest Ecology and Management 262: 331–338.

Dorman M, Svoray T, Perevolotsky A, Moshe Y, Sarris D. 2015. What determines tree mortality in dry environments? A multi-perspective approach. Ecological Applications 25: 1054–1071.

Duursma RA, Blackman CJ, Lop´ez R, Martin-StPaul NK, Cochard H, Medlyn BE. 2019. On the minimum leaf conductance: its role in models of plant water use, and ecological and environmental controls. New Phytologist221: 693–705. Eimil-Fraga C, Rodrı´guez-Soalleiro R, S´anchez-Rodrı´guez F, P´erez-Cruzado C,

´Alvarez-Rodrı´guez E. 2014. Significance of bedrock as a site factor

determining nutritional status and growth of maritime pine. Forest Ecology and Management 331: 19–24.

Elliott KJ, Swank WT. 1994. Impacts of drought on tree mortality and growth in a mixed hardwood forest. Journal of Vegetation Science 5: 229–236.

Estrada-Medina H, Santiago LS, Graham RC, Allen MF, Jim´enez-Osornio JJ. 2013. Source water, phenology and growth of two tropical dry forest tree species growing on shallow karst soils. Trees 27: 1297–1307.

Fensham RJ, Fairfax RJ. 1997. Drought-related tree death of savanna eucalypts: species susceptibility, soil conditions and root architecture. Journal of Vegetation Science 18: 71–80.

Geekiyanage N, Goodale UM, Cao K, Kitajima K. 2019. Plant ecology of tropical and subtropical karst ecosystems. Biotropica 51: 626–640. Gortan E, Nardini A, Gasc´o A, Salleo S. 2009. The hydraulic conductance of

Fraxinus ornus leaves is constrained by soil water availability and coordinated with gas exchange rates. Tree Physiology 29: 529–539.

Hahm WJ, Dralle DN, Rempe DM, Bryk AB, Thompson SE, Dawson TE, Dietrich WE. 2019. Low subsurface water storage capacity relative to annual rainfall decouples Mediterranean plant productivity and water use from rainfall variability. Geophysical Research Letters 46: 6544–6553.

Hammond WM, Yu K, Wilson LA, Will RE, Anderegg WRL, Adams HD. 2019. Dead or dying? Quantifying the point of no return from hydraulic failure in drought-induced tree mortality. New Phytologist 223: 1834–1843. Hanewinkel M, Cullmann DA, Schelhaas MJ, Nabuurs GJ, Zimmermann NE.

2013. Climate change may cause severe loss in the economic value of European forest land. Nature Climate Change 3: 203–207.

Hubbert KR, Graham RC, Anderson MA. 2001. Soil and weathered bedrock: components of a Jeffrey pine plantation substrate.Soil Science Society of America Journal 65: 1255–1262.

Hughes SW. 2005. Archimedes revisited: a faster, better, cheaper method of accurately measuring the volume of small objects. Physics Education 40: 468–474.

Jayawickreme DH, Jobb´agy EG, Jackson RB. 2014. Geophysical subsurface imaging for ecological applications. New Phytologist 201: 1170–1175. Jol HM. 2009. Ground penetrating radar: theory and applications. UK: Elsevier. Jones DP, Graham RC. 1993. Water-holding characteristics of weathered

granitic rock in chaparral and forest ecosystems. Soil Science Society of America Journal 57: 256–261.

Jurkovˇsek B, Biolchi S, Furlani S, Kolar-Jurkovˇsek T, Zini L, Jeˇz J, Tunis G, Bavec M, Cucchi F. 2016. Geology of the Classical Karst Region (SW Slovenia–NE Italy). Journal of Maps 12: 352–362.

Kiorapostolou N, Da Sois L, Petruzzellis F, Savi T, Trifil`o P, Nardini A, Petit G. 2019. Vulnerability to xylem embolism correlates to wood parenchyma fraction in angiosperms but not in gymnosperms. Tree Physiology 39: 1675–1684.

Klein T, Zeppel MJB, Anderegg WRL, Bloemen J, De Kauwe MG, Hudson P, Ruehr NK, Powell TL, von Arx G, Nardini A. 2018. Xylem embolism refilling and resilience against drought-induced mortality in woody plants: processes and trade-offs. Ecological Research 33: 839–855.

Kubota A, Bordon J, Hoshiba K, Horita T, Ogawa K. 2005. Change in physical properties of “Terra Rossa” soils in Paraguay under no-tillage. Soil Science Society of America Journal 69: 1448–1454.

Kukowski K, Schwinning S, Schwartz B. 2013. Hydraulic responses to extreme drought conditions in three co-dominant tree species in shallow soil over bedrock. Oecologia 171: 819–830.

Lehto T, Zwiazek JJ. 2011. Ectomycorrhizas and water relations of trees: a review. Mycorrhiza 21: 71–90.

Li X, Blackman CJ, Rymer PD, Quintans D, Duursma RA, Choat B, Medlyn BE, Tissue DT. 2018. Xylem embolism measured retrospectively is linked to canopy dieback in natural populations of Eucalyptus piperita following drought. Tree Physiology 38: 1193–1199.

Li Y, Xu H, Cohen S. 2005. Long-term hydraulic acclimation to soil texture and radiation load in cotton. Plant, Cell & Environment 28: 492–499.

Liu W, Li P, Duan W, Liu W. 2014. Dry-season water utilization by trees growing on thin karst soils in a seasonal tropical rainforest of Xishuangbanna, Southwest China. Ecohydrology 7: 927–935.

Lopez OR, Kursar TA, Cochard H, Tyree MT. 2005. Interspecific variation in xylem vulnerability to cavitation among tropical tree and shrub species. Tree Physiology 25: 1553–1562.

MacGregor SD, O’Connor TG. 2002. Patch dieback of Colophospermum mopane in a dysfunctional semi-arid African savannah. Austral Ecology 27: 385–395. Mackay DS, Savoy PR, Grossiord C, Tai X, Pleban JR, Wang DR, McDowell

NG, Adams HD, Sperry JS. 2019. Conifers depend on established roots during drought: results from a coupled model of carbon allocation and hydraulics. New Phytologist 225: 679–692.

Marusig D, Petruzzellis F, Tomasella M, Napolitano R, Altobelli A, Nardini A. 2020. Correlation of field-measured and remotely sensed plant water status as a tool to monitor the risk of drought-induced forest decline. Forests 11: 77. Matzner SL, Rice KJ, Richards JH. 2001. Intra-specific variation in xylem

cavitation in interior live oak (Quercus wislizenii A. DC.). Journal of Experimental Botany 52: 783–789.

McDowell NG, Beerling DJ, Breshears DD, Fisher RA, Raffa KF, Stitt M. 2011. The interdependence of mechanisms underlying climate-driven vegetation mortality. Trends in Ecology and Evolution 26: 523–532.

McDowell NG, Fisher RA, Xu C, Domec JC, H¨oltt¨a T, Mackay DS, Sperry JS, Boutz A, Dickman L, Gehres Net al. 2013. Evaluating theories of drought-induced vegetation mortality using a multimodel-experiment framework. New Phytologist 200: 304–321.

Menezes-Silva PE, Loram-Lourenc¸o L, Ferreira RD, Alves B, Ferreira Sousa L, da Silva Almeida SE, Santos Farnese F. 2019. Different ways to die in a changing world: consequences of climate change for tree species performance and survival through an ecophysiological perspective. Ecology and Evolution 9: 11979–11999.

In: Loidi J, ed. The vegetation of the Iberian Peninsula. Plant and vegetation, vol. 13. Cham, Switzerland: Springer, 277–354.

Nardini A, Battistuzzo M, Savi T. 2013. Shoot desiccation and hydraulic failure in temperate woody angiosperms during an extreme summer drought. New Phytologist 200: 322–329.

Nardini A, Casolo V, Dal Borgo A, Savi T, Stenni B, Bertoncin P, Zini L, McDowell NG. 2016. Rooting depth, water relations and non-structural carbohydrate dynamics in three woody angiosperms differentially affected by an extreme summer drought. Plant, Cell & Environment 39: 618–627.

Nardini A, Salleo S, Trifil`o P, Lo Gullo MA. 2003. Water relations and hydraulic characteristics of three woody species co-occurring in the same habitat. Annals of Forest Science 60: 297–305.

Nardini A, Salleo S, Tyree MT, Vertovec M. 2000. Influence of the ectomycorrhizas formed by Tuber melanosporum Vitt. on hydraulic conductance and water relations of Quercus ilex L. seedlings. Annals of Forest Science 57: 305–312.

Nie Y, Chen H, Wang K, Ding Y. 2014. Rooting characteristics of two widely distributed woody plant species growing in different karst habitats of southwest China. Plant Ecology 215: 1099–1109.

Nie Y, Chen H, Wang K, Yang J. 2012. Water source utilization by woody plants growing on dolomite outcrops and nearby soils during dry seasons in karst region of Southwest China. Journal of Hydrology 420: 264–274. Padilla FM, Pugnaire FI. 2007. Rooting depth and soil moisture control

Mediterranean woody seedling survival during drought. Functional Ecology 21: 489–495.

Pagliarani C, Casolo V, Beiragi MA, Cavalletto S, Siciliano I, Schubert A, Gullino ML, Zwieniecki MA, Secchi F. 2019. Priming xylem for stress recovery depends on coordinated activity of sugar metabolic pathways and changes in xylem sap pH. Plant, Cell & Environment 42: 1775–1787. Pantin F, Simonneau T, Rolland G, Dauzat M, Muller B. 2011. Control of leaf

expansion: a developmental switch from metabolics to hydraulics. Plant Physiology 156: 803–815.

Petit G, Savi T, Consolini M, Anfodillo T, Nardini A. 2016. Interplay of growth rate and xylem plasticity for optimal coordination of carbon and hydraulic economies in Fraxinus ornus trees. Tree Physiology36: 1310–1319.

Petrucco L, Nardini A, von Arx G, Saurer M, Cherubini P. 2017. Isotope signals and anatomical features in tree rings suggest a role for hydraulic strategies in diffuse drought-induced die-back of Pinus nigra. Tree Physiology 37: 523–535. Petruzzellis F, Nardini A, Savi T, Tonet V, Castello M, Bacaro G. 2019. Less

safety for more efficiency: water relations and hydraulics of the invasive tree Ailanthus altissima (Mill.) Swingle compared with native Fraxinus ornus L. Tree Physiology 39: 76–87.

Pierret A, Maeght JL, Cl´ement C, Montoroi JP, Hartmann C, Gonkhamdee S. 2018. Understanding deep roots and their functions in ecosystems: an advocacy for more unconventional research. Annals of Botany 118: 621–635.

Preisler Y, Tatarinov F, Gr¨unzweig JM, Bert D, Og´ee J, Wingate L, Rotenberg E, Rohatyn S, Her N, Moshe Iet al. 2019. Mortality versus survival in drought-affected Aleppo pine forest depends on the extent of rock cover and soil stoniness. Functional Ecology 33: 901–912.

Pulido-Bosch A, Motyka J, Pulido-Leboeuf P, Borczak S. 2004. Matrix hydrodynamic properties of carbonate rocks from the Betic Cordillera (Spain). Hydrological Processes 18: 2893–2906.

Querejeta JI, Estrada-Medina H, Allen MF, Jim´enez-Osornio JJ. 2007. Water source partitioning among trees growing on shallow karst soils in a seasonally dry tropical climate. Oecologia 152: 26–36.

Querejeta JI, Estrada-Medina H, Allen MF, Jim´enez-Osornio JJ, Ruenes R. 2006. Utilization of bedrock water by Brosimum alicastrum trees growing on shallow soil atop limestone in a dry tropical climate. Plant and Soil 287: 187–197. Reed WP, Kaye MW. 2020. Bedrock type drives forest carbon storage and

uptake across the mid-Atlantic Appalachian Ridge and Valley, U.S.A. Forest Ecology and Management 460: 117881.

Rempe DM, Dietrich WE. 2018. Direct observations of rock moisture, a hidden component of the hydrologic cycle. Proceedings of the National Academy of Sciences, USA 115: 2664–2669.

Renne RR, Schlaepfer DR, Palmquist KA, Bradford JB, Burke IC, Lauenroth WK. 2019. Soil and stand structure explain shrub mortality patterns following global change-type drought and extreme precipitation. Ecology 100: e02889.

Rice KJ, Matzner SL, Byer W, Brown JR. 2004. Patterns of tree dieback in Queensland, Australia: the importance of drought stress and the role of resistance to cavitation. Oecologia 139: 190–198.

Rodrı´guez-Robles U, Arredondo JT, Huber-Sannwald E, Vargas R. 2015. Geoecohydrological mechanisms couple soil and leaf water dynamics and facilitate species coexistence in shallow soils of a tropical semiarid mixed forest. New Phytologist 207: 59–69.

Rose KL, Graham RC, Parker DR. 2003. Water source utilization by Pinus jeffreyi and Arctostaphylos patula on thin soils over bedrock. Oecologia 134: 46–54.

Rowland L, da Costa ACL, Galbraith DR, Oliveira RS, Binks OJ, Oliveira AAR, Pullen AM, Doughty CE, Metcalfe DB, Vasconcelos SSet al. 2015. Death from drought in tropical forests is triggered by hydraulics not carbon starvation. Nature 528: 119–122.

Sapes G, Roskilly B, Dobrowski S, Maneta M, Anderegg WRL, Martinez-Vilalta J, Sala A. 2019. Plant water content integrates hydraulics and carbon depletion to predict drought-induced seedling mortality. Tree Physiology 39: 1300–1312.

Savi T, Casolo V, Dal Borgo A, Rosner S, Torboli V, Stenni B, Bertoncin P, Martellos S, Pallavicini A, Nardini A. 2019a. Drought-induced dieback of Pinus nigra: a tale of hydraulic failure and carbon starvation. Conservation. Physiology 7: coz012.

Savi T, Casolo V, Luglio J, Bertuzzi S, Trifil`o P, Lo Gullo MA, Nardini A. 2016. Species-specific reversal of stem xylem embolism after a prolonged drought correlates to endpoint concentration of soluble sugars. Plant Physiology and Biochemistry 106: 198–207.

Savi T, Petruzzellis F, Martellos S, Stenni B, Dal Borgo A, Zini L, Lisjak K, Nardini A. 2018. Vineyard water relations in a karstic area: deep roots and irrigation management. Agriculture, Ecosystems and Environment 263: 53–59. Savi T, Petruzzellis F, Moretti E, Stenni B, Zini L, Martellos S, Lisjak K,

Nardini A. 2019b. Grapevine water relations and rooting depth in karstic soils. Science of the Total Environment 692: 669–675.

Schachtman DP, Goodger JQD. 2008. Chemical root to shoot signalling under drought. Trends in Plant Science 13: 281–287.

Schwantes AM, Parolari AJ, Swenson JJ, Johnson DM, Domec JC, Jackson RB, Pelak N, Porporato A. 2018. Accounting for landscape heterogeneity improves spatial predictions of tree vulnerability to drought. New Phytologist 220: 132–146.

Schwinning S. 2010. The ecohydrology of roots in rocks. Ecohydrology 3: 238–245.

Sevanto S, McDowell NG, Dickman LT, Pangle R, Pockman WT. 2014. How do trees die? A test of the hydraulic failure and carbon starvation hypotheses. Plant, Cell & Environment 37: 153–161.

Shangguan W, Hengl T, Mendes de Jesus J, Yuan H, Dai Y. 2017. Mapping the global depth to bedrock for land surface modelling. Journal of Advances in Modeling Earth Systems 9: 65–88.

Skelton RP, Anderegg LDL, Papper P, Reich E, Dawson TE. 2019. No local adaptation in leaf or stem xylem vulnerability to embolism, but consistent vulnerability segmentation in a North American oak. New Phytologist 223: 1296–1306.

Sternberg PD, Anderson MA, Graham RC, Beyers JL, Tice KR. 1996. Root distribution and seasonal water status in weathered granitic bedrock under chaparral. Geoderma 72: 89–98.

Sucre EB, Tuttle JW, Fox TR. 2011. The use of Ground-Penetrating Radar to accurately estimate soil depth in rocky forest soils. Forest Science 57: 59–66. Swaffer BA, Holland KL, Doody TM, Li C, Hutson J. 2013. Water use

strategies of two co-occurring tree species in a semi-arid karst environment. Hydrological Processes 28: 2003–2017.

Tai X, Mackay DS, Anderegg WRL, Sperry JS, Brooks PD. 2017. Plant hydraulics improves and topography mediates prediction of aspen mortality in southwestern USA. New Phytologist 213: 113–127.

Tetegan M, Korboulewsky N, Bouthier A, Samou¨elian A, Cousin I. 2015. The role of pebbles in the water dynamics of a stony soil cultivated with young poplars. Plant and Soil 391: 307–320.

Tomasella M, Casolo V, Aichner N, Petruzzellis F, Savi T, Trifil`o P, Nardini A. 2019. Non-structural carbohydrate and hydraulic dynamics during drought and recovery in Fraxinus ornus and Ostrya carpinifolia saplings. Plant Physiology and Biochemistry 145: 1–9.

Tomasella M, Petrussa E, Petruzzellis F, Nardini A, Casolo V. 2020. The possible role of non-structural carbohydrates in the regulation of tree hydraulics. International Journal of Molecular Sciences 21: 144.

Urbanc J, Mezga K, Zini L. 2012. An assessment of capacity of Brestovica– Klariˇci Karst water supply (Slovenia). Acta Carsologica 41: 89–100. Wang ZQ, Newton M, Tappeiner JC. 1995. Competitive relations between

Douglas-fir and Pacific madrone on shallow soils in a Mediterranean climate. Forest Science 41: 744–757.

Zhao W, Forte E, Colucci RR, Pipan M. 2016. High-resolution glacier imaging and characterization by means of GPR attribute analysis. Geophysical Journal International 206: 1366–1374.

Zhao W, Forte E, Fontolan G, Pipan M. 2018. Advanced GPR imaging of sedimentary features: integrated attribute analysis applied to sand dunes. Geophysical Journal International 213: 147–156.

Zini L, Calligaris C, Cucchi F. 2015. The challenge of tunnelling through Mediterranean karst aquifers: the case study of Trieste (Italy). Environmental Earth Sciences 74: 281–295.

Zwieniecki MA, Newton M. 1995. Roots growing in rock fissures: their morphological adaptation. Plant and Soil 172: 181–187.

Supporting Information

Additional Supporting Information may be found online in the Supporting Information section at the end of the article.

Fig. S1 Photographs of roots in karst cracks and caves.

Fig. S2 Temperature and precipitation in the study site during the experimental period.

Fig. S3 Photographs of trees and bedrock in the study sites. Fig. S4 Potting procedure for the glasshouse experiment. Fig. S5 Close-up view of ground-penetrating radar profiles in the study sites.

Fig. S6 Canopy dieback observed at the end of summer at the site on dolostone bedrock.

Table S1 Output of statistical analysis of soil moisture release curves.

Please note: Wiley Blackwell are not responsible for the content or functionality of any Supporting Information supplied by the authors. Any queries (other than missing material) should be directed to the New Phytologist Central Office.

New Phytologist is an electronic (online-only) journal owned by the New Phytologist Foundation, a not-for-profit organization dedicated to the promotion of plant science, facilitating projects from symposia to free access for our Tansley reviews and Tansley insights.

Regular papers, Letters, Research reviews, Rapid reports and both Modelling/Theory and Methods papers are encouraged. We are committed to rapid processing, from online submission through to publication ‘as ready’ via Early View – our average time to decision is <26 days. There are no page or colour charges and a PDF version will be provided for each article.

The journal is available online at Wiley Online Library. Visit www.newphytologist.com to search the articles and register for table of contents email alerts.

If you have any questions, do get in touch with Central Office (np-centraloffice@lancaster.ac.uk) or, if it is more convenient, our USA Office (np-usaoffice@lancaster.ac.uk)