Rooting depth and soil moisture control Mediterranean woody seedling survival during drought


†Author to whom correspondence should be addressed. E-mail:


  • 1Seedling survival is one of the most critical stages in a plant's life history, and is often reduced by drought and soil desiccation. It has been hypothesized that root systems accessing moist soil layers are critical for establishment, but very little is known about seedling root growth and traits in the field.
  • 2We related seedling mortality to the presence of deep roots in a field experiment in which we monitored soil moisture, root growth and seedling survival in five Mediterranean woody species from the beginning of the growing season until the end of the drought season.
  • 3We found strong positive relationships between survival and maximum rooting depth, as well as between survival and soil moisture. Species with roots in moist soil layers withstood prolonged drought better, whereas species with shallow roots died more frequently. In contrast, biomass allocation to roots was not related to establishment success.
  • 4Access to moist soil horizons accounted for species-specific survival rates, whereas large root : shoot (R:S) ratios did not. The existence of soil moisture thresholds that control establishment provides insights into plant population dynamics in dry environments.


Seedling recruitment is a critical stage in plant life history because high mortality rates are often associated with the seedling phase (Fenner 1987). Seedlings of different species die from a wide variety of causes (Fenner & Kitajima 1999; Moles & Westoby 2004), including many biotic and abiotic factors such as pathogens, herbivory, high or low temperatures and radiation, allelopathy and competition. Drought and soil desiccation are primary limits to establishment in many environments (Moles & Westoby 2004). One such environment is Mediterranean-type ecosystems, where establishment after germination is severely limited by long, dry summer periods (Herrera 1992). In such areas, seedlings are very drought-sensitive and recruitment processes are often restricted to sporadic rainfall periods (Holmgren & Scheffer 2001; Pugnaire et al. 2006b) or wet microsites (Padilla & Pugnaire 2006).

Deep roots may improve water uptake and increase the probability of survival in Mediterranean communities (Donovan, Mausberg & Ehleringer 1993; Canadell & Zedler 1995; Lloret, Casanovas & Peñuelas 1999) as they can access stable water reserves, allowing plant growth to extend into the dry season (Nepstad et al. 1994). However, differences in seedling survival during drought are often due to varying tolerance to low soil moisture (Hasting, Oechel & Sionit 1989; Ackerly 2004). For instance, Davis (1989) found in the California chaparral that seedlings of drought-tolerant species, often shallow-rooted, survived water shortage better than seedlings of drought-sensitive species.

Although the role of deep roots in plant establishment has long been acknowledged (Davis 1989; Enright & Lamont 1992; Canadell & Zedler 1995; Pugnaire, Chapin & Hardig 2006a), establishment failure due to a lack of deep roots has rarely been quantified. Very little is known about root growth in the field in relation to soil water, and many of the mechanisms controlling establishment success remain poorly understood (Hanley et al. 2004; but cf. Lloret, Casanovas & Peñuelas 1999). This question is particularly significant in dry environments and under global change scenarios with longer drought spells (IPCC 2001). Drier conditions may alter the regeneration niche of many species (sensu Grubb 1977), and species richness may be limited if seedlings are unable to deal with lower water availability (Brown, Valone & Curtin 1997; Schenk & Jackson 2002).

Here we address how drought affects the establishment of five woody species, and relate establishment to rooting depth and access to soil moisture in a field experiment. We hypothesized that deep-rooted seedlings would attain higher survival after summer than shallow-rooted individuals, by keeping roots in moister soil horizons.

Materials and methods


We selected five native perennial woody species occurring in Mediterranean shrublands in semiarid south-east Spain. Two of the species were nearly leafless shrubs with photosynthetic stems, including a gymnosperm, Ephedra fragilis Desf., and a legume, Retama sphaerocarpa (L.) Boiss. The other species included a succulent C4, Salsola oppositifolia Desf., a large C3 shrub, Olea europaea var. sylvestris Brot., and a tree species, Pinus halepensis Mill. Hereafter we refer to these species by their generic names only. Ephedra, Olea and Pinus are frequent in late-successional communities, whereas Retama and Salsola successfully colonize disturbed areas (Peinado, Alcaraz & Martínez-Parras 1992). Our selected species differ in drought-tolerance strategy based on minimum predawn water potential (Ψpd). Salsola tolerates low water potentials (Ψpd≈−5 MPa, Pugnaire, Armas & Valladares 2004), as does Ephedrapd≈−5·2 MPa, F.I.P., unpublished data). Our other species could be considered as non-drought-tolerant based on their less negative Ψpd: ≈−1·5 MPa for the deep-rooted Retama (Haase et al. 1999), −2·5 MPa for Pinus (Oliet et al. 2002), and −2·25 MPa for Olea (Faria et al. 1998). Germination of Olea, Pinus and Retama under Mediterranean conditions begins in winter (Rey & Alcántara 2000; Nathan & Ne’eman 2004; Pugnaire et al. 2006b). There are no accurate data for germination patterns of Ephedra and Salsola, but specific traits suggest that they also germinate in winter, because seeds disperse in late autumn or early winter ( Rodríguez-Pérez, Riera & Traveset 2005) and seeds do not show dormancy (Navarro & Gálvez 2001).

field site and experimental design

We tested our hypothesis by conducting a transplant experiment in a semi-natural field site rather than by monitoring seedling occurrence in nature. This minimized environmental heterogeneity and root losses at harvest, allowed roots to grow without soil impediments, and also allowed comparisons of potential root growth under the same set of abiotic conditions. The experiment was set up in a flat, homogeneous 15 × 15-m terrace for vegetable crops in the foothills of the Sierra Alhamilla range (Almería, Spain, 37°99′ N, 02°99′ W, 600 m elevation). The silt soil had been ploughed regularly for years, was free of rocks, and reached 2 m in depth over a mica–schist bedrock. Fertility and water-holding capacity were very low (Pérez-Pujalte 1989). Neither pesticides nor fertilizers had been applied at the site for at least 5 years. The climate is typically Mediterranean semiarid with a mean annual temperature of 17·3 °C, mean annual precipitation of 282 mm, and a marked drought period from May to September. Temperatures are mild in winter and high in late spring and summer.

In late winter 2004, eight 3 × 3-m plots spaced 1·5 m apart were laid out on the terrace in a 3 × 3 design. To homogenize soil and facilitate root growth, the soil in each plot was completely dug up to a depth of 0·5 m, using an auger (BT120C, Stihl AG & Co. KG, Waiblingen, Germany) to drill adjacent 30-cm-wide holes. Seedlings 1–2 months old were transplanted in early April after heavy spring rainfalls. Seedlings were provided by local nurseries, and seeds were collected in areas with similar ecological conditions. Care was taken to follow the natural recruitment dynamics of all species, and transplanting was done when seedlings of all species had already emerged in the field.

In each plot, 10 bare-root seedlings of each species, similar in size and with intact root systems, were planted 35 cm apart from each other and from the plot borders. The spatial arrangement of different species in each plot was fully random. The terrace was fenced to prevent herbivory, and each plot was watered once with 5 l (0·1 l water per plant) immediately after transplanting. One plot was harvested (H) every 3 weeks on average between April and September, encompassing the spring growing period and the summer drought. Initial data (H0) consisted of 10 randomly harvested seedlings of every species before transplanting. The remaining harvests (H1 to H8) were carried out 13, 28, 48, 66, 81, 97, 121 and 153 days after transplanting. At each harvest, all living individuals in a randomly chosen plot were dug out carefully and maximum root depth was recorded. Root recovery was maximized by digging a 3-m-long trench around the periphery of the plot. Initially the trench was 40 cm wide and 30 cm deep, but the depth of the trench increased in successive harvests until it reached 100 cm on the last sampling date. The front of the trench was gently crumbled from side to side with a hoe, which was replaced by a small punch when close to the base of the plant. Roots were carefully brushed, then manually extracted and stored in paper bags. Soil containing roots that could not be separated in the field were processed in the laboratory. Roots could be matched with individuals because all species had one or several major tap roots, grew vertically, and none spread horizontally. Fine roots attached to major tap roots were also collected. In the laboratory, roots and soil were repeatedly submerged in water and finely sieved to retain fine roots. Shoots and roots were dried at 71 °C for 48 h.

Survival on each harvest date was calculated as the proportion of plants alive after the first week. Soil moisture (ECH2O, Decagon Devices, Inc., Pullman, WA, USA) and temperature (Onset Computers, Pocasset, MA, USA) at depths of 5, 15, 30, 45 and 60 cm were monitored continuously during the experiment in the last plot harvested. Readings were taken every 10 minutes and averaged daily. Soil water content at any given depth within an interval was determined through interpolation between neighbouring readings, assuming that water content in the interval changed linearly. Similarly, we inferred soil depth corresponding to a particular moisture content by interpolating from readings of probes immediately above and below that depth.

growth analysis and statistics

Mean relative growth rate (RGR) for each species during the monitoring period was calculated from observed values between H8 and H0 (Hunt et al. 2002). Growth curves were analysed using ancova on log-transformed observed values with number of days after transplanting as a covariate. Differences among species were considered significant when the species–time interaction was significant. Relative root extension rates (RER) between two consecutive harvests were obtained for each species from fitted polynomial curves (HP curves ver. 3·0, A. Pooley et al.). Differences in biomass and maximum rooting depth among species at H8 were tested using one-way anova followed by Scheffépost hoc comparison tests. Heteroscedastic variables were log-transformed to meet anova assumptions. Differences in seedling survival among species in September were compared through simple binary logistic regression, where survival was the dependent variable and species the predictor factor (Agresti 2002). Regression analyses were performed to test correlation strength between variables, using adjusted R2 to correct for the degrees of freedom. All analyses were conducted with spss ver. 13·0 (SPSS Inc., Chicago, IL, USA) and differences were significant at P < 0·05. Sample size in all analyses was four to 10 for each species, with the exception of Pinus at H8, when only two plants remained alive. Data are presented as means ± 1 SE.


A rainy spring (205 vs 101 mm average in the 1967–97 period, Confederación Hidrográfica del Sur) was followed by a summer with no rainfall (Fig. S1 in Supplementary material). Survival in spring (April–June) was 100% for all species except Ephedra; in this case seedling survival was 87·5% by mid-June and 60% in September. In contrast, Retama and Salsola had complete survival throughout the season. Survival of Olea at the end of the drought period was 80%, and of Pinus 20% (Fig. 1). Moisture decreased quickly in the top soil layers as the drought period progressed, reaching values in September of 1·5 and 11·5% at 5 and 15 cm, respectively; soil moisture remained at 21% from mid-June onwards at 45 and 60 cm. Thus soil moisture in September showed a strong gradient, increasing with depth (Fig. 1).

Figure 1.

Maximum rooting depth (solid bars) and isoclines of soil moisture (shaded areas) on left y-axis and seedling survival (white dots) on secondary right y-axis. Soil layers with moisture >20%, dark grey; 14–20%, mid-grey; 8–14%, pale grey; <8%, white. Values of rooting depth are means ± 1 SE.

Species differed significantly in the maximum depth reached by roots in September (one-way anova, F4,23 = 11·7, P < 0·001; Table 1). The two early colonizers, Salsola and Retama, rooted deepest and also had the highest mean root extension rates (RERmean; Table 1). In contrast, the shallowest roots were found in Ephedra and Pinus, which also had the lowest RERmean. In September, at least one tap root of Olea, Retama and Salsola reached well below 35 cm, whereas roots of Ephedra and Pinus did not surpass 25 cm (Fig. 1). While the R:S ratio in all species was <0·5, it varied significantly among species (one-way anova, F4,23 = 29·3, P < 0·001). Pinus allocated the most to roots, followed by Olea, Retama and Ephedra. In contrast, allocation to roots relative to shoots was rather small in Salsola (R:S < 0·1). Root : shoot ratio did not increase in response to increasing drought but, on the contrary, decreased over the course of the season in Olea, Pinus and Salsola, and remained relatively constant in Ephedra and Salsola (Fig. S2).

Table 1.  Final plant mass, mean relative growth rate (RGR) and mean root extension rate (RERmean), maximum root depth and root : shoot ratio (R:S) of five woody species
  1. Significant differences among species are indicated by F values: ***, P < 0·001. Different letters in a row show significant differences among species at P < 0·05 (one-way anova, Scheffé's test). Values are means ± 1 SE.

Total mass (g)0·28 ± 0·06ac1·57 ± 0·33b0·15 ± 0·02c1·22 ± 0·29ab75·01 ± 17·01d67·96***
RGR (mg g−1 day−1)17·1 ± 3·413·7 ± 0·5 4·4 ± 1·925·5 ± 7·8 48·0 ± 5·1
RERmean (mm cm−1 day−1) 5·3 ± 3·3   5 ± 1·3 1·9 ± 1·2 9·3 ± 3·8 10·8 ± 2·5
Rooting depth (cm)20·6 ± 3·1a35·2 ± 2·3ab15·9 ± 2·7a47·3 ± 6·9b 59·5 ± 6·7b11·65***
R:S ratio0·22 ± 0·03a0·43 ± 0·02a0·49 ± 0·07a0·31 ± 0·04a 0·08 ± 0·01b29·32***

At the end of the drought season there were significant differences in seedling establishment among species (logistic regression, χ2 = 24·2, df = 4, P < 0·001). There were strong positive relationships between seedling survival and maximum rooting depth (logistic function, inline image = 0·99, P < 0·01; Fig. 2a), and between survival and the soil moisture estimated at the maximum rooting depth of the species at final harvest (logistic function, inline image = 0·97, P < 0·02; Fig. 2b), showing that the probability of establishment success was strongly related to increasing rooting depth and therefore soil moisture. Species with roots accessing soil deeper than 45 cm had 100% survival (Salsola and Retama), whereas much lower rates (20–40%) were found in species that rooted in shallower, drier soil layers (22·5 cm, Ephedra and Pinus). Species rooting in layers with <12% soil moisture established poorly (Pinus, Ephedra). In contrast, species with roots reaching soil with moisture >18% (Retama and Salsola) had complete survival. There was, however, no relationship between survival at final harvest and R:S ratio (linear regression, inline image = 0·12, P > 0·38; Fig. 2c). Initial plant size did not correlate with maximum rooting depth at H8 or RGR (linear regression, P > 0·57 and P > 0·3, respectively), or survival at H8 (P > 0·3 for all fitted functions).

Figure 2.

Relationships between survival and maximum rooting depth (a); moisture at the deepest soil layer reached by roots (b); and root : shoot (R:S) ratios in September (c) after the summer drought. Values are means ± 1 SE, with the exception of survival. Ef, Ephedra; Oe, Olea; Ph, Pinus; Rs, Retama; So, Salsola; ns = No significant correlation.

Species also differed in root-growth patterns over spring and summer (ancova species × time, F4,333 = 18·4, P < 0·001; Fig. S3). Roots of Salsola displayed a parabolic growth curve characterized by rather low RER values early in the season, followed by a period of increasing growth rate until the onset of the drought season in mid-June, at which point RER started to decrease. In contrast, Ephedra, Olea and Retama grew at a constant rate from April to September. Ephedra and Olea shared nearly identical RER, whereas Retama exhibited larger values. Growth rate of Pinus decreased from the beginning, showing the highest value in the first harvest and the lowest in the last one.


The effect of summer drought on seedling establishment has long been acknowledged in Mediterranean environments (Herrera 1992) but, to our knowledge, direct links between rooting depth, soil moisture and establishment have never been quantified. The ability to develop deep roots and access soil moisture was decisive for seedlings’ survival of summer drought, regardless of species-specific drought tolerance. Deep-rooted seedlings from either a drought-tolerant species (based on minimum Ψpd reported) such as Salsola or a drought-sensitive species such as Retama had consistent access to moist soil layers and showed the greatest survival rates. In contrast, shallow-rooted seedlings of Ephedra (a drought-tolerant species) and Olea and Pinus (more drought-sensitive species) relied on water from shallower soil layers and died as summer drought progressed.

Climate-change scenarios for the western Mediterranean predict reduced annual precipitation, shifts in seasonal rainfall patterns (decreasing in spring, summer and autumn), and extended drought periods (IPCC 2001). Here we looked at species’ ability to extend their roots quickly enough to keep pace with retreating soil moisture, and showed that deep-rooted seedlings were best able to establish during a very dry growing season, suggesting that these species may be favoured over species of shallow-rooted seedling during extended droughts. Whether shallow-rooted species would decrease in abundance or be confined to more mesic patches or microsites remains unknown (Schenk & Jackson 2002); however, it is worth noting that shifts in regional climates are currently leading to changes in vegetation type dominance, for example in the encroachment of shrubs into grasslands (Brown, Valone & Curtin 1997), probably because new conditions favour the establishment of deep-rooted species (Schenk & Jackson 2002).

The relationship between soil moisture and survival suggests the existence of a threshold of soil moisture that controls plant establishment. In our system, very low establishment rates were achieved by species that kept roots in shallow soil layers with moisture around 12% (e.g. Ephedra and Pinus) and, according to our data, no establishment would occur for plants rooting in layers drier than 8%. On the other hand, higher establishment rates were found for deep-rooted species reaching soil layers wetter than 15% (Retama and Salsola), and full establishment would be related to rooting in soil layers moister than 20%. It is likely that something similar occurs in natural systems, where the threshold will vary depending on soil properties and the species involved. Rainfall-dependent recruitment dynamics reported in dry environments can be interpreted under such operating thresholds. Kitzberger, Steinaker & Veblen (2000) and Holmgren et al. (2006) showed that recruitment in dry years is almost zero, whereas rainy years constitute a window of opportunity for establishment. We suggest that plants in dry habitats may establish easily in wet years without deep roots because soil moisture remains above the critical threshold along the soil profile (Sala & Lauenroth 1982). Conversely, when moisture in the soil profile is below the threshold, it is not enough to maintain seedlings alive. Soil moisture between both extremes would pivot around the critical threshold, and rooting ability and drought tolerance of the different species would explain variation in establishment patterns.

The lack of rain during our experiment produced a vertical soil-moisture gradient, but summer rains could have replenished soils and altered the gradient, probably changing the final outcome. Supplying water during summer drought boosts establishment success in Mediterranean environments (Castro et al. 2005), and summer irrigation considerably increases survival in all our study species (Sánchez et al. 2004). Small rainfalls (like watering) keep soil moisture above certain thresholds and improve survivorship (Sala & Lauenroth 1982).

Plants may adjust to resource imbalance by allocating biomass to organs that acquire the limiting resource (Chapin et al. 1987), so that a higher R:S ratios are expected under water stress. However, we found no relationship between survival and R:S ratio; paradoxically the most successful survivor, Salsola, allocated relatively the least to roots, and the species with most failure, Pinus, had the highest R:S ratio. Overall we did not find substantial changes in R:S during the growth period, in contrast to reports that found shifts in dry-mass partitioning between shoots and roots during plant growth (Klepper 1991). Biomass allocation to roots did not increase in any species in response to drought, suggesting that large R:S ratios may not be enough to compensate for the deepening of moisture along the soil profile in summer. Rather, the ability to alter rates, timing and placement of root proliferation may be more important for plant success than changes in biomass allocation between roots and shoots (Reynolds & D’Antonio 1996). Lloret, Casanovas & Peñuelas (1999), however, reported a large positive correlation between R:S and seedling survival in a Mediterranean shrubland, but in their field site roots rarely reached below 10 cm, and the high summer rainfall in the study years kept shallow soil horizons moist. Under such circumstances, greater biomass allocation to roots may increase water uptake.

Some species from dry environments have dual root systems with shallow lateral roots that exploit small rainfall events that hardly penetrate the soil, and deep roots that tap deep water sources (Canadell & Zedler 1995). However, dual systems often develop as the plant matures, and the absence of lateral branches is frequent in seedlings from xeric habitats (Canadell et al. 1999; Nicotra, Babicka & Westoby 2002). Our observations agree with these patterns, as roots of the five species grew vertically and none spread horizontally. This suggests a primary investment to develop root systems that penetrate into deeper, more reliable water sources, rather than allocating biomass to develop both surface and deep roots, because moisture in the soil surface is unreliable (Ehleringer & Dawson 1992).

In conclusion, our work underlines the importance of rooting depth for seedling survival. In the absence of other constraints on establishment, such as dispersal, seed-germination triggers or herbivory, the ability to reach deeper, moister soil horizons is critical in coping with water stress at such an early stage and becoming established. Our data suggest that species able to keep roots in moist soil layers are better prepared to withstand drought. Also, soil moisture thresholds appear to control plant survival, so plant establishment may be restricted if soil moisture is below certain levels, with direct implications for population and community dynamics.


We are grateful to A. Moreno, J. Padilla, J.D. Miranda, M.J. Jorquera, M.P. Sánchez and Pepe del Cortijo La Sierra for helping with the field work. Serfosur SL, Viveros Retamar and Junta de Andalucía provided seedlings. Comments by Ragan M. Callaway, Heather L. Reynolds, Scott D. Wilson and two anonymous reviewers greatly improved the manuscript. Funds were provided by the Spanish Ministry of Education and Science (grant CGL2004-00090/CLI), and F.M.P. was supported by an I3P fellowship (CSIC-European Social Fund).