Juvenile shrubs show differences in stress tolerance, but no competition or facilitation, along a stress gradient


Dr L. A. Donovan (fax 706 542 1805; donovan@dogwood.botany.uga.edu).


1 We investigated experimentally differences in abiotic stress tolerance and the effects of plant–plant interactions for two desert shrubs, Chrysothamnus nauseosus and Sarcobatus vermiculatus, along a soil salinity (NaCl) and boron (B) gradient at Mono Lake, California, USA. Based on differences in natural distribution, and the classical expectation of a trade-off between competitive ability and stress tolerance, we hypothesized that (i) Chrysothamnus would have greater competitive ability than Sarcobatus at the low salinity end of the gradient, and that (ii) Sarcobatus would be more stress tolerant than Chrysothamnus.

2 Juvenile target plants of Chrysothamnus and Sarcobatus were planted into four sites along the gradient. Biomass was determined by destructive harvests over two growing seasons. At each site, interspecific relative competitive ability was assessed as the effect of Sarcobatus neighbours on Chrysothamnus targets compared to the effect of Chrysothamnus neighbours on Sarcobatus targets. Stress tolerance was assessed as the ability of each species to survive and grow, in the absence of neighbours, at different sites along the gradient.

3 The two species did not differ in the relative strength of plant–plant interactions, providing no support for the expectation that Chrysothamnus had greater competitive ability than Sarcobatus. Furthermore, there was no evidence for competition or facilitation, either interspecific or intraspecific, at any site in either year of the study. However, fertilization treatments demonstrated nutrient limitations, soil water reached limiting levels and root systems of targets and neighbours overlapped substantially. It is therefore surprising that plant–plant interactions among juveniles apparently play little role in the growth and survival of shrubs in this saline desert habitat.

4Sarcobatus was more stress tolerant than Chrysothamnus and the two species performed optimally at different sites along the gradient. Sarcobatus juveniles grew best at the two most saline sites and survived at all sites, whereas Chrysothamnus juveniles grew best at a low-salinity site and did not survive at the most saline site. The difference in site of optimal performance may be due to differences in nutrient limitations or to interactions between nutrient availability and sodium (Na) and B tolerance.


Patterns of plant species zonation have long been described for macro- and micro-environmental gradients in factors such as aridity, altitude, salinity and inundation. The debate on the relative importance of abiotic and biotic determinants of plant distribution and performance along these gradients is often based on the discussion of the general hypothesis that there is a trade-off between stress tolerance and competitive ability of a species (Grime 1977; Wisheu & Keddy 1992; Ungar 1998). The conceptual framework for such a trade-off comes from the expectation that stress-tolerant plants that are adapted to low resource or other harsh abiotic environmental conditions (non-resource factors such as salt or other toxic ions) will have inherently low maximum relative growth rates (RGR) and conservative resource use traits. Low RGR would then be expected to result in reduced competitive ability because of a low capacity to affect neighbours by depleting resources (Grime 1977, 1979; Goldberg 1990; Chapin et al. 1993; Ungar 1998). However, in other theories of competitive success (Tilman 1988; Goldberg 1990, 1997; Grace 1990), low RGR is not necessarily associated with poor competitive ability. In addition, more recent conceptual models emphasize that plant–plant interactions in harsh habitats may tend towards facilitation (positive) instead of competition (negative) (Bertness & Callaway 1994; Callaway 1995; Callaway & Walker 1997).

Empirical studies have documented individual species’ optima and tolerance limits along abiotic stress gradients. In the absence of interspecific interactions, species from different places along a gradient are often found to have the same preferred habitat or physiological optimum, e.g. mangroves and Typha species along inundation gradients (Rabinowitz 1978; Grace & Wetzel 1981), grass species along artificial nutrient gradients (Mahmoud & Grime 1976; Austin & Austin 1980; Aerts et al. 1990), herbaceous species along a freshwater shoreline gradient (Wilson & Keddy 1985), and salt marsh species along salinity and elevation gradients (Snow & Vince 1984; Badger & Ungar 1990; but see Bertness 1991). Although their optima often coincide, such species generally differ in abiotic stress tolerance, as reflected in differential performance at the harsher end of a gradient.

Gradient studies incorporating interspecific interactions have demonstrated varying biotic effects, including competitive suppression, competitive displacement and facilitation (Mahmoud & Grime 1976; Grace & Wetzel 1981; Snow & Vince 1984; Vince & Snow 1984; Gurevitch 1986; Bertness & Ellison 1987; Bertness 1991; Pennings & Callaway 1992). Several gradient studies have indicated a species-level trade-off in competitive ability and stress tolerance: species’ distribution towards the optimal end of the gradient is determined by interspecific competition, whereas abiotic stress tolerance limits distribution towards the harsher end of the gradient (Grace & Wetzel 1981; Vince & Snow 1984; Gurevitch 1986; Rahman & Ungar 1994). This trade-off has also been invoked to explain the vegetation differences that occur across sharp abiotic discontinuities, such as pockets of hydrothermally altered substrates (Goldberg 1985; DeLucia & Schlesinger 1990). However, not all studies have found the expected trade-off (Reader & Bonser 1993; Nernberg & Dale 1997), and stress tolerance and plant–plant interactions along abiotic gradients therefore warrant further investigation.

We determined stress tolerance and plant–plant interactions experimentally for two desert shrub species that differ in their distribution along a topographic salinity gradient at Mono Lake, California, USA. Similar landscape-scale zonation of these shrubs occurs throughout the Great Basin biogeographic region of North America. Natural and anthropogenically induced variation in the level of Mono Lake over the last c. 1000 years has exposed substrates at different times. Differential leaching by precipitation has resulted in a strong gradient of soil salinity (NaCl) and boron (B). Along this gradient, initial colonization of the highly saline substrates is by the shrub Sarcobatus vermiculatus (Hook.) Torrey (Chenopodiaceae; greasewood). Upslope, where soil Na and B decrease, Sarcobatus co-occurs with and thus declines in dominance relative to Chrysothamnus nauseosus (Palla.) Britt. ssp. consimilis (E. Greene) H.M. Hall & Clements (Asteraceae; salt rabbitbrush). These species have similar phenology and life-history characteristics (Roundy et al. 1981; Donovan et al. 1996), and seeds of both species are widely dispersed along the entire gradient (Fort & Richards 1998). Growth chamber and glasshouse studies have shown that both species grow best with low soil Na and B, but that Sarcobatus is a Na-accumulating halophyte that is much more tolerant of both Na and B than Chrysothamnus (Glenn & O’Leary 1984; Richards 1994; Schaber 1994; Dodd & Donovan 1999). Although seedlings of these species have similar RGR under optimal conditions (Dodd & Donovan 1999), Chrysothamnus juveniles have higher RGR under non-saline field conditions (Brown 1997).

Soil resources are limiting in the sandy desert habitat of the Mono Lake site, creating the potential for below-ground resource competition. Soil total N and bicarbonate P are very low, with typical values of 0.02% and 4 p.p.m., respectively (Donovan et al. 1997). The site is also characterized by very alkaline soils (pH > 9.5) and low mean annual precipitation (160 mm; Toft 1995), received mostly in the winter. Soil moisture availability during the growing season is extremely limited in the surface soil layers where roots of seedling and juvenile plants are most prevalent (Dobrowolski et al. 1990; Muller et al. 1995; Toft 1995). Evidence suggesting competition in this habitat, however, is incongruent. Adults of Chrysothamnus and Sarcobatus are rooted down to the groundwater capillary fringe (≤ 7 m depth at the study site), making it unlikely that adults are competing for water. The random spacing of co-occurring Chrysothamnus and Sarcobatus adults suggests that competitive thinning does not occur (Toft 1995). However, seasonal declines in predawn plant water potentials suggest that soil moisture limitations do occur (Donovan & Ehleringer 1994a; Donovan et al. 1996). In addition, 22% of adult Chrysothamnus at the study site died during a 5-year period of groundwater decline (Toft 1995). At another Great Basin site, removal of neighbouring vegetation resulted in greater biomass and flower production for some adult Chrysothamnus, suggesting competition (McKell & Chilcote 1957). Competition may be more likely among juvenile plants that have not yet rooted deep enough to access the stable groundwater source (Fowler 1986; Toft 1995). High mortality has been found for Chrysothamnus and Sarcobatus seedlings and juveniles at this and other sites (K.P. Fort, J.H. Richards & C.A. Toft, unpublished data; Donovan et al. 1993; Donovan & Ehleringer 1994b).

The broad objective of this study was to assess the relative strength of stress tolerance and plant–plant interactions for juveniles of Chrysothamnus and Sarcobatus along the abiotic stress gradient. Although we focused on competition, the experimental design also allowed for detection of facilitation. Based on the classical expectation of a trade-off between stress tolerance and competitive ability, we hypothesized (i) that Chrysothamnus would be a better competitor than Sarcobatus at the lower salinity sites, and (ii) that Sarcobatus would be more stress tolerant than Chrysothamnus. We defined competitive ability as the effect of neighbouring plants on growth and survival of a target, presumably mediated through below-ground resources because canopies did not overlap. For interspecific relative competitive ability we compared the effect of Chrysothamnus neighbours on Sarcobatus targets to the effect of Sarcobatus neighbours on Chrysothamnus targets. This incorporates both response and effect competition (sensuGoldberg 1987, 1990; Goldberg & Scheiner 1993). Stress tolerance was defined as the ability of a species to survive and maintain growth along the soil salinity gradient in the absence of neighbours. The initial size, phenology and spacing of transplants in the experimental gardens were representative of naturally occurring 1-year-old plants in the surrounding populations.

Materials and methods

Study area

This research was conducted on the north shore of Mono Lake, California, USA (38°5′N, 118°58′W 1944–66 m a.s.l.), where previous studies have investigated geomorphology, soils, climate, vegetation, plant nutrition and plant water relations (Schaber 1994; Toft 1995; Donovan et al. 1996, 1997; Fort & Richards 1998). Precipitation was determined using records from Simis and Cain Ranches (2 km and 24 km from the study area, respectively). Where data from Simis Ranch were missing, values were estimated from Cain Ranch data (r2 = 0.66 for mean monthly precipitation; Toft 1995).

Experimental garden sites and treatments

Four experimental garden sites, Diverse Dunes, Sand Flat, Transverse Dune and Shoreline Berm, were located along a permanent, previously reported, transect extending from a mature dune complex toward the lakeshore (Donovan et al. 1997; Fort & Richards 1998). At the study area, lake level changes cause large horizontal displacements of the actual shoreline, necessitating use of the dune end of the gradient as the origin. The sites were located 150, 400, 750 and 1800 m, respectively, along the surveyed north–south transect. Adult Chrysothamnus and Sarcobatus co-occurred at the Diverse Dunes and Sand Flat sites and juveniles of both species were present (Fig. 1). Only Sarcobatus occurred, as adults and juveniles, at the Transverse Dune and Shoreline Berm sites. At 50-m intervals along each of three replicate, parallel, north–south transects, we estimated the percentage of juvenile plants in the population within a 20-m radius of that point. Values were assigned to categories at 10% intervals between 0% and 100%, and means and SE were calculated using the midpoints of these categories.

Figure 1.

Percentage of shrubs that are juveniles (pre-reproductive plants) along the north–south transect at the study site. Adults are the remaining portion of the population at each location. Total plant density (adults plus juveniles) in large mapped populations representative of each experimental garden site are indicated under each garden name as plants per m2 (C. A. Toft, unpublished data). Chrysothamnus occurs along the transect from 0 m to 500 m, with cover ranging from 2.2% to 0%. Sarcobatus occurs from 0 m to 1200 m, with cover ranging from 5.2% to 0.4%, and in an isolated population at 1850 m with cover of 0.05% (C.A. Toft, unpublished data; Donovan et al. 1997).

Each garden site contained 10 replicate plots: five harvested in August 1995 and five harvested in August 1996. Within each plot, treatments were randomly assigned to subplots and coded according to the target species (C = Chrysothamnus and S = Sarcobatus) followed by the treatment: –C, control target with no neighbours; –F, fertilized target with no neighbours; –IH, interspecific high-density neighbours (different target and neighbour species); –IL, interspecific low-density neighbours; –MH, monospecific high-density neighbours (target and neighbours same species); –ML, monospecific low-density neighbours. Each plot at the Diverse Dunes site contained all 12 treatments. At all other sites, the low-density treatments were omitted, and at the Shoreline Berm site the C–MH treatment was replaced by a second S–C subplot. The fertilized plants received approximately 6 g of granular NPK fertilizer (10–10–10) at transplanting in March 1995 and an additional 6 g of slow-release fertilizer (14–14–14; Osmocote™, Scotts Co., Marysville, OH) on both 5 May 1995 and 15 May 1996.

The target was planted in the centre of a 40 × 40-cm subplot. In the high-density treatments (–IH, –MH), four neighbours were planted at the corners of a 15 × 15-cm square, each 11 cm from the target. In the low-density treatments (–IL, –ML), four neighbours were planted at the corners of the 30 × 30-cm square, each 21 cm from the target. Subplots were separated from each other and from the plot edge by a 30-cm buffer zone. A cage of 2.5-cm mesh wire protected each plot from larger herbivores (jackrabbits and mule deer). The sites were located in the natural shrub matrix. Prior to planting at each site, all naturally occurring plants, especially Distichlis spicata (L.) E. Greene, were removed. Regrowth was discouraged with localized herbicide application and weeding.

Soil and plant measurements

Soils were sampled adjacent to the C–MH and S–MH subplots in five plots at each site prior to planting in March 1995 and at final harvest in August 1996. At the Diverse Dunes, Sand Flat and Transverse Dune sites, where soils were sands, the initial sampling for soil chemistry was at 0, 25 and 50 cm. The final sampling for soil chemistry and soil moisture was at 25, 50, 75 and 100 cm. At the Shoreline Berm site, a shallow (30–50-cm depth to groundwater) and highly stratified soil profile necessitated sampling the distinct soil layers of surface crust, buff granular (5–20 cm depth) and organic material (20–34 cm depth) in March 1995. Final samples were taken from the buff granular layer, consisting of pumice particles, and from the organic layer, derived primarily from aquatic insects, brine shrimp and bird feathers deposited by Mono Lake when the berm was formed by a rising lake in 1983–84. Soils were weighed wet and after drying at 100 °C for determination of gravimetric soil moisture. Dried soils were sieved (2-mm screen), and soil electrical conductivity (EC) and pH were determined on 1 : 5 soil : water extracts. Extracts were analysed for Na, P and B using an inductively coupled argon plasma emission spectrometer (ICP; Thermo Jarrell-Ash 965, Franklin, MA). Soil total N was determined by micro-Dumas combustion (NA1500 C/H/N Analyser, Carlo-Erba, Milan, Italy). Soil chemistry data for each year were analysed with anova, with site as the main effect, and depth nested within site, followed by mean separation with Tukey's multiple range tests (SAS 1989).

Juvenile plants were transplanted with attached soil (3 : 1 fritted kaolinitic clay/sand) into the garden sites on 29–31 March 1995. The Chrysothamnus and Sarcobatus plants had been raised from seed collected in autumn 1994 from multiple mothers across the site and germinated in November 1994. Seedlings grew for 5 months in a glasshouse at the University of California, Davis, using supplemental lighting and 1/10 strength Hoagland's nutrient solution, and were then frost hardened outside before transplanting. Each seedling received approximately 1 litre of water at transplant but no further watering occurred during the study. Seedlings were scored for height (longest stem length) and health status (0–5, see legend to Fig. 5) throughout the two growing seasons of the study. Plants were considered dead when stems were brittle and no regrowth occurred on subsequent census dates. G-tests were used to compare end-of-season mortality (Zar 1996).

Figure 5.

Average health code of living plants (targets and neighbours), and percentage survival for each species and site throughout the study period. Health codes: 1 = stem alive but no leaves; 2 = leaves unhealthy and only at shoot tips; 3 = leaves healthy but only at shoot tips; 4 = leaves at shoot tips and along stems, but some leaves unhealthy; 5 = healthy leaves at tips and along stems.


On 10–15 August 1995, we harvested five replicate plots at each site. On 15–29 August 1996, we harvested the remaining five replicate plots at each site. At each harvest, shoots of all target and neighbour plants were severed at the root crown and dried at 60 °C for stem, leaf and flower dry weight. In 1995, the root systems of targets were excavated in entirety for root biomass. In 1996, the root systems of targets were partially excavated for estimation of total biomass. A 10-cm diameter core cylinder was excavated to approximately 15-cm depth around the taproot and all roots were recovered. Beyond and below the core cylinder, a subsample (1/4 except for 1/2 at Shoreline Berm) of the remaining root system was excavated. Roots in the subsample were recovered to 1-m depth and to 50-cm radius, washed free of adhering soil, dried at 60 °C and weighed. For 1996, leaves were ground, ashed, extracted with weak acid solution, and analysed for leaf Na and B using the ICP. For both harvests, additional data on plant water status, nutrient status and shoot and root characteristics (root length, lateral root architecture, mycorrhizal infection) will be in a future publication (L.A. Donovan & J.H. Richards, unpublished data).

Plant–plant interactions

The effects of neighbours on targets were assayed for each garden site in each year (1995 and 1996). The data were analysed using PROC GLM analysis of covariance (ancova) procedures in SAS (SAS Institute Inc. 1989). At each site, the analyses for high-density interspecific effects used S–C, S–IH, C–C and C–IH treatments, and the analyses for intraspecific high-density effects used S–C, S–MH, C–C and C–MH treatments. Because the Diverse Dunes site also contained low-density treatments, low-density interspecific effects were also analysed using the appropriate treatments and controls. For each analysis, the dependent variable was log10 target biomass, and model effects were Plot, Neighbour Biomass (covariate), Target Species and Neighbour Biomass × Target Species. A significant interaction term in these analyses indicates a significant difference in relative competitive ability between the species (modified from Goldberg & Scheiner 1993). Visualizing the data for the interspecific comparisons clarifies the interpretation of the interaction term. When target Sarcobatus and target Chrysothamnus biomass (y-axis) are plotted as a function of interspecific neighbour biomass (x-axis), then a comparison of slopes is the test for species’ relative competitive ability. A non-significant interaction term (no difference between slopes) would indicate no difference in the effects of interspecific neighbours on Chrysothamnus and Sarcobatus targets. A significant interaction term (different slopes) would indicate a difference in competitive (if slopes are negative) or facilitative (if slopes are positive) effects on the target species. Analyses were run using either total biomass (roots, leaves, stems and flowers), shoot biomass (leaves, stems and flowers) or leaf biomass as the dependent variable.

A non-significant interaction term (Neighbour Biomass × Target Species) in the plant–plant interaction ancovas could indicate either plant–plant interactions of equal strength (equivalent slopes) or no plant–plant interactions (no significant slopes). To distinguish between these alternatives, each relationship between target and neighbour biomass (for each target species, high or low density, site, date, and interspecific or intraspecific analysis) was analysed using regression to determine if slopes were significantly different from zero.

Stress tolerance

For each species in each year, target plant total biomass and the ratio of root/total biomass were compared across sites with an anova (model effects were Plot and Site) and Tukey's multiple range comparisons. The analyses included all targets, not just controls with no neighbours, because neither competition nor facilitation effects were detected at any site in either year (see the Results). Initial plant height (at transplant) was considered in analyses of biomass. Initial height tended to differ among sites for Chrysothamnus (F = 2.5, site d.f. = 3, error d.f. = 419, P = 0.057) and differed significantly among sites for Sarcobatus (F = 3.3, site d.f. = 3, error d.f. = 449, P = 0.028). Sarcobatus initial seedling height at Transverse Dune (7.1 ± 0.02) was significantly lower than at Diverse Dunes, Sand Flat and Shoreline Berm (7.5 ± 0.2, 7.8 ± 0.2 and 7.8 ± 0.1, respectively). However, initial height was not a significant covariate when included in the comparisons of biomass across sites. Therefore, results of the simpler analysis, without initial height as a covariate, are presented.

Average health code of living plants and percentage survival were calculated for each site for each sampling date.


Precipitation and soils

The winter preceding the study (January–March 1995) was much wetter than normal, with 17.8 cm precipitation, which was 247% of the long-term (58-year) average for that period (Fig. 2). The first few months after transplanting the shrubs into the field gardens (April–June 1995) were also wetter than normal, with precipitation of 5.2 cm, 232% of the average. Precipitation during the following soil moisture recharge season (October 1995–March 1996) was 10.3 cm, 82% of the average for that period. The second growing season (April–August 1996) was drier than normal (2.3 cm, 59% of the average).

Figure 2.

Precipitation (cm) for Simis Ranch, California, USA, approximately 2 km from study site.

Soil salinity (EC and Na), B and pH were lowest at the Diverse Dunes site, furthest from the lake, and increased significantly to maximum values (pH 10.1; 41–255 p.p.m. B; 12,158–16,134 p.p.m. Na) indicative of severe soil stress at the Shoreline Berm site (Table 1). The saturated paste soil Na concentration at Shoreline Berm was estimated to be c. 250 mm based on the conversion indicated in Table 1. In addition to large differences in soil stress factors among the four sites, there were differences of much lower magnitude in soil nutrients. Soil N and P were higher at the Shoreline Berm than at the other sites (Table 1). Sampling depth, a nested factor within site, was also significant for most soil variables, but the variation accounted for by depth was generally much smaller than that accounted for by site.

Table 1.  Characteristics of experimental garden site soils at study initiation in March 1995 and of soils and plant leaves at final harvest in August 1996. Soil analyses were done on 1 : 5 soil : water extracts and N, P, B and Na are expressed relative to dry weight soil. Saturated extract values can be approximated by multiplying these values by 14.1 (linear r2 = 0.99; Schaber 1994). Soil values (mean ± SE) are averages across depths (see the Methods) for five depth profiles at each garden site. For soils d.f. = 3 and 90–117 for Site and Error, respectively. Where data were log transformed to meet statistical analysis assumptions (1995 EC, P, B, Na, and 1996 N, B, Na), back-transformed means and the larger SE are presented. Leaf Na and B are expressed relative to dry weight. For Chrysothamnus leaves d.f. = 2 and 30 for Site and Error, respectively. For Sarcobatus leaves d.f. = 3 and 49 for Site and Error, respectively. Different letters in a row indicate significantly different mean values as determined by anovas and subsequent multiple range tests. Significant results are indicated by bold P values
 Diverse DunesSand FlatTransverse DuneShoreline BermF (P)
March 1995 soils
pH9.3 ± 0.1d9.6 ± 0.1c9.9 ± 0.1b10.1 ± 0.1a40.6 (0.001)
EC (dS m–1)0.17 ± 0.02c0.27 ± 0.02b0.35 ± 0.03b1.51 ± 0.82a85.7 (0.001)
N (%)0.031 ± 0.002b0.032 ± 0.001b0.036 ± 0.003b0.053 ± 0.013a4.3 (0.007)
P (p.p.m.)0.75 ± 0.10c0.98 ± 0.35bc1.84 ± 0.44b5.85 ± 4.94a12.3 (0.001)
B (p.p.m.)0.7 ± 0.1d2.3 ± 0.3c8.2 ± 1.1b41.4 ± 25.3a274.5 (0.001)
Na (p.p.m.)132 ± 17c285 ± 31b371 ± 43b16134 ± 781a125.3 (0.001)
August 1996 soils
N (%)0.028 ± 0.002b0.027 ± 0.001b0.030 ± 0.001b0.266 ± 0.183a62.7 (0.001)
P (p.p.m.)0.48 ± 0.40b0.18b0.78 ± 0.17b26.6 ± 412.7a31.6 (0.001)
B (p.p.m.)1.5 ± 0.1d3.8 ± 0.2c10.0 ± 0.7b255.3 ± 91.8a428.6 (0.001)
Na (p.p.m.)232 ± 23c489 ± 23b526 ± 16b12158 ± 3687a268.7 (0.001)
August 1996 Chrysothamnus leaves
Na (p.p.m.)2268 ± 254b3037 ± 269b7271 ± 1862aNone survived17.6 (0.001)
B (p.p.m.)265 ± 30b191 ± 21b583 ± 35aNone survived23.6 (0.001)
August 1996 Sarcobatus leaves
Na (p.p.m.)24577 ± 8752d61687 ± 7851c113592 ± 3397b147556 ± 2576a83.2 (0.001)
B (p.p.m.)70 ± 20d115 ± 11c422 ± 34b795 ± 40a112.7 (0.001)

In August 1996, the soil moisture profile was similar for the Diverse Dunes, Sand Flat and Transverse Dune sites (Table 2). Soil moisture was lowest at shallow depths and increased slightly with depth. At the Shoreline Berm, the high organic matter in the substrate acted like a sponge, resulting in extremely high soil moisture content.

Table 2.  Mean (± SE) gravimetric soil moisture content (% soil dry weight) in August 1996. n = 10 for each mean except for Shoreline Berm where n = 6. For Shoreline Berm, 25 and 50 cm represent the buff granular and organic layers, respectively
Depth (cm)Diverse DunesSand FlatTransverse DuneShoreline Berm
255.5 ± 0.24.8 ± 0.35.8 ± 0.835.4 ± 8.5
508.4 ± 0.86.7 ± 0.45.6 ± 0.4191.0 ± 74.0
7512.4 ± 1.511.0 ± 1.56.3 ± 0.4 
10011.3 ± 1.111.5 ± 0.99.8 ± 1.0 

Plant leaf toxic ions

To confirm the soil gradient, leaf tissue was analysed for Na and B in August 1996. For each species, leaf Na and B paralleled soil Na and B along the gradient (Table 1). Sarcobatus leaves had 14.8% Na and 795 p.p.m. B at the most saline site, Shoreline Berm. Chrysothamnus leaves had 0.7% Na and 583 p.p.m. B at the most saline site where transplanted Chrysothamnus seedlings survived, Transverse Dune.

Plant–plant interactions

All sites in 1995 and the Diverse Dunes and Sand Flat in 1996 were analysed for high-density interspecific plant interactions. In all cases, there were no differences between species in strength of interspecific plant–plant interactions (either competition or facilitation), as indicated by lack of significant interaction terms (Neighbour Biomass × Target Species) (Table 3 and Fig. 3). These analyses used total biomass of targets and shoot biomass of neighbours. The same statistical results were obtained from additional analyses of (i) shoot biomass for both targets and neighbour and (ii) shoot biomass for targets and leaf biomass for neighbours (statistical results not shown). For some sites, a significant target species effect indicated a difference in x-intercept (average biomass for each species in the absence of neighbours) but this term is independent of plant–plant interactions. The analysis of high-density interspecific plant–plant interactions could not be done for Transverse Dune or Shoreline Berm in 1996 because Chrysothamnus mortality resulted in insufficient sample size. Interspecific plant–plant interactions at lower density were examined for the Diverse Dunes in 1995. There was no difference in species competitive ability at this lower density (P for interaction term was 0.70, data not shown).

Table 3. ancova results for August 1995 and August 1996 harvests, assessing species differences for interspecific and intraspecific plant–plant interactions in each garden. Significant results are indicated by bold P-values. Log10 target total biomass was the dependent variable, and neighbour above-ground biomass was the covariate. For each experimental garden, d.f. are 4 for Plot, 1 each for Neighbour Biomass, Target Species, and Neighbour B × Target S. Error d.f. are 8–12 in 1995 and 3–5 in 1996. A dash (–) indicates insufficient sample size for analysis and NA indicates treatment not done. Data are presented in Fig. 3
 Diverse Dunes
Sand Flat
Transverse Dune
Shoreline Berm
Interspecific high density (C–C, S–C, C–IH and S–IH)
August 1995 harvest
 Plot0.078, 0.180.026, 0.430.436, 0.240.069, 0.147
 Neighbour Biomass0.001, 0.950.006, 0.640.007, 0.870.002, 0.787
 Target Species0.004, 0.770.705, 0.0010.004, 0.911.297, 0.001
 Neighbour B × Target S0.039, 0.350.001, 0.830.022, 0.780.016, 0.489
August 1996 harvest
 Plot0.188, 0.340.246, 0.40
 Neighbour Biomass0.231, 0.240.238, 0.32
 Target Species0.193, 0.272.821, 0.013
 Neighbour B × Target S0.231, 0.240.248, 0.315
Intraspecific high density (C–C, S–C, C–MH and S–MH)
August 1995 harvest
 Plot0.047, 0.39NANANA
 Neighbour Biomass0.004, 0.76   
 Target Species0.010, 0.63   
 Neighbour B × Target S0.014, 0.57   
August 1996 harvest
 Plot0.046, 0.9810.250, 0.605
 Neighbour Biomass0.052, 0.7920.274, 0.482
 Target Species1.999, 0.2041.978, 0.216
 Neighbour B × Target S0.027, 0.8500.307, 0.468
Figure 3.

Target plant total biomass (log10) by interspecific neighbour plant above-ground biomass for August 1995 and August 1996 harvests. ancova results are presented in Table 3.

The lack of significant interaction terms allowed us to reject the hypothesis that Chrysothamnus and Sarcobatus differed in relative strength of interspecific plant–plant interactions. However, these results did not adequately distinguish between the possibilities of (i) equal strength effects of neighbours on targets (competition or facilitation), or (ii) no effects of neighbours on targets. Therefore, the slope of each target and neighbour biomass relationship was assessed to see if it was different from zero, for each target species, density, site and year. In no case was the slope of any relationship significantly different from zero, indicating the lack of any competitive or facilitative interspecific interactions, regardless of which neighbour–target combination was examined (Table 4, statistical results for low-density treatments not shown).

Table 4.  Regression analysis for each slope (total target biomass dependent variable and neighbour above-ground biomass independent variable) for all high-density treatments. Dash (–) indicates insufficient sample size and NA indicates treatment not done
SiteCompetitionTarget Species1995
n, r2, P
n, r2, P
Diverse DunesInterspecificChrysothamnus9, 0.01, 0.787, 0.12, 0.44
Sarcobatus10, 0.09, 0.414, 0.02, 0.85
 IntraspecificChrysothamnus8, 0.12, 0.37
Sarcobatus9, 0.13, 0.335, 0.06, 0.70
Sand FlatInterspecificChrysothamnus10, 0.02, 0.737, 0.06, 0.90
Sarcobatus10, 0.12, 0.326, 0.01, 0.90
 IntraspecificChrysothamnusNA4, 0.24, 0.51
SarcobatusNA6, 0.03, 0.73
Transverse DuneInterspecificChrysothamnus7, 0.01, 0.93
Sarcobatus9, 0.01 0.79
 IntraspecificChrysothamnusNA3, 0.99, 0.44
SarcobatusNA5, 0.14, 0.53
Shoreline BermInterspecificChrysothamnus7, 0.18, 0.34
Sarcobatus10, 0.02, 0.67
SarcobatusNA12, 0.01, 0.64

Analyses of intraspecific plant–plant interactions yielded no species differences either for the Diverse Dunes (analysed in August 1995) or at Sand Flat and Transverse Dune (for which August 1996 data were available) (Table 3). Furthermore, regression analyses of target biomass and con-specific neighbour biomass relationships indicated no intraspecific competition or facilitation for any target species, site or year (Table 4, statistical results for low-density treatments not shown).

Mortality was also examined to determine any effect of neighbours on target survival. During the 1995 growing season, target mortality was less than 20% for each species, treatment and site combination (excluding fertilized controls). Mortality for each treatment of targets with neighbours (–IH, –IL, –MH, –ML) was compared with that of the appropriate control treatment (C–C or S–C) for each species and site. Out of the 19 G-tests performed, in no case was there a significant difference in mortality of targets with or without neighbours (P > 0.05 for all 2 × 2 cell G-tests, n = 20–25 for each test). By the harvest date in 1996, target mortality for each species, treatment and site (excluding fertilized controls) ranged from 20% to 100%. Insufficient sample size precluded individual tests of mortality for each treatment (relative to control) for all species and sites (Zar 1996), so data were combined across sites for high-density treatments. For Chrysothamnus (all sites pooled), a G-test indicated no statistical difference in probability of mortality for targets with (combined C–IH and C–MH) or without (C–C) neighbours (G = 0.56, Gcritical = 3.84, 2 × 2 analysis, n = 55). For Sarcobatus at all sites, a G-test indicated no statistical difference in probability of mortality for targets with (combined S–IH and S–MH) or without (S–C) neighbours (G = 0.014, Gcritical = 3.84, 2 × 2 analysis, n = 67). Overall, therefore, there was no effect of neighbours on target mortality.

Stress tolerance

For each species, stress tolerance was assessed by comparison of total biomass of targets across sites (Fig. 4 and Table 5). All living targets were used because neighbours were shown to have no significant effect on targets (Tables 3 and 4). Chrysothamnus total biomass was higher at the Sand Flat than at all other sites, in both 1995 and 1996. Each year, Chrysothamnus root/total biomass ratio, which varied less than twofold, was higher at the sites with lower total biomass. Total biomass of Chrysothamnus plants increased approximately 10-fold from 1995 to 1996 at the Sand Flat, with the root/total biomass ratio remaining about the same. By August 1996, there were no Chrysothamnus survivors at the Shoreline Berm.

Figure 4.

Total biomass (mean ± SE) and root/total biomass (mean ± SE) for each species and site in August 1995 and August 1996. Different letters indicate significant differences among sites within species.

Table 5. anova results for each species in 1995 and 1996, comparing total plant biomass and root/total biomass across sites, using all living target plants. Data are presented in Fig. 4. Significant results are indicated by bold P values

 August 1995
August 1996
August 1995
August 1996
Total biomassSite0.956, 0.0012.035, 0.0021.114, 0.0019.453, 0.001
Plot0.051, 0.2970.428, 0.0480.259, 0.0410.222, 0.243
(d.f. Site, Plot, Error)(3, 4, 37)(2, 4, 19)(3, 4, 45)(3, 4, 36)
Root/total biomassSite0.070, 0.0010.053, 0.0180.084, 0.0010.442, 0.001
Plot0.090, 0.0010.006, 0.6880.002, 0.9070.022, 0.017
(d.f. Site, Plot, Error)(3, 4, 37)(2, 4, 19)(3, 4, 45)(3, 4, 36)

Total biomass of Sarcobatus targets showed a very different pattern across the gradient. In both years Sarcobatus total biomass was higher at the high-salinity sites closer to the lake (Fig. 4 and Table 5). In each year, Sarcobatus root/total biomass ratio, which varied more than threefold across the gradient in 1996, was higher at the sites with lower total biomass. Sarcobatus total biomass increased approximately 10-fold at Shoreline Berm and greater than 40-fold at Transverse Dune from 1995 to 1996, with root/total biomass ratio remaining about the same.

A few plants flowered in 1996. For Chrysothamnus, 3.6% and 12.3% of the living plants were flowering at the Diverse Dunes and Sand Flat, respectively. For Sarcobatus, 3.5% and 6.2% of the living plants were flowering at the Transverse Dune and Shoreline Berm, respectively. For each species, the flowering plants were found at the sites with the greatest biomass.

A plot of health code for living plants of each species, averaged by site, indicated that plants at all of the sites followed the same basic pattern through the study. Most plants lost their leaves immediately after transplant and then recovered (Fig. 5). The rate of mortality during the growing seasons was not substantially greater than during the intervening winter. No significant evidence of herbivory was recorded during the study. At Transverse Dune substantial mortality was caused by blowing sand covering some of the plots (Fig. 5).

A fertilized treatment was included for each species in each site (C–F and S–F) to examine nutrient-limitation stress. By the August 1996 harvest, fertilization had significantly enhanced plant growth for both species at the Diverse Dunes and Sand Flat sites (Table 6); sevenfold and threefold for Chrysothamnus and 51-fold and sevenfold, for Sarcobatus, respectively.

Table 6. anova results for comparison of fertilized targets plants (–F) and non-fertilized targets (all other treatments), for Chrysothamnus nauseosus and Sarcobatus vermiculatus at each site in each year. Significant results are indicated by bold P values
 Diverse Dunes
Sand Flat
Transverse Dune
Shoreline Berm
 Fertilization0.001, 0.9030.213, 0.4780.088, 0.4160.280, 0.209
 Plot0.027, 0.7381.706, 0.0300.134, 0.4150.087, 0.599
 (d.f. Fert., Plot, Error)(1, 4, 20)(1, 4, 9)(1, 4, 6)(1, 3, 4)
 Fertilization4.621, 0.00293.687, 0.0010.113, 0.6991.388, 0.387
 Plot0.3380.193, 0.7181.15, 0.2841.386, 0.529
 (d.f. Fert., Plot, Error)(1, 4, 23)(1, 4, 7)(1, 4, 5)(1, 4, 5)
Chrysothamnus 1996    
 Fertilization2740.3, 0.0016510.1, 0.048
 Plot56.7, 0.0222.798.6, 0.139
  (d.f. Fert., Plot, Error)(1, 4, 9)(1, 4, 6)
 Fertilization277.7, 0.001125.9, 0.0041669.9, 0.310
 Plot0.167, 0.0267.078, 0.5271884.7, 0.334
 (d.f. Fert., Plot, Error)(1, 4, 5)(1, 4, 9)(1, 4, 14)


There were steep gradients in pH and toxic ions (Na and B) across the experimental garden sites, and thus an increasing potential for plant toxic ion stresses with increasing proximity to the lake. The soil gradients in Na and B were paralleled by the concentrations in the leaves of transplants at the end of the study and were consistent with expectations based on previous sampling at this study area (Schaber 1994; Donovan et al. 1997). The response of plants to NPK fertilization indicated that nutrients significantly limited growth of both species at the Diverse Dunes and Sand Flat sites, but that NPK were not limiting at the other two sites. The first growing season of the study followed a particularly wet winter, and soil was at field capacity at planting in March 1995 (data not shown). The large amount of stored soil water plus higher than normal precipitation from April to June 1995 indicates that water was probably not limiting during the first few months of growth of the transplants. However, the virtual lack of precipitation from July to November 1995, drier than normal winter 1996, and again a virtual lack of precipitation from May to the final harvest in August 1996 make it likely that water limitations were important during the majority of the growth of the plants in this study. These precipitation patterns and the soil moisture profile in August 1996, with very dry shallow soil (all sites except the Shoreline Berm), are typical for Great Basin habitats (Caldwell 1985; Dobrowolski et al. 1990; Muller et al. 1995).

Contrary to our expectation, we found no difference in relative competitive ability of Chrysothamnus and Sarcobatus, and in fact no evidence of any plant–plant interactions either interspecific or intraspecific. Although larger sample sizes would have been desirable, particularly in 1996 when mortality reduced sample sizes for some comparisons, there were no obvious trends in the data that might have been bolstered by increased sample size. This held true for both the ancovas assessing relative competitive ability (comparing slopes) and the regression analyses testing whether individual slopes were different from zero. Nor was mortality of targets affected by presence of neighbours. Therefore, neither competition nor facilitation played a large role in the growth and survival of juvenile plants during this study. It is remotely possible that concurrent or consecutive episodes of competition and facilitation during the growing season resulted in no net effect being observable at end-of-season harvest. However, it is unlikely that such counteracting effects could have been large and yet equally balanced in all treatments sites, and years.

One potential critique of this study is that the design used for comparison of species’ interspecific competitive ability incorporated both effect and response competition (sensuGoldberg 1990; Goldberg & Scheiner 1993) and did not allow for their independent quantification. However, this study did include intraspecific treatments with the same target and neighbour species. Given that there was no significant slope detected in any of the interspecific or intraspecific combinations that we could analyse (Table 4), there is no evidence for competition in any treatment. Thus, the separation of effect and response competition, which would have required additional controls, is a moot point for this study.

The lack of net plant–plant interactions in our study is surprising. Other studies of plant–plant interactions with desert shrubs, mostly neighbour removal experiments, support the expectation that desert shrubs are competing with each other or with herbaceous plants for soil resources (McKell & Chilcote 1957; Friedman 1971; Friedman & Orshan 1975; Fonteyn & Mahall 1978, 1981; Ehleringer 1984; Parker & Salzman 1985; Fowler 1986; Manning & Barbour 1988). In addition, reviews addressing plant interactions in many habitats report that the majority of experimental manipulation studies provide evidence for one or both of competition or facilitation (Aarssen & Epp 1990; Goldberg & Barton 1992; Callaway 1995; but see Underwood 1986). Only a few field manipulative experiments report results indicating no plant–plant interactions (Rabinowitz & Rapp 1985; Manning & Barbour 1988; Goldberg & Barton 1992; McPherson 1997; Briones et al. 1998).

Our study concentrated on juvenile plants because we expected them to experience competition for limited available water and nutrients in the shallow soil layers. At study initiation, juvenile plants had a rooting depth of only 0.1 m. However, roots of both species grew through the surface soil layers relatively quickly, and by the final harvest surviving plants were generally rooted to at least 1 m depth (or to groundwater at Shoreline Berm) (L. A. Donovan & J. H. Richards, unpublished data). This fast root growth minimized the duration of exposure to limiting soil moisture conditions present in the shallow layers and may have minimized interactions with other plants for moisture. In contrast to water, the lack of competition for nutrients cannot be readily explained by the quick attainment of deep rooting depth. In addition to the deep roots, plants of both species had abundant fine root growth in the surface soil layers, with root systems of targets and neighbours intertwining thoroughly. Available nutrients were low throughout the soil profiles (with the exception of the Shoreline Berm) and were also undetectable in the groundwater throughout most of the year (J.H. Richards, unpublished data). At the Transverse Dune and Shoreline Berm sites, where the comparison of NPK fertilized and control plants indicated no nutrient limitations (Table 6), high N concentrations in leaves of mature Sarcobatus (2.0–4.5% June–October) are also indicative of high levels of available N (Donovan et al. 1997). However, we still cannot explain the lack of competition for nutrients at Diverse Dunes and Sand Flat, where nutrients were limiting and both species survived.

Plant–plant interactions in this natural community may be more prevalent between juveniles and adults than among juveniles (Grace 1985; Fowler 1986; Callaway & Walker 1997). Unless juveniles find themselves in an open space created by disturbance or senescence of an older individual, the established adult shrubs have the potential both to facilitate juveniles by modifying a harsh above-ground micro-environment and to compete with juveniles by depleting resources (Holmgren et al. 1997). Future experiments are planned where juveniles will be planted into existing shrub matrices and neighbour removal designs will be used to assess juvenile–adult interactions. However, preliminary spatial analysis at our site provides only weak evidence of competition between adults and juveniles (C.A. Toft, unpublished data).

Our study indirectly contributes to the ongoing debate on the importance of competition along productivity gradients (Twolan-Strutt & Keddy 1996; Goldberg & Novoplansky 1997; Peltzer et al. 1998, references therein). Our study was not designed directly to address the debate. However, it contributes a case study for juvenile woody plants in a low productivity saline desert habitat and apparently weighs in on the side of competition playing little role. One interesting development in this debate is the consideration that resource dynamics or interpulse period length may determine the relative importance of competition. Goldberg & Novoplansky (1997) suggest that competition will be a relatively unimportant determinant of overall plant performance in unproductive environments where (i) individual plant survival is primarily determined by long intervals between resource pulses, and (ii) soil resource availability during interpulse intervals is largely independent of plants, i.e. abiotically driven. This scenario may apply to our Mono Lake site, where soil resources are most readily available in a short pulse in the spring and subsequent survival of plants exposed to soil toxicity, long dry growing seasons and cold winters is apparently more dependent on abiotic stress tolerance. However, for competition to be relatively unimportant, plant size must also be unrelated or negatively related to plant survival during the interpulse interval (Goldberg & Novoplansky 1997), and this does not appear to be true for Chrysothamnus seedlings or adults (Donovan et al. 1993; Toft 1995).

For stress tolerance, we did find the expected difference between Chrysothamnus and Sarcobatus. Sarcobatus was much more stress tolerant; it survived and grew better in the higher soil Na and B regions of the gradient, close to the lake. However, our very definition of the stress gradient is called into question by examining sites where each species performed best. Chrysothamnus grew best at a low Na and B site along the gradient (Sand Flat) and survived poorly or not all at the higher Na and B sites. In contrast, Sarcobatus grew best at the high Na and B sites. This is unlikely to be due to a requirement for Na because controlled environment studies indicate that increasing NaCl decreases growth of Sarcobatus when all else is equal (Glenn & O’Leary 1984; Richards 1994). Other abiotic factors apparently contribute to the poorer performance of juvenile Sarcobatus at the low Na and B sites, perhaps greater sensitivity to nutrient limitations, or salinity and nutrient interactions. Preliminary data indicate extremely low P and N acquisition by Sarcobatus at the two low-salinity sites. In contrast, Chrysothamnus, while apparently N limited at these sites, had much higher leaf P concentrations compared with the high-salinity sites. Details of nutrient limitations and potential nutrient interactions with Na and B tolerance will be addressed in a forthcoming paper (L.A. Donovan & J.H. Richards, unpublished data).

The lack of any competitive effects means that a trade-off between competitive ability and stress tolerance could not be assessed for juveniles in this study, even though one might have been expected for these species. Overall, abiotic stress tolerance limits Chrysothamnus distribution to non-saline areas and permits Sarcobatus distribution to extend into more saline regions of the stress gradient. A lack of interspecific competition among juveniles is also consistent with the observed co-dominance of Chrysothamnus and Sarcobatus in the less saline areas. Species’ differences in stress tolerance and physiological requirements appear to be most important in determining their distribution along the gradient.


The authors thank C. Toft and D. Elliott-Fisk for collaboration at the field sites, D. Chirman, A. Cook, J. P. Davis, G. Dodd, K. Fort, C. Miller, C. Hinkson, M. Ho, G. Kyser, S. Matzner, M. Muller, K. O’Keefe, R. Pappert, N. Pergam, B. Richards, B. Rojas, E.J. Schaber, J. Wall and J. Wooten for assistance with field and laboratory work, and Dave Grisé for generous assistance with statistical analyses. This research was supported by USDA NRICGP grants 92–37101–7419 (C.A Toft, J.H. Richards and D. Elliott-Fisk) and 94–37101–1144 (L.A. Donovan), NSF grants IBN-9816670 (L.A. Donovan) and IBN-9903004 (J.H. Richards), and by the California AES. Fieldwork operated out of the Sierra Nevada Aquatic Research Laboratory, University of California. The US Forest Service provided access to the study site through a Special Use Permit. The co-operation of L. Ford and the Mono Lake Ranger District was appreciated.

Received 21 January 1999revision accepted 8 June 1999