Present address and correspondence: Graeme T. Hastwell, Cooperative Research Centre for Australian Weed Management, Alan Fletcher Research Station, Department of Natural Resources and Mines, PO Box 32, Sherwood, Queensland 4075, Australia (tel. +61 73375 0723; fax +61 73379 6815; e-mail firstname.lastname@example.org).
1Although plants growing below tree or shrub canopies in xeric ecosystems often exhibit higher growth or survival than conspecifics growing in the open, few studies have sought to identify the processes that facilitate plant performance.
2We tested the relationship between the importance of facilitation and environmental severity in a field experiment in which seedlings of the perennial shrub Enchylaena tomentosa were shaded at different times of year. We measured plant growth and survival on six occasions between June 1999 (winter) and February 2000 (summer).
3Although seedling relative growth rates (RGR) changed with growing conditions, shade always increased mean RGRs. The intensity of facilitation (the difference in mean RGRs due to shade) did not differ between measurement periods.
4Shade reduced survival during winter and spring, but had a strong positive effect during summer. Larger seedlings had lower mortality rates during summer, suggesting that prior facilitation of growth indirectly aided survival.
5Our results provide only partial support for the hypothesis that the importance of facilitation increases with environmental severity. Whereas the importance of facilitation of seedling survival was high under severe conditions, the importance of facilitation of seedling growth did not increase under more severe conditions.
6We conclude that the relationship between facilitation and environmental severity is more complex than previously recognized, and that the effects of plant interactions can vary between different aspects of plant performance.
The usage of the terms ‘importance’ and ‘intensity’ in the literature has been inconsistent. Here, ‘intensity’ is the change in plant performance that occurs as a result of interactions with neighbouring plants. This is analogous to ‘absolute competition intensity’ (Grace 1995; Weigelt & Jolliffe 2003), and in the context of the experiment we describe here is identical to Holzapfel & Mahall's (1999) ‘net effect’.
Emphasizing that importance is not necessarily correlated with the intensity of interactions, Welden & Slauson (1986) proposed that importance could be measured as the proportion of the total variance in data that anova attributes to the effects of plant interactions, i.e. the relative degree to which they contribute to changes in growth rate, metabolism fecundity, survival or fitness. Although much discussed, this measure is infrequently reported. More often, relativized measures (Grace 1995) are used as de facto measures of importance (e.g. Tielborger & Kadmon 2000; Pugnaire & Luque 2001).
Shading has multiple effects that may simultaneously benefit and impair plants; it reduces plant and soil temperatures (Turner et al. 1966; Franco & Nobel 1989), decreases evaporation rates (Valiente-Banuet & Ezcurra 1991; Breshears et al. 1997; Breshears et al. 1998), and has complex effects on plants through photosynthesis and morphological plasticity (Holmgren 2000; Ryser & Eek 2000). At any given time, the overall balance of these positive and negative effects will determine the observable change in plant performance due to shading (Holmgren et al. 1997; Holzapfel & Mahall 1999). Furthermore, the balance between positive and negative effects may change between seasons. For example, during months when air temperatures are high and water availability is low, the positive shade effects of reduced soil temperatures and evaporation rates may outweigh any negative effects of reduced light levels. Conversely, if periods of high water availability coincide with cloudy weather or short day lengths, light may become limiting, resulting in the overall balance becoming negative.
Enchylaena tomentosa R. Br., a small perennial chenopod shrub found throughout much of southern Australia (Jessop & Toelken 1986), occupies contrasting microhabitats across its range. It grows in the open on coastal foredunes, but is restricted to under-canopy habitats in arid regions (Barker 1972; Tester et al. 1987). The habitat preference in arid regions has previously been attributed to seed dispersal by birds (Tester et al. 1987) and higher emergence and growth rates in under-canopy soils (Facelli & Brock 2000). However, these factors do not explain why microhabitat distributions differ between climatic zones.
Few studies have attempted to unravel the causative factors affecting plant performance in under-canopy microhabitats, possibly because of the complex, multiple effects that trees have on local environments (Belsky 1994). Nevertheless, such studies may offer insights into the effects of heterogeneity and the processes underlying plant interactions.
We hypothesize that shading enhances plant performance in arid systems, and predict that shade-induced facilitation of seedling survival helps determine the microhabitat distribution of E. tomentosa. In accordance with the Callaway & Walker (1997) and Brooker & Callaghan (1998) models, we further predict that the effects of shade on seedling growth and survival will be more important during summer, when conditions are most severe, than during winter. We tested these hypotheses while controlling for non-shade effects of trees, such as their role in creating fertile islands in substrates (Zinke 1962), by mimicking natural shade with artificial canopies.
The study was conducted at the T.G.B. Osborn Vegetation Reserve on Koonamore Station (32°07′ S, 139°22′ E, altitude 200 m a.s.l.), 60 km north-west of the nearest official weather station at Yunta, South Australia. The climate is arid, with a mean annual rainfall of approximately 200 mm (Osborn & Paltridge 1935; Carrodus et al. 1965). While median rainfall does not differ between seasons (P = 0.3902, Kruskal–Wallis), winter (June to August) rainfall is less variable between years than that in any other season. Mean daily minima and maxima at Yunta (1962–94) range from 3.0 and 15.5 °C, respectively, in July to 15.1 and 32.8 °C in January (data courtesy Australian Commonwealth Bureau of Meteorology). Koonamore maxima are higher, although the data were not collected under standard conditions (Carrodus et al. 1965; R. Sinclair, unpublished data).
We established two 100 × 250 m plots near the north-western corner of the Reserve. Both plots had similar slopes and aspect, with solonized brown soils (Northcote Gc 1.12) (Carrodus et al. 1965). They had a sparse tree cover of scattered Myoporum platycarpum and Alectryon oleifolius, with E. tomentosa common in the understorey vegetation. Atriplex vesicaria and Eremophila spp. were the most common shrubs, and the herbaceous layer was primarily composed of Sclerolaena spp., Danthonia caespitosa and Stipa spp. Nomenclature follows Jessop & Toelken (1986).
We used artificial canopies so that shade treatments were applied independently of the fertile islands associated with trees (Vetaas 1992; Facelli & Brock 2000). Each artificial canopy consisted of a 3.6 × 3.6 m piece of black nylon 80% WeathashadeTM shadecloth suspended 1.2 m above the ground from wooden posts. In each plot we erected two replicate canopies at randomly generated co-ordinates.
We planted E. tomentosa seedlings in winter (15–17 June 1999). Seedling height and width at time of planting were 90 ± 19 mm and 50 ± 9 mm, respectively (mean ± SD). We made four 2-m long east–west rows at each canopy, the ends of each row being 0.8 m inside the eastern and western edges of the canopies (Fig. 1). Five seedlings were planted along each row, with the distance between each seedling varying randomly between 10 and 50 cm. We took this precaution to reduce the potential effects of spatial periodicity in any uncontrolled factors that might affect survival or growth (Quinn & Keough 2002). Twenty seedlings were planted under and adjacent to each canopy, giving a total of 80 seedlings across the four canopies. Wire mesh cages provided protection from vertebrate herbivores.
The time of year that rows were shaded at noon was determined by their position (Fig. 1). The northernmost rows received direct noon sunlight all year (hereafter no shade), while the rows inside the northern canopy edges received direct sunlight at noon in winter, but were shaded in summer (summer shade). The rows inside the southern canopy edges were shaded at noon all year (all shade), while the southernmost rows received direct sunlight in summer but were shaded at noon in winter (winter shade). Shading between seedlings was restricted to brief periods after dawn and before dusk, when light levels were very low. All rows received direct sunlight during early morning and late afternoon.
The noon shade status of the summer shade and winter shade rows changed when the solar elevation moved through 50° above horizontal. Using formulae for solar declination and elevation (public communication, http://www.uniwinnipeg.ca/~blair/physclim/lab2.htm), we calculated that the summer shade rows were sunlit, and the winter shade rows shaded, from the start of our experiment until 3 September.
Winter and early spring 1999 (June to September) were dry, so we watered the seedlings at the time of planting, and on 30 June, 24 July and 29 July 1999, to ensure successful establishment. No further supplementary watering was provided.
Seedling size was recorded as the maximum height and width of each seedling. Seedling data were recorded on 17 June, 24 July, 13 October, 24 November 1999, 18 January and 22 February 2000. The experiment was terminated on 22 February 2000 after 36 weeks.
physical effects of artificial canopies
Although we did not attempt to emulate the shade of any particular species, we sought to impose a treatment that was biologically plausible. To verify this, we compared light transmission and soil temperatures below the artificial canopies with those below A. oleifolius trees.
We measured PAR with two quantum sensors (Li-Cor LI-190SZ), one in full sunlight and the other 0.5 m south of an A. oleifolius trunk, or in the centre of the all-shade row under an artificial canopy. Instantaneous measurements were made every 15 s, and the mean of four consecutive readings was logged every minute (Data Electronics Datataker 50). Soil temperatures were measured using Type T thermocouples at a depth of 5.5 cm at the quantum sensor locations. Instantaneous temperature measurements were logged at 5-minute intervals (Grant Instruments 1203 Squirrel logger). Limited equipment availability prevented simultaneous measurements under A. oleifolius trees and artificial canopies, so we collected data on consecutive days with comparable weather conditions.
Canopy light transmission was derived by calculating shade PAR as a percentage of sun PAR. As A. oleifolius data were skewed, means were bootstrapped (bias-corrected and adjusted percentiles, 10 000 permutations) (MathSoft 1999).
On 24 July 1999, we found that 13 E. tomentosa seedlings had suffered herbivore damage. These seedlings showed large reductions in height during June–July, and large increases during July–October. We excluded these data from analyses of plant growth for June–July and July–October, as the growth patterns differed from those of undamaged plants (ln RGRheight: F1,80 = 6.7802, P = 0.0102, anova). Subsequent growth data from surviving plants were not excluded as there was no evidence of ongoing effects. Survival was not strongly affected (P = 0.0794, Fisher Exact test) so data from damaged plants were included in survival analyses.
Wind damage to one canopy left a shaded row partially exposed to direct noon sun during November–January. Data from these seedlings were excluded from all analyses after the October–November period. The summer shade and winter shade rows were excluded from the July–October data because of the change in their noon shade status.
We tested whether the intensity of shade-induced facilitation changed during the experiment, whether the importance of facilitation increased with environmental severity, and whether size-related effects indirectly facilitated survival.
We calculated mortality rates for each row in each period as
where M is the mortality rate, D is the number of deaths in the row during the period, I is the number of live plants in the row at the start of the period and t is the duration of the period in days.
The mortality rates could not be transformed to fulfil the assumptions of anova, so we used permutation methods (Anderson 2001), with Period and Shade as factors. Shade had two levels, shaded and sunlit. Missing values were replaced by mean mortality rates within each treatment level. Periods in which shade affected mortality rates were identified using post hoc one-way anovas on data for each period.
We measured environmental severity as the mortality rates in sunlit rows, equating higher mortality rates with increased severity. E. tomentosa shows few, if any, indications of phenological patterns in growth or survival under both field and glasshouse conditions. Consequently, we argue that direct measures of E. tomentosa performance under ambient conditions quantify environmental severity more reliably than indirect measures such as rainfall or temperature.
We determined whether severity differed between periods with pairwise permutation tests. Importance was calculated using sums of squares from one-way permutation tests (Anderson 2001) for each period with Welden & Slauson's (1986) formula:
The effects of plant size on survival were analysed by the Cox proportional hazards method (Venables & Ripley 1999), using the size of the plant when last alive as a covariate. We tested whether shade may directly facilitate survival by comparing the survival of plants in sunlit rows with that of similarly sized plants in shaded rows during the periods of highest mortality (July–October and January–February, Fisher Exact Test).
We tested whether the intensity of shade-induced facilitation of growth changed during the experiment, and whether the importance of facilitation increased with environmental severity. We calculated relative growth rates (RGR) for plant height and width with a version of Fisher's (1948) relative growth rate adapted for discrete time intervals
with Time given as the duration of the period in weeks. RGRs were ln transformed, and we used repeated measures anovas, treating the subject (i.e. the individual plant) as a random effect (SAS Institute 1997). The transformed June–July RGRheight data were excluded as they departed from normality.
If the effects of shade on seedling RGRs differed between periods, the intensity of facilitation must have changed during the experiment. We tested this by including the interaction term Period × Shade in the statistical model for the repeated measures anovas.
We measured environmental severity as the RGRs in sunlit rows, equating lower RGRs with increased severity. We determined whether severity differed between periods by conducting repeated measures anovas on the sunlit RGRheight and RGRwidth data, followed by post hoc comparisons between all periods.
We calculated importance with the sums of squares from anovas comparing sunlit RGRs with shaded RGRs for each period, using Canopy, and Shade nested within Canopy, as factors. Importance was calculated as
Following a dry winter and early spring, above average rain fell from mid-spring to early summer (October–December). However, mid-summer (January–February) was particularly hot and dry, with daily maxima exceeding 45 °C several times. Heavy rain then fell in the final week of the experiment.
Light transmission by artificial canopies (15.76%) was comparable with that of similarly sized A. oleifolius canopies (15.71%). However, the artificial canopies had less effect than A. oleifolius in reducing near-surface soil temperatures. Whereas shading by artificial canopies reduced soil temperatures by up to 15 °C, A. oleifolius reduced soil temperatures by more than 20 °C. This suggests that our experimental results are likely to underestimate the facilitative effects of shade by trees during summer, particularly with respect to survival.
Mortality rates differed between periods (F4,79 = 21.6108, P = 0.0001, NPmanova), and were affected by shade (F1,79 = 10.6242, P = 0.0005). Moreover, the intensity of shade-induced facilitation of survival changed through time (Period × Shade, F4,79 = 11.5486, P = 0.0001). Shade had significant effects on survival during July–October and January–February (Fig. 2). However, these periods differed in that shade had negative effects on survival in cool conditions (July–October), and positive effects in very hot summer conditions (January–February).
Environmental severity, as measured by ambient mortality rates, was highest during January–February (Fig. 3a). The remaining periods did not differ from each other. Importance was high during January–February, reflecting the positive effects of shade on survival under adverse conditions. Importance was low when conditions were more benign, except during July–October, when shade had negative effects on survival. Note that by definition importance is always positive, regardless of the direction of interactions, because it is calculated from sums of squares.
Larger plants survived longer than smaller plants (Table 1a,b), so shade-induced facilitation of growth could subsequently indirectly increase survival. Shade also had direct effects, decreasing survival in winter and spring but increasing it in summer. Survival of shaded plants was lower than survival of similar-sized sunlit plants in July–October (P = 0.0471, two-tailed Fisher Exact Test), but during January–February survival of shaded plants was higher than that of similar-sized sunlit plants (P = 0.0041, similar heights; P = 0.0456, similar widths; one-tailed Fisher Exact Tests).
Table 1. Effects of (a) plant height and (b) width on E. tomentosa seedling survival (Cox proportional hazards)
P > χ2
The RGRs of seedlings in shaded rows were higher than those of seedlings in sunlit rows, and RGRs differed between measurement periods (Table 2a,b, Fig. 4). However, the interactions between Period and Shade were not significant, so there was no evidence that the intensity of shade-induced facilitation of growth changed during the experiment (Table 2c,d).
Table 2. Effects of shade on relative growth rates of E. tomentosa seedlings*
P > F
Repeated measures anova (SAS Institute 1997), with Subject as a random effect. June–July RGRheight data are excluded. (c) and (d) show the non-significant time–treatment interaction terms that were dropped from the final models.
(a) ln RGRheightr2 = 0.2563
(b) ln RGRwidthr2 = 0.4861
(c) ln RGRheight
Period × Shade
(d) ln RGRwidth
Period × Shade
There was no evidence to support the prediction that the importance of shade-induced facilitation of growth increases with environmental severity (Fig. 3b,c). Rather, the RGRheight data suggest that the importance of facilitation of growth decreased as conditions became more severe. The RGRwidth data are consistent with a unimodal relationship, importance being highest when conditions were moderately unfavourable, and lowest when conditions were either benign or particularly severe.
Our results reveal a complex relationship between facilitation and environmental severity. Part of this complexity arises from differences between the effects of shade on growth and survival. Further, the effects of seedling size on survival show that facilitation of growth during one period could increase survival during subsequent stressful periods, thus propagating facilitation through time across different aspects of plant performance.
environmental severity and the intensity of facilitation
Although shading facilitated both growth and survival of E. tomentosa seedlings, the temporal patterns of facilitation of these two aspects of plant performance differed. The intensity of shade-induced facilitation of growth did not differ during the course of the experiment, whereas its effect on survival varied in both intensity and direction: shading reduced survival in winter–spring, and increased survival at the height of summer. Thus the effects of shade diverged during winter and spring, increasing growth but simultaneously reducing survival.
Although the literature has primarily focused on the relationship between environmental severity and the importance of facilitation, some attention has also been given to changes in the intensity of facilitation. For example, Callaway (1997, p. 145) discusses experimental evidence showing ‘that positive effects become stronger as abiotic stress increases’. Elsewhere a model shows that, for a given nurse plant size (or density), the facilitative ‘strength’ of the nurse plant effect increases as abiotic stress increases (Callaway & Walker 1997,Fig. 1, p. 1962). Our results provide partial support for these ideas; the survival data concur with them while our results for growth do not. Given that growth requirements probably differ from those for survival, there is little reason to expect these two aspects of plant performance to show similar patterns of response to positive interactions (Goldberg et al. 2001). Growth responses are also likely to change over a larger range of resource levels than survival responses. It is also possible that some mortality-inducing factors may act over such short time-scales that their net effects on growth could be negligible. Consequently, the effects of facilitation on growth should probably be considered separately from its effects on survival.
environmental severity and the importance of interactions
We found no evidence to support the prediction that the importance of facilitation of growth increases with environmental severity (see Callaway & Walker 1997; Brooker & Callaghan 1998), but the importance of shade-induced facilitation of survival was high under severe conditions as predicted. Whereas the intensity of facilitation of growth did not vary with environmental severity, the importance of facilitation of growth varied greatly. Moreover, the pattern of variation was irregular, with some suggestion that importance may decrease with environmental severity.
Our results concur with Goldberg et al.'s (2001) conclusions that the effects of plant interactions on survival differ from their effects on growth. Goldberg et al. (2001) also hypothesized that the facilitative effects of density-dependent interactions primarily affects survival, and that exploitation competition primarily affects growth. Our experiment was not intended to address density-dependent phenomena, although the shade treatment may have simulated some positive density-dependent effects. However, shade is unlikely to have reduced below-ground resource availability. This might explain why our results show facilitative effects of shade on both growth and survival, instead of only facilitating survival as hypothesized by Goldberg et al. (2001). Whether this was indeed a consequence of the minimal below-ground competition for resources in our experiment is an important question that requires verification.
shading and the spatial distribution of e. tomentosa
The high summer mortality rate of sunlit seedlings suggests that E. tomentosa may not be able to establish in chenopod shrublands in the absence of trees, suggesting a factor in addition to seed dispersal (Tester et al. 1987) and soil properties (Facelli & Brock 2000) to explain the microhabitat distribution of E. tomentosa in this system. A glasshouse trial showed that emergence of E. tomentosa would occur in soils from open areas providing soil moisture levels were adequate, but that both emergence and growth were higher in soil from below Acacia papyrocarpa (Facelli & Brock 2000). Thus although it is lack of shade that is responsible for the absence of E. tomentosa from exposed habitats, soil characteristics could also affect distribution. Given that higher seedling growth rates enhance survival, the effects of shade under tree canopies are likely to be amplified by the positive effects of increased water retention and higher nutrient availability in soils under tree canopies.
Our results show that the higher survival of shaded plants during summer is a result of both the direct effects of habitat amelioration by shade and the indirect effects of shaded plants attaining greater size before the onset of extreme conditions. We surmise that the effects of shade in lowering soil and air temperatures and transpiration may reduce mortality rates in shaded plants during summer. Elevated soil temperatures have previously been implicated in increased mortality rates in cacti (Turner et al. 1966; Valiente-Banuet & Ezcurra 1991), and the differences we found between sunlit and shaded soil temperatures are comparable with those reported in earlier studies (Valiente-Banuet et al. 1991). The only negative effects of shading on survival occurred during July–October, when the level of shade we applied presumably reduced light to limiting levels.
Our finding that the intensity of facilitation of growth by shade did not vary with environmental conditions may not be inconsistent with other studies that have shown that the effects of shade on growth varied with soil type (Franco-Pizana et al. 1996; Weltzin & McPherson 1999; Schenk et al. 2003) and rainfall (Belsky 1994). These conclusions were based on studies of extant vegetation (Belsky 1994; Weltzin & McPherson 1999; Schenk et al. 2003) or pot experiments where seedlings were grown at relatively high densities (Franco-Pizana et al. 1996), so it is likely that interactions between individuals of the target species influenced the outcomes. While we cannot altogether discount the possibility of intraspecific interactions between our E. tomentosa seedlings, the relatively low planting density and absence of adjacent vegetation should have reduced intraspecific interactions to low levels.
Most studies on the effects of trees in arid systems have focused on spatial heterogeneity in resource levels (Vetaas 1992; Belsky et al. 1993) and, to a lesser extent, temporal variability (Belsky et al. 1989; Facelli & Brock 2000). Our results shed light on the temporal dynamics of the effects of one component of tree-induced heterogeneity on understorey vegetation. They show that even in the absence of elevated soil nutrient levels (Vetaas 1992; Facelli & Brock 2000) and hydraulic lift (Caldwell et al. 1998), levels of shade similar to that produced by trees increased E. tomentosa growth rates across a range of seasonal conditions. Somewhat surprisingly, shading was as advantageous for growth in winter as it was in summer, presumably because ambient light levels in this region tend to be relatively high year round. The negative effects of shading on survival during winter were puzzling given the positive effects on growth. It is possible that biomass allocation to roots was very low in the shade treatment during winter, in which case the trade-off between shade and drought tolerance for E. tomentosa seedlings may manifest during winter rather than summer in this system (Holmgren et al. 1997).
The effects of long-lived trees and shrubs on their local environments are multifaceted and context-dependent (Belsky et al. 1993; Belsky 1994; Facelli & Brock 2000). By examining the effects of a single canopy-related process in a highly variable system, our experiment constitutes an initial step towards a detailed mechanistic understanding of the relationships between plant environmental modification and plant interactions. Most of all, our results indicate that plant interactions are complex and that their outcomes are sometimes counterintuitive.
The results of this and other experiments suggest that current models of facilitation require elaboration, and that they need to differentiate between survival and growth responses. Clearly the relationship between facilitation and environmental severity is more complex than previously thought, for neither the intensity nor the importance of facilitation necessarily increases as conditions become more severe. Moreover, the effects of facilitation on survival and growth may differ simultaneously, presumably as a consequence of differences in resource requirements. We contend that understanding and predicting plant interactions requires consideration of the physiological characteristics of the interacting plants and the environmental modifications that plants produce.
We thank J. Trice, J. Prider, D. Ladd, J. Moyle-Croft, T. Lenz, C. Hodgson, G. Fogarty and N. Wilczinski for help in the field, and R. Sinclair for access to the T.G.B. Osborn Vegetation Reserve records and much more. We also thank L. MacLachlan and B. Pumper for generously providing access to Koonamore Station, and for supporting arid zone research over many years. M.J. Anderson provided advice on non-parametric methods of calculating importance. R. Callaway and J. Prider provided valuable comments on early versions of this paper. We were greatly assisted by comments from L. Haddon, M.J. Hutchings and three anonymous reviewers.