Competitive effect and response of ten Namaqualand ephemeral plant species at two nutrient levels

Competitive effect and response hierarchies between Namaqualand pioneer plant species, across two nutrient levels, were constructed. The ten species investigated were: Arctotheca calendula, Dimorphotheca sinuala, FoveoHna a/bida, Gazania IichtensteinH, HefiophHa variabilis, Leysera fenella, Oncosiphon grandiflorum, Osteospermum hyoseroides, Senecio arenarius and Ursinia cakilefolia. The target species used to construct the ran kings were Dimorphotheca sinuata, Gazania lichtensteinii, Heliophila variabilis and Ursinia cakilefolia. Competitive effect as well as response ran kings were found to be concordant between the two nutrient levels, Le. soil fertility did not have a significant effect on the competitive effect or response hierarchy, as well as between the target species. Competitive eHect and response were significantly negatively correlated only at the low nutrient level. Competition intensity for each pairwise interaction showed no significant difference between target species however, differences were found between neighbour species and nutrient levels, competition intensity being higher at low nutrient levels.


Introduction
The arid Namaqualand covers an area of approxi mate ly 55 000 km 2 and is situated in the North-western corner of the Republic of South Africa. The climate is characterised by a hot, dry summer and a sparse and erratic rainfall, falling mainly in winter (Schulze 1965). Namaqu aland owes its fame mainly to the display of annual wil d flowers. whi ch transform s the normally barren landscape into a land of co lour in the spring following a good rainy season (Van Rooyen et al. 1992).
Ephemeral plant populations in Namaqualand vary considerably in spec ies co mposition and abundance from year to year. This variati on is primarily due to the unpredictability of the timing of the lirst rains. In high rainfall ye ars , Namaqualand ephemeral plan t species form dense stands and probably compete intensely for limi ted resources (Va n Rooyen 1988;Van Rooyen et al. 1992;Oosthuizen elal. 1996aOosthuizen elal. , 1996b. Mass ive floral displays of ephemerals in Namaqualand are visited by many touri sts eac h year. Competition between these species in flu ences thei r performance and display (Oosthui zen et ai. 1996a). Understanding the facto rs that in fl uence competition are necessary for optimal utili zat ion and management of the ephemeral vegetation as a tou ri st attraction. Any factor that can give one species a competiti ve advan tage over anot her has the potential of changing the noral d ispl ay.
The importance, and even existence, of competition among plants in arid ecosystem s has o ften been questioned (Fow ler 1986). Shm ida el al. (t986) argued th at, under the harsh and unpredictable conditions charac terisi ng desert environnnents, the probability is very low that densities increase up to levels in whi ch co mpetition becomes important. Other studies (KUkorr 1966;Friedman & Ors han 1974;Inouye el al. 1980;Kadmon & Shmida 1990a, 1990b suggest that co mpetition may play a maj or role in determining the dynamics of desert annual plant populations. 'Competitive ability' has two co mponents (Goldberg 1990): competi ti ve effec t (t he ab ility to dep ress the growt h or reproducti on of neighbours) and competitive response (the ability to withstand the negative effects o f neighbours). These t.:an both be estimated by growing species in additive mixtures and measuring the reduction in performance of species in mixtures relative to control s (Keddy el al. 1994). In thi s study a phytometer or indica£or approac h based on a modified additive design was appli ed in which the relat ive competitive performance of a species is evaluated by measuring its relati ve ability to suppress the growth of a common indi cator species (the phytometer) (Gaudet & Keddy 1995).
The aim of this study was to determine the competi ti ve effects an d responses of ten prominently displaying plant species that occur in dense stands in Namaqua land at two fe rtility levels. The questions to be answered were: Are the co mpetitive effect and response hierarchies consistent a) among targe t species and b) across nutrient leve ls?
Diaspores of the ten species were sown. out of doors, in sand filled pots with a volume of 1000 em' in April 1995 at the University of The a bove~gro und parts of the plants were harvested 119 days (17 weeks) after sowing and the dry mas s of each plant was determined after being dried for one week at 60· C to a constant mass . Five n::petiti ons of al l treatments were harvested . The following indices were calculated: a) RYP, rel ati ve yield per plant (Harper 1977): It should be noted that these measures of relative yield are based on an additi ve design and not a re placement seri es design ( Kedd y et al. 1994) b) I, co mpetition intensity (Kedd y el al. 1994) : J is the intensity of the interaction between species i and j.
A one way analysis of variance (Bonferroni) as well as a multi factor ANOYA were used to test for significant differences at a = 0.05 (Sakal & Rohlf 1982). Target species RYP values were used to det e r~ mine diffe rences between nutrient levels for co mpetitive effect abil~ ity while neighbour species RYP values were used for competitive response ability. Differences in competiti on intensity between nutrient levels, target and neighbour species were also determined using Bonferroni .
Kendall 's rank correlations (S teyn elal. 1987) were used 10 lest for concordance of ranking of competi tive effect and response between nutrient levels and among target species w ithin a nutrie nt level. Mean RYP values were used to establi sh one effect and one response matrix for each treatment. The effec t matrices include the mean effect of each neighbour species on each target species, mean effect of each nei ghbour species o n all target s pecies (row means of include mean response of each neighbour species to each target sre· ci es, mean res ponse of each neighbour species to all target species (row me ans of matrix) and the mean response of all nl!ighbour species to each targe l species (column means of matrix ) (Keddy er af. 1994). Species within each effect amI response matrix arc ranked with one com::sponding 10 Ihe species with the highesl co mpetitive performance (i .c. a neighbour species with a low mean competi ti ve dfect value OJ' high mean co mpetitive response value) .
Statistical results werc obtained with the aid of the STATGRAPHICS computer program (STATGRAPHICS 6.0 1992, Inc. USA.)

Results and Di s cussion Relative yield per plant values for competitive effect and response
The RYP values for all the pairwise combinations were used to establish one effect and one response matrix for each nutri en t level (Table 1). A muitifac[Or analysis of variance showed no significant difference in effect RYP values between the two nutrie nt le ve ls, however, a very highly significant difference (P < 0.001) between targe t species and between neighbours was found ( Table  2). Effect RYP values between target species differed significantly between D. S;lIItatfl and C. lichtensteini;, betwee n D. sinuala and H. variabilis. and between U. cakile/o/;(l and G. lichtensteinii. No significant difference in competitive effect or re sponse could be demon strated between Dimorphorheca sinlfata and Ursillia cnkilejofia within a nutrient level. Local fa rme rs maintain that at high nutri e nt leve ls D. silllfata is fa voured whereas U. cakifejolia has an advantage at low nutrient leve ls.
Although not significant, results in Table I lend support to this theo ry by indicating that D. sill uara was the stronge r e ffect competitor at the hig h nunie nl leve l while U. cakile/ olia was the stronger competitor at the low nutri ent level.
In the case of the response RYP values, a multi factor ANOYA showed a highly significant difference (P < 0.001) between target species and between neighbour species and a signilicant difference (P < 0.05) between the different nutrient levels (Table 2).
Be tween the target species, competitive response of D. sinU(l1n differed significantly from the three other target species. At the low nutrient level the measured response (RYP) of all four target species was less than at the high nutrient level. Competitive effects of the ten species were significantly correlaled (P < 0.05) among the two nutrient levels ( Figure I). Mean RYP at a high nutrient level Figure 1 Correlation between relative competitive effect fo r len Namaqualand pioneer plant species at two nutrient levels.
Similarly, there was a signifi cant correlation between the competitive responses ut the two nutrient levels (P < 0.05, Figure 2).
Thus nutrient level does not affect the status of the species i.e. the strong competitors at the low nutrient level were also the strong compcli(Ors at the hi gh nutrient leve l and the same applies for the weaker competitors. In their study on wetland plants, Keddy et at. (1994) found competitive effeci to be significantly correlated wheras co mpetiti ve response was not. At the high nutrient level competitive effect and response were not significantly (P > 0.05) negativel y correlated (Figure 3), whereas the negative correlation was significant (P < 0.05) at the lower nutrient level (Figure 4). Therefore, at a low nutrient level a strong effect competitor is a weak respon se competitor and vice versa although thi s is not the case at the high nutrie nt level. Non -sign ificant correlations bdween competitive effect and response were reported by Keddy ef al. (1994) and Goldberg and Landa (1991), although other ex periments have revealed different results: a posi tive relationship was found by Goldberg and Fleetwood (1987) and a negalive re lationsh ip was found by Miller and Werner (1987).

Competitive hierarchy
Kendall's rank correlation coefficient indicated that the mean competitive effect and response rankings of each target species  between th e two nutrient levels were concordant (P < 0.001, Table 3), The mean effec t and response of all neighbours on each target species. across the two lreatmenls were also concordant (P < 0.05, Tabl e 3). When separated into target species, the rankings were concordant among four targets across the two nutrient leve ls (P < 0.05, Tabl e 3).
As was the case in this study, Goldberg and Landa (199 1) found hierarchies of competitive effect to be highly concordant amon g neighbour species, suggesting that rankings of competitive effects are independent of the target species. The hierarchy found in this study agrees with the hierarchies produced in other stud ies on Namaqualand pioneer plant species (Oosthuizen 1996a, I 996b;Rosch 1997a;Rosch 1997b). Harper ( 1977) states (hat competit ive hierarchies are consistent and Keddy et al. ( 1994) have also shown that co mpetitive effec t hierarchies are unaffec ted by soil fertility. Because of the con sistency of competiti ve effect hierarchies Keddy er af. (1994) have suggested concentrating on determining which traits enable some plants to compete better than others. This was done in a study of fifteen Namaqualand pio neer plant species by Rosch er al. ( 1997b). It was found that the traits best related to competitive effect ability were all size related indicating that the larger the plant the stronger it acts as a co mpetitor (Rosch el al. 1997b). However, several studies have shown that compet itive hiera rchies change over time and within the same environment (Connolly er (/1. 1990;Menchaca & Conn Olly 1990), and therefore 213 which traits determine competitive ability mu st depend on factors such as relative sizes or stages of the life cycle in the co mpeting plants (Go ldberg & Landa 199 1). Because a large component of dep letion ab ility (co mpetitive ability) is total biomass or surface area of resource acquiring organs, per plant effects shou ld be strongly re lated to plant size and species shou ld be more sim ilar in competitive effec t on a per-unit size bas is than on a per-individual basi s (Goldberg & Werne r 1983) .
Competitive response in thi s stud y showed the same pattern as competi tive effect Le. perfect agree men t between treatmems. Goldberg and Landa (1991 ) found that hi erarchies in competitive response among target species were si milar regardless of neighbour species. Positions in competiti ve response hierarchies should depend on either relati ve abi lili es to tolerate depleted resource levels due to the presence of neighbo urs or relative abilit ies to avoid experiencing depicted resource leve ls because of pre-emption of resources from neighbours (Goldberg & Landa 1991). Which of these is more important should be related to relative sizes of targets and neighbours (Goldberg 1990) .
In contrast, Keddy et al. ( 1994) found that response rankings were not concordant across env ironments when the rankings were based on all three indicator species and there was no concordance across the environments for any of the species analyzed se parately.
According to the Kendall's rank correlation values competitive effect and response makings (usi ng mean effect and response on alilarget species) for both nutrient levels are in perfec t disagreement (not concordant) with one another (P < 0.001, Table 3). That is. if a species is a good effect competitor it is also a weak response competitor. Across targets. co mpetitive effect and response (in both treatments) are also in perfect di sagreement (P < 0.05, Table 3). Results by Keddy er al. (1994), however, indicate that one cannot generalize from competitive effect to competitive response.
In some studies it has been found (hat the choice of indicator species had no effect on the results (Gaudet & Keddy 1988). Others have found that the choice of indi cator species affects the magnitude of competition (Wilson & Keddy 1986;DiTommasio & Aarssen 1989;Wilson 1993), the relative importance o f below and above ground competition (Putz & Canham 1992), and the importance of competition (Reader & Bonser 1993). Keddy er af. (1994) suggest that when choosing target species it is probably best to avoid both strong and weak co mpetitors, since thi s tends to produce many species with similar competitive performances. A species of intermediate competitive performance may be the best choice as it will produce the best spread of relative competitive performances (Keddy er af. 1994). In this case the use of  species with a range of competitive abilities produced effect and response hierarchies that were consistent amo ng the four species and the two treatments thus the choice of target species did not affect the resulting hierarchies.
Competition intensity Mean competition intensity for each pairwise int eraction showed no significan t difference between target species (Table 4). Howeve r, between nutrient levels and neighbours there was a significant difference in competition intensity (P < 0.05, Table 4).
Competition intensity was greater at the low nutrient level than at the high nutrient leve l. However, it was found by other authors (Campbell & Grime 1992; Wil son & Tilman 1993) that competition intensity does not vary with nitrogen addition.

Conclusion
Individual competitive ability can be compared between species in two di stinct ways: in their co mpetitive effect or ability to suppress other individuals and in their competitive response or ability to avoid being suppressed (Goldberg & Landa 1991). Relative yield per plant (RYP) values for competitive effect differed significantly between target species and between neighbour species but not between nutrient levels. In contrast RYP values for competitive response differed significantly between target species and neighbour species as well as between nutrient levels.  Afr. 1. Bol. 1997, 63(4) Competitive response therefore seemed more sensitive to nutrient levels than competitive effect. This study concluded that co mpetitive effect hie rarchies as well as competiti ve re sponse hierarchi es across two nutrient levels and between targets within a nutrient treatment are concordant. Competitive effect and respo nse hierarch ies with in nutrient leve ls were found to be in perfect di sagreement. Competition between Namaqualand ephemeral plant species is such that the hierarchy is unaffected by the choice of target species :.!Od unaffected by nutrient level i.e. soil fertility will not l:hangc a species' ranking. Competitive intensity for each pairwise interaction showed no significant difference within target species, however differences were found between nutrient levels and neighbou r species.
Extrapolating from experimental res uhs to field conditions should always be done with caution, since there are many more factors interacting in the field. It can however, be assumed that the status of these Namaqualand ephemeral plant species is not affected by nutrient level and a strong competitor at a low soil fertility should remain a strong competitor at a high soil fertility. At a low soil fertility a strong effect competitor also acts as a weak response competitor, while this is not the case at a high soil fertility. Competition intensity is stronger at the low nutrient level and species show less competitive response than at the high nutrient level Le. their ability to be suppressed by neighbours is reduced.