Opposing effects of floral visitors and soil conditions on the determinants of competitive outcomes maintain species diversity in heterogeneous landscapes

Theory argues that both soil conditions and aboveground trophic interactions are equally important for determining plant species diversity. However, it remains unexplored how they modify the niche differences that stabilise species coexistence and the average fitness differences driving competitive dominance. We conducted a field study in Mediterranean annual grasslands to parameterise population models of six competing plant species. Spatially explicit floral visitor assemblages and soil salinity variation were characterized for each species. Both floral visitors and soil salinity modified species population dynamics via direct changes in seed production and indirect changes in competitive responses. Although the magnitude and sign of these changes were species specific, floral visitors promoted coexistence at neighbourhood scales while soil salinity did so over larger scales by changing the superior competitor's identity. Our results show how below and aboveground interactions maintain diversity in heterogeneous landscapes through their opposing effects on the determinants of competitive outcomes.


Introduction
where N i,t+1 N i,t is the per capita population rate, and N i,t is the number of seeds of species i in the soil 136 prior to germination in winter of year t. The germination rate of species i, g i , can be viewed as a 137 weighting term for an average of two di erent growth rates: the annual survival of ungerminated seed 138 in the soil (s i ), and the viable seeds produced per germinated individual (F i ). In past work, F i , was 139 expanded into a function describing how the average fecundity of each germinated seed that becomes 140 an adult (i.e. per germinant fecundity) declines with the density of competing number individuals in 141 the system (Godoy & Levine 2014). Now, we slightly modify this function to include the additional 142 e ect of floral visitors and soil conditions on the per germinant fecundity as follows: (2) 145 where ◊ i,s and " i,f v control the e ect of soil salinity (S t ) and floral visitors (A t ) respectively on the per 146 germinant fecundity of species i in the absence of competition (⁄ i ). In addition, ⁄ i is modified by the 147 germinated densities of other species including its own (g j N j,t ). To describe the per capita e ect that 148 species j is mediating on species i, we multiplied these germinated densities by a sum of three 149 interaction coe cients (ij +Â ij,s +Ê ij,f v ), which describes the additional direct e ect of soil salinity 150 and the apparent e ect of floral visitors on the competitive interactions between species. Notice that 151 we considered only explicitly in our study the e ect that soil salinity and floral visitors have on species' 152 fecundity (F i ), but the model could be easily extended to include the e ect of these two factors on the 153 other two vital rates, germination (g i ) and seed soil survival (s i ). 154 With the direct and apparent dynamics of competition described by this population model, we followed 155 the approach of Chesson (2012) to determine fitness and niche di erences between species pairs. Our 156 procedure here parallels previous work described in Godoy & Levine (2014), and allows us to define  As an opposing force to stabilising niche di erences, average fitness di erences drive competitive 171 dominance, and in the absence of niche di erences, determine the competitive superior between a pair 172 of species. Addressing the modifications done in the annual population model described by eqns (1)   173 and (2) to include the e ect of floral visitors and soil conditions, we define average fitness di erences 174 between the competitors ( k j k i ) as: When the ratio k j k i >1 this indicates that species j has a fitness advantage over species i. Both soil 181 salinity and floral visitors can be seen as equalising mechanisms promoting coexistence because they 182 can reduce fitness di erences between a species pair by two contrasted pathways. They can modify the 183 'demographic ratio' ( ÷ j ≠1 ÷ i ≠1 ) which describes the degree to which species j produces more seeds 184 (g j ⁄ j (1 + ◊ j,s S t + " j,f v A t )) per seed loss due to death or germination (1-(1-g j )s j ) than species i, and 185 they can also modify the 'competitive response ratio' ( describes the degree to which species j is less sensitive to competition than species i (eqn (4)). Notice 187 that these modifications can produce the opposing e ect and promote species' competitive dominance 188 by a combination of high demographic rates and low sensitivity to competition.

189
Competitors can coexist when niche di erences overcome fitness di erences, allowing both species to distinguish which of two following models were the best fit for our observations for each target species i. 245 The first model (model 1) assumes that competitive interactions between species are pairwise specific 246 but the e ects of salt and floral visitors on competitive interactions are common across species.
The second model (model 2) assumes that competitive interactions between species are pairwise 249 specific, as are the e ects of salt and floral visitors on competitive interactions.
It is also likely that soil salinity and floral visitors may not be a ecting the competitive dynamics 252 between species pairs. Therefore, we evaluated an additional model (model 3) that did not account for 253 these abiotic and biotic factors.
For all three models, individuals of the 10 species surveyed apart from our six focal species were 256 grouped together and their competitive e ect on the six focal species was summarized as a single 257 parameter. The average viable seed production per species (F i ) was estimated by counting the number  these species. While higher number of visits to C. fuscatum increased its potential fecundity and 282 reduced the negative e ect of both intra and interspecific competition on seed production, the opposite 283 pattern was observed for L. maroccanus and P. paludosa ( Fig. 2 and 3). Soil salinity, in contrast, had a 284 similar e ect across species increasing seed production in the absence of neighbours and promoting 285 weaker competitive interactions. For the other three non Asteraceae species, AIC values did not help to 286 distinguish unambiguously whether soil salinity and floral visitors had a common e ect on ⁄ i andij 287 (i.e. di erences in AIC between model one and three were lower than 10). In neither case, model 288 selection did support the view that the e ect of floral visitors and soil salinity on the species' responses 289 to competition was pairwise specific (i.e. model two showed consistently higher AIC values) (Appendix 290 S4). 291 the determinants of competitive outcomes in opposite and specific directions (Fig. 4). While floral 293 visitors tend to maintain stable coexistence (5 out of 15 species pairs) or to promote coexistence by 294 equalising fitness di erences (5 out of 15 species pairs moved closer to the coexistence region), soil 295 salinity tend to promote competitive exclusion (4 species pairs moved out of the coexistence region) 296 and increase competitive asymmetries between species pairs. As a result, floral visitors reduced on 297 average the niche di erences needed for coexistence (estimated from the mutual invasibility, eqn (5)) 298 across species pairs (paired t-test, t = 2.15, P = 0.049), while soil salinity increased significantly the 299 niche di erences needed for coexistence (paired t-test, t = 5.51, P < 0.001). Although soil salinity 300 reduced the likelihood of species coexistence at neighbourhood scales for all except one species pair, 301 this abiotic factor also determined changes in the identity of competitive winners (6 out of 15 species 302 pairs), which suggest that soil salinity can promote species coexistence over larger scales by turnover of 303 the dominant competitor (Fig. 4). At the neighbourhood scale, floral visitors consistently promoted species coexistence by reducing the 320 niche di erences needed to overcome fitness di erences between species pairs (Fig. 4) Francisco Rodriguez-Sanchez for his useful tips and templates to develop all this work in markdown. 405 We thank Curro Molina and Oscar Aguado for their help in field work and identifying insect species. 406 We also thank the Doñana NP sta for granting access to Caracoles Ranch, and Manu Saunders for her 407 manuscript review.   Table 1