Journal of Vegetation Science 19: 417-424, 2008 doi: 10.3170/2008-8-18384, published online 14 March 2008 © IAVS; Opulus Press Uppsala. - Testing the intermediate disturbance hypothesis in species-poor systems - 417 Testing the intermediate disturbance hypothesis in species-poor systems: A simulation experiment for mangrove forests Piou, Cyril1,2*; Berger, Uta1,2,3; Hildenbrandt, Hanno1,4 & Feller, Ilka C.5 1Center for Tropical Marine Ecology, Fahrenheitstrasse 6, 23859 Bremen, Germany; address: Institute of Forest Growth and Forest Computer Sciences, Technical University Dresden, Postfach 1117, 01735 Tharandt, Germany; 3E-mail uta.berger@forst.tu-dresden.de 4Present address: Theoretical Biology, Centre for Ecological and Evolutionary Studies, University of Groningen, Biological Centre, Kerklaan 30, 9751 NN Haren, The Netherlands; E-mail h.hildenbrandt@rug.nl; 5Smithsonian Environmental Research Center, PO Box 28, Edgewater, MD 21037, USA; E-mail felleri@si.edu; * Corresponding author; E-mail cyril.piou@forst.tu-dresden.de 2Present Abstract Questions: What factors inluence tree species diversity of mangrove forests, an example of species-poor systems? What are the respective importance and interactions of these factors? Is the intermediate disturbance hypothesis applicable to such systems? Methods: We used the spatially explicit individual-based model KiWi to investigate the effects on species diversity of perturbation frequency and intensity, different abiotic conditions, and interspeciic competition simulated at the individual level. The simulation system considered the three dominant Caribbean mangrove species: Rhizophora mangle, Avicennia germinans and Laguncularia racemosa, applying species-speciic growth and mortality characteristics. Firstly, effects on species dominance of the abiotic conditions nutrient availability and porewater salinity were tested with two competition scenarios. Secondly, the effect of perturbation frequency and intensity were investigated with selected abiotic conditions. Results: Abiotic conditions inluenced species dominance and, in extreme cases, excluded one or two species. Abiotic and competition settings controlled the successional dynamics and the response of species dominance to perturbation regimes. A response consistent with the intermediate disturbance hypothesis was observed only with a coniguration of plant interaction in which one species behaved as a pioneer so that succession occurred by competitive exclusion. Conclusions: We suggest that successional dynamics interact with the intensity and timing of perturbations and determine whether or not mangrove tree diversity conforms to predictions of the intermediate disturbance hypothesis. For mangroves, these successional dynamics are site-speciic depending on abiotic conditions and species conigurations. Keywords: Individual-based modeling; Interspeciic competition; KiWi model; Perturbation regime; Species dominance; Succession. Abbreviations: FON = Field of Neighbourhood; IDH = Intermediate disturbance hypothesis; ISDH = Index of species dominance heterogeneity; psu = Practical salinity units; RNA = Relative nutrient availability. Introduction For several decades, plant ecologists have tried to understand the processes implicated in variations in species diversity (e.g. Chust et al. 2006 see reviews by Loreau et al. 2001; Barot & Gignoux 2004; Vellend & Geber 2005). Among these processes, perturbations have been considered of high importance and have led to an ongoing debate on the intermediate disturbance hypothesis, which states that species richness is maximized at intermediate levels of disturbance (Grime 1973; Connell 1978; see reviews by Mackey & Currie 2001; Sheil & Burslem 2003; Shea et al. 2004). The situation of mangroves along tropical coastlines favours potential damage by major destruction events such as hurricanes or tropical storms (Imbert et al. 1998). Smith & Duke (1987) addressed the question of disturbance effects on mangrove tree diversity in Northern Australia. They showed that species richness decreased with increasing hurricane frequency. However, very few studies have analysed changes in mangrove species composition in relation to perturbation regime (Baldwin et al. 2001; Piou et al. 2006), and none have evaluated the implicated processes behind these effects. A straightforward explanation for this lack of consideration is the low number of tree species on mangrove systems. For example, in the Caribbean region, which is a hot spot of hurricane activity, only three to four true mangrove species are found. Thus, studies on tree species diversity are mostly seen as superluous in this system. However, considering species diversity as an expression of species richness and evenness (Kempton 1979), systems with only three species could also vary in species diversity. Piou et al. (2006) used an adaptation of the Simpson’s reciprocal index of species diversity (Simpson 1949; Hill 1973) to determine that the destruction intensity at different mangrove sites in Belize had an effect on the heterogeneity of species dominance. Although the patterns in Belize differed from other situations (e.g. Baldwin et al. 2001), it indicated 418 Piou, C. et al. that the effects of large destruction on species diversity also exist for species-poor mangrove systems. Based on these indings, we chose to use the Caribbean mangrove system as an example for analysing factors and processes inluencing species diversity in species-poor systems. Our irst hypothesis was that the succession of species dominance depends on both the interspeciic competition coniguration and the abiotic conditions. The second and more general hypothesis was that the resulting succession trajectories determine the type of response of the system to perturbations, and the eventual production of a bell-shape pattern of species diversity with intermediate perturbation regime. To test these hypotheses, we investigated the effects on species diversity of perturbation frequency, perturbation intensity, different abiotic conditions, and interspeciic competition by means of simulation experiments with an individual-based model. Methods KiWi model: General settings The experiments were carried out with the spatially explicit mangrove model KiWi (Berger & Hildenbrandt 2000, 2003), developed as dynamic library software written in C++ and using an interface in Microsoft ® Visual Basic ® (DLL and examples available from the corresponding author). The KiWi model describes resource competition on the level of individuals and simulated growth of mangrove stands composed of the three main Caribbean species, Rhizophora mangle, Avicennia germinans and Laguncularia racemosa. The gap model FORMAN (Chen & Twilley 1998) provided the growth formulas, multipliers for nutrient and salinity effects and respective parameters. It is important to note that the KiWi model is not a gap model since it describes trees individually and is spatially explicit. We used Berger & Hildenbrandt’s (2000) innovation of the ield of neighbourhood (FON) approach, which simulated inter-individual competition for space and resources. We assumed that the FON described the area where a tree inluenced its neighbours and was inluenced by them by sharing limiting resources such as light or nutrients. The FON was deined as a circular intensity ield that decreased from the center (stem position) out to the boundary. It speciied the intensity of competition exerted by a tree at any position within its neighbourhood. The growth of each individual tree was calculated with the following formula (Berger & Hildenbrandt 2000):  DBH × H  G × DBH ×  1 −  DBH max × H max  ∆DBH = × fs (SALT ) × fn ( RNA) × fc ( FA ) (1) 274 + 3 × b2 × DBH − 4 × b3 × DBH 2 ∆t where: DBH was the stem diameter at breast height (cm); H was tree height (cm); DBHmax and Hmax were maximum values of diameter and height for a given tree species; G, b2 and b3 were species-speciic growth constants and the three f-functions were growth multipliers (see App. 1 for details or Chen & Twilley 1998). The growth multipliers fs(SALT) and fn(RNA) considered the effects of the porewater salinity and relative nutrient availability, respectively (Chen & Twilley 1998). The function fc(FA) was the growth multiplier for the FON effect on growth (Berger & Hildenbrandt 2000):  1  fc ( FA ) = max {0 ; 1 − ϕ × FA } = max 0 ; 1 − ϕ ×  ∑ ∫ FON n ( x, y ) do   A n≠ k O    (2) where: ϕ was an arbitrary maximum value of effect of competition simulating resource sharing capacity, A was the FON area of the focus tree k, n were the neighbours of k, belonging to the focal and neighbour tree n, and the FONn function was the intensity of competition of the neighbour n at each point of O. The FON function was calculated as: for 0 ≤ r < RBH 1     ln ( Fminn )   FON n ( r ) = exp  −   × ( r − RBH ) for RBH ≤ r ≤ R  R RBH −       for R < r 0 (3) where: RBH was the radius of the stem at breast height of n, r was a distance from the stem position of n, and Fmin was the minimum intensity of the FON (0.1, Berger & Hildenbrandt 2000) at the FON radius (R). This FON radius (R) depended on the size of the tree: (4) R = a × RBH b where: a and b were scaling parameters (cf. ‘setting interspeciic competition’ and App. 2). The value of b determined inversely the competition intensity of individuals (see Berger & Hildenbrandt 2003 for variations of model behavior depending on these two parameters). According to the assumption of Chen & Twilley (1998), an overall availability of recruits was considered as RNtot = 18 saplings.100m–2.yr–1 (Chen & Twilley 1998). However, the annual number of recruits varied randomly, and the number from each species (RNi) was proportionally set according to the occurrence of mature trees (height >5m) of each species: RN i = int ( rnd1 × pi × RN tot + rnd2 × RN tot ) (5) where: rnd1 was a uniform random number between 0.5 and 1.5; rnd2 a uniform random number between 0.1 and 0.3; and pi the proportion of mature trees of species i over the total number of mature trees in the plot. The range of variation of rnd1 was chosen to describe a natural luctuation in the availability of recruits per species (+/-50%). The range of variation of the rnd2 provided an occasional reappearance of an already excluded species. These two - Testing the intermediate disturbance hypothesis in species-poor systems ranges of variations were set arbitrarily, but a sensitivity analysis showed a low effect of these parameters on the main results (App. 3). The recruits were installed randomly on the simulation area, but were removed if the FON intensity (sum of FON(x,y) of all trees at the point of installation x, y) was higher than the species-speciic threshold (FAmax). This threshold was set to FAmax = 0.5 for R. mangle (Berger & Hildenbrandt 2000) and assumed as FAmax = 0.0 for the two other species to simulate the shade intolerance of seedlings of L. racemosa and A. germinans (Ball 1980; McKee 1993). Mortality of individual trees not due to external perturbations was growth-rate dependent as described by Berger & Hildenbrandt (2000). Settings of interspeciic competition parameters The growth parameters and effects of salinity and nutrient availability (DBHmax, Hmax, G, b2, b3, fs and fn, Eq. A1.1, see App. 1) created species-speciic differences in growth response at the stand level. For additional variation in interspeciic competition, we considered two ways of simulating spatial competition at the individual level. The irst considered an equal effect of neighbouring competition for trees of the same size for the three species. Thus, they had the same resource sharing tolerance ( ϕ = 2.000, Berger & Hildenbrandt 2000) and identical a and b parameters (11.0, 0.64, respectively, cf. App. 2, Fig. A2.1). Since the interspeciic competition in this parameterization was only through the relative growth rate of each species, it is hereafter referred to as species homogeneous spatial competition. The second parameterization considered that each species had spatially-speciic competition strength. Particularly, L. racemosa, which was described as heliophilic (Wadsworth 1959; Ball 1980; Roth 1992) was set to have a lower sharing tolerance ( ϕ = 2.222, assuming that the maximum FA was 10% lower than the other species, i.e., maximum FA = 0.45). Additionally, species-speciic a and b parameters (App. 1, Table A1.1) were used to describe the canopy and rootsystem differences for the three species. These parameters were tuned (App. 2, Fig. A2.1) to reproduce ield data of monospeciic stands of tree size / density relationships from Belizean offshore mangroves, and to set L. racemosa as less competitive than the two other species. This lower competition capacity of L. racemosa was at DBH < 80 cm; while the a and b values also determined that A. germinans was more competitive than R. mangle at DBH > 20 cm. This second parameterization is hereafter referred to as species heterogeneous spatial competition. Effects of abiotic conditions Our irst exercise was set to analyse the effect of abiotic conditions on species diversity without any perturbations. We also investigated the effect of interspeciic 419 competition on succession of species dominance in this exercise. Five salinities (0, 20, 40, 50 and 60 psu) and four relative nutrient availabilities (RNA) (100%, 80%, 60% and 40%) were considered. Ten replicates of all possible salinity/RNA scenarios on the two competition parameterizations were simulated on a 6000-m² plot and over 1000 years. The number of trees and basal area per species were used to calculate relative abundance and dominance for each time step and transformed into importance values (IV) according to Cintrón & Schaeffer-Novelli (1984): IVi = 100 × Densi q ∑ Dens j + 100 × BAi q ∑ BA j =1 j (6) j =1 where: IVi , BAi and Densi were the importance values, basal area and density of trees of the species i, and q was the number of species. As a measure of species diversity, we used the index of species dominance heterogeneity (ISDH) from Piou et al. (2006). It was adapted from the reciprocal index of Simpson (Hill 1973) and computed as follows:  q  q ∑ IV ×  ∑ IV − 1 i i I SDH = i =1 q i =1 ∑ ( IV × ( IV − 1)) i (7) i i =1 This index indicated relative species dominance in our three-species system and was given a value of 0 if no trees could grow because of harsh abiotic settings. If trees could grow, the ISDH were given values from 1 (only one species present on the plot) to 3 (the three species representing each 33% of importance on the plot). Since this index was not mathematically independent from species richness, we decided not to use the term ‘evenness’ to avoid confusion with its calculations in community studies (Smith & Wilson 1996). However, through the variation of relative species dominance, this index could indicate if different threespecies conigurations of our system were relatively rich or not in term of species diversity. As indicators of salinity/ RNA effect on species diversity, the median, 1st. and 3rd. quartiles of ISDH for each scenario over the last 400 yr of simulations were calculated. Effects of perturbation regimes The second exercise was set to analyse the effects of perturbation regimes on species diversity. Massive killing events, which simulated mortality induced by a tropical storm or hurricane, were applied at different mortality rates (intensity) and frequencies. Because there is inconsistency in the literature on the way authors described storm resistance capacity according to species or tree size (e.g. Vermeer 1963; Stoddart 1963; Bardsley 1984; Roth 420 Piou, C. et al. 1992, 1997; Smith et al. 1994; Imbert et al. 1998; Sherman & Fahey 2001; Baldwin et al. 2001; Imbert 2002), we could not consider the mortality events related to size or species in our simulations. The applied intensities were probabilities of 30%, 50%, 70%, 90% and 99% of mortality for each tree at the event times. The perturbation frequencies (1 /100 yr, 1 /80 yr, 1 /60 yr, 1 /40 yr, 1 /20 yr and 1 /10 yr) determined the exact number of years between two events. To achieve a stabilized system in term of number of trees, we excluded the irst 400 simulation years. Perturbations were applied only on the following 400 years so that the total simulated time was 800 years. The role of abiotic conditions on system response to perturbation was considered by selecting scenarios from the results of the previous exercise. Benign (salinity 0 psu and 100% RNA) and medium (salinity 50 psu and 80% RNA) conditions were analysed, but extreme ones were not considered because they resulted in a system overwhelmed by one species. Ten replicates were simulated for each abiotic scenario (benign or medium) for each competition parameterization (homogeneous or heterogeneous spatial competition) and all mortality rate/perturbation frequency scenarios. Similar to the previous exercise, the median of ISDH was calculated over the last 400 yr for all cases. To analyse the signiicance of perturbation intensity with selected perturbation frequency, Kruskal-Wallis non-parametric analysis of variance (ANOVA) by ranks were applied on the last ISDH values of each simulation. To analyse the effect of perturbation frequency with selected perturbation intensity, identical non-parametric ANOVAs were done considering all the ISDH values of the simulated perturbation time. Mann-Whitney U tests were used to assess signiicant differences of extremes and intermediate ISDH values in order to validate disturbance effect patterns such as U-shaped, linear increase or decrease, irregular or bell-shaped. Results First exercise: effects of abiotic conditions Extremely low relative nutrient availabilities (40% RNA) and extremely high salinities (60 psu) decreased signiicantly the index of species dominance heterogeneity (ISDH) for both spatial competition parameterizations (App. 4, Fig. A4.1). These extreme abiotic conditions caused species exclusion through the parameterization of R. mangle and A. germinans growth characteristics to be non-adapted to high salinities and low nutrient availabilities, respectively. At the worst condition (salinity 60 psu and 40% RNA), no species grew at all, resulting in ISDH = 0. Considering the rest of the abiotic scenarios, Fig. 1. Dynamical variations of the two competition parameterizations with selected abiotic scenarios (medium = Salinity 50 psu and RNA 80%) in species relative importance (IV) and ISDH (thin lines = respective irst and third quartiles). highest ISDH values in both spatial competition parameterizations were found at intermediate levels of salinity and RNA. The median ISDH values over the last 400yr were relatively similar between the two spatial competition parameterizations. However, ISDH and species importance values varied more importantly during the irst 400yr for all abiotic scenarios. For medium abiotic scenario (e.g., salinity 50 psu and 80% RNA), variations of species importance values showed a cycling of species dominance (Fig. 1). The species heterogeneous spatial competition parameterization created a quick succession from L. racemosa to A. germinans (Fig. 1b) during the irst 50yr of simulations. With homogeneous spatial competition, the dominance of L. racemosa varied but stayed always higher than the two other species (Fig. 1a). Identically, for benign abiotic scenarios, the species heterogeneous spatial competition created species succession, while the homogeneous spatial competition showed importance values variation without shift of species dominating. - Testing the intermediate disturbance hypothesis in species-poor systems - Fig. 2. Dynamical variations in species relative importance (IV) and ISDH for the heterospeciic competition parameterization and medium case of abiotic scenario (salinity 50psu, 80% RNA), for different perturbation regimes. a. frequency= 1 / 100 yr, intensity = 30% mortality; b. frequency = 1 / 60 yr, intensity = 70% mortality; c. frequency = 1 / 10 yr intensity = 99% mortality). (thin lines = respective irst and third quartiles). Second exercise: effects of perturbation regimes For the analysis of phenomena explaining the response pattern, we concentrated only on the medium abiotic scenario. Massive mortality altered the temporal dynamic of ISDH (Fig. 2). The low perturbation regime (Fig. 2a) did not modify the general trend of variation of species importance values and ISDH compared to non-disturbed dynamics (Fig. 1b). With an intermediate perturbation regime (more frequent and stronger disturbances, Fig. 2b compared to 2a), L. racemosa gained in importance although still less important than the two other species. This reduced the difference in relative importance of the three species and thus led to an overall higher ISDH than with the low perturbation regime. The most frequent and destructive perturbation regime (Fig. 2c) switched the system quickly from A. germinans to L. racemosa dominance. At this level of perturbation regime, each disturbance had an effect of 421 Fig. 3. Median ISDH variations according to perturbation frequency and intensity for the two competition parameterizations (a and b) with medium abiotic scenario (Salinity 50 psu and 80% RNA). Dashed lines represent selected pattern illustration for Fig. A4.2-2 (App. 4). keeping L. racemosa as the most important species on the plot. This corresponds to the original succession situation at the beginning (irst 10 years) of the simulations, as if L. racemosa were the pioneer species of the system. However, this change of dominance did not modify signiicantly the ISDH values compared to low or absent perturbations because the ratios of species importance values were conserved. In this case, the high frequencies stabilized these ratios and ISDH values over time. Variations in perturbation regimes always had an effect on the species dominance heterogeneity of the simulated stands (Fig. 3). However, the overall patterns of simulation results depended on the different competition parameterization. The species homogeneous spatial competition parameterization (Fig. 3a) showed lower ISDH values at intermediate perturbation regimes than at lower and higher perturbation frequencies and intensities. This U-shaped curve pattern was clearly observable as to the inluence of frequency regime with a selected perturbation intensity (following line on Fig. 3a or App. 4, Fig. A4.2a), although the values showed high variation among simulations (1st. and 3rd. quartile variations). The species heterogeneous spatial competition parameterization resulted in an overall increase in ISDH values with increasing disturbance regimes until non-extreme intensity and frequency followed then by a small decrease (Fig. 3b). Thus, this trend led to an 422 Piou, C. et al. overall bell-shaped pattern, which was also more visible for the inluence of frequency regime with selected perturbation intensity (following line on Fig. 3b or App. 4, Fig. A4.2b). Analysing the patterns for all scenarios, we found a clear difference of patterns between the two spatial competition parameterizations. The homogeneous spatial competition parameterization led to some cases of U-shaped patterns while the heterogeneous spatial competition on the contrary showed bell-shaped patterns. However, not all perturbation regimes led to these U- or bell-shaped patterns, but also included cases of linear increasing or decreasing patterns or even non-signiicant or irregular patterns. These trends were repeated with the two selected abiotic conditions (cf. App. 4, Fig. A4.3 for detailed results). Discussion This study illustrated that even for species-poor systems, the dynamics and the processes that could explain variations in species diversity are diverse and interconnected. The interplay of abiotic conditions and interspeciic competition produces a set of potential vegetation dynamics. Depending on the perturbation regime, a system will follow a particular trajectory of this set, and eventually test the expectations of the intermediate disturbance hypothesis pattern. Our simulations integrate the actual knowledge on Caribbean mangrove species of species-speciic parameterization of growth, adaptations to abiotic conditions, settlement, and spatial competitive strength. The results of the irst exercise illustrate that abiotic conditions inluence the dominance distribution of these species, up to eventually excluding one or more species. On the contrary, intermediate conditions of porewater salinity and nutrient availability favorable to all three species lead to higher coexistence. The setting of species-speciic growth parameters of our model is thus able to re-create the diversity of species dominance observed in the Caribbean. Other factors that were not considered in this study, such as tidal regime, temperature, soil physico-chemical properties (e.g., redox potential or sulide contents), could have similar effects on species richness and dominance in mangrove systems (Ball 1980; McKee 1993). The results of the irst exercise also show that changes in the characteristics of species-speciic spatial competition do not modify signiicantly the overall measure of species diversity. However, at a given abiotic condition, a change in the settings for spatial competition drastically alters the temporal variations of relative species dominance. Our parameterization of homogeneous spatial competition leads to a cycling dynamic but with L. racemosa dominating all the time because of its faster growth rate. The hypothesis behind this parameterization is that species differ in their resource use capacity but not in a spatially explicit way. For example, trees of the same size would have the same spatial extent of resource use disregarding their species. In contrast, the heterogeneous spatial competition parameterization is derived from the hypothesis that individuals of L. racemosa are less competitive for spatially distributed resources than individuals of other species (Wadsworth 1959; Ball 1980; Roth 1992). The reduction of resource-sharing tolerance for the L. racemosa trees increased the effects of neighbours on their growth rates. Additionally, species-speciic changes in the FON radius inluenced species interactions by conferring lower competitive strength to L. racemosa individuals than equal-sized A. germinans or R. mangle trees. Thus, after the irst years of fast growth of L. racemosa trees this heterogeneous spatial competition parameterization produced a shift in dominance. Thus, this succession resulted from the switch in the importance of two forces: (a) the primary growth rate of L. racemosa, which is known to be faster than for the other species under low salinity conditions, high nutrient, and light availability (McKee 1995; Sherman et al. 1998; Lovelock & Feller 2003); and (b) the low strength of spatial interspeciic competition of L. racemosa (as hypothesized by Berger et al. 2006). These characteristics are typical of pioneer-like species in any plant system. In mangrove forests, such successions were described in some secondary recovery areas (Ball 1980; Berger et al. 2006), which suggests that our second spatial competition parameterization is supported by ield observations. These differences in the dynamics between the two parameterizations become especially important when considering the effects of perturbations. The simulations with perturbations illustrated that species dominance of our system depended on the frequency of the destruction events and their intensities. However, we have seen that the pattern of response changed mainly depending on the competition parameterization and thereby the successional dynamic. Perturbations created gaps that would take the same trajectory as the system’s dynamics observed without perturbations. For each gap recovery, the seedling availability depended on the dominant species in the rest of the stand. In the case of homogeneous spatial competition parameterization, if the system was perturbed each time when the majority of gaps were in the cycling phase of highest dominance of L. racemosa, the dominance of this species would increase more and more, as in a resonance phenomenon. This situation was created at intermediate perturbations regimes, leading to the lowest ISDH values. With extreme disturbance regime, the system would achieve the cycling phases earlier and, thus, would return to a more even species distribution. This scenario led to higher ISDH values, and overall created the observed Ushaped patterns. With the heterogeneous spatial competition - Testing the intermediate disturbance hypothesis in species-poor systems parameterization, perturbations caused the system to return reiteratively to conditions seen during the initial succession phases. Since L. racemosa was the most pioneer-like of the three species, it obtained higher importance with stronger and more frequent perturbations, which created a more homogeneous species dominance. Eventually, with extreme perturbation regimes, L. racemosa dominated completely, reducing the index of species dominance heterogeneity. In mangrove forests, it is therefore possible to observe the bell-shaped pattern typically described by the intermediate disturbance hypothesis (IDH) (Connell 1978) if we have a biotic coniguration where L. racemosa is pioneer and succession happens during stand recovery or establishment. However, in addition to bell-shaped or U-shaped patterns, our results also revealed many cases of linear increases or decreases due to perturbation regimes not itting exactly the resonance of the recovery dynamics. This diversity of responses to perturbation its the observations of Mackey & Currie (2001) and the prediction of the IDH axioms detailed by Sheil & Burslem (2003). Speciically, to have an IDH pattern one needs: (1) a dominance successional sequence when no perturbations occur; (2) succession due to competitive exclusion of fastest growing trees; and (3) perturbations that bring the system back to earlier successional stages. The results of our individual-based model simulating competition at individual-level conirm these axioms. The homogeneous competition parameterization of our study did not create succession and therefore did not exhibit a pattern predicted by the IDH. However, this dynamic is possible in nature (e.g. in understorey species systems as in Beckage & Stout 2000) and in mangrove ecosystems particularly. Only few studies have observed a real species succession in mangrove forests (e.g., Ball 1980; Berger et al. 2006). Lugo (1980) concluded that zonation was a steady state result of abiotic conditions and refuted Davis’ (1940) hypothesis that zonation was the result of succession and land building processes. Since Lugo’s paper, succession in mangroves has been cautiously attributed to changes in abiotic conditions because of external factors, but rarely to species-induced modiications of abiotic conditions (e.g., Bertrand 1999). Because the IDH pattern is the expression of the dynamics of species succession, it can be used to compare species succession at different disturbance levels, or conversely, to compare the recovery dynamics of sites that exhibit different succession dynamics. Both aspects have never been considered in mangrove ecology. Such studies could support our simulation results that in some cases succession could be due to plant-plant interactions and not always exclusively to changes in abiotic conditions. Finally, our study at the individual level demonstrates that even if abiotic conditions strongly inluence species composition in mangrove forests, spatial plant-plant interactions also play an important role. We showed that the 423 successional dynamic is dependent on the capacities of individuals of different species to compete spatially for resources, and that these dynamics determine the way species diversity will increase or decrease in case of perturbations. Thus, we demonstrate that variations of mangrove species diversity due to perturbation regime will depend on a series of interacting factors, including succession coniguration, actual dynamic phases, plant spatial interactions, and abiotic settings. Additionally, ield studies show that changes of abiotic settings after perturbations (e.g., Cahoon et al. 2003), recruitment patterns (e.g. Baldwin et al. 2001; Clarke 2004; Piou et al. 2006) and also differences of resistance of species to the considered perturbations (Baldwin et al. 2001; Imbert 2002) could inluence species composition of mangroves. Hence, forecasting a general trend of evolution of species diversity of mangrove forests only considering the perturbation regime seems risky. It could be possible only in a site-speciic case, knowing not only the abiotic conditions of a particular site, but also the type of species interactions and succession phenomenon that could occur. Acknowledgements. The authors are very grateful to Volker Grimm, Martha Liliana Fontalvo-Herazo, Elizaveta Pachepsky and an anonymous reviewer for valuable comments on an earlier version of this manuscript. This study was inanced in the frame of the MADAM project, a cooperation between the ZMT, Bremen, Germany and the UfPa and MPEG, both Belém, Brazil, inanced by the German Ministry of Education, Science, Research and Technology (BMBF) [MADAM – Mangrove Dynamics and Management (Project number: 03F0154A)], and the Conselho Nacional de Pesquisa e Tecnologia (CNPq). This is MADAM contribution number 116. References Baldwin, A., Egnotovich, M., Ford, M. & Platt, W. 2001. Regeneration in fringe mangrove forests damaged by Hurricane Andrew. Plant Ecology 157: 149-162. Ball, M.C. 1980. Patterns of secondary succession in a mangrove forest in southern Florida. Oecologia 44: 226-235. Bardsley, K. 1984. The effects of Cyclone Kathy on mangrove vegetation. In: Bardsley, K.N., Davie, J.D.S. & Woodroffe, C.D. (eds.) Coasts and tidal wetlands of the Australian monsoon region, pp. 167-185. North Australia Research Unit, ANU Press, Canberra, AU. Barot, S. & Gignoux, J. 2004. Mechanisms promoting plant coexistence: can all the proposed processes be reconciled? Oikos 106: 185-192. Beckage, B. & Stout, I.J. 2000. Effects of repeated burning on species richness in a Florida pine savanna: A test of the intermediate disturbance hypothesis. Journal of Vegetation Science 11: 113-122. Berger, U. & Hildenbrandt, H. 2000. A new approach to spatially explicit modelling of forest dynamics: spacing, ageing and neighbourhood competition of mangrove trees. Ecological Modelling 132: 287-302. 424 Piou, C. et al. Berger, U. & Hildenbrandt, H. 2003. The strength of competition among individual trees and the biomass-density trajectories of the cohort. Plant Ecology 167: 89-96. Berger, U., Adams, M. & Hildenbrandt, H. 2006. Secondary succession of neotropical mangroves: causes and consequences of growth reduction in pioneer species. Perspectives in Plant Ecology, Evolution and Systematics 7: 243-252. Bertrand, F. 1999. Mangrove dynamics in the Rivières du Sud area, West Africa: an ecogeographic approach. Hydrobiologia 413: 115-126. Cahoon, D.R., Hensel, P., Rybczyk, J., McKee, K., Profitt, C.E. & Perez, B.C. 2003. Mass tree mortality leads to mangrove peat collapse at Bay Islands, Honduras after Hurricane Mitch. Journal of Ecology 91: 1093-1105. Chen, R. & Twilley, R.R. 1998. A gap dynamic model of mangrove forest development along gradients of soil salinity and nutrient resources. Journal of Ecology 86: 37-51. Chust, G., Chave, J., Condit, R., Aguilar, S., Lao, S. & Pérez, R. 2006. Determinants and spatial modeling of tree β-diversity in a tropical forest landscape in Panama. Journal of Vegetation Science 17: 83-92. Cintrón, G. & Schaeffer-Novelli, Y. 1984. Methods for studying mangrove structure. In: Snedaker, S.C. & Snedaker J.G. (eds.) The mangrove ecosystem: research methods, pp. 91-113. UNESCO Monographs in Oceanographic Methodology 251. UNESCO, Paris, FR. Clarke, P.J. 2004. Effects of experimental canopy gaps on mangrove recruitment: lack of habitat partitioning may explain stand dominance. Journal of Ecology 92: 203-213. Connell, J.H. 1978. Diversity in tropical rain forests and coral reefs. Science 199: 1302-1310. Davis, J.H. 1940. The ecology and geologic role of mangroves in Florida. Publications of the Carnegie Institute. Washington, DC, US. Grime, J.P. 1973. Control of species density in herbaceous vegetation. Journal of Environmental Management 1: 151-167. Hill, M.O. 1973. Diversity and evenness: a unifying notation and its consequences. Ecology 54: 427-432. Imbert, D. 2002. Impact des ouragans sur la structure et la dynamique forestières dans les mangroves des Antilles. Bois et Forêts des Tropiques 273: 69-78. Imbert, D., Rousseau, A. & Labbé, P. 1998. Ouragans et diversité biologiques dans les forêts tropicales. L’exemple de la Guadeloupe. Acta Oecologica 19: 251-262. Kempton, R.A. 1979. The structure of species abundance and measurement of diversity. Biometrics 35: 307-321. Loreau, M., Naeem, S., Inchausti, P., Bengtsson, J., Grime, J. P., Hector, A., Hooper, D. U., Huston, M. A., Raffaelli, D., Schimd, B., Tilman, D. & Wardle, D. A. 2001. Biodiversity and ecosystem functioning: current knowledge and future challenges. Science 294: 804-808. Lovelock, C.E. & Feller, I.C. 2003. Photosynthetic performances and resource utilization of two mangrove species coexisting in a hypersaline scrub forest. Ecophysiology 134: 455-462. Lugo, A.E. 1980. Mangrove ecosystems: successional or steady state? Biotropica 12: 65-72. Mackey, R.L. & Currie, D.J. 2001. The diversity-disturbance relationship: is it generally strong and peaked? Ecology 82: 3479-3492. McKee, K.L. 1993. Soil physicochemical patterns and mangrove species distribution: reciprocal effects? Journal of Ecology 81: 477-487. McKee, K.L. 1995. Interspeciic variation in growth, biomass partitioning, and defensive characteristics of neotropical mangrove seedlings: Response to light and nutrient availability. American Journal of Botany 82: 299-307. Piou, C., Feller, I.C., Berger, U. & Chi, F. 2006. Zonation patterns of Belizean offshore mangrove forests 41 years after a catastrophic hurricane. Biotropica 38: 365-374. Roth, L.C. 1992. Hurricanes and mangrove regeneration: effects of Hurricane Joan, October 1988, on the vegetation of Isla del Venado, Blueields, Nicaragua. Biotropica 24: 375-384. Roth, L.C. 1997. Implications of periodic hurricane disturbance for the sustainable management of caribbean mangroves. In: Kjerve, B., Lacerda, L.D. de & Diop, E.H.S. (eds.) Mangrove ecosystem studies in Latin America and Africa. UNESCO & ISME, Paris, FR. Shea, K., Roxburgh, S.H. & Rauschert, E.S.J. 2004. Moving from pattern to process: coexistence mechanisms under intermediate disturbance regimes. Ecology Letters 7: 491-508. Sheil, D. & Burslem, D.F.R.P. 2003. Disturbing hypotheses in tropical forests. Trends in Ecology and Evolution 18: 18-26. Sherman, R.E. & Fahey, T.J. 2001. Hurricane impacts on a mangrove forest in the Dominican Republic: damage patterns and early recovery. Biotropica 33: 393-408. Sherman, R.E., Fahey, T.J. & Howarth, R.W. 1998. Soil-plant interactions in a neotropical mangrove forest: iron phosphorus and sulfur dynamics. Oecologia 115: 553-563. Simpson, E.H. 1949. Measurement of diversity. Nature 163: 668. Smith, B. & Wilson, J. 1996. A consumer’s guide to evenness indices. Oikos 76: 70-82. Smith III, T.J. & Duke, N.C. 1987. Physical determinants of inter estuary variation in mangrove species richness around the tropical coastline of Australia. Journal of Biogeography 14: 9-19. Smith III, T.J., Robblee, M.B., Wanless, H.R. & Doyle, T.W. 1994. Mangroves, hurricanes, and lightning strikes. Bioscience 44: 256-262. Stoddart, D.R. 1963. Effects of Hurricane Hattie on the British Honduras reefs and cays, Oct. 30-31, 1961. Atoll Research Bulletin 95. Vellend, M. & Geber, M.A. 2005. Connections between species diversity and genetic diversity. Ecology Letters 8: 767-781. Vermeer, D.E. 1963. Effects of Hurricane Hattie, 1961, on the cays of British Honduras. Zeitschrift für Geomorphologie 7: 332-354. Wadsworth, F.H. 1959. Growth and regeneration of white mangrove in Puerto Rico. Caribbean Forester 20: 59-71. Received 15 February 2007; Accepted 17 August 2007; Co-ordinating Editor: M. Pärtel. For App. 1-4, see below (online version) also available at JVS/AVS Electronic Archives; www.opuluspress.se/ I App. 1. Details on growth multipliers and parameters. In this appendix, we give the details of the growth multipliers and parameters (Table A1.1) entering in Eq. A1.1. The function fs(SALT) was the growth multiplier considering the effect of the pore water salinity on growth (Chen & Twilley 1998): 1 fs ( SALT ) = (A1.1) 1 + exp ( d × ( S0.5 − S )) where: S was the salinity at tree position and S0.5 and d were species speciic constants (Table A1.1). The function fn(RNA) was the growth multiplier considering the effect of the relative nutrient availability (RNA) on growth (Chen & Twilley 1998): fn ( RNA ) = c1 + c2 × RNA + c3 × RNA 2 (A1.2) where: c1, c2 and c3 were species speciic constants (Table A1.1). Table A1.1. Growth and spatial competition species-speciic parameters used in the KiWi model. Sources: (1) Chen & Twilley 1998, (2) see App. 2. Parameter Description DBHmax Hmax G b2 b3 d S0.5 c1 c2 c3 a b Maximum diameter at breast height Maximum height Growth constant Constant in height to dbh relationship Constant in height to dbh relationship Salinity effect constant Salinity effect constant RNA effect constant RNA effect constant RNA effect constant FON radius scaling parameter for heterospeciic competition parameterization FON radius scaling parameter for heterospeciic competition parameterization Apps. 1-4. Internet supplement to: Piou, C. et al. 2008. Testing the intermediate disturbance hypothesis in species-poor systems: A simulation experiment for mangrove forests Journal of Vegetation Science 19: 417-424; doi: 10.3170/2008-8-18384 A. germinans L. racemosa R. mangle 140 3500 162 48.04 0.172 -0.18 72.0 -0.50 2.88 -1.66 13.7 0.72 80 3000 243 71.58 0.447 -0.20 65.0 -1.00 4.42 -2.50 17.0 0.95 100 4000 267 77.26 0.396 -0.25 58.0 0.00 1.33 -0.72 18.0 0.83 (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (2) (2) II App. 2. Parameterization of the FON radius calculation. In the KiWi model, the FON radius R of a tree depends on its size: (A2.1) where: a and b are scaling parameters. The parameterization of a and b can be effectuated to reproduce the DBHdensity trajectories of a self-thinning phenomenon. R = a × RBH b Demonstration: In equation A2.1, the RBH is half the DBH, so A2.1 becomes: 1 R = a × b × DBH b (A2.2) 2 The FON approach has been seen as reproducing the self-thinning trajectory very well (Berger & Hildenbrandt 2003). During the self-thinning in KiWi model, because of the mortality function, the total FON area of all individuals can be considered as constant since the dead individuals are replaced by growth of the remnant. This corresponds to a constant maximum resource use. Let assume this constant be FONtot. We could simplify its calculation as: (A2.3) FONtot = N × FONind where FONind is the mean area of the FON area of the individuals deined as: 1 × DBH 2 b (A2.4) 22b where R and DBH are respective mean values assuming they represent the entire community. Assuming that during self-thinning we have the relationship of the DBH-density trajectory: FON ind = π × R 2 = π × a 2 × log ( N ) = α + β´log ( DBH ) (A2.5) or N = exp (α ) + DBH β Interchanging Equation A2.4 in A2.3 and comparing to A2.5 we get: N = exp (α ) + DBH β =  22b  FON tot × DBH −2 b = FON tot ×  FON ind  π × a 2  (A2.6)  22b  Since exp (α) and FON tot ×  are not dependents on DBH, we can link the β parameter directly to the FON  π × a 2  b parameter: β = –2b (A2.7) Identically we can derive the value of a: a= 2 2 b × FON tot π × exp (α ) (A2.8) We determined with the KiWi model that FONtot is constant ~215% and not depending on a nor b. These relationships are conirmed by simulation experiments with monospeciic stands (Fig. A2.1). Apps. 1-4. Internet supplement to: Piou, C. et al. 2008. Testing the intermediate disturbance hypothesis in species-poor systems: A simulation experiment for mangrove forests Journal of Vegetation Science 19: 417-424; doi: 10.3170/2008-8-18384 III Parameterization of species-speciic values Data from monospeciic stands of Belizean mangroves (I.C. Feller, F. Chi and C. Piou unpubl.) at different density were used to create regressions and calculate the parameters a and b for Rhizophora mangle and Avicennia germinans. Fig. A2.1 shows the ield data, linear regression and results of monospeciic simulation without recruitment with the corresponding FON a and b. Fig. A2.1. Field DBH-density (cm and stem/ha) data on natural logarithmic scale with corresponding linear regression (plain lines) and simulation results (dashed lines) of monospeciic stand of Rhizophora mangle (black) and Avicennia germinans (grey) without recruitment. Fig. A2.2. CARICOMP DBH-density (cm and stem/ha) data of mixed forests on natural logarithmic scale with corresponding linear regression (plain line). For Laguncularia racemosa, not enough monospeciic ield data were available, so we estimated that this species was less competitive in Belize in terms of spatially distributed resources such as light. This was then considered in the a and b parameter giving a larger b-value (0.95) and smaller a-value (17.0) than for R. mangle (e.g. Berger & Hildenbrandt 2003). Parameterization of species-identical values To use the same approach for the tuning of the a and b parameter in the irst parameterization (species homogeneous spatial competition), data of density and mean diameter from mixed stands of the three species were considered. We used the data from plots of the CARICOMP program (CARICOMP 2002, http://www.ccdc.org.jm/mangrove_data. html) over the entire Caribbean region to create the regression and calculate the parameters a and b considering all three species (Fig. A2.2). Apps. 1-4. Internet supplement to: Piou, C. et al. 2008. Testing the intermediate disturbance hypothesis in species-poor systems: A simulation experiment for mangrove forests Journal of Vegetation Science 19: 417-424; doi: 10.3170/2008-8-18384 IV App. 3. Sensitivity analysis on random variables affecting recruitments Since there are no solid data on the variation of sapling numbers, the ranges of the two random variables rnd1 and rnd2 (Eq.5) were arbitrarily chosen (respectively [0.5 to 1.5] and [0.1 to 0.3]). They described a natural luctuation and occasional reappearance of saplings in the plot respectively. In order to analyze the suitability of these parameterizations, a sensitivity analysis was conducted to test the effect of the variation of these ranges on the variation of the median values of index of species dominance heterogeneity (ISDH). For this analysis, we selected simulations with the species heterogeneous spatial competition parameterization, the intermediate abiotic conditions and three selected cases of perturbation regimes that should present the so-called bell-shape pattern characteristic of the intermediate disturbance hypothesis. For each case, we tested 11 different new ranges for each random variables by multiplying the limit values of these 2 ranges by variation factors of Δrnd = 0.5 to 1.5. We measured the median ISDH results over the last 400 years for each new range (ISDH-new) and analyzed the variation comparing it to the original value (ΔISDH= ISDH-new /ISDH-0). The results of this sensitivity analysis are presented in Fig. A3.1. Variations of up to 10% of the ranges of rnd1 and rnd2 change with less than 10% the ISDH results and not generally the original pattern of system answer to perturbation regime. Actually, only with the extreme perturbation regime the rnd2 variation lead to variations of ISDH higher than 5% but increasing then the trend of bell-shape answer of the system to perturbation regime. Based on these results, we considered the selected ranges of rnd1 and rnd2 adequate for our study. Fig. A3.1. Results of sensitivity analysis of rnd1 and rnd2 on ISDH variations, (a) with low perturbation regime (intensity = 30%, frequency = 1 / 100 yr), (b) with intermediate perturbation regime (intensity = 50%, frequency = 1 / 40 yr) and (c) with extreme perturbation regime (intensity = 99%, frequency = 1 /10 yr) (original values of ISDH: 2.551, 2.781 and 2.766 respectively). All sensitivity analysis simulations were done with the species heterogeneous spatial competition parameterization and intermediate abiotic conditions. Apps. 1-4. Internet supplement to: Piou, C. et al. 2008. Testing the intermediate disturbance hypothesis in species-poor systems: A simulation experiment for mangrove forests Journal of Vegetation Science 19: 417-424; doi: 10.3170/2008-8-18384 V App. 4. Complementary results In this appendix we present complementary results of the simulation exercises. Fig. A4.1 shows the general results of the irst analysis: relative nutrient availability and salinity conditions on species dominance heterogeneity. Fig. A4.2 shows speciic results of the second analysis: effects of perturbation regimes on species dominance heterogeneity with selected abiotic scenarios. Fig. A4.1. ISDH variations according to salinity and relative nutrient availability (RNA) conditions for the two competition parameterizations. Points are median values of replicate simulations, error bars represent irst and third quartiles. Fig. A4.2. Median ISDH variations following perturbation frequency for the two competition parameterizations (a and b) with selected mortality intensity (70%) and medium abiotic scenario (Salinity 50 and RNA 80) (N = 30 for each point, boxes represent irst and third quartiles, error bars represent minimum and maximum). Apps. 1-4. Internet supplement to: Piou, C. et al. 2008. Testing the intermediate disturbance hypothesis in species-poor systems: A simulation experiment for mangrove forests Journal of Vegetation Science 19: 417-424; doi: 10.3170/2008-8-18384 VI The analysis of variations in species diversity (ISDH) of the system depending on the perturbation regime showed different type of patterns for the different parameterizations (Fig A4.3). The homogeneous spatial competition parameterization with benign abiotic conditions led to 4 U-shaped patterns out of 11 analyses. The heterogeneous spatial competition parameterization with benign abiotic conditions led to 4 bell-shaped patterns out of 11 analyses. In both cases, the rest of the analyses showed irregular, increasing or decreasing pattern of ISDH variations. With medium abiotic conditions the patterns were more often U-shaped or bell-shaped, but with an identical trend: the homogeneous spatial competition parameterization led to 6 U-shaped patterns and the heterogeneous spatial competition led to 7 bell-shaped patterns out of 11 analyses in both cases. Fig. A4.3. Patterns of system response in median ISDH variations according to frequency effect with selected intensity of disturbance (a) or according to intensity effect with selected frequency of disturbance (b). Apps. 1-4. Internet supplement to: Piou, C. et al. 2008. Testing the intermediate disturbance hypothesis in species-poor systems: A simulation experiment for mangrove forests Journal of Vegetation Science 19: 417-424; doi: 10.3170/2008-8-18384