“O espaço concreto foi extraído das coisas. Elas não estão nele, é ele que está nelas”
45 Functional redundancy in phytoplankton communities in different spatial scales
Letícia Barbosa Quesado1, Juliana B. O. Santos2, Camila Rodrigues Cabral, Fabíola da Costa Catombé Dantas, André M. Amado3, Adriano Caliman1, Luciana Silva Carneiro1
Addresses:
1Universidade Federal do Rio Grande do Norte (UFRN) – Lab. of Aquatic Ecology, Natal – Brazil. ZipCode: 59.072-970
2Universidade Federal do Rio de Janeiro (UFRJ) – Lab. of Phycology (Museu Nacional), Rio de Janeiro – Brazil. ZipCode: 20940-040
3Universidade Federal do Rio Grande do Norte (UFRN) – Lab. of Limnology, Natal– Brazil. ZipCode: 59.014-002
46 Abstract
Increasing scientific effort has been devoted to testing for patterns and mechanisms regulating functional-taxonomic diversity relationship. Spatial scale is recently being considered as a way to better understand it, because larger spatial scale incorporates distinct environmental filters, as a possible result a constant positive relationship is established. Here, we explore the influence of the grain of spatial scale in the functional-taxonomic relationship for phytoplankton community metrics (richness, evenness, diversity and functional redundancy). Our specific objectives were to test (1) whether and how the functional-taxonomic diversity relationship varies according to the spatial scale grain (lake and watershed); (2) if there is a change of importance of the component (functional and taxonomic diversity and richness) of the functional redundancy with increasing the spatial scale grain. For that, we sampled 95 lakes distributed in 13 watersheds in northeast Brazil. We found that there was a positive linear relationship among all functional and taxonomic community metrics at both levels of spatial scale grain, contrasting to what we expected. However, a visually asymptotic relationship was only observed with richness at the lake level of spatial scale grain. The linear relationship as the main pattern was due to the addition of the abundance weight, showing a high functional complementarity guided by evenness at the lake level and influenced by richness regionally strengthening diversity pattern. As a result, functional redundancy decreased at both spatial scale grain with changes in the metrics responsible for it. At the lake level, functional redundancy was correlated with an increase in functional diversity, while at the watershed level with richness, taxonomic and functional ones. These results suggest a substantial role of functional complementary for structuring phytoplankton communities and increasing importance across the spatial scale. More studies with distinct levels of spatial scale grain and even time can help elucidate biodiversity patterns and behaviors and how local and regional factors structure communities.
47 Introduction
Understanding biodiversity patterns and their consequences for ecosystem processes and services through a functional approach has become a central topic at the intersection of community and ecosystem ecology (Hooper et al. 2005; Cardinale et al. 2012). This scientific inquiry has been motivated mainly by the fact that biodiversity is a concept which integrates taxonomic, functional and evolutionary aspects of species, and may be correlated in many complex ways through space and time (Ricklefs 2004; Ricklefs 2007). In current functional ecology studies, there is a search to understand how taxonomic and functional diversity co-vary in space and time, and what are the determinants of this co-variation (Guillemot et al. 2011). The crucial point regarding this covariance lies in how the loss of taxonomic diversity impacts functional diversity (Petchey & Gaston 2006). As many species shares functional characteristics to a greater or lesser degree, it is important to understand at what level the loss of species can have meaningful results in the loss of functional traits in the community. Studies, so far, show how this relation behaves at the local level of spatial scale grain and recently revisited trait-based approach including functional traits in a continuous multivariate trait-space (Violle et al. 2007). However, most decisions for the conservation of biodiversity are taken on a broader scale.
Although this relationship has been subject of discussion in the literature, important aspects that drive it remains unclear, especially concerning what determines the form and strength of the relationship (Devictor et al. 2010; Guillemot et al. 2011). For example, regarding the biodiversity facets, species and functional diversity not weighted by abundance, which we can consider as functional richness, have been the most common metrics used to test this relationship (see Petchey & Gaston 2006; Petchey et al. 2007; Schmera et al. 2009; Cianciaruso et al. 2009; Cadotte et al. 2011; de Bello et al. 2012; Schriever et al. 2015). However, as one does not consider the relative abundance of individuals and traits, the interpretations of the relationship may present significant limitation on its qualitative effects (see Devictor et al. 2010; Guillemot et al 2011; Gallardo et al. 2011; Pillar et al. 2013; Weithoff et al. 2015). When
48 abundance is not considered, the increase in species richness increases the possibility to add function to the ecosystem, as new species bring new traits and variability to the present ones. When abundance is considered, changes can be due to two reasons, or related to abundant species (changes in taxonomic diversity) or trait abundance (changes in functional diversity), both as a response to environmental filtering (Vogt et al. 2010). However, changes in trait abundance can be species-related, as trait can vary along with species (Clark 2016), or trait- related when a trait is common for more than one species (Loreau 2004). In this case, an increase in taxonomic diversity and richness with the same functional diversity due to an increase of species with the same traits leads to functional redundancy (Loreau 2004).
Functional redundancy gives a potential for functional stability to the local community in time or space, defined as resistance when the community structure “avoids” any change in its functions when losses species (Carpenter et al. 2001). This stability can occur from within community through another species functionally redundant, or from a metacommunity point of view, among neighbor communities through dispersal and colonization (Cottenie 2005). In the first, community stays the same functionally, while in the second, it gains the ability to return its functions pre-species lost (Pillar et al. 2013).Despite its importance, most researchers have often inferred functional redundancy indirectly through a saturation in the relationship between functional diversity (or richness) and species richness (Petchey et al. 2007). In these studies, if functional redundancy is high, the relationships between taxonomic and functional richness will become asymptotic and weaker (Cadotte et al. 2011). So far, in a direct way, there is no consensus on how to measure functional redundancy (Ricotta et al. 2016). Recently, a new methodology and definition were proposed (de Bello et al. 2007) and improved (Ricotta et al. 2016). This new method is a partition of the Simpson index into Rao's quadratic entropy (functional diversity) and redundancy. Therefore, there is no functional redundancy when species are entirely different functionally, and functional redundancy is maximum when all species are functionally identical.
49 Assuming that biodiversity can be maintained by resource partitioning when distinct species complement each other functionally (paradox of the plankton’s solution: Hutchinson 1961; Tilman 1977), communities with dominant species have few species responsible for most functions (asymptotic relationship), while in diverse communities most species are rare and have a similar influence on the ecosystem functioning (linear relationship). As more species are added locally, more they will contribute with new functions enhancing complementarity till a saturation point (asymptotic relationship). Here, functional redundancy will persist, as one increases species diversity and does not affect the functional diversity (Dolédec & Bonada 2013). Above this level of “optimal” local biodiversity, the environment supporting capacity does not tolerate more species and the new ones would increase the competition, thus competitive exclusion would begin removing both species and functions (Dolédec & Bonada 2013). A curve similar to a Gaussian curve is formed in the relationship of functional and taxonomic diversities. However, with increasing spatial scale grain, species from different locations of the region complement each other functionally, so species richness increases are directly related to new functions (linear relationship) (Dolédec & Bonada 2013). Therefore, understanding how environmental and anthropogenic factors impacts species loss and loss of ecosystem functions at different spatial scale grain may tell us a better way to define areas of high conservation interest, such as biodiversity hotspots (Devictor et al. 2010).
In this context, we hypothesize that there will be a positive asymptotic relationship between functional and taxonomic diversity at lake level (local). With the increasing spatial scale grain and the possibility of new habitats with distinct environmental filtering forces acting upon communities, there will be a constant increase of functions in the metacommunity (watershed level) with the addition of a new species (Dolédec & Bonada 2013). Our specific objectives were to test (1) whether and how the functional-taxonomic diversity relationship varies according to the spatial scale grain (lake and watershed); (2) if there is a change of importance of the component (functional and taxonomic diversity and richness) of the functional redundancy
50 (index) with increasing the spatial scale grain. To answer these questions, we chose phytoplankton communities, because they are a group of microorganisms with well-known taxonomy and functional traits, and with a high potential dispersion (responding to regional factors), thus allowing exploring spatial questions related to the form and strength of the relationship between taxonomic and functional metrics.
Materials and methods
Study area
The field survey was carried out in Rio Grande do Norte (Brazil). The state has an area of 53,007 km2, with a precipitation gradient ranging from the humid coast in the East towards the semi-arid land in the West (Fig. 1). The area covered 33,562 km2, equivalent to 63% of the total area of the state.
51 Figure 1 – Map of the 95 lakes distributed in 13 watersheds and in two climatic regions (separated by a red dotted line; semi-arid and humid) sampled in Rio Grande do Norte state, Brazil, in September 2012.
52
Sampling design
We sampled a total of 95 lakes in 13 watersheds of two different climatic regions. The aim of our sampling design was to study the influence of spatial scale grain in the functional-taxonomic diversity relationship. We considered two different spatial scale grains: local when considering lakes and regional when considering watershed, which comprehended a set of lakes. From the 13 watersheds, one out of ten watersheds is located exclusively in semi-arid area (Piranhas/Açú), five exclusively in the humid region (Humid-east, Curimataú, Maxaranguape, Pirangi, and Rio Doce), and four in between the two regions (Ceará-Mirim, Jacú, Potengi, and Trairí) (IDEMA 2015; Fig. 1).
Sampling and sample analysis
All samplings were made during the dry season (INPE 2015) in September 2012, a year marked by a severe drought (EMPARN 2016). For each lake, we sampled water on the subsurface of the water column for characterization of phytoplankton community. Phytoplankton samples (100 mL) were preserved with lugol solution (0.2%, final concentration). Phytoplankton was counted and identified in random fields using inverted microscopy (Utermölh 1958; Uhelinger 1964). Thresholds for counting species were based on two criteria: up to 100 individuals (i.e. cells, colonies or filaments) of the most abundant species (Lund et al. 1958) and until saturation of the species accumulation curve. Species were classified using the major taxonomic schemes from the Integrated Taxonomic Information System (itis.gov), Tree of Life database (tolweb.org) and Adl et al. (2012), except for Cyanobacteria (Komárek et al. 2014). Community data were expressed in species abundances (number of individuals/mL). For each species present in a sample, when possible, 30 (or all if <30) random individuals were measured (maximum linear dimension of individual cells, colonies or filaments) to assess the local average species size, for all sites in which they occurred. Organism surface was measured by approximating their shapes to the most similar geometric form (Hillebrand et al. 1999; Sun & Liu 2003; Fonseca et al.
53 2014). Mucilage presence and type of life-forms (colonial, filamentous or unicellular) were noted during the counting procedure. For pigments composition, we used the classification proposed by Longhi & Beisner (2010), four main spectral groups according to the peripheral antennae. Therefore, green pigments were the ones with chlorophyll-a, -b and xanthophyll; blue with phycocyanin; brown with chlorophyll-a, -c and xanthophyll (fucoxanthin or peridinin); and mixed with chlorophyll-a, -c and phycoerythrin.
Measuring biodiversity facets
Phytoplankton diversity was accessed through taxonomic (richness and diversity) and functional metrics (Rao’s quadratic entropy weighted and not by species abundance – functional diversity and richness) (Laliberté & Legendre 2010). We estimated taxonomic richness as the species number per unit in each spatial scale grain (species/unit), while taxonomic diversity as Simpson diversity (D), considering all species as equally and maximally dissimilar (Ricotta et al. 2016).
To measure phytoplankton functional diversity, we considered the ones suggested by Litchman & Klausmeire (2008). We pre-selected the best combination of traits through a highest explained variability (Nock 2016) in a principal coordinate analysis (PCoA) of mean pairwise dissimilarities (Swenson 2014). For that, we tested all possible combination (4093 possible combinations) from the all traits measured (surface, size, volume, sphericity index, presence of flagella, silica, aerotopes, nitrogen-fixation, toxic potential, mucilage, pigment composition and mixotrophy). The traits combinations with highest explained variability were composed by distinct types: a) continuous: (1) individual size (maximum linear dimension) and (2) surface; and b) categorical: (3) presence of mucilage, (4) life-forms (unicellular, chains-filaments, colonies) and (5) pigment composition (blue, green, brown, and mixed). We used a multiple process approach to best functionally characterize the community, as the overall functioning is a joint effect of many ecosystem processes, and species contribute to more than one process at the same time (Reiss et al. 2009). The functional traits matrix was standardized and transformed into
54 a distance functional matrix through the mixed-variables coefficient of Gower distance, both proposed by Pavoine et al. (2009).
We, then, calculate the functional diversity and richness with the widely used Rao’s quadratic entropy, defined as the average of all dissimilarities, weighted (diversity) or not (richness) by their relative abundance in a multidimensional space-trait (de Bello et al. 2016). Rao’s index is one of the few functional diversity indexes which considers abundance, multiple- trait approach and has low sensitiveness to species richness (Petchey & Gaston 2006). Rao’s index is equal to the taxonomic diversity Simpson index (expressed as 1–D, D = dominance), therefore when all species are functionally different (unique), Rao is the maximum Simpson index value. When including relative abundance into the index, we can interpret as if the loss of individuals will translate in the loss of particular functions, which appears to be the main threat to the ecosystem processes and services (Mouilllot et al. 2014). On the other hand, trait similarity (redundancy) will explain the maintenance of a process (de Bello et al. 2010). Rao and Simpson’s indexes were calculated with melodic function proposed by de Bello et al. (2016).
Functional redundancy (FR) is considered the number of species sharing the same space trait (de Bello et al. 2007; Pillar et al. 2013; Ricotta et al. 2016). We obtained FR through two methods: a visual observation of the scatter plot of the functional-taxonomic diversity relationship (common method) and as a standardized index that relates the observed functional diversity (FDQ – which takes the species dissimilarities into account) to the value of a maximally distinct community with the same abundance distribution (Simpson diversity):
FR= ((1-D)-FDQ)/(1-D) (Ricotta et al. 2016). FR varies from 0 to 1, if FR = 0 the species have completely different traits, and if FR = 1 all species have identical traits.
To assist the functional and taxonomic diversity indexes interpretation, we also calculated the functional and taxonomic evenness indexes. The functional evenness was calculated considering its maximum value, when all species relative abundances are perfectly evenly
55 distributed and all dissimilarities are equal (perfect evenness), method proposed by Ricotta et al. (2014). Taxonomic evenness was calculated through the Simpson index of evenness, which compares Simpson's diversity with a distribution of observed species that maximize diversity. Except for taxonomic richness, all other metrics range from zero to 1.
Testing diversity relationship
To assess whether and how functional-taxonomic diversity relationship varies according to the spatial scale grain (lake and watershed level), we performed Pearson’s correlation between functional-taxonomic diversity, richness, and evenness for each spatial scale grain. At the lake level of spatial scale grain, all indexes were calculated for each individual lake (n = 95). For the watershed level, because the number of sampled lakes varied across watersheds (2 till 17; Fig. 1) and diversity is expected to increase with sampling effort (i.e., in our case the number of sampled lakes per watershed), we standardized calculations within watershed as the average of all possible pair-wise combination of lakes communities of a given metric. We used lakes pair-wise combination because it assumes the minimum possible number of lakes (i.e. two) for calculations, what allowed us to consider a great number of watersheds in the analyses and standardizes the sampling effort per watershed. Therefore, we summed all individuals of a given species in each pair of lakes to recalculate the relative abundance and all the indices of this new "community". We also fitted a linear regression to obtain the slope of the taxonomic and functional metrics relationship and, then, being able to examine whether the observed functional redundancy increased, decreased or remained stable with increasing spatial scale grain (Guillemot et al. 2011), in the scatterplot graphics. A steep slope indicates a fast increase of new functions, whereas a gentle slope indicates a greater redundancy of existing functions (Guillemot et al. 2011). To see if there is a change of importance of the component of the functional redundancy with increasing the spatial scale grain, we performed Pearson’s correlation between
56 functional-taxonomic diversity and richness with functional redundancy calculated by the index for both spatial scale grain.
Testing random effects
To test if differences in species richness biased the pattern observed in the functional-taxonomic relationship, we compared the observed pattern to the one obtained when functional diversity is randomized (Manly & Sanderson 2002; Swenson 2014). For that, we used the standardized effect size (SES) for functional diversity and evenness (Gotelli & McCabe 2002; Mason et al. 2013). In each spatial scale grain, abundances values were randomized with the independent swap algorithm (Gotelli 2000) maintaining species occurrence frequency and lake species richness (Kembel et al. 2016). SES values were obtained by dividing the difference between the observed and the mean of the null distribution by the standard deviation of the respective null distribution (Laureto & Cianciaruso 2015). Then, we performed Pearson’s correlation between taxonomic diversity and evenness with SES Rao’s quadratic entropy for each spatial scale grain. We interpreted SES values relationship with taxonomic diversity and evenness as, when observing the same pattern, functional diversity and evenness are independent of species richness, if not functional diversity and evenness are dependent. We ran 10000 randomizations for all null model analyses to ensure accurate estimates of SES values. However, for this analysis, we had to remove three watersheds with only two sampled lakes (Humid-north, Boqueirão, and Punaú) because the randomization of the abundance of two lakes would not change the result of the combined community representative of the watersheds, remaining 89 lakes for this test. All analyses were performed in the statistical software R (R Development Core Team) using the packages vegan (Oksanen et al. 2017), ade4 (Dray et al. 2017) and picante (Kembel et al. 2016).
57 Results
Functional-taxonomic relationships across spatial scales
All taxonomic and functional metrics increased with increasing spatial scale grain (Appendix A) and none of the taxonomic metrics were 1:1 or more distributed with their functional respective at any spatial scale grain (Fig. 2). At the lake level of spatial scale grain, functional and taxonomic diversity were strongly positively correlated (r = 0.924, p < 0.0001; Fig. 2A), as well as the standard effect size (SES) of functional diversity and taxonomic diversity (Appendix B: Fig. S2C, SES: r = 0.916, p < 0.0001). Taxonomic richness showed a moderate correlation with functional richness when more than one species is needed to increase one functional trait (r = 0.473, p < 0.0001; Fig. 2C). Functional and taxonomic evenness were strongly correlated, even though high values of taxonomic evenness were not translated in high values of functional evenness, as we could observe with the low slope (β = 0.418, r = 0.777, p < 0.0001; Fig. 2E). The strong correlation was corroborated with the standard effect size (SES) of functional evenness and taxonomic evenness (Appendix B: Fig. S2G, SES: r = 0.779, p < 0.0001).
The increase in the spatial scale grain revealed the importance of richness on influencing another metrics’ relationship. From the lake to watershed level of spatial scale grain, richness’ relationship showed the highest increase in the slope (β = 0.004 to 0.938; Fig. 2D), influencing a moderate increase in diversity (β = 0.759 to 0.938; Fig. 2B), while evenness decreased the slope as well as explanation power (β = 0.418 to 0.29, r = 0.597, p = 0.0312; Fig. 2F). Taxonomic richness presented also a greater variability in its values within watershed (Fig. 2D). The watershed level of spatial scale grain preserved the diversity relationship, although the functional diversity was less explained by species diversity, the remaining relationship was stronger (r = 0.735, p = 0.004; Fig. 2B). Although diversity showed a similar pattern as evenness, richness affected diversity significance at the watershed level, as SES functional diversity comparison was not significant (Appendix B: Fig. S2D). Therefore, when observing the functional-
58 taxonomic metrics relationship, spatial scale grain only influenced greatly richness patterns and maintained increasing functionally with increasing taxonomic evenness, while diversity lost significance due to its dependency with richness.
Figure 2 – Functional-Taxonomic diversity (A-B), richness (C-D), and evenness (E-F) correlations at the lake (95) and watershed (13) level of spatial scale grain. Error bar represents