Spatio-temporal nested patterns in macroinvertebrate assemblages across a pond network with a wide hydroperiod range

Spatio-temporal nested patterns in macroinvertebrate assemblages across a pond network with a wide hydroperiod range Margarita Florencio • Carmen D´ıaz-Paniagua • Laura Serrano • David T. Bilt on Abstract Nestedness has been widely used to measure the structure of biological communities and occurs when species-poor sites contain subsets of species-rich ones. Here, we examine nested patterns across the macroinvertebrate assemblages of 91 ponds in Don˜ ana National Park, Spain, and explore temporal variation of nestedness and species richness in 19 temporary ponds over 2 years with differing rainfall. Macroinvertebrate assemblages were significantly nested; both pond spatial arrangement and environmental variation being important in driving nested patterns. Despite the nested structure observed, a number of taxa and ponds deviate from this pattern (termed idiosyncratic), by occurring more frequently than expected in species-poor sites, or having assemblages dominated by species largely absent from species-rich sites. Aquatic adults of winged insects, capable of dispersal, were more highly nested than non-dispersing taxa and life-history stages. Idiosyncratic taxa were found in ponds spanning a wide range of hydroperiods, although nestedness was higher in more permanent waterbodies. Monthly sampling Communicated by Jonathan Shurin. M. Florencio ( & ) · C. D´ ıaz-Paniagua Don˜ ana Biological Station-CSIC, P.O. Box 1056, 41080 Seville, Spain e-mail: [email protected] L. Serrano Department of Plant Biology and Ecology, University of Seville, P.O. Box 1095, 41080 Seville, Spain D. T. Bilton Marine Biology and Ecology Research Centre, University of Plymouth, Drake Circus, Plymouth PL4 8AA, UK demonstrated a gradual increase of species richness and nestedness from pond filling to April–May, when the most temporary ponds started to dry. Although the degree of nestedness of individual pond assemblages varied from month to month, the overall degree of nestedness in the two study years was practically identical despite marked differences in hydroperiod. Our results suggest that differential colonization and environmental variation are key processes driving the nested structure of Don˜ ana ponds, that macroinvertebrate assemblages change in a predictable manner each year in response to cycles of pond wetting and drying, and that connectivity and environmental variability maintain biodiversity in pond networks. Keywords Colonization · Community composition · Dispersal · Nestedness · Temporary ponds Intro duction Nested systems occur when species-poor sites contain subsets of the assemblages found in species-rich sites, and the degree of nestedness thus quantifies the overlap in species composition between high and low diversity areas (McAbendroth et al. 2005). Nestedness has been widely used to measure the structure of biological communities (Fleishman and Murphy 1999; Hylander et al. 2005; Meyer and Kalko 2008; Elmendorf and Harrison 2009), including lentic freshwaters (Baber et al. 2004; McAbendroth et al. 2005; Angeler et al. 2008; Wissinger et al. 2009), although few studies have explored the factors which may drive this pattern. Factors that lead to consistent differences amongst species in immigration or extinction rates cause strong patterns of nestedness across species assemblages (Wright and Reeves 1992). Although, originally, historical extinction was assumed to be the main cause of nestedness (Atmar and Patterson 1993), local immigration through the differential dispersal abilities of species can also be very important in generating nestedness under particular conditions (Cook and Quinn 1995). Whilst extinction is important in structuring biological communities over long timescales (Patterson 1990; Hausdorf and Hennig 2003), differential colonization can be important over shorter-term scales (Patterson 1990; Atmar and Patterson 1993), particularly when colonization drives community assembly in originally empty sites such as new and ephemeral habitat patches (Norton et al. 2004). In addition to selective immigration and extinction, variation in environmental conditions, including habitable area, across sites can also generate nested patterns (Lomolino 1996; Baber et al. 2004; McAbendroth et al. 2005; Heino et al. 2009), although studies of nestedness which assess the importance of immigration–extinction dynamics and environmental heterogeneity are rare in any habitat. Temporary ponds are fluctuating waterbodies with recurrent seasonal phases of flooding and desiccation in most years (Grillas et al. 2004; Bro¨ nmark and Hansson 2005; Williams 2006). Although some macroinvertebrates persist in ephemeral habitats as resting stages in dry sediment (see Batzer and Wissinger 1996), dispersal to more permanent waterbodies is the main strategy through which they survive dry phases in temporary aquatic habitats (Wiggins et al. 1980; Bilton et al. 2001; Garrido and Munilla 2008). As a consequence, much of the invertebrate biota of temporary ponds is assembled through colonization on each re-wetting cycle, making these an ideal system in which to explore the role of colonization in generating nestedness. Such a dispersal-driven system can have important implications for nestedness patterns, and isolated ponds would show lower nestedness if the regional species pool is dominated by poor dispersers (Cook and Quinn 1995; McAbendroth et al. 2005). In lentic freshwaters, hydroperiod duration is well established as a key factor driving assemblage structure (e.g., Baber et al. 2004; Urban 2004; Waterkeyn et al. 2008). As nestedness can be related to assemblage stability (Atmar and Patterson 1993), this could be expected to increase with hydroperiod length due to the presence of species which require more time to complete their life cycles which are not able to survive even occasional dry phases. The temporal variability of nested patterns in biological communities can shed light on the role of colonization and extinction in generating nestedness (Patterson 1990; Loo et al. 2002; Bloch et al. 2007), and the temporal cycle of wetting and drying in ephemeral waters makes them an ideal system in which to explore such processes. Despite this, no studies to date have explored temporal variation in nestedness in these systems over their wetting/drying cycles. Idiosyncratic species and sites are those that depart from the general nested pattern, reducing the value of matrix nestedness (Atmar and Patterson 1993). Idiosyncratic species are those which occur more frequently than one would expect in species-poor sites, and/or less frequently in species-rich ones, something which could result through habitat specialization or competition avoidance, as well as being a locally infrequent habitat generalist (see McAbendroth et al. 2005). Individual sites are considered idiosyncratic when they contain a higher proportion of idiosyncratic species than one would predict from their species richness––in other words, their assemblages contain a number of species which are largely absent from the most species-rich sites. The detection of idiosyncratic species and sites is of potential interest for conservation (McAbendroth et al. 2005; Heino et al. 2009) as idiosyncratic sites may harbor specialist taxa not present in species-rich localities, and idiosyncratic species may require specialized conditions, present in relatively few sites in a region (Atmar and Patterson 1993). Here we examine nestedness and its temporal variation, across a natural pond network in southern Spain, spanning a wide range of hydroperiods. We determine the degree of nestedness across the pond network as a whole by analyzing data from close to 100 ponds, collected over a single season. We use this dataset to determine whether nestedness increases with: (1) the dispersal ability of macroinvertebrates; (2) the degree of isolation of ponds; and (3) the length of hydroperiod. We go on to examine whether variation in environmental conditions across sites may contribute to nestedness, and evaluate the relative importance of colonization and environmental variation in generating observed patterns. Additionally, we explore whether species richness and nestedness vary in a predictable manner over the hydrological cycle in a subset of ponds sampled monthly across 2 years with differing rainfall regimes. Finally, we consider the possible implications of our findings for the conservation of lentic biodiversity in the region. Materials and method s Study area Our study was performed in Don˜ ana National Park, located between the mouth of the Guadalquivir River and the Atlantic Ocean in south-western Spain (see Siljes tr o¨ m et al. 1994). The climate of this area is Mediterranean subhumid, with most rainfall in autumn and winter, hot and dry summers, and mild winters. In this area, two main regions can be differentiated: an extensive marsh and an adjacent sandy area where numerous temporary waters form in natural depressions; more than 3,000 occurring in wet years. Ponds in the park differ widely in surface area and hydroperiod, and constitute a heterogeneous network of aquatic habitats which has been listed under the RAMSAR Convention since 1982. Temporary ponds in Don˜ ana are seasonally flooded after heavy rains, usually filling in autumn or winter, and persisting until late spring or early summer. There are also two large semi-permanent ponds which only occasionally dry in years of severe drought, and some natural ponds have been artificially deepened to supply water for cattle and wild mammals, making them permanent in most years. A detailed description of the environmental characteristic of these ponds is given in Go ´ mez-Rodr´ ıguez et al. (2009), Espinar and Serrano (2009) and D´ ıaz-Paniagua et al. (2010). Sampling surveys Two linked sampling surveys were carried out, and these provide the data which form the basis of this study. Extensive macroinvertebrate survey In order to assess the degree of nestedness across the pond network as a whole, we sampled a total of 91 ponds which ranged from 24 to 122,672 m2 in area between end-March and mid-June 2007. Pond assemblages in Don˜ ana were relatively stable over this time period (no consistent change in species richness between March and June; linear regression: R2 = 0.017, p = 0.22), which was the minimum window over which all sites could be visited. Ponds were chosen to span the full range of hydroperiod variation present within the study area, and cover all areas of the park where discrete lentic waterbodies occur (Fig. 1). Based on their depth, area, and previous observations (see Florencio et al. 2009), ponds were assigned to one of three categories based on their degree of permanence: short hydroperiod (n = 28), intermediate hydroperiod (n = 22), and long hydroperiod (including permanent ponds; n = 41). Pond surface area varied considerably within each hydroperiod category, and did not increase with pond duration. Although short and intermediate hydroperiod categories did not differ in area, long hydroperiod ponds were actually significantly smaller (one-way ANOVA, F (2,88) = 10.919, p \ 0.0001; Tukey HSD, p \ 0.0001; see Table 1), often being relatively small pools which had been artificially deepened (see above). Monthly macroinvertebrate survey To analyse temporal variation in nestedness over wet–dry cycles, we sampled aquatic macroinvertebrates in 19 temporary ponds during their wet phases, from February 2006 to August 2007. All these ponds were distributed within the central area of the Park (Fig. 1). The first study year was relatively dry (468 mm rainfall; hereafter referred to as the dry year), and most ponds were only wet from February to May. The second study year was wetter (716.9 mm rainfall; hereafter referred to as the wet year), and most ponds were flooded from November to July. Due to the increase in the number of ponds formed in the second year (as a result of higher rainfall), we sampled three additional ponds replacing three ponds only sampled in the dry year in order to include ponds spanning the highest range of hydroperiod observed in the area. For the monthly survey, we calculated hydroperiod as the number of months ponds were flooded in a given year, and classified this in relation to the longest hydroperiod seen in each year. In the dry year, hydroperiod Fig. 1 Location of the ponds sampled in Don˜ ana National Park, identified by a hydroperiod categories and b idiosyncratic or nested character of their macroinvertebrate assemblages. The 19 ponds sampled for the monthly survey are within the Don˜ ana Biological Reserve area marked by a dashed line. Marsh, where discrete ponds are largely absent during flooding periods, is indicated by light shading 4 4 Table 1 Environmental characteristics of ponds in the extensive survey, showing the mean, minimum (min.) and maximum (max.) v alues Environmental variables Mean (min–max) Pools (%) 3.8 (0–80) Maximum depth (cm) 74.2 (9–126) Pond area (m2) 4,103 (24–122,672) Short hydroperiod 3,427 (25–23,097) Intermediate hydroperiod 4,517 (25–37,544) Long hydroperiod 4,370 (24–122,672) Pond number 4.4 (0–16) NH ? (mg l - 1 ) 1.34 (0.04–3.59) - 1 Diptera, Oligochaeta and saldid bugs were identified only to family. We kept the different taxonomic levels reached for adults and larvae, using these stages of taxa separately (hereafter referred to as taxa/stages), to assess the role of each stage in generating nested patterns. Environmental variables In order to determine whether pond characteristics influence macroinvertebrate nestedness, we measured a range of environmental variables across the ponds (see Table 1). In the field, we measured the proportion of the pond surface divided into separate pools, maximum water depth (with a i-P (mg l [O2] (mg l ) 0.11 (0.01–0.75) - 1 ) 2.2 (0.2–16.8) - 1 graduated pole at the deepest point of the pond), conductivity (on bed using HI 9033), dissolved oxygen concen- EC ( l S cm ) 962.8 (82–8,800) tration (on bed using YSI 550A) and turbidity (in water Turbidity (NTU) 35.0 (1.1–975.5) Organic matter (%) 5.43 (0.17–27.74) SO 2 - (meq l - 1 ) 0.82 (0.04–23.24) Pond number The number of ponds [ 150 m2 within 200 m of each pond, i-P dissolved inorganic phosphate, [O2] dissolved oxygen concentration, EC electrical conductivity categories were short ( \ 2.5 months), intermediate (2.5–3.5 months) and long ( [ 3.5 months). In the wet year, short hydroperiod was \ 7 months, intermediate 7–8 months and long [ 8 months (for further details on hydroperiod of these ponds, see Florencio et al. 2009). Macroinvertebrate sampling and taxon identification We sampled macroinvertebrates with a dip-net (39 9 21 cm, 1 mm mesh), netting a stretch of water of approximately 1.5 m length in each sampling unit. In each pond, we sampled all different available microhabitats, based largely on differences in aquatic plant cover and depth (Heyer et al. 1994). As the efficiency of dip-netting increases in small ponds (Heyer et al. 1994), we took more samples in larger ponds, which also typically contained a higher number of microhabitats, in order to achieve comparable effort in detecting rare species. In all ponds, sampling was concentrated amongst vegetation, where most macroinvertebrate species are located. In most cases, a minimum of three samples were taken (maximum 13), with the exception of five small ponds sampled at the end of their dry phase, which were so small that three discrete 1.5-m sampling areas could not be identified. Most macroinvertebrates were identified in situ and then returned to the pond, with those taxa which could not be determined in the field preserved in 70% ethanol for examination in the laboratory. Identifications were performed to the lowest taxonomic level possible, in general to species (most adults) or genus (most larvae), although Basomatophora, column using HANNA HI93703). Surface water (500 ml) was also collected to determine nutrient [dissolved inorganic phosphate and ammonium––using an Auto Analyzer (Bran ? Luebbe)] and SO 42 - anion concentrations [using inductively coupled plasma mass spectrophotometer (ICP)]. Surface sediment samples (5 cm depth) were collected and organic matter measured in the laboratory (mean of three replicates via loss on ignition, 450°C, 5 h). Pond area and the number of ponds [ 150 m2 within 200 m of each pond were extracted from a GIS-based map constructed at the time of maximum inundation (see Go ´ mez- Rodr´ ıguez et al. 2008). Statistical analyses We constructed several matrices using presence–absence data of macroinvertebrate occurrence by pond: For the extensive survey we built: (1) the taxon matrix, pooling adults and larvae of individual taxa, to the highest taxonomic resolution possible; (2) the stage matrix, with adults and larvae separated; (3) the disperser matrix, including only adults of taxa capable of flight; and (4) the non-disperser matrix, including only larvae and taxa with non-flying adults. To evaluate whether longer hydroperiod increased nestedness, we built three sub-matrices of the stage matrix which included the ponds catalogued as short, intermediate or long hydroperiod for each matrix. A v2 test was used to evaluate whether idiosyncratic taxa or stages were associated with ponds of any of the hydroperiod categories. A v2 test was also used to evaluate whether idiosyncratic ponds were more common in some hydroperiod categories than others. In order to evaluate how the degree of nestedness observed across ponds changed through their wetting– drying cycle, we built a stage matrix as described above for each month when [ 2 ponds were wet (4 months in the dry year; 9 months in the wet year). To compare nestedness between the two study years, we also built two annual presence–absence matrices with the occurrence of taxa/ stages pooled across months (one matrix per year). The size of monthly matrices varied between sampling months with the gradual desiccation of ponds, as not all ponds were flooded in all months. Also, different numbers of taxa/ stages were recorded each month. For nestedness analysis, we used ANINHADO (see http://www.guimaraes.bio.br; Guimaraes and Guimaraes 2006), a recent package with advantages over the more widely used nestedness temperature calculator (Atmar and Patterson 1995). We considered two nestedness metrics: (1) temperature (T), widely used to analyse nested patterns (Atmar and Patterson 1993), and (2) the recently developed NODF, which differs from T in being independent of both matrix size and shape (Almeida-Neto et al. 2008). Atmar and Patterson (1993) interpret T as a thermodynamic measure of disorder which ranges from 0 to 100°, being low in ordered (i.e., nested) systems and high in disordered systems. The procedure to calculate both T and NODF requires a presence–absence matrix ordered by species presences (top-to-bottom), automatically generated in ANINHADO (Guimaraes and Guimaraes 2006). Using this ordered matrix, T is calculated through the number of departures from a perfectly nested matrix (Atmar and Patterson 1993). In order to scale the degree of nestedness, or order, to values between 0 and 1 (maximum nestedness), we used N instead T, calculated as N = (100 - T )/100 (Bascompte et al. 2003). As we compared nestedness between matrices with different sizes, and N is dependent on matrix size (Atmar and Patterson 1993; Almeida-Neto et al. 2008), we calculated the relative nestedness [N* = ( N - N R )/ N R ] where N is the degree of nestedness of the original matrix and NR is the average nestedness of 1,000 null model matrices (Bascompte et al. 2003). NODF, the recently developed metric to calculate nestedness, varies from 0 to 100 (maximum nestedness). NODF is obtained through the percentage overlap of presences for each pair of columns and for each pair of rows in an ordered matrix (Almeida-Neto et al. 2008). We include reference to NODF and N here to allow comparison with previous studies, most of which have employed variations of the latter metric. ANINHADO allows the degree of nestedness present within a system to be assessed against four separate null models, all of which were tested here: (1) presences assigned at random across the matrix; (2) a fixed–fixed null model, with both column sums and row sums fixed; (3) only column sums fixed; and (4) only row sums fixed (see http://www.guimaraes.bio.br). The significance of observed values of nestedness was assessed using 1,000 permutations of each null model. Using N and NODF, whenever significant nestedness was obtained with the fixed–fixed null model, it was also obtained using the other three, and we therefore used the fixed–fixed null model (the most restrictive; incurring fewer Type I errors; Gotelli 2000; Ulrich and Gotelli 2007a, b; Almeida-Neto et al. 2008) to assess the degree of nestedness. To detect idiosyncratic ponds and taxa we calculated N for each pond and taxon/stage, and considered as idiosyncratic those taxa/stages whose N values were lower than the value of the original matrix (Atmar and Patterson 1993). Since the degree of nestedness can be affected by sample size, and the number of samples taken differed between ponds, we explored the effect of sample number on nestedness via sample-based rarefaction (Gotelli and Colwell 2001) with the minimum number of samples per pond and month using EcoSim Version 7 (Gotelli and Entsminger 2004). In all cases, we obtained similar nested patterns to those produced with raw data, and consequently present only the latter here. Pond spatial isolation was used to assess the contribution of differential colonization to nestedness patterns, following Lomolino (1996). We evaluated whether the degree to which a pond departed from nestedness (N) was affected by the spatial arrangement of sampled ponds (i.e., whether the level of nestedness observed depended on relative isolation), using distance-based Moran’s eigenvector maps (MEM; see Dray et al. 2006), a general framework of principal coordinates of neighbour matrices (PCNM; see Borcard and Legendre 2002). We also compared the results for disperser and for non-disperser matrices in order to evaluate the role of dispersal in generating the nested pattern. We used the Delaunay triangulation criterion (see Legendre and Legendre 1998) which was appropriated to connect large distances, but did not consider edge effects since the Park boundaries are not the limit of the aquatic systems in the region. The number of eigenvectors was reduced using the ‘ortho.AIC’ command in R software (‘spacemakeR’ package; Dray et al. 2006). Only significant eigenvectors were used directly as explanatory variables in a multiple forward stepwise regression with the nestedness of ponds (in STATISTICA 6.0). Forward stepwise regression was performed following Blanchet et al. (2008), in which all eigenvectors were included to analyze global significance (preselected alpha = 0.05) considering the adjusted coefficient of multiple determination (Adj. R2). This complete procedure is an effective way of controlling for Type I error (Peres- Neto and Legendre 2010). To explore the role of inter-pond environmental variation in driving nestedness in the macroinvertebrate community, we used the Monte Carlo approach of Lomolino (1996). Analyses here were restricted to a subset of 80 ponds for which full environmental variables were available. The ponds of the stage matrix were ranked by each environmental variable, and the number of unexpected absences followed by a presence quantified as the departure (D) from perfect nestedness (Honnay et al. 1999). Statistical significance was estimated as the number of randomizations of the fixed–fixed null model giving D \ observed (see Lomolino and Davis 1997). Environmental variables with identical values in two or more ponds (proportion of pools, maximum depth, nutrient concentrations and the number of ponds [ 150 m2 within 200 m) were not analyzed, as the approach requires unique values for each site (Lomolino 1996). To assess the relative role of pond spatial arrangement and local environmental variation in generating nested patterns, we used a variance partitioning approach. Urodele predators (Triturus pygmaeus, Lissotriton boscai and Pleurodeles waltl) and exotic fish (Gambusia holbrooki), were ubiquitous and highly infrequent, respectively, had no influence on macroinvertebrate nestedness, and were therefore excluded from further analyses. As a measure of the spatial arrangement of ponds, we used spatial descriptors extracted from MEMs; the effect of environmental variation being summarized using the first axis scores from a principal components analysis (PCA) of all environmental variables (Primer v.6; Clarke and Warwick 2001). Variance partitioning was performed using the ‘varpart’ command in R software (‘vegan’ package; Oksanen et al. 2008), which obtains an adjusted multiple coefficient of determination (Adj. R2, 0–1) in order to compare explanatory variables (Peres-Neto et al. 2006). Significance was tested using the complete procedure described above, through a multiple forward stepwise regression for the spatial descriptors, and a linear regression for the PC1 scores. Result s Nestedness in the Don˜ ana ponds network The macroinvertebrate assemblages of Don˜ ana ponds were highly nested overall (stage matrix: N = 0.81, 16.43% fill; taxon matrix: N = 0.83, 17.65% fill; Table 2). The taxon matrix was more highly nested than the stage matrix (including adults and larvae separately), in which we detected a higher number of idiosyncratic taxa and stages (Table 2). The disperser matrix exhibited higher nestedness than the non-disperser matrix (Table 2). Out of the 135 taxa/stages recorded, 59 were idiosyncratic (Table 3), occurring in 1–53 ponds, across all hydroperiod classes. Some idiosyncratic beetles [Yola bicarinata (Latreille, 1804), Hygrobia hermanni (Fabricius, 1775), Hygrotus confluens (Fabricius, 1787)] and damselflies [Lestes virens (Charpentier, 1825)] were significantly more abundant in long hydroperiod ponds, whereas some idiosyncratic flies (Culicidae and Chaoborus spp.) were significantly more frequent in short hydroperiod ponds. Sympetrid dragonflies [Sympetrum fonscolombei (Selys, 1841), S. meridionale (Selys, 1841), S. striolatum (Charpentier, 1840) and S. sanguineum (Mu¨ ller, 1764)] were significantly more frequent in short and intermediate hydroperiod ponds rather than those with long hydroperiods (Table 3). Of the 91 studied ponds, 34 were idiosyncratic at the assemblage level (Table 4). These idiosyncratic ponds tended to cluster in both northern and southern areas of the park, suggesting an effect of isolation, but did not correspond to a particular hydroperiod category (Fig. 1). Four significant spatial descriptors extracted from MEMs analyses were strongly correlated with the degree of nestedness of the ponds (N) of the stage matrix (Adj. R2 = 0.35, F (24,66) = 2.987, p \ 0.001); i.e., the spatial arrangement of ponds, based on their inter-pond distances (pond isolation), can partially drive the macroinvertebrate nested structure. This effect was much higher in the disperser matrix with six significant spatial descriptors (Adj. R2 = 0.40, F (21,68) = 3.849, p \ 0.0001) than in the non-disperser matrix with only a single significant spatial descriptor (Adj. R2 = 0.13, F (10,79) = 2.354, p \ 0.05). This means dispersers were much more affected by the spatial structure and isolation of ponds than non-dispersers. We detected significant differences in the proportion of idiosyncratic ponds in each hydroperiod category (v2 = 11.99, df = 2, p \ 0.01). Table 2 Degree of nestedness (N, N* and NODF) detected in the extensive macroinvertebrate survey for different matrices: the taxon matrix, the stage matrix (which included adult and larval stage separately), matrix of dispersers (adults with capability of flight) and non-dispersers (larvae and non-flying adults) N N* NODF Columns (taxa) Idio-ponds Idio-taxa Stage matrix 0.81 0.43 34.52 135 34 59 Taxon matrix 0.83 0.46 45.20 102 37 32 Disperser matrix 0.82 0.48 42.32 66 40 24 Non-disperser matrix 0.88 0.34 41.55 68 30 24 Columns are the number of taxa/stages; idio-taxa the number of idiosyncratic taxa/stages; idio-ponds the number of idiosyncratic ponds. All nested values were significant at p \ 0.001 Table 3 Idiosyncratic taxa/stages (adults or larvae) with their degree of nestedness (N), the number of ponds in which they occurred (Pond occurrence) and hydroperiod category of these ponds: Short, Intermediate (Interm) and Long Order Taxa/stages N Pond occurrence Short Interm Long Bassomatophora Adults Physa spp. 0.42 33 10 6 17 Coleoptera Gyrinus dejeani Brulle ´, 1832 0.47 15 4 3 8 Heteroptera Trichocorixa verticalis (Fieber, 1851) 0.50 13 5 4 4 Coleoptera Hydrobius fuscipes (Linnaeus, 1758)/Limnoxenus niger (Zschach, 1788) 0.54 36 12 13 11 Coleoptera Hydroglyphus geminus (Fabricius, 1792) 0.54 28 5 5 18 Coleoptera Rhantus hispanicus Sharp, 1882 0.62 31 15 6 10 Heteroptera Notonecta meridionalis Poisson, 1926 0.63 25 8 6 11 Coleoptera Ochthebius dilatatus Stephens, 1829 0.64 5 3 0 2 Coleoptera Helochares lividus (Forster, 1771) 0.66 10 4 1 5 Heteroptera Notonecta viridis Delcourt, 1909 0.67 23 4 5 14 Coleoptera Agabus nebulosus (Forster, 1771) 0.67 19 8 3 8 Coleoptera Hydroporus lucasi Reiche, 1866 0.67 41 12 13 16 Coleoptera Yola bicarinata (Latreille, 1804) 0.67 6* 0 0 6 Coleoptera Hygrotus confluens (Fabricius, 1787) 0.70 21* 3 3 15 Coleoptera Hygrobia hermanni (Fabricius, 1775) 0.71 29* 5 4 20 Heteroptera Gerris thoracicus Schummel, 1832 0.71 55 18 16 21 Coleoptera Berosus signaticollis (Charpentier, 1825) 0.71 15 4 6 5 Heteroptera Notonecta glauca Linnaeus, 1758 0.72 18 6 6 6 Coleoptera Enochrus fuscipennis (Thomson, 1884) 0.73 32 7 11 14 Coleoptera Hygrotus lagari (Fery, 1992) 0.73 34 5 9 20 Bassomatophora Planorbidae 0.74 13 2 3 8 Heteroptera Sigara scripta (Rambur, 1840) 0.74 12 3 2 7 Heteroptera Plea minutissima Leach, 1817 0.75 26 4 6 16 Heteroptera Nepa cinerea Linnaeus, 1798 0.75 8 2 2 4 Coleoptera Anacaena lutescens (Stephens, 1829) 0.75 49 16 12 21 Heteroptera Micronecta scholzi (Fieber, 1860) 0.77 6 1 0 5 Coleoptera Helophorus spp. 0.78 36 10 9 17 Coleoptera Noterus laevis Sturm, 1834 0.78 6 1 1 4 Coleoptera Ilybius montanus (Stephens, 1828) 0.78 11 1 4 6 Coleoptera Laccobius revelierei Perris, 1864 0.79 1 1 0 0 Heteroptera Sigara lateralis (Leach, 1817) 0.79 39 11 8 20 Coleoptera Berosus affinis Brulle ´ , 1835 0.79 17 4 6 7 Heteroptera Notonecta maculata Fabricius, 1794 0.80 14 3 4 7 Coleoptera Laccophilus minutus (Linnaeus, 1758) 0.80 29 6 9 14 Heteroptera Sigara stagnallis (Leach, 1817) 0.80 15 4 2 9 Coleoptera Coleoptera Colymbetes fuscus (Linnaeus, 1758) Larvae Hydrobius fuscipes (Linnaeus, 1758)/Limnoxenus niger (Zschach, 1788) 0.80 0.44 35 23 10 5 10 10 15 8 Diptera Culicidae 0.49 27* 15 6 6 Coleoptera Hyphydrus aubei Ganglbauer, 1892 0.55 22 4 4 14 Diptera Chironomus plumosus (Linnaeus, 1758) 0.61 53 11 12 30 Odonata Sympetrum fonscolombei (Selys, 1841) 0.61 46* 17 16 13 Odonata Ishnura pumilio (Charp., 1825) 0.62 33 9 12 12 Diptera Chaoborus spp. 0.64 5* 2 3 0 Heteroptera Corixidae 0.72 52 16 13 23 Coleoptera Cybister lateralimarginalis (De Geer, 1774) 0.73 21 7 6 8 Coleoptera Haliplus spp. 0.73 7 2 4 1 Table 3 continued Order Taxa/stages N Pond occurrence Short Interm Long Coleoptera Laccophilus minutus 0.73 47 15 10 22 Coleoptera Colymbetes fuscus/Rhantus spp. 0.75 17 7 6 4 Odonata Sympetrum meridionale (Selys, 1841) 0.77 16* 8 6 2 Odonata Sympetrum striolatum (Charpentier, 1840) 0.78 21* 10 7 4 Odonata Aeshna mixta Latreille, 1805 0.78 7 2 2 3 Diptera Tanypodinae 0.78 8 2 2 4 Odonata Lestes virens (Charpentier, 1825) 0.78 8* 0 0 8 Odonata Lestes barbarus (Fabricius, 1798) 0.79 13 3 5 4 Coleoptera Agabus spp. 0.79 9 1 3 5 Heteroptera Nepa cinerea 0.79 8 1 1 6 Odonata Sympetrum sanguineum (Mu¨ ller, 1764) 0.79 20* 7 9 4 Coleoptera Berosus spp. 0.80 22 8 7 7 Odonata Ischnura elegans (Vander Linden, 1820) 0.80 24 4 10 10 * Significant over-representation in one pond hydroperiod category at level p \ 0.05, v2 test Fifty-four percent of short hydroperiod ponds were idiosyncratic, 23% of intermediate and 34% of long hydroperiod ponds (Table 4). Although the macroinvertebrate assemblages of ponds with different hydroperiod categories were similarly nested using NODF, with N* we observed the highest nestedness in the macroinvertebrate assemblages of long hydroperiod ponds (N* = 0.41, NODF = 33.66, both indices p \ 0.001) whilst short and intermediate hydroperiod ponds showed similar degrees of nestedness (N* = 0.38, NODF = 32.81, both indices p \ 0.001; and N* = 0.34, NODF = 35.17, both indices p \ 0.001, respectively). All measured environmental variables made a significant contribution to the nested pattern according to the number of Lomolino’s departures (D) from perfect nestedness (Table 5). Pond area produced the lowest D, making the greatest contribution to the macroinvertebrate nestedness (Table 5). Despite the contribution of environmental variables to the nested pattern, variance partitioning showed that spatial descriptors were more strongly associated with nestedness (pure explained variation Adj. R2 = 0.117, p \ 0.001) than environmental variables (pure explained variation Adj. R2 = 0.041, p \ 0.01). Monthly variation in species richness and nestedness Most of the ponds which were sampled monthly differed in hydroperiod between dry and wet years (Table 6). In the dry year, the monthly variation in taxon/stage richness per pond in each of the three hydroperiod categories reached a maximum in April, except for short hydroperiod ponds which had already dried up (Fig. 2). In the wet year, maximum richness occurred in March–April for short hydroperiod ponds, but in May for those with intermediate and long hydroperiods (Fig. 2). The maximum taxon/stage number per pond occurred in long hydroperiod ponds with up to 30 taxa and/or stages (Fig. 2). In ponds of all three hydroperiod categories, we observed an initial increase in the proportion of adult dispersers with respect to larvae just after ponds filled. From then on, adult dispersers and larvae remained in similar proportions until April, when the relative proportion of dispersers increased again (Fig. 2). A higher total number of taxa and/or stages were recorded in the wet year (n = 132) than in the dry year (n = 108). In both years, we observed a similar seasonal shift in the degree of nestedness, with a gradual increase from the time at which ponds first became wet (November and February in the wet and dry years, respectively) to April–May in both years when nestedness decreased or disappeared (Fig. 3). Significantly nested assemblages were found in more months using N* than NODF (Fig. 3). The nested pattern observed was similar using N* and NODF, these differing only when NODF was non-significant (Fig. 3). We obtained a highly similar degree of nestedness for the pool of macroinvertebrate taxa and stages recorded every year (dry year: N* = 0.65, NODF = 61.62, both indices p \ 0.001; wet year: N* = 0.64, NODF = 60.15, both indices p \ 0.001). With one exception, the idiosyncratic character of ponds varied between months as well as between study years (Table 6). While a lower number of ponds were detected as idiosyncratic in the wet year (Table 6), a higher number of taxa/stages were detected as idiosyncratic in the wet year (n = 53) than in the dry year (n = 38). 3011 0.34 5 Short Cry 0.36 8 Long 429 0.43 9 Short 1502 0.47 12 Short 3425 0.49 11 Long Trp 0.50 32 Short 3411 0.57 7 Long 313 0.61 7 Long 308 0.62 12 Short 3431 0.64 12 Long 140 0.64 13 Intermediate 1300 0.65 13 Short 389 0.66 15 Short 3339 0.67 11 Short Fre 0.69 16 Long 3362 0.70 14 Short 630 0.70 16 Short 3286 0.71 26 Long 3413 0.71 15 Short 3345 0.74 9 Intermediate 3432 0.74 18 Long 3315 0.74 26 Short 3410 0.75 7 Long Trj 0.76 15 Long Arm 0.76 2 Short 310 0.77 18 Intermediate Acm 0.77 18 Intermediate 604 0.78 22 Intermediate 321 0.79 18 Short 311 0.79 7 Long Orf 0.79 12 Long 3371 0.80 25 Long 299 0.80 16 Short 3398 0.80 19 Long 2 - Table 4 Idiosyncratic ponds in which macroinvertebrate assemblages departed from the nested pattern obtained with the stage matrix Table 5 Number of departures (D) by Lomolino’s approach calculated with a subset of 80 ponds of the stage matrix Pond N Taxon/stage richness Hydroperiod Ponds ranked by D Pond area 1,083 Richness 1,091 SO4 1,145 O2 1,146 EC 1,149 O.M. 1,155 Turbidity 1,165 Ponds were ranked by the value of each environmental variable. We did not find any randomization of the fixed–fixed null model giving D [ observed (1,000 randomizations, all p = 0.001) EC Electrical conductivity; O2 the dissolved oxygen; O.M. the organic matter Degree of nestedness (N), taxon/stage richness and hydroperiod category of the ponds are also shown Discussion Nestedness in the macroinvertebrate assemblages of Don˜ ana ponds We detected a high degree of nestedness in the macroinvertebrate assemblages of Don˜ ana ponds overall, when surveyed in a single season, despite this being a highly dynamic aquatic system (Go ´ mez-Rodr´ ıguez et al. 2009; Florencio et al. 2009; D´ ıaz-Paniagua et al. 2010). This finding is in line with some previous studies of lentic macroinvertebrate assemblages (e.g., McAbendroth et al. 2005; Baber et al. 2004), but such a pattern is not universal. Urban (2004), for example, did not detect nestedness in the macroinvertebrates of the temporary ponds of a 200-ha region. Nestedness of Don˜ ana macroinvertebrate assemblages implies that a high proportion of taxa (including adult and larval stages) may be found in most of the ponds in the park, with the exception of those species and ponds which are idiosyncratic. The high degree of nestedness found contrasts with the high proportion of idiosyncratic taxa/stages (43%) and the presence of idiosyncratic assemblages in 37% of the sampled ponds. Soininen (2008) detected a similar proportion of idiosyncratic taxa amongst nested diatom assemblages, suggesting good dispersal amongst sites may be a major driver of nestedness. In our highly nested pond network, we also suggest that nestedness is partly driven by differential colonization amongst these temporary aquatic systems; whilst the high number of idiosyncratic taxa/stages and assemblages may be partly due to the presence of specialized environments in some sites. Consequently, high connectivity among sites (i.e., close enough proximity to facilitate dispersal) and high environmental variability are highlighted in this study as important factors maintaining biodiversity in nested systems. Nestedness can be used as a measure of stable conditions in biological communities (Atmar and Patterson 1993), and in our study, such stability can be related to the degree of unpredictability in the composition of macroinvertebrate assemblages which is related to hydroperiod (Florencio et al. 2009). Hence, short duration ponds would be expected to generally exhibit lower nestedness, since only species adapted to sudden desiccation events could persist, than long hydroperiod ponds, where more stable environmental conditions allow more sensitive species to