Ecoregional Characteristics Drive the Distribution Patterns of Neotropical Stream Diatoms
Abstract
We assessed the relative influence of ecoregional features in explaining diatom distribution in the Orinoco river basin. Ecoregions in the Colombian Orinoco can be seen as imprints of the evolutionary history of the basin, for their current biodiversity and physiographic features are the result of the geological and climatic shifts that have occurred since the Tertiary. Thus, they represent an ideal testing ground for studying the interplay between ecological and evolutionary processes shaping diversity patterns of microorganisms, such as diatoms, in the present day. To study this interplay, we compared diatom community composition variance within and among seven ecoregions and assessed the explanatory power of environmental, spatial and historical drivers. This was done by a combination of correlation analyses, multivariate methods and constrained ordinations. We also deconstructed the whole community data set into ecological guilds (low- and high-profile, and motile) to explore their individual response to the contemporary and historical drivers. Taken together, these analyses indicated that contemporary constraints to species occurrence and dispersal, as well as the legacies of historical events, can provide an explanation for the contemporary distribution of diatoms in the Colombian Orinoco. Specifically, we provided evidence showing that both historical legacies and contemporary environmental conditions (temperature, pH, and phosphorus concentration) are interacting to determine diatoms’ distribution. Our results suggest the need to consider ecoregional gradients for unraveling the mechanisms shaping tropical diversity as well as for designing conservation plans.
Abbreviations
-
- CAP
-
- Analysis of Principal Coordinates
-
- dbMEM
-
- Distance-based Moran Eigenvector Map
-
- [E]
-
- Effect of the environmental component
-
- [E + S]
-
- Joint effect of the Environmental and the Spatial component
-
- [Eco + E]
-
- Joint effect of the Ecoregional and the Environmental component
-
- [Eco + S]
-
- Joint effect of the Ecoregional and the Spatial component
-
- [Eco]
-
- Effect of the ecoregional component
-
- GFF
-
- Glass Microfiber Filters
-
- IndVal
-
- Species Indicator Value analysis
-
- N2tQ
-
- Number of days since the last flood event
-
- nEvents2t
-
- Number of flood events
-
- nMDS
-
- non-Metrical Multidimensional Scaling
-
- PRS
-
- Reactive soluble phosphorus
-
- Q.max/Q
-
- Ratio between the maximum and basal flow discharges
-
- RDA
-
- Redundancy Analysis
-
- S]
-
- Effect of the spatial component
-
- SPH
-
- Evolutionary Species Pool Hypothesis
-
- TOC
-
- Total Organic Carbon
-
- TSS
-
- Total Suspended Solids
In the Neotropics, the available literature indicates that microorganism distribution is mostly limited by altitude-driven shifts in the local environment (Nottingham et al. 2018), as it was formerly described by Humboldt and Bonpland in plant communities (von Humboldt and Bonpland 1807). However, a recent study brought to light the existence of clusters or “provinces” of freshwater lakes with unique combinations of diatom species (Benito et al. 2018). Such a pattern hinted that historical events could have constrained species’ dispersal in an evolutionary timeframe, particularly through isolations caused by the orogenesis of the Andean uplifts. Similar distribution patterns associated with these historical events have previously been described in reptile, fish, and plant communities (Hughes and Eastwood 2006, Hoorn et al. 2010). Nevertheless, further evidence is needed to determine if there is a general pattern in Neotropical freshwaters, particularly in river networks, where greater connectivity could erase the clustered pattern because of greater propensity to disperse within these landscapes.
The effect of dispersal limitation on species distribution can occur in ecological and evolutionary timeframes. When dispersal constraints are weak, or occur within an ecological timeframe, the distributional patterns may reflect both the spatially structured local environment and the structure of dispersal pathways. Conversely, when strong and longstanding constraints take place, species dispersal may be impeded within an evolutionary timeframe, which could create distributional patterns that are footprints of the historical contingencies (Brown 1995). Stream networks are an example of structural constraints that can shape the distribution of microorganisms within ecological timeframes (Liu et al. 2013, Widder et al. 2014), but which are themselves also affected by historical contingencies. At the same time, streams are also affected by contemporary environmental factors (e.g., changing hydrology, water chemistry, altered temperature regimes), which comprise additional potential modulators of microorganisms’ distribution (Martiny et al. 2006, Soininen 2012).
Few studies have addressed the relative influence of all of these constraints on diatom distributional patterns (but see Heino et al. 2017). In this regard, Neotropical streams constitute an appropriate setting to disentangle the interplay between contemporary and historical factors. Firstly, compared to Europe and other areas with large and long-lasting anthropogenic impact, the pressures received by this region are still relatively small (though widening Sabater et al. 2017). Secondly, Neotropical basins have complex evolutionary histories that have been determined by major geological and climatic events, of which the Andean uplift played a particularly strong role shaping the current distribution of the ecoregions (Hoorn et al. 2010, Hazzi et al. 2018). Lastly, this diverse ecoregional mosaic is drained by several stream networks that are interconnected within large basins, and therefore, local communities that had once been isolated by past events may currently be re-connected through dispersal, leading to the onset of metacommunity dynamics within an ecological timeframe.
In order to evaluate the major drivers of stream diatom diversity, we compared the variance of community composition among and within ecoregions and assessed the explanatory power of habitat-(stream-) and ecoregion-level variables in diatoms’ compositional patterns. Ecoregions in the Colombian Orinoco can be seen as imprints of the evolutionary history of the basin, for their current biodiversity and physiographic features are the result of the geological and climatic shifts that have occurred since the Tertiary (Goosen 1971, van der Hammen 1974, Hughes and Eastwood 2006, Hoorn et al. 2010); for example, events such as Andean uplifts and glacier retreats shaped unique combinations of river forms and riparian ecosystems at the ecoregional scale (Rull 2008, Hoorn et al. 2010). Therefore, by including the ecoregion type and characteristics, we are able to account for the potential legacies of past geological and climatic events.
We used constrained and unconstrained ordination methods to evaluate the relative influence of contemporary constraints (environmental- and dispersal-related) and possible legacies of historical events on the assembly process and, consequently, on the distribution of stream diatoms. Our predictions about the interplay among these drivers were based on the framework proposed by Martiny et al. (2006). According to this framework, a random assembly is expected if we find a non-spatially structured pattern and lower explanatory power in the three sets of variables (Fig. 1A). Constraints to species dispersal are expected to be the main underlying factors if diatom communities are more similar between neighboring streams or within ecoregions, and if diatom communities’ species compositions are significantly explained by the spatial or ecoregional structure (Fig. 1B). Environmental filtering is expected to occur if diatom communities’ species compositions are mostly explained by the environment data set (Fig. 1C). Finally, it is also recognized that distribution patterns could result from a combination of environmental filtering and dispersal-driven effects. Under this scenario, communities’ species compositions are expected to be explained by an interaction between the environmental and spatial and/or ecoregional components (Fig. 1D).
Methods
Study sites
The Orinoco basin is the third largest basin in South America, and covers an area of about 990,000 km2 in Venezuela and the eastern part of Colombia (Romero Ruíz et al. 2004). The complex geological and climatic history of the basin has created a heterogeneous landscape containing a broad range of ecosystems (Romero Ruíz et al. 2004). An intricate network of rivers and tributaries drains down three large forms of relief that grow in altitude from 100 to 3,500 m a.s.l.: the Orinoco's ancient massifs and shields; its recently raised ridges; and its tectonic depressions and accumulation plains (Stallard 1985). These geological features are the main determinants of the streams’ geomorphological chemical characteristics.
Our sampling was performed at 26 stream segments within an area of about 40,000 km2, in the Colombian Orinoco (Fig. 2; geographical center 3°34′ N, 73°31′ W). The area encompasses an elevation gradient from 300 to 3,400 m a.s.l. and includes a heterogeneous assembly of ecoregions and landscapes. From three to five pristine or near-pristine streams were selected and sampled in seven of these ecoregions, which were a priori identified based on their physiographic features (i.e., the age of major landforms, altitude, vegetation and climate, channel shape and slope, and the main streambed substratum; Table 1).
| Páramo (n = 3) | High-Andean (n = 5) | Piedmont (n = 5) | Alluvial fans (n = 4) | Alluvial terraces (n = 3) | High plains (n = 3) | Guiana shield (n = 3) | References | |
|---|---|---|---|---|---|---|---|---|
| Age of major landform | Early Pleistocene to Holocene | Early Pleistocene to Holocene | Pliocene to early Holocene | Pliocene to Holocene | Pliocene to Holocene | Early Holocene | Miocene | (Van der Hammen 1958, 1974, Goosen 1971, Restrepo and Toussaint 1988, Flórez 2003) |
| Altitude(m.a.s.l) | >3000 | 2000–3000 | 300–1000 | <300 | <300 | <400 | <500 | (Van der Hammen 1974, J.D. González-Trujillo, pers. obs.) |
| Vegetation type | Páramo and subpáramo | Andean forest | Lower tropical forest | Lower tropical forest | Gallery forest | Savanna and Gallery forest | Dry tropical woodland | (Van der Hammen 1974) |
| Climate | Tundra (ET) | Oceanic (Cfb) | Rainforest (Af) | Rainforest (Af) | Rainforest (Af) | Monsoon (Am) | Monsoon (Am) | Köppen-Geiger classification(Kottek et al. 2006) |
| Channel form | Gorge | Mountain | Mountain | Braided | Meandering | Meandering | Meandering | Sensu Petts and Amoros (1996) |
The larger number of streams corresponded to ecoregions on which variability was expected to be higher (J.D. González-Trujillo, pers. obs.). The ecoregions included in the study were (in order of decreasing altitude): (i) The Páramo, endemic to the Neotropics, it is the most recent ecoregion in the Andean hills and is globally recognized as a diversity hotspot as well as a “natural water fabric” (Buytaert et al. 2006); (ii) the High-Andean ecoregion, located immediately below the Páramo, is a belt of cloud forest abundant in fertile zones, with comparatively high precipitation and low temperatures; (iii) the Piedmont ecoregion, reaching down the Andean foothills, it encompasses a belt of rainforest growing on steep slope zones with a warm and humid climate; (iv) the Alluvial fans ecoregion; where the streams leave the Andean hills and join the greater mainstreams, such as the Meta or Guaviare rivers; (v–vii) the remaining three ecoregions are distributed throughout the lowlands (locally known as “Llanos Orientales”). During the early Holocene, the Meta fall divided the lowlands into an elevated zone to the east, which is characterized by a savanna-like climate and vegetation (the High plains ecoregion), and a lower zone to the west that includes the Alluvial terraces and the Guiana shield ecoregions. These two ecoregions are elevated zones with a comparatively flat morphology and distinct vegetation. They were both shaped during the Andean uplifts, although the former is more ancient, and has a geology that is largely influenced by the Guiana shield reliefs.
Environmental characterization
Each stream segment, spanning from 20 to 60 m long, was characterized by its hydrology, substratum type and distribution, physical water conditions, and water chemistry. Long-term hydrological variables were estimated following the modified rational method (Témez 2003). The variables calculated were: i) the number of days elapsed after the last flood event (defined as twice the basal flow discharge (“Nt2Q”); ii) the number of flood events (“nEvents2t”); and iii) the ratio between the maximum and basal flow discharges (“Q.max/Q”). These variables were derived from the daily discharge estimation, which was a function of the total precipitation, the basin area and associated land uses, the concentration time, and the runoff coefficient (see Appendix S1 in the Supporting Information for an extended explanation on the computation).
Physical and chemical variables were measured before (November 2016), during (January–February 2017) and after (January–February 2018) the sampling campaign. Samplings were performed during the low water period, as diversity is maximized during this period. Instantaneous water flow was estimated in the three riffles using water depth and flow velocity, which were measured by means of three cross-sections with measurement intervals of 15 cm. Flow velocity was measured using a digital flow meter (Schiltknecht—MiniAir 20). Conductivity, pH, oxygen, and temperature were recorded using a Hanna HI98194 water quality meter. Canopy shading (%) was estimated analyzing a series of upward photographs taken using a fisheye lens. The luminosity and white balance of each photo were manually adjusted to color leaves in black and sky in white. The proportion between black and white pixels was used to estimate the percentage of canopy shading. Finally, considering the high heterogeneity among ecoregions in terms of substratum type, we reclassified them as low- (bedrock and boulders), mid- (cobbles and pebbles), and high mobility (gravel and sand) substratum before including them in RDA models.
One liter of water was collected, filtered through 0.7 μm glass fiber filters (Whatman GF/F, Kent, UK) and stored in a frozen state to be used for chemical analyses. Ammonium and nitrate concentrations were determined in a Dionex ICS-5000 ion chromatograph (Dionex Corporation, Sunnyvale, CA, USA). PRS concentration was determined colorimetrically using a fully automated discrete analyzer Alliance Instruments Smartchem 140 (AMS, Frépillon, France). The total suspended solids (TSS) were determined by filtering 500 mL of water in a pre-weighted GFF filter, drying the filter for 1 h at 105°C, and weighing the resulting contents. The environmental variables’ means and coefficient of variation in each ecoregion are summarized in Table S1 in the Supporting Information.
Diatom sampling
We sampled diatom communities at three different riffle sections of each stream segment. At each riffle section, we collected 8 cm2 of surface, brush-scraped algal material from 30 boulders and cobbles. In the case of streams from Guiana shield and High plains, where boulders and cobbles were scarce, we also took samples from bedrock, pebbles, and sand. Algal material was pooled by riffle section (= 3 samples per stream segment) and subsequently preserved in a Transeau solution. In the laboratory, the organic material from samples was cleaned using hydrogen peroxide. Clean diatom frustules were mounted on permanent slides using a Naphrax® medium, the slides were then observed under a 1000x light microscope and identified at the finest level possible using specialized monographs (Krammer and Lange-Bertalot 1986, 1991, Metzeltin and Lange-Bertalot 2007, Bellinger and Sigee 2015). At least 400 valves were counted in each slide, and counts processed as relative abundances.
We identified a total of 297 diatom taxa (Table S2 in the Supporting Information). In several cases, the identity of the taxa did not match the monograph descriptions (presumably new species); these were classified as affine to (aff.) and were recorded as a separate taxonomic group. To alleviate taxonomic bias between samples, we assembled the 297 morphospecies into 72 taxonomic groups (Table S3 in the Supporting Information). Finally, we assigned each identified taxa to their corresponding ecological guild (low-profile, high-profile, motile; Passy 2007, Rimet and Bouchez 2011).
Data analysis
We used a Hellinger-transformed community dataset for all the analyses, and standardized all environmental variables to mean = 0 and variance = 1, as recommended by Legendre and Legendre (2012). To avoid the possible influence of singletons, we summed up sample counts and used a matrix of relative abundances per stream in the subsequent analyses (abundance-data matrix of 26 rows/streams and 72 columns/taxonomical entities). We calculated beta diversity using the Hellinger distance for abundance-based data to maintain comparability with the redundancy analysis. Additionally, we performed all analyses for the whole community dataset, as well as in the separated datasets of the low-profile, high-profile, and motile guilds. All analyses and graphical outputs were done using the R Statistical software v 3.5.1 (R Core Team 2018) and the “ggplot2” package (Wickham 2016).
Elevation-driven diversity and local environment patterns
In mountain systems, changes in species richness and community composition are generally linked to elevation, following a monotonical function (i.e., Nottingham et al. 2018). We therefore tested the existing monotonic relationships between elevation and diatom alpha and beta diversity, as well as between elevation and water chemistry and physiographical characteristics. We used the Spearman′s rank tests for taxa richness, and the Shannon index (alpha diversity) and Mantel tests (9,999 permutations) for environmental and community dissimilarity (beta diversity).
Ecoregional diversity patterns
We used two analyses to evaluate community composition dissimilarity among and within ecoregions. First, we performed a non-Metrical Multidimensional scaling (nMDS) ordination, followed by an ANOSIM analysis (Clarke 1993) to explore differences in diatom communities between ecoregions. Second, we used a canonical analysis of principal coordinates (CAP; Anderson and Willis 2003) to test for differences among ecoregions. The CAP test finds axes along the multivariate space that best discriminate a priori groups. In our case, stream ecoregion types were tested in order to find among-ecoregion differences in community composition, using 9,999 permutations. Finally, we used an analysis of indicator species (IndVal; Cáceres and Legendre 2009) to look for the preferences of species in each ecoregion. The IndVal analysis is based on the concepts of specificity (highest values when the species is present in only one ecoregion) and fidelity (highest when the species is present in all streams of a given ecoregion). A high indicator value, which ranges from 0 to 1, is obtained by the combination of high specificity and fidelity. We performed IndVal analysis using the function “multipatt” from the indicspecies package (Cáceres and Legendre 2009). CAP calculations were performed using the function ‘CAPdiscrim’ from the BiodiversityR package (Kindt and Coe 2005) and the ANOSIM test was done using the function “anosim” from the VEGAN package (Oksanen et al. 2013).
Drivers of diversity and distribution patterns
We used a Redundancy Analysis (“RDA”; Legendre and Anderson 1999) to explore the main drivers (environmental, spatial, and biogeographical) of community composition. We used Moran's eigenvector maps (MEM; Legendre and Legendre 2012) to model the spatial structure, using the function “create.dbMEM.model” from the “adespatial” package (Dray et al. 2018). We also created a matrix of seven dummy variables, representing each ecoregion type, to model the dependence of community structure on the ecoregional characteristics.
After constructing the first models, we conducted a forward-selection procedure to reduce the number of explanatory variables and obtain more parsimonious models. We included variables with a variance inflation factors (VIF) inferior to 3 to prevent multicollinearity. The selection procedure was performed separately for the variables of the environmental and the spatial components, as recommended by Borcard et al. (2018), using the “ordiR2step”. The significance of all models was tested using a permutation-based ANOVA (999 permutations). Finally, we conducted a variatance partitioning analysis (Borcard et al. 1992) using the most parsimonious models.
Results
Our sampling captured a representative fraction of the Orinoco diatom flora, as evidenced by the species accumulation curves performed when streams were grouped in their corresponding ecoregions (Fig. S1 in the Supporting Information), which approached asymptotes in most of the ecoregions.
Elevation-driven diversity and local environment patterns
The diatom community did not exhibit any increasing or decreasing alpha diversity trends along the altitudinal gradient. Neither their taxa richness (Spearman test: rho = −0.144, S24 = 3347.1, P = 0.482) nor their Shannon diversity (Spearman test: rho = −0.101, S24 = 3221.1, P = 0.423) were significantly correlated with elevation. The dissimilarity among diatom communities (beta diversity) increased as differences in elevation increased (Mantel test: ρ = 0.468, P < 0.0001; Fig. S2a in the Supporting Information). The dissimilarity of low-profile (Mantel test: ρ = 0.468, P < 0.0001; Fig. S2b) and high-profile guilds (Mantel test: ρ = 0.425, P < 0.0001; Fig. S2c) also increased with elevation. Motile diatom species were the group which varied less as differences in elevation increased (Mantel test: ρ = 0.215, P < 0.01; Fig. S2d).
Contrary to the observed patterns in community dissimilarity, environmental dissimilarity between streams was not significantly linked to elevation differences (Mantel test: ρ = 0.0034, P = 0.48). Streams from different ecoregions at similar elevation ranges can exhibit very distinct environmental characteristics. This was consistent in the separate chemical (Mantel test: ρ = 0.0093, P = 0.284) and physiographic variables (ρ = 0.0085, P = 0.414) evaluations. By analyzing each variable, we found that only five variables were significantly correlated with elevation. Temperature significantly decreased (Spearman test: rho = −0.855, S24 = 5425.4, P < 0.0001; Fig. S3a in the Supporting Information), while saturated oxygen (Spearman test: rho = 0.849, S24 = 442.58, P < 0.0001; Fig. S3b) and pH (Spearman test: rho = 0.531, S24 = 1370.7, P < 0.01; Fig. S3c) increased significantly. A significant increase was also observed in water conductivity (Spearman test: rho = 0.429, S24 = 1668.8, P = 0.028) and PRS (Spearman test: rho = 0.416, S24 = 1708.9, P = 0.035).
Ecoregional diversity patterns
The occurrence of distinct regional diatom pools within the basin was evidenced by the ANOSIM and CAP results, which partially supported hypotheses H2b and H4b (Fig. 1). Most ecoregions included in the analysis hosted a distinctive set of diatom taxa (Figs. 3 and 4). Differences were consistent when considering the complete dataset (ANOSIM R = 0.6323, P < 0.001, 9,999 permutations; Fig. 3a), as well in the low-profile (ANOSIM R = 0.4675, P < 0.001, 9,999 permutations; Fig. 3b), high-profile (ANOSIM R = 0.5962, P < 0.001, 9,999 permutations; Fig. 3c) or motile (ANOSIM R = 0.2505, P < 0.001, 9,999 permutations; Fig. 3d) guild datasets.
Figure 4 summarizes the results of the CAP analyses. The percentages of correct classification were similar (about 80%) for the environmental and the community datasets, suggesting a partial match between the community composition and the environmental conditions. The differences were observed when the contemporary environmental conditions in the ecoregions (Fig. 4a) were compared to their community composition(Fig. 4b). Some ecoregions can be confidently classified according to their diatom communities but not according to their physiography and/or water chemistry (e.g., Páramo, High Andean or Piedmont), while others (High plains and Alluvial fans) can be better classified according to their environmental conditions.
The IndVal analysis revealed 52 morphospecies to be potential indicators of ecoregional distribution (Table 2), being Hannaea arcus and Fragilaria capucina var. vaucheriae the most indicative for the Páramos, Planothidium sp. pl. and Cocconeis sp. pl. for the High-Andeans, Synedra sp. pl. and Gomphonema micropumilum for the Piedmonts, Frustulia rostrata and Navicula cryptocephala for the Alluvial terraces, several Eunotia and Gomphonema gracile for the High plains, and Cymbellopsis sp. pl for the Guiana shield. There were not species assigned as indicators for the Alluvial fans. Additionally, 17 and 2 genera were potential indicators of combinations of two or three ecoregions, respectively (Table 2).
| Ecoregion | Morphospecies | Specificity | Fidelity | IndVal | P |
|---|---|---|---|---|---|
| Páramo | Hannaea arcus | 0.9057 | 1 | 0.952 | 0.001 |
| Fragilaria capucina var. Vaucheriae | 0.8703 | 1 | 0.933 | 0.001 | |
| Cymbella sp. | 0.9412 | 0.8889 | 0.915 | 0.001 | |
| Tabellaria floculosa | 0.6466 | 1 | 0.804 | 0.001 | |
| Reimeria uniseriata | 0.8262 | 0.7778 | 0.802 | 0.001 | |
| Gomphonema sp. | 0.7185 | 0.8889 | 0.799 | 0.009 | |
| Encyonema minutum | 0.6718 | 0.8889 | 0.773 | 0.001 | |
| Achnanthidium affine | 0.7362 | 0.6667 | 0.701 | 0.032 | |
| Fragilaria sp. | 0.942 | 0.4444 | 0.647 | 0.007 | |
| Epithemia cf. turgida | 0.9655 | 0.3333 | 0.567 | 0.01 | |
| High-Andean | Planothidium sp. | 0.8263 | 0.6429 | 0.729 | 0.002 |
| Cocconeis sp. | 0.8198 | 0.5714 | 0.684 | 0.004 | |
| Piedmont | Synedra sp. | 0.8797 | 0.4615 | 0.637 | 0.006 |
| Gomphonema micropumilum | 0.7712 | 0.4615 | 0.597 | 0.011 | |
| Caloneis sp. | 1 | 0.3077 | 0.555 | 0.005 | |
| Alluvial terraces | Frustulia rostrata | 0.8719 | 0.8333 | 0.852 | 0.001 |
| Navicula cryptocephala | 0.9761 | 0.25 | 0.494 | 0.03 | |
| High plains | Eunotia curvula | 1 | 0.6667 | 0.816 | 0.001 |
| Eunotia ventricosa | 1 | 0.6667 | 0.816 | 0.001 | |
| Gomphonema gracile | 0.8739 | 0.6667 | 0.763 | 0.001 | |
| Aulacoseira sp. | 1 | 0.5 | 0.707 | 0.001 | |
| Stenopterobia densestriata | 0.9091 | 0.5 | 0.674 | 0.001 | |
| Aulacoseira granulata | 1 | 0.3333 | 0.577 | 0.008 | |
| Eunotia rabenhorstiana | 1 | 0.3333 | 0.577 | 0.01 | |
| Guiana shield | Cymbellopsis sp. | 0.8566 | 0.5556 | 0.69 | 0.001 |
| Páramo & High-Andean | Encyonema silesiacum | 0.9214 | 0.8261 | 0.872 | 0.001 |
| Fragilaria capucina | 0.9952 | 0.6522 | 0.806 | 0.001 | |
| Diatoma mesodon | 1 | 0.3043 | 0.552 | 0.007 | |
| High-Andean & High plains | Eunotia exigua | 1 | 0.35 | 0.592 | 0.01 |
| Piedmont & Alluvial fans | Gomphonema lagenula | 0.9446 | 0.9048 | 0.924 | 0.001 |
| Nitzschia palea | 0.8795 | 0.9524 | 0.915 | 0.001 | |
| Gomphonema parvulum | 0.8401 | 0.5714 | 0.693 | 0.004 | |
| Amphora cf. montana | 0.8846 | 0.4286 | 0.616 | 0.012 | |
| Alluvial terraces & Guiana shield | Encyonopsis frequentis | 0.8747 | 0.9524 | 0.913 | 0.001 |
| Encyonema sp. | 0.8247 | 0.7619 | 0.793 | 0.002 | |
| Eunotia rhomboidea | 1 | 0.2381 | 0.488 | 0.045 | |
| Alluvial terraces & High plains | Frustulia sp. | 0.8093 | 0.9444 | 0.874 | 0.001 |
| Chamaepinnularia sp. | 0.8066 | 0.6111 | 0.702 | 0.001 | |
| Guiana shield & High plains | Eunotia bilunaris | 0.8615 | 0.3333 | 0.536 | 0.018 |
| Piedmont & High plains | Adlaphia sp. | 0.8257 | 0.3684 | 0.552 | 0.014 |
| Alluvial fans & Piedmont & High-Andean | Achnanthidium eutrophilum | 0.9893 | 0.5714 | 0.752 | 0.002 |
| Pinnularia sp. | 0.9398 | 0.4839 | 0.674 | 0.007 |
Drivers of diversity and distribution patterns
The diatom community structure was partially modeled by the three sets of explanatory variables (environmental, spatial structure and ecoregional identity; Table 3, Table S4, and Fig. S4 in the Supporting Information). The RDA model that included all diatom taxa had the largest variance (Radj 9,61 = 0.526, P < 0.0001), followed by the models of the high-profile (Radj 9,61 = 0.526, P < 0.0001), low-profile (Radj 9,61 = 0.451, P < 0.0001), and motile (Radj 9,61 = 0.309, P < 0.0001) taxa. In all four models, the largest constrained variance was explained by the ecoregional component [Eco], and by its joint effect with the environmental component [E + Eco]. The pure effects of the environmental [E] and spatial [S] components, and its respective interaction [E + S], were relatively low compared to the former components (Fig. 5).
| Whole community | Low-profile | High-profile | Motile | |
|---|---|---|---|---|
| Spatial structure | dbMEM – 1,2,5,6 | dbMEM1,2,4,7 | dbMEM1,5 | dbMEM1,6 |
| Environmental conditions |
pH, Temperature Qmax/Q Low mobility substrate PRS N2tQ Conductivity TOC |
pH, Temperature N-NH4 Low mobility substrate PRS N-NO3 N2tQ Conductivity |
pH, Temperature Qmax/Q NH4 N-NO3 TOC |
pH, Temperature Q Low mobility substrate nEvents2t PRS |
- dMEM, distance-based Moran Eigenvector maps; PRS, reactive phosphorus concentrations; Qmax/Q, the ratio between the maximum and basal flow discharges; N2tQ, the number of days elapsed after the last flood event (defined as twice the basal flow discharge); nEvents2t, the number of flood events; TOC, total organic carbon.
The forward selection procedure retained the first and last MEM eigenvectors, indicating that community structure was explained to a similar degree by the streams’ spatial structure [S] at large and local scales (Table 3). Regarding the environmental component [E], the forward selection procedure retained pH and temperature in all the datasets (Table 3). The first axis represented a gradient of pH, while the second and third axes were distinctly correlated to temperature. The hydro-morphological variables and nutrients were more highly correlated with the second and third axes respectively (Table S4). The other environmental variables were highly correlated with the second and the subsequent axes, depending on the dataset used.
Discussion
Contrarily to previous findings, changes in diatom diversity and distribution were partially uncoupled from the Andean elevational gradient, but greatly explained by the basin ecoregional gradient. Overall, our study provides several evidences in support of Martiny et al.'s (2006) fourth hypothesis (Fig. 1d, H4b) as the primary explanation for the observed patterns in community composition. It means that the interplay between contemporary processes and past historical events is driving the distribution of diatoms in the Colombian Orinoco basin. Consequently, the species composition of a given community is shaped by the effect of dispersal limitation within an evolutionary timeframe and the effect of environmental filtering within an ecological timeframe. Particularly, our findings suggest that the historical legacies associated with the Andean uplifts would have shaped distinct regional pools of diatom taxa, promoting diversity at the regional scale.
Since von Humboldt and Bonpland (1807), elevation-driven changes in temperature have been stated as one of the major drivers of biodiversity patterns (Jackson 2009). Nottingham et al. (2018), for instance, recently showed that alpha and beta diversity patterns of microbes, fungi, and plant communities coincide with an altitudinal gradient in the Peruvian Andes and concluded that temperature was the main driver of microorganism diversity and distribution. Our results partially conformed to these findings. Contrarily to the Nottingham et al. (2018) study, stream environmental features were spatially constrained according to the ecoregion type, rather than changing monotonically with elevation.
Spatially constrained environments can obscure longitudinal, latitudinal, and/or elevational patterns of diversity when the environmental variability does not change linearly in an spatial dimension(Passy et al. 2018). Our results are in line with this statement, since we found that ecoregional, rather than elevational drivers, are underlying the distribution of diatoms in Neotropical streams. Ecoregion type outperformed elevation in explaining the environmental features of the streams (i.e., CAP 83.1% of correct stream classification based on ecoregions). We observed that, regardless of their spatial proximity, diatom communities from streams within the same ecoregions were more similar to each other than to those from different ecoregions.
Uncovering this pattern provides a potential mechanism for a better understanding of diatom distribution, and possibly diatom biogeography in the Neotropics. The ecoregionaly constrained distribution of diatom taxa suggests that historical constraints of dispersal may have been the major determinants of their present-day distribution. Although a similar distribution has already been observed in Neotropical lakes diatoms (Benito et al. 2018), it is still surprising to observe this level of discreteness in stream networks, since their dispersal is limited to a lower degree than lakes (Vilmi et al. 2017).
The observed concordance between the distributions of diatoms and ecoregions can be explained by the evolutionary species pool hypothesis (SPH; Pither and Aarssen 2005), which states that the longer a set of environmental conditions persists in time and space, the greater the opportunity for the evolution of adapted species (Pither and Aarssen 2005). This theory predicts that several pools of species might result from the conjoint action of events limiting dispersal and the filtering of species according to their tolerances. This interplay has already shown to be the main driver of biogeographic provincialism of Lupinus plants (Hughes and Eastwood 2006). Although further evidences on diatom phylogenetic relationships and radiation events would be needed to confirm this mechanism, the higher variance of the [Eco + E] interaction (Fig. 5), and the occurrence of ecoregionally constrained pools of diatom taxa, provide initial supporting evidences. We have recently observed a similar pattern in the Orinoco basin assessing invertebrate communities (González-Trujillo et al. 2020). In this regard, the concordance in distributional patterns between two biological communities (diatoms and invertebrates) with distinct dispersal abilities may imply that historical legacies play a key role determining the present-day distribution of freshwater biota in Neotropical basins.
Historical legacies alone cannot fully explain diatom distribution. This is evident when comparing the models for the different diatom guilds, which suggests that different drivers may be underlying the distribution of each guild. For example, the non-spatially structured distribution of motile diatoms suggests that other underlying processes, such as mass effects or stochastic forces (Jamoneau et al. 2018), must have important roles. The motile guild collects species able to use the available resources and support disturbances (Dong et al. 2016, Passy 2016, Falasco et al. 2019), which makes them potentially successful under a multiplicity of circumstances, and rather independent of any spatial structure. On the other hand, the distribution of low-profile species is expected to be weakly structured in the space because of their high dispersal ability, which is mostly relevant at large scales (Passy 2016). Still, these species have preferences for high mountain sites because of their tight attachment abilities and preferences for poor nutrients and capacity to resist fast water velocities (Bottin et al. 2014, Dong et al. 2016), and this may support their preference for high-altitude ecoregions. The high-profile species, which assembles species remaining erect or colonial, spanned mostly through low altitude sites, as much as it has been observed under other situations (Bottin et al. 2014). Overall, there is a certain degree of discreteness in the distribution of the different guilds which may support the occurrence of biogeographical boundaries on the contemporary distribution of diatoms in the Neotropics. Nonetheless, further research is needed to confirm if the drivers of diatom biogeography can change as a function of other functional traits (e.g., growth forms).
Ecoregional characteristics can also exert a high influence on ecological timeframes, acting as large-scale drivers of local environmental conditions (Varanka and Luoto 2012, Neff and Jackson 2013, Vilmi et al. 2017). In this study, the stream's relevant physical and chemical characteristics, such as water pH, temperature, conductivity, and soluble reactive phosphorus (Table 3) were related to some of the ecoregions. Of these, pH provided the higher degree of differentiation, and separated the High Andean and Páramo regions from the alluvial terraces and the Guiana shield (Fig. S4). The RDAs identified diatom taxa such as Hannaea arcus, Fragilaria capucina, or Tabellaria flocculosa, which are common in moderately low pH conditions, and others, such as Frustulia rostrata or Navicula cryptocephala, which are more common in circumneutral pHs (Whitmore 1989, Poulíčková et al. 2010), a Result that is also supported by the IndVal analysis, which found the former morphospecies as the best indicator taxa of páramos streams and the latter as the best indicator taxa of streams draining alluvial terraces. These findings are in consonance with the fact that algae living under low pH conditions have developed adaptative systems, such as the use of active pumps to prevent H+ ions entering the cell (Seckbach and Oren 2007).
Concluding remarks
Here we have provided evidence that ecoregional characteristics drive the community composition of Neotropical stream diatoms. Historical legacies, particularly, might explain the observed ecoregionally constrained distribution at the regional scale. Analogous findings have already been found for microbial biogeography in high-latitude streams (Vyverman et al. 2007, Verleyen et al. 2009, Liu et al. 2013, Widder et al. 2014). However, to our knowledge, our study is one of the firsts to assess the main drivers underlying microorganism distribution in Neotropical streams. Understanding the main drivers for a basal-diversity pattern may help define the subsequent conservation steps not only for the algal and diatom flora but also for the ecosystems hosting them. Our overall results suggest that, not only future studies, but also the delimitation of zones of conservation priority should consider the type and the extent of the different ecoregions.
We would like to thank Liz Alonso, Katterin Rincón, Gabriela Córdoba, and Juan Pablo Álvarez for their help in the field and the laboratory. We extend our gratitude to the “Playa Güio” family, to “El Paraiso” family, and to Miguel Rodriguez and family, for hosting us during our field work. JDGT wants to thanks to Secretariat of Universities and Research from Generalitat de Catalunya and European Social Fund for his FI fellowship (2019 FI_B1 00210) as well as the funding from CERCA program. This work was supported by the Dirección de Investigación Sede Bogotá and the “Convocatoria nacional de proyectos para el fortalecimiento de la investigación, creación e innovación de la Universidad Nacional de Colombia 2016-2018” of the Universidad Nacional de Colombia (DIB project no. 34856); Departamento Administrativo de Ciencia, Tecnología e Innovación “Colciencias”, PhD project “Linking functional diversity patterns of algae and invertebrates to scale-dependent constrains of rivers from the Orinoco basin”.
Author contributions
JDGT and SS conceived the research; JDGT gathered the data at field; EDPG identified diatom taxa; JDGT analyzed the data; JDGT wrote the first draft. All authors discussed the research as it developed and edited the manuscript.
