Disturbances – temporally discrete events that disrupt the structure of an ecosystem by altering conditions and resource availability – have a major influence on ecological patterns (Sousa, 1984). The ecological consequences of a disturbance will depend on the balance between its magnitude and the sensitivity of the species involved. For gradual and cumulative disturbances, such as habitat loss, climate change, pollution and overexploitation, species are expected to show different vulnerability levels to these impacts. That is, extinction is not random regarding species’ ecological features so that the presence of certain traits makes some species more prone to extinction than others (Olden et al., 2008; Purvis et al., 2000). However, for abrupt, massive disturbances, species loss may be random, because the animals cannot adapt to sudden, unpredictable impacts. The mass extinctions are examples of such non-selective extinctions (Heard and Mooers, 2002; Jablonski, 2005). The main differences between historical catastrophes that led to mass extinctions and local disasters are their extent, but in both cases there is potential to cause non-selective extinctions.

Either way, the different facets of biodiversity (taxonomic, functional, genetic and phylogenetic) (Pool et al., 2014) are likely to be affected, and the impact of disturbance may be reflected across a range of spatial scales, including local (within-community; α), among community diversity (β-diversity), and regional (γ) diversity (Collins et al., 2000; Melo et al., 2011). Many studies have asked how disturbance alters local taxonomic α-diversity (e.g. species richness) (Magurran and McGill, 2011; McKinney and Lockwood, 1999; Reis et al., 2016). However, the question of how abrupt disturbance alters the compositional similarity (taxonomic β-diversity) at larger spatial scales, and the consequences of such shifts for functional β-diversity, remains less explored.

Changes in β-diversity arise as a result of two processes – species replacement (turnover) and differences in local richness (nestedness) (Baselga, 2010). These processes also underpin shifts in functional β-diversity, or the replacement (turnover) and loss (nestedness) of functional traits contributed by species in neighbouring communities. Taxonomic β-diversity can increase in response to disturbance through (i) the loss of shared species from two or more communities, resulting in more diverse communities, an outcome known as subtractive heterogenization; or (ii) by the addition of species in only one community in a pairwise comparison, an outcome known as additive heterogenization (Socolar et al., 2016). Alternatively, disturbance can cause a decrease in taxonomic β-diversity if there is a loss of (iii) rare species from one community, subtractive homogenization, and/or (iv) rare species become more widespread, additive homogenization (Socolar et al., 2016). These different patterns of increase or decrease in taxonomic β-diversity will have varying consequences for ecosystem function, depending on the functional roles of the species involved (McGill et al., 2015; Meynard et al., 2011). For instance, extinction (i.e., subtractive) scenarios, imply reduced functional redundancy or reduced overall function (subtractive heterogenization), whereas subtractive homogenization can lead to a loss of ecosystem function that was exclusive to one of the communities (Villéger et al., 2010). Empirical evidence shows that the loss of functional dissimilarity typically exceeds the loss of taxonomic dissimilarity, with assemblages becoming functionally similar to one another (Villéger et al., 2014).

The primary approach to quantifying shifts in regional taxonomic and functional β-diversity is to examine species composition in the constituent assemblages before and after the disturbance. However, this exercise requires extensive and comparable information on species composition pre- and post-disturbance, which may be unavailable within the reasonable timeframe. An alternative approach is to draw on existing information about the species composition of the regional assemblages, and then simulate the impact of the disturbance, assuming different scenarios of change. An emblematic case of local, destructive disturbance was the Bento Gonçalves dam burst, known as the ‘Mariana disaster’, that occurred in November 2015, in the Doce River Basin, Minas Gerais, southeastern Brazil (Fig. 1). This environmental catastrophe released c.50 billion litres of heavy metal-rich waste into the Doce Basin. Villages were destroyed and the river ecosystem downstream collapsed (Escobar, 2015; Fernandes et al., 2016). It seems likely that the disaster resulted in the local extinctions of many riverine species, such as fishes. As such, it provides a case study for investigating the consequences of abrupt changes in regional taxonomic and functional β-diversity in relation to the regional species pool.

Fig. 1.

Seven hydrographic basins used in this study showing the Doce River (blue line) and its neighbour basins. BA ES coast: this basin comprises part of the coast from Bahia (BA) and Espírito Santo coasts. ES coast: Espírito Santo coast. The blue map shows the variations in species richness and the green map shows the variations in functional richness. The Pfafstetter drainage basin classification has been broadly used to classify freshwater environments for research and management purposes, e.g. “Agência Nacional das Águas – ANA” in Brazil (National Agency of Waters) and by the International Union for Conservation of Nature – IUCN (ANA, 2017; IUCN, 2016).

(0.35MB).

In this study, we use a stochastic simulation to evaluate how the components of regional taxonomic and functional β-diversity (turnover and nestedness), and functional richness of freshwater primary ray-finned fishes (Actinopterygii fishes restricted to freshwater environments) may have changed in response to the ‘Mariana disaster’. We consider different scenarios of species loss. Primary freshwater fishes are defined as those intolerant to a salinity level above 0.5 grams per litre of water (Myers, 1949), and as such, they are highly sensitive to environmental changes of that magnitude. Therefore, because of the extent, pervasiveness and lethality of the ‘Mariana disaster’, we consider that it may have caused non-selective extinctions of fishes, different from continuous impacts that are more likely to be biased towards some specific traits of species.

Our simulations suggest that under abrupt and stochastic levels of extinction, a process known as subtractive heterogenization – i.e. an increase in β-diversity by the loss of shared species from two or more communities, when many native species become rarer and a few become extinct (Socolar et al., 2016) – could have affected the patterns of taxonomic β-diversity between those seven basins. Meanwhile, for functional beta diversity (functional β-diversity), we predicted a process of functional homogenization, in which the basins would become more functionally similar (Clavel et al., 2011). We found that the existence of exclusive species within each basin is common (57% of all species) and that basins share different proportions of functional diversity. For instance, before the disaster, the Doce Basin was fourth in species richness (n=139), but the second basin in functional richness (69%). This pattern accounts for differences in how species losses affected the amount and direction of both taxonomic and functional diversity through a differential contribution of nestedness (67–81% for the functional diversity) and turnover (87–81% for taxonomic) components.

The change in taxonomic dissimilarity that might have resulted from local extinctions is dependent on the level of rarity or commonness of the species present in the communities (Baiser and Lockwood, 2011). For example, the Doce Basin had 31 species that were shared between one other basin (less ubiquitous species, Fig. S1; column 2). Therefore, a removal of these species from the Doce Basin would make them exclusive species in other communities. The predicted increase of the turnover component of taxonomic β-diversity from zero extinction (before the disaster) to 50% of extinction scenario (Fig. 3B) could be explained by the removal of the less ubiquitous species (Fig. S1; columns 2 and 3) that were present in the Doce Basin before the disaster. However, increasing the possibility of losing the more widespread species, e.g. the species that occur in Doce Basin and also occur in 4 or more neighbour basins (Fig. S1; columns 4–7), decreased the turnover, as the gradient of extirpations intensified.

Functional diversity aids the understanding of the ecological consequences of environmental disturbances by considering changes in the diversity of biological species traits and their ecosystem function. The lower dissimilarity levels of functional β-diversity (0.33) compared with taxonomic β-diversity (0.75) could be explained by the presence of species with similar combinations of traits and life history strategies (Pool and Olden, 2012). The predicted change in frequency (i.e., density) of functional β-diversity values found for all simulations (decreasing for lower density values and increasing for higher density values, Fig. 5) indicates that pairs of basins that possessed lower values of functional β-diversity before the disaster could have higher values if the disaster actually led to the simulated extinctions. These results suggest that catastrophic disturbances may remove exclusive functional attributes from affected areas, whilst functions that are common in adjacent communities or basins may remain. Therefore, the loss of unique species occupying specific places at the multidimensional functional space from the Doce Basin may disrupt the ecosystem services provided by freshwater systems.

We found a greater contribution of functional nestedness (increased by 15%) than taxonomic nestedness (increased by 6%). As functional diversity is a more reliable indicator of ecosystem stability and functioning (Villéger et al., 2014), the predicted increase in the contribution of nestedness to the total taxonomic and functional β-diversity emphasizes the contribution of local diversity (α) to regional diversity (γ). The net result is that if the environment loses rare or ubiquitous species (a shift in taxonomic diversity) functional diversity is impacted. Thus, any extinctions that may have resulted from the Mariana disaster will have permanently affected several aspects of biodiversity at a regional scale.

Taxonomic and functional changes of fish faunas over time are highly dynamic, due to both natural and anthropogenic processes (Pool and Olden, 2012). Freshwater systems, such as in Brazil and many other regions worldwide, are threatened by biological invasions (Assis et al., 2017), water salinization (Brito and Magalhães, 2017), pollution (Costa and Barletta, 2016), habitat degradation (Dudgeon et al., 2006), fisheries overexploitation (Alho et al., 2015), and climate change (Ficke et al., 2007). Predicting change using multiple facets of biodiversity, as we propose here, may be key to anticipate management measures of these irreplaceable freshwater resources, and to protecting global food security, given that riverine communities are the main source of protein in many developing countries (McIntyre et al., 2016). The particular case of the ‘Mariana disaster’, one of the greatest environmental disasters in the history of Brazil (Escobar, 2015) is typical of this scenario. The disaster had far-reaching environmental, social and economic impact, as entire human populations relied exclusively on those resources for a living (Ramos et al., 2017).

Massive and stochastic impacts are thought to cause both random and selective extinctions of species because species do not evolve adaptation to such abrupt events (Heard and Mooers, 2002; Jablonski, 2005). Therefore, in the particular case of a lack of consistent information post-disturbance, our random simulation on fish extinctions following the ‘Mariana disaster’ has the value of pointing to likely consequences of the loss in compositional changes among basins at a geographical scale. In fact, it is unthinkable that the inrush of c.50 billion litres of heavy metal-rich waste in the Doce Basin happened without severe consequences to species’ populations (Escobar, 2015; Fernandes et al., 2016). Fishes are ecological engineers (Helfman et al., 2009) and the primary species are intolerant to salt or brackish water (Albert and Reis, 2011; Myers, 1949). Therefore, all species were likely to have been affected by a disturbance of such magnitude. Selective extinction owing to trait filtering may also be expected, but mainly in continuous, long-term impacts (Olden et al., 2008, 2007). Accordingly, our approach can be applied to different types of anthropogenic impacts, either those which have happened or are due to, and which are either abrupt or gradual. In the latter case, however, the simulation should allow for a non-random extinction, such as by pre-defining different levels of susceptibilities among species based on relevant traits. Clear articulation of the assumptions used in such simulations is essential.

Disturbances are analysed mainly at the level of populations (Melo et al., 2011; Ricklefs, 2008), but little is known about how these changes rebound at the level of communities and among communities, as explored here. In this exercise, we address this question by probing an emblematic environmental catastrophe in Brazil, the ‘Mariana disaster’ in the Rio Doce River Basin. Specifically, we asked how random extinctions could have altered the pattern of taxonomic and functional β-diversity of ray-finned fishes. We show that, whichever the level of extinction, the functional diversity losses would surpass taxonomic diversity. We reach these findings through a stochastic simulation approach based solely on pre-impact data available. In this regard, we argue that risk assessments or management strategies could benefit from this sort of analytical approach in order to better understand the likely consequences of different treatments on local and regional patterns of biodiversity and ecosystem functions.

The authors declare no conflicts of interest.