Extreme weather events threaten biodiversity and functions of river ecosystems: evidence from a meta-analysis

Extreme weather events threaten biodiversity and functions of river ecosystems: evidence from a meta-analysis Sergi Sabater 1,2, *,Anna Freixa 1,2 ,Laura Jiménez 1,3 ,Julio Lopez-Doval 1,3 , Giorgio Pace 4,5 ,Claudia Pascoal 4,5 ,Núria Perujo 1,3 ,Dylan Craven 6 and Juan David Gonzalez-Trujillo 7 1 Catalan Institute of Water Research (ICRA), Carrer Emili Grahit 101, 17003, Girona, Spain 2 GRECO, Institute of Aquatic Ecology, University of Girona, Campus de Montilivi, 17071, Girona, Spain 3 University of Girona, Plaça de Sant Domènec 3, 17004, Girona, Spain 4 Centre of Molecular and Environmental Biology (CBMA), Department of Biology, University of Minho, Campus of Gualtar, 4710-057, Braga, Portugal 5 Institute of Science and Innovation for Bio-Sustainability (IB-S), University of Minho, Campus of Gualtar, 4710-057, Braga, Portugal 6 Centro de Modelacion y Monitoreo de Ecosistemas, Universidad Mayor, Santiago, Chile 7 Museo Nacional de Ciencias Naturales (CSIC-MNCN), C. de José Gutiérrez Abascal, 28006, Madrid, Spain ABSTRACT Both gradual and extreme weather changes trigger complex ecological responses in river ecosystems. It is still unclear to what extent trend or event effects alter biodiversity and functioning in river ecosystems, adding considerable uncertainty to predictions of their future dynamics. Using a comprehensive database of 71 published studies, we show that event –but not trend –effects associated with extreme changes in water flow and temperature substantially reduce species richness. Furthermore, event effects –particularly those affecting hydrological dynamics –on biodiversity and primary productivity were twice as high as impacts due to gradual changes. The synthesis of the available evidence reveals that event effects induce regime shifts in river ecosystems, particularly affecting organisms such as invertebrates. Among extreme weather events, dryness associated with flow interruption caused the largest effects on biota and ecosystem functions in rivers. Effects on ecosystem functions (primary production, organic matter decomposition and respiration) were asymmetric, with only primary production exhibiting a negative response to extreme weather events. Our meta-analysis highlights the disproportionate impact of event effects on river biodiversity and ecosystem functions, with implications for the long-term conservation and management of river ecosystems. However, few studies were available from tropical areas, and our conclusions therefore remain largely limited to temperate river systems. Further efforts need to be directed to assemble evidence of extreme events on river biodiversity and functioning. Key words:flow interruption, floods, temperature, species richness, abundance, river ecosystem. CONTENTS I. Introduction .......................................................................451 II. Methods ..........................................................................451 (1) Data collection ..................................................................451 (2) Data analysis ...................................................................452 III. Results ............................................................................453 IV. Discussion .........................................................................455 *Author for correspondence (Tel.: +34 972 4183380; E-mail: [email protected]). Biological Reviews 98 (2023) 450–461 © 2022 The Authors. Biological Reviews published by John Wiley & Sons Ltd on behalf of Cambridge Philosophical Society. This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. Biol. Rev. (2023), 98, pp. 450–461. 450 doi: 10.1111/brv.12914 (1) Impacts on biodiversity of rivers ....................................................455 (2) Impacts on ecosystem functions of rivers ..............................................456 (3) Caveats of the analysis ............................................................457 V. Conclusions ........................................................................457 VI. Acknowledgements ..................................................................458 VII. Author contributions .................................................................458 VIII. References .........................................................................458 IX. Supporting information ...............................................................461 I. INTRODUCTION Extreme weather events are becoming more frequent and unpredictable as global warming and human activities jointly alter the water cycle (Evans & Boyer-Souchet, 2012; Ummenhofer & Meehl, 2017; Best & Darby, 2020). Changes in rainfall and temperature patterns have led to prolonged heatwaves, abrupt decreases in water flow, or large floods, which are subsequently altering temporal patterns of water flow in river networks over large areas of the Earth (Messager et al., 2021). Low water flow periods have become longer (Döll & Zhang, 2010) and, as a result, river systems experience progressively stronger transitions towards slowflowing phases (Sabater, 2008) or are completely interrupted (Messager et al., 2021). Furthermore, these hydrological changes have altered sedimentological processes (Inman & Jenkins, 1999), and increased water temperatures (Morrill, Bales & Conklin, 2005; Kaushal et al., 2010). At the other extreme of hydrological events, large floods have become more frequent (Hirabayashi et al., 2013; Arnell & Gosling, 2016), and already produce long-term effects in larger rivers of the world (e.g. Espinoza et al., 2022). Large floods may transform the geomorphological setting of river networks and, when catastrophic, produce long-lasting impacts on river biodiversity (Woodward et al., 2015). Thus the type and intensity of extreme events, principally hydrological and climatic events, likely affect the biodiversity and ecosystem functioning of river ecosystems, yet the magnitude and direction of their impacts remain to be comprehensively synthesized. Extreme events, which frequently overlap in time and space (Bowler et al., 2020), vary in their impacts on biodiversity and ecosystem functions (Rillig et al., 2019). Ecosystems impacted by frequent disturbances, such as river networks, host ecological communities and provide ecosystem functions that respond directly to water flow patterns (Thorp, Thoms & Delong, 2006). River systems are increasingly affected by gradual changes (trend effects) and extreme events (event effects; Jentsch, Kreyling & Bejerkuhnlein, 2007) in their hydrology and climate patterns. While gradual changes in water flow and temperature can be classified as press disturbances, extreme events are equivalent to pulse disturbances (Lake, 2000; Harris et al., 2018). The impacts of press disturbances on biodiversity may be seen as continuous and gradual, leading to shifts in species distributions and community structure (e.g. Mantua, Tohver & Hamlet, 2010). The pace of changes due to press disturbances may allow for some degree of adaptation, at least for some species (BenedettiCecchi et al., 2006; Heino, Virkkala & Toivonen, 2009). By contrast, the impacts of pulse disturbances may lead to abrupt changes in environmental conditions, disproportionately affecting species with limited opportunities for adaptation (Harris et al., 2018; Ledger & Milner, 2015). Press or pulse disturbances may differentially affect the capacity of ecosystems to resist and recover structurally and functionally (Jentsch & White, 2019). Changes to community structure due to trend and event effects may cascade to ecosystem functions, with unknown consequences for biodiversity–ecosystem functionrelationships(Biggsetal.,2020).Interrestrialecosystems such as experimental grassland communities, biodiversity consistentlybuffersvariationinecosystemfunctioning(productivity) for both extreme wet and dry events (Isbell et al., 2015). Functional redundancy, i.e. the extent to which multiple species perform similar roles in a community, may moderate theimpacts of extreme climate events on ecosystem functions (Woodward et al., 2015). Quantifying the relative importance of the impacts oftrend-andevent-drivendisturbancesonbiodiversityandecosystem functions of river ecosystems is essential for developing data-driven adaptation and mitigation strategies. Herein we perform a comprehensive meta-analysis that evaluates the relative impacts of trend and event weather effects on the biodiversity and ecosystem functions of river ecosystems. We compiled a database of 199 effect sizes, including both observational and experimental studies, to determine whether: (i) event effects have stronger, negative impacts than those of trend effects on diversity, abundance, and biomass of river communities, based on the assumption that pulse effects are stronger than press effects; (ii) hydrological anomalies lead to stronger or weaker impacts than temperature anomalies on river ecosystems; (iii) the impacts of trend and event effects on ecosystem functions (decomposition, respiration, primary production) are weaker than impacts on biodiversity, since ecosystem functions may be maintained at pre-disturbance levels where functional redundancy is high. II. METHODS (1) Data collection We used the Web of Science and Scopus databases to perform a comprehensive search of relevant studies (Gusenbauer & Haddaway, 2020). Entries up to March 2021 were included using a composite search string with hydrological and Biological Reviews 98 (2023) 450–461 © 2022 The Authors. Biological Reviews published by John Wiley & Sons Ltd on behalf of Cambridge Philosophical Society. Extreme events threaten biodiversity of river 451 1469185x, 2023, 2, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/brv.12914 by Cochrane Portugal, Wiley Online Library on [20/10/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License climatic descriptors considering different ecological communities and functions of river ecosystems (the complete set of Boolean terms is provided as online supporting information in Appendix S1). Our search returned 30,847 references, which was reduced to 1,235 unique records after removing duplicates (PRISMA diagram, Appendix S2). We only retained peer-reviewed papers testing the effects of climate change on biodiversity or ecosystem functions. These papers had to use standardized sampling methods (in the field or the laboratory), with at least one control and impact treatment and more than three samples for each. We included species richness, density (number of individuals per unit area), and biomass as descriptors of biodiversity, while organic matter decomposition, respiration, and primary production were included as descriptors of ecosystem functions (Hooper et al., 2012). We extracted mean values, standard deviations, standard errors, and the number of samples in control and impact treatments from tables, text, or figures. If the mean abundance or standard deviation were not reported, we derived these from the total abundance, sample size, and standard error, if possible. Data from figures were extracted using Webplot Digitizer 3.4 (Rohatgi, 2015). We also retrieved information on four moderators that we anticipated could affect biodiversity and ecosystem functions (Table S1). Following the information provided by the authors, for each extracted data point we collected information on the ecological community, type of effect (trend or event), anomaly type (increase or decrease in water flow; increase or decrease in water temperature), and study type (experimental or observational). In total, we built a database of 199 effect sizes from 71 studies (each observational or experimental treatment was considered as a replication unit), of which only 21 had both biodiversity and ecosystem function data. All studies included in the meta-analysis are identified in the reference list with an asterisk. Effect sizes were calculated as standardized mean differences (Hedges’g) using means, standard deviations, and sample sizes retrieved from each study. We used the escalc function in the metafor package in R (Viechtbauer, 2010)to compute Hedges’gand its respective variance. To account for differences in sampling variance, we weighted the estimated effect sizes by their sample size as recommended by Hamman et al.( 2018). Effect sizes are presented as units of the pooled standard deviation, such that a value of 0.5 represents a difference equivalent to half of a standard deviation. Thus, negative values indicate that climate change, whether trend or event, had a negative impact on the variables, while positive values indicate the opposite. The complete list of data sources, effect sizes, and moderators is available at https://doi.org/10.5281/zenodo.7004412. (2) Data analysis We used hierarchical Bayesian models to analyse variation in effect sizes, as implemented in the brms package in R (Bürkner, 2017). To distinguish between the effect of trend versus event effects and the anomaly type, we fitted a univariate model for each predictor variable, excluding the intercept. We chose this approach because intercept models for categorical predictors with more than two levels use only one as the reference against all others (Doherty et al., 2020). We considered predictor variables important for all models when the 95% credible intervals (CIs) of the posterior estimates did not overlap with zero. We did not evaluate interactions between fixed effects because some combinations of variables (e.g. extreme temperature decrease) were poorly replicated. We ran four chains of 10,000 iterations each for each model, with a burn-in of 1000 iterations, resulting in a total of 36,000 samples. We assessed convergence by inspecting trace plots and ensuring that the Gelman–Rubin statistic was <1.1 (Gelman & Rubin, 1992). We specified non-informative normal priors (μ=0, σ=10) for the fixed effects and weakly informative half Cauchy priors (μ=0, σ=1) for the random effects (Bürkner, 2018). We fitted models assuming a normal distribution and included random effects for the study identity, type, and duration to account for non-independence between effect sizes from the same study or study type. By incorporating these categorical factors as random effects, we assume that the biodiversity and ecosystem function responses will be more similar within the same study (ID) and similar types of study (observational or experimental). To identify an appropriate structure of random effects, we fitted and ranked five nested models containing different combinations of study ID and type as random effects. The ecological community could not be included as a random effect factor because the number of samples per class was highly unbalanced. Model ranking using leave-one-out cross-validation (Vehtari, Gelman & Gabry, 2017) indicated different random effects structures for each response variable (Table S2). To avoid potential confounding effects of unusually adverse events, we decided to exclude data points falling outside the ranges of the biodiversity (−10.3 to 9.1) and ecosystem function (0.04 to −7.7) aspects. These ranges accounted for the 1st percentile of the distribution of effect sizes in biodiversity (effect sizes less than −10) and ecosystem function variables (effect sizes less than −7). Specifically, we did not include some of the results from Romero et al. (2019) and Truchy et al.(2020) (Hedges’gfor biomass between −14 and −244), Geraldes, Pascoal & Cassio (2012) (Hedges’gfor decomposition greater than −10), and Oprei, Zlatanovic & Mutz (2019) (Hedges’gfor respiration greater than −10). The inclusion of these studies increased the magnitude of the effect size for extreme events although did not alter the reported trends. We used a combination of funnel plots and Egger’s tests (Egger et al., 1997) to assess potential publication bias. Visual inspection of the funnel plot suggested no systematic asymmetry in effect sizes (Fig. S1), a result also supported by the non-significance of Egger’s tests (Table S3). To assess the sensitivity of our analysis to the use of Hedges’g, we also ran a parallel meta-analysis using the log-response ratio (LRR) as Biological Reviews 98 (2023) 450–461 © 2022 The Authors. Biological Reviews published by John Wiley & Sons Ltd on behalf of Cambridge Philosophical Society. 452 Sergi Sabater and others 1469185x, 2023, 2, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/brv.12914 by Cochrane Portugal, Wiley Online Library on [20/10/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License the effect size. To calculate the LRR we divided values from impact treatments by those in the control treatments and took the logarithm of this number. We used the escalc function for this computation. Table S4 shows that our results are robust to the selection of the effect size, as the magnitude and direction of the effect sizes did not change significantly when comparing Hedges’gand LRR results. All analyses and figures were performed using R v. 4.0.3 (R Core Team, 2020). Code and data are available at https://doi.org/10.5281/zenodo.7004412. We provide a completed PRISMA for Ecology and Evolutionary Biology (PRISMA-EcoEvo; O’Dea et al., 2021) checklist in Table S5. III. RESULTS We analyzed 199 effect sizes for components of biodiversity (62 studies) and 100 effect sizes for ecosystem functions (26 studies) (see Table S4 for a summary of the effect sizes and their respective 95% credible intervals). The geographical distribution of the studies is shown in Fig. 1. Most studies were distributed in temperate and boreal regions in North America and Europe; very few were from tropical or arctic regions. Our analysis of multiple components of river biodiversity revealed negative impacts of weather changes on species richness (Fig. 2). A univariate model for weather anomalies (trend versus event) revealed that these negative impacts were mainly due to extreme events [posterior mean of event: −1.08 (95%CI: −1.78 to −0.42), posterior mean of trend: 0.07 (95%CI: −0.73 to 0.9)]. The between-study heterogeneity variance of this model was estimated at τ=0.75 (95%CI: 0.06–1.5). A separate model for anomaly type showed that decreases in water flow caused the most dramatic impacts on species richness (posterior mean: −1.48, CI: −2.42 to −0.60). The between-study heterogeneity variance of this model was estimated at τ=0.83 (95%CI: 0.18–1.46). Fig. 1. Geographical distribution of the studies included in the meta-analysis. Different colours indicate whether studies contained data on biodiversity, ecosystem functions, or both. Fig. 2. Impacts of trend and event effects and anomaly type on species richness of river ecosystems. The distributions display the posterior probability of different effect sizes, dots corresponding to posterior means, and their horizontal bars to 95% credible intervals. The dashed line corresponds to an effect size of zero. Numbers in parentheses represent the sample size (i.e. number of effect sizes) for each predictor variable. Weather events, particularly flow decrease, had a significant impact on riverine species richness. Biological Reviews 98 (2023) 450–461 © 2022 The Authors. Biological Reviews published by John Wiley & Sons Ltd on behalf of Cambridge Philosophical Society. Extreme events threaten biodiversity of river 453 1469185x, 2023, 2, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/brv.12914 by Cochrane Portugal, Wiley Online Library on [20/10/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License Neither extreme increases in temperature nor water flow had detectable effects on species richness. In contrast to species richness, community density (N=85) did not change consistently in response to trend or event effects (Table S4). As was the case for species richness, community biomass showed an overall decrease in response to event effects and anomaly types (Fig. 3A, model’sτ=−0.74, 95%CI: −1.17 to −0.31). The decrease in community biomass was not significantly associated with trend effects (posterior mean: −0.62, CI: −1.28 to 0.04), but the relationship was significant for event effects (posterior mean: −1.13, CI: −1.84 to −0.41). A second model revealed that reduction in biomass is associated with increases in temperature and decreases in flow (Fig. 3A, model’sτ=−1.04, 95%CI: −1.89 to −0.19). However, the response of community biomass was not consistent across taxonomic groups. Independent models indicate that trend changes do not impair benthic algal (biofilm) biomass (Fig. 3B), and that only extreme weather events affect fungal (posterior mean: −1.75, CI: −3.41 to −0.09; model’sτ=1.08, 95%CI: 0.04–2.96) (Fig. 3C) and invertebrate biomass (posterior mean: −1.46, CI: −2.37 to −0.54; model’sτ=1.34, 95%CI: 0.89–3.45) (Fig. 3D). We also found significant impacts of temperature increases and flow decreases on invertebrate biomass (model’sτ=−1.48, 95%CI: −2.53 to −0.43) (Fig. 3D). Finally, our results did not show effects of anomaly type on most of the ecosystem functions (Fig. 4). Respiration (Fig. 4C) and organic matter decomposition (Fig. 4B) were not affected by either event or trend effects or anomaly type. The between-study heterogeneity variance of these models (type of effect: trend versus event) was estimated at τ=0.98 (95%CI: 0.06–2.20) for respiration; τ=1.62 (95%CI: 1.03–2.46) for decomposition; and τ=0.83 (95%CI: 0.03–2.61) for primary productivity. However, we found that Fig. 3. Impacts of trend and event effects and anomaly type on the biomass of communities of river ecosystems (A), and on the biomass of biofilm (B), fungi (C) and invertebrate (D) communities. The distributions display the posterior probability of different effect sizes, with dots corresponding to posterior means and their horizontal bars to 95% credible intervals. The dashed line corresponds to an effect size of zero. Numbers in parentheses represent the sample size (i.e. number of effect sizes) for each predictor variable. Weather events (flow decrease) had a significant impact on biomass, mostly on that of invertebrates and fungi, but not on biofilm. Biological Reviews 98 (2023) 450–461 © 2022 The Authors. Biological Reviews published by John Wiley & Sons Ltd on behalf of Cambridge Philosophical Society. 454 Sergi Sabater and others 1469185x, 2023, 2, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/brv.12914 by Cochrane Portugal, Wiley Online Library on [20/10/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License primary productivity was negatively affected by abrupt flow decreases (Fig. 4A, posterior mean: −2.73, CI: −4.12 to −1.3). The between-study heterogeneity variance of these models (anomaly type) was estimated at τ=1.05 (95%CI: 0.05–2.73) for respiration; τ=1.62 (95%CI: 1.04–2.45) for decomposition; and τ=0.88 (95%CI: 0.04–2.46) for primary productivity. IV. DISCUSSION Ongoing changes in weather patterns may lead to transformative impacts on the biodiversity and ecosystem functions of river ecosystems across the globe. Here, we present a comprehensive meta-analysis showing that extreme weather events negatively affect the biodiversity of rivers, while of the ecosystem functions assessed, only primary productivity responded significantly, and only to weather events. (1) Impacts on biodiversity of rivers We found that extreme weather events had consistent, negative impacts on biodiversity (species richness and biomass) of river ecosystems. We determined that, on average, event effects on biodiversity were twice as strong as trend effects, which adheres to the definition of extreme event disturbances as localized, high-energy, rapid changes in environmental conditions (Peters et al., 2011). Our results raise concerns about the future of riverine biodiversity since extreme events are increasingly frequent and will challenge the adaptation capacity of the biota (Ummenhofer & Meehl, 2017; Jentsch & White, 2019). Event effects have driven considerable changes in a variety of ecosystems and biological compartments, spanning from soil microbiota (Fierer & Schimel, 2002), to meadowdwelling insects (Piessens et al., 2009), to fish foraging in coral reefs (Stuart-Smith et al., 2021; Richardson et al., 2018). In rivers, abrupt changes in water flow and temperature directly affect the habitat extent and suitability for organisms as they cause rapid variations in hydraulic and chemical conditions (Guse et al., 2015; Petrovic et al., 2011; Ponsatí et al., 2015). These variations challenge the adaptability of many species, with potential effects on the composition, diversity, and abundance of biological communities. Our results show that such effects differ for different biological compartments, with some being more vulnerable than others. Our synthesis of the available evidence reveals that river biota do not exhibit consistent responses to extreme events. While the species richness and biomass of invertebrates were consistently negatively affected by water flow interruption, microbial communities showed inconsistent responses, ranging from non-significant effects for benthic algal (biofilm) biomass to decreases for fungal biomass. Rivers undergoing complete or abrupt flow interruptions experience dramatic declines in available habitat (Suren & Jowett, 2006; Pace, Bonada & Prat, 2013), leading to habitat fragmentation and to the prevalence of colonization–extinction processes. Only taxa with traits conferring resistance to low-flow conditions (usually associated with physical and chemical stressors, such as high temperatures, organic matter and fine sediment accumulations, and low dissolved oxygen levels) or prolonged droughts, will remain after an extreme event occurs (Townsend & Hildrew, 1994). The different responses obtained in our meta-analysis suggest that organisms with shorter life cycles (microbes) may resist and recover from Fig. 4. Impacts of trend and event effects and anomaly type on ecosystem functions of rivers: (A) primary productivity, (B) decomposition, and (C) respiration. The distributions display the posterior probability of different effect sizes, with dots corresponding to posterior means, and their horizontal bars to 95% credible intervals. The dashed line corresponds to an effect size of zero. Numbers in parentheses represent the sample size (i.e. number of effect sizes) for each predictor variable. Only weather events, particularly flow decrease, had a significant effect, and only on primary productivity. Biological Reviews 98 (2023) 450–461 © 2022 The Authors. Biological Reviews published by John Wiley & Sons Ltd on behalf of Cambridge Philosophical Society. Extreme events threaten biodiversity of river 455 1469185x, 2023, 2, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/brv.12914 by Cochrane Portugal, Wiley Online Library on [20/10/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License pulse disturbances (event effects) to a greater extent than organisms with longer lifespans. Shorter generation times, higher dispersal ability and greater heat, abrasion, and desiccation tolerance may confer microorganisms with a greater resistance to hydrological extremes (Sabater et al., 2016; Chester & Robson, 2014). By contrast with specific studies considered in our analysis (Gonçalves et al., 2019; Romero et al., 2019), we conclude that neither changes in water flow nor temperature affected biomass of microbes (except for fungi). The absence of detectable effects on microbial communities can be attributed to high heterogeneity among studies or perhaps to the inaccuracy of species richness as a measure of the impacts of environmental changes on river biodiversity (Hillebrand et al., 2018). Losses of aquatic microbial species may favour the arrival of terrestrial species that are adapted to resist desiccation (Truchy et al., 2020), which may alter the composition of the bacterial community, but not necessarily its richness (following a rather general pattern across ecosystems; Blowes et al., 2019). Alternatively, the limited response of microbial biomass to hydrological disturbances suggests that other factors, such as the availability of resources or biological interactions, may enhance resistance (Krauss et al., 2011). Water flow reduction, for example, does not reduce benthic algal biomass because lower flow primarily reduces grazing pressure (Truchy et al., 2020). Invertebrates are the biological group for which the most detailed evidence is available regarding responses to environmental effects. While several studies have determined that changes in the distribution, diversity and abundance of invertebrates are associated with long-term increases in water temperature or flow patterns (Durance & Ormerod, 2007; Domisch et al., 2013), our analysis concludes that invertebrate richness and biomass are mostly sensitive to sudden (event) changes in water temperature and flow. Abrupt changes such as extreme droughts may induce effects on diversity (Bogan, Boersma & Lytle, 2015), mostly by affecting the most sensitive taxa (Piggot et al., 2015a; Madji et al., 2020). Ephemeroptera, Plecoptera and Trichoptera are the most affected groups of invertebrates, likely because they tend to be adapted to cold and highly oxygenated water, and they adapt poorly to warmer temperatures, lower dissolved oxygen levels, and shrinking waters (Céréghino, Boutet & Lavandier, 1997; Tierno de Figueroa et al., 2010). It has been observed that unseasonal or supraseasonal flow interruptions (Lake, 2003) can cause unpredictable effects on macroinvertebrate richness and biomass, especially in permanent systems (Hill et al., 2019). Several studies included in our meta-analysis show that flow interruptions may result in extensive loss of individuals and species (Calapez et al., 2014; Dewson, James & Death, 2007), thereby directly affecting species richness and biomass. Direct effects of water flow interruption not only challenge the resistance of invertebrate communities, but also their recovery because the colonization rate of drifting and flying invertebrates is a function of the distance to drought refuges (Vander Vorste, Malard & Datry, 2015). In contrast to interruptions in water flow, floods did not show consistent impacts on either species richness or biomass. While flood events may lead to losses of algae and macroinvertebrates via physical disturbance and washout (the scouring effect; Feeley et al., 2012) and cause important changes in the geomorphological structure of rivers, they do not necessarily affect all taxonomic groups similarly (Piniewski et al., 2017). This variability may reflect the differential ability of organisms to find shelter and the differential capacity of populations to recover quickly after floods subside (response diversity; Hershkovitz & Gasith, 2013). Our meta-analysis included several studies describing the effects of catastrophic floods (Tsai et al., 2014; Feeley et al., 2012; Foord & Fouché, 2016), which indeed reported large changes in the biomass of algae or invertebrates. Species loss of macroinvertebrates is most likely related to substratum movement and associated drift, and taxa inhabiting riffle areas (e.g. blackflies) are the most affected (Milner et al., 2013). However, with certain exceptions (e.g. Hynes, 1970; Woodward et al., 2015), studies monitoring the responses of invertebrate communities after floods have shown rapid invertebrate recovery (Herbst et al., 2019; Baillie et al., 2020), with an initial decline in abundance usually followed by a rapid increase in both abundance and species richness. Moderate floods may even enhance the mobility of nutrients and sediment, and the complexity of the river habitat (Death, Fuller & Macklin, 2015). Floods may therefore confer a competitive advantage to invertebrates with rapid development of aerial stages, asynchronous egg hatching, and synchronized metamorphosis with flood timing (Lytle & Poff, 2004). Compared to water flow, we found that temperature changes induced few direct and indirect effects on the biomass of stream biota. Moderate short-term warming has not been found to cause substantial effects on the biomass of biofilm communities (Romero et al., 2019), but sustained warming may increase the functional richness and diversity of benthic microbial communities (Ylla, Canhoto & Romaní, 2014). Warming generally enhances microbial activity (Diaz Villanueva, Albariño & Canhoto, 2011) but may cause changes in resource acquisition (Romero et al., 2019). In macro-organisms, warming may produce direct effects on the metabolism, leading to increased growth rates and smaller size at maturity. Indirect effects on macroorganism survival occur through changes in the availability and quality of basal resources (Piggot et al., 2015b; Hogg & Williams, 1996), which may account for our observed decrease in invertebrate biomass in response to warming (Fig. 3D). (2) Impacts on ecosystem functions of rivers Contrasting with responses in biodiversity, the ecosystem functions of rivers did not exhibit consistent, negative responses to weather variability. Only primary productivity exhibited a negative response to event effects, largely due to a negative response to extreme water flow decreases. Neither Biological Reviews 98 (2023) 450–461 © 2022 The Authors. Biological Reviews published by John Wiley & Sons Ltd on behalf of Cambridge Philosophical Society. 456 Sergi Sabater and others 1469185x, 2023, 2, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/brv.12914 by Cochrane Portugal, Wiley Online Library on [20/10/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License respiration nor organic matter decomposition were affected by trend or event effects. This synthesis of available evidence reveals that extreme water flow reduction consistently reduces primary productivity. Low flow, particularly during certain periods (e.g. summer), consistently decreases stream width and water velocity, favours higher water temperature, and reduces nutrient uptake (Riis et al., 2017). The physical constraints imposed by low water velocity (and associated higher thickness of the boundary layer over the stream bottom) have been found to lead to consistent decreases in net and gross primary production (Riis et al., 2017; Arias-Font et al., 2021), moving the balance from autotrophic to heterotrophic (Acuña et al., 2015; Romero et al., 2019). This shift towards heterotrophy is probably due to collapse of the photosynthetic capacity of primary producers under dry conditions (Colls et al., 2021). The lack of response of respiration and decomposition to trend or event effects in our meta-analysis does not support results of previous case studies showing that extreme events affect organic matter decomposition in river ecosystems. A recent study showed that the effects of drying on shredders did not affect leaf litter decomposition (Carey, Chester & Robson, 2021), highlighting the variable or uncertain outcomes of interactions among climate, water quality, and decomposers (Bernabé et al., 2018). The response of microbial decomposition to changes in water stress is highly context dependent, reflecting differential responses of microbial communities to temperature, nutrient, and oxygen levels (Duarte et al., 2017). Our meta-analysis indicates that extreme events can reduce fungal biomass, which is associated with decomposing substrates, while decomposition itself remained unaffected. Gonçalves et al.( 2019) showed that fungal biomass and diversity decreased after experimentally reducing water flow by 74–88%, but that ecosystem functions were maintained, suggesting that fungal communities have high functional redundancy. Similarly, Arias-Real et al.( 2022) found that a reduction in species richness did not alter decomposition rates in fungal communities with initial higher richness. Species-rich communities exhibit a wider range of ecological and functional responses to stressors than less-diverse communities, making them more resistant to disturbances (McLean et al., 2019). Overall, the apparent discrepancy between biodiversity and functional responses of river ecosystems to hydrological and climate trends and events highlights the role that functional redundancy (Biggs et al., 2020) may have on conserving the provisioning of ecological functions. (3) Caveats of the analysis While our meta-analysis revealed stronger effects from extreme weather events, we recognize that our results may reflect limitations associated with the distribution, type, and number of studies. We found few studies from Asia, Africa, and South America, whose river ecosystems show specific differences (e.g. Dudgeon, 2008; Petsch et al., 2020; GonzalezTrujillo et al., 2021) compared with the temperate ecosystems that comprise most effect sizes in our study. Secondly, only a small number of studies provided initial conditions to be compared with data after disturbances (Peters et al., 2011). This is an essential requirement for quantifying effect sizes. Due to this constraint, we included studies from a variety of contexts, ranging from field observations, to manipulative field studies, to experimental studies in the laboratory. Including this variety of studies to enable sufficient statistical power may, however, restrict the transferability of our results to real-world ecosystems, as the magnitude of climate change treatments frequently exceeds model-based projections (Korell et al., 2020). Thirdly, the data gaps for ecosystem functions, which are pervasive (von Schiller et al., 2017), did not allow us to define the temporal frame or period under which trend or event effects are more (or less) damaging (Trisos, Merow & Pigot, 2020). It should be kept in mind that most studies conducted so far on ecosystem functions have been performed under controlled conditions, at relatively short spatial and temporal scales (Estes et al., 2018) and exclude biotic components that may also contribute to ecosystem functions (e.g. detritivores in decomposition studies and grazers in primary production studies). Finally, the low sample size also affected our ability to quantify precisely the impacts of weather anomalies, particularly for groups such as fish or bacteria. There is need for a larger number of studies to reduce between-studies heterogeneity (our measured τ) and provide more accurate estimates. We dealt with imperfect data availability by conducting a Bayesian meta-analysis, which allows direct modelling of the uncertainty of τestimates and improves estimations of the pooled effects when the number of included studies is small (McNeish, 2016). Bayesian methods produce full posterior distributions that allow comparison of the whole spectrum of impact levels caused by trend or event effects. Despite these caveats, our meta-analysis is the most comprehensive to date on the impacts of hydrology and climate trends and events on the biodiversity and ecosystem functions of river ecosystems. Future research should aim to expand how the impacts of trend and event effects on river ecosystems are quantified in terms of biodiversity or ecosystem functions in order to test the differential effects of extreme stressors and their consequences for the future of rivers. We call on future investigators to measure the impacts of statistically unusual events (i.e. above the 99th percentile) using standardized protocols. V. CONCLUSIONS (1) Our study provides insights into the impacts of extreme weather events on the maintenance of biodiversity and ecosystem functions in river ecosystems. (2) We found evidence that substantial changes may occur at population and community levels, but mostly for larger organisms such as invertebrates. Biological Reviews 98 (2023) 450–461 © 2022 The Authors. Biological Reviews published by John Wiley & Sons Ltd on behalf of Cambridge Philosophical Society. Extreme events threaten biodiversity of river 457 1469185x, 2023, 2, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/brv.12914 by Cochrane Portugal, Wiley Online Library on [20/10/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License (3) Even though impacts on ecosystem functions are less evident, primary productivity was particularly impacted by extreme conditions. (4) Strong, negative impacts of certain types of anomalies, especially water flow interruption, cause effects that may shift biodiversity and (at least some) ecosystem functions irreversibly beyond their dynamic stable states (Trisos et al., 2020). (5) Evidence that weather extremes may have long-lasting impacts on the biodiversity and ecosystem functioning of river ecosystems already exists. The Millennium drought in Australia (1997–2009) led to some rivers being unable to recover to pre-drought flow regimes (Peterson et al., 2021), with these river ecosystems providing examples of ‘no turning back’in response to weather extremes that cause irreversible effects. VI. ACKNOWLEDGEMENTS This research was funded by the CLIMALERT ERA4CS project (CGL2017-88640-C2-1-R of the Spanish Ministry of Economy and Competitiveness; ERA4CS/0004/2016 of the Portuguese Foundation for Science and Technology). Additional support was received through the projects RIVSTRESS (PID2020-115708RB-C22) and STREAMECO (PTDC/CTA-AMB/31245/2017). ICRA researchers acknowledge funding from the CERCA program and the support through the Consolidated Research Group ICRA-ENV 2017 SGR 1124. The CBMA researchers acknowledge the Contrato-Programa UIDB/04050/2020 funded by Fundaç˜ ao para a Ciência e a Tecnologia. A.F. acknowledges the Juan de la Cierva program (IJC2019-039181-I) funded by MCIN/AEI. VII. AUTHOR CONTRIBUTIONS S. S. and J. L.-D. designed the study. S. S., J. L.-D., A. F., L. J. and N. P. extracted data from the text, tables, figures, or supplementary information in the reviewed papers. J. D. G.-T. performed the statistical analyses. S. S. and J. D. G.-T. led the writing, and all other authors discussed and contributed to revisions. VIII. REFERENCES References used in the meta-analysis are identified with an asterisk (*). Acun ˜a, V.,Casellas, M.,Corcoll, N.,Timoner, X. &Sabater, S. (2015). Increasing extent of periods of no flow in intermittent waterways promotes heterotrophy. Freshwater Biology 60(9), 1810–1823. *Archdeacon, T. P.,Diver-Franssen, T. A.,Bertrand, N. G. &Grant, J. D. (2020). Drought results in recruitment failure of Rio Grande silvery minnow (Hybognathus amarus), an imperiled, pelagic broadcast-spawning minnow. Environmental Biology of Fishes 103(9), 1033–1044. *Arias-Font, R.,Khamis, K.,Milner, A. M.,Sambrook Smith, G. H. & Ledger, M. E. (2021). Low flow and heatwaves alter ecosystem functioning in a stream mesocosm experiment. Science of the Total Environment 777, 146067. Arias-Real, R.,Gutie ´rrez-C anovas, C.,Mun ˜oz, I.,Pascoal, C. & Mene ´ndez, M. (2022). Fungal biodiversity mediates the effects of drying on freshwater ecosystem functioning. Ecosystems 25, 780–794. Arnell, N. W. &Gosling, S. N. (2016). The impacts of climate change on river flood risk at the global scale. Climate Change 134, 387–401. *Arroita, M.,Flores, L.,Larran ˜aga, A.,Martı ´nez, A.,Martı ´nezSantos, M.,Pereda, O.,Ruiz-Romera, E.,Solagaistua, L. &Elosegi, A. (2016). Water abstraction impacts stream ecosystem functioning via wettedchannel contraction. Freshwater Biology 62(2), 243–257. *Baillie, B. R.,Evanson, A. W.,Kimberley, M. O. &Bergin, D. O. (2020). Combined effects of an anthropogenic (forest harvesting) and natural (extreme rainfall event) disturbance on headwater streams in New Zealand. Freshwater Biology 65(10), 1806–1823. *Batista, D.,Pascoal, C. &C assio, F. (2012). Impacts of warming on aquatic decomposers along a gradient of cadmium stress. Environmental Pollution 169,35–41. Benedetti-Cecchi, L.,Bertocci, I.,Vaselli, S. &Maggi, E. (2006). Temporal variance reverses the impact of high mean intensity of stress in climate change experiments. Ecology 87(10), 2489–2499. Bernabe ´,T.N.,De Omena, P. M.,Dos Santos, V. P.,De Siqueira, V. M.,De Oliveira, V. M. &Romero, G. Q. (2018). Warming weakens facilitative interactions between decomposers and detritivores, and modifies freshwater ecosystem functioning. Global Change Biology 24(7), 3170–3186. Best, J. &Darby, S. E. (2020). The pace of human-induced change in large rivers: stresses, resilience, and vulnerability to extreme events. One Earth 2(6), 510–514. Biggs, C. R.,Yeager, L. A.,Bolser, D. G.,Bonsell, C.,Dichiera, A. M., Hou, Z.,Keyser, S. R.,Khursigara, A. J.,Lu, K.,Muth, A. F.,Negrete, B. Jr. &Erisman, B. A. (2020). Does functional redundancy affect ecological stability and resilience? A review and meta-analysis. Ecosphere 11(7), e03184. Blowes, S. A.,Supp, S. R.,Antao, L. H.,Bates, A.,Bruelheide, H., Chase, J. M.,Moyes, F.,Magurran, A.,Mcgill, B.,Myers-Smith, I. H., Winter, M.,Bjorkman, A. D.,Bowler, D. E.,Byrnes, J. E. K., Gonzalez, A.,ET AL. (2019). The geography of biodiversity change in marine and terrestrial assemblages. Science 366(6463), 339–345. *Boersma, K. S.,Bogan, M. T.,Henrichs, B. A. &Lytle, D. A. (2013). Invertebrate assemblages of pools in arid-land streams have high functional redundancy and are resistant to severe drying. Freshwater Biology 59(3), 491–501. Bogan, M. T.,Boersma, K. S. &Lytle, D. A. (2015). Resistance and resilience of invertebrate communities to seasonal and supraseasonal drought in arid-land headwater streams. Freshwater Biology 60(12), 2547–2558. *Boule ˆtreau, S.,Lyautey, E.,Dubois, S.,Compin, A.,Delattre, C.,Touron Bodilis, A.,Mastrorillo, S. &Garabetian, F. (2014). Warming-induced changes in denitrifier community structure modulate the ability of phototrophic river biofilms to denitrify. Science of the Total Environment 466-467, 856–863. Bowler, D. E.,Bjorkman, A. D.,Dornelas, M.,Myers-Smith, I. H., Navarro, L. M.,Niamir, A.,Supp, S. R.,Waldock, C.,Winter, M., Vellend, M.,Blowes, S. A.,Bo¨hning-Gaese, K.,Bruelheide, H.,Elahi, R., Antao, L. H.,ET AL. (2020). Mapping human pressures on biodiversity across the planet uncovers anthropogenic threat complexes. People and Nature 2,380–394. Bu ¨rkner, P. C. (2017). brms: An R package for Bayesian multilevel models using Stan. Journal of Statistical Software 80(1), 1–28. Bu ¨rkner, P. C. (2018). Advanced Bayesian multilevel modeling with the R package brms. The R Journal 10(1), 395–411. *Calapez, A. R.,Elias, C. L.,Almeida, S. F. P. &Feio, M. J. (2014). Extreme drought effects and recovery patterns in the benthic communities of temperate streams. Limnetica 33(2), 281–296. Carey, N.,Chester, E. T. &Robson, B. J. (2021). Flow regime change alters shredder identity but not leaf litter decomposition in headwater streams affected by severe, permanent drying. Freshwater Biology 66(9), 1813–1830. Cereghino, R.,Boutet, T. &Lavandier, P. (1997). Abundance, biomass, life history and growth of six Trichoptera species under natural and hydropeaking conditions with hypolimnetic releases in a Pyrenean stream. Archiv für Hydrobiologie 138(3), 307–328. Chester, E. T. &Robson, B. J. (2014). Do recolonisation processes in intermittent streams have sustained effects on benthic algal density and assemblage composition? Marine and Freshwater Research 65(9), 784–790. Colls, M.,Timoner, X.,Font, C.,Acun ˜a, V. &Sabater, S. (2021). Biofilm pigments in temporary streams indicate duration and severity of drying. Limnology and Oceanography 66(9), 3313–3326. *Covich, A. P.,Crowl, T. A. &Scatena, F. N. (2003). Effects of extreme low flows on freshwater shrimps in a perennial tropical stream. Freshwater Biology 48(7), 1199–1206. Death, R. G.,Fuller, I. C. &Macklin, M. G. (2015). Resetting the river template: the potential for climate-related extreme floods to transform river geomorphology and ecology. Freshwater Biology 60(12), 2477–2496. *Delgado, C.,Almeida, S. F. P.,Elias, C. L.,Ferreira, V. &Canhoto, C. (2017). Response of biofilm growth to experimental warming in a temperate stream. Ecohydrology 10(6), e1868. Biological Reviews 98 (2023) 450–461 © 2022 The Authors. Biological Reviews published by John Wiley & Sons Ltd on behalf of Cambridge Philosophical Society. 458 Sergi Sabater and others 1469185x, 2023, 2, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/brv.12914 by Cochrane Portugal, Wiley Online Library on [20/10/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License