Introduction

The latest IPCC report (IPCC, 2021) stresses that climate change imprints ramping changes in the global water cycle, reductions in snow and ice cover, and higher frequency of extreme climatic events. Impacts of climate change will be more stressful in some regions than others. In Europe, Northern regions will progress towards wetter conditions, while the Mediterranean region will suffer from more acute drying. Overall, a runoff decrease is expected in southern Europe, which will add to strong human pressures on water resources (Sabater et al., 2018), either directly through flow regulation and water withdrawal for irrigation, or indirectly through effects of landscape change (Jaramillo & Destouni, 2015; Gudmundsson et al., 2021).

Climate change effects on hydrological conditions will directly affect freshwater ecosystems and their biological communities (Dudgeon et al., 2006). While impacts will concern all biological components of the aquatic ecosystem, some are highly sensitive and perform as early warning indicators. This is the case of the diatoms, siliceous microscopic algae making up the largest fraction of algal assemblages in river systems (Pan et al., 1999). Diatoms are highly sensitive organisms to flow reduction and desiccation, this affecting their typology and community composition (Tornés et al., 2021). We have recently observed that under extreme non-flow conditions, only aerophilic diatoms persist, while sensitive species disappear after a few days of non-flow (Tornés et al., 2021). Diatoms may then provide an ideal model group on which to explore the response of biological communities to water flow reduction associated to climate and global change.

We here use diatom assemblages from reference sites in river networks spread throughout several ecoregions in continental Spain. Reference conditions in freshwaters refer to the absence or minimal anthropogenic impact (Wallin et al., 2003). According to the European Water Framework Directive (WFD-European Commission, 2000), reference sites should consistently show this minimal impact in biological, hydromorphological, and physical and chemical characteristics. A central aspect of the WFD is accounting for type-specific ecological assessment and classification of waterbodies, using reference sites as a contrast to impacted sites (Feio et al., 2013). Biological communities, and diatoms amongst them (Tornés et al., 2012), play a central role under this scheme.

We here aim to characterize the distribution of diatom assemblages in reference sites of the main river basins of continental Spain and associate them to the present physiographic conditions of temperature and runoff. We subsequently associate the current assemblages’ distribution to predicted climate changes affecting the different ecoregions within Spain. We used an extensive dataset of reference sites distributed throughout continental Spain, and covering a wide range of fluvial typologies, to test the working hypotheses that (a) river types with lower runoff and associated high temperature and high mineralization may harbour the most tolerant diatoms to climate change; and (b) that differences between river types are reduced because of the strong decrease in water runoff caused by climate change, which may lead diatom assemblages towards homogenization.

Material and methods

Study sites

The study was conducted in 346 reference stream sites (Fig. 1) distributed over the Spanish territory and the most relevant watersheds, covering the fluvial typologies recognized in the country (Toro et al., 2009). We used 32 national river types, originally classified using progressive segregation of fluvial network subsets by establishing thresholds for each of the variables. This is an open hierarchical classification based on a GIS tool map. The variables used to determine the river types were altitude (meters a.s.l.), annual thermal amplitude (°C), catchment area (km2), mean annual discharge (m3/s), mean annual specific discharge (m3/s km2), estimated baseline conductivity (μS/cm), latitude (UTM 30) and longitude (UTM 30), stream order (according to the Strahler classification), mean catchment gradient (%), mean annual air temperature (°C), and the fraction of months with non-flow (%). This last descriptor derives from the SIMPA model (after its Spanish acronym, Integrated System for Rainfall-Runoff Modelling, Álvarez et al. 2004).

Fig. 1
figure 1

Study area and location of reference stream sites in the continental Spain. Coloured areas represent the three biogeographic areas present (Alpine ALP, Atlantic ATL, and Mediterranean MED), following the classification of the European Environment Agency (Codelist for bio-geographical regions, Europe 2011; www.eea.europa.eu). Sampling sites are labelled following the new classification of river types (Fig. 2) obtained in this study

Full size image

The area of study has a large spatial heterogeneity regarding its geomorphological and climatic diversity. River sites included in the study range from high, middle mountain, and lowland rivers, headwaters and large rivers, floodplain rivers, and coastal streams. These rivers and their basins include the Mediterranean, Atlantic, and Alpine ecoregions in Spain. Those under Mediterranean climate show a strong seasonality and interannual rainfall variability (Gasith & Resh, 1999) and are submitted to low-flow periods during late spring and summer, and high-flow periods in late autumn and winter. Headwaters and middle-order streams in the Mediterranean ecoregion often become intermittent or have intermittent reaches during long periods of the hydrological cycle (Colls et al., 2019). Rivers in the Cantabrian-Atlantic ecoregion are short in length and receive abundant rainfall, even in summer, due to the dominance of northern winds, and a narrow yearly temperature range (Amigo et al., 2017). The climate in the Cantabrian-Atlantic region is oceanic, with cool winters and warm summers, and with a mean annual rainfall of 1000 –1200 mm distributed throughout the year. Althouh these systems do not dry up, the lowest precipitation occurs during summer (Gartzia de Bikuna et al., 2015). The Alpine ecoregion mostly has mountain rivers, with steep gradients along most of the channel length. These systems experience minimum temperatures below 0 ºC, annual rainfall of above 1000 mm, and heavy snowfall in winter (Wohl, 2010; Sabater et al., 2022). Although dry periods are characteristic from the Mediterranean ecoregion, drying events increasingly occur in the humid and temperate regions. A general decrease in snow cover, along with an increase in the annual temperature and air pressure has been observed in the Alpine ecoregion, leading to a significant decreasing trend in the annual runoff (Falasco et al., 2018).

We used diatom data from two nation-wide surveys (TRAGSATEC, 2020), complemented with others obtained in regional surveys (CHE, 2017; Tornés et al., 2021). Overall, most of the sites were sampled more than once, spanning from 2014 to 2020, and providing a total of 603 samples (Supplementary Table S1).

Environmental characterization

The environmental variables of the sites were extracted from the characterization of Spanish river types (Toro et al., 2009). The ones selected were altitude (meters a.s.l.), catchment area (Km2), mean annual specific discharge (m3/s/km2), estimated baseline conductivity (μS/cm), mean catchment gradient (%), months of non-flow (%), and mean annual air temperature (°C) (Table 1). We did not use annual thermal amplitude (°C), mean annual discharge (m3/s), latitude (UTM 30), longitude (UTM 30), and stream order (Strahler), due to their redundancy with the other variables.

Table 1 Mean annual values, coefficient of variation (CV, %), and range of the environmental variables defining the Spanish national river types
Full size table

Actual hydrological data in the reference sites were sought using the QGIS software (version 3.16.0; QGIS Development Team, 2020). We constructed a layer with the gauging stations of the region, using data available from different water authorities (www.miteco.gob.es; www.aca.gencat.cat; www.juntadeandalucia.es). We then used the Distance Matrix algorithm from QGIS Network Analysis Toolbox (QNEAT3) plugin, to find out the nearest gauging station to each reference site, using the fluvial systems as the network layer. QNEAT3 computes the network route-based cost of Origin–Destination relations between the points of two layers and calculates the shortest paths between all the points. We considered appropriate gauging stations that ones located upstream of a site and at a distance less than 10 km, with natural flow regime, and where dams and/or reservoirs, or tributaries between the gauging station and the sampling site, did not occur. Results from the QGIS QNEAT3 plugin indicated that only few sites fulfilled those criteria. Even cases where the gauging station existed, data were not available for the period of the diatom sampling. Then, from the 346 reference stream sites selected for the analysis, less than 50 accomplished the fixed criteria, representing less than 100 samples (from a total of 603), not evenly distributed across the river types.

Overall, hydrological data for each river site could not be used in the subsequent analyses and forced us to define theoretical values after those of Toro et al. (2009). For that, we used the extensive dataset of reference sites in the continental Spain, and applied the predictions of the Representative Concentration Pathways (RCP) 4.5 and 8.5 to the present characteristics of runoff, temperature, and water conductivity of the different river types. With these, we defined new values under the two RCP scenarios based on the IPCC 2013 (CEDEX, 2017). A conductivity-temperature relationship was applied after Hayashi (2004) to estimate future values of water conductivity in the different types.

Diatom sampling and laboratory analyses

Diatom samplings were conducted mostly during summer, although some of the sites were sampled in spring and autumn. Sampling periods, therefore, covered low-flow (summer) and high-flow (spring, autumn) periods. Sampling followed standard procedures (Kelly et al., 1998; MAGRAMA, 2013). In short, diatoms were sampled from riffle sections and obtained from several colonized cobbles. From each cobble, the total algal materials were scraped with a brush into 20 mL of distilled water. A composite sample was created from the samples of the five cobbles from each stream, and each composite sample was preserved in 4% formaldehyde or ethanol until analysis. Materials were later cleaned for the organic material using boiling hydrogen peroxide, and cleaned frustules were mounted on permanent slides using Naphrax (r.i. 1.74; Brunel Microscopes Ltd., Chippenham,Wiltshire, UK). Up to 400 valves were counted on each slide by performing random transects under light microscopy using differential interference contrast optics at a magnification of ×1000.

Diatom assemblages were examined for their taxonomical composition. Diatoms were identified using reference floras (Krammer & Lange-Bertalot, 1991a, b, 1997a, b; Cantonati et al., 2017), and complemented through the monographs of “Diatoms of Europe” and “Bibliotheca Diatomologica”. As diatom data derived from different datasets, they were checked for their homogeneity in the identification. Then, last accepted names were validated for each taxon, and same taxa with different names (synonymies) were combined in a single entry. Teratological forms were summed to the normal forms as they were not considered in all surveys.

Data analyses

Sites ordination

We departed from the original classification of national river types of Toro et al. (2009) and then proceeded to group them according to their physiography and hydrology. Classification of the sites was conducted by means of hierarchical agglomerative clustering. Pearson's correlation was performed to define co-linear variables (R2 > 0.8) which were then not included in the cluster analysis. As such, catchment gradient was removed because of its high correlation with specific discharge (R2 = 0.83). The cluster analysis was performed using group average as cluster mode, on previously calculated Euclidean distances for the available environmental variables. The statistical significance of between-group differences was tested using a Permutational MANOVA (PERMANOVA). The PERMANOVA was based on the Euclidean distances scores. The PERMANOVA operates on a resemblance matrix and is similar to traditional parametric multivariate analysis of variance (MANOVA) but does not require a multivariate normal distribution. It provides a pseudo-F statistic by expectations of mean squares, and a P-value based on permutations of the data (Anderson, 2001).

The environmental variability of the resulted end-groups of the cluster analysis was explored by means of a Principal Components Analysis (PCA). Before performing the PCA, variables were normalized by subtracting the mean and dividing it by the standard deviation.

Diatom analyses

Principal Coordinates Analysis (PCO) was used to ordinate the taxonomical composition of diatom assemblages in the 603 stream samples available. PCO can be based on any (symmetric) resemblance matrix like non-metrical multi-dimensional scaling, and it is a projection of the points onto axes that minimize residual variation in the space of the resemblance measure chosen (Anderson et al., 2008). The PCO was performed on previously calculated Bray–Curtis similarities for the taxonomical composition of diatom assemblages. We used the complete diatom taxa list to avoid arbitrary decisions associated to the removal of rare taxa; downweighing the contributions of rare taxa is an inherent property of the computation of Bray–Curtis similarities (Capone & Kushlan, 1991; Hansen & Ramm, 1994). Afterwards, and independently of the environmental variables, the PCO was linked to the cluster by indicating membership of each sample to a determined cluster end-group using symbol factors.

We finally used the indicator value method (IndVal; Dufrêne & Legendre, 1997) to identify the indicator species of the resulted groups of types in cluster dendrogram. IndVal is a simple and useful method to identify indicator species and species assemblages in groups of samples (Dufrêne & Legendre, 1997). The relevance of this method lies in the way it combines information on the specificity and the fidelity of occurrence of a species in a group. It produces indicator values for each species in each group expressed as the product of the specificity and fidelity. Therefore, indicator species are defined as the most characteristic species in each group. Taxa which were mostly observed in only one type of stream were nominated as type-specific. The statistical significance of the species indicator values is evaluated using a randomization procedure. The indicator value of a species i is the largest value of IndValij observed amongst all groups j. The indicator value is at its maximum when all individuals of a species are found in a single group of sites (high specificity) and when the species occurs in all sites of that group (high fidelity) (Dufrêne & Legendre, 1997).

Cluster analysis, PCA, PCO, and PERMANOVA (999 permutations) were performed with PRIMER-E 6 v.6.1.11 and PERMANOVA + v.1.0.1 (PRIMER-E Ltd., Plymouth, UK). Analyses were carried out with square-root-transformed diatom data further converted into a resemblance matrix using Bray–Curtis similarity. Environmental data (except those expressed as ranked variables) were logarithmically transformed (x + 1) before analyses to reduce skewed distributions. IndVal analysis was conducted using R package “labdsv” (version 4.0.3; R Core Team, 2020).

Results

The stream sites we used in this study derive from a larger group of sites previously validated as references by the corresponding water authorities (CHE, 2017; ACA, 2020; TRAGSATEC, 2020). The selected sites accomplished the reference conditions defined by the WFD REFCON guidelines (European Commission, 2003). Setting the high ecological status required for the reference conditions, in terms of biological, hydromorphological, and physical and chemical quality elements, may be difficult in the Mediterranean region, due to the long history of human presence and the strong human exploitation of water resources (Feio et al., 2013). This is a general problem in European Mediterranean rivers, where almost all river types have experienced hydromorphological changes of anthropological origin (Feio et al., 2013; Almeida et al., 2014). Our reference sites adjusted to minimally disturbed conditions (i.e. absence of significant human disturbance), or least disturbed conditions (i.e. sites with the lowest signs of human disturbance in areas with extensive human disturbance), following the categories defined by Stoddard et al. (2006). Reference conditions could be attributed to each of the river types considered in the classification of Toro et al. (2009). However, reference sites could not be nominated in large areas of some river basins (Fig. 1). In particular, the northern and southern parts of the Central Plateau, and the Ebro Depression, have large areas onto which prevail disturbed conditions (Sabater et al., 2022). The absence of reference sites was the most notorious in the main axes of rivers flowing through these areas.

The cluster analysis lead to the definition of 7 end-group river types (Fig. 2). The groups were significantly different (PERMANOVA; pseudo-F6,25 = 13.548, P = 0.001). The physiographic and hydrological characteristics of each of the 7 end-groups are detailed in Table 2. Cluster group 1 (4 national river types) consisted of siliceous lowland rivers. A second cluster group (6 national river types) included mineralized rivers of low altitude. Cluster group 3 (4 national river types) accounted for mineralized rivers having low specific discharge. The fourth cluster group (4 national river types) encompassed both Mediterranean and Mediterranean-continental axes. Cluster group 5 (3 national river types) included rivers with the smallest catchment area and the lowest mineralization. Groups 1 to 5 were situated in the Mediterranean biogeographic region (Fig. 1). The sixth group (3 national river types) consisted of mountain rivers, with the highest altitude and high specific discharge. This group was mostly located in the Alpine and the Atlantic regions. Finally, cluster group 7 (8 national river types) included the Cantabrian-Atlantic rivers, with streams short in length and high specific discharge. Group 7 was only found in the Atlantic biogeographic region. A Principal Components Analysis (PCA) based on the available environmental variables of the groups of types in the cluster dendrogram (Fig. 3), stressed the prevailing environmental characteristics for the different cluster groups.

Fig. 2
figure 2

Group average clustering dendrogram of river types based on their Euclidean distances using data from Table 1. The resulted seven end-groups are identified. Codes correspond to those of Table 3. Table 2 and Fig. 3 show the environmental variables characterizing each end-group and their distribution in an ordination space, respectively

Full size image
Table 2 Mean values, coefficient of variation, and ranges of environmental variables characterizing the different cluster groups of Spanish national river types, with the number of river types per group indicated in brackets
Full size table
Fig. 3
figure 3

Principal components analysis (PCA) based on available environmental variables of the resulted groups of types in cluster dendrogram. The vector length and direction reflect the importance of each variable's contribution to each of the two axes

Full size image

We performed the ordination of the diatom assemblages by means of a Principal Coordinates Analysis (PCO) and labelled them according to the different groups of sites (Fig. 4). Even though the comparison of the two analyses has inherent limitations, the ordination of the diatom assemblages is largely coherent with that of the cluster groups. The subsequent IndVal analysis revealed that all cluster end-groups had significant indicator diatom taxa (Supplementary Table S2), and nearly all type groups had species with IndVals higher than 25% (Dufrêne & Legendre, 1997). Higher IndVals mostly occurred in groups 1, 4, and 6. Planothidium frequentissimum (Lange-Bertalot) Lange-Bertalot var. frequentissimum showed a high IndVal (48%) in the siliceous lowland rivers (Group 1). Sellaphora nigri (De Not.) C.E. Wetzel et Ector comb. nov. emend., Melosira varians Agardh, Nitzschia palea (Kützing) W. Smith, Cocconeis euglypta Ehrenberg, and Navicula cryptocephala Kützing showed an indication between 20 and 35%. The species with the highest indicator value in Group 2 were Halamphora tenerrima (Aleem & Hustedt) Levkov, Nitzschia inconspicua Grunow, Halamphora coffeaeformis (Agardh) Levkov, and Brachysira aponina Kützing, characteristic in mineralized lowland rivers. The highest IndVals for mineralized mid-altitude rivers with low specific discharge (Group 3) corresponded to Encyonopsis microcephala (Grunow) Krammer and Encyonopsis minuta Krammer & Reichardt. Aerophilic taxa such as Sellaphora stroemii (Hustedt) Kobayasi in Mayama Idei Osada & Nagumo, Luticola mutica (Kützing) D.G. Mann, Humidophila contenta (Grunow) Lowe, Kociolek, Johansen, Van de Vijver, Lange-Bertalot & Kopalová, and Halamphora montana (Krasske) Levkov emerged as indicator species in Group 3. These taxa had low IndVals (≤ 11%) but showed a high specificity to this group. Diatoma vulgaris Bory var. vulgaris, Navicula cryptotenella Lange-Bertalot, Cymbella affinis Kützing, Diatoma moniliformis Kützing, Achnanthidium druartii Rimet & Couté in Rimet & al., Pantocsekiella costei (Druart et F. Straub) K.T. Kiss et Ács, and Nitzschia dissipata subsp. dissipata (Kützing) Grunow var. dissipata were indicator species in Mediterranean and Mediterranean-continental main axes (Group 4), with an indication between 21 and 38%. The best indicators for Group 5 (poorly mineralized rivers) were Gomphonema rhombicum M. Schmidt, Cocconeis pseudolineata (Geitler) Lange-Bertalot, and Encyonema minutum (Hilse in Rabh.) D.G. Mann in Round Crawford & Mann. The species with the highest indicator value in Mountain rivers (Group 6) were Gomphonema angustivalva E. Reichardt, Achnanthidium lineare W. Smith, Diatoma ehrenbergii Kützing, Achnanthidium pyrenaicum (Hustedt) Kobayasi, and Gomphonema pumilum (Grunow) Reichardt & Lange-Bertalot. The highest IndVals for Cantabrian-Atlantic rivers (Group 7) were for Navicula angusta Grunow, Surirella roba Leclercq, and Achnanthidium subatomus (Hustedt) Lange-Bertalot. Some terrestrial (Humidophila laevissima (P.T.Cleve) Lowe, Kociolek, Johansen, Van de Vijver, Lange-Bertalot & Kopalová) and aerophilic (Eunotia minor (Kützing) Grunow in Van Heurck, and Nitzschia hantzschiana Rabenhorst var. hantzschiana) taxa also appeared as indicators for this group, having a low IndVals (6, 15, and 5%, respectively) but high specificity.

Fig. 4
figure 4

Principal coordinates analysis (PCO) performed with the diatom assemblages of the 603 samples available. Independent to this ordination, the membership of each sample to one of the cluster end-groups of Fig. 2 was indicated with symbol factors

Full size image

Discussion

There is general agreement that future climate change will cause higher variability in water regimes and further reduction in water flow over large areas of the planet (Messager et al., 2021). Low water flow periods may become longer (Döll & Zhang 2010), and rivers will experience lentification, i.e. a progressive conversion towards slower flowing condition (Sabater, 2008). In many cases, streams may be driven towards desiccation for long periods (Stubbington et al., 2017) and even become terrestrialized. Under lower-than-normal dilution conditions, concentrations of nutrients and other dissolved materials increase (Petrovic et al., 2011), water warms up, and the mobility of materials and organisms becomes compromised (Liu et al., 2013; Cañedo-Argüelles et al., 2015; Death et al., 2015; Dong et al., 2016). These are conditions for significant changes in biological communities. In the case of diatom assemblages, climate change might favour taxa less sensitive to warmer and lentic conditions while affect those preferring colder and fast-moving waters (Tornés et al., 2021).

There is a clear prediction of strong runoff reductions in river ecosystems of the Iberian Peninsula under the worst climate change scenarios (IPCC, 2021). These models contemplate a 20% of runoff reduction, accompanied with an increase in temperature and conductivity for most river types. We present the associated variations in runoff, air temperature, and conductivity under the two climate change scenarios. While the predictions for the RCP 4.5 in the period 2070–2100 are moderate (Table 3), those under the RCP 8.5 scenario show significant changes to the main physiographic characteristics of many river types. Runoff (19 to 36% decrease), air temperature (increase of 4–5.3 C), and water conductivity (increase between 10 and 13%) changes will be striking at the 2070–2100 horizon. The outputs from the RCP 8.5 simulation led to the worst-case situation in the southern part of the Peninsula, with runoff reductions reaching nearly 40% in the river types of the Southeast. The hydrological alterations under this scenario may lead to longer low-flow periods and drastic changes in water chemistry, particularly in the mineralized river types of Groups 2, 3, and 4.

Table 3 Variations in runoff, air temperature, and conductivity predicted under the Representative Concentration Pathways (RCP) 4.5 and 8.5 for the impact period 2070–2100, for each river type in the region based on the IPCC 2013 (CEDEX, 2017)
Full size table

Since hydrological data were not available for most of the sites, and therefore, they could not be used for modelling, the future effects of runoff reduction had to rely on the autecology of the diatom taxa. That is, we used the current environmental attribution of diatom taxa to define which changes would occur in the diatom assemblages in the event of the environmental changes predicted under the RCP 8.5. Then, using the projected physiographic values of the national river types in Table 3, we propose a new configuration of end-group types, which contemplates a displacement in the inter-group classification (Fig. 5).

Fig. 5
figure 5

Simulated displacement of end-group types of reference sites, based on the projected physiographic values of the national river types of Table 3 for the impact period 2070–2100. Displacements are indicated with dotted-lined boxes and red arrows

Full size image

This simulation (Fig. 5) shows that differences between the diatom assemblages in the Northern and Southern Spain, and between siliceous and calcareous regions (roughly corresponding to the Western and Eastern parts, respectively; Sabater et al., 2022), will become more pronounced. In this new arrangement, Groups 2, 3, and 4 become more alike to each other than presently are. Mediterranean and Mediterranean-continental axes (Group 4), and mineralized rivers (Groups 2 and 3), account for the highest impact (Table 3), and will become similar under future climate change conditions. In this future, scenario terrestrial and aerophilic diatoms will be the best adapted taxa to the harsh conditions imposed by runoff reduction. Also, brackish diatoms will be more frequent due to the increase in water mineralization. As such, the presence of aerophilic taxa such as S. stroemii, L. mutica, H. contenta, and H. montana (Delgado et al., 2013; Novais et al., 2014; Falasco et al., 2016; Tornés et al., 2021) and brackish taxa such as H. tenerrima, N. inconspicua, H. coffeaeformis, and B. aponina (Ros et al., 2009; Heudre et al., 2020), might become more prevalent. On the other hand, sensitive freshwater species to warmer and lentic conditions such as A. pyrenaicum, G. angustivalva, and D. vulgaris may become less common (Centis et al., 2010; Cantonati et al., 2012; Falasco et al., 2020).

In the new arrangement derived from the RCP 8.5 scenario, mountain rivers (Group 6) may become closer to small and poorly mineralized mid-altitude rivers (Group 5), then becoming more distinct than the present Cantabrian-Atlantic rivers (Group 7). The river types in the Atlantic biogeographic region (Group 7) and those of Group 1 which assemble poorly mineralized lowland streams in the Mediterranean region, will likely become apart from all other groups. Such a singularity is under the assumption that these river types will be the least affected under the RCP 8.5 conditions. However, even in the Cantabrian-Atlantic rivers, presently the wettest parts of the Iberian Peninsula, terrestrial, and aerophilic diatoms are already present, with a high group specificity. This may indicate a potential risk if the lentification of these systems is progressing, which needs to be confirmed by future investigations. In these systems, lentic (planktic) species might be also abundant, replacing large-sized taxa in presently lotic areas (B-Béres et al., 2019).

Overall, our predictions may be summarized by saying that climate change may likely act by reducing the differences between river types, and causing that diatom assemblages could simplify their composition towards higher homogenization. Biotic homogenization has been described as one of the most likely effects promoted by climate change (e.g. Tornés & Ruhí, 2013; Petsch, 2016; Falasco et al., 2021), with likely implications for loss of resistance and resilience of biological communities, and alteration of related ecosystem functions.

Finally, the limitation of hydrological data on the reference sites leads to a general concern. The scarcity of hydrological data in our study highlights a global problem, especially in the case of non-perennial rivers and streams (Messager et al., 2021). Hydrological characterization of reference conditions is limited in many river networks, due to their poor instrumentation in many tributaries. In fact, most gauging stations are installed on large, perennial rivers worldwide (Zimmer et al., 2020), while headwaters are underrepresented. A temporal bias is added to this spatial constraint, since in some monitoring schemes, gauging stations may change location or data frequency over time. Advancing on effects of climate change on water flow and biological communities in reference sites requires shifting from the current scarcity of primary hydrological data to a planned, long-term monitoring including hydrological and biological data (Zimmer et al., 2020; Messager et al., 2021). These are essential requisites to produce reliable predictions on the effects of climate change in riverine biota.