1. Introduction
Biodiversity in streams and rivers is being impacted by multiple anthropogenic stressors (Jackson et al., 2016). To understand these impacts, their functional consequences, and management effectiveness taxonomically highly resolved information with high spatial and temporal resolution is important. However, such information is difficult to obtain through traditional morphological assessments as many invertebrate species are small or present only in juvenile stages that are difficult to identify. Molecular taxonomic approaches, in particular metabarcoding of environmental DNA (eDNA) collected from water, offer a fast and cost-effective way to assess biodiversity and are routinely used in aquatic bioassessments around the world (Deiner et al., 2017). eDNA metabarcoding is based on extracted DNA shed by organisms via sloughed cells, feces, gametes or other particles into the water and is thus a non-invasive method to assess community composition because assessment is based on water rather than organismal bulk samples (Taberlet et al., 2012; Valentini et al., 2016). DNA metabarcoding uses high-throughput sequencing methods to generate comprehensive taxa lists (Brantschen et al., 2022; Leese, Sander et al., 2021). However, since the reference databases used to assign taxonomic names to the obtained sequences are still incomplete (Weigand et al., 2019), not all sequences can be assigned to species level. Therefore, molecular Operational Taxonomic Units (OTUs) that are generated according to genetic distance-based similarity thresholds can be used as surrogates for species. Using OTUs in addition to species can reveal further insights into ecological processes (e.g. Beermann et al., 2018).
Despite the obvious advantages, several factors hinder the direct interpretation of eDNA data (Barnes & Turner, 2016; Harrison et al., 2019). First, the possibility to detect a present lotic macroinvertebrate community can be strongly affected by the selected sampling position in the water. Similar to stratified aquatic environments where non-mixed layers need to be considered in the sampling design (Jeunen et al., 2019; Lawson et al., 2019), eDNA molecule distribution may also differ between different positions in lotic environments, such as the water surface versus the riverbed. Accordingly, small scale differences in sampling position both vertically and horizontally may recover different lotic communities (Berger et al., 2020; Macher & Leese, 2017; Thalinger et al., 2021), producing different perspectives of biodiversity change depending on where sampling is conducted. Alternatively, sampling position may have no effect given that eDNA can be transported over long distances of >12 km in streams (Deiner & Altermatt, 2014), which may homogenize eDNA community signals across sampling positions (Macher et al., 2021). Second, several abiotic factors can influence DNA transport and detectability and may thus distort the inferred community (Barnes & Turner, 2016; Harrison et al., 2019), such as discharge and water temperature. Discharge is an important factor influencing eDNA detectability in streams because high discharge could lead to more species being detected by eDNA signals from transported DNA or whole organisms being swept downstream (Fremier et al., 2019; Shogren et al., 2017; Carraro et al., 2018). In contrast, high discharge can also dilute the eDNA signal thus making it more difficult to detect all present species (e.g., Thalinger et al., 2021), which may particularly impede the detection of rare species that are already at low abundance. In addition to discharge, temperature also affects eDNA detectability (Strickler et al., 2015), either negatively if higher temperatures reduce eDNA persistence due to increased enzymatic activity or positively if higher temperatures increase DNA shedding rates (Jo et al., 2019; Kasai et al., 2020; Strickler et al., 2015). These potentially contrasting effects of discharge and temperature make it difficult to predict how they will affect estimates of community composition .
As a consequence of the phenology of organisms, eDNA detectability in streams might also be influenced by sampling season. Depending on the stream type and region, characteristic abundance patterns can be found for different macroinvertebrate orders, genera and species throughout the year and across years (Cowell et al., 2004, Wagner et al., 2011; Wagner, 2004). In addition, the biology of the different macroinvertebrate taxa has a strong effect on seasonal community composition. One important factor are differences in organismal life cycles. While hololimnic species (species with a fully aquatic life cycle) are presumed to be present in the water the whole year, merolimnic species (species with aquatic larvae and aerial adults) leave the water after hatching and have distinct emergence periods, often lasting up to few months, which can lead to a sudden decline in sampled benthic communities (Füreder et al., 2005; Jackson & Füreder, 2006). Besides life cycle based community composition changes, streams also differ systematically with respect to their functional feeding groups (FFG) in both space and time (Vannote al., 1980). For example, shredders are typically more abundant in autumn, when the amount of allochthonous material in streams is highest (Cummins et al., 1989) and grazers in spring and summer, due to sun exposure supporting the growth of large biofilms. The functional composition of macroinvertebrate communities in the form of different FFG affects ecosystem functioning and is therefore also included in bioassessment approaches (Šporka et al., 2006). Detecting these community dynamics patteres is important in aquatic ecology. There is ample evidence that these types of seasonal differences are reflected in eDNA metabarcoding data (Bista et al., 2017; Dunn et al., 2017; Zizka et al., 2020) and that season or even month of sampling lead to different biological assessment results with eDNA metabarcoding (Jensen et al., 2021; Zizka et al., 2020).
While biodiversity studies addressing larger spatial or temporal scales often suffer from an insufficient resolution (Jackson & Fuereder, 2006; Pilotto et al., 2020), studies using high resolution spatiotemporal data at smaller scales are still scarce. Especially for macroinvertebrates, eDNA metabarcoding data has the potential to assess small temporal and spatial changes in community composition time- and cost-effectively to complement future long-term bioassessment of streams.
Using time series data from a Long-Term Ecological Research (LTER; Mirtl et al., 2018) site, the aim of this study was to test the effect of the sampling position (i.e., different positions in the river’s cross section), discharge and temperature, and sampling season on stream macroinvertebrate community composition determined from eDNA. The time series comprises 102 total eDNA samples taken every two weeks for 15 months (from 24.05.2017 to 29.08.2018) at three sampling positions at the same location: (i) the river surface; (ii) the river bottom; and (iii) the river bank. Community composition was determined using high-throughput mitochondrial cytochrome c oxidase subunit I (COI) gene metabarcoding. We tested three hypotheses:
  1. Change in community composition will be driven by seasonality but, because of turbulent flow and mixing, not or less the sampling position. Differences in seasons will follow a cyclic pattern throughout the year (‘seasonal clock’) irrespective of the sampling position.
  2. Differences in community composition will reflect the diverging life cycles of different mero- and hololimnic taxa. For example, the species numbers of merolimnic taxa like Ephemeroptera, Plecoptera and Trichoptera (EPT) will decline in summer after emergence, while differences in species numbers will be less pronounced for hololimnic taxa like Annelida and Coleoptera.
  3. eDNA metabarcoding will detect seasonal differences in community composition for different functional feeding groups (FFG). In particular, more grazer species will be present in spring and summer, whereas more shredder species will be found in autumn and winter based on seasonal food availability. For parasite species that are dependant on other organisms, overall species occurrence will not follow any seasonal pattern, as it is linked to the presence of different host taxa.
Moreover, we also tested the effects of discharge and water temperature on community composition, although we had no a prioriexpectations of what these effects would be.