Community size and disturbance history jointly explain the interplay between stochastic and deterministic community variation in benthic invertebrates

Community ecology has so far struggled to integrate both deterministic and stochastic processes into a global model of community variation. To address this issue, we aimed to characterise the ecological conditions that determine the relative importance of deterministic and stochastic community variation in benthic macroinvertebrates. We sampled macroinvertebrate and algal periphyton communities, and microhabitat conditions at monthly intervals over a 5 year period in two sampling sites in the regulated Durance River and one site in the Asse River (a natural tributary of the Durance). The relative importance of stochastic and deterministic processes was estimated over time as functional and taxonomic spatial β-deviation (i.e. β-diversityobs – mean β-diversitynull / SD β-diversitynull) for each sampling date and site. We additionally quantified deterministic variation as the trait-environment relationship and predicted a positive correlation with β-deviation metrics. We found evidence of overdispersal in both functional and taxonomic β-diversities compared to null expectations. Spatial β-deviation was highly variable over temporal scales and was jointly explained by invertebrate community size, disturbance (i.e. flooding), periphyton diversity and to a lesser extent, microhabitat diversity. The key factor that explained β-deviation was recent river discharge (< 90 days), which had a strong negative effect on community size but was also directly associated with higher β-deviation, likely related to recolonization processes post-flood. Lastly, we found that when β-deviation was high, the hydraulic trait-environment relationship was stronger. This indicates that spatial heterogeneity in hydraulic conditions was the main driver of deterministic variation in invertebrate community structure. This study demonstrates that the importance of stochastic processes can vary greatly over short (~months) and long (~years) temporal windows with significant consequences for trait-environment relationships. Our study provides an example of how stochastic community variation can be modelled to better interpret deterministic patterns of community structure in natural communities.

Methods
Hydrological conditions

We obtained hourly hydrological data (Q m3s-1) for the period of December 2012 to August 2018. Hydrological data for the two Durance sites was obtained from the Escale hydroelectrical dam (28.3km & 34.7km upstream of the Dur-1 and Dur-2 sites, respectively). We also obtained hydrological data for the Bléone River (906 km² drainage basin) as both Durance sites were situated downstream from its confluence with the Durance (Figure 1). The Bléone River is also regulated by a dam but as discharge measurements were not available. We obtained simulated discharge values inferred from reservoir water levels for the sample period from the French electrical company (EDF). These simulated discharge values were then added to the Durance hydrograph. For the Asse site, we obtained hydrological data from the Clue Chambrières limnology station (14.7km upstream from the Asse study site). We also added the input of the Asse to the Dur-2 hydrograph, as Dur-2 is located downstream from the confluence between the Asse and Durance rivers. Lastly, we standardised river discharge (Q m3s-1) to account for differences in the size of catchment basins between our sampling sites. River discharge was converted to mm/h and then standardised by catchment area using the following formula: Q(l/s)*T(s) / (S km² *1000*1000). Where Q is discharge (in l/s), T is the number of seconds in the desired time unit (i.e. 3600 seconds in an hour) and S is the surface area of the catchment basin (Dur-1 = 8131km², Dur-2 = 9185km² and Asse = 538km²).

Macroinvertebrate and periphyton sampling protocol

We performed monthly sampling beginning in late winter (~February) and ending in early summer (~July) from 2014 to 2018 (Table 1). We selected this period in order to capture the main growth period of invertebrate and periphyton communities in our study sites (Cazaubon & Giudicelli, 1999).

Sampling was designed to be representative of the various hydrologic ambiances found in riffle habitats. To this end, we defined three hydraulic ambiances: 30-50 cm s-1, 70-90 cm s-1 and > 100 cm s-1, wherein we obtained three macroinvertebrate and periphyton samples at each sampling date (i.e. three samples per ambiance = nine samples total). Samples were distributed haphazardly within each ambiance with a minimum of three meters between two samples. The length of the sampling reach varied according to the availability of the three ambiances, ranging from 50m to ~1000m in length. Macroinvertebrate samples were obtained using a Surber sampler (0.10 m²) with a 0.25 mm mesh. Samples were immediately stored in 70% ethanol, awaiting morphological identification. In the laboratory, macroinvertebrate samples were rinsed and further sieved (0.63 mm mesh) before being counted and identified under binocular microscope. Species- or genus-level Identification was obtained for the majority of specimens with the exception of Diptera (which were identified to the family-, sub-family- or tribe-levels) and other groups like Oligochaeta and Hydrachnidia which could not be reliably identified. Nine periphyton samples were obtained in parallel to each of the nine macroinvertebrate samples in the same sample area. A periphyton sample was constituted of three medium-sized stones found beside the Surber sample. The biofilm on each stone was scrapped from a standardised 25 cm² area (drawn on by pencil), using a scalpel and/or nylon brush and transferred to a container with a solution of formaldehyde 2-5%. The resulting solution from each of the three stones (total area sampled = 75 cm²) was then pooled into a single container. Note that the area sampled changed after 2014 (when 100 cm2 area was sampled on each stone) to account for the large amount of material that was collected. In the laboratory, each sample was brought to a fixed volume of liquid (50 mL) and put into suspension to homogenize the samples. Counting and identification of the periphyton was performed using a microscope and expressed as the number (of cells) per cm2.

Micro-habitat description

We described microhabitat conditions that are pertinent to the ecology of benthic organisms (granulometry, clogging, depth, velocity, etc.) at the scale of each Surber sample (i.e. 0.10 m²). The granulometry of the riverbed was quantified using semi-quantitative size-classes described by Malavoi & Souchon, 1989. The substrate size-class scale ranged from 0 (silt: 0.0039 – 0.0625mm diameter) to 10 (bedrock: > 1024mm diameter; Supporting information). We used substrate size-classes to estimate the surface and underlayer substrate richness, largest substrate size and dominant substrate size. Substrate clogging was evaluated visually using a semi-quantitative scale ranging from 1 to 5 (Supporting information) designed to describe the level of substrate embeddedness and clogging of the interstitial space by silt and algae (Archambaud et al., 2005). Current velocity was measured at regular intervals beginning at 3cm from the river bottom and at approximately 0.20, 0.40 and 0.80 water depth, using an electromagnetic current meter (OTT MFT Pro).