A major goal of community ecology is understanding the processes responsible for generating biodiversity patterns along spatial and environmental gradients. In stream ecosystems, system specific conceptual frameworks have dominated research describing biodiversity change along longitudinal gradients of river networks. However, support for these conceptual frameworks has been mixed, mainly applicable to specific stream ecosystems and biomes, and these frameworks have placed less emphasis on general mechanisms driving biodiversity patterns. Rethinking biodiversity patterns and processes in stream ecosystems with a focus on the overarching mechanisms common across ecosystems will provide a more holistic understanding of why biodiversity patterns vary along river networks. In this study, we apply the Theory of Ecological Communities (TEC) conceptual framework to stream ecosystems to focus explicitly on the core ecological processes structuring communities: dispersal, speciation, niche selection, and ecological drift. Using a unique case study from high elevation networks of connected lakes and streams, we sampled stream invertebrate communities in the Sierra Nevada, CA to test established stream ecology frameworks and compared them to the TEC framework. Local diversity increased and β-diversity decreased moving downstream from the headwaters, consistent with the river continuum concept and the small but mighty framework of mountain stream biodiversity. Local diversity was also structured by distance below upstream lakes, where diversity increased with distance below upstream lakes, in support of the serial discontinuity concept. Despite some support for the biodiversity patterns predicted from the stream ecology frameworks, no single framework was fully supported, suggesting “context dependence”. By framing our results under the TEC, we found species diversity was structured by niche selection, where local diversity was highest in environmentally favorable sites. Local diversity was also highest in sites with small community sizes, countering predicted effects of ecological drift. Moreover, higher β-diversity in the headwaters was influenced by dispersal and niche selection, where environmentally harsh and spatially isolated sites exhibit higher community variation. Taken together our results suggest that combining system specific ecological frameworks with the TEC provides a powerful approach for inferring the mechanisms driving biodiversity patterns and provides a path toward generalization of biodiversity research across ecosystems.

Study Area

The study area was located in the Sierra Nevada Mountains of eastern California (USA) and encompasses portions of Inyo National Forest and Sequoia-Kings Canyon National Park. Over the ice-free seasons (June-September), we sampled five distinct lake-stream networks, where each network was within a spatially distinct catchment and were treated as independent replicate systems (Fig. 3). The Kern (n=24) and Bubbs (n=26) networks were sampled in 2011, the Evolution (n=21) and Cascades (n=11) networks in 2018, and Rock Creek (n=36) in 2019. For each lake-stream network, streams were sampled throughout the network along a spatial gradient from headwaters downstream as well as along a spatial gradient downstream from lakes. Because the spatial distances of the river networks and the distance separating lakes naturally vary among networks as well as backcountry sampling constraints, the number of sites sampled along the distance from headwaters gradient varied (n=11 to n=36) and the downstream lake gradient varied (n=1 to n=9). This field system and the data collected naturally provide spatial gradients relevant to test stream ecology theories. In addition, this data is ideal for testing TEC processes because of the naturally varying gradients of community size, connectivity, and environmental heterogeneity present in our sampling design.

Field Methods

At each sampling location, we established transects in riffle sections of streams. At five equally spaced points along transects we measured stream depth and current velocity at mid-depth using a portable flow meter (Marsh-McBirney Flow Mate 2000). We then calculated stream discharge as the sum of the product of average depth x current velocity x width/5 over all transect points (Gordon et al. 2010; Herbst et al. 2018). A calibrated YSI multiparameter device was placed above transects to measure temperature, dissolved oxygen, conductivity, and pH. Benthic chlorophyll data was collected by scrubbing the entire surface area of three randomly selected cobble sized rocks (64-255 mm) of benthic algae (periphyton) with a toothbrush for 60 seconds (Herbst and Cooper 2010). Chlorophyll measurements were taken using a handheld fluorometer (Turner Designs Aquafluor), which measures raw fluorescence units. Florescent measurements were calibrated to chlorophyll concentration using a known concentration of Rhodamine. We standardized chlorophyll measurements by accounting for both the surface area of rocks and volume of water used to remove algae.

Eight to twelve macroinvertebrate samples at each site were collected using a D-frame kick net (250 mm mesh, 30cm opening, 0.09m2 sample area) in riffle habitats, depending on the density of macroinvertebrate samples collected. We took samples by placing the net on the streambed, then turning and brushing all substrate by hand in the sampling area (30cm x 30cm) immediately above the net, with dislodged invertebrates being carried by currents into the net. All macroinvertebrate samples were preserved in 75% ethanol within 48 hours of sampling. Samples were sorted, identified, and counted in the laboratory. Taxa were identified to the finest taxonomic level possible, usually to genus or species for insects (excluding Chironomidae) and order or class for non-insects (Merritt, Cummins, and Berg 2019). The replicate samples taken at each site were pooled together and divided by the number of replicates and the area sampled to determine the density of invertebrate communities.

Spatial Data

Stream distance measurements were taken using the R package “riverdist”, which utilizes data from the USGS National Hydrological Dataset Flowline in order to determine pairwise distances from sampling sites along the river network (Tyers 2020). We determined distance below upstream lakes, with the closest upstream lake location being the outlet of the lake determined by the USGS Watershed Boundary Dataset. For sites where multiple upstream lakes were draining into streams, we defined the upstream lake as the closest upstream lake to sites that was also along the mainstem of the flowline. We determined distance from headwaters as the streamwise distance from sites to the uppermost portion (headwaters) of the mainstem of streams, where the headwaters of streams was determined by the endpoint (beginning) of the flowline in the USGS NHD Flowline Dataset (U.S. Geological Survey 2016). In cases where multiple headwater stream reaches corresponded to downstream sites, we defined the headwaters as the particular reach that accounted for the most discharge which was determined using USGS Flowline Dataset. Upstream lake area and perimeter measurements were determined using the USGS Watershed Boundary Dataset. Land-cover proportions were computed using the 2016 USGS National Land Cover Database (Jin et al. 2019).