Environmental and spatial effects on co-occurrence network size and taxonomic similarity in stream diatoms, insects, and fish

Aim: The influences of environmental and spatial processes on species composition have been at the center of metacommunity ecology. Conversely, the relative importance of these processes for species co-occurrences and taxonomic similarity has remained poorly understood. We hypothesized that at a subcontinental scale, shared environmental preference would be the major driver of co-occurrences across species groups. In contrast, co-occurrences due to shared dispersal history were more likely in dispersal-limited taxa. Finally, we tested whether taxa co-occurring due to similar responses to environmental and spatial processes were more taxonomically similar than expected by chance.

Location: The conterminous United States

Time Period: 1993-2019

Major taxa studied: Stream diatoms, insects, and fish

Methods: We generated co-occurrence networks and developed methodology to determine the proportions of nodes and edges explained by pure environment alone (after accounting for space), pure space alone (after accounting for the environment), pure environment and pure space together, and spatially structured environment. Taxonomic similarity of taxa co-occurring because of environmental and/or spatial controls or because of unmeasured processes was compared to that of a null model.

Results: Pure environment alone, spatially structured environment, and pure environment and pure space together explained the greatest proportion of nodes and edges in the co-occurrence networks of diatom species and genera, and insect genera. Conversely, pure environment and pure space together best explained the nodes and edges in the co-occurrence network of fish species and genera. Co-occurring taxa were more closely related than the random expectation in all 30 cases.

Main Conclusions: The environment controlled co-occurrences of all groups, while the influence of space was the strongest in fish, the most dispersal-limited group in our study. All co-occurring taxa were more taxonomically related than expected by chance due to environmental or spatial overlap or unaccounted factors.

Methods
We obtained data from Passy et al., (2023), including diatoms, insects, and fish, collected from, respectively, 1698, 1700, and 1700 stream sites across the conterminous United States (Fig. S1). Sites were compiled from both the National Water-Quality Assessment (NAWQA) program of the US Geological Survey and the National Rivers and Streams Assessment (NRSA) of the US Environmental Protection Agency, which used similar collection methods (Moulton II et al., 2002; US Environmental Protection Agency, 2013). Streams were sampled between 1993 and 2019, with the majority of samples collected between 2007 and 2010. Diatoms and insects were collected from a predetermined area of substrate from May to September. Fish were sampled throughout the year using backpack electrofishing and seining. Community data consisted of counts of species or a lower taxonomic category in diatoms and fish, but genera in insects. To standardize the sampling effort, diatom counts were sampled down to 400 cells, and the insect and fish counts, to 100 individuals. We accounted for differences in taxonomic resolution by examining all datasets at a genus level in addition to the species level for diatoms and fish.

We selected our 1698-1700 sites from larger datasets of 2278 diatom, 2270 insect, and 2296 fish samples, ensuring similar environmental conditions, a minimum pairwise distance between any two sites of 1 km, and comparable average pairwise distance among datasets (1522, 1497, and 1491 km, respectively), which we provide here. Pairwise distances between all sites were calculated in km using the ‘distm’ function in the R package ‘geosphere’ (Hijmans et al., 2019). Each site had available physicochemical data on pH, specific conductance (µS/cm), water temperature (C°), nitrate (NO3, µg/L), and total phosphorous (TP, µg/L) from NAWQA or NRSA. Elevation (m) was obtained from the WorldClim database (Fick & Hijmans, 2017) and used to calculate slope (% grade), averaged across a 5 km buffer around each site with the package ‘raster’ (Hijmans et al., 2020). There were 19 climatic variables that described temperature and precipitation minima, maxima, averages, ranges, and seasonality across a 5 km buffer around each site (WorldClim V1.4) (Hijmans et al., 2005). Environmental variables were ln-transformed if normality was improved. The slope was arcsine square root-transformed. All variables were standardized (mean = 0, standard deviation = 1).

We also assembled higher-order taxa for each metacommunity, consisting of phylum, class, order, family, genus, species, subspecies, variety, and form for diatoms; phylum, class, order, family, and genus for insects, and finally, phylum, class, order, family, genus, and species for fish.