Assessing Site-Selection Strategies for Modeling the Influence of Landscape Factors on Stream Fish Assemblages
Lindsey A. Bruckerhoff and Keith B. Gido
Abstract.—Linking landscape features, both natural and human-altered, to aquatic ecosystem structure and function is a fundamental objective in landscape ecology and freshwater science, but this process is data- and resource-intensive. Quantifying how landscape stressors influence aquatic communities requires balancing logistic and financial constraints with effectively sampling the landscape to capture gradients of interest. There are a variety of ways to balance these constraints, such as using existing data, handpicked site selection, or a statistical site-selection scheme. Poor sampling design reduces statistical power; however, we do not know how differences in site-selection designs influence our ability to measure ecological responses to landscape gradients. We quantified how the distribution of sample sites across landscape gradients affected the measured responses of stream fish assemblages to these gradients at different sample sizes. Specifically, we used randomization tests to compare the variability in the responses of fish assemblage structure (species richness and composition) to catchment area and land use (agricultural land) with manipulated distributions (random, highly skewed, and uniform) of sites across these landscape gradients. Assemblage composition was more sensitive than species richness to sampling design, and we observed less variability in the detected response of assemblage composition when samples were distributed uniformly across landscape gradients, especially when sample sizes were small. Although strong responses to environmental gradients, such as species richness to catchment area, are robust to sampling distributions, large sample size and a uniform distribution of samples might help elucidate more subtle responses to environmental gradients.