Observational data are frequently used to better understand the effects of changes in P and N on stream biota, but nutrient gradients in streams are usually associated with gradients in other environmental factors, a phenomenon that complicates efforts to accurately estimate the effects of nutrients. Here, we propose a new approach for analyzing observational data in which we compare the effects of changes in nutrient concentrations in time within individual sites and in space among many sites. Covarying relationships between other, potentially confounding environmental factors and nutrient concentrations are unlikely to be the same in both time and space, and, therefore, estimated effects of nutrients that are similar in time and space are more likely to be accurate. We applied this approach to diatom rbcL metabarcoding data collected from streams in the East Fork of the Little Miami River watershed, Ohio, USA. Changes in diatom assemblage composition were consistently associated with changes in the concentration of total reactive P in both time and space. In contrast, despite being associated with spatial differences in ammonia and urea concentrations, diatom assemblage composition was not associated with temporal changes in these nitrogen species. We suggest that the results of this analysis provide evidence of a causal effect of increased P on diatom assemblage composition. We further analyzed the effects of temporal variability in measurements of total reactive P and found that averaging periods greater than ~1 wk prior to sampling best represented the effects of P on the diatom assemblage. Comparisons of biological responses in space and time can sharpen insights beyond those that are based on analyses conducted on only 1 of the 2 dimensions.
Keywords: Bayesian hierarchical model; DNA metabarcoding; agriculture; algae; biomonitoring; nitrogen; periphyton; phosphorus; rbcL; temporal variability.