Modeling Regional Variation in Riverine Fish Biodiversity in the Arkansas–White–Red River Basin

Abstract

The patterns of biodiversity in freshwater systems are shaped by biogeography, environmental gradients, and human-induced factors. In this study, we developed empirical models to explain fish species richness in subbasins of the Arkansas–White–Red River basin as a function of discharge, elevation, climate, land cover, water quality, dams, and longitudinal position. We used information-theoretic criteria to compare generalized linear mixed models and identified well-supported models. Subbasin attributes that were retained as predictors included discharge, elevation, number of downstream dams, percent forest, percent shrubland, nitrate, total phosphorus, and sediment. The random component of our models, which assumed a negative binomial distribution, included spatial correlation within larger river basins and overdispersed residual variance. This study differs from previous biodiversity modeling efforts in several ways. First, obtaining likelihoods for negative binomial mixed models, and thereby avoiding reliance on quasi-likelihoods, has only recently become practical. We found the ranking of models based on these likelihood estimates to be more believable than that produced using quasi-likelihoods. Second, because we had access to a regional-scale watershed model for this river basin, we were able to include model-estimated water quality attributes as predictors. Thus, the resulting models have potential value as tools with which to evaluate the benefits of water quality improvements to fish.

Received July 1, 2010; accepted March 7, 2011

ACKNOWLEDGMENTS

We thank Jason McNees (NatureServe) for providing us with fish richness data for the AWR aggregated by HUC8. We are grateful to Latha Baskaran, Oak Ridge National Laboratory (ORNL), who kindly provided value-added geographic data including LULC, elevation, and hydrologic information and ran the SWAT model to produce water quality and quantity predictors used in this study. Craig Brandt (ORNL) provided advice with integrating and analyzing our data. Mark Bevelhimer (ORNL) provided us with a thorough review, and we are grateful to the anonymous reviewers whose comments helped us to improve the manuscript substantially. This research was sponsored in part by the Department of Energy Office of Biomass Programs and in part by the Laboratory Directed Research and Development Program of ORNL, which is managed by UT-Battelle, LLC for the U.S. Department of Energy under Contract No. DE-AC05–00OR22725. The U.S. Government retains and the publisher, by accepting the article for publication, acknowledges that the U.S. Government retains a nonexclusive, paid-up, irrevocable, world-wide license to publish or reproduce the published form of this manuscript, or allow others to do so, for U.S. Government purposes.

Notes

^aAll land cover values are percentages of total subbasin area.

^aThe model terms for fixed effects are log-transformed mean annual discharge (LQ; m³/s), elevation of the subbasin outlet (ElDr; m), number of dams downstream to the most-downstream discharge point in the basin (NDams), percentages of the total subbasin covered by forest (%For) and shrubland (%Sh), and SWAT-derived estimates for nitrate (NO₃N; mg/L), total phosphorus (TP; mg/L), and sediment concentration (Sed; g/L).