Toward bioassessment without reference sites using species distribution modeling
Authors: Dean A. Hendrickson, Benjamin Labay, Adam E. Cohen, Timothy H. Bonner, Ryan S. King, Leroy Kleinsasser, Gordon Linam, Kirk O. Winemiller
Year: 2012 (xliv)
Abstract
Use of least-disturbed sites for reference conditions in bioassessment can be subjective, is region-specific, and can contribute to management for steadily declining ecosystem health (shifting baselines). We explore an alternate approach to bioassessment that uses species distribution models to construct a fish community model across an environmentally diverse landscape (the entire state of Texas) to provide landscape-wide (at 1 km2 resolution) reference state benchmarks of taxonomic completeness of fish communities. Using museum specimen-based occurrence records we constructed distribution models for 100 fish species to compare model-based predicted community composition against empirical fish community data from four independent surveys that independently sampled 269 sites broadly distributed across the study area. Two surveys had repeated-sampling protocols allowing for more rigorous model evaluation, and two had associated multimetric-based index of biotic integrity (IBI) scores and used methodologies characteristic of state and federal agency bioassessment efforts. Numbers of species predicted by the models were moderately to strongly correlated with observations from all surveys. Deviations were correlated in ways that indicate influence of species-specific prevalence, environmental quality, geographic patterns of species richness, and survey sampling intensity. We found significant, though weak, relationships between observed/predicted ratios and IBI scores. Site-specific values from the National Fish Habitat Assessment Plan (NFHAP) index showed weak and non-significant relationships to both our observed/predicted ratios and to IBI scores. The results of model-based assessment were similar to those of traditional methods based on reference sites, however, confounding influences of modeling artifacts contributed to over-prediction of richness that warrants further exploration. The model-based approach, especially in light of rapidly improving modeling methods, offers a promising heuristic platform for investigation and application of assessments at broad scales, and is immediately useful as a complement to traditional assessment methods.