Modeling to Improve Vegetation-Based Wetland Biological Assessment

Document Type


Publication Date



The accurate and precise biological assessment of wetland ecosystems has proven to be a significant challenge to natural resource managers. Biological assemblages in wetland ecosystems are highly variable and this variability can confound inferences of biological condition resulting from biological assessments. Efforts to control for this natural biological variation have led to the development of many different biological assessment indices that are based on classification. Classification-based indices often lack broad applicability and may not adequately control for natural sources of biological variation. Biological variation is often associated with natural environmental gradients that modeling techniques may be able to account for. The general goal of my thesis research was to develop a model-based biological assessment index for wetlands in Ohio, to determine if modeling could improve the performance of the wetland assessment indices that are currently available. I developed two types of model-based biological indices for Ohio wetlands, a vegetationbased index of biological integrity (MVIBI), and several indices of plant assemblage taxonomic completeness (O/E). The MVIBI exhibited significant improvements in performance over previously developed vegetation-based indices of biological integrity (VIBIs). Modeling also iv accounted for enough biological variation to permit the assessment of three wetland types with a single index. Use of the MVIBI should increase manager’s confidence in plant-based wetland assessments and improve wetland assessment comparability. The plant-based O/E indices performed poorly relative to O/E indices that have been developed for other types of assemblages (i.e. macroinvertebrates, fish), indicating that the plant-based O/E indices are unlikely to detect biological degradation. The poor performance of these indices was related to poor predictability of individual plant taxa. Plant taxa occurrence is strongly related to the timing and intensity of stochastic disturbance events and complex biotic interactions that are difficult to quantify. These factors present challenges to predicting the presence and absence of individual plant taxa. My results provide insight into the ways that modeling may and may not be used to predict plant assemblage composition and should help index developers improve the performance of plantbased biological indices.


roceedings of the Society of Wetland Scientists Annual Meeting


June 2-6, 2013.