Comparing surface effective saturation and depth-to-water-level as predictors of plant composition in a restored riparian wetland

Depth-to-water-level (DTWL) measurements in shallow groundwater piezometers are commonly used to develop predictive relationships between wetland plant composition and the water regime. These analyses, however, implicitly use DTWL as a surrogate for the soil water regime within the root zone. We collected bi-weekly field measurements of DTWL and surface effective saturation (SES) at a riparian wetland (34 years after restoration) in southwestern Wisconsin during the 2009 and 2010 growing seasons. Plant species composition was also sampled at the same locations (N = 62). Nonmetric multidimensional scaling (NMS) and nonparametric multiplicative regression (NPMR) were used to compare how effectively the two hydrological metrics explain the overall plant community ordination space (NMS) and predict the probability of presence of certain dominant species (NPMR). In addition, we performed each statistical method on each year's data separately in order to estimate the robustness of the metrics to predict species composition. Both SES and DTWL metrics were significantly correlated with one NMS axis for both years, although correlations were higher and more significant (p < 0.001) with the SES metrics. NPMR-generated models were created for six representative species using the SES and DTWL metrics as predictor variables. Models created using the SES metrics had consistently higher evaluation metrics for both years as compared with the DTWL metrics. While the metrics that consistently led to models with the highest evaluation metric (log beta > 2.2) for each of the species were mean and minimum SES, models that used mean SES were more temporally consistent and, therefore, more generally applicable. Copyright (C) 2011 John Wiley & Sons, Ltd.