Effects of environmental change on plant species density: Comparing predictions with experiments

Ideally, general ecological relationships may be used to predict responses of natural communities to environmental change, but few attempts have been made to determine the reliability of predictions based on descriptive data. Using a previously published structural equation model (SEM) of descriptiv e data from a coastal marsh landscape, we compared these predictions agains t observed changes in plant species density resulting from field experiment s (manipulations of soil fertility, flooding, salinity, and mammalian herbi vory) in two areas within the same marsh. In general, observed experimental responses were fairly consistent with pre dictions. The largest discrepancy occurred when sods were transplanted from high- to low-salinity sites and herbivores selectively consumed a particul arly palatable plant species in the transplanted sods. Individual plot resp onses to some treatments were predicted more accurately than others. Indivi dual fertilized plot responses were not consistent with predictions (P > 0. 05), nor were fenced plots (herbivore exclosures; R-2 = 0.15) compared to u nfenced plots (R-2 = 0.53). For the remaining treatments, predictions reaso nably matched responses (R-2 = 0.63). We constructed an SEM for the experimental data; it explained 60% of the va riance in species density and showed that fencing and fertilization led to decreases in species density that were not predicted from treatment effects on community biomass or observed disturbance levels. These treatments may have affected the ratio of live to dead biomass, and competitive exclusion likely decreased species density in fenced and fertilized plots. We conclud e that experimental validation is required to determine the predictive valu e of comparative relationships derived from descriptive data.