GRADIENT ANALYSIS OF DIATOM ASSEMBLAGES IN WESTERN KENTUCKY WETLANDS 1

Diatom and water chemistry data from 35 wetland sites in western Kentucky were used to assess diatoms as indicators of ecological conditions in wetlands. The wetlands were affected by different degrees of acid mine drainage and agriculture. Canonical correspondence analysis indicated that the distribution of diatoms was highly correlated with conductivity and total phosphorus (TP), two variables commonly associated with acidic mine drainage and agriculture, respectively. Diatom‐based inference models were developed for use as quantitative indicators of two important environmental variables in wetlands: conductivity and TP. Diatom‐inferred conductivity and TP values were highly correlated with measured values. Cross‐validation with jackknifing procedures suggested that high apparent r 2 between inferred and measured conductivity was overly optimistic and should be treated with caution. Jackknifing‐derived TP inference models performed poorly in predicting TP toward the ends of low and high TP concentrations. In general, the conductivity inference models based on plankton had better predictability than those based on epiphyton. Epiphyton‐based inference models can predict TP better than plankton. Therefore, diatoms in planktonic and epiphytic assemblages could provide complementary information on ecological conditions. Our data suggest that diatoms can reflect major regional environmental gradients and therefore can be used as indicators of the ecological conditions in wetlands. Quantitative inference models with known predictive power can aid the development of realistic and ecologically sound biotic indices for this region.