A statistical procedure that integrates physical, chemical and biological data is used to operationally define and analyze ecological health. Benthic macroinvertebrates collected from May through September from eight locations along the St. Croix River in Minnesota and Wisconsin were analyzed by canonical correspondence analyses. A sequential multivariate analysis procedure was developed that included: 1) determining which environmental variables explained the most variation in the taxonomic data, individually and collectively, 2) running canonical correspondence analyses on the taxonomic data along with environmental variables chosen from a stepwise forward selection procedure, 3) performing a cluster analysis of the taxonomic data in sampling times to determine artificial biological seasons, and 4) analyzing the seasonal data (three time intervals) separately, as in steps 1 and 2. Changes in community structure were significantly associated with changes in environmental factors (physical and chemical) that varied in time and space. The variables found to be significant were largely those expected to be associated with the variations in the structure of benthic macroinvertebrate community in an unimpacted river. A slight influence of nitrogen could be detected in agricultural areas.