Executive Summary: Summary: Multivariate statistical techniques were used to define a method for establishing a water quality index (WQI) for use in protecting the stream environment in a high mountain watershed. The purpose of the WQI was to aggregate water quality parameters in such a way that the effects of low level increments in mining, grazing, logging and other activities could be related to a change in the value of a single entity, aquatic environmental aquality, in a linear programming (LP) management model. Several data aggregation methods were explored, using water quality data collected over 5 years (1975-1979) by the USDA Forest Service in the upper Blackfoot River watershed in southeastern Idaho. The WQIs thus generated were compared with indices of benthic invertebrate community composition as determined from samples collected late in the summer of 1981. Community composition indices were based on emergent community properties (biomass and diversity) and on taxonomic composition as revealed by principal componenets analysis. Significant results of the study include the following: 1. Existing determinisitic general purpose WQIs (such as the National Sanitation Foundation Index) proved useless for guidance in protecting water quality in these high mountain watersheds, because stream water quality often remains excellent by drinking water standards, even though subtle changes in water quality parameters may significantly affect instream habitat. 2. An increasing scale, multivariate statistical WQI was created for the study area using 5-year May-October averages of ten water quality variables in eight streams. Removal of some streams from the data set, as well as aggregating or replacing some variables, did not significantly alter the rank order of the stream WQI values. 3. Changes in the calculation time step to 5-year bimonthly (May-June, July-August, September-October) or monthly averages, or to annual averages for four water years, provided little additional information, and resulting in decreasing sensitivity to changes in water quality variables because of larger standard deviations in the data sets. 4. The WQI was composed of four principal components that were easily interpretable as common factors (e.g., nutrient sources, suspended sediment sources, groundwater, and discharge) that affected groups of variables. These principal components, or subindices, were positively correlated with the presence of certain benthic invertebrate taxa or groups of taxa. 5. Cluster analysis was useful in reducing the dimensionality of water quality data and in revealing relationships among invertebrate communities (Q-type analysis). However, R-type cluster analysis of the study streams showed no similar groups of streams based on water quality variables. 6. The WQI was highly negatively correlated (r^2=0.93) with benthic invertebrate standing stock biomass, a relationship described by a decreasing power function. There was no apparent relationship between benthic invertebrate diversity and the WQI values. 7. The WQI-biomass relationship may be useful in setting a constraint value on the WQI in the LP model. Additional data and information from more sites should be collected and analyzed, however, to stengthen the confidence in the correlation, and to establish causality between the variables contributing strongly to the WQI and community biomass. 8. The multivariate WQI was found to be heavily influenced by the relative standard deviations of the variables used to form the index. Inclusion of only similar (pristine) streams in a baseline data set will result in a lower standard deviation for each variable. The result is higher sensitiviely to a given polluting factor than will be found in a mixed group of streams in which some are already impacted by anthropogenic activities. 9. High standard deviations for individual variables may mask relationships between environmentally significant parameters and biological communities. 10. Multivariate WQI indexing provides valuable insights into the relationship between water quality and biological community composition, even if appliaction of the WQIs in predictive settings is premature or ultimately proves to be unacceptable. Suggestions for Further Research: 1. Collect additional invertebrate data at other seasons and on Upper Angus and Mabie Creeks in order to reinforce or refine the relationships reported here. 2. Collect more detailed habitat data in order to elucidate the relative importance of water quality and physical habitat in controlling benthic community composition. Artificaial substrates may be useful in reducing physical habitat dissimilarity in order to focus on water quality effects. 3. Update the WQI using 1980-1982 data from the Forest Service and look for recent trends that would reinforce or alter the conclusions based on the older data. 4. Examine the use of standardized extreme values, rather than standardized means, to create a WQI. 5. Employ cannonical correlation as a means of further elucidating water quality-physical substrate-benthic community relationships. 6. Monitor changes in the WQI and benthic invertebrate community in one of the study streams in response to changing management practices (e.g., erosion control or additional phosphate mining). 7. Investigate the effects of a proposed change in management practice on a WQI using Monte-Carlo analysis to account for simultaneous changes in many variables. 8. Investigate the response of the invertebrates in principal components 1 and 4 to nutrients and suspended sediments in controlled (artificial) ecosystems to test their suitability as water quality indicators in the study area.
Mahmood, Ramzi; Messer, Jay J.; Nemanich, Frank J.; Liff, Charles I.; and George, Dennis B., "A Multivariate Water Quality Index for Use in Management of a Wildland Watershed" (1982). Reports. Paper 231.