Tim Driskell
Fitchburg State College
M.S. Thesis Proposal
Dr. Neal Anderson/
Dr. Meg Hoey
March 9, 1998

Traditional methods of analyzing benthic water quality have usually involved numerous methods of chemical analysis resulting in various levels of success.  For many years biologists have sought to develop more realistic biotic indices.  Limnologists have often studied aquatic macroinvertebrates in lakes and streams, attempting to correlate total species distribution, diversity,  and known tolerance to pollution with measurable habitat parameters and environmental factors (Jacobson, 1997).  Correlation of physical and biological data seems to suggest that quality and quantity of macroinvertebrate communities may indeed be an indicator of environmental health, although the accuracy of analyses are often questioned (Faith, 1989).  Substrate, vegetation, overhead canopy, annual and seasonal climatic swings, local disturbance,  and other factors create an infinite and ever-changing possibility of habitats.  Taxa richness may or may not indicate overall water quality (Minshall, 1985).  Ratio of tolerant verses sensitive species is also quite variable (Minshall, 1985), often contridicting initial conclusions.  Several methods of data analysis have been investigated to ordinate observations in a reliable and consistent manner (Culp,  1980).  Data often must be clustered, frequently in simple linear or numerical categories with meaningless  endpoints (Matthews, 1991b).  One possible solution involves testing a variety of traditional clustering algorithms.  Many of the methods work well for simple two dimensional data (Fox, 1977).  For macroinvertebrate data collected over a typical temperate-zone growing season, a definite third dimension is evident (Green,  1974).   Communities exhibit an obvious temporal variation  as the season advances (Matthews, 1991a).   This third dimension dramatically increases the difficulty of determining conclusive correlations in temperate-latitude year-long biomonitoring programs.

The recent development and utilization of machine learning language and software may help overcome some of these seasonally imposed limitations.   The computer program known as "Riffle" is one such tool that has shown very promising possibilities (Matthews, 1991a).  It utilizes the principle  of nonmetric conceptual clustering, or the division of datasets into clusters independent of any preconceived differentiations or random endpoints.  My thesis proposes to test this software and its potential for further limnological applications in regional streams.   Many seasons of data have been collected and recorded over the years by members of the Nashua River Watershed Association (NRWA),  but adequate statistical analysis has always proven to be difficult and  frequently inconclusive.  Riffle-type software and its methodology of nonmetric conceptual clustering using machine learning language is quite novel, and has not yet been utilized on local waterways. This new software can be allowed to run for many hours to help determine patterns that previously were nearly impossible to detect using traditional statistical models.  This upcoming sampling season will provide much new data from both Department of Environmental Protection (DEP) biologists and numerous NRWA volunteers (including myself).  Data will be collected from parallel tributaries of numerous streams.  This presents a unique opportunity to investigate aquatic ecosystems using adequate controls,  to attempt to  further develop this tool of the future, and test it's ability to correctly identify known and unknown parameters.
 
 

Literature cited:

Culp, J.M.,  1980. Reciprocal averaging and polar ordination as techniques for analyzing macroinvertebrates communities.  Canadian Journal of Fisheries and Aquatic Sciences. Vol. 37,  1358-1365

Faith, D.P., 1989.  Correlation of environmental variables with patterns of distribution and abundance of command and rare freshwater macroinvertebrates. Biological Conservation, Vol. 50,  77-98

Fox, L.R., 1977. Species richness in streams; an alternative mechanism. The American
Naturalist, Vol. 111,  1017-1021

Green, R.H., 1974.  Multivariate niche analysis with temporally varying environmental factors.  Ecology, Vol. 55,  73-83

Jacobson, D., 1997.  Structure and diversity of stream invertebrate assemblages: the influence of temperature with altitude with latitude. Freshwater Biology, Vol. 38,  247-261

Landis,  W.G., 1993.  Multivariate Analyses of the Impact of the Turbine Fuel Jet-A Using a Standard Aquatic Microcosm Toxicity Test. Environmental Sciences, Vol. 2,  113-130

Matthews, G.B.,   1991a.  Mathematical Analysis of Temporal and Spatial Trends in the Benthic Macroinvertebrate Communities of a Small Stream. Canadian Journal of Fisheries and Aquatic Sciences  Vol. 48,  2184-2190

Matthews, R.A., 1991b. Classification and ordination of limnological data; a comparison of analytical tools. Ecol. Modeling, Vol. 53,  167-187

Minshall, G.W., 1985. Species richness in streams of different size from the same drainage basin. The American Naturalist, Vol. 125,  16-38

Sheldon, A.L., 1981  Habitat selection and association of stream insects; a multivariate analysis. Freshwater Biology, Vol. 11,  395-403