Eklund, N. Rensselaer Polytechnic Institute, email@example.com
Embrechts, M. Rensselaer Polytechnic Institute, firstname.lastname@example.org
Shaw, W. Sullivan College, shaw
Momen, B. Rensselaer Polytechnic Institute, email@example.com
Zehr, J. P. Rensselaer Polytechnic Institute, firstname.lastname@example.org
APPLICATION OF AN ARTIFICIAL NEURAL NETWORK FOR INVESTIGATION OF EFFECTS OF ACIDIC DEPOSITION ON ADIRONDACK LAKE COMMUNITIES.
Reduction in emissions and deposition of acidic sulfur compounds over the past two decades have raised the issue of whether biological communities of acidified Adirondack lakes will now recover. The evaluation of recovery is dependent upon understanding the complex factors that control aquatic community structure. Biological data are often poorly suited for parametric statistical analyses. We used artificial neural networks (ANNs) technique to relate biological and chemical data in 30 lakes that range in acidity. ANNs are advantageous because ANN is free of several major assumptions required in regression analysis, and can be performed even when the number of independent variables (predictors) exceed the number of observations (cases). Results indicated that in the study lakes, (1) fish presence is positively correlated with two types of adult crustaceans (Leptodiaptomus minutus and Mesocyclops edax), but negatively corelated with an unidentified rotifer (Rotifer 133), total aluminum and MRP, and (2) pH is correlated positively with a different rotifer (Rotifer 126) and Diaphanosoma, but negatively with Bosmina and Rotifer 133. The results of this study indicate that the ANN approach is extremely useful for identifying indicator species for assessing impacts and recovery from perturbations, such as acidification.
Day: Wednesday, Feb. 3
Location: Sweeney Center