An Alternative Approach for Assessing Sediment Impact on Aquatic Ecosystems Using Single Decision Tree (SDT)

Elfimov Valery Ivanovich, Khakzad Hamid





The purpose of this study was to identify factors affecting scale of the severity (SEV) of ill effect for fishes and to evaluate the derived prediction model. This study is based on 303 data about aquatic ecosystem quality over a wide range of sediment concentrations (1-500,000 mg SS/L) and durations of exposure (1-35,000 h). The independent variables were concentration of suspended sediment, species, life stage and duration of exposure. A single decision tree (SDT) analysis was done to identify factors for predicting a model of SEV and the CART algorithm was employed for building and evaluating regression trees. Results show that a single decision tree is better than traditional regression models with higher recognition rate and forecast accuracy and strong practical value.



Keywords: Scale of the severity; concentration of suspended sediment; species; life stage; duration of exposure; single decision tree