ABSTRACT Title of Document: USE OF MACROINVERTEBRATE PREDICTIVE MODELS TO EVALUATE THE STREAM RESTORATION EFFECT

Multivariate analysis was used to build macroinvertebrate predictive models for stream assessment in Britain, Australia, and the west coast of the United States. The philosophy behind these predictive models was similar, but variations exist and have been adapted for different regions. The macroinvertebrate predictive model in Maryland has been improved using Region-style models, including the Assessment by Nearest Neighbour Analysis (ANNA), the Burn’s Region of Influence (BROI), and the New Datum Region of Influence (NROI) predictive schemes. For better prediction precision, different parameter selection methods (stepwise AIC, exhaustive AIC, and exhaustive BIC) and rational multiple regression function checking have been used to prevent overfitting. Root mean squared error (RMSE) was used to select the final best model. The calibration results from the Region-Style models are better than those from previously built River InVertebrate Prediction And Classification System (RIVPACS)-style model. The different parameter selection criteria along with rational regression function checking discourage overfitting and improve the prediction results. Region–style methods can be alternative methods for buildin