Title: Ecological modelling using artificial neural network for macroinvertebrate prediction in a tropical rainforest river
Authors: Joyce Osarogie Odigie; John Ovie Olomukoro
Addresses: Department of Animal and Environmental Biology, University of Benin, Nigeria ' Department of Animal and Environmental Biology, University of Benin, Nigeria
Abstract: In this study, artificial neural network was applied to predict the benthic macroinvertebrates fauna of Obueniyomo River using 75% of the dataset for model testing and 25% for training, scaled between 0 and 1 and implemented using R statistical. Thirty-nine predictors (physical and chemical variables) served as the inputs from which 25 output parameters successfully predicted the presence or absence of macroinvertebrates fauna in the study stations in the visualised neural network model. Sensitivity analysis, an essential test was applied to ascertain the influence of the output parameters in the prediction of the macroinvertebrates fauna and to outline which variables significantly determined the model output. The model showed that depth, flow rate, transparency and pH, appeared uniformly segregated than other selected input parameters, which served as a good predictor. We conclude that ANN constitutes a practical model for predicting macroinvertebrates fauna of freshwater ecosystems under alliance with some environmental conditions.
Keywords: artificial neural network; benthic macroinvertebrates; ecological modelling; freshwater ecosystem; sensitivity analysis.
International Journal of Environment and Waste Management, 2020 Vol.26 No.3, pp.325 - 348
Received: 16 Sep 2018
Accepted: 04 Jun 2019
Published online: 01 Sep 2020 *