Modelling water quality parameters of Lower Usuma Dam Reservoir, Abuja, using artificial neural network.

Water quality parameters of Lower Usuma Dam, Abuja was analyzed and modeled using Artificial Neural Network (ANN). Monthly water quality parameters of pH, turbidity, and electrical conductivity, total dissolved solids and total hardness for a duration of 6 years (2017-2022) was obtained from the Water Laboratory Department, of Federal Capital Territory Nigeria (F.C.T) Water Board, Abuja. Microsoft excel was used to analyze the trends variation of this parameters.  Artificial Neural Network (ANN) was used to develop three model equations for the prediction of electrical conductivity, total dissolved solids and total hardness respectively, with pH and turbidity as input parameters. F-test and t-test were used to validate each model using Microsoft excel. The error analysis and performance evaluation of the applied models were also done to evaluate the goodness/suitability of each of the models. The coefficients of determination (R2) between the actual parameters were 0.89085, 0.83156, and 0.86931 for testing, training and validation respectively.  A very strong relationship between the predictors (pH and turbidity) and the response variables (Electrical conductivity, total dissolved solids and total hardness) was established. The Root Mean Square Error were 11.2, 13.8 and 5.54. Thus, the total hardness model is the best among them because it has the lowest predictive error. The model validation carried out through the F-test and t-test for each of electrical conductivity, total dissolved solids, and total hardness, respectively, shows that F critical is greater than F, as well as t critical is greater than t-stat. This further shows that the ANN model is fit for prediction of water quality parameters. Water quality parameters of Lower Usuma Dam, Abuja was analyzed and modeled using Artificial Neural Network (ANN). Monthly water quality parameters of pH, turbidity, and electrical conductivity, total dissolved solids and total hardness for a duration of 6 years (2017-2022) was obtained from the Water Laboratory Department, of Federal Capital Territory Nigeria (F.C.T) Water Board, Abuja. Microsoft excel was used to analyze the trends variation of this parameters.  Artificial Neural Network (ANN) was used to develop three model equations for the prediction of electrical conductivity, total dissolved solids and total hardness respectively, with pH and turbidity as input parameters. F-test and t-test were used to validate each model using Microsoft excel. The error analysis and performance evaluation of the applied models were also done to evaluate the goodness/suitability of each of the models. The coefficients of determination (R2) between the actual parameters were 0.89085, 0.83156, and 0.86931 for testing, training and validation respectively.  A very strong relationship between the predictors (pH and turbidity) and the response variables (Electrical conductivity, total dissolved solids and total hardness) was established. The Root Mean Square Error were 11.2, 13.8 and 5.54. Thus, the total hardness model is the best among them because it has the lowest predictive error. The model validation carried out through the F-test and t-test for each of electrical conductivity, total dissolved solids, and total hardness, respectively, shows that F critical is greater than F, as well as t critical is greater than t-stat. This further shows that the ANN model is fit for prediction of water quality parameters.