Assessment of dissolved oxygen in Shatt Al-Arab River by other quality parameters of water using Artificial Neural Networks

Abstract

The Major sources of water are surface and subsurface. Surface waterincludes Rivers, Reservoirs, Creek, Streams, etc. This paper deals withusing a neural network model to recognize dissolved oxygen in Shatt Al-Arab. Within the present study, Shatt Al-Arab River (Basrah-Iraq) isconsidered as the study area with monthly observed data from 2009-2014.Artificial Neural Network (ANN) has been applied to pattern the relationsamong eight (8) water quality parameters which are devoted for predictingone parameter (1) so that to decrease the load of long experimentalprocedure. Physical and chemical parameters that are inserted in the modelare: pH, total dissolved solids, electrical conductivity, sulphate, phosphate,calcium, magnesium and nitrate. Dissolved oxygen (DO) is included in theoutput models. The three layered feed-forward model with back-propagationmulti-layer perception (MLP) models architecture of 8-8-1 for DO. Theartificial neural network has got training successfully and has been testedwith 70% and 30% of the data groups. Statistical criteria of correlationcoefficient (R2) and mean square error (MSE) are used to evaluateperformance of the models. The correlation coefficients of the artificialneural network model for predicting DO have been 0.99354 and 0.98237,and mean square error for the model are 0.007698 and 0.00122 respectively.It can be concluding that these techniques provide similar accuracy inestimating DO concentration and predicting the dissolved oxygen (DO) inShatt Al-Arab