Measurement and mathematical modelling of nutrient level and water quality parameters M. Kasheﬁ Alasl, M. Khosravi, M. Hosseini, G. R. Pazuki and R. Nezakati Esmail Zadeh ABSTRACT Physico-chemical water quality parameters and nutrient levels such as water temperature, turbidity, saturated oxygen, dissolved oxygen, pH, chlorophyll-a, salinity, conductivity, total nitrogen and total phosphorus, were measured from April to September 2011 in the Karaj dam area, Iran. Total nitrogen in water was modelled using an artiﬁcial neural network system. In the proposed system, water temperature, depth, saturated oxygen, dissolved oxygen, pH, chlorophyll-a, salinity, turbidity and conductivity were considered as input data, and the total nitrogen in water was considered as output. The weights and biases for various systems were obtained by the quick propagation, batch back propagation, incremental back propagation, genetic and Levenberg–Marquardt algorithms. The proposed system uses 144 experimental data points; 70% of the experimental data are randomly selected for training the network and 30% of the data are used for testing. The best network topology was obtained as (9-5-1) using the quick propagation method with tangent transform function. The average absolute deviation percentages (AAD%) are 2.329 and 2.301 for training and testing processes, respectively. It is emphasized that the results of the artiﬁcial neural network (ANN) model are compatible with the experimental data. M. Kasheﬁ Alasl R. Nezakati Esmail Zadeh Department of Environment, North Tehran Branch, Islamic Azad University, Tehran, Iran M. Khosravi M. Hosseini (corresponding author) Department of Chemistry, North Tehran Branch, Islamic Azad University, Tehran, Iran E-mail: mhosseini87@gmail.com G. R. Pazuki Department of Chemical Engineering, Amirkabir University of Technology, Tehran, Iran Key words | artiﬁcial neural network, different depths, mathematical modelling, nutrient level, water quality INTRODUCTION Several papers present the results of the determination of water quality parameters in different seasons, which is a major problem in environmental engineering. During the past 10 years, new agricultural policies and environmental regulations have been developed in several European countries in order to improve water quality (Raed et al. ). Other studies have conﬁrmed that agriculture is fre- quently the main source of diffuse nitrogen and phosphorus pollution in water bodies in Western Europe (European Environment Agency ). Various anions are considered as the most important parameters to character- ize water quality, because the concentrations of the nutrients nitrogen and phosphorous are included among them (Reiche ). An important scientiﬁc and environ- mental task is quantifying nutrient retention in natural and artiﬁcial treatment of water quality (Scholz ). The bio- logical importance of nitrogen compounds in living organisms has been well recognized. For instance, fertilizers are the major cause of nitrate pollution in different sources of water. High levels of nitrate are toxic to humans. The determination of total nitrogen is important in the character- ization of potentially polluted waters. There are different analytical methods for nitrate analysis, such as UV detection. The failure of water mains has been analyzed using mainly statistical models; these models depend on data, col- lected over several years, which is used for predicting water pipe failure (Day et al. ). Many statistical models use his- torical data for predicting changes of different parameters in water (Brouwer et al. ). Day et al.() used linear stat- istical models. The main weakness of the statistical models is a wide variation in performance of individual assets (Eisenbeis et al. ). Among different models, artiﬁcial neural networks (ANNs) have been popular in modelling non-linear data including river ﬂow, surface ﬂow and ground water forecasting (Kleiner & Rajani ). The 1962 © IWA Publishing 2012 Water Science & Technology | 66.9 | 2012 doi: 10.2166/wst.2012.333