Limits...
Prediction and assessment of drought effects on surface water quality using artificial neural networks: case study of Zayandehrud River, Iran.

Safavi HR, Malek Ahmadi K - J Environ Health Sci Eng (2015)

Bottom Line: The models used in this research include multi-layer perceptron and radial basis function.Finally, these two models are compared in terms of their performance using the time series of electrical conductivity at eight monitoring-hydrometric stations during drought periods between the years 1997-2012.It is further shown that radial basis function works better for the upstream stretches of the river while multi-layer perceptron is more efficient for the downstream stretches.

View Article: PubMed Central - PubMed

Affiliation: Department of Civil Engineering, Isfahan University of Technology, Isfahan, Iran.

ABSTRACT
Although drought impacts on water quantity are widely recognized, the impacts on water quality are less known. The Zayandehrud River basin in the west-central part of Iran plateau witnessed an increased contamination during the recent droughts and low flows. The river has been receiving wastewater and effluents from the villages, a number of small and large industries, and irrigation drainage systems along its course. What makes the situation even worse is the drought period the river basin has been going through over the last decade. Therefore, a river quality management model is required to include the adverse effects of industrial development in the region and the destructive effects of droughts which affect the river's water quality and its surrounding environment. Developing such a model naturally presupposes investigations into pollution effects in terms of both quality and quantity to be used in such management tools as mathematical models to predict the water quality of the river and to prevent pollution escalation in the environment. The present study aims to investigate electrical conductivity of the Zayandehrud River as a water quality parameter and to evaluate the effect of this parameter under drought conditions. For this purpose, artificial neural networks are used as a modeling tool to derive the relationship between electrical conductivity and the hydrological parameters of the Zayandehrud River. The models used in this research include multi-layer perceptron and radial basis function. Finally, these two models are compared in terms of their performance using the time series of electrical conductivity at eight monitoring-hydrometric stations during drought periods between the years 1997-2012. Results show that artificial neural networks can be used for modeling the relationship between electrical conductivity and hydrological parameters under drought conditions. It is further shown that radial basis function works better for the upstream stretches of the river while multi-layer perceptron is more efficient for the downstream stretches.

No MeSH data available.


Related in: MedlinePlus

A four-layer MLP-NN for predicting EC
© Copyright Policy - OpenAccess
Related In: Results  -  Collection

License 1 - License 2
getmorefigures.php?uid=PMC4597443&req=5

Fig9: A four-layer MLP-NN for predicting EC

Mentions: In the MLP-NN model, two hidden layers were used while the numbers of neurons varied from five to fifty for each station (Fig. 9). Because there is no general rule for determining the properties of hidden layers and the neurons, trial-and-error procedures recommended by many researchers [2] were used to construct the hidden layer and the neurons. The number of hidden layer neurons significantly influences the performance of a network: if the number is small, the network may not achieve the acceptable level of accuracy, but if there are too many, training may be lengthy and the model may over-fit the data. Two loops were used to build the first and the second hidden layers and the efficient results obtained were stored at each point in time. These results were procured based on the assumption that the network should not run into over-fitting and that the error should have decreased by increasing number of neurons. MLP-NN was trained by the back-propagation rule and the Levenberg-Marquardt optimization of weights and bias values. The transfer function for the first hidden layer was tangent-sigmoid while it was log-sigmoid for the second. Based on the data sets, 70 % of the data sets were used for training and 30 % for both testing and validation of the network.Fig. 9


Prediction and assessment of drought effects on surface water quality using artificial neural networks: case study of Zayandehrud River, Iran.

Safavi HR, Malek Ahmadi K - J Environ Health Sci Eng (2015)

A four-layer MLP-NN for predicting EC
© Copyright Policy - OpenAccess
Related In: Results  -  Collection

License 1 - License 2
Show All Figures
getmorefigures.php?uid=PMC4597443&req=5

Fig9: A four-layer MLP-NN for predicting EC
Mentions: In the MLP-NN model, two hidden layers were used while the numbers of neurons varied from five to fifty for each station (Fig. 9). Because there is no general rule for determining the properties of hidden layers and the neurons, trial-and-error procedures recommended by many researchers [2] were used to construct the hidden layer and the neurons. The number of hidden layer neurons significantly influences the performance of a network: if the number is small, the network may not achieve the acceptable level of accuracy, but if there are too many, training may be lengthy and the model may over-fit the data. Two loops were used to build the first and the second hidden layers and the efficient results obtained were stored at each point in time. These results were procured based on the assumption that the network should not run into over-fitting and that the error should have decreased by increasing number of neurons. MLP-NN was trained by the back-propagation rule and the Levenberg-Marquardt optimization of weights and bias values. The transfer function for the first hidden layer was tangent-sigmoid while it was log-sigmoid for the second. Based on the data sets, 70 % of the data sets were used for training and 30 % for both testing and validation of the network.Fig. 9

Bottom Line: The models used in this research include multi-layer perceptron and radial basis function.Finally, these two models are compared in terms of their performance using the time series of electrical conductivity at eight monitoring-hydrometric stations during drought periods between the years 1997-2012.It is further shown that radial basis function works better for the upstream stretches of the river while multi-layer perceptron is more efficient for the downstream stretches.

View Article: PubMed Central - PubMed

Affiliation: Department of Civil Engineering, Isfahan University of Technology, Isfahan, Iran.

ABSTRACT
Although drought impacts on water quantity are widely recognized, the impacts on water quality are less known. The Zayandehrud River basin in the west-central part of Iran plateau witnessed an increased contamination during the recent droughts and low flows. The river has been receiving wastewater and effluents from the villages, a number of small and large industries, and irrigation drainage systems along its course. What makes the situation even worse is the drought period the river basin has been going through over the last decade. Therefore, a river quality management model is required to include the adverse effects of industrial development in the region and the destructive effects of droughts which affect the river's water quality and its surrounding environment. Developing such a model naturally presupposes investigations into pollution effects in terms of both quality and quantity to be used in such management tools as mathematical models to predict the water quality of the river and to prevent pollution escalation in the environment. The present study aims to investigate electrical conductivity of the Zayandehrud River as a water quality parameter and to evaluate the effect of this parameter under drought conditions. For this purpose, artificial neural networks are used as a modeling tool to derive the relationship between electrical conductivity and the hydrological parameters of the Zayandehrud River. The models used in this research include multi-layer perceptron and radial basis function. Finally, these two models are compared in terms of their performance using the time series of electrical conductivity at eight monitoring-hydrometric stations during drought periods between the years 1997-2012. Results show that artificial neural networks can be used for modeling the relationship between electrical conductivity and hydrological parameters under drought conditions. It is further shown that radial basis function works better for the upstream stretches of the river while multi-layer perceptron is more efficient for the downstream stretches.

No MeSH data available.


Related in: MedlinePlus