Water Quality Prediction of Gangapur Reservoir (India) Using LS‐SVM and Genetic Programming. Issue 4 (December 2015)

• Record Type: Journal Article
• Authors: Jadhav, Mrunalini Shivaji; Khare, Kanchan Chandrashekhar; Warke, Arundhati Suresh
• Is Part Of: Lakes & reservoirs. Volume 20:Issue 4(2015)
• Page Start: 275
• Page End: 284
• Publication Date: 2015-12
• Subjects: genetic programming -- least square support vector machines (LS‐SVM) -- performance indicators -- water quality parameters; Lakes -- Periodicals; Reservoirs -- Periodicals; Gestion de l'environnement; Gestion écologique; Lac; Lac de barrage; Périodique électronique (Descripteur de forme); Ressource Internet (Descripteur de forme); 333.9163
• DOI: 10.1111/lre.12113
• Languages: English
• ISSNs: 1320-5331

Abstract

Abstract: Water quality analysis involves analysis of physio‐chemical, biological and microbiological parameters that reflect the abiotic and biotic status of ecosystems. This assessment facilitates planning for the utilization, antipollution and conservation strategies for sustainable use of aquatic ecosystem. Many mathematical models are available for predicting water quality. They have complex structures and require detailed information about sources and receptors, which are difficult and non‐economical. Difficulties in applying mathematical models promote the application of alternative approaches for data‐driven techniques for analysis of the results. The present study focuses on water quality predictions for the Gangapur Reservoir for a 30 days in advance scenario, using genetic programming (GP) and least square support vector machines (LS‐SVMs). A data period of 11 years (2000–2011) of Gangapur Reservoir temporal water quality was evaluated. The data were taken from a single sampling point representing climatological, hydrological and surface water quality measurements. One of the most important steps in application of data‐driven technique is selection of significant input parameters. Genetic programming equations were used for selecting significant input parameters. These significant input parameters are used for 30 days advance predictions of faecal coliform. A performance analysis of GP and LS‐SVM models was carried out with the help of coefficient of determination, root‐mean‐square error and correlation coefficient. In the absence of availability of data, a typical situation for Indian case studies, the model runs were conducted with the use of available parameters. The developed models, along with their performance indicators, also are discussed.