Development of sediment load estimation models by using artificial neural networking techniques

• Published in: Environmental monitoring and assessment Vol. 187; no. 11; p. 686
• Main Authors: Hassan, Muhammad, Ali Shamim, M., Sikandar, Ali, Mehmood, Imran, Ahmed, Imtiaz, Ashiq, Syed Zishan, Khitab, Anwar
• Format: Journal Article
• Language: English
• Published: Cham Springer International Publishing 01.11.2015; Springer Nature B.V
• Subjects: Algorithms; Artificial intelligence; Atmospheric Protection/Air Quality Control/Air Pollution; Canals; Civil engineering; Earth and Environmental Science; Ecology; Ecotoxicology; Efficiency; Environment; Environmental Management; Environmental monitoring; Environmental Monitoring - methods; Environmental Pollutants - analysis; Environmental Pollutants - chemistry; Geologic Sediments - chemistry; Irrigation; Linear Models; Load; Mathematical models; methodology; Models, Theoretical; Monitoring/Environmental Analysis; Neural networks; Neural Networks (Computer); Pakistan; pollution load; Sediment load; sediment yield; Sedimentation & deposition; Sediments; Soil erosion; Statistical models; Studies; temperature; Water resources; Watersheds; Pakistan; Dead storage; Physical parameters; Gamma test; Sedimentation; Artificial neural networks
• ISSN: 0167-6369; 1573-2959; 1573-2959
• DOI: 10.1007/s10661-015-4866-y

This study aims at the development of an artificial neural network-based model for the estimation of weekly sediment load at a catchment located in northern part of Pakistan. The adopted methodology has been based upon antecedent sediment conditions, discharge, and temperature information. Model input and data length selection was carried out using a novel mathematical tool, Gamma test. Model training was carried out by using three popular algorithms namely Broyden-Fletcher-Goldfarb-Shanno (BFGS), back propagation (BP), and local linear regression (LLR) using forward selection of input variables. Evaluation of the best model was carried out on the basis of basic statistical parameters namely R -square, root mean squared error (RMSE), and mean biased error (MBE). Results indicated that BFGS-based ANN model outperformed all other models with significantly low values of RMSE and MBE. A strong correlation was also found between the observed and estimated sediment load values for the same model as the value of Nash-Sutcliffe model efficiency coefficient ( R -square) was found to be quite high as well.