Using the Combined Model of Gamma Test and Neuro-Fuzzy System for Modeling and Estimating Lead Bonds in Reservoir Sediments Publisher Pubmed



Mohammadi AA¹ ; Yousefi M² ; Soltani J³ ; Ahangar AG⁴ ; Javan S¹
Source: Environmental Science and Pollution Research Published:2018

Abstract

Heavy metals attract a great deal of attention nowadays due to their potential accumulation in living creatures and transference in the food chain. Sediments of water reservoirs are considered to be a source of accumulation of these metals that develop in response to human activities and soil erosion. This study collected 180 samples of the surface sediments of water reservoir 1 at Chahnimeh in Sistan. Efficiency of the ANFIS model was evaluated to estimate the five bonds following the measurement of parameters in the laboratory. The following results were obtained for the parameters: organic carbon (OC) %, 0.31; cation exchange capacity (CEC), 37.07 Cmol kg; total Pb, 25.19 mg/kg; clay %, 45.87; and silt %, 39.02. These parameters were used as input for the training model. In the output layer, lead bonds were chosen as modeling targets in the following way: Pb f1 (4.61); Pb f2 (0.54); Pb f3 (16.28); Pb f4 (3.42); and Pb f5 (0.38) mg/kg. The best input compound in this model was chosen using the gamma test. From a total of 180, 88 data were considered for the model training section. Eventually, the neural-fuzzy model (subtractive clustering), developed for the prediction of lead bonds in the studied region, was able to account for over 99% of lead bonds in the sediments; considering statistical criteria of root mean squares error or RMSE (0.0337–0.0813) and determination coefficient or R2 (0.92–0.99), this model showed good performance with regard to prediction. © 2018, Springer-Verlag GmbH Germany, part of Springer Nature.