Abstract
The transport characteristics of bulk solutions of seven heavy metal compounds (NaAsO2, Cd(NO3)2, Pb(NO3)2, Ni(NO3)2, ZnCl2, CuSO4 and Co(NO3)2), two pesticides (Cartap and Carbendazim) and a simple salt (CaCl2) through eight agricultural soils of seven agro-ecological zones (AEZ) of Bangladesh were investigated. The transport experiments for these chemicals (hereafter called solutes) were conducted in repacked soil columns in laboratory under unsaturated steady-state conditions. The relevant physico-chemical properties of the soils were measured. Breakthrough curves (BTCs) of the solutes were measured by time-domain reflectometry (TDR) and analyzed by a transfer-function method to determine the solute-transport parameters. Four of the soil columns were then conditioned by applying municipal wastewater of Mymensingh town for several days in an alternate wetting and drying manner. The BTCs of the solutes (except for Co(NO3)2 and CuSO4) were measured again through these soil columns; the experimental conditions were kept unchanged. Pedo-transfer functions that correlated solute-transport parameters and soil properties were developed through multiple linear regression (MLR) and artificial neural network (ANN) models. The transport velocity through the soils was the highest (0.938 cm h-1) for CaCl2 and lowest (0.411 cm h-1) for NaAsO2. According to ranked dispersivities and associated soil textural classes, Co(NO3)2 had the maximum dispersivity (0.347 cm), obtained for Satkhira silt loam and Carbendazim had the minimum value (0.137 cm), obtained for Mymensingh loamy sand. The dispersion coefficient of the solutes increased with depth; the increasing rate over 5 - 25 cm soil profile was the highest (136.46 to 281.21%) for CaCl2 and lowest (6.05 to 26.11%) for Co(NO3)2. The increasing rate of dispersivity, on the other hand, decreased with increasing travel depth. Based on retardation factor, the accumulation rate of the solutes in the soils, in order of preference, was: As > Co > Cd > Carbendazim > Pb > Cu > Cartap > Zn > Ni > Ca. The solute-transport parameters were highly dependent on clay content of the soils and less dependent on soil organic carbon. The transport behaviors of the two pesticides were similar to that of the heavy metals. Municipal wastewater significantly altered the soil properties and hence the transport parameters of the solutes through the soil. The bulk density and clay content were the most sensitive input parameters for both the MLR and ANN models. The average root mean square error (RMSE) of MLR model was 0.096 and that of the ANN model was 0.015. The ANN model provides a way of significantly enhancing pollution transport prediction through soil.