Artificial neural network model to predict transport parameters of reactive solutes from basic soil properties

Area of Science:

• Environmental Science
• Soil Science
• Computational Science

Background:

• Direct measurement of solute transport parameters in diverse soils is impractical due to time, cost, and labor.
• Pedo-transfer functions offer an indirect estimation approach for these parameters.
• Accurate prediction of chemical transport in soils is crucial for environmental management.

Purpose of the Study:

• To develop and evaluate an Artificial Neural Network (ANN) model for estimating solute transport velocity (V), dispersion coefficient (D), and retardation factor (R).
• To predict these parameters using basic soil properties as input.
• To assess the model's accuracy and efficiency in estimating transport parameters for various chemicals and soil types.

Main Methods:

• Developed an ANN model using breakthrough data from 14 agricultural soils in Bangladesh.
• Measured solute breakthrough curves using time-domain reflectometry (TDR) under unsaturated steady-state flow.
• Determined soil properties including bulk density, organic carbon, clay content, pH, median grain diameter, and uniformity coefficient.

Main Results:

• The ANN model reliably predicted V, D, and R with high accuracy (RRMSE 0.028-0.363, EF >0.99).
• Soil clay content and bulk density were identified as the most influential input variables for the ANN model.
• The model demonstrated low errors (ME -0.00004 to 0.0005, BOE% 0-0.003).

Conclusions:

• The developed ANN model provides a cost-effective and efficient method for estimating solute transport parameters.
• This approach significantly enhances the prediction of pollution transport through soils.
• The study highlights the potential of ANN models in soil science and environmental risk assessment.