Different modelling approaches for predicting titanium dioxide nanoparticles mobility in intact soil media

Area of Science:

• Environmental Science
• Soil Science
• Nanotechnology

Background:

• Understanding engineered nanoparticle (ENP) transport is vital for environmental risk assessment.
• Predicting ENP mobility in soils can reduce the need for extensive experiments.

Purpose of the Study:

• Investigate factors influencing titanium dioxide nanoparticle (nTiO2) mobility in soil.
• Evaluate machine learning models for predicting nTiO2 transport in soil.

Main Methods:

• Assessed nTiO2 mobility in real soil under various conditions.
• Compared Multiple Linear Regression (MLR), Classification and Regression Tree (CART), Random Forest (RF), and Artificial Neural Network (ANN) models.

Main Results:

• ANN and RF models demonstrated good predictive performance for nTiO2 mobility.
• ANN models were slightly superior to RF, showing fewer generalization errors.
• MLR and CART models exhibited poor performance, especially in validation datasets.

Conclusions:

• Soil properties, experimental conditions, soil fractures, and preferential flow significantly impact nTiO2 transport.
• ANN and RF are promising tools for predicting ENP mobility in soil environments.