Using machine learning to predict the efficiency of biochar in pesticide remediation

Abstract

Pesticides have remarkably contributed to protecting crop production and increase food production. Despite the improved food availability, the unavoidable ubiquity of pesticides in the aqueous media has significantly threatened human microbiomes and biodiversity. The use of biochar to remediate pesticides in soil water offers a sustainable waste management option for agriculture. The optimal conditions for efficient pesticide treatment via biochar are aqueous-matrix specific and differ amongst studies. Here, we use a literature database on biochar applications for aqueous environments contaminated with pesticides and employ ensemble machine learning models (i.e., CatBoost, LightGBM, and RF) to predict the adsorption behavior of pesticides. The results reveal that the textural properties of biochar, pesticide concentration, and dosage were the significant parameters affecting pesticide removal from water. The data-driven modeling intervention offers an empirical perspective toward the balanced design and optimized usage of biochar for capturing emerging micro-pollutants from water in agricultural systems.