A comparative study of response surface methodology (RSM) and artificial neural networks (ANN) based optimization and modeling of paper mill effluent treatment using Lemna minor

Industrial effluents are a major source of water pollution, often leading to ecological imbalance and human health risks. Among these, paper mill effluent is particularly challenging to manage due to its high pollution load. This study aimed to evaluate the phytoremediation potential of Lemna minor for paper mill effluent and optimize operational conditions using Response Surface Methodology (RSM) and Artificial Neural Networks (ANN). For this, effluent samples were collected from a industry located in Saharanpur, India, and exposed to L. minor at varying biomass densities (10, 20, 30 g) in 20 L containers under controlled greenhouse conditions for 16 days. Three wastewater concentrations (0%, 50%, 100%) were tested. Both RSM and ANN models were developed to predict BOD and COD removal and biomass production, and statistical validation was performed using selected metrics. Results indicated severe pollution in the effluent, with maximum pollutant removal and biomass production occurring at moderate wastewater concentration (50–70%) and higher plant density (20–30 g). Specifically, optimized conditions of 69.03% wastewater and 29.89 g plant density yielded a model-predicted near-complete BOD removal (~100%; the polynomial model extrapolated to 100.81% and is therefore reported as ~100%), COD removal of 88.23%, and biomass of 102.95 g (predicted), closely relating to experimental values (BOD 96.48%, COD 85.21%, biomass 97.42 g). In this, ANN consistently outperformed RSM in predictive accuracy (R2 > 0.999, lower ME, and lower MSE and RMSE), effectively capturing nonlinear interactions among variables. The study showed that L. minor is an effective and sustainable candidate for industrial effluent treatment, contributing significantly to pollutant removal and biomass yield.