This study employs an integrated modeling approach to elucidate the mechanisms of nitrate ion transport through nanofiltration (NF) and reverse osmosis (RO) membranes. The investigation first applied models from irreversible thermodynamics, specifically the Kedem–Katchalsky and Spiegler–Kedem models, to describe convective/diffusive contributions and the impact of the initial nitrate concentration (50–150 mg/L) on phenomenological parameters (reflection coefficient σ, and solute permeability P_s). The results revealed a marked sensitivity of NF membranes to the initial nitrate concentration, in contrast to the stable performance of RO membranes. To deepen this analysis, Response Surface Methodology (RSM) was used as a robust statistical tool to systematically model and quantify the synergistic effects of the initial concentration and other key operational parameters, transmembrane pressure (TMP) and recovery rate (Y) on NF performance. The results highlight the complementarity between transport modelling and statistical approaches for analysing nitrate rejection and permeate flux. The proposed approach provides useful insight into NF membrane-specific behaviour and relative sensitivity to operating conditions, within the scope and limitations of the studied membrane and experimental configurations.

This study evaluates the performance of electrodialysis (ED) using the CMX/ACS membrane pair, nanofiltration (NF) with the NF90 or NF270 membranes, and reverse osmosis (RO) with the BW30 membrane for fluoride removal in groundwater from Ben Guerir. This region is characterized by a salinity of 1.2 g/L and a fluoride concentration of 2.32 ppm. The results indicate that ED achieved good water quality with a demineralization rate of 40% at a cost of 0.2275 AMD/m3. NF using the NF270 membrane produced water of acceptable quality for a cost of 0.1033 AMD/m3. In contrast, both RO with the BW30 membrane and NF with the NF90 membrane yielded water with a highly aggressive chemical profile, necessitating remineralization. Remineralization, achieved by adding filtered water, reduced the cost to 0.0650 AMD/m3 for the NF90 membrane and 0.0658 AMD/m3 for the BW30 membrane, whereas remineralization using lime resulted in costs of 0.126 AMD/m3 for NF90 and 0.1343 AMD/m3 for BW30. In conclusion, the findings suggest that the most efficient and cost-effective solution for fluoride removal in this context is either NF using the NF90 membrane or RO with the BW30 membrane, combined with remineralization by adding filtered water.