| Abstract | FSM is essential for maintaining hygienic sanitation conditions in highly populated humanitarian settlements particularly within the Rohingya refugee settlements in Cox’s Bazar, Bangladesh. The performance of faecal sludge treatment plants (FSTPs) remains insufficiently evaluated, and data-driven prediction approaches are limited. This study evaluates four FSTP technologies, namely Anaerobic Baffled Reactor (ABR), Up Flow Filter (UFF), Decentralized Wastewater Treatment System (DEWATS), and Biological Process (BP). It also analyzes parameter interactions and develops machine learning (ML) models to predict effluent quality. A 61-day monitoring campaign collected paired influent and effluent samples. Parameters included BOD, COD, TSS, NO3 − , PO4 3−, pH, temperature & Escherichia coli (E. coli). Statistical analyses included Analysis of Variance (ANOVA) and Pearson correlation. ML Models included MLR, RFR, SVR and XGB. All systems showed significant reductions (p < 0.001) in TSS, BOD, COD, NO3 − , PO4 3−, and E. coli., however pH and temperature remained stable (p > 0.05). UFF achieved the highest removal of BOD (89.39%), TSS (74.68%), NO3 − (72.99%), and E. coli (85.29%). DEWATS showed the highest removal of COD (83.05%) and PO4 3− (76.59%). ABR performed strongly for BOD (88.73%) and E. coli (84.37%). BP showed comparatively high COD removal (81.06%). Strong correlations existed among BOD, COD, and TSS (r > 0.75). RFR demonstrated the best predictive accuracy, with an R² value reaching 0.98 and XGB showed overfitting signs. This study provides one of the first continuous, high-resolution datasets for FSTP performance in the Rohingya camps and highlights the potential of ML algorithms for predicting BOD, COD, and E. coli using faster physicochemical parameters, reducing the need for their time-consuming direct measurement while maintaining reliable performance assessment and helps sanitation management, offering insights for optimizing design, improving operation, and supporting data-driven decisions. |
