Multivariate Statistical Assessment of Organic, Nutrient, and Microbial Removal in a Tropical AAO–Based Sewage Treatment Plant

This study investigated the spatiotemporal variability in influent wastewater quality at the Dasherkandi Sewage Treatment Plant, Dhaka’s largest advanced biological treatment facility. Using eight months of daily monitoring data, principal component analysis revealed three dominant pollution modes: organic loading driven by biochemical oxygen demand (BOD), chemical oxygen demand (COD), ammonia, and suspended solids; microbial contamination defined by fecal coliform bacteria; and a secondary gradient linked to residual nutrient variation. Together, these three components explain nearly 79% of the total variance, underscoring the complex interplay of physicochemical and microbial pressures. Partial least squares regression indicated that influent BOD5 and fecal coliform were the strongest predictors of BOD5 removal, with COD and NH3–N playing secondary roles; the modest cross-validated R2 underscores the challenge of predicting removal under tropical variability. K-means clustering distinguished three clear operating regimes: persistent high organic loads during the dry months of October to January, a marked shift to pathogen-dominated inflow during late dry and premonsoon months from February to April, and a return to stable baseline conditions in May. Three operating regimes were identified via k-means on z-scored influent features, aligning with seasonal mass-loading patterns. Average removal efficiencies were high—BOD5 97%, COD 95%, NH3–N 99%, TSS 98%, and P 71%. PLS indicated influent BOD5 and fecal coliform as primary drivers of BOD5 removal (VIP>1). These results demonstrate the value of multivariate diagnostics for resolving load regimes and linking influent signals to treatment performance.