Turbidity-associated and warm-season manganese patterns at river and canal drinking-water intakes in Hai Phong City, Vietnam: Statistical evidence and same-day alert rules from routine monitoring data

Manganese (Mn) often has to be removed at treatment plants that draw water from lowland rivers and canals, but particle-bound Mn and dissolved Mn(II) require different treatments, and only total concentrations are reported by routine monitoring.It was examined whether records already kept by utilities can separate turbidity-associated from warm-season Mn patterns and support daily raw-water decisions.Routine data (2021-2024; 3,451 Mn and 2,022 iron determinations) from seven intakes on four sources in Hai Phong City, Vietnam, were analyzed with fixed-effects, generalized additive, and segmented regressions and gradient-boosted trees.Total Mn exceeded 0.1 mg L -1 in 47.6% of samples (10.7% in February, 87.5% in September).Mn and iron (Fe) rose almost equally with turbidity, leaving the Mn/Fe ratio, on average, unchanged (standardized β = +0.004,95% CI -0.017 to 0.023).The ratio increased instead with a 30-day water-temperature index (T 30 ) marking the warm season (β = +0.036,0.019 to 0.055) but not with temperature anomalies, so the pattern is seasonal rather than demonstrably temperature-driven.Above a transition zone near 25 °C, Mn rose 14.2% per °C of T 30 .Five routine variables reproduced Mn at withheld intakes, sources, and years (log-scale Nash-Sutcliffe efficiency 0.73-0.79,linear model).In leave-one-year-out tests, same-day alert rules combining a seasonal term with turbidity raised the critical success index by 0.08-0.09over a fixed calendar.As only routine variables are needed, likely exceedance of the 0.1 mg L -1 limit can be flagged in a spreadsheet on the sampling day, before the laboratory Mn result is reported.