Monitoring‐Network Composition and Analytical‐Method History Shape Temporal Inference of Fecal‐Coliform Exceedance in an Urban River–Lake Network, Taihu Basin, China

ABSTRACT

Changing station composition and analytical methods can complicate microbial water‐quality trends. We analyzed 2573 fecal‐coliform (FC) observations from 64 reconciled sections in the Wuxi river–lake network, Taihu Basin, between December 2021 and August 2026. Records were available for 53 of 57 calendar months; four monthly files were irretrievably lost following a system error. Overall, 853 observations (33.15%) exceeded the common Class III FC benchmark of 10,000 counts L ⁻¹ . A fixed panel of 21 sections contributed 1113 observations. Pooled and fixed‐panel Sen slopes were 0.503 and 0.344 percentage points per month; their paired difference was 0.158 (section‐bootstrap 95% interval 0.011–0.327). Binomial mixed models with calendar‐month effects and section random intercepts gave annual odds ratios of 1.59 and 1.51, respectively. Adding an indicator for the reported October 2023 change from the rapid paper‐strip method to the enzyme‐substrate method reduced these estimates to 1.26 (section‐clustered 95% interval 1.08–1.47) and 1.36 (1.02–1.81). Within the enzyme‐substrate period alone, annual odds ratios were 1.17 (0.96–1.43) and 1.35 (0.93–1.98). Thus, the full‐period association remained positive under seasonal and method‐period adjustment, but the shorter, single‐method record did not resolve a temporal association. Between‐section heterogeneity and sensitivity to the monitoring population support reporting station coverage and analytical‐method history alongside pooled trends. The method‐period contrast cannot separate analytical effects from environmental changes occurring at the same time.