Monitoring-network change reshapes long-term Water Quality Index inference: A dual-frame analysis from Ghana

Abstract Purpose. Long-term water-quality records are widely used to judge whether rivers and other water bodies are improving or deteriorating, but interpretation becomes difficult when the monitoring network itself changes through time. Using Ghana’s national Water Quality Index (WQI) record, we investigate two sources of apparent long-term change: variation over time at consistently monitored locations and changes in the set of stations contributing to the national record. Methods. We analysed 41 surface-water stations across 20 reported years from 2005 to 2025 using a dual-frame design that compares the changing national network with a fixed panel of 19 long-running stations. The analysis combined robust station-level trends, multiplicity control, persistence screening, endpoint sensitivity, and leave-one-station-out analysis. Results. After 2010, the full-network median exceeded the fixed-panel median in 13 of 16 reported years, with a median difference of + 1.55 WQI units. The fixed panel itself also showed a broad low-score period from 2012 to 2018, with the strongest negative departure in 2013. Station-level trends were heterogeneous, none remained significant after multiplicity correction, and four stations met the primary persistent-poor criterion. Conclusions. Monitoring-network change and temporal variation within a fixed station panel can coexist. We therefore propose Dual-Frame Monitoring Inference (DFMI), which reports contemporary network conditions alongside a stable sentinel panel and station-specific evidence. DFMI offers environmental agencies a practical way to make long-term monitoring records more comparable, transparent, and useful for management as observation systems evolve.