Predicting water quality parameters in a subtropical estuary under different scenarios: Evidence from the Gold Coast, Australia.

Estuarine water quality is shaped by complex interactions between hydrodynamics, climate, and catchment inputs, making accurate prediction challenging. This study develops machine learning models to predict chlorophyll-a (chl a), total nitrogen (TN), total phosphorus (TP), and total suspended solids (TSS) in the Gold Coast Broadwater, a subtropical micro-tidal estuarine lagoon in Australia. Using a six-year dataset (2016-2021) from 18 monitoring sites, five algorithms were trained and tested, with Random Forest delivering the highest and most consistent predictive performance across all variables (chl a: R2 = 0.60, RMSE = 1.07 μg/L; TN: R2 = 0.72, RMSE = 0.043 mg/L; TP: R2 = 0.80, RMSE = 0.004 mg/L; TSS: R2 = 0.67, RMSE = 3.26 mg/L). Scenario-based simulations reveal that precipitation and temperature are more strongly associated with chl a dynamics than nutrient enrichment alone, while TN, TP, and TSS show pronounced sensitivity to spatial variability in rainfall. These results highlight the potential importance of climate-driven processes in shaping estuarine water quality responses under the conditions evaluated. The study provides a novel framework that integrates data-driven prediction with observation-based modelling, spatial mapping, and scenario analysis. Together, these components enable high-resolution environmental predictions and provide a practical tool for adaptive water-quality management in dynamic estuarine systems.