Real-time data driven forecast system for coastal algal blooms

We present a new modeling system for prognostic daily forecasting of algal bloom (Chlorophyll-a > 10 mg/m3) risks in weakly flushed coastal bays. According to a vertical stability theory (Wong et al. 2009), a necessary condition for an algal bloom is determined by the ratio of a bulk vertical turbulent diffusivity (E) relative to a critical value (Ec) dependent on the algal growth rate and photic depth. Based on 3D hydrodynamic modeling and data driven models, the turbulent diffusivity E for a given semi-enclosed water on any given day can be estimated from the predicted tidal range and hydro-meteorological data - and accounting for effects of density stratification. The critical diffusivity Ec can be determined from water temperature and light extinction data. This bloom forecast framework has been validated against extensive biweekly and monthly water quality data (1986 to 2018). Using high-frequency data (10 min) on salinity and temperature at various depths, together with readily available hydro-meteorological data (solar radiation, air temperature, wind speed and rainfall) the vertical density gradient on the next day can be predicted using an artificial neural network (ANN) model - and hence the water column stability risk (E/Ec). Combining the stability risk with a nutrient availability factor estimated from water quality monitoring, the algal bloom risk on the next day can be predicted. The forecast system has been validated against 4 years of high frequency data for the Yim Tin Tsai marine fish culture zone in Tolo Harbour, Hong Kong.