Algal Sentinel: a novel web application for early warning and forecasting of Ostreopsis cf. ovata blooms in the Italian Southern Adriatic Sea

The toxic benthic dinoflagellate Ostreopsis cf. ovata causes harmful algal blooms (HABs) in the Apulia region of southern Italy. These HABs pose real threat to human health. Therefore, it was crucial to develop a model capable of predicting HABs before they occur, rather than afterwards as in conventional monitoring programmes. To this end, we applied a Quantile Regression Forest (QRF) non-parametric regression method, known for its ability to provide accurate estimation of conditional quantities. Then, we used a machine learning approach, training this QRF model with data collected by the Regional Agency for Environmental Protection of Apulia. These data are O. ovata concentrations measured from 2010 to 2024 in four towns of Apulia, namely Molfetta, Bisceglie, Giovinazzo and Bari. Only a few meteorological parameters (seawater temperature, air temperature, and dew point) proved useful for predicting blooms. The values of these three meteorological parameters in the days preceding the HABs allowed the model to predict O. ovata concentrations with an average accuracy of 90%. During summer 2025, Algal sentinel, a dynamic and interactive web interface, provided real-time visualization of bloom risk guaranteeing citizens the safe access to the coast. Noteworthy, Algal Sentinel is the first application worldwide capable of forecasting O. ovata blooms and, since it is based on weather data stored in open access archives, it can be scaled up at national and international levels.