Human activities have induced recurrent hypoxic events in coastal oceans, posing a formidable challenge to marine ecosystems and fisheries. This study makes use of extensive long-term (1986-2022) monthly in situ observation data from the Hong Kong Environmental Protection Department, and reveals a distinctive east-high-west-low spatial gradient of dissolved oxygen (DO) concentrations in waters surrounding Hong Kong (HKw), progressively rising from the nearshore to offshore realms. Notably, dissolved oxygen (DO) concentrations reach their peak in winter. Subsequently, they decline and reach their lowest levels in summer, accompanied by localized hypoxia. Employing the Empirical Orthogonal Function (EOF) analysis, we discern the prevailing modes of DO variability. Intriguingly, our analysis identifies tipping points in the annual DO concentration fluctuations within HKw over the 1986-2022 period, marked by a discernible decreasing trend post-2000. Compared to the 1990-2000 decade, the frequency and severity of hypoxic events in HKw surged significantly during 2010-2020, likely attributable to the proliferation of population density. Consequently, anthropogenic pressures are poised to exacerbate hypoxia events in HKw, underscoring the urgency for predictive capabilities. To address this need, we harnessed the Informer model to devise a predictive framework for DO concentrations in HKw, enabling forecasts of DO levels up to three months ahead, thereby facilitating proactive identification of impending hypoxic events. Validation against in situ observations confirms the accuracy of our predictions. These findings not only deepen our comprehension of DO dynamics in HKw but also lay a foundation for refining future water quality predictions and devising effective management strategies, vital for safeguarding marine ecosystems.
 

Dissolved Oxygen,Deep Learning,Hong Kong