Excessive exposure to ultraviolet (UV) radiation can lead to detrimental health effects such as sunburn, wrinkling, increased risk of skin cancers, and cataracts. Therefore, accurate future predictions of the Ultraviolet Index (UVI) are critical to raise community awareness of UV exposure and mitigate the associated health risks. The Hong Kong Observatory (HKO) has been providing a day-ahead maximum UVI forecasting service since 2006, based on empirical clear-sky UVI multiplied by a predicted weatherdependent modification factor. To enhance this service, this project aims to establish novel forecasting methods that harness multimodal data and advanced deep learning algorithms. The project not only seeks to enhance the accuracy of the current day-ahead maximum UVI prediction but also strives to forecast 1-hour averaged UVI time series for the subsequent day across the Hong Kong Region with a spatial resolution of 2 km. This shift from single-location, single-value forecasts to high-resolution spatiotemporal predictions will empower the public by providing critical information on when and where to safeguard themselves against UV exposure. The technology being developed within this project has significant potential for real-time deployment, paving the way for a more comprehensive and effective UVI forecasting service for the Hong Kong community.
Development of High-fidelity and High-Spatiotemporal-Resolution Day-ahead Ultraviolet Index (UVI) Forecasting Method -- a Deep Learning Approach
