2.00 PM - 3.30 PM
"Visibility and low ceiling (lowest layer of clouds) are often a matter of concern during aircraft take-off and landing operations because they obstruct the pilot's view and have other vital safety and economic functions of air transportation. Therefore, accurate forecasting of both properties is essential for the safe operation of the aviation industry and travel safety of passengers and crew. Despite the availability of physical models elsewhere, there is a dearth of any such studies for the Fiji Islands, making this proposal and study region of paramount importance.
This Master of Research proposal considers the development of predictive models based on artificial intelligence methods, particularly for aviation stakeholders. It aims to develop short-term (hourly) forecasts models for Visibility and low ceiling using key atmospheric datasets. To pursue this, several innovative and data-driven deep learning (DL) models, which are state-of-the-art in the atmospheric modelling field, will be designed for short-term visibility and ceiling forecasting.
We propose the N-BEATS model, which will be used for univariate forecasting of the targeted parameters promising state-of-the-art results using a pure deep learning architecture. Google's TabNet will be used as a built-in DL architecture for targeted predictions using multiple meteorological variables. The outcomes of this research project will enable a better understanding of visibility and ceiling parameters for the study regions and provide an additional approach for making reliable and consistent forecasts of meteorological parameters useful for the aviation sector. It will not only benefit airline operators, flight planners, airport operators, and air traffic controllers but importantly pilots for the safe and efficient conduct of flights."
For more information, please email the Graduate Research School. For the zoom link, please refer to the latest GRS Bulletin.