An artificial intelligence system that can detect extreme weather and landslides, and issue forecast warnings at an earlier stage may come into use after a decade as the Hong Kong University of Science and Technology has been granted about HK$91 million in funding.
The university said with the approval of the funding, the study, which will be led by a team of international professionals from multiple disciplines, will aim to push the forecast period of heavy rainfall to six hours earlier - currently the warning mechanism is three hours prior to the adverse weather.
Charles Ng Wang-wai, the university's associate vice-president for research and development, said with the help of the system, citizens can have more time to prepare for potential disasters.
"We would work closely with the Hong Kong Observatory to run a model to double the forecast lead time of heavy rainfall with more accurate locations and intensity," he said.
Ng said many cities, including the SAR, are located in mountainous regions but are not prepared for the potentially devastating impacts of unprecedented extreme rainstorms, which will occur more frequently due to climate change.
He cited a heavy rainstorm in 2008 that lashed Lantau Island to drive home his point.
"If it had affected Hong Kong Island instead, the damage to infrastructure and potential loss of lives would've been unimaginable," Ng said.
"By increasing the forecast lead time for heavy rainfall, our research findings will give decision makers, authorities, and citizens more time to prepare, thereby enhancing the resiliency of the local community."
The research idea originated from parents complaining about late announcements of heavy rain forecasts, and the Education Bureau not being able to act quickly enough in announcing class suspensions ahead of thunderstorms that could trigger red or black rainstorm warnings.
In 2012, the observatory launched a feature to announce location-based rain forecasts on its mobile application, so that users could get hold of the forecast of selected districts two hours prior. The research team will also work on a slope failure mechanism and develop an early landslide warning system.
"The system will apply deep learning on data obtained from stereoscopic sources, including remote sensing satellites," Ng said. "Another newly designed multiple flexible barrier system and other green mitigation measures will be developed to prevent landslides."
The project also aims to establish a Center for Slope Safety, which would create an immersive environment to educate citizens on slope safety by using virtual reality technology and narrative visualization.