With climate change, Hong Kong looks to artificial intelligence

By Medha Basu

As climate change brings more erratic weather, the city is looking to AI to predict disasters faster and more accurately.

The hills around Hong Kong offer great hiking trails - a chance for locals to get away from the hustle of the city and enjoy wonderful views of the skyline.

But the hills also pose a peril to the people working and living around them. The hillsides have been cut into and refilled to create more space in the dense city, changing the natural topography and increasing the risk of landslides. A series of landslides in 1972 killed about 150 people and destroyed several buildings on Hong Kong island.

To reduce the impact of such disasters, the government set up the Geotechnical Engineering Office to predict landslides on both natural and manmade slopes. Ivan Ho, senior geotechnical officer in the team, tells GovInsider how the territory is preparing for more extreme weather in the future, and plans to use AI to make predictions faster.

Climate change

Climate change is bringing more extreme weather patterns, with heavy rainfall becoming more frequent. This increases the risk of landslides, and Ho’s team is looking for ways predict them faster and more accurate.

One of the most time-consuming efforts in this is identifying and mapping the hills across Hong Kong. This is done manually, with geologists poring over aerial images of the territory and using their own expertise to identify landslide risk.

The office plans to use artificial intelligence to assist the geologists. AI could do a first screening of the images to identify areas with potential landslide risks, and shortlist areas that engineers and geologists should pay closer attention to, Ho says. “We are now planning to develop AI and machine learning to see whether we can identify the landslides faster by use of the machines,” he says.

Ho’s team is also looking to increase the accuracy of its predictions, which are calculated based on previous landslides caused by rainfall. “We are exploring whether, in the future, this kind of work can be done by AI,” Ho says. Already, such mathematical calculations are largely automated in industries like finance, where algorithms predict and place stock trades.

The office is growing its repository of data on past landslides, which are crucial to ‘train’ such machine learning algorithms. Engineers in the field use an app to make new data on landslides available immediately to decision-makers at headquarters.

“We try to use IoT and cloud technology to make everything real-time, faster and make more data available to people,” Ho says.

Giving early warnings

Unfortunately, Hong Kongers have little choice in whether or not to live near the hills, despite the danger: “We have too many people and too little flat land, [so] we have to live very close to the hillside,” Ho says. Boulders and debris from landslides can cause casualties, fatalities and extensive damage to property.

Man-made slopes are reinforced with steel bars to prevent landslides, but the same cannot be done for the hills covering 60% of the territory. Landslides along the hill slopes are a natural phenomenon and cannot be prevented altogether, Ho says.

Instead, the city has tried to minimise casualties and damage. It has built a massive geospatial repository of 60,000 slopes, and uses this information along with rainfall data to give residents an early warning on landslides. These can be predicted up to three hours in advance, giving residents time to evacuate. “One of our chief tasks is that we set up an early warning system to warn people to keep away from slopes when there is heavy rainfall,” Ho says.

Preparing for the worst

Even with warnings and predictions in place, the city must prepare for the worst. For instance, it plans to build new communication networks that will allow officials to receive disaster data even if the entire city blacks out.

Ho’s team currently relies on the city’s 3G network to get real-time data from rain gauges, allowing it to forecast rain patterns. This information is crucial to predicting landslides as “most of the serious landslides in Hong Kong happen during very heavy rainfall”, Ho says.

However, public telecommunications infrastructure come under a great deal of pressure during natural disasters like typhoons, and officials risk being left without access to information when it’s most needed. In extreme cases, power stations could also flood, Ho says, leaving the entire city in the dark.

His team is exploring a dedicated network to get data from the rainfall sensors, which would be completely separate from ones used by residents. It would be powered by solar energy with a backup source, allowing officials to continue working in the midst of a power shutdown.

In areas particularly prone to landslides, the office also predicts which way the rocks, mud and trees would fall, allowing it to plan how to build barriers to stop or deflect the debris. Hong Kong is now piloting sensors which will alert officials if these barriers are damaged or weakened, giving engineers time to repair or reinforce them. “We can keep an eye on the barriers to see whether the debris is going to build up further,” Ho adds.

Climate change is our new reality, making natural disasters more unpredictable with each passing year. Hong Kong is doing everything it can to minimise the impact - and save lives in the process.