Simulating Disaster with a Digital Twin: Kumamoto City
Japan has experienced a typhoon annually in the last ten years, compared with the world’s ten-year average of only two. In addition, the country is also highly exposed to flooding due to heavy rainfall and earthquakes. The combination of climate-related impacts can be paralyzing for a nation that is regularly recovering from climate events.
In 2016, two large earthquakes struck Kumamoto Prefecture in southern Japan, with the second earthquake occurring only 28 hours after the first. The 7+ magnitude earthquakes collapsed more than 8,000 buildings and significantly damaged an additional 35,000 buildings. The initial devastation from these back-to-back earthquakes was immense.
Two months later, heavy rains caused a break in the levees of nearby rivers, resulting in flooding. The earthquakes and heavy rains made an already challenging recovery even more daunting.
Given the country’s high exposure to climate risk, Japan is a world leader in identifying, quantifying, and reporting climate risk using artificial intelligence models and machine learning to pinpoint climate-related impacts on critical infrastructure and dependencies. In addition, in 2019, 200 Japanese companies embraced the recommendation of the Task Force on Climate-related Financial Disclosures and began disclosing their climate risks and vulnerabilities.
Using digital twins to predict flood damage
For the last three years, One Concern has worked with Kumamoto City to increase its resilience to climate-related disasters. To provide the necessary insights to mitigate impacts before a disaster strikes, One Concern built a digital twin to measure overflows from rivers (external flooding), rainfall that exceeds the drainage capacity in urban areas (internal flooding), and floods led by storm surges in coastal areas. The system estimates how the flood water spreads in urban areas and develops different inundation models for river, inland water, and coastal storm surges.
For example, in August of 2021, during dangerously heavy rainfall, One Concern’s platform correctly predicted that the river would not flood, and the river water level prediction results were close to the observed effects.
Caption: Sample image of One Concern platform
How the platform works
Weather forecast data enables One Concern to estimate where and when flooding is likely to occur. The system also monitors weather information provided periodically by Weathernews, a local weather monitoring organization, and sets thresholds for river water levels, coastal tide levels, and precipitation. If the necessary data is available for large rivers, such as first-class rivers, the system can also predict the river water level for the next 72 hours and display the points where water overflow may occur.
The flood damage prediction systems use models for each individual use case, namely a riverine flood model, inland flood model, coastal flood model, inundation prediction model that have been peer-reviewed and validated.
Flooding is highly influenced by local factors, such as detailed topography and land formation. For this reason, One Concern builds models based on natural environment data, such as elevation and land use and regularly calibrates them based on locally accumulated flood damage data.
The public data used to build conventional flood models is insufficient for the comprehensive predictions One Concern produces. One Concern applies artificial intelligence and machine learning technology to publicly available data to interpolate missing information. The model is calibrated using past river flow and water level data, tide level data, and data on the extent and depth of flooding in an actual flood event.
According to One Concern’s Technical Working Group, which periodically reviews our modeling, One Concern’s predictions are at the “cutting edge of the integrated river and coastal flood hazard simulation.”
Simulating extreme heavy rain
Flooding from heavy rainfall is a climate-related disaster that occurs with a degree of regularity and can be anticipated days in advance using commonly implemented weather monitoring technology. Instead of relying on limited traditional warning systems to provide intel and scrambling to prepare for incoming devastation, simulating these events with a digital twin can enable flood resilience by providing decision-makers insights to take action before the disaster strikes.
Our climate is changing rapidly — faster than we have been able to adapt. Year after year, climate-related disasters are setting records, meaning that historical disasters can no longer provide the necessary insights to plan for events yet to come. So rather than look to the past, we must model the future with digital twins capable of assessing vulnerabilities case by case, block by block, and even as granularly as building by building.