Seven promising innovations of artificial intelligence across the DRM cycle
We are living in interesting times: simulataneously playing catch-up with the accelerating pace of climate-related disasters alongside the tremendous speed of technological growth. These two dizzying trends present an opportunity - to
We are living in interesting times: simulataneously playing catch-up with the accelerating pace of climate-related disasters alongside the tremendous speed of technological growth. These two dizzying trends present an opportunity - to use the latter to deal with the former.
The abundance of available data and sensor technology poses a unique challenge: we often have more data than we can effectively use.
Key sources of disaster risk and loss data include the Sendai Framework Monitor, where Member States report disaster impacts and progress in risk reduction. EM-DAT provides a comprehensive database tracking global disasters from 1900 to the present. Another data source is DesInventar Sendai which is an open-source system that allows countries to record detailed, sub-national data on disaster losses, including human impacts and economic indicators.
A major challenge, however, is making sense of the abundance of data - and tackling the challenge of heterogeneous data sources.
In this context, artificial intelligence (AI) emerges as a transformative technology. AI can enhance the speed and accuracy of data collection and analysis, enabling rapid and informed decision-making through real-time data analysis. In a nutshell, AI could allow us to tackle complex problems that were previously deemed too difficult or impossible to solve.
Read on to explore seven diverse examples of how AI can reduce disaster risks and build resilient communities - from harnessing satellite data to predict natural hazards, to improving multilingual weather alerts and streamlining evacuation procedures:
Simulating risk scenarios with a digital twin
The Digital Twin Earth Hydrology Platform is pioneering a virtual model of Earth's water cycle using advanced satellite data and modeling. This initiative aims to predict and manage water-related disasters by simulating scenarios with unprecedented detail.
As human activities and climate change increasingly affect the water cycle, accurately predicting floods and droughts becomes crucial but challenging. Traditional methods struggle with the complexity of interactions across diverse landscapes.
Led by Dr. Luca Brocca, the project utilizes high-resolution satellite data measuring soil moisture, precipitation, evaporation, river flow, and snow depth. This detailed data is crucial for developing models that simulate real-world conditions accurately.
Accessible via a cloud-based platform, these models enable interactive simulations and visualizations. They empower decision-makers and communities to assess risks like floods and landslides and optimize water resource management strategies.
Explore digital twin technology
Improving monsoon predictions with machine learning
There are numerous examples of how artificial intelligence has improved weather forecasting, such as typhoons, and recent studies are looking at new uses, such as predicting dust storms. Another example concerns rainfall predictions.
Every year, the South Asian monsoon brings crucial rainfall to over a billion people, impacting agriculture and urban planning. Predicting these rainfall patterns, crucial for crop harvesting and flood preparedness, has traditionally been challenging beyond a few days. However, a new machine-learning-based forecast, developed by Eviatar Bach and collaborators, significantly enhances predictions 10 to 30 days in advance.
This innovative approach combines machine learning with traditional numerical models to better understand and predict monsoon intraseasonal oscillations (MISOs). By integrating these technologies, the accuracy of rainfall predictions has improved by up to 70%, aiding in climate adaptation and disaster preparedness efforts.
Crowdsourcing bushfire prevention with an AI-powered app
The NOBURN app, developed by the University of Adelaide, leverages AI to predict and prevent bushfires by analyzing user-submitted photos of fire-prone areas. This tool assesses potential bushfire fuel loads and estimates the severity and spread of fires
By mimicking human experts, the AI empowers citizens like bushwalkers and campers to help prevent bushfires. Collecting quality data through crowdsourcing and training the AI will take up to two years, but the app aims to raise awareness about AI's role in disaster risk reduction. The goal is to create a real-time command center for bushfire management, enhancing situational awareness and resource deployment.
The initiative, sparked by the devastating 2019-2020 Black Summer bushfires, showcases the potential of AI in bushfire research and prevention. NOBURN represents a significant step in combining different disciplines to make bushfire research accessible to the public, ultimately aiming to reduce the devastating impacts of future bushfires.
Inclusive weather alerts save lives
NOAA's National Weather Service (NWS) is leveraging artificial intelligence to enhance disaster risk reduction by translating weather forecasts and warnings into multiple languages. For the past 30 years, these translations were done manually, primarily in Spanish.
Now, through a series of pilot projects, NWS has trained AI software to accurately translate weather, water, and climate terminology into Spanish and Simplified Chinese, with plans to include Samoan and Vietnamese next, and additional languages in the future. This initiative aims to provide timely, lifesaving information to non-English speakers, ensuring inclusivity and safety during severe weather events.
This project intends to improve service equity for traditionally underserved and vulnerable populations with limited English proficiency. By providing weather forecasts and warnings in multiple languages, NWS aims to enhance community readiness and resilience as climate change drives more extreme weather events, ultimately protecting more lives.
Read about the multilingual alerts
AI in space: cutting-edge wildfire detection
Australian scientists are advancing bushfire detection using cube satellites equipped with onboard AI, detecting fires 500 times faster than traditional methods. This innovation compresses and processes hyperspectral imagery in space, enabling early detection before fires intensify.
Led by the University of South Australia and funded by SmartSat CRC, the Kanyini mission pioneers bushfire and smoke detection. Equipped with a hyperspectral imager, the satellite generates detailed maps crucial for disaster risk reduction, including bushfire monitoring and water quality assessment.
The researchers developed a lightweight AI model to detect smoke onboard, reducing data volume and energy consumption significantly. The AI detects smoke faster than ground-based systems, crucial for early warning systems and disaster response. Planned for operational deployment in 2025, this technology aims to revolutionize fire detection and disaster preparedness globally.
Learn about AI satellite technology
Using AI to enhance evacuation processes
In Iwate Prefecture, Japan, initiatives are underway to bolster disaster preparedness using artificial intelligence (AI) technologies, spurred by lessons from the devastating 2011 Great East Japan Earthquake. Rikuzentakata City has pioneered an AI-driven system aimed at enhancing evacuation procedures.
This system automatically contacts registered residents during emergencies, querying their evacuation status and condition via telephone. Responses are converted into text and relayed to emergency centers, addressing challenges in disseminating timely information, particularly to elderly and non-social media users.
Meanwhile, Visnu, an IT firm in Morioka City, is developing AI-equipped surveillance cameras designed to monitor anomalies like tidal fluctuations and human presence in evacuation zones. Collaborating with Kamaishi City and local civil engineering firms, Visnu aims to deploy these cameras at strategic locations such as ports and roadways.
Assessing damage with the support of deep learning
In response to the slow process of assessing post-disaster damage to buildings and infrastructure, researchers, in collaboration with Simon Fraser University, have developed a new deep-learning model named DAHiTrA. This model utilizes high-resolution satellite imagery to classify the extent of destruction following natural disasters, such as Hurricane Harvey.
DAHiTrA employs advanced algorithms to analyze geographic features and detect changes over time. By comparing pre- and post-disaster satellite images, the model accurately determines the severity of damage, including collapse, partial damage, and water damage. This capability extends beyond buildings to encompass roads and bridges.
The model's strength lies in its ability to precisely delineate building boundaries, facilitating more accurate damage assessments. Its rapid processing of large volumes of satellite data enables swift response and recovery actions post-disaster, aiding in efficient resource allocation by governments and relief organizations.
Artificial intelligence is revolutionizing disaster risk reduction by enhancing data collection, analysis, and response across various phases of the disaster management cycle.
By leveraging AI's capabilities in real-time data processing and decision-making, we can mitigate risks more effectively and allocate resources efficiently during emergencies - saving lives and reducing the impact of disasters globally.