In recent decades, urban areas have seen a significant increase in population, tripling over the last 50 years. As cities grow, they face heightened risks from natural disasters like earthquakes and extreme weather events, which are becoming more frequent and intense due to climate change. This scenario has prompted global efforts to enhance disaster preparedness and response through advanced technologies such as artificial intelligence (AI).
On 6th November, the Barcelona Supercomputing Center in Spain will host the inaugural meeting of the Global Initiative on Resilience to Natural Hazards through AI Solutions. This new initiative, launched by the United Nations, seeks to advise governments, organisations, and communities on employing AI for efficient disaster management. This initiative is founded on work commenced by the International Telecommunications Union, the World Meteorological Organization (WMO), and the U.N. Environment Programme. This collaboration began in 2021 to establish best practices for the utilisation of AI in disaster management, focusing on enhancements in data collection, forecasting, and communication.
Monique Kuglitsch, who chaired the initial focus group, highlighted the beneficial applications of AI across various natural hazards. For instance, AI was recently utilised to optimally place traffic sensors in Florida to identify road blockages following hurricanes. Additionally, AI forecast models accurately predicted Hurricane Milton's landfall near Siesta Key, Florida. The integration of AI into weather services helps streamline warnings; as demonstrated by the U.S. National Weather Service's partnership with AI translation company Lilt, reducing the time to translate hurricane warnings into Spanish and simplified Chinese from an hour to just 10 minutes.
Moreover, AI is pivotal in coordinating response efforts post-disaster. Non-profit organisation GiveDirectly utilised Google's machine-learning models to assess satellite images and prioritise aid distribution in areas severely affected by hurricanes Milton and Ian. AI-assisted aerial image analysis has been employed following events such as Cyclone Freddy in Mozambique and the 7.8 magnitude earthquake in Adıyaman, Turkey.
Operating early warning systems is a core public duty, yet the involvement of private industry in AI-driven climate modelling is expanding. Companies like SeismicAI collaborate with Mexican civil defence agencies to employ AI-enhanced seismic sensors for real-time earthquake detection. Tech giants including Google, Nvidia, and Huawei are collaborating with European meteorological services to develop models that provide rapid and precise forecasts. Similarly, IBM, in partnership with NASA, recently launched an open-source model designed for a variety of climate modelling applications, which can be operated on a desktop.
AI's integration into weather forecasting has long been underway but recent advancements have enabled the creation of entirely AI-driven models, boosting forecast speed and accuracy. AI models derive patterns from extensive historical climate data, allowing for more detailed and efficient predictions. For example, the European Centre for Medium-Range Weather Forecasts employs a high-resolution model with a grid of 5.5 miles, whereas startup Atmo's models can achieve resolutions finer than one square mile. Such granularity proves crucial for cities managing extreme weather impacts, as seen in projects in the Philippines and Tuvalu.
Despite the promising potential, AI models are only as accurate as the data fed into them. In regions with inadequate meteorological infrastructure, data can be sparse, resulting in less dependable AI predictions. Furthermore, as AI models grow complex, they become 'black boxes', making it harder to trace the input-output process. The U.N. initiative aims to establish guidelines to ensure responsible AI use, including transparency about model limitations and cross-regional operability.
The initiative plans to apply these guidelines practically. One project aims to predict wildfire occurrences near Athens, Greece, while another seeks to improve flood and landslide warnings in Tbilisi, Georgia, using AI. Concurrently, companies like Tomorrow.io are addressing data gaps by deploying satellites equipped with meteorological sensors to gather additional data, improving the accuracy of their AI forecasts used by various cities and industries.
Another U.N. project, the Systematic Observations Financing Facility (SOFF), focuses on closing global weather data gaps, providing financial and technical support to disadvantaged regions. While engaging with private AI developers such as Google and Microsoft, the WMO underscores the importance of maintaining human oversight in AI applications, ensuring that local meteorological services remain central to decision-making processes.
Source: Noah Wire Services