‘AI for Humanitarian Action’ with Microsoft

Know more How is Artificial Intelligence
helping Satyavati prepare for
impending cyclones? Know more Bridging gaps between risk
information and vulnerable
communities Know more Push for better and inclusive
climate analytics for all

AI powered disaster impact model

Context
IPCC code red has prompted governments at national and state levels to urgently develop climate change adaptation and mitigation pathways, making it an important aspect of the current development discourses. There is also an understanding pledged by world leaders at COP26 in Glasgow that the financing for climate must be substantially increased. While the science, investment and activism part grow on one side, conversion of these facts, resources, and voices into context specific solutions at scale remains hard. In urban areas, this gap is starker. Urban areas harbor a multitude of complexities in terms of livelihoods, populations, inequalities, governance, thus making it tricky to detect and understand climate induced disaster risks. Deciphering risks is also difficult with the current hazard warnings as they generalise information for complete districts or cities at best. Thus, leaving neighbourhoods with little information about their specific vulnerabilities that is needed in order to adapt to the amplifying risks, making informed pre-emptive action by individuals/local communities almost impossible.

Need
There is a need to collate hazard risk information at a neighbourhood level to get clarity on the driving factors as risk is a culmination of hazard, vulnerability and exposure. Its experience varies with the coping capacity of the family, structural strength of the building and other factors such as slope, elevation and built-up density of the neighbourhood. The available risk information should not only be localised, but needs to be made easily decipherable for at-risk communities and other actors involved in disaster response and planning.

Our Solution
SEEDS, in collaboration with Microsoft and technology partner Gramener, has developed “Sunny Lives”, an AI model that bridges this gap so that communities can be better prepared and governments can plan better at a local scale. The model, which has been developed under Microsoft’s global program ‘AI for Humanitarian Action’, processes large volumes of data to provide risk information at a hyper-local (building cluster) level. With the underlying principle that the roof condition of a building says a lot about the condition of the house, the model uses high resolution satellite imagery to detect and assign risk scores to buildings according to their roof types. Other location and hazard specific attributes are also used in calculating the risk scores.

Impact
The model was piloted for risk assessment of cyclone induced flooding during cyclone Nivar and cyclone Burevi in 2020 and showcased promising results. In 2021, we scaled the dissemination to reach out to over 50,000 families vulnerable to heatwaves (Delhi and Nagpur), cyclones (parts of Odisha and Gujarat) and monsoon flooding (Puri and Mumbai). During cyclone Yaas, we deployed the model for coastal parts of Puri and shared advisories and pre-emptive measures with over thousand vulnerable families, helping reduce their losses.

Way forward
As the model matures, the model’s deployment will be scaled for multiple cities across the country. We are also looking to collaborate with various city and state governments, to aid them in developing risk informed planning. Our vision for the model is to use it for climate change adaptation and disaster management in a way that the hyper-local risk of the communities is understood and pathways for their protection and resilience are put into practice.