1. General
Category
SDG 11: Sustainable Cities and Communities
SDG 13: Climate Action
Category
Government Employees
2. Project Details
Company or Institution
Dymaxion Labs
Project
Adress Informal settlements' exposure to climate-related hazard
General description of the AI solution
An AI tool, specifically a Machine Learning (ML) tool, that is designed to monitor the rate of growth, expansion and change of Informal Settlements in fast-urbanizing environments especially at the fringe of the city while identifying flash flooding and other climate-related risks. This AI tool will produce risk heat-maps where the flood-flash risks intersect with informal settlements. Through an integrated analysis of a large variety of data including satellite imagery, local topographic, climate, ecological, socio-economic, informal settlements and flood data, is designed to be used as a wide-range planning tool. This tool will add value to planning practices, municipalities and communities.
Website
Organisation
Dymaxion Labs
3. Aspects
Excellence and Scientific Quality: Please detail the improvements made by the nominee or the nominees’ team or yourself if your applying for the award, and why they have been a success.
The starting point for the project was a geospatial database combining regional and local scale data sources. Once the dataset was built, we used convolutional networks, as they consider the information not only in a single point (pixel) but also in adjacent ones. We used a RESNET-50 network, to develop a classifier of image tiles that distinguishes tiles by likelihood to flood across the entire satellite image, based on the “ground truth” dataset showing actually flooded areas and their associated characteristics (topography, land cover and precipitations among others)
The working procedure is:
a. Instantiate the convolutional base of ResNet-50
b. Add a fully-connected model on top, with a standard SGD optimizer and validating with the binary cross-entropy loss function
c. Freeze the layers of the model up to the top 70 layers
d. Retrain the model
Climate change predictions are regional in scope but mitigation, adaptation and resilience-building actions are local, applying to specific sites with similar features and risks.
Scaling of impact to SDGs: Please detail how many citizens/communities and/or researchers/businesses this has had or can have a positive impact on, including particular groups where applicable and to what extent.
Despite selecting only SDGs 11 and 13, our solution intersects with more objectives and scopes of the Sustainable Development Goals such as reducing inequalities and no poverty.
The urban population of the developing countries is expected to double between 2000 and 2030 while the built-up area of their cities can be expected to triple. Much of this growth will take place through the expansion of informal settlements and there are more than one billion residents of vulnerable communities at physical and economic risk from climate change. This tool identifies and monitors location and growth patterns of informal settlements through satellite or aerial images and that, combined with topography and local precipitation data, can identify exposure to climate-related risks, shocks, and stresses within hours once the system is set up. The solution is able to scale economically and quickly, enabling use at the regional and local levels. This solution is replicable in all Central America and every country that suffers climate change consequences regarding water hazards.
Quick data analysis and visualization can be done more cheaply, frequently, and transparently than current approaches through an inclusive process for the collection and dissemination of regional and local information, and for the development of effective interventions, through a mobile interface, harnesses the collective intelligence of communities, local authorities and planning specialist.
Scaling of AI solution: Please detail what proof of concept or implementations can you show now in terms of its efficacy and how the solution can be scaled to provide a global impact ad how realistic that scaling is.
The model was required by and created with the help of the INSTITUTE FOR INTERNATIONAL URBAN DEVELOPMENT, due to the lack of downscaled information that difficult the comparison on how individual communities, households and small businesses adapt to severe and extreme weather events. This creates challenges for local governments and other public and non-governmental organizations in implementing appropriate measures to increase resilience in poor urban areas.
A proof of concept was made in Tegucigalpa and Valle de Sula, Honduras where we trained and tested the expansion/flash flooding-monitoring tool. The ML model output lets us identify and monitor location and growth patterns of informal settlements and, combining satellite imagery with topography and local precipitation data, we were able to identify exposure to climate-related risks, shocks, and stresses within hours once the system is set up.
We are finding it difficult to replicate the study in other countries like Guatemala, which is exposed to a similar severe climate and has a similar location; we have already identified informal settlements in Guatemala City through the same algorithm used in Tegucigalpa. As the model was built on the cloud, its able to process large amounts of information from different countries along the globe. Its quick data processing and association to satellite imagery, enable governments and NGO to plan rapid-action measures to mitigate water hazard consequences over informal settlements.
Ethical aspect: Please detail the way the solution addresses any of the main ethical aspects, including trustworthiness, bias, gender issues, etc.
The application of AI technologies in this project is ethical and equitable. Dymaxion Labs is an Argentinian startup that seeks to contribute private companies, NGOs and governments with ML models and GIS platforms to develop a more equal world and enhace countries economic growth through a more cheap, frequent, and transparent methodology than current approaches.
The tool created in this project bring together several layers of information such as topography, land cover, precipitations and city infraestrcture deploying them in a user-friendly way to accelerate decision-making.