PLACE Data - The Spatial Foundation of AI
Applications for Climate Change
Artificial intelligence (AI) and machine learning (ML) are ushering in a technological revolution that will fundamentally impact work, economies, and societies in ways we have only begun to imagine. With the ability to detect, observe, and monitor patterns at scale, AI has the potential to transform how Africa addresses its most pressing challenges. However, there is a conspicuous problem. Unlike the Global North, much of Africa lacks the high-resolution spatial data foundation needed for AI, placing the continent at risk of being left behind in the data revolution.
PLACE has curated a series of use cases to demonstrate how ML can utilize spatial data to rapidly detect and map real-world objects to address common challenges. By having the ability to almost instantly detect objects in an observed landscape, decision-makers can quickly answer questions like "How green is my city?", "How many people live here?" or “How resilient is our housing”, as they work to improve livelihoods, mitigate the effects of climate change, and advance the UN Sustainable Development Goals (SDGs).
How green is my city?
Trees and green spaces are visible on PLACE imagery. Using a pre-trained model (meaning detection only gets better with more labels as someone works across the imagery tagging what they observe to be trees making the model better at finding trees) it’s possible to identify and map all trees in an area of interest, and in some cases, identify species.
Within one section of the city (measuring just 0.3 sq km), the model detected well over 1,000 trees with an estimated canopy cover of 0.05 sq km. That’s 17% green and it doesn’t include parks and open spaces. City planners may be able to infer from this type of analysis whether a city has green lungs or not – and can use the vegetation data for other use cases.
Who will be impacted by infrastructure expansion?
Abidjan, Cote d’Ivoire is one of the largest and fastest-growing cities in Africa and has consequently prioritized improvement of urban transport. PLACE Community Members, Atlas AI and Esri, supporting analysis by the Ivorian geospatial agency, applied several ML models to PLACE data to detect cars, vegetation, and building outlines. The resulting datasets can be used across a variety of urban planning scenarios, including identifying buildings impacted by road widening and quantifying resettlement needs. The data is also crucial for visualization and can be the basis of a 3D model for a new urban transport system, as one of many examples.
How resilient is our housing?
The World Bank Global Program on Resilient Housing supports governments to build and retrofit homes, so they are better able to withstand the impacts of natural disasters and climate change. The World Bank considers roof condition and material as key proxies for determining housing resilience.
In Accra, Ghana, The World Bank ran its housing models on PLACE imagery. Within the community of Dansoman (an area to the west of Accra), over 25,000 dwellings were detected, of which more than 15,000 (60%) had a metal roof. Slightly under 15,000 (60%) had a roof deemed to be of fair condition with just over 7,000 (30%) having a roof of poor condition, allowing the city to categorize the housing stock.
In March 2023, Cyclone Freddy inundated southern Malawi with record rainfall, displacing over 650,000 citizens according to the UN. The local government of Zomba, Malawi, at the foot of the Zomba Plateau, recognized the ongoing risk they face and prioritized a flood analysis of the city. Using PLACE imagery, ML models identified building outlines (bottom left). We then derived a Digital Surface Model (DSM) from the collected imagery, identifying where water will flow. Finally, we modeled the rise in the water level based on simulated rainfall events, identifying which buildings will flood depending upon the water level rise. The City Council can now work to implement mitigation plans that will limit flood damage or work with occupants to relocate those most at risk.
Who is at risk of flood?
In March 2023, Cyclone Freddy inundated southern Malawi with record rainfall, displacing over 650,000 citizens according to the UN. The local government of Zomba, Malawi, at the foot of the Zomba Plateau, recognized the ongoing risk they face and prioritized a flood analysis of the city. Using PLACE imagery, ML models identified building outlines (bottom left). We then derived a Digital Surface Model (DSM) from the collected imagery, identifying where water will flow. Finally, we modeled the rise in the water level based on simulated rainfall events, identifying which buildings will flood depending upon the water level rise. The City Council can now work to implement mitigation plans that will limit flood damage or work with occupants to relocate those most at risk.
How do we transition to renewable energy?
Many African countries are grappling with an inability to effectively provide sufficient electricity to their citizens. Africa recently reversed positive trends in improving access to modern energy, with 4% more people living without electricity in 2021 than in 20191. With much of the continent lying within the tropical belt, solar power presents a scalable, cost-effective mechanism for closing the energy gap.
As the Government is currently planning investment in solar panels, ML can be used to help identify the optimum location for these new installations. By measuring the roof height, the direction the roof is facing, and the angle of the roof, as well as an assessment of the roof condition, it can prioritize those roofs best suited for solar panels as well as estimate the power that can expected to be produced.
Is my city susceptible to overheating?
Excessive temperatures, especially in an urban setting, can be dangerous. As higher temperatures put vulnerable populations at an increased risk of death, city planners need to know which areas are likely to be most impacted and guide interventions like tree planting and the use of cooler building materials.
A heat risk index or a Heat Vulnerability Index (HVI) can be used to map neighborhoods whose residents are more at risk during and immediately following extreme heat. Heat risk is determined by environmental factors like surface temperature, vegetation cover, and population density.
A heat index was generated by combining three (3) datasets for Zomba, Malawi. Normalized Difference Vegetation Index (NDVI) and Land Surface Temperatures (LST) data from Landsat 8 satellite data with population density derived from PLACE data.
The result is a HVI map that shows at-risk areas in Zomba ranked from 1 - 5, with 5 being the most at risk of excessive heat and 1 the least, represented by a 60m-by-60m grid, and identifying those locations in the city that may require additional services during a heat wave.
Additional Applications with PLACE Data
These examples highlight just a few of the uses of PLACE data that have been shared by the PLACE Community, and we look forward to continued innovation by our membership.
