Bill & Melinda Gates Foundation

Malaria Program

Problem: Mosquito breeding grounds are difficult and costly to identify

Solution: Forests are a suitable habitat for malaria vectors in tropical regions. Monitoring these habitats using machine learning using Synthetic Aperture Radar (SAR) measurements from Sentinel-1 satellite, Radiant Earth Foundation developed a data pipeline to process Sentinel-1 imagery at scale developed (a) random forest and (2) convoluted neural network models architecture and training data generation.

Status: In Progress

Next Steps: Improve U-Net training with more samples, and optimize U-Net architecture and training parameters based on the results from new samples.

Post-Hurricane Vegetation Analysis

Mapping vegetation cover damage due to a hurricane

Highlights: Major reduction in vegetation cover; particularly, near rivers and flood inundated regions.

Description: The goal of this application is to evaluate the extent of vegetation cover loss after the two major hurricanes in the Caribbean Islands in Sep 2017. The Normalized Difference Vegetation Index (NDVI) is used to assess vegetation greenness, and subsequently coverage. Satellite imagery at each location prior and after the hurricane are retrieved from Planet.

Catholic Relief Services (CRS)

Bednet Distribution Program

Problem: Nigeria-- highest Malaria burden globally, but population distribution not captured by maps. CRS has to distribute 30 million bed nets.

Solution: Radiant Earth Foundation houses Gates Foundation settlement databases, which provides up-to-date high resolution imagery, village boundaries, transportation networks and population estimates.

Status: Completed

Benefit/Impact: CRS were able to distribute bed nets to targeted communities faster and more accurately, resulting in significant staff time and money saved. There was also an increased percentage of all communities served, due to this provision of data.