Tree Detection for Reforestation in Madagascar Using Satellite Imagery
The Omdena team of engineers developed a machine learning model to detect tree patches and approximate the number of trees.
The partner of this project, Madagascar-based startup Bôndy, aims to build innovative reforestation solutions to create social and environmental impacts.
Madagascar is the 5th poorest country in the world and faces terrific consequences due to climate change.
Madagascar-based impact startup Bôndy aims to build innovative reforestation solutions to create social and environmental impacts. They plant high-value tree species on farmer lands to develop substitutes for the remaining primary forest and to help the rural populations have more sustainable revenues. Bôndy’s community-based reforestation program in Madagascar is spread across 5 regions from agroforestry to mangroves.
The project outcomes
The goal of this project has been to build an AI model using satellite and drone imagery to monitor reforestation success based on the first 5 years of tree growth and survival.
The data has been collected from satellite images (Planet satellite monthly data and Sentinel 2 Level 3A cloud-free data) and meteorological data (ERA-5 data) apart from drone images and field data provided by Bondy. Three vegetation indices namely Normalized Difference Vegetation Index (NDVI), Normalized Difference Water Index (NDWI), and Modified Soil Adjusted Vegetation Index Values (MSAVI2) were calculated. The indices calculation has been done using the Python library Rasterio.
The drone image preprocessing resulted in the extraction of useful information like centroid, date and time, flight altitude, and camera information (height, width, focal length). The data was further pre-processed to estimate the number of trees in an image.
To monitor the impact of the environment on the health of trees, three environmental parameters were chosen; temperature, precipitation, and evaporation. Time series models for each of the parameters for three regions of Madagascar were built using the FaceBook prophet library. The team also tested a number of pre-trained super-resolution models on drone and satellite data.
The best-performing model can detect the tree patches and approximate the number of trees. The results from this project continue to be developed through a dedicated Omdena team of engineers.
This challenge has been hosted with our friends at