Mangrove Restoration Optimisation
A proposed approach for deciding where to plant mangroves, combining machine learning with multi-objective optimisation, and a planning prototype to go with it.
Where should mangroves be planted, when several goals pull in different directions?
The proposal weighed three goals against each other: survival of the planted mangroves, societal impact, and economic impact.
The proposed method
Machine learning to assess candidate sites, then Pareto multi-objective optimisation to show the trade-offs between them, rather than a single answer.
The method was designed and proposed, but not yet fully tested.
My role combined technical development with coordinating a team whose members brought different experience. I worked with Google Cloud and Vertex AI, while dividing responsibilities and helping bring our contributions together. Much of the work was turning an ambitious proposal into tasks we could actually tackle.
The planning tool
A web prototype for restoration planners: set ecological and financial constraints, compare candidate sites on a map, and follow planted sites through monitoring and field verification.
Behind it, a Google Earth Engine service reads vegetation health from Sentinel-2 and land-surface temperature from Landsat 9 for any point on the map.
React · Flask · Google Earth Engine · Google Cloud · Vertex AI · Sentinel-2 · Landsat 9
Looking back
What began with spotting a competition and submitting an application with three teammates who had no programming background took us all the way to Saudi Arabia. It was a rare opportunity to experience the country, meet participants from around the world, and learn together as a team from different disciplines. I had not imagined where filling in that first form would take us.