Back to Homepage
NASA Space Apps 2025 - Global Nominee

BloomBuzzer: AI-Powered Pollination Optimisation

Selected as a Global Nominee in NASA’s 2025 International Space Apps Challenge. I architected BloomBuzzer, an autonomous pollination optimisation platform synthesising NASA MODIS satellite imagery, hyper-local meteorological data, and Google Gemini AI. The system delivers actionable, high-precision agronomic intelligence for the Californian almond industry, translating complex earth observation data into prescriptive operational strategies.

Project I

~85% OpEx Reduction
4-Band Spectral View
120hr Forecast
Sub-Acre Precision

Technical Stack

React TypeScript TailwindCSS Leaflet Google Gemini AI NASA GIBS WMS Open-Meteo API

NASA Satellite Data Integration

Integrated NASA's GIBS (Global Imagery Browse Services) WMS endpoint to fetch MODIS Terra NDVI (Normalised Difference Vegetation Index) 8-day composite imagery. The NDVI values quantify orchard vigour, providing a key metric for assessing plant health and bloom potential across the defined orchard polygon.

Interactive Geospatial Mapping

Built an interactive mapping interface using Leaflet with custom drawing controls for orchard boundary definition. Users draw their exact orchard polygon on high-resolution satellite imagery, with automatic area calculation (in acres) using geodesic measurements and reverse geocoding for county identification.

Hyper-Local Weather Integration

Implemented real-time 5-day weather forecasting using the Open-Meteo API, fetching temperature, wind speed, wind direction, and precipitation data for the orchard's centroid coordinates. A custom algorithm calculates optimal "Bee Flight Hours" based on temperature thresholds, wind conditions, and rain probability.

AI-Powered Optimisation Engine

Developed a sophisticated prompt engineering system for Google Gemini AI that synthesises all quantitative inputs (NDVI, weather summary, orchard details) into a prescriptive "Pollination Plan". The AI acts as an expert agronomist, generating bloom phenology predictions, hive dosage recommendations, and detailed reasoning grounded in the input data.

Actionable Operational Planning

The output includes a 5-day operational plan with daily recommendations based on predicted bee activity levels, a pollination schedule with phase-specific guidance (pre-bloom prep, hive placement, peak pollination, hive removal), and key metrics including recommended hive count and hives-per-acre ratio.

Challenge Response: BloomWatch

BloomBuzzer addresses NASA's "BloomWatch" challenge, which calls for tools that harness NASA Earth observation data to monitor and visualise plant blooming events and address specific vegetation monitoring, prediction, or management needs. The application demonstrates how satellite-derived NDVI can be translated into actionable insights for commercial agriculture.