BenjaminAbabio/SpatioTemporal-Air-Quality-Data-Cube-Accra-Ghana-

This project consolidates multi-source satellite observations and ground-station measurements into a unified SpatioTemporal Data Cube. The resulting dataset is specifically engineered for Machine Learning applications in urban air quality forecasting and PM2.5 estimation. ## Data Sources & Variables The cube integrates the following parameters at a synchronized 1km spatial resolution: - **Ground Truth:** PM2.5 concentrations (Low-cost sensor network). - **Aerosols:** Aerosol Optical Depth (AOD) at 550nm (MODIS/Sentinel-5P). - **Meteorology:** 2m Temperature and Relative Humidity (ERA5-Land). - **Dynamics:** Planetary Boundary Layer Height (PBLH) (ERA5). - **Hydrology:** Daily Precipitation (CHIRPS/GPM). ## Technical Implementation The project utilizes Python and Google Earth Engine to perform: 1. **SpatioTemporal Matching:** Aligning asynchronous satellite overpasses with hourly ground station readings. 2. **Feature Engineering:** Calculation of Julian days, hour of day, and coordinate encoding. 3. **Data Imputation:** Filling ground-station gaps using Random Forest regressor logic. 4. **Coordinate Harmonization:** Projecting all data to a common CRS (EPSG:32630) for spatial consistency. ## Data Structure The final output is a flattened GeoDataFrame/CSV containing `[timestamp, location_id, latitude, longitude, pm25, aod, temp, rh, pbl, precip]`. ## Requirements - Python 3.8+ - Geopandas - Scikit-learn - Scipy (cKDTree for nearest-neighbor matching)