Produce spatially aggregated, cloud-masked Sentinel-2 (S2) surface reflectance data for one million fields over six years.
After an evaluation of the impact of clouds on aggregated S2 data previously delivered to a modeling team, showing that > 40% of aggregated observations were severely impacted by clouds, we worked to deliver the aggregated S2 data using the surface reflectance level processing and with cloudy pixels masked.
Cloud-masked, surface reflectance aggregated S2 data for ~1 million fields, over six years. Stored on AWS S3 as a parquet file with indexed partitions.
Descartes Labs (DL) Python APIs to retrieve and mask remote sensing data at scale; sen2cor atmospheric correction of top-of-atmosphere S2 data; spatial aggregation and cloud masking with numpy and xarray, AWS CLI to upload very large quantities of data.
Geospatial remote sensing data pipeline for atmospheric correction of retrieving S2 data, retrieval, cloud masking using the native Sentinel-2 L2A Scene Classification Layer (SCL) and storage of data on AWS S3 for all available S2 observations in ~1 million unique farm fields over 6 years.
