The EarthCODE catalog now includes the Yield Data at Field Level product from the Earth Observation (YIPEEO) project.
The project aims to enhance crop yield forecasts at the field level by integrating high-resolution satellite data from sources like Copernicus Sentinel-1, Sentinel-2, Sentinel-3, and Proba-V with advanced machine-learning techniques and crop growth models.
The new dataset covers in-situ crop yield data at field level obtained through a network of collaborating farms and research contacts across Europe.
Open Science Catalog
In addition to the Yield Data above, you can now find the Sentinel 1 and Sentinel 2 features used in the project - ESA Open Science Catalog | ESA Open Science Catalog.
Sentinel 1 features for crop yield prediction and estimatation from 2015 to 2022. The processing workflow consists of the following steps:1. Apply precise orbit data 2. Border-noise removal, 3. Radiometric calibration, 4. Radiometric terrain-flattening, 5. Range-Doppler terrain correction, For steps 4. and 5. the 30 m Copernicus Digital Elevation Model (DEM) was used. To extract time series on field level from the pre-processed Sentinel-1 data, several further processing steps were performed to mitigate the impact of the viewing geometry and undesired objects within or near the fields. In a first step, an incidence angle normalization to 40° was performed. Afterwards, all pixels below a standard deviation of 5dB within one year were filtered out as they are typically stemming from radar shadow pixels or are no crop pixels. Finally, the cross-ratio was calculated by subtracting VV and VH polarized backscatter.
Sentinel 2 features based on the provided bands as well as various vegetation indices. The Sentinel-2 L2A data cube is dynamically created by utilising the STAC API. The datacube is pre-filter with scenes of a cloud cover less than 80%. The following features are extracted per field and timestamp: Band Medians and Standard Deviations: B02, B03, B04, B05, B06, B07, B08, B8A, B11, B12; Vegetation indices based on median bands of NDVI, EVI, NDWI, NMDI. An outlier removal was added on a field scale level utilising the SCL band and outlier removal based on 2 x inter quartile range (IQR).
1 Like