Open tristangrupp opened 1 week ago
Approximately 80k boundaries.
Some boundaries might encapsulate multiple fields so could be split by road line geometries to improve them.
Otherwise, filter them out by size OR potentially filter with randomly sampled points within each polygons, extracting NDVI time series to each point, filtering out boundaries whose points are too different in their time series of NDVI
Use public data from CONAB from Brazil: