In this example zonal statistics are calculated using a big raster and Python rasterstats library. The idea is that we have multiple zones split across several raster files (and some zones also covering more than one raster) and we want to compute zonal statistics for these zones. The way we handle the multiple raster files in this example is to construct a single virtual raster after which we don't have to worry about which polygon covers which raster. Here the virtual raster for 10M DEM available in Puhti is used directly.
The script uses geoconda module in Puhti.
Two a little bit different examples are given here:
zonal_stats_serial.py is the more basic version, here the work is done on one core.
zonal_stats_parallel.py is the more advanced version, here the work is done on 4 cores. To make processing multiple polygons faster we divide the calculation into parts and process them in parallel using multiprocessing library.
Additionally a batch job scripts are provided, for running this script on CSC's Puhti supercluster. For submitting the job to Puhti:
sbatch batch_job_XX.sh