CCMP parsing

This branch is for taking upstream CCMP data and reducing it to weekly averages in an intermediate set of netcdf files; see branch db-population for writing these weekly averages to mongodb.

Downloading data

• this Argovis product based on the 6h wind measurements at https://www.remss.com/measurements/ccmp/
• build and push container image described in Dockerfile-ftp as argovis/ccmp:download
• put REMSS FTP credentials in pod-template.yaml
• edit and run dl.sh to download data as desired.
• a large openshift volume named ccmp is expected to be available.

Computing weekly averages

• weekly averages aligned to match https://psl.noaa.gov/data/gridded/data.noaa.oisst.v2.html
• build and push image described in Dockerfile as argovis/ccmp:reduce and run via pod-reduce.yaml after choosing the appropriate year. Resource intensive, takes about 14h on openshift.

Proofreading

• build and push image described in Dockerfile-proofread as argovis/ccmp:reduce-proofread and run via pod-prrofread.yaml
• Randomly checks for consistency between upstream and reduced data until interrupted.
• Use dump.py from the same container to manually inspect the upstream values for a given lat/lon in a given upstream file.