I have a large NOAA satellite dataset. It consists of visual band and infrared data from one of the NOAA's oldest datasets. We are trying to replicate a product produced by the government, adapted to our needs.
Alternatively, we are amenable to a coauthor as this is an academic project.
The task is as follows:
(1) for each scan, read the dataset using a standard Java library made available for this file format.
(2) Follow a standard paper's procedure for manipulating the data. Of these steps, of significant note is the need to clean out northern lights (approximately). This was done originally by hand, but obviously is infeasible here.
(3) overlay cloud data or temperature data (we have cloud data for a large subset of the period).
The deliverable is a script that basically completes all of those steps to produce global snapshots of the Earth's surface using the satellite data. Then, benchmarking against the government dataset.
Since the satellites are calibrated differently, it'd be nice to know that the code doesn't break for the various versions of the satellite.
You can use any software you want. I don't care.